Continuous monitoring of Synthetic Biology

In 2025, as part of the Synthetic Biology monitoring programme, a total of 4,950 publications were identified through literature searches in scientific databases and search platforms [1]. From these, the ZKBS working group on Synthetic Biology selected 155 publications for a more detailed analysis. Of these, 24 publications were chosen to be presented as short summaries on the ZKBS-homepage. They provide an overview of the global developments in the field of Synthetic Biology.

Monitoring of the original articles published in international peer-reviewed journals and categorized by the ZKBS as Synthetic Biology revealed that all research approaches and methods described fall within the scope of existing legislation, in particular the German Genetic Engineering Act (GenTG).

In the course of the monitoring, an increase in publications on the generation of protocells and minimal cells using bottom-up approaches compared to 2024 was noted. Furthermore, as was observed in 2024, the number of publications on the application of artificial intelligence (AI) and laboratory automation increased. Additionally, the ZKBS became aware of a preprint publication by King et al. (2025)[2] on the bioRxiv server. As this paper has not yet been peer-reviewed, it was not included in the monitoring process as a standalone contribution. Nevertheless, this publication marks a significant advancement in the field of Synthetic Biology. The authors describe the first successful generation of functional bacteriophage genomes using an AI model. The ZKBS has examined this work in detail and will closely monitor further developments in the field of AI-based de novo generation of nucleic acid sequences and genomes. In addition, on the 19th of June 2026, a commentary on this subject was published on the ZKBS-homepage: ZKBS commentary on the challenges posed by de novo-generated nucleic acid sequences and proteins in the assessment of genetic engineering work.

[1]: The search was conducted in the PubMed and Google Scholar databases. In PubMed, the search terms “Synthetic Biology”, “Mirror Life”, “Mirror Image” and “Mirror Bacteria” were used. In Google Scholar, searches were carried out for “Mirror Life” and “Mirror Bacteria”. 

[2] King et al. (2025) bioRxiv 2025.09.12.675911.

In 2024, the ZKBS working group on Synthetic Biology selected 225 publications from the monitoring process for detailed analysis. Of these, 28 were chosen to be presented as short summaries on the ZKBS-homepage, offering an overview on and insight into the global developments in the field of Synthetic Biology.

The review of original articles published in international peer-reviewed journals, categorized as Synthetic Biology by the ZKBS, revealed that all reported approaches and methodologies fall within the scope of existing legal frameworks, particularly the GenTG.

Mirror Life

The ZKBS took notice of the comment on the development of “mirror bacteria” by Adamala et al., published as a policy forum in Science in December 2024 and authored by 38 scientists from different countries and disciplines, as well as of the accompanying technical report on mirror bacteria.

The comment outlines potential risks posed by artificially created mirror bacteria composed of mirror-image amino acids, sugars, etc. According to the authors, the technical realization of mirror life is still at least a decade away. However, they wish to engage early with the scientific community, policymakers, research funders, industry, civil society, and the public to discuss potential negative aspects of mirror bacteria, e. g. immune evasion or harmful effects to ecosystems. The authors recommend that mirror organisms should not be created unless there is evidence that mirror life would not pose extraordinary dangers. For related technologies, which might be used to create potential therapeutic substances, the authors do not recommend any restrictions.  

The ZKBS Working Group on Synthetic Biology has intensively discussed the comment and has published a first assessment on its homepage. As “mirror life” is going to be a matter of scientific as well as societal debate, ZKBS will consecutively and considerately monitor mirror life’s worldwide development and report on this issue at ZKBS’ website.

Two Highlight Publications

The ZKBS would like to consider two original articles as Synthetic Biology highlights in 2024.

At first, Chen et al. reported on the first multi-cellular organism with a partially artificial chromosome. The terrestrial moss Physcomitrium patens´ chromosome 18 was chemically synthesized and significantly reduced in size. Following this reduction, the moss grew without discernible phenotyping effects. The work is planned to be continued in the synthetic moss genome project that aims at ultimately replacing the full genome with a synthetic one.

The second paper the ZKBS would like to highlight is the work of Gao et al., in which photosynthetic cyanobacteria modified to secrete glucose and ATP were introduced as endosymbionts into yeast cells. The yeast cells being mitochondria-deficient could only survive without external carbon sources when the cyanobacteria supplied glucose and ATP. These chimeras were stably formed, grew using solely CO₂ and light, and were able to produce natural biosynthetic products under photosynthetic conditions.

 PDF with the highlights and summaries of the 2024 monitoring report on Synthetic Biology (not accessible) [PDF, 347 KB]

The screening process identified 217 publications that were analysed in more detail. 37 of these now appear as a short summary on the ZKBS-homepage.

The ZKBS considers the work of Inda-Webb et al. to be one highlight paper from 2023. The authors have generated an application which combines biological and technical engineering to a special degree and is therefore a prime example of an application of synthetic biology. The authors have developed a capsule the size of a tablet that contains live bacteria and can be ingested to detect inflammatory markers in the gastrointestinal tract. The information obtained is sent directly to a smartphone. The tablet has already been tested on pigs and could become a valuable diagnostic tool.

Another paper the ZKBS would like to highlight is the work of Belluati et al. Here, the authors have developed a bioreactor in which various biological processes can take place. To produce the bioreactor, the authors used a method in which myoglobin acts as a biocatalyst for the assembly of synthetic polymers into GUVs (giant unilamellar vesicles), i.e. artificial, cell-like structures. This represents a major step towards functional artificial cells. 

Semantic design of functional de novo genes from a genomic language model (Merchant et al. 2026)

In 2024, a research collaboration presented the genomic language model Evo 1 in the paper by Nguyen et al. It was based on training data comprising 2.7 million bacterial and phage genomes, with a focus on the co-design of molecular systems, such as new CRISPR-Cas approaches. It represents a paradigm shift by designing functional biological systems directly at the DNA level, rather than merely modelling protein structures. Here, the authors present the upgrade to Evo 1.5. and SynGenome, an AI-generated genome database containing 120 billion base pairs of synthetic DNA sequences derived from prompts comprising 9,000 functional terms.

Through even more extensive training with DNA datasets, Evo 1.5 learns the genomic grammar and semantics of genomes and successfully generates novel genes with specific functions. These include, among others, toxin-antitoxin pairs, including a toxic gene with no similarity to known bacterial toxins, as well as functional multi-component systems such as CRISPR-Cas systems and RNAs, demonstrating the ability to synthetically create complex biological functions. For example, Evo 1.5 was able to learn the rule of the type II toxin-antitoxin system, in which the gene for the toxin and the gene for the antitoxin are always located next to each other, even though the sequences vary greatly from one bacterial species to another. It was found that Evo 1.5 understands the spatial arrangement and logic of genes in the genome and generates sequences that encode protein pairs for which in silico predictions suggest complex formation. The authors conclude that semantic design enables the exploration of synthetic genomics, revealing biological insights that both complement and exceed discoveries made in natural organisms. 

A designer synthetic chromosome fragment functions in moss (Chen et al. 2024)

The authors have constructed the first living multi-cellular organism, the terrestrial moss Physcomitrium patens, that carries a partial artificial chromosome. A region of 155 kilobases (kb), which spans about one third of an arm of chromosome 18 was replaced with a redesigned, simplified, chemically synthesized fragment of roughly 68 kb. The semi-syn18L moss shows normal wildtype growth, produces spores and maintains an epigenetic landscape similar to the wildtype. Chen et al. substantially simplified the genome without discernible phenotypic effects, implying that many transposable elements may minimally impact growth. They also introduced other sequence modifications, such as PCR tags, gene locus swapping and stop codon substitutions. The results lay the foundation for the synthetic moss genome project (SynMoss). The next phase of the project will focus on the complete replacement of chromosome 18 with chemically synthesized sequences, followed by full synthetic genome replacement. 

Open-ended molecular recording of sequential cellular events into DNA (Loveless et al. 2025)

The authors present the DNA recorder peCHYRON (prime editing Cell HistorYRecording by Ordered Nucleotide insertion), that inserts durable mutations in a cell´s genome upon transient biological events, that can later be reconstructed by DNA sequencing. The system uses a prime editor, a nickase Cas9-reverse transcriptase fusion that can insert precise mutations specified by a prime editing guide RNA (pegRNA). peCHYRON functions as follows: a pegRNA targets a 20 bp locus in a cell´s genome and inserts a variable 3-nt sequence encoding up to 6 bits of information together with a 17-nt propagator sequence that serves as the target for the next insertion step. After each insertion, the previous propagator sequence is no longer adjacent to the PAM-sequence and thus inactive, while the new propagator sequence is adjacent to the PAM and active. Each editing step inserts a new 3-nt sequence with the new propagator sequence as a record of an event. Insertions are sequential and can be propagated without end. This allows the temporally resolved recording of multiple cellular signals in mammalian cells. The authors showed that the constitutive expression of pegRNA collections generates insertion patterns for the straightforward reconstruction of cell lineage relationships whereas the inducible expression of specific pegRNAs results in the accurate recording of exposures to specific biological stimuli.

Convenient synthesis and delivery of a megabase-scale designer accessory chromosome empower biosynthetic capacity (Ma et al. 2025)

Reconstructing complex traits in new host organisms using large-scale genetic information remains a significant challenge. In their work Ma et al. developed a CRISPR/Cas9-mediated haploidization method that bypasses the natural process of meiosis. Based on the programmed haploidization in yeast, they further developed a method designated HAnDy (Haploidization-based DNA Assembly and Delivery in yeast) that enables efficient assembly and delivery of large synthetic DNA, with no need for elaborate in vitro manipulations. Using HAnDy, a de novo designed 1,024 Mb synthetic accessory chromosome (synAC) encoding 542 exogenous genes was parallelly assembled and then directly transferred to six phylogenetically diverse yeasts. The synAC enhances host adaptation and expands the metabolic network, enabling the production of valuable compounds. This approach should facilitate the assembly and delivery of largescale DNA for expanding and deciphering complex biological functions.

Building synthetic chromosomes from natural DNA (Coradini et al. 2023)
The authors report the method CReATiNG (Cloning, Reprogramming, and Assembling Tiled Natural Genomic DNA) for constructing synthetic chromosomes from cloned segments of natural DNA in the yeast S. cerevisiae. This is a faster and cheaper alternative to de novo chromosome synthesis, and could be used in research when complete chromosome reprogramming is not required. In essence, natural chromosome segments are cloned such that unique adapter sequences are appended to their termini, specifying how these molecules will recombine with each other later when they are assembled. Then, cloned segments are co-transferred into cells and assembled by homologous recombination. The method was used to synthetically recombine chromosomes between different strains and species, to modify chromosome structure, and to delete many linked, nonadjacent regions accounting for 39 % of one chromosome. The multiplex deletion experiment reveals that CReATiNG also enables correcting flaws in synthetic chromosome design via recombination between a synthetic chromosome and its native counterpart.

Continuous synthesis of E. coli genome sections and Mb-scale human DNA assembly (Zürcher et al. 2023)
The authors developed the tool BASIS, Bacterial artificial chromosome (BAC) Stepwise Insertion Synthesis, a method for megabase-scale assembly of DNA in Escherichia coli episomes. They used BASIS to assemble 1.1 Mb of human DNA. BASIS provides a powerful platform for building synthetic genomes for diverse organisms. The authors also developed the method CGS, Continuous Genome Synthesis, useful for continuously replacing sequential 100 kb stretches of the E. coli genome with synthetic DNA; CGS minimizes crossovers between the synthetic DNA and the genome in a way that the output for each 100 kb replacement provides without sequencing the input for the next 100 kb replacement. Using CGS, a 0.5 Mb section of the E. coli genome was synthesized constituting a key intermediate in its total synthesis from five episomes in 10 days. By parallelizing CGS and combining it with rapid oligonucleotide synthesis and episome assembly and rapid methods for compiling a single genome from strains bearing distinct synthetic genome sections the authors anticipate that it will be possible to synthesize entire E. coli genomes of functional designs in less than two months.

Synthetic Yeast Genome Project (Sc2.0)
In this project a global consortium is working to develop the first synthetic eukaryote genome from scratch having a number of specific molecular tags. The final goal of the project is to assemble a fully synthetic yeast organism and to facilitate synthetic biology and engineering research in eukaryotes. Status in 2023: a total of 14 Sc2.0 chromosomes plus a bonus tRNA neochromosome were generated, with two more chromosomes to be constructed. Here, you can find general information and the current status of the Sc2.0 project. In a cluster of papers dedicated to individual synthetic chromosomes, the Sc2.0 consortium describes new findings on aneuploidy, extrachromosomal DNA regulation, chromosome fusion, and many other aspects of yeast genome biology (Blount et al., Foo et al., Lauer et al., Shen et al., Williams et al., Luo et al.). Highlights in this collection include the creation of a yeast strain containing seven synthetic chromosomes, which equals ~ 50 % of the yeast genome, that functions similarly to wild-type yeast (Zhao et al. 2023), a strain with all tRNA genes re-located to an entirely synthetic tRNA neochromosome (Schindler et al. 2023), and the re-design and 3D structural genomic characterization of the largest yeast chromosome (Zhang et al. 2023). 

Multiplex base editing to convert TAG into TAA codons in the human genome (Chen et al. 2022) the authors take the first steps in whole-genome recoding in human cells by demonstrating that the stop codon TAG can be simultaneously converted into TAA in dozens of genes in a single transfection experiment. One goal of whole-genome recoding is to generate virus-resistant cells that could be applicable for biomedicine, especially for making cell therapies or therapeutic production lines resistant to most natural viruses.
The approach is based on previous Escherichia coli recoding projects in which all 314 TAG stop codons were replaced with TAA codons genome-wide (Isaacs et al., 2011). To adapt the approach to the size of the human genome, the authors developed a Python-based software called Genome Recoding Informatics Toolbox (GRIT) that is tailored to recoding and can automate the process of part design. GRIT identified a total of 6700 TAG codons in the human genome of which 1937 sites are located in essential genes and are editable using cytosine base editors.
For the multiplex base editing strategy gBlocks, with each gBlock containing five individual gRNA cassettes, were designed and synthesized. The strategy was optimized by using single-cell RNAseq. Applying 10 gBlocks at a time the authors were able to edit 33 of 47 target sites within the HEK293T cells in a single transfection event while observing ~40 C-to-T off-target events. Nonetheless, this work demonstrates the feasibility of TAG to TAA conservations in the human genome and provides a framework for large-scale engineering of mammalian genomes.

Imaging cell lineage with a synthetic digital recording system (Chow et al. 2021) the authors developed a system to record different cell lineages in higher organisms by permitting an imaging-based in situ-readout. They used site-specific serine integrases such as Bxb1 for their “integrase-editable memory by engineered mutagenesis with optical in situ readout (intMEMOIR)” system. The cells were marked with a genetic “barcode” flanked by attP and attB sites that allow Bxb1-recombination leading to either a deletion or an inversion of the barcode. The cells´ state can be read out using fluorescence in situ hybridisation (FISH). Ten orthogonal attP and attB sites can be used in the same organism allowing for the recording of 310 states. To test the system in vivo the authors created the Drosophila line memoiphila, which harbours 10 barcodes and whose neuronal cells can be integrase-edited when heat-shocked. Editing was induced four hours after egg laying so that neuroblasts labelled with distinct arrays can pass on their editing to all progeny in the adult brain. Adult flies were then dissected and their brains subjected to several rounds of FISH analysis reading out the induced editing as well as the expression of eight endogenous genes that mark different neuronal cell types. The combination of gene expression and cell clones identified by barcode editing allowed to determine several known cell types. The possibility to visualize the cell lineage directly in the native tissue could provide insights into the roles of different factors in cell fate development.

Belcher et al. (2020) built libraries of synthetic promoters and corresponding transcription factors (TFs) for plants. These shall enable a more precise control of transgene expression and thus provide for easier engineering in plants. The authors started with a plant minimal promoter to which they added different concatenated cis-elements from the well-characterised yeast GAL4 regulon. These cis-elements are bound by the GAL4 TF. The cis-element-containing promoters showed different expression strengths that allowed to be used for tuning gene expression. 25 such promoters were successfully tested in vivo in Arabidopsis thaliana. The number of TF was then expanded to TFs from other families and their respective cis-elements. Finally, transactivators and repressors were added to the promoter library by fusing known transactivation domains from Zea mays or Herpes simplex virus type 1 or the repressor domain SRDX to the TFs or truncated versions thereof that consisted only of the DNA-binding domain. The hybrid promoters bind to different TFs coupled to different activator or repressor domains, thus allowing multi-gated logic operations in plant cells. - to the original literature

Kotopka et al. (2020) established an in silico based method to generate artificial promoters for Saccharomyces cerevisiae. In order to generate the basic data for this method, conserved motifs from known promoters in combination with randomized spacer sequences were used to create two plasmid based libraries in S. cerevisiae. These libraries comprise over 675,000 constitutive and over 327,000 promoters inducible by an artificial transcription factor (ZEV). The yeast libraries were analysed by fluorescence-activated cell sorting combined with high-throughput DNA sequencing (FACS-seq). In the next step an in silico model was build, predicting promoter activity as a function of sequence. By implementing a convolutional neural network (CNN) as a deep learning technique, the model was able to handle the very large data sets. To validate the predictions made with this model, sets of artificial promoters were generated in silico with three different sequence designs: 1) screening: generation of random sequences based on the original libraries, 2) evolution: mutagenesis of specific sequences and 3) gradient ascent: iterative modification of initially random sequences. In addition, a GC constraint was implemented when estimating promoter activities. These sets of artificial promoters were then tested against control-promoter-sets by transformation of yeast and FACS-seq analysis. The analysis revealed that screening and evolution strategies produced promoter sets with comparable diversity to the strongest promoters in the initially generated data. The gradient ascent promoter strategy led to a loss in diversity. Further optimisation of evolution and gradient ascent strategies finally led to high-performing promoter sequences even when applying a GC content constraint. This model can be used to not only generate sequences with useful rare properties but also large and sequence-diverse sets of promoters exhibiting high activities. - to the original literature

Lin et al. (2020) established a single strand (ss) double strand (ds) DNA architecture for DNA-based dynamic operations and reusable information storage (DORIS). Systems to access information from DNA data storage should heed three basic criteria: 1) scalability, 2) compatibility with efficient and dense encodings and (3) reusability. In order to meet these criteria dsDNA with a single-stranded overhang (ss-dsDNA) were generated by PCR and hybridization. The overhang sequence („file address“, 20 nt) was followed by the dsDNA containing the T7 promotor (23 nt) and the encoded data (data payload, 117 nt) were designed and generated by PCR. In this design, all strands encoding for one specific file have the same „file address“. By using oligonucleotides complementary to the „file address“ and coupled to magnetic beads, it was possible to separate file related strands out of a pool of mixed DNA molecules under room temperature conditions. In contrast to PCR-based separation methods, the data payload remains annealed under these conditions, blocking any undesired oligonucleotide binding to any similar sequences in the data payload regions. Since unwanted binding is blocked, the sequence of the data payload can be more flexible, thereby increasing density (information per nt) and capacity (storage maximum of the system) of DORIS. To perform as a storage device, DORIS was designed to be capable of in-store file operations such as locking, unlocking, renaming, and deleting files. In a proof of principle, high temperatures (98°C) were used to place a so-called lock, a 20 nt long complementary oligonucleotide, on the sequence of the „file address“. To unlock the system, a key-oligonucleotide, which is complementary to the lock, was added. Renamings of files were performed by oligonucleotides consisting of the merged sequence of the new „file address“ and the old „file address“, creating a new strand overhang. Oligonucleotides blocking the „file address“ were used to delete files from the storage. These in-storage modifications combined with the increased density and capacity of the system are a first step towards DNA based highly parallel processing of extreme levels of information (e.g. medical, genomic and financial data). - to the original literature

Fredens et al. (2019) have synthesized the biggest synthetic genome known so far, a 4 Mbp Escherichia coli-genome. They used the REXER (replicon excision for enhanced genome engineering through programmed recombination) technique to exchange the whole length genome of E. coli MDS42 with a synthetic genome. To create this synthetic genome, only 61 of the possible 64 codons were used in protein-coding genes. Two codons for serine (TCG, TCA) and one stop codon (TAG) were replaced with synonymous codons, resulting in the change of 18,214 codons in total. - to the original literature

Wang et al. (2019) developed and tested a system for the precise allocation of the unmodified genome of E. coli to defined, circular artificial chromosomes. The subdivision of E. coli genome thereby took place in the cell by means of Cas9, the lambda-red recombination system and an artificial bacterial chromosome (BAC), which functioned as an acceptor for a part of the genome. A reduced genome-chromosome and a BAC chromosome with the residual genome thus resulted. The subdivision of the genome had only a small influence on the growth behaviour of cells and remained stable across generations. Additional modifications of the system permitted undertaking targeted inversions or translocations in the separated genome and to combine defined sections of genomes of various E. coli strains with one another. As a result, a system was created that made rapid, precise engineering of large genomes possible. - to the original literature

Quantum spin resonance in engineered proteins for multimodal sensing (Abrahams et al. 2026) 

The authors have developed magneto-sensitive fluorescent proteins (MFPs), i.e. proteins whose fluorescence changes in response to an external magnetic field. The proteins are derived from a LOV2 (light-oxygen-voltage) domain protein, a natural light sensor originally found in plants, where it ensures that the plant grows towards the light. MFPs function as quantum sensors by allowing magnetic fields and free radicals to be measured directly within a single living cell via changes in a fluorescence signal. This function is based on processes that depend on electron spin or nuclear spin effects, i.e. on the intrinsic angular momentum of an electron or the total angular momentum of an atomic nucleus. The signal, an optically detected magnetic resonance (ODMR), is thought to arise from an interaction between the LOV2 domain and a flavin cofactor chromophore (a vitamin molecule that absorbs light). Specifically, the two components share two electrons. The combination of these two electrons is referred to as a radical pair. The two electrons spin like tiny compasses and, as they are linked, they influence one another (they are spin-correlated). An external magnetic field disrupts this rotation. This causes the protein’s luminescence to change, which can be detected using a simple fluorescence microscope. The MFPs were specifically improved through directed evolution. By combining different variants of MFPs, various biological processes can be measured simultaneously. The method works in three-dimensional space, which means it is possible to pinpoint, for example, exactly where a signal is located within a cell, and whether free radicals are present in the cell. MFPs can thus serve as biological sensors, indicating changes in the magnetic field via a light signal, or as biological switches, whereby targeted reactions within the cell are triggered by altering the magnetic field. 

Programmable trans-splicing riboregulators for complex cellular logic computation (Gao et al. 2025) 

The authors developed a novel class of riboregulators called split-intron-enabled trans-splicing riboregulators (SENTRs). Riboregulators are RNA molecules that act as molecular switches to control gene expression in cells. SENTRs exploit the catalytic activity of group I introns to reassemble separately expressed RNA fragments into a functional RNA through trans-splicing. The system is based on the well-characterized group I intron from Tetrahymena thermophila (TT intron), which was split into two halves. These halves were fused to synthetic external guide sequences (EGSs) that promote the assembly of the RNA fragments and thereby trigger the trans-splicing reaction.

The functionality of SENTRs was evaluated using a fluorescence assay in which the split intron was inserted into the sfgfp (superfolder GFP) gene. For this purpose, a library of 56 EGS variants was generated and characterized. The two best-performing variants achieved more than 10,000-fold activation while exhibiting extremely low background activity. To improve the predictability of SENTR performance, the authors analysed numerous EGS combinations and used the resulting dataset to train machine-learning models. In addition, orthogonal SENTRs were developed that function largely independently of one another, making them suitable for the construction of complex genetic circuits.

The authors further demonstrated that SENTRs can be flexibly positioned at different locations within genes. Moreover, they were employed to regulate a CRISPR activation system and were engineered into RNA sensors capable of detecting specific intracellular mRNAs. SENTRs were also applied to construct biological logic circuits. Using SENTRs, various two-input logic gates (AND, NAND, OR, NOR, IMPLY, and NIMPLY) as well as more complex three- and four-input systems were implemented. By combining SENTRs with split intein-mediated protein trans-splicing (Pinto et al. 2020), the authors constructed six-input logic circuits. SENTRs have the potential to be applied in a variety of organisms, including bacteria, yeast, and mammalian cells, making them a versatile tool for post-transcriptional gene regulation.

Electromagnetic wireless remote control of mammalian transgene expression (Lin et al. 2025)

The authors developed a wireless electronic device for minimal invasive biomedical applications that is controlled via an electromagnetic field. To achieve this, they combined specialized nanoparticles with genetically engineered cells. The nanoparticles respond to the electromagnetic field and generate reactive oxygen species (ROS), which serve as a signal to activate the cells. 

The engineered cells are designed to respond to ROS through a cellular signaling pathway in which the ROS-sensitive Kelch-like ECH-associated protein 1 (KEAP1) releases the transcription factor nuclear factor erythroid 2 p45-related factor 2 (NRF2). NRF2 then binds to synthetic antioxidant-response elements in promoter regions, thereby initiating the expression of target genes. 

The system was validated for blood-glucose management of type 1 diabetes in mice. The engineered cells were encapsulated with the nanoparticles into alginate microcapsules to shield the cells from the hosts immune system. Upon exposure to an electromagnetic field, the cells produced insulin as a therapeutic target protein and blood-glucose levels in the mice decreased. 

Engineering yeast multicellular behaviors via synthetic adhesion and contact signaling (Meng et al. 2025)

The authors developed modular toolkits to establish contact-dependent signaling and specific cell-cell adhesion in yeast multicellular systems. Contact-dependent signaling is achieved via surface-displayed peptides and G protein-coupled receptors and termed mating-peptide anchored response system (MARS), while cell-cell adhesion is done with adhesion-protein pairs (termed SATURN, Saccharomyces Adhesion Toolkit for mUlticellular patteRNing). This resembles juxtracrine signaling, which requires direct cell-cell contact. Multicellular logic circuits were designed with the two systems combined and a genetic sensor for probing protein-protein interaction and the selection of high-affinity nanobody binders was created. 

For MARS, display and sensor cells were generated that display a signal peptide and a G-protein coupled receptor on their cell surfaces. The strain used for both cell types was engineered with 17 gene knockouts to abolish natural cell-cell adhesion. When the cells displaying the GCPR recognize the membrane-bound signal peptide, a phosphorylation cascade starts that leads to the expression of a fluorescent protein. Eight orthogonal MARS pairs were created. A multicellular logic chain based on MARS was created, starting with a display cell that activates a sensor cell, which then expresses and displays a new MARS peptide. Signal transmission was achieved across chains of up to six sequentially connected cells. Chains could also be programmed for logic operations such as OR, NOT or AND. 

For SATURN, a set of high-affinity protein pairs, including SpyTag-SpyCatcher, was displayed on different yeast cell populations. The cells were created based on the MARS strain background, which contains 17 gene knockouts and was further engineered with six additional CRISPR-mediated gene deletions removing remaining adhesin genes and the FLO genes (genes encoding flocculation proteins and adhesins). Eight pairs were identified that are orthogonal. 

SATURN and MARS were then combined so that yeast cells interacted via SATURN and then showed altered gene expression via MARS signaling. The two systems were also combined to study protein-protein interaction, where the binding strength of two proteins of interest is linked to intracellular gene expression. Display and sensor cells then co-display a protein-of-interest together with the signal peptide or the G-protein coupled receptor. The system was applied to screen antigen-nanobody interactions. All systems could be used to construct user-defined multicellular yeast systems in the future.

Engineered receptors for soluble cellular communication and disease sensing (Piraner et al. 2025)

The authors modified the so-called synthetic intramembrane proteolysis receptor (SNIPR) (Zhu et al. 2022), a receptor that can be activated not only by membrane-bound ligands, but also by soluble ones. SNIPR uses domains from the regulated intramembrane proteolysis proteins Notch and Robo1 coupled to a ligand binding domain such as single-chain variable fragment or nanobody and a transcription factor that is activated when a ligand is bound. It binds natural and synthetic soluble factors. SNIPR was shown to localize chimeric antigen receptor T cells to solid tumours expressing soluble disease-associated factors. Furthermore, the authors developed orthogonal synthetic signalling networks for cellular communication and environmental interactions. SNIPRs could be used in cells for therapeutic purposes or to study biological interactions.

Synthetic protein circuits for programmable control of mammalian cell death (Xia et al. 2024)

The authors built a genetic circuit that induces cell death in mammalian cells. Cells can regulate cell death amongst others via either apoptosis or pyroptosis. Unlike apoptosis, which does not alert the immune system, pyroptosis releases damage-associated molecular patterns that stimulate the immune system and cause inflammation. The circuit termed “synpoptosis” circuit was built using modified apoptosis- (caspase-3) and pyroptosis (gasdermin)-inducing proteins. Both proteins were modified for protease-controlled expression/repression and showed the desired action in different cell types. The synpoptosis circuit was also modified to selectively recognize and eliminate specific target cells or as a tool for the establishment of synthetic killer cells. For this purpose, sender cells were engineered to express virus-like particles containing the synpoptosis circuit to enter and eliminate receiver cells. The synpoptosis circuit is thus a step forwards to programmable control of mammalian cell death, for example in a future context of treating cancer, autoimmune diseases and infection.  

Programmable mammalian translational modulators by CRISPR-associated proteins (Kawasaki et al. 2023)
The authors repurposed Cas proteins as translational modulators in mammalian cells to build artificial logic circuits. A Cas-binding motif was introduced into the 5‘ untranslated region of a gene of interest. When the respective Cas-protein was present in the same cell, the translation of the gene of interest was repressed by the binding of Cas to the binding-motif present in the mRNA. They also constructed an ON switch, in which the Cas-protein binding prevents the mRNA from non-sense mediated decay. Using a split-Cas9, genetic logic gates (NAND) were constructed as well.

Engineered bacterial swarm patterns as spatial records of environmental inputs (Doshi et al. 2023)
The authors have modified Proteus mirabilis, a bacterium that naturally forms centimetre-sized bullseye swarm patterns on solid agar, to respond to specific environmental stimuli and form characteristic swarm patterns. These swarm patterns serve as a visual representation of environmental conditions and allow complex environmental information to be captured and visualised in a simple manner. Specifically, a P. mirabilis strain was developed that visually recorded the presence of up to 50 mM copper in water samples, with a graded change from 0 mM to 50 mM in the ring widths and colony radii at the spot locations. This work provides an approach for the construction of macro-scale bacterial recording devices.

Tailored Synthetic sRNAs Dynamically Tune Multilayer Genetic Circuits (Velazquez Sanchez et al. 2023)
The authors have designed novel small RNAs (sRNAs) that dynamically modulate gene expression of genetic circuits with a broad spectrum of repression (high, medium, and low), by utilising the intrinsic RNA interference pathway in E. coli. The sRNAs were designed to bind to their respective target sequences with different binding affinities, whereby the binding energy correlates with the repression strength. The sRNAs do not bind directly to the mRNA of the target gene, but to the start codon of the mRNA of a transcription factor that controls the corresponding target gene. The results suggest that gene expression can be dynamically modulated across different conditions by incorporating synthetic sRNA to existing genetic constructs without modification of the genetic parts, e. g. exchanging promoters.

Co-opting signalling molecules enables logic-gated control of CAR T cells (Tousley et al. 2023)
The authors developed a chimeric antigen receptor (CAR) engineering approach in which the traditional CD3ζ domains for T cell activation were replaced by intracellular proximal T cell signalling molecules. It is shown that some proximal signalling molecules such as ZAP-70 and PLCγ1 after their conversion as surface receptors are themselves sufficient to initiate CAR T cell signalling and eradicate tumours in vivo without the need for CD3ζ. The main role of ZAP-70 is to phosphorylate LAT and SLP-76, which form a scaffold for signalling propagation. The co-operative role of LAT and SLP-76 has been exploited to develop a logic-gated intracellular network (LINK) CAR, a rapid and reversible Boolean-logic AND-gated CAR T cell platform that outperforms other systems in both efficacy and prevention of on-target, off-tumour toxicity. LINK CAR will broaden the range of molecules that CAR T cells can target and will enable these powerful therapeutics to be used in solid tumours and various diseases such as autoimmunity and fibrosis.

Engineering a modular double-transmembrane synthetic receptor system for customizing cellular programs (Zhou et al. 2023)
The authors developed a synthetic receptor system based on two receptor domains, one of which carries a protease and the other a synthetic transcription factor (TF) which is released after cleavage by the protease. The receptor chains are located at the extracellular side and dimerize after binding a ligand, whereupon the protease is able to release the TF, which then can transcribe the customised downstream genes. The receptor system was further engineered to be sensitive to extracellular protein signalling with minimal background signals and positive loop switching. It is shown that this synthetic receptor system can be easily customised to respond to various inputs such as interleukin-1 (IL-1), programmed death ligand 1 (PD-L1) and HER2 and to release tailored outputs, including fluorescent signals and the therapeutic molecule IL-2.

Conditional Control of Universal CAR T Cells by Cleavable OFFSwitch Adaptors (Kvorjak et al. 2023)
The authors have refined universal chimeric antigen receptors (CARs). Universal CARs provide better control over T cell function by not binding the target antigen directly compared to conventional CAR T cells, but instead the CAR binds to a co-administered antibody adaptor that is specific for the target antigen. The authors have further improved the universal CARs by developing OFF-switch adapters. These carry a cleavable biotin tag that can be removed and thus controlled by the addition of a small molecule (phosphine-2-(diphenyl-phosphanyl)-benzamide (2DPBM)) or a light stimulus (UV light). In addition, OFF-switch adapters were able to simultaneously control several antigens orthogonally according to Boolean logic in adaptor combination tests. OFF-switch adaptors represent a new approach for the precision targeting of universal CAR T cells with potential for enhanced safety.

A synthetic switch based on orange carotenoid protein to control blue-green light responses in chloroplasts (Piccinini et al. 2022) the authors have developed a synthetic orthogonal photoreceptor in plant chloroplasts whose spectral range does not overlap with endogenous plant photoreceptors. They split the orange carotenoid protein (OCP) 2 from cyanobacterium Fischerella thermalis, which is activated by blue-green light, into two domains. The domains can join when a prosthetic keto-carotenoid is present and their interaction was detected by fusing them to split nanoluciferase fragments that produce a fluorescent signal when interaction occurs. The photoreceptor was functional in Arabidopsis thaliana plants that also expressed the bacterial enzyme β-carotene ketolase from Agrobacterium aurantiacum in its chloroplast to provide the prosthetic keto-carotenoid. The system was then applied to regulate transcription of plastid genes by fusing the OCP2 domains to the Sigma2 factor, which is part of the plastid-encoded polymerase, and an anti-sigma factor from T4 phage. That way, the transcription of a plastid gene was inhibited when the two OCP2 domains joined in the dark and was activated by exposure to green light. The photoreceptor can be used to control gene expression in prokaryotes and eukaryotes upon blue-green light exposure.

Photoswitching of feedback inhibition by tryptophan in anthranilate synthase (Bhagat et al. 2022) the authors created a system to control feedback inhibition of an enzyme by light. In feedback inhibition an enzyme of a biosynthetic pathway is inhibited when the end-product binds to an allosteric site. Here, the anthranilate synthase from Salmonella typhimurium was used as a proof-of-concept. The enzyme catalyses the first step in the tryptophan-producing pathway and its synthase subunit TrpE is feedback inhibited by tryptophan. The authors inserted the bulky noncanonical amino acid o-nitrobenzyl-O-tyrosine (ONBY) into TrpE by expanding the genetic code. ONBY possesses an o-nitrobenzyl caging group that hinders the binding of tryptophan. This inhibition can be reversed by irradiation with UV that removes the o-nitrobenzyl and allows tryptophan binding to its allosteric site. Such systems can be used for light-dependent activation or inactivation of enzymes in different biological applications such as biotherapeutics.

Synthetic genetic circuits as a means of reprogramming plant roots (Brophy et al. 2022) the authors have developed a set of transcriptional regulators to build genetic circuits in plants, that will allow the specific expression of genes in different cell types to control plants´ responses to a changing environment. They constructed transcriptional activators, repressors and activatable and repressible synthetic plant promoters. Activators were composed of a DNA-binding domain, an activation domain and a nuclear localisation signal (NLS) and repressors of DNA-binding proteins and a NLS. For an activatable promoter one to six copies of the DNA sequence bound by the transcription factors were fused to a minimal cauliflower mosaic virus (CaMV) 35S promoter, the repressible promoter was made up of a full-length CaMV 35S promoter and one DNA sequence fused to its 3´ end. In the model plant Arabidopsis thaliana the synthetic input transcription factors were placed under control of different tissue-specific promoters allowing the circuits to produce different spatial pattern of gene expression in the plant root. As an application, the authors modified the plants´ lateral root branch density. They designed a logic gate to express a mutated gene called slr-1 that eliminates root branching only in lateral root stem cells where the generally negative side-effects of this mutation are not observed. The gates allowed for an expression of the mutated gene at varying levels and thus for a predictable lateral root development.

Spatiotemporal control of engineered bacteria to express interferon-γ by focused ultrasound for tumor immunotherapy (Chen et al. 2022) the authors have developed a system to control the spatiotemporal expression of therapeutic genes in bacteria that target and reduce tumors. They introduced the gene for the cytokine interferon-γ (IFN-γ) into Escherichia coli MG1655 and placed it under control of the temperature-regulated leftward and rightward phage lambda promoter. When delivered into tumor-bearing mice the bacteria can be heated with ultrasound to stimulate the expression of IFN-γ. The ultrasound-activated bacteria can penetrate into deep-seated tumors and are non-invasively activated by ultrasound.

Tunable and modular miRNA classifier through indirect associative toehold strand displacement (Chen & Chen 2022) the authors use microRNAs (miRNAs) as inputs for logic-gated classifiers. They refined the catalytic hairpin assembly (CHA), in which two complementary strands are prepared as hairpins so that their interaction is blocked. If a single-stranded nucleic acid such as a miRNA is added, the two hairpins are opened successively. The strand displacement is detected by a reporter duplex consisting of a fluorophore and a quencher strand, that will emit fluorescence if the fluorophore strand is displaced by the opened hairpin. By changing the lengths of toeholds and adding variable clamps to the hairpin strand, any miRNA could be adapted to the system. The system was designed as AND, NOT, and ANDNOT gates with up to four inputs and can be used as a simple diagnostic assay.

Computing within bacteria: programming of bacterial behaviour by means of a plasmid encoding a perceptron neural network (Becerra et al. 2022) the authors showed that trained neural networks can be implemented in silico in a colony of synthetic bacteria. Without performing wet-lab experiments, they programmed bacteria with different neural networks either solving an optimization problem or responding to changing environmental conditions. The neuronal network was first tested offline, implemented into a genetic circuit using the programming language Gro and finally integrated into a plasmid using the Cello platform and the programming language Verilog. The authors conclude that the use of algorithms from computer science in synthetic biology will give bacteria artificial intelligence to solve a range of different problems from medicine to bioremediation.

Orthogonal control of gene expression in plants using synthetic promoters and CRISPR-based transcription factors (Kar et al. 2022) the authors have constructed three mutually orthogonal synthetic plant promoters. They used a minimal Cauliflower mosaic virus (CaMV) 35S promoter and added synthetic guideRNA-binding sequences that can be activated through a guideRNA and a dCas9:VP64 transcription factor. When putting gRNA expression under control of the endogenous signalling molecule ethylene, circuits can be tied to the cellular metabolism. Plant expression vectors to test the promoters were constructed with the modular cloning framework MoClo and tested in Nicotiana benthamiana and Arabidopsis thaliana. The orthogonal promoters can be used for computing operations in plants limiting crosstalk with endogenous pathways and could also be applied to non-model plants.

Developments of mammalian cell logic gates controlled by unnatural amino acids (Mills et al. 2021) the authors used genetic code expansion in mammalian logic gates. They expanded the genetic code by incorporating unnatural amino acids (UAA) into proteins by using a quadruplet genetic code. Therefore, they screened 11 quadruplet-decoding pyrrolysyl tRNA variants described in the literature for Escherichia coli for their function in mammalian cells and identified the variants decoding CUAG or AGGA as functional. For the construction of a double-input logic gate, an E. coli derived tyrosin aminoacyl-tRNA synthetase/tRNA pair was identified as a second orthogonal pair. Both pairs (PylRS/tRNA and TyrRS/tRNA) functioned orthogonally in the same cell and incorporated their specific UAA. The two codons for the pairs were inserted in two split GFP polypeptides that need to assemble for fluorescence to construct AND and OR gates in which either both or at least one of the UAA has to be incorporated to reconstitute the GFP fluorescence. These gates are an alternative approach for logic gate construction in mammalian cells and employ biologically inert molecules.

A logically reversible double Feynman gate with molecular engineered bacteria arranged in an artificial neural network-type structure (Srivastava and Bagh 2022) the authors built an artificial neural network (ANN) that acts as a double Feynman gate. The double Feynman gate has three inputs (X1, X2, X3) and three outputs (O1, O2, O3) and one input leads to one defined output. O1 is a control and is defined as O1 = X1. O2 is defined as X1 + X2 and O3 as X1 + X3. The in- and outputs are calculated as logical operations in so-called truth tables that address a weight to each individual input/output combination and together built five artificial neuro-synapses. These logic operations were implemented into Escherichia coli bacterial cells as bactoneurons where extracellular chemical inducers act as inputs and fluorescent proteins as outputs. Weights are applied as gene activation or repression of synthetic promoters and biases were added. The authors constructed and co-cultured five bactoneurons (bacterial cultures) and obtained a double Feynman gate function at the population level. This kind of ANN-type architecture can be used for cellular computation operations.

Synthetic multistability in mammalian cells (Zhu et al. 2022) the authors created multistability, in which genetically identical cells can exist in molecularly distinct and mitotically stable cell states, with the help of a circuit architecture composed of transcription factors (TF). These TF can heterodimerize competitively and will only activate their own genes as homodimers (auto-activation). The authors first developed a set of zinc finger transcription factors with homodimer-dependent self-activation and heterodimer-dependent inhibition. When stably integrated into mammalian cell lines, the cells would express either one of the TF or both thus exhibiting three different states. The states could be changed by adding external inducers. When protein stability was reduced, the cells transitioned from tristable to bistable. The system was then expanded by adding a third transcription factor, which resulted in cells exhibiting seven distinct states. An expansion beyond three TF was modelled and seems to be possible. The paper is a step towards understanding and using multistability, for example for engineered cell therapies.

LED control of gene expression in a nanobiosystem composed of metallic nanoparticles and a genetically modified E. coli strain (Aratboni et al. 2021) the authors have established a photothermal control of gene expression using non-toxic metallic nanoparticles that convert light into heat. Free conducting electrons in gold nanoparticles (AuNPs) oscillate upon illumination and part of this energy is emitted as heat increasing the temperature of the surrounding medium. The authors activated AuNPs by a light-emitting diode in microorganisms expressing the fluorescent protein mCherry under control of a U6 RNA thermometer. RNA thermometers can be naturally found in the 5´-untranslated regions of mRNAs and control the expression of downstream genes by temperature-induced changes in their conformation. At low temperatures the RNA thermometer masks the ribosome binding site, at higher temperatures the RNA secondary structure melts locally and allows for ribosome binding. It was shown that mCherry was only expressed at room temperature when the AuNPs were illuminated. This work provides another option to control gene expression in biological systems by illumination.

De novo design of a reversible phosphorylation-dependent switch for membrane targeting (Harrington et al. 2021) the authors developed a minimal de novo peptide-based molecular switch that is based on a heterodimeric coiled-coil assembly and changes between monomer and dimer upon phosphorylation and dephosphorylation. When coupled to a membrane-targeting peptide the switch can shift between a membrane bound and a solution state. Such a switch can be used for transcriptional regulation and construction of orthogonal, protein-based circuits.

T cell circuits that sense antigen density with an ultrasensitive threshold (Hernandez-Lopez et al. 2021) the authors designed T cell circuits to prevent „off-tumour“ toxicities, frequently occurring during cancer treatment with chimeric antigen receptor (CAR)-T cells. The authors envisioned CAR-T cells which reliably distinguish cancer cells from normal cells by antigen density and enable an increased cancer cell killing. Using the human epidermal growth factor receptor 2 (HER2) as a target, ultrasensitive T cells with a two-step recognition circuit were designed by combining a synNotch receptor with a CAR. The synNotch receptor detects the antigen with low affinity, thereby acting as a high-(antigen)density-filter. When the synNotch receptor becomes activated it induces the expression of a high-affinity CAR which upon activation induces T cell proliferation and T cell mediated killing of tumour cells. The designed T cells were tested on human leukaemia (K562) cells, expressing different amounts of HER2, for their ability to discriminate between cells expressing >10^6.5 and 10^4.5 HER2 molecules per cells. These HER2-expression levels correspond to those on HER2-expressing cancer cells and normal HER2-expressing tissues respectively. The T cells were able to meet the designated discrimination parameters. In a mixed culture, with high-density (10⁷) and low-density (10^4.8) HER2 cells, the designed T cells selectively eliminated cells with the high HER2 density, leaving the low-density cells unharmed. To evaluate how these T cells behave in a more complex environment, immunocompromised mice were injected subcutaneously with a high-HER2-density K562 tumour in one flank on the one side and a low-HER2-density K562 tumour in the other flank. After the tumours were established, the designed T cells were administered and showed a strong discrimination: while high-density tumours were cleared from mice, low-density tumours retained growth. This ultrasensitive antigen-density discrimination provides a very important tool for treating solid tumours with T cells.

Spatiotemporally confined red light-controlled gene delivery at single-cell resolution using adeno-associated viral vectors (Hörner et al. 2021) the authors engineered an adeno-associated viral (AAV) vector system for the transfer of genetic information into cells controlled by illumination with red light. AAV vectors transduce both dividing and non-dividing cells and are widely used as gene therapy vehicles. The light-control of the systems controls transduction at the level of cell entry. The AAV vector is engineered in a way that it cannot recognize its natural cell entry receptor heparin sulfate proteoglycan, but expresses the phytochrome-interacting factor 6 (PIF6) from Arabidopsis thaliana. To enable interaction with the target cell an adapter protein is needed. The adapter protein consists of phytochrome B of A. thaliana and a designed ankyrin repeat protein specific for a cell surface protein. Upon illumination with red light, the phytochrome B part of the adapter protein interacts with PIF6 on the viral vector to recruit the vector to the target cell and initiate transduction. Illumination with far-red light dissociates the viral vector from the cell. By switching the adapter protein different cell types can be targeted and the system could easily be applied, for example, to in vivo gene therapy.

An engineered protein-phosphorylation toggle network with implications for endogenous network discovery (Mishra et al. 2021) the authors developed a bistable toggle switch consisting of reversible protein-protein phosphorylation interactions. The switch contains two branches that repress each other and two inputs to switch between the two states of the system. The protein network is built in S. cerevisiae from eleven proteins either stemming from the endogenous MAKP-pathway or being constructed as exogenic chimeric proteins combining endogenous protein domains with domains from Arabidodsis thaliana and Mus musculus. The interaction between the proteins is regulated by phosphorylation: the protein of one branch will interact with the protein of the next branch it represses only when it is phosphorylated. The resulting toggle network is sensitive, can respond quickly to the input signals, and shows long-term bistability across cell divisions. It was shown to control outputs such as fluorescence or abrogation of cell division. Furthermore, the authors developed a computational algorithm that identified 109,401 potential endogenous bistable pathways. 186 of these pathways were tested and five exhibited bistability unknown before. The protein-based toggle-switch could be used as a sensitive environmental sensor or a micro-electronic device for delivery into the gut where it could detect medical conditions such as internal bleeding.

Light-controllable RNA-protein devices for translational regulation of synthetic mRNAs in mammalian cells (Nakanishi et al. 2021) the authors created a system for optogenetic control of mRNA translation of modified RNAs, in which modified nucleosides were incorporated. The authors developed two light-responsive activation systems based on the caliciviral VPg-based translational activator (CaVT). One system is a split CaVT, in which the RNA binding domain and the translational activation domain can only interact if a ligand is uncaged by light. The second system uses a destabilizing domain-fused CaVT that is rapidly degraded if the ligand is absent and is stabilized if the ligand is uncaged by light. The split CaVT has the advantage to be more robust towards shorter exposure times with the uncaged ligands, while the destabilizing domain-fused CaVT is capable not only of activation but also of repression. The controllable expression of modified RNAs will be useful for gene therapy as these RNAs show enhanced translation efficiency, decreased immunogenicity and cytotoxicity and no risk of insertional mutagenesis.

Wearable materials with embedded synthetic biology sensors for biomolecule detection (Nguyen et al. 2021) the authors embedded cell-free, freeze-dried genetically engineered polynucleotide circuits in materials such as textiles or silicones creating wearable sensors for the detection of small molecules, nucleic acids or toxins. They first established a sensor with the lacZ-operon as an output that was used to detect different inputs: (1) detection of anhydrotetracycline via a transcription factor-regulated circuit, (2) detection of Ebola virus RNA via a toehold switch and (3) detection of the small molecule theophylline via a riboswitch. Sensor circuits were also embedded in fibre optic textiles to allow the detection of viral (HIV) or bacterial (Borellia burgdorferi) RNA via a fluorescent or luminescent output. A sensor based on a programmable CRISPR system was used for direct nucleic acid detection. In this sensor, a Cas12a protein-gRNA combination detects a target dsDNA and subsequently cleaves a quenched ssDNA fluorophore probe resulting in a fluorescent output. The CRISPR-system was successfully applied to the detection of Staphylococcus aureus resistance markers. The authors also used their system to construct a sensor integrated into a face mask for the detection of SARS-CoV-2 RNA that was as sensitive as the standard laboratory RT-PCR assay.

Chen et al. (2020) explored the possibility of designing logic gates using de novo designed protein heterodimers. In order to build such logic gates, they first defined specific properties of protein building blocks: 1) high number of orthogonal pairs to prevent gate complexity limitations, 2) blocks to be modular and similar in structure to simplify gate construction, 3) single blocks to be able to bind to different partners with different and tuneable affinities, 4) the interaction of the blocks to be cooperative to ensure that gate activation is not sensitive to stoichiometric imbalances in the inputs. To meet these properties in their design, Chen et al. used existing datasets of orthogonal designed heterodimers (DHD) which all share the same four-helix-structure. The monomers were fused with flexible linkers in a way that the interaction domains (surfaces) are buried within the fusion domains and would require free energy to get exposed. This type of construction ensures that the logic gate is only activated by the sum of the binding energy of cooperating DHD pairs. Purified circuit elements were used to demonstrate the functionality of the DHD logic gates under variable experimental conditions in a cell-free system and in yeast cells, using gene expression control of a fluorescent reporter as a read out. Additionally, multi input gates based on DHD have been tested as well, revealing the system’s scalability. As a proof of concept, Chen et al. aimed to repress the immune checkpoint gene for the T cell immunoglobulin and mucin domain-containing protein 3 (TIM3) in native T-cells. Therefore, an “OR” DHD logic gate was constructed, in which one monomer of the DHD is linked to the transcription activator like effector (TALE) DNA-binding domain and the other to the Krüppel-associated box (KRAB) repressor domain. If the monomers are brought into proximity to each other, the expression of TIM3 could be repressed. Thus, a protein logic gate was designed that, due to its flexibility, composition, and scalability, could enable post-translational control over a variety of biological functions. - to the original literature

Frei et al. (2020) generated miRNA-based incoherent feedforward (iFF) circuits to rescue the expression level of genes of interest despite changes in endogenous cellular conditions. Thereby, this work contributes to the development of robust-by-design mammalian synthetic circuits. To generate the necessary data for designing the iFF circuits, Frei et al. started by investigating the burden imposed by transiently expressed synthetic circuits on mammalian cells. It was shown that exogenous gene expression in HEK293T and H1299 cells had a negative impact on host cell protein expression. Additionally, the authors demonstrated that miRNA mediated downregulation can help mitigating the metabolic burden of a genetic circuit. Based on these data, a model for synthetic circuits was established, in which expression of the gene of interest is uncoupled from endogenous environmental changes. The model was able to adapt to existing circuit topologies such as the open-loop and iFF loop to include pools of shared and limited resources. As a proof of concept, an iFF circuit with a miRNA mediated downregulation using miRNA target sites complementary to an endogenous or synthetic miRNA was tested in silico and in vitro. As expected, the iFF circuit was less sensitive to changes in available resources. - to the original literature

Krawczyk et al. (2020) have linked electrical stimulation of human cells to either transgene expression or secretion of protein therapeutics from intracellular vesicles. The electronic information is converted into protein production and release via depolarization of the cell membrane. The authors introduced L-type voltage-gated calcium channels into the membrane of mammalian cells. The channels open when the membrane is depolarized by an electric pulse, enabling a calcium influx which finally leads to the induction of target genes by the nuclear factor of activated T cells (NFAT). As a proof of concept, human β cells that secrete insulin upon calcium influx were made electrosensitive by expressing the ion channel. Such cells were placed into a bioelectronic implant that can be stimulated wirelessly. When implanted into type 1 diabetic mice, the electrosensitive cells were able, upon electrical stimulation, to secrete insulin and decrease blood glucose levels to normal. - to the original literature

Lajoie et al. (2020) designed colocalization-dependent protein switches (Co-LOCKR) that perform AND, OR, and NOT Boolean logic operations to enable a specific cell targeting. For example, the system can be used to recognize a specific antigen combination on tumour cells to induce a directed anti-tumour action. Based on the original design of the LOCKR-Switch (Langan et al. 2019), where a protein cage is kept inactive by a latch protein until a key protein binds and enables the interaction with the effector protein, a new LOCKR-version with shorter helices, an improved hydrophobic packing and an additional hydrogen bond network was generated. To target cells with a combination of specific surface antigens, targeting domains were added that recruit the Co-LOCKR cage and key proteins to a cell expressing specific target antigens. The authors showed that the cage and key could be recruited to Her2- and EGFR-expressing cells by the targeting domains in an AND logic. When adding a second key, OR logic was introduced. A NOT logic was established by adding a decoy protein acting as a molecular sponge for the key and fused to a targeting domain against a surface marker to be avoided. In a proof of principle experiment the authors generated chimeric antigen receptor (CAR)-T-cells targeting an exogenously expressed protein on cells. The CAR-T-cells were specifically directed towards tumour cells expressing Her2 and EGFR when adding the Co-LOCKR, the key, and the targeting domain. In experiments, in which CAR-T cells were co-cultured with AND/OR or AND/NOT Co-LOCKRs and tumour cells displaying the designated antigen, the Co-LOCKR carried out the expected logic, resulting in T-cell proliferation at a specific antigen combination. Co-LOCKR was able to identify a specific tumour cell line in a co-culture of different tumour cells based on the antigen combination on its cell surface. - to the original literature

Pinto et al. (2020) characterized a library of 34 split inteins (internal proteins). Inteins are auto-catalytic protein segments excising themselves from a larger precursor peptide whose flanking residues are then ligated through the formation of a new peptide bond, a process known as protein splicing. The intein itself is excised. Split inteins are inteins joining two protein halves that are expressed from different genes. The authors used a split mCherry protein to evaluate the cis-splicing (both protein halves are on the same plasmid) and trans-splicing (protein halves and split inteins are on different plasmids) capacity of split inteins under similar conditions. The authors found 15 mutually orthogonal split intein pairs capable of cis- and trans-splicing of which 10 could be simultaneously used in a cell-free assay. The split inteins were then used for logic circuit operations, two-input/two-output as well as three-input/three-output logic circuits, where they were coupled to split transcriptional regulators. In another application, split inteins were used to assemble large repetitive proteins from multiple peptides and may thus be applied in biomaterial manufacturing. - to the original literature

Wiechert et al. (2020) built and tested synthetic promoters for precise inducible expression systems based on the bacterial mechanism of xenogenic silencing and counter-silencing. Xenogenic silencers are nucleoid-associated bacterial proteins that preferentially binding to horizontally acquired AT-rich DNA. The authors used the xenogenic silencer CspS, a prophage-encoded protein from Corynebacterium glutamicum. CspS acts as a silencer when it binds to a target promoter, oligomerizes and forms a nucleoprotein complex inhibiting transcription. To regulate this inhibition a synthetic counter-silencer was included in the promoter sequences, the operator site of the transcription factor GntR. When GntR binds to the synthetic counter-silencer promoter it interferes with the nucleoprotein complex allowing transcription to start. 44 synthetic counter-silencer promoters were identified. In addition, the system was used to build a toggle switch in C. glutamicum switching between cell growth and L-valine production.- to the original literature

Williams et al. (2020) have worked on combinatorial antigen pattern recognition on tumour cells to enable more specific cancer-cell targeting. The authors used multiple synthetic Notch (synNotch) receptors to flexibly link a range of receptors and outputs into circuits. In a three-input-AND gate, the first synNotch receptor recognises antigen 1, becomes activated and induces the expression of a second synNotch receptor. If this synNotch receptor is activated by the recognizing antigen 2, a chimeric antigen receptor (CAR) is expressed that binds the third antigen and leads to target cell killing. As an alternative to the three-input-AND-gate, the authors also tested a three-input AND-NOT gate, in which a synNotch leads to activation of a CAR and a third synNotch receptor, if activated, induces the expression of a proapoptotic protein. In this circuit, the presence of only the first two antigens leads to the killing of a tumour cell, whereas the presence of antigen 3 will destroy the T cell itself. - to the original literature

Aoki et al. (2019) - A closed biological control circuit was integrated into E. coli cells for the first time, based on the engineering principle of integral regulators. For this, Bacillus subtilis-σ- and anti-σ-factors (SigW and RsiW) were used. SigW activates the expression of araC and the gene of interest (GOI), which should be held stable by the control circuit. The anti-σ-factor RsiW binds and inactivates SigW, so that less araC and GOI are expressed. Because RsiW is simultaneously under the control of an araC dependent promoter, a lower araC concentration also leads to less RsiW. From this, a control circuit resulted that also withstood disturbances such as temperature variations in experiments. As proof of principle the authors used the methionine-synthetase gene as the GOI, producing an enzyme that is essential in methionine-free media for the growth of the cells. The control circuit led to a stable growth rate of cells. - to the original literature

Chung et al. (2019) developed a genetic circuit that detects a protein (ErbB) specifically overexpressed on the surface of tumor cells and thereupon induces apoptosis in those cells. The system consists of two proteins: a protease, which is fused to a phosphotyrosine-binding domain (PBD), and an intracellular membrane associated carrier-protein with flexible cargo. The protease only binds constitutively phosphorylated ErbB, which is found mainly in cancer cells, and thus gets in close proximity to the cell membrane. The protease is then able to cleave off the cargo and e. g. induce apoptosis. As a proof of concept the signaling circuit has been introduced into healthy hepatocytes and pancreatic cancer cells with an adeno-associated viral vector, resulting in a cancer cell-restricted apoptosis. - to the original literature

Huang et al. (2019) - Oncolytic viruses are increasingly be used in the therapy of solid tumours. In order to achieve a higher tumour specificity and an improved effect Huang et al. created programmable oncolytic adenoviruses. These contain a circuit, comprising a tumour-specific promoter, a tumour- specific microRNA and a microRNA expressed mainly in healthy cells. If the tumour-specific promoter is activated, the tumour-specific microRNA increases and the microRNA specific to healthy cells is little present, the adenovirus expresses the factor E1 and an immune effector. An improved effect and increased tissue specificity for hepatocellular carcinoma was thereby achieved. - to the original literature

Langan et al. (2019), Ng et al. (2019) - To implement a synthetic feedback control for endogenous signal paths and synthetic genetic circuits, in the first step a de novo protein system (LOCKR switch) was developed, which can undergo conformational changes (Langan et al.). The system comprises a bundle of six helices, of which five form a protein cage and the sixth helix represents a protein trap with integrated peptide sequences to bind the target molecule. In the initial conformation the sixth helix is bound to the cage and the trap is inaccessible for the target molecule. The addition of a key that is homologous competitive to the cage is associated with a change in the conformation of the helices, whereby the binding site of the target molecule is exposed such that the protein trap becomes functional. Building upon this system, Ng et al. implemented a synthetic feedback control for endogenous signal paths and synthetic genetic circuits. In the two-part work, a degronLOCKR switch (Langan et al.) was used first in order to modulate the native MAPK signal cascade in yeast such that an increased or diminished output of the signal cascade can be induced. The second part of the work concerns the use of the degronSwitch in synthetic genetic circuits. For this, a simple hormonally influenced, synthetic transcriptional cascade in yeast was tested with and without feedback-control, comprised of two inducible degronSwitch-merged transcription factors (GEM, Z3PM9). In a direct comparison with the circuit without feedback control, the circuit with feedback control was more adaptable to outside influences (hormone concentration changes) and thus represents the basis for implementing complex synthetic functions in cells. - to the original literature: Langan et al. (2019), Ng et al. (2019).

Liao et al. (2019) - Complex circuits are often lost in cells through evolutionary processes, because they are associated with a loss of fitness. In order to avoid this loss, the authors use a co-culture with three sequentially employed E. coli strains that respectively include a toxin-antitoxin system (TA-system). In addition to the respective individual TA system, such a system also codes for an antitoxin for one of the other strains. It is thereby assured that bacteria without the suitable antitoxin in the co-culture are rapidly killed and replaced by those with the antitoxin. The newly added bacterial strain with the antitoxin is thus resistant compared to the previously grown bacteria and can now dominate the culture. In a rotation principle, the second strain could then be replaced by a third and this, in turn, through the first, before mutations lead to undesirable effects. Through this principle the second circuit, which is contained in all three strains on the same plasmid as the TA system, remains stably maintained in the culture. As proof-of-principle, a circuit for population-dependent lysis was used that, at a sufficiently high concentration of a quorum sensing molecule, triggers the lysing of cells. In this circuit high selection pressure normally prevails, so that the cells adapt after two days and lysis no longer takes place. The rotation principle in the TA system allowed the formation of mutants to be prevented and the longevity of the circuit to be clearly extended. - to the original literature

Pandi et al. (2019). In order to generate advanced synthetic biological circuits, Pandi et al. first developed combined transducer-actuator-models in silico and tested these in Escherichia coli. These models consist of a transducer layer containing the gene for an enzyme converting an input molecule such as hippurate or cocaine into the output molecule benzoate and an actuator layer in which benzoate directs the expression of an output signal such as GFP. Two or more transducers can be combined into an analogous, metabolic concentration adder when the genes for two or more converting enzymes, e.g. hippurate- and benzaldehyde-converting enzymes, are placed in the same operon. The fluorescent output then will increase if either metabolite’s concentration is increased. The transducer-actuator system was also tested in a cell-free system, in which the transducers can be weighted by adding different concentrations of DNA. The more enzyme coding DNA was added, the more GFP-fluorescence was observed, demonstrating a direct correlation. This weighted transducer-actuator system was then defined as a single layer metabolic perceptron. The perceptron is a biological computation algorithm that mimics the neuron’s ability to process information, learn, and make decisions. Like a neuron, a metabolic perceptron should be able to receive multiple input signals in the form of different substrates and, as an output, trigger the specific transcription of a target gene, depending on the weighted sum of the inputs. The perceptron per se is a binary classifier, meaning: if the calculated weighted sum of the inputs reaches a predefined threshold, the decision ON is reached and fluorescence is established, otherwise the systems decision is OFF. The authors created a metabolic 4-input binary logic gate perceptron that can classify different amounts of input substrates by applying different weights to each transducer. In vitro experiments confirmed that the in silico model accurately predicted weights to obtain the full OR logic gate behavior with short execution times. This work combines analog information processing with a digital output, thus laying the groundwork for more advanced metabolic circuits for rapid and scalable multiplex sensing. - to the original literature

Saltepe et al. (2019) - Nanoparticles (NP) are playing an increasingly greater role in clinical applications and can, among others, be used to transfer chemotherapeutic agents into tumour tissue. In the development of new NP, above all the cytotoxicity and the biocompatibility are of fundamental significance. To generate a reliable and rapid toxicity sensor for a first test for NP compound screening, Saltepe et al. established a genetic circuit in E. coli, based on the heat shock response mechanism (HSR) of Mycobacterium tuberculosis. This mechanism reacts to the cellular stress induced by toxic NP with the activation of the HSR promoter and the expression of a downstream fluorescent reporter gene. In the result, a test was created that makes rapid, simple optical selection possible. - to the original literature

Purcell et al. (2018) present two approaches to encrypt synthetic gene circuits in order to protect the user’s intellectual property. The major challenge is to prevent the decryption of the organization of the synthetic gene circuit through full genome sequencing. The first strategy is based on site-specific unidirectional recombinases and their recognition sites to scramble circuit topology, while the circuit is dormant. A second approach follows the incorporation of additional genes, which are irrelevant for the gene circuit but serve as a camouflage. Furthermore, Purcell et al. present a “molecular key” to decrypt the data by removing or repressing the effect of the proteins expressed by the camouflage genes, leading to correctly reassembled data. As an example, the key could be a plasmid, containing components for a CRISPR-interference, enabling the repression of camouflage gene expression. - to the original literature

Segall-Shapiro et al. (2018) designed an incoherent feed-forward loop with a stabilized promoter to enable a constant gene expression decoupled from internal and external influences, such as plasmid copy number or genomic localization. The system shall be used to stabilize the expression of different genes in customized synthetic pathways. Since cells are subject to dynamic changes during growth or differentiation, stable gene expression after introduction of a new pathway can be challenging and generally needs laborious adjustments. To regulate the expression of the gene of interest (GOI), Segall-Shapiro et al. implemented a repressor that binds to the GOI´s promoter. The gene encoding this constitutively expressed repressor is cloned upstream of the GOI, ensuring that both genes are equally affected by changes in the host cell. The GOI is thus expressed independently of copy numbers. plasmid copy numbers. - to the original literature

Toda et al. (2018) present a synthetic network of cell-cell communication that creates customized complex as well as asymmetric multicellular structures, the basis of any synthetic self-organizing tissue and bio-materials. The approach follows a synthetic signaling circuit based on the synNotch juxtacrine signaling platform. Within the circuit specific cell-cell-contacts change the cadherine cell adhesion, resulting in cell differentiation and production of further cell-cell signals. - to the original literature

No publications were selected for the year 2025.

Advancing the scale of synthetic biology via cross-species transfer of cellular functions enabled by iModulon engraftment (Choe et al. 2024)

The authors analysed transcriptome databases and identified so-called iModulons. These are sets of co-expressed but potentially independently regulated genes that are responsible for a specific cellular metabolic function. Genes that have not yet been annotated are also recorded. The identification of iModulons enables the precise identification of genes that are required for the transfer of complex cellular functions between different species. iModulons can be further improved with adaptive laboratory evolution, so that the new host organism makes optimal use of the transferred metabolic pathway under selection pressure and that host factors which need to be adapted can be identified. In the present approach, three metabolic pathways and one antibiotic resistance trait were transferred from Pseudomonas to E. coli. This cross-species transfer of metabolic pathways is fast and precise and could be a valuable tool to generate specific production strains.

De novo design of proteins housing excitonically coupled chlorophyll special pairs (Ennist et al. 2024)

The authors have functionally incorporated the central reaction centre of the native photosynthetic LH1−RC complex from purple nonsulfur bacteria (Blastochloris viridis), consisting of a synthetic ‘special pair’ of chlorophyll, into designer proteins. The special pair is excited by light energy, and the resulting charge separation initiates the electron transfer cascade that leads to the production of energy-rich compounds in the cell. The authors synthesised a c2-symmetrical protein that binds two chlorophyll molecules close together and in the same orientation as in the native special pair. Spectroscopy showed that the chlorophylls are excitonically coupled, and fluorescence lifetime imaging demonstrated energy transfer. Due to its symmetry, the synthetic special pair could also be successfully incorporated into larger subcomplexes. An octahedral nanocage generated with the synthetic protein, 24 synthetic chlorophyll molecules, and a special pair at each end is a first step towards a de novo design of photosynthetic compartments corresponding to thylakoids or chromatophore vesicles. The study lays the foundation for investigating the photophysics of special pairs, regardless of the complexity of native photosynthetic proteins, and represents the first step towards the development of synthetic photosystems for new energy conversion technologies.

New-to-nature CO₂-dependent acetyl-CoA assimilation enabled by an engineered B12-dependent acyl-CoA mutase (Schulz-Mirbach et al. 2024)

Acetyl-CoA is a key intermediate in C1 carbon assimilation, in which C1 is incorporated into higher compounds and biomass after its formation. In aerobic metabolic pathways, a part of the previously fixed CO₂ is released again during further processing. In the present study, the authors have developed the synthetic lactyl-CoA mutase (Lcm) module to prevent this loss. The Lcm module establishes a direct connection between acetyl-CoA and pyruvate without releasing CO₂. To achieve this, a new synthetic coenzyme B12-dependent mutase was used, which converts 3-hydroxypropionyl-CoA into lactyl-CoA. The 2-hydroxyisobutyryl-CoA mutase from Bacillus massiliosenegalensis was used as a basis. The catalytic efficiency of Lcm in E. coli was improved via adaptive evolution and hypermutation with the eMutaT7 system. The Lcm module still needs improvement of its efficiency, stability, and oxygen tolerance, but is thought to be used as an alternative route for various microbial production applications in the future.

ATP production from electricity with a new-to-nature electrobiological module (Luo et al. 2023)
The authors constructed a new-to-nature electrobiological module, the acid/aldehyde ATP cycle (AAA cycle) that enables a cell to directly convert electrical energy into ATP. The AAA cycle consists of four enzymes and does not require a membrane-based charge separation. It can be coupled to other in vitro processes such as transcription/translation systems. The conduction of electrons into the module is accomplished by hexamethyl viologen, enabling the direct storage of electricity as energy in biological systems.

Complete integration of carbene-transfer chemistry into biosynthesis (Huang et al. 2023)
The authors integrated the unnatural carbene transfer reaction for the first time into a biosynthetic pathway of a bacterium. The carbene transfer reaction is a prime example of a reaction that is available to synthetic chemists but has been lacking in biology, resulting in a narrower range of accessible products in biosynthesis than in synthetic chemistry. The biosynthetic gene cluster for the natural α-diazoester azaserine from Streptomyces fragilis is introduced into Streptomyces albus. The authors show that azaserine serves as a carbene donor to cyclopropanate, an intracellularly produced styrene, to generate unnatural amino acids with a cyclopropyl group. The reaction is catalysed in S. albus by an evolved, mutant cytochrome P450 enzyme with its native cofactor. The study creates a scalable microbial platform for performing intracellular abiological carbene transfer reactions to functionalise a range of natural and new-to-nature products and expands the range of organic products that can be produced by cellular metabolism.

Manipulation of sterol homeostasis for the production of 24-epi-ergosterol in industrial yeast (Jiang et al. 2023)
The authors constructed a yeast cell factory for the scalable production of 24-epi-ergosterol. 24-epi-ergosterol is an unnatural sterol that can be used as a precursor for the semi-synthesis of brassinolide, a plant hormone that has the potential for a wide range of agricultural applications limited by its extremely low natural occurrence and the lack of synthetic precursors. The authors constructed the artificial metabolic pathway by first introducing a Δ²⁴⁽²⁸⁾-sterol reductase (DWF1) from plants in Saccharomyces cerevisiae, followed by enzyme-directed evolution to enhance the catalytic activity of DWF1 and enable de novo biosynthesis of 24-epi-ergosterol. The sterol fluxes towards 24-epi-ergosterol were further strengthened by the engineering of sterol homeostasis, via overexpression of three endogenous genes (YEH1, YEH2, and ARE2). The sterol homeostasis engineering strategy can be used for mass production of other economically important phytosterols.

Construction and modular implementation of the THETA cycle for synthetic CO₂ fixation (Luo et al. 2023)
The authors have constructed a new-to-nature CO₂-fixation pathway, the THETA (tricarboxylic acid branch/4-hydroxybutyryl-CoA/ethylmalonyl-CoA/acetyl-CoA) cycle. The THETA cycle comprises 17 enzymes from 9 organisms and is centred around two of the most efficient CO₂-fixing enzymes described in nature, crotonyl-CoA carboxylase/reductase and phosphoenolpyruvate carboxylase. Rational and machine learning-based optimisation approaches have improved the yield of the cycle by two orders of magnitude and demonstrated the formation of various biochemical building blocks directly from CO₂. The THETA cycle was divided into three modules that were implemented in vivo in Escherichia coli. This represents the first step towards the realisation of highly orthogonal and complex CO₂-fixation pathways in the background of living cells.

Construction of an artificial phosphoketolase pathway that efficiently catabolizes multiple carbon sources to acetyl-CoA (Yang et al. 2023)
The authors have designed and constructed an artificial phosphoketolase (APK) pathway that consists of only 3 types of biochemical reactions and has the potential to achieve 100% carbon yield to acetyl-CoA from any monosaccharide. The central enzyme in this pathway is phosphoketolase, while phosphatase and isomerase act as auxiliary enzymes. For the conversion of fructose-6-phosphate the APK pathway included phosphoketolase (PK) from Actinobacteria bifidobacterium (BbPK), haloacid dehalogenase (HAD)-like hydrolase (EcHAD) from E. coli, and L-rhamnose isomerase (Ps-LRhI) from Pseudomonas stutzeri, while for the conversion of xylulose-5-phosphate BbPK, triose phosphate isomerase (EcTIM) from E. coli, sugar phosphatase from Candida parapsilosis (CpHAD), and formolase (FLS), a synthetic enzyme, were involved. The APK pathway was tested in vitro and it was shown that typical C1-C6 carbohydrates were efficiently metabolised to acetyl-CoA with a yield of 83 % to 95 %. In addition, the authors have developed E. coli strains that are able to grow via the APK pathway when glycerol is used as a carbon source. The novel APK pathway has the potential for using multiple carbon sources with higher efficiency in biomanufacturing in the future.

Engineering α-carboxysomes into plant chloroplasts to support autotrophic photosynthesis (Chen et al. 2023)
The authors generated morphologically correct carboxysomes in tobacco chloroplasts by transferring to them the genetic components from the well-studied α-carboxysome of proteobacterium Halothiobacillus neapolitanus. The fully functional α-carboxysome consists of components encoded by nine carboxysome genes and replacing the endogenous tobacco rbcl gene that encodes the ribulose-1,5-biphosphate carboxylase/oxygenase (Rubisco), an important enzyme involved in carbon fixation. α-carboxysome supports tobacco plant growth and autotrophic photosynthesis at elevated CO₂ levels. This work paves the way to improve crop plant photosynthesis and productivity.

Engineering a new-to-nature cascade for phosphate-dependent formate to formaldehyde conversion in vitro and in vivo (Nattermann et al. 2023)
The authors developed a new pathway to reduce formate to formaldehyde using two enzymes and inorganic phosphate (Pi). First, formate is phosphorylated and activated to formyl phosphate with acetate kinase from Escherichia coli followed by reduction to formaldehyde with a variant of N-acetyl-γ-glutamyl phosphate (NAGP) reductase (ArgC) from the bacterium Denitrovibrio acetiphilus. Using a semirational enzyme engineering approach, the authors have developed a novel formyl phosphate reductase variant from ArgC. The enzyme is a triple substitution variant of the enzyme that exhibits a complete loss of native enzyme activity (i. e. NAGP reduction), a 300-fold shift in specificity towards formyl phosphate and reduced side reactivity with acetyl phosphate. The functionality of the pathway was demonstrated both in vitro and in vivo in E. coli. In addition, the Pi-based route was linked to a recently developed formaldehyde assimilation pathway, the FORCE pathway (Chou et al. 2021). This ultimately enables the production of C2 compounds in vitro and in vivo from formate as the sole carbon source.

A biobricks metabolic engineering platform for the biosynthesis of anthracyclinones in Streptomyces coelicolor (Wang et al. 2022) the authors have found a way to synthesize clinically relevant anthracyclines from the actinomycete Streptomyces coelicolor. They created a genetic toolbox called BIOPOLYMER (BIOBricks POLYketide Metabolic EngineeRing) consisting of engineered strains, vectors, promoters, and biosynthetic genes for the synthesis of anthracyclinones. They engineered four different anthracyclinone pathways and explored their effectivity. BIOPOLYMER is thought to serve as a platform for the synthesis of designer anthracycline analogues.

Conversion of CO₂ into organic acids by engineered autotrophic yeast (Baumschabl et al. 2022) the authors used the synthetic autotrophic yeast strain Komagataella phaffii (Pichia pastoris) described by Gassler et al. as a host to engineer organic acid producing strains. In a previous study the K. phaffii strain was provided with the Calvin-Benson-Bessham cycle, enabling it to grow continuously with CO₂ as a sole carbon source. In the present work two genes from Aspergillus terreus, a cis-aconitate decarboxylase (cadA) and a mitochondrial tricarboxylic acid transporter (mttA), were introduced into the autotrophic K. phaffii strain for the production of itaconic acid. For lactic acid synthesis ldhL from Lactobacillus plantarum was implemented into the K. phaffii genome and the CYB2 gene was deleted by using CRISPR-Cas9. The resulting strains are able to produce organic acids solely from CO₂ as carbon source. They could be used for the production of value-added chemicals and act as CO₂-neutral or CO₂-negative chassis organisms.

Microbial synthesis of vanillin from waste poly(ethylene terephthalate) (Sadler & Wallace et al. 2021)theauthors designed an engineered Escherichia coli that converts post-consumer plastic waste into vanillin. The polyethylenglycol (PET)-derived monomer terephthalate (TA) was shown to be converted into vanillin in a 5-step enzymatic pathway, consisting of enzymes from bacteria (Comamonas sp., Nocardia iowensis and Bacillus subtilis) and a mammal (Rattus norvegicus, rat).

A new-to-nature carboxylation module to improve natural and synthetic CO₂ fixation (Scheffen et al. 2021) theauthors engineered a new-to-nature reaction for the CO₂-dependent assimilation of the C₂ compound glycolate into the central carbon metabolism without carbon loss. The few natural pathways converting C₂ into C₃ metabolites, e. g. photorespiration, result in carbon loss. The authors engineered the so far hypothetical tartronyl-CoA (TaCo) pathway, a synthetic photorespiratory bypass that represents an alternative to natural photorespiration. In this pathway, glycolate is metabolized to glycerate by a three-enzyme-reaction transforming glycolate into glycolyl-CoA, then into tartronyl-CoA, and finally into glycerate. An engineered glycolyl-CoA synthetase from Erythrobacter sp., NAP1, was found to catalyse the first reaction, while a malonyl-CoA reductase from Chloroflexus aurantiacus catalyses the third step and converts tartronyl-CoA into glycerate. The key enzyme of the pathway, a glycolyl-CoA carboxylase (GCC), had yet to be identified. The authors screened propionyl-CoA carboxylases for their potential as a GCC, because they exhibit a similar structure. They identified the propionyl-CoA carboxylase of Methylorubrum extorquens as a candidate, mutated the gene and applied a directed evolution approach to obtain a protein with similar efficiency to naturally occurring biotin-dependent acyl-CoA carboxylases.
The synthetic GCC enzyme proved to be functional in in vitro experiments in combination with the two other enzymes of the TaCo pathway and could be combined with other synthetic CO₂ fixation cycles such as the CETCH (crotonyl-coenzyme A (CoA)/ethylmalonyl-CoA/hydroxybutyryl-CoA) cycle. 

Optogenetic control of plant growth by a microbial rhodopsin (Zhou et al. 2021)the authors established an optogenetic control system in plants which is able to control their growth. Microbial rhodopsins are proteins often expressed as optogenetic tools, but are difficult to express in plants due to a lack of the essential cofactor retinal. The authors used the green light-gated anion channel-rhodopsin ACR1 from the unicellular algae Guillardia theta in tobacco plants. To provide the necessary retinal, the bacterial ß-carotene 15,15`-dioxygenase MbDio as a fusion protein with the chloroplast transit peptide RC2, that allows retinal location to the chloroplast, was co-expressed. The expression of these proteins allowed for a green light-induced anion efflux through the plants´ cell membrane and membrane potential depolarization.
In a proof-of-principle experiment, the authors used the MbDio-RC2-ACR1-construct to test the hypothesis that pollen tube growth is controlled by a local anion channel activation and the resulting voltage gradient. While global illumination led to a massive anion efflux which either did not affect or stopped the tube growth, local illumination changed the tube´s growth direction away from the side in which ACR1 was activated, thereby confirming the hypothesis. This work provides the basis for a powerful tool in studying signalling in plants and could also enable the testing of chemical-electrical signalling.

Miller et al. (2020) developed a semi-synthetic chloroplast by encapsulating and operating photosynthetic membranes in cell-sized droplets. In order to establish a module for the light-driven regeneration of adenosine triphosphate (ATP) and nicotinamide adenine dinucleotide phosphate (NADP+), thylakoid membranes from the chloroplasts of Spinacia oleracea were isolated and initially tested for their ability to perform a light-dependent reduction of NADP+ to NADPH and regeneration of ATP from adenosine diphosphate (ADP). Since the thylakoid membranes were able to produce NADPH and ATP as predicted, Miller et al. used these membranes not only to provide energy for individual enzymatic reactions for CO₂-fixation but also for a complete synthetic metabolic cycle for the continuous fixation of CO₂. This synthetic metabolic cycle consists of three different parts: 1) the artificial 16 enzyme crotonyl-CoA/ethylmalonyl-CoA/hydroxybutyryl-CoA (CETCH) pathway for in vitro fixation of CO₂, 2) the glyoxylate/hydroxypyruvate reductase from Escherichia coli and 3) a crotonase from Pseudomonas aeruginosa. In order to create a semi synthetic chloroplast, a microfluidic platform was developed, in which the metabolically active microcompartments can be automatically assembled into water-in-oil micro droplets. The optimisation of the droplet resulted in a complex system that not only possesses essential characteristics of photosynthesis but also outperforms systems using only single enzymes in energy production. This work provides the basis for the development of a completely self-sustaining synthetic organelle. - to the original literature

Reifenrath et al. (2020) use endoplasmic reticulum-derived vesicles to compartmentalise the enzymes of a metabolic pathway into membrane surrounded organelles. That way, potential toxic side effects of unwanted reactions caused by a heterologously expressed protein can be avoided. The three enzymes of the cis,cis-muconic acid (CCM) pathway (toxic for yeast) were fused to a synthetic peptide containing the self-assembling region of the maize storage protein gamma-Zein (Zera). Zera can induce so-called protein bodies derived from the endoplasmic reticulum or the vacuole in plants as well as in heterologous systems. The Zera-induced vesicles with heterologous CCM-enzymes were generated in the yeast Saccharomyces cerevisiae and the functionality of the metabolic pathway in the vesicles was shown. - to the original literature

Gleizer et al. (2019) - The authors have produced an E. coli strain that, for the first time, uses CO₂ as the sole source to produce biomass. The conversion of atmospheric CO₂ into foods, fuels and biochemicals should thereby be possible. For carbon fixation the bacteria use the Calvin cycle, whereby the electrochemically produced molecule formate (HCOO-) serves to gain energy. The Calvin cycle drives the conversion of formate into ATP, converting CO₂ into sugar and other organic molecules. The autotrophy was achieved through evolution under laboratory conditions: the xylose concentration was reduced in a chemostat, while adequate formate and CO₂ were available. The bacteria became autotrophs through various mutations, of which many were in genes with a metabolic connection to the Calvin cycle. - to the original literature
 

Luo et al. (2019) established a complete biosynthetic pathway for the production of several complex cannabinoids in the yeast Saccharomyces cerevisiae. To achieve this the mevalonate-pathway of yeast was modified by introducing genes from different organisms and Cannabis sativa necessary for the hexanoyl-CoA pathway into the yeast genome. The genetically modified yeast enables a controlled industrial production of cannabinoids. - to the original literature

South et al. (2019) increased growth and productivity of the C3 plant tobacco by introducing a synthetic, more effective photorespiration pathway. When facing high temperatures and dry periods, C3 plants like wheat, rice and soybean increasingly produce glycolate, which is toxic for the plant and has to be converted to non-toxic molecules by photorespiration. This process can reduce the harvested biomass by up to 50 %. By introducing a synthetic glycolate pathway in the chloroplast and simultaneously inhibiting the native photorespiration pathway, biomass production in homozygous transgenic lines was increased by up to 40 %. - to the original literature

Engineering Artificial Mitochondria with Self-Amplifying Proton Generation for Autonomous Energy Supply and Metabolic Coupling in Artificial Cells (Fu et al. 2025)

The authors have created a nanofactory for artificial cells that, like a real mitochondrion, ensures an autonomous energy supply. To achieve this, the enzymes glucose oxidase (GOx) and catalase (CAT) were co-encapsulated in silica nanocapsules (GC@SiO₂), whose membrane allows exchange with small molecules via pores. GOx converts glucose and oxygen (O₂) into gluconic acid, releasing protons (H⁺) in the process. The protons lead to a strong proton gradient across the membrane. The H₂O₂ produced as a toxic by-product is converted by CAT into O_₂ and H₂O. GC@SiO₂ was then embedded in liposomes, into whose lipid bilayer the enzyme ATP synthase (ATPase) was inserted (GC@SiO₂@Lips-ATPase). To enable ATP-dependent synthesis of NADH in a subsequent step, the ATP nanofactory GC@SiO₂@Lips-ATPase, along with the building blocks for NADH production - the enzymes hexokinase and glucose-6-phosphate dehydrogenase (G6PDH), as well as NAD⁺ - were encapsulated into giant unilamellar vesicles (GUVs). Within the GUVs, glucose is converted into glucose-6-phosphate by hexokinase, consuming ATP in the process. This intermediate is further metabolised by G6PDH, converting NAD⁺ into NADH. The system thus enables continuous glucose-driven ATP production, which allows for ATP-dependent NADH production and thereby mimics the oxidative phosphorylation of mitochondria, which is the main process by which energy is generated in aerobic cells. 

Metabolite-Responsive Control of Transcription by Phase Separation-Based Synthetic Organelles (Jerez-Longres et al. 2025)

The authors have produced the first synthetic organelle that responds to a specific metabolite, in this case pyruvate, and thereby regulates DNA transcription. The artificial organelle is based on droplets (coacervates) formed by liquid-liquid phase separation. In this process, proteins in an aqueous solution coagulate to form dense liquid droplets in which specific molecules can be encapsulated. The phase separation of a protein can be triggered by the N-terminal intrinsically disordered region (IDR) of the fused in sarcoma protein. In this case, the IDR is fused with the bacterial repressor protein PdhR (pyruvate dehydrogenase complex regulator). PdhR has the property of binding to a specific DNA sequence (pdhR operator), depending on whether the metabolite pyruvate is present or not. DNA segments containing the pdhR operator, as well as the DNA sequence of the RNA aptamer SdBroccoli are inserted into the PdhR-IDR droplets. Upon contact between SdBroccoli and the ligand DFBHI a light green fluorescence signal can be measured. This results in a pyruvate-dependent reversible switch with two possible states:

Off-state without pyruvate: The artificial PdhR-IDR protein forms droplets that act as synthetic organelles. In the absence of pyruvate, a strong bond forms between the DNA and PdhR, so that the DNA is contained within the droplets. The DNA is thus blocked, preventing it from being transcribed by the enzyme T7 RNA polymerase. The reporter SdBroccoli is not transcribed and no fluorescence signal can be measured.

On-state with pyruvate: When pyruvate is added, it binds to PdhR, causing it to lose its binding affinity for the DNA. The DNA is released from the droplets and can be transcribed by T7 RNA polymerase. Transcription is measured by monitoring the production of the SdBroccoli reporter, its binding to the DFHBI, and the resulting fluorescence signal.

This work represents an important step towards a fully synthetic protocell containing various organelles that perform different functions.

A synthetic cell phage cycle (Levrier et al. 2025)

The authors present the first complete viral infection cycle (adsorption, genome entry, replication, packaging, host cell lysis) of a T7 phage recreated in vitro within artificial cells. Through the use of liposomes with specific receptors and cell-free gene expression (CFE) systems, T7 phages were able to infect the synthetic cells, replicate within the cells and release infectious progeny phages. 

To produce stable liposomes capable of supporting CFE, a shorter variant was used instead of the commonly used lipopolysaccharides (LPS). This shorter “rough” variant, known as RdLPS, has a lower molecular weight and remains stable within the membrane of the artificial cell. However, wild-type T7 phages cannot bind to the short RdLPS, which is why the T7-split-S-phage was engineered with a new method. This phage features RdLPS-specific tail fibers – a thread-like protein structure at the tail end of the phage. Lysis of the synthetic cell occurs via an osmotic shock caused by the transfer from an isoosmotic to a hypoosmotic environment. DNA replication results in one- to twofold amplification compared to 50- to 300-fold amplification in bacterial hosts. 

This work provides a defined platform for investigating virus-host interactions and phage therapies.

Monodisperse Giant Unilamellar Niosomes as Minimal Synthetic Cells (Luo et al. 2025)

Liposomes, polymersomes and fatty acid vesicles, collectively referred to as giant unilamellar vesicles (GUVs), are used in the development of synthetic cells. The authors have developed so-called giant unilamellar niosomes (GUNs) as an improvement, which consist of non-ionic surfactants. GUNs show a high membrane fluidity and a permeability for molecules up to 500 Da, meaning they do not require transport proteins to take up glucose, for example. To mimic subcellular membrane-less organelles, an internal liquid-liquid phase separation was induced. To achieve this polyallylamine hydrochloride and ATP were introduced; at a specific external pH, these form droplets (coacervates) and can take up molecules. 

By introducing glioma cell extracted mitochondria and the enzymes glucose dehydrogenase, gluconate dehydrogenase and 2-keto-3-deoxygluconate aldolase, ATP could be generated within the GUNs. The additional introduction of actin monomers and the supplementation of glucose to the external environment led to the formation of actin filaments within the GUNs. The study establishes GUNs as a new approach for the creation and study of synthetic cells.

Genetic encoding and expression of RNA origami cytoskeletons in synthetic cells (Tran et al. 2025)

Replicating gene expression (DNA to RNA to protein) poses a major challenge in synthetic systems. The authors propose a possible solution involving the shortening of the process to a single step from genotype to phenotype. Here, artificial cytoskeletons were autonomously formed from a DNA template using RNA origami, a method from nanotechnology in which a single, long strand of RNA is designed to fold itself into a complex two- or three-dimensional shape at the nanometre scale. The cytoskeleton is produced within synthetic cells, known as giant unilamellar vesicles (GUVs), through the consumption of ribonucleoside triphosphates ATP, GTP, UTP and CTP (rNTPs). The surface membrane of the GUVs contains pores or ionophores through which the rNTPs and salts enter the artificial cell from the external environment.

This results in an active, energy-dependent system that continuously self-assembles, in contrast to classical, equilibrium DNA origami structures, which remain largely static objects after their one-time self-assembly. 

As a proof-of-concept RNA origami nanotubes were produced. During transcription under isothermal conditions, RNA origami tiles fold into nanotubes measuring several micrometres in length. Mutations were also introduced in the DNA template that led to incorrect folding and the formation of rings instead of tubes. The advantages of RNA origami over DNA origami according to the authors are: RNA is genetically encodable, meaning that cells can produce their own building blocks in an active, non-equilibrium process, whereas in DNA origami, DNA is used merely as a building material and not as a template; functions such as membrane binding can be integrated; and small changes to the sequences or helices can lead to significant morphological changes. Furthermore, only one enzyme, the RNA polymerase, is required instead of the 150 genes needed to build a natural cytoskeleton. The RNA origami method developed here is a new tool that henceforth will enable the manipulation cells and investigate questions in cell biology.

Responsive DNA artificial cells for contact and behavior regulation of mammalian cells (Wang et al. 2025) 

The authors have developed artificial cells that are DNA-based and have the ability to regulate mammalian cells (stimulable artificial cells designated to regulate mammalian cells, STARMs). STARMs are produced through temperature-controlled DNA self-assembly. In this process, the STARMs spontaneously arrange themselves from DNA-encoded copolymers into a cell-like structure, which is divided into subcompartments mimicking artificial cell surfaces and the cytoplasm. The cell surface contains DNA-based sensors capable of responding to specific stimuli. Depending on the sensor they carry, STARMs can thus be activated by stimuli such as ions or light. This leads to the release of ligands capable of interacting with specific receptors on a mammalian cell. Upon contact, signaling pathways are activated in the natural target cells. A modular design of the sensors allows for the simple rewiring of different stimuli to different outputs of mammalian cell regulation. A mouse model for skeletal muscle injuries was used as an example of an application as a cell-based therapy in vivo. First, tissue in the mouse was damaged by the administration of cardiotoxin into the anterior tibialis muscle. STARMS were injected into the damaged tissue, that are activated by near-infrared (NIR) light, which penetrates deep into the tissue, and release aptamer ligands. The ligands activate the proliferation of Pax7-positive muscle stem cells via the fibroblast growth factor receptor 1. Pax7 is a transcription factor involved in myogenesis. Histological examinations revealed regeneration of the injured tissue following three cycles of injection and NIR irradiation. The artificial cells thus represent an innovative tool for the development of smart therapeutics in biomedicine. 

Integrating incompatible tandem photobiocatalysis in artificial cells enables metabolic modulation of natural cells (Wang et al. 2025)

The authors developed a new method for creating artificial cells with organelles that assist natural cells in metabolic processes as helper cells. The artificial cells were constructed as giant unilamellar vesicles (GUVs) and contain two different silica-based organelles (SiNOs) that spatially separate chemically incompatible processes. In one organelle (SiNO@PC), light-driven (photocatalytic) regeneration of the cofactor NAD⁺ takes place. In the other organelle (SiNO@ADH/ALDH), ethanol is converted into acetate by the enzymes alcohol dehydrogenase and aldehyde dehydrogenase, a process that consumes NAD+. Continuous cofactor regeneration sustains alcohol degradation, while the spatial separation protects the enzymes from reactive oxygen species generated during photocatalysis. The cells were used as helper cells for liver cells. In vitro experiments demonstrated that these artificial cells can reduce oxidative stress caused by alcohol metabolism in co-cultured mouse liver cells. This work shows that artificial cells can function as metabolic helper cells through the integration of multiple functional compartments. The proposed concept may also be applied in the future to other biocatalytic processes and to the targeted modulation of biological functions.

Genetically programmed synthetic cells for thermo-responsive protein synthesis and cargo release (Monck et al. 2024)

The paper describes the development of synthetic cells that control protein expression through a temperature-sensitive mRNA element. The authors used an RNA thermometer, consisting of a ribosome binding site (RBS) and a complementary anti-RBS sequence located in the 5´ untranslated region between promoter and gene of interest (Jia et al. 2019). The complementary sequences form a hairpin that is only opened when a certain temperature is reached. This system was used in combination with a cell-free protein expression system in giant unilamellar vesicles. As a proof-of-concept, the pore protein α-hemolysin from Staphylococcus aureus was expressed under control of the thermometer. Upon temperature elevation, the vesicles synthesized the pore protein and a small fluorescent cargo protein was released from the vesicles. This technique could be used for example in biomedicine to create customizable synthetic cells that release their cargo upon a specific temperature. 

3D DNA origami pincers that multitask on giant unilamellar vesicles (Zhan et al. 2024)

The authors developed artificial, programmable DNA origami structures, so-called DNA origami pincers (DOPs), that enable the targeted control of the morphology of giant unilamellar vesicles (GUVs). The DOPs consist of three DNA bundles, with their inter-bundle angle precisely tuneable via the incorporation of specific DNA locking strands. Depending on the programmed pinching angle, varying degrees of membrane deformation were induced in the GUVs. Furthermore, multiple DOP units were interconnected on the vesicle surface, leading to the self-assembly of higher-order cage-like DNA architectures. During this oligomerization process, transient membrane pores were formed, allowing the influx of small molecules into the vesicle interior. The DNA cages were also capable of capturing lipid fragments and subsequently detaching from the membrane. This work demonstrates how DNA nanostructures could be employed in the future to enable precise membrane remodelling, controlled molecular transport, or the construction of synthetic cell-like systems.

Artificial membraneless organelles and compartments :

Spatiotemporal regulation of cell fate in living systems using photoactivatable artificial DNA membraneless organelles (Zhang et al. 2024)

The authors developed synthetic, DNA-based coacervates that function as membraneless organelles. These structures consist of long single-stranded DNA molecules generated via rolling circle amplification (RCA), which condense into stable coacervate droplets through liquid–liquid phase separation. By incorporating photoactive palladium nanoparticles, the coacervates enabled targeted drug release, such as doxorubicin, upon near-infrared irradiation. The particles were produced in vitro and introduced into murine tumor cells (4T1 cells) via endocytosis or injected intravenously into mice. There, they allowed for light-dependent, spatiotemporal release of the drug and selectively induced programmed cell death in tumor tissue.

Dipeptide coacervates as artificial membraneless organelles for bioorthogonal catalysis (Cao et al. 2024)

The authors developed artificial coacervate droplets based on the dipeptide FF-OMe, a diphenylalanine (FF) derivative bearing a methoxy group (OMe) at its C-terminus. These droplets form via pH-dependent self-association and create a hydrophobic microenvironment that facilitates the encapsulation and catalytic activity of enzymes and hydrophobic catalysts in aqueous media. Used as microreactors, they significantly enhanced the efficiency of enzymatic and metal-catalyzed reactions in water. By combining them with polyelectrolyte-based coacervates, cell-like multicompartment systems were generated, in which the FF-OMe coacervates functioned as artificial organelles. Through chemical dimerization, the coacervates were stabilized and successfully introduced into HeLa cells, where they exhibited bio-orthogonal catalytic activity without affecting cell viability. The study demonstrates that minimalist peptide coacervates can serve as versatile, cell-compatible reaction compartments and artificial organelles in both cell-free systems and living cells.

Co-transcriptional production of programmable RNA condensates and synthetic organelles (Fabrini et al. 2024)

The authors developed a modular system for the formation of programmable RNA condensates based on synthetic RNA nanostructures. These so-called RNA nanostars are composed of a single RNA strand that folds into a four-armed, star-shaped structure during transcription. Through specific base-pairing interactions, these nanostars self-assemble into larger complexes. Both in vitro and in synthetic cells (water-in-oil droplets), they formed membraneless RNA organelles with defined size, number, and composition. Embedded RNA aptamers enabled the targeted capture of fluorophores and proteins such as EYFP and streptavidin. The introduction of specialized linker RNA altered the interactions between condensates and allowed for the controlled formation of multiphase, structured assemblies. This study demonstrates how RNA nanotechnology can be used to create cell-like compartments suitable for the selective recruitment of functional molecules.

Towards chloroplastic nanofactories: formation of proteinaceous scaffolds for metabolic engineering (Dwyer et al. 2024)

The authors established a synthetic bacterial microcompartment (BMC) within the chloroplasts of Arabidopsis thaliana. Due to its resemblance to a porous plastic ball, an icosahedral hollow structure with regular pores, it was termed a minimal wiffle ball. To construct it, two shell proteins (BMC-H and BMC-T1) from Haliangium ochraceum were genetically engineered to autonomously assemble into 40 nm icosahedral structures after import into chloroplasts. These structures resembled BMC shells and did not impair plant growth or chloroplast function. The study demonstrates that BMCs can be stably assembled as protein-based sub-compartments in plant cells and may be used to specifically modulate and optimize plant metabolic processes.

Nucleated synthetic cells with genetically driven intercompartment communication (Ioannou et al. 2024)

The authors developed an easy-to-implement method for constructing nucleated synthetic cells (nSynCells) in the form of vesicle-in-vesicle structures. The nuclear and cytoplasmic compartments could be selectively loaded with different biochemical components. Within the inner compartment, the pore-forming protein α-hemolysin was constitutively expressed using the commercial PURExpress® TX-TL system and an α-hemolysin DNA plasmid. This protein formed functional pores exclusively in the inner membrane, enabling the directed diffusion of small molecules from the outer to the inner compartment. There, the signalling molecules could react with encapsulated enzymes and trigger an enzymatic reaction. The method requires no microfluidics and is based on a straightforward emulsion-centrifugation protocol. According to the authors, this represents the first demonstration of genetically programmed communication between compartments in synthetic cells, in which communication is enabled only after in situ expression of a membrane pore protein.

Artificial endosymbiosis:

In 2024, three studies were published that describe the artificial reconstruction of endosymbiotic processes in eukaryotic cells. In these studies, both yeast and mammalian cells were made photosynthetically active by incorporating cyanobacteria or chloroplasts of algae. In Gao et al., genetically engineered cyanobacteria (Synechococcus elongatus PCC7942) capable of secreting ATP and glucose were introduced into mitochondria-deficient yeast cells, resulting in stable yeast–cyanobacteria chimeras that could grow solely with light and CO₂. Hu et al.first introduced Synechocystis PCC 6803 into murine macrophages and later into non-phagocytic mammalian cells, with human HEK293 cells proving particularly suitable for establishing long-term endosymbiotic relationships. In Aoki et al., chloroplasts from the red alga Cyanidioschyzon merolae were incorporated into hamster cells (CHO-K1) through co-cultivation, where they retained their internal thylakoid structure and photosynthetic activity for up to two days.

Artificial cell synthesis using biocatalytic polymerization-induced self-assembly (Belluati et al. 2023)
The authors have developed a more robust biocatalytic polymerization-induced self-assembly (bioPISA) for the generation of giant unilamellar vesicles (GUVs) in order to synthesize cell-like structures. Myoglobin was selected as biocatalyst because of its facilitation of atom transfer radical polymerization, peroxidase activity as well as working at a biologically relevant pH of 7.4. Depending on the monomer:initiator ratio, spherical, worm-like or vesicular morphologies could be observed, as well as their stability in relevant aqueous solutions like Luria-Bertani broth or RPMI-1640. Differentially sized molecules including enzymes could be encapsulated and functional enzyme cascades could be shown. The ability of GUVs to respond to external stimuli by altering their shape and structure was demonstrated by magnesium-triggered changes in the internal architecture and the conversion of dissolved calcium glycerophosphate to insoluble calcium phosphate by alkaline phosphatase, imitating the production of the mineral matrix of bones by osteoblasts. Encapsulated plasmids and Escherichia coli lysate mimic protein expression, shown by the production of mClover, F-actin and Calcium phosphate. The authors conclude that bioPISA will pave the way for hybrid systems combining synthetic polymers with natural molecules for applications in synthetic biology.

Cell-sized asymmetric phospholipid-amphiphilic protein vesicles with growth, fission, and molecule transportation (Suzuki and Kamiya 2023)
The authors generated cell-sized asymmetric phospholipid-amphiphilic protein vesicles composed of a lipid membrane on the outer leaflet and an oleosin membrane on the inner leaflet. Oleosins are plant proteins that stabilize the oil body. Insertion of the outer membrane protein G into the membrane led to nanopore formation and transport of ions and small molecules. Addition of lysophosphatidylcholine micelles facilitated an increase of vesicle size, eventually leading to deformation and fission. The authors conclude that this new vesicle has the potential as a novel cellular model that combines the advantages of lipid and protein leaflets, including the self-replication system, communication of living cells, and complex metabolic pathways.

Programmable fusion and differentiation of synthetic minimal cells (Gaut et al. 2022) the authors developed a combinatorial genetic circuit assembly technology, where populations of synthetic cells (liposomes) can control individual gene components and can be combined via fusion (“mating”) to construct complex biological pathways (in synthetic cells). Before, the ability to control expression of enough genes to build complex genetic pathways and the ability to mate and differentiate populations into separate lineages were yet elusive in synthetic cells. The synthetic cell mating system of Gaut et al. is based on a programmable, fully orthogonal liposome fusion via surface DNA tags. More precisely, they used liposomes with Escherichia coli cell-free protein expression systems that carry different DNA tags and contain a series of RNA polymerases and recombinases. Liposome fusion is targeted via the DNA tags that have a counterpart on another liposome population and as a result of different fusion events genetic circuits are activated. The authors demonstrated the utility of this technology to create lineages of synthetic cells, differentiating “ancestral pluripotent” populations into independent lineages upon sensing of a small molecule signal (or upon fusion with another population of synthetic cells).

A DNA origami rotary ratchet motor (Pumm et al. 2022) the authors developed a nanoscale rotary motor built from DNA origami that is driven by ratcheting where energy is needed to overcome physical obstacles. The motor has the potential to drive uphill synthesis reactions. Its mechanical capabilities approach those of biological motors such as F₁F₀-ATPases. After attachment of the motor to a microscope glass coverslip fluorescent dyes at the tips of each rotary arm enable experimental observation of the motion. To reach a biased movement a non-rotating electric alternating current (AC) field was applied. The field causes an alternating ion current flowing through the sample chamber along a fixed axis. Depending on the nature and location of energetic minima in the motor relative to the axis of the electric field, the modulation can produce a kinetic asymmetry per field cycle that causes the motor to move with a preferred rotation direction. The authors successfully encoded the motor in DNA sequences and self-assembled it in reaction mixtures. They showed a rotation with directional bias, where the direction is randomly set. The maximum velocity recorded was comparable to the power of natural molecular machines such as the ATP synthase. The direction of rotation and the effective angular speed of each motor particle could be controlled by the orientation of the AC field axis relative to the motor particles (which are fixed on the substrate).

Implanted synthetic cells trigger tissue angiogenesis through de novo production of recombinant growth factors (Chen et al. 2022) the authors generated synthetic cells (SCs, autonomous protein-manufacturing cell-like particles) that produce proangiogenic factors and are able to trigger the physiological process of neovascularization in mice. The cells could serve as a tool replacing diseased natural cells and addressing medical needs. The SCs were constructed of giant lipid vesicles and were optimized to facilitate enhanced protein production. When introduced with the appropriate genetic code, the SCs synthesized a recombinant human basic fibroblast growth factor (bFGF), reaching expression levels of up to 9 x 10⁶ protein copies per SC. The release of the protein took place via natural membrane permeability. In culture, the SCs induced endothelial cell proliferation, migration, tube formation, and angiogenesis-related intracellular signalling, confirming their proangiogenic activity. Integrating the SCs with bioengineered constructs bearing endothelial cells promoted the remodelling of mature vascular networks, supported by a collagen-IV basement membrane-like matrix. In vivo, prolonged local administration of the SCs in mice triggered the infiltration of blood vessels into implanted Matrigel plugs (mimicking the physiological cell matrix and used to study angiogenesis) without recorded systemic immunogenicity.

Building artificial plant cell wall on lipid bilayer by assembling polysaccharides and engineered proteins (Notova et al. 2022) the authors achieved the assembly of complete plant cell wall mimics by using an engineered chimeric protein designed for bridging pectin to the cellulose/hemicellulose network. Such artificial cell walls can serve as a basis for the development of (experimental) plant cell mechanical models. So far existing models lacked the pectin component. Here, the authors used the carbohydrate-binding module from Ruminococcus flavefaciens engineered for high affinity binding to polygalacturan, a main component of the pectin network, and combined it with the lectin of Ralstonia solanearum, engineered for strong binding of hemicellulose. Thereby a protein glue is created that stabilizes the interaction between hemicellulose and pectin for the assembly of an artificial cell wall. A two-dimensional assembly of an artificial plant cell wall was built first on synthetic polymer and afterwards on supported lipid bilayer.

Living material assembly of bacteriogenic protocells (Xu et al. 2022) the authors demonstrate a new approach to generate artificial cells by using components of living systems as the basis. They organized pieces of broken-down bacteria cells on a synthetic scaffold of a coacervate thereby producing artificial cells with functional and compositional complexity reminiscent of living cells. The work shows that coacervates have a huge potential as platforms to localize and integrate diverse biomaterials, including living cells, to make artificial cells. First, coacervates (membrane-free droplets made of dense liquid phase that form and separate spontaneously from aqueous solutions) to make the synthetic scaffold where generated. The coacervates were formed through the associative interactions between a synthetic polymer and ATP. This platform was then used to capture two types of bacteria (Escherichia coli and Pseudomonas aeruginosa (PAO1 strain)), one on the droplet’s surface and the other type inside the droplets. The rupture of the bacteria cells than led to artificial cells consisting of a coacervate with an outer, bacterium-derived membrane and bacterium derived subcompartments/container comprising active enzymes, a functional protein synthesis machinery, vacuoles and plasmid DNA. The authors demonstrated that the cell core carries out transcription and translation using the bacterial components. They also achieved subcellular organization by using an enzyme to cleave the bacterial DNA into short strands and adding a negatively charged polymer plus histones which caused the DNA to condense into a nucleus-like structure. Furthermore, actin proteins could be introduced and assembled into fibres, providing a rudimentary cytoskeleton. As a step towards self-sustainable energization E. coli cells were implanted as surrogate mitochondria for energization by ATP. This resulted in the artificial cell’s morphing into an amoeba-like shape.

Dendrimersome synthetic cells harbor cell division machinery of bacteria (Wagner et al. 2022) the authors show the reconstitution of a rudimentary/active bacterial divisome in a xenobiotic system based on dendrimersomes. The dendrimersomes were assembled from Janus dendrimers (JD) and are particularly stable compared to other synthetic compartments such as liposomes. The authors used two Janus dendrimers: JD^PC contains a zwitterionic phosphocholine headgroup and confers membrane stealth to proteins and JD^PG has a phosphoglycerol headgroup for interaction with proteins. The MinCDE protein system, which self-assembles into protein patterns, directs the assembly of FtsZ protein filaments at mid-cell in Escherichia coli to form a contractile division ring. The JDs interact with the Min proteins and this interaction can be varied by the JD^PG:JD^PC ratio. That way, Min protein patterns in dendrimersomes were regulated from irreversible binding or no attraction to a dynamic pattern formation. The compartmentalised Min proteins showed either pole-to-pole oscillations, pulsing or traveling waves, with the first one being characteristic for living cells. The authors showed that the Min system could direct the dynamic assembly and positioning of FtsZ filaments at the membrane, thus representing the first example of an active cell division machinery incorporated into a fully synthetic vesicle. This modular system could be used to develop functional synthetic cells.

In vitro assembly, positioning and contraction of a division ring in minimal cells (Kohyama et al. 2022) the authors fully reconstituted the cell division machinery of Escherichia coli in lipid vesicles. This presents a major step towards autonomous self-replication in artificial cells/towards developing a synthetic cell. The E. coli divisome consists of the MinCDE protein system guiding assembly and positioning of a presumably contractile ring based on FtsZ and its membrane adaptor FtsA. The authors used two approaches to successfully demonstrate Min wave-assisted FtsZ-ring assembly within lipid vesicles (giant unilamellar vesicles (GUVs), more than ten times the size of bacterial cells): (1) a fully controlled system with purified proteins and (2) a specifically tailored assay with cell-free protein expression in vesicles. In the first assay they could show that crowded environments are essential to form FtsZ-ring structures. The second assay supported a time-sensitive sequential expression of all the five components and showed that the timing of events is crucial. In contrast to the system with purified proteins, the time-controlled series of events of FtsZ self-assembly and MinCDE oscillation resulted in a noticeable shape transformation of the spherical vesicle along with the Min induced centric condensation of an originally isotropic FtsZ-FtsA meshwork.

Bottom-up assembly of synthetic cells with a DNA cytoskeleton (Jahnke et al. 2022) the authors realized the de novo assembly of entirely artificial DNA-based cytoskeletons with programmed multifunctionality inside vesicles and by this showed the potential of DNA nanotechnology to mimic the diverse functions of a cytoskeleton in synthetic cells. More precisely, the authors used giant unilamellar lipid vesicles (GUVs) in which the DNA cytoskeletons are repeatedly and reversibly assembled and disassembled through exposure to light using the cis−trans isomerization of an azobenzene moiety positioned in the DNA tiles. They induced ordered bundling of hundreds of DNA filaments into more rigid structures using molecular crowders and their persistence length is tuned via the choice of crowder. That way, they could achieve the formation of ring-like cortical structures inside GUVs, resembling actin rings that form during cell division. Additionally, the authors show that DNA filaments can be programmably linked to the compartment periphery using cholesterol-tagged DNA as a linker. The linker concentration determines the degree of the cortex-like network formation, and they demonstrated that the DNA cortex-like network can deform GUVs from within. In the future, it is envisaged to equip DNA filaments with molecular motors for intracellular cargo transport, force generation and contractility.

Cell-mimic directional cargo transportation in a visible-light-activated colloidal motor/lipid tube system (Ghellab et al. 2022) the authors designed a system mimicking the subcellular traffic system in living cells that is composed of colloidal motors and self-assembled lipid tubes. The colloidal motors are composed of asymmetrically gold-coated hematite and display light-activated self-propelled motion upon green-light activation when hydrogen peroxide as a fuel is present. The motion is dependent on light intensity and fuel concentration. Importantly, the motors show light-switchable binding with lipid cargoes and attachment to the lipid tubes, whereby the latter act as the motor highways. Upon assembly, the colloidal motor/lipid tube system demonstrates directional delivery of lipid vesicles, emulating intracellular transportation. The lipid tubular track allows light-switchable motion and the transportation and release of cargo thereby resembling the kinesin/microtubule system. This might be a promising platform for various applications, for example in material science or in synthetic cells.

Light-powered reactivation of flagella and contraction of microtubule networks toward building an artificial cell (Ahmad et al. 2021): the authors tried to build a self-sustained system with an ATP-driven directed movement. To achieve this, the authors constructed switchable photosynthetic liposomes that contain bacteriorhodopsin and an ATP synthase from Escherichia coli that can convert ADP to ATP upon light illumination. They coupled the module to the functional flagella module isolated from Chlamydomonas reinhardtii and observed the conversion of light into mechanical work. In another approach, the ATP regeneration module was encapsulated with microtubules and kinesin-1 molecular motors and contractions of the filamentous network were observed upon light stimulation. This work is a further step towards advanced synthetic cells which are able to use a broad range of cytoskeleton dependent motor-driven functions.

Programmable microbial ink for 3D printing of living materials produced from genetically engineered protein nanofibers (Duraj-Thatte et al. 2021) the authors developed a microbial ink that is produced by genetically engineered E. coli and can be functionalized. The ink consists of nanofibers made up of bacterial curli fibres. The curli fibres gene CsgA was modified so that the curli proteins are fused at its ends to a knob and a hole protein domain derived from fibrin. These domains allow the self-assembly into nanofibers that are printable into a 3D hydrogel. The hydrogel can be functionalized by incorporating e. g. genetically engineered E. coli that produce an anticancer drug upon induction, sequester the toxic chemical bisphenol A, or regulate their own growth.

Reconstitution of contractile actomyosin rings in vesicles (Litschel et al. 2021) the authors induced the formation of membrane-bound actin rings in giant unilamellar vesicles that resemble the contractile division ring in many cells. The authors encapsulated G-actin with the actin-bundling proteins talin and vinculin and linked the actin filaments to the phospholipid bilayer via biotin-neutravidin bonds. When muscle myosin was included, contractile actomyosin rings were formed that can be seen as a first step towards a minimal divisome for protocells. The rings can direct contractile forces directly to the membrane leading to a shape deformation of the vesicle.

Compacting a synthetic yeast chromosome arm (Luo et al. 2021) the authors have reduced the genome of Saccharomyces cerevisiae by using the loxP-sites integrated into synthetic chromosomes during the synthetic yeast genome project Sc2.0. These sites were integrated to allow for a process called SCRaMbLE (Synthetic Chromosome Rearrangement and Modification by LoxPsym-mediated Evolution) that generates genetically diverse yeast cells via inversion, deletion, duplication, and translocation upon expression of the Cre recombinase. The authors integrated selectable marker genes in the sequence between two loxP-sites, so-called loxP-units (LU), allowing the identification of strains in which an LU is lost. They tested this method using a yeast strain containing the synthetic left arm of chromosome XII (synXIIL) and identified a strain that lost 26 % (45 kbp) of synXIIL. Greater reduction of chromosome synXIIL was difficult to achieve because some nonessential genes were located in the same LU as an essential gene. To overcome this, an episomal gene array carrying essential genes was constructed to compensate for the loss of these genes. That way, some LUs containing essential and non-essential genes could be deleted resulting in a yeast strain in which 58 % (100 kb) of synXIIL was deleted. The reduction of the yeast genome is thought to produce better chassis organisms for the production of metabolites as less energy is consumed for unnecessary processes.

Programmable Aggregation of Artificial Cells with DNA Signals (Qiu et al. 2021) the authors studied programmable communication and coordination between vesicles using DNA strands. They created giant unilamellar and small unilamellar vesicles (GUV and SUV) that interact with each other upon a DNA signal. The GUVs possess pores made from DNA origami that are closed with planar DNA origami caps. The caps can be removed by a toehold-mediated strand displacement opening the pores for the incorporation of an external DNA hairpin. The hairpin DNA can bind both the GUV and the SUV, but only when cleaved by an enzyme encapsulated in the GUV. After cleavage the hairpin diffuses out of the GUV and binds both vesicles so that SUVs aggregate on the GUV. This aggregation can be reversed by a “releaser strand” complementary to the transduced hairpin. The system developed here can be used for cell-cell communication and coordinated cell behaviour in bottom-up synthetic biology.

Biner et al. (2020) constructed and characterized energy-regenerating nanovesicles (erNV) mimicking the oxidative phosphorylation of mitochondria to regenerate the pool of adenosine triphosphate (ATP). In order to generate erNVs, a lipid mixture was mixed with a minimal set of respiratory chain relevant units, such as ubiquinone-10 (Q10), ATP synthase complex (Ec-F1F0) and complex I (mito-CI) from common cattle (Bos taurus) and the alternative oxidase (AOX) from Trypanosoma brucei brucei. In theory, the erNVs should be able to use electrons from the oxidization of the substrate nicotinamide adenine dinucleotide (NADH) to NAD+ to generate a electrochemical proton motive force (Δp) and synthesize ATP. Experimental data from different assays confirmed that all components are inserted predominantly in their functional active orientation within the vesicle membrane and that the erNVs are able to produce ATP. Compared to sub-mitochondrial or sub-bacterial vesicles prepared from native mammalian or bacterial membranes, the erNVs are able to maintain a similar and high Δp, meaning only 50 % of the protons are lost due to leakage and ATP can be stably synthesized. This work represents the first example of a semi-synthetic system to drive ATP regeneration by NADH oxygenation. - to the original literature

Buddingh’ et al. (2020) constructed an intercellular allosterically activated communication network between artificial cells. In order to implement the communication platform, two populations of artificial cells (giant unilamellar vesicles) were designed. The “sender” cells respond to an external trigger by producing adenosine 5′-monophosphate (AMP) as a small signalling molecule that is released into the medium. The “receiver” cells recognize and amplify the signal, using the allosteric AMP dependent activation of the glycogen phosphorylase b (GPb). Compared to the background activity in absence of AMP, even low concentrations of AMP lead to a conformational change in GPb in “receiver” cells, resulting in a high rate of glycogen phosphorolysis and an 80-fold increase of nicotinamide adenine dinucleotide (NADH) production. This is the first allosterically activated communication network, which not only fully relies on the protein machinery and small molecules as chemical information agents but is also able to facilitate the propagation of signals over long distances within the artificial cell consortia. - to the original literature

Fan et al. (2020) generated a simplified synthetic biology chassis by creating chromosome-free simple cells (SimCells) from Escherichia coli, Pseudomonas putida and Ralstonia eutropha. To remove the native chromosomes, the heterologous I-CeuI endonuclease and the degradative activity of endogenous nucleases were used to induce several double strand breaks within the chromosomes and the complete breakdown of the genome. The generated SimCells remained functional and stable. In fact, SimCells were able to express introduced synthetic pathways for at least ten days. In a proof of principle experiment, SimCells, synthesizing anti-cancer drugs (e.g. catechol) could be used to significantly decrease cancer cell viability in cell culture after incubation with these SimCells. SimCells represent a synthetic tool that can be programmed to manufacture and deliver medicinal products safely without interference with the host genome. - to the original literature

Yang et al. (2020) developed protocells as minimalistic communication models for cell-cell communication studies based on the model published in 2019 by Joesaar et al.. The authors built two different proteinosome-based semipermeable protocells containing biotinylated DNA complexes capable of either sending or receiving. Upon laser irradiation a photocleavable nitrobenzyl linker is cleaved off and releases a DNA strand. This DNA strand can reach the receiver cells where it results in a fluorescent response. By changing the experimental conditions, the authors showed that the signal range is determined by several factors such as density and permeability of receivers, extracellular signal degradation, signal consumption and catalytic regeneration. They also created receiver cell with an AND receiver that was activated spatiotemporally by two sender cells.- to the original literature

Berhanu et al. (2019) developed an artificial cell system with a self-supplying energy production for protein synthesis. They created giant unilamellar vesicles with an artificial organelle that contained the membrane proteins bacteriorhodopsin and F-type-ATP-synthase. The bacteriorhodopsin pumps protons into the organelle which can be used by the ATP synthase to produce ATP. When a modified transcription-translation system was added the produced ATP can be used as a substrate for messenger RNA (mRNA), for the phosphorylation of GDP and for the aminoacylation of tRNA. As a result, bacteriorhodopsin and subunits of the ATP synthase could be synthetized de novo in this system thereby increasing photosynthetic ATP production in the artificial organelles. - to the original literature

Diederichs et al. (2019) - The authors developed molecules for the formation of nanopores with a diameter of 20.5 nm and an average height of 31.5 nm, through which folded proteins such as fluorescent proteins can also be transported. The nanopores comprise squarely arranged DNA duplexes that are provided with cholesterol-lipid anchors for insertion in the membrane. There thus result 50 nm2-sized pores that could be inserted into artificial cells. - to the original literature

Ghosh et al. (2019) - A diffuse movement during substrate turnover can be observed with many enzymes, increasing even more with an increase in the substrate concentration. Making use of this fact, Ghosh et al. produced phospholipid vesicles that contained Na+/K+-ATPase as a transmembrane molecule. These vesicles were then tested in culture under physiological conditions and showed directed movement, followed by randomised reorientation. These vesicles represent a first step in the direction of autonomous nanovesicles and can be used e.g. as transport vesicles for biomolecules. - to the original literature

For the further development of synthetic cells, mechanosensitive mechanisms for various uses were implemented by two different research groups, as follows:

Garamella et al. (2019) constructed protocells that react to environmental influences and can independently counteract a stimulus. For this, synthetic cells with a phospholipid membrane were produced that are mechanosensitive through the incorporation of MscL ion channels. The intracellular space of the synthetic cells contains plasmids for expression of a bacterial cytoskeletal protein (MreB) and also an in vitro transcription- and translation system. By changing the outer environmental conditions toward a hypoosmolar exterior there arises a mechanical stress to the membrane and thus pore formation through the MscL. A regulator molecule located in the extracellular medium subsequently flows in, which makes the expression of the MreB possible. MreB then migrates to the membrane and increases the mechanical robustness there. An adaptive synthetic cell system could herewith be created that showed resistance to osmolar stress. - to the original literature

Hindley et al. (2019) established a mechanosensitive signal path based on the sPLA2-membrane-MscL network by Charalambous et al. (2012). The calcium-dependent soluble phospholipase A2 (sPLA2) thereby served as the catalyst to synthesize lysophosphatidylcholine, which is incorporated into the membrane of artificial cells. This in turn leads to a change in the membrane mechanics and to the opening of the ion channel MscL. Building on this, Hindley et al. created artificial cells that contained mechanosensitive vesicles with the fluorescent dye calcein as well as inactivated sPLA2. The subsequent extracellular addition of alpha-haemolysin led to the formation of pores in the outer membrane, then to calcium influx and to activation of sPLA2. The changed mechanics of the vesicle membrane and the resulting opening of the MscL was verified by means of increased calcein fluorescence. A controlled calcium inflow for future signal transduction in artificial cells was thus established. - to the original literature

Joesaar et al. (2019) developed consortia of protocells that are able to correspond with each other through orthogonal channels. These protocells possess a proteinosome that is permeable for short (< 100 bases) single-stranded DNA (ssDNA) and that protects the intra-protocellular DNA from degradation in the culture medium. The protocells are able to communicate via those ssDNA, by using the following scheme: cell one releases an input-ssDNA, this input-ssDNA enters the next cells and replaces another ssDNA with a short mismatch. This second ssDNA is subsequently released and can diffuse into another cell. This system demonstrates a bidirectional communication between different protocell-populations. - to the original literature

Venetz et al. (2019) synthesized a minimal bacterial genome of the model organism Caulobacter ethensis. Already known essential genes were newly assembled with the help of an algorithm, while keeping the original gene organization and orientation. In addition, the genome was streamlined for synthesis. For this purpose, 10,172 base substitutions were introduced to eliminate 1233 repeats, 93 homopolymeric regions, 4342 regions of high GC content and 1045 endonuclease restriction sites. On top, another 123,141 base substitutions were introduced into protein coding regions. The newly synthesized genome (C. eth-2.0) was condensed to 785 kb with 676 protein-coding, 54 non-coding and 1015 intergenic sequences. 56.1 % of all codons were replaced by synonymous codons to eliminate 87.4 % of alternative ORFs, 95.3 % of predicted internal transcription initiation sites and 76.7 % of ribosome stalling motives. 81.5 % of C. eth-2.0 genes demonstrated functionality, indicating that the primary sequence of the mRNA, the secondary structure, or a codon´s context in the genome seem to have no significant impact on the biological function of the coded protein. - to the original literature

Niederholtmeyer et al. (2018) manufactured a synthetic protocell with a porous membrane and a primitive “nucleus”. This “nucleus” is a hydrogel compartment built from clay minerals and contains the cellular DNA. The synthetic cells can take up all components of a cell-free transcription and translation system and express the DNA from their “nucleus”. The expressed proteins are able to diffuse to neighboring synthetic cells as a form of cell-cell communication, enabling an artificial quorum sensing. - to the original literatur

Shao et al. (2018) merged all 16 Chromosomes of S. cerevisiae into one so-called super-chromosome, in which all non-essential, telomeres, centromeres as well as 19 repetitive sequences were removed by using CRISPR-Cas9 and the residual chromosome-parts were fused. In a similar approach, Lou et al. (2018) were able to reduce the number of yeast chromosomes to two. These experiments were based on the observation that the number of chromosomes in eukaryotes seems to be rather arbitrary and independent of the quantity of genetic information.
Indeed, the reduction of chromosomes had only little impact on most of the yeast´s characteristics. The cells showed a similar phenotype and only modest transcriptome changes. Growth was slowed-down only in the yeast with a single chromosome. Both yeasts showed a significant reduction in the formation of ascospores as part of sexual reproduction. By reorganizing the genome the new organism was unable to cross-breed with wild-type yeast and is thus reproductively isolated. Hereby, the classic criterion for the assignment to different biological species is met. - to the original literature: Shao et al. (2018), Luo et al. (2018)

Design and evolution of artificial enzyme with in-situ biosynthesized non-canonical amino acid (Huang et al. 2025)

The authors present a new method for generating and optimising artificial enzymes (designer enzymes) directly within living cells. The method combines the in situ-biosynthesis of non-natural amino acids (ncAA) and genetic code expansion. Using this strategy, an ‘S-functionalised, cysteine-dependent enzyme‘ (SFC enzyme), containing S-(4-aminophenyl) -L-cysteine (pAPhC) was developed in the bacterium Escherichia coli. After three rounds of directed evolution, the enzyme catalyses a Friedel-Crafts alkylation reaction with high yield and selectivity and exhibits reversed stereoselectivity compared to known enzymes. Friedel-Crafts alkylation is one of the most important reactions in organic chemistry and serves to attach alkyl groups to an aromatic ring (e.g. benzene or indole). Using the engineered designer enzyme, this catalytic activity, which is not known to occur in nature, was efficiently realized in living cells for the first time. 

Efficient genetic code expansion without host genome modification (Costello et al. 2024)

Quadruplet codon usage is a strategy for genetic code expansion with non-native codons that is often less efficient than triplet decoding. The authors developed a strategy to improve incorporation of non-canonical amino acids (ncAAs) into proteins without altering the Escherichia coli host cell. They used a superfolder GFP reporter, in which the permissive Y151 was replaced by a quadruplet codon. This design will only allow GFP expression when the quadruplet codon is read by a tRNA. Furthermore, the codon usage of the five adjacent residues upstream and downstream was modified. With this approach, it could be shown that quadruplet decoding has a codon usage bias with a strong preference for a high-usage codon 3´ to the quadruplet codon and that off-target frameshifts could be avoided when high-usage codons were used. Furthermore, 12 mutually orthogonal transfer RNA (tRNA)-synthetase pairs were identified to incorporate ncAAs into proteins and five were evolved so that they can incorporate a repertoire of ncAAs at orthogonal quadruplet codons. Using these strategies, the authors built an in vivo biosynthesis platform to create > 100 new-to-nature macrocyclic peptides with up to three unique ncAAs. This could be a step towards building chemically diverse polymers in a programmable manner.

Adding α,α-disubstituted and β-linked monomers to the genetic code of an organism (Dunkelmann et al. 2024)

To incorporate non-canonical monomers (ncMs, non-canonical amino acids or other molecules that are normally not incorporated into proteins) into proteins an orthogonal tRNA synthetase has to acylate the ncM to an orthogonal tRNA, which will then be bound by the ribosome for translation. ncM are often poor ribosomal substrates and cannot be acylated to the orthogonal tRNA efficiently. The authors developed a strategy to identify tRNA variants capable of ncM incorporation. They established a method that detects the acylation status of a tRNA, thus detecting whether the ncM is loaded onto the tRNA. Next, a fusion of a split tRNA and the mRNA of a pyrrolysyl-tRNA synthetase (PylRS) was created that allowed correlating the mRNA synthetase sequence with its activity. To identify new PylRS variants capable of incorporating ncMs, large libraries of these hybrid RNAs with mutated residues in the PylRS active site were generated and cultured with ncMs that are either non-translatable or poor ribosomal substrates. The acylated hybrid RNAs were then isolated and reverse transcribed to obtain the PylRS gene responsible for acylation. Using this approach, PylRS for the ncM classes β-amino acids, α,α-disubstituted amino acids and β-hydroxy acids were identified and some of them incorporated into proteins. This incorporation of new ncMs into proteins may be useful in creating new drug-like molecules in living cells.

Designing artificial fluorescent proteins and biosensors by genetically encoding molecular rotor-based amino acids (Hu et al. 2024)

Fluorophores in fluorescent proteins like GFP are encoded by the primary amino acid sequence, that folds into a fluorescent rotor structure upon protein folding followed by spontaneous cyclization. The authors designed fluorogenic non-canonical amino acids with this rotor structure, the so-called fluorescent molecular rotor amino acids (FMR-AAs). When integrated into a protein via expanded genetic code, the fluorophores transform proteins into fluorescent proteins. The technique was used to establish biosensors in mammalian cells, that detect, for example, changes in intracellular calcium concentration or protein interactions. That way, customized fluorescent proteins for molecular visualization and investigation can be created.

Expanding the substrate scope of pyrrolysyl-transfer RNA synthetase enzymes to include non-α-amino acids in vitro and in vivo (Fricke et al. 2023)
The authors describe the first aminoacyl-tRNA synthetase enzyme that accepts α-thio acids and α-carboxy acids that could support carbon-carbon bond formation within the ribosome, increasing the number of non-L-α-amino acids that can be incorporated into proteins in vivo. The structure of Methanomethylophilus alvus pyrrolysyl-tRNA synthetase (MaPylRS) variant MaFRSA was investigated, showing the ability to support biosynthesis of proteins with internal aromatic α-hydroxy acid monomers. The authors conclude that combined with synthetic genomes, ribosomes capable of carbon-carbon bond formation would set the stage for template-driven biosynthesis of unique hybrid biomaterials and sequence-defined polyketide-peptide oligomers.

For the year 2022, no publications from the field of xenobiology were selected for publication on the ZKBS homepage.

Multiplex suppression of four quadruplet codons via tRNA directed evolution (DeBenedictis et al. 2021) the authors used directed evolution to establish a small set of quadruplet tRNAs (qtRNAs) that incorporate amino acids under multiplexing conditions. By introducing a quadruplet codon as a reporter into the β-galactosidase gene, codon-anticodon interactions were first validated in E. coli. With this strategy, 24 qtRNAs were identified that would decode the introduced codon and, as a result, charge a defined amino acid onto the protein. The identified qtRNAs were further evolved by directed evolution in phages. The phages could only propagate if a certain quadruplet codon was decoded. The authors showed the translation of six UAGA-codons in a row as well as the translation of four different quadruplet codons in the same protein. The work is a further step towards an exclusively quadruplet codon translation system.

Sense codon reassignment enables viral resistance and encoded polymer synthesis (Robertson et al. 2021) the authors liberated sense codons to incorporate multiple distinct non canonical amino acids (ncAAs). Using an engineered tRNA-aminoacyl-tRNA synthetase (aaRS) pair, ncAAs can be incorporated as a reaction to a specific codon. The authors used the E. coli strain Syn61 developed by Fredens et al., in which two serine codons (TCG and TCA) as well as the stop codon UAG were exchanged for synonymous codons and evolved the strain so that the tRNAs and release factor 1, which decodes TCG, TCA, and TAG, were deleted. The deletion of these tRNAs and the release factor made the cells resistant to phages that harbour the three codons in their genome. The codons were also reassigned for the incorporation of three distinct ncAAs into a protein. Incorporation of ncAAs into proteins can be used for the development of new biomaterials or biotherapeutics.

Chen et al. (2020) created an Escherichia coli strain, which autonomously biosynthesizes 5-hydroxytryptophan (5HTP) as a 21st noncanonical amino acid (ncAA) and incorporates it into proteins. This represents a tool for the evolution of novel proteins containing ncAAs and may be used to generate new therapeutic proteins. In order to use 5HTP as a ncAA in prokaryotes the hydroxylation machinery of tryptophan (the substrate for 5HTP-generation) had to be introduced into E. coli. In a first step, a plasmid harbouring a specific hydroxylase (e.g. XcP4H from Xanthomonas campestris), an artificial recycling pathway for the hydroxylase co-factor tetrahydromonapterin (MH4), and an expression cassette for the green fluorescence protein (GFP) containing one amber stop codon at Tyr151 was generated. A second plasmid which contains a modified tryptophanyl-tRNA synthetase (ScTrpRS)/tRNATrp from Saccharomyces cerevisiae provides the bioorthogonal machinery for incorporating 5HTP at the amber codon. Both plasmids were transferred into E. coli which led to a hydrolase- and MH4 recycle genes-dependent expression of GFP with an efficient and specific 5HTP incorporation. As a proof concept, Chen et al. used E. coli to express the 5HTP containing single-chain variable fragment (scFv) of the anti-human epidermal growth factor 2 (HER2). - to the original literature

Fischer et al. (2020) identified further unnatural codons that could be inserted into the semi synthetic organism (SSO) generated by Zhang et al. (2017). Since the SSO (Escherichia coli) is able to replicate and transcribe the unnatural base pair (UBP) NaM and TPT3 in different sequence contexts, Fisher et al. used this UBP to systematically analyse unnatural codons. Inserting the NaM/TPT3 base pair has the potential to create new codons. In order to test these unnatural codons, GFP-expressing plasmids with NaM or TPT3 at the first, second or third position of codon 151 and tRNAs with cognate unnatural anticodons to enable the incorporation of non-canonical amino acids (ncAAs) in the protein were constructed. Plasmids were then used to transform E. coli. With the help of GFP fluorescence, nine functional codon-anticodon pairs were identified. All of them are stably encoded in the DNA and are transcribed into mRNA and bound by their specific tRNA, thereby efficiently mediating the decoding at the ribosome. Moreover, three of these codon-anticodon pairs were further characterized in E. coli by inserting them simultaneously into the GFP gene along with the corresponding anticodons that were inserted into three tRNAs capable of charging different ncAAs. The three codons were orthogonal to each other and could be simultaneously decoded within the SSO, thereby enabling the cell for the first time to decode 67 codons. - to the original literature

Flamme et al. (2020) have enzymatically created unnatural base pairs (UBPs) on the basis of metal ions. These metal base pairs are orthogonal to the natural Watson-Crick base pairs and integrate easily into the DNA´s structure. The incorporation of metal base pairs expands the chemical diversity of DNA and RNA and can be used for novel nanomaterials.
The authors used two synthetic analogues for the enzymatic synthesis of a silver-mediated UBP: the imidazole containing nucleoside dIM^C and the 6-pyridylpurine nucleoside dPur^P. This silver-mediated UBP (dIM^C - Ag^I - dPur^P) was then incorporated into a DNA duplex using a two-step synthesis protocol. - to the original literature

Lee et al. (2020) addressed how the incorporation of non-canonical, backbone extended amino acids (amino acis analogs) into natural polypeptides can be improved and present a further step towards synthesizing non-canonical sequence-defined polymers. The charging of a tRNA with a long chain carbon structure is currently difficult and was studied in the flexizyme system (Fx, an aminoacyl tRNA synthetase-like ribozyme). It was shown that charging of amino acid analogs with linear carbon chains was hindered by the formation of lactams. To address this, intramolecular lactam formation was prevented by steric restriction of the amino and activated ester functionalities. This architecture used a rigid spacer like vinyl or a cyclic centre structure in the central region of the amino acid and increased the overall charging efficiency of the Fx-system. To transfer these findings to a more natural RNA translation, the Fx-charged tRNAs were applied to a cell-free protein synthesis system to incorporate the amino acid analogs at the C- and N-terminus during protein synthesis. It was shown that every substrate that could be loaded onto tRNAs was successfully incorporated into a peptide by wild-type as well as by engineered ribosomes. - to the original literature

Yang et al. (2020) have expanded the studies of Eremeeva et al. (2017), who showed the amplification of DNA with four non-canonical base pairs (DZA). These chemical modifications can enhance the stability and versatility of nucleic acids for new applications, e. g. in biomedical research. The authors tested fully base-modified RNA (RZA) in replication, transcription and reverse transcription and established a flow of genetic information between natural and unnatural genetic information. In vitro, the PCR-synthesized DZA could be transcribed into RZA or into unmodified RNA, and reverse transcribed into DNA or DZA. In vivo, DZA was transcribed and translated into the red fluorescent protein mTangerine in Escherichia coli. - to the original literature

Calles et al. (2019) produced a genetic fail safe code that can be read by the natural translation machinery and features a protective effect against spontaneous mutations. Each amino acid is thereby coded by a single codon, all other codons for this amino acid were non-translatable "null-codons". Ideally, the mutation of one of these sense codons should not lead to another sense codon. This is only possible with a 3-base code if only 16 amino acids are coded. However, if a 4-base code is used, all 20 amino acids and, additionally, non-natural amino acids can be coded without a mutation resulting in another sense codon. In a first application of the fail safe code an in vitro transcription and translation system was used. In this system all natural tRNAs were depleted and replaced by 20 tRNAs, that recognized the 20 codons of the fail safe code. No tRNAs were thus available that could recognize the "null-codon“. -to the original literature

Hoshika et al. (2019) expanded the genetic code by adding four synthetic nucleotides und thus created an 8-letter code (Hachimoji). The synthetic 8-letter DNA is largely similar to the natural DNA: it forms a double helix in which interaction occurs via hydrogen bonds and can be transcribed into RNA. However, there are no organisms so far that use and stably pass on the 8-letter code to the next generation. - to the original literature

Kneuttinger et al. (2019) - The authors inserted non-natural amino acids (nnamino acids) in the bienzyme imidazole-glycerol phosphate synthase to achieve the activation of the enzyme by means of light. In the original mechanism the activation of the HisH subunit took place exclusively through the binding of a substrate to the HisF subunit. By incorporating the three light-sensitive nnamino acids phenylalanine-4'-azobenzene, o-nitropiperonyl-O-tyrosine and methyl-o-nitropiperonyllysine in the substrate binding enzyme subunit a light-inducible and reversible increase in activation of the HisH activity was achieved, so that spatial and temporal control of enzyme activity is made possible through light. - to the original literature

Koh et al. (2019) produced auxotrophic Escherichia coli bacteria in which the β-subunit of DNA polymerase III-holoenzyme at position 273 contained synthetic p-benzoylphenyl-alanine (pBzF) instead of a leucine and that therefore could grow only in the presence of pBzF in the medium. The β-subunit that is also described as a sliding clamp is strictly conserved in Gram-negative bacteria, so that the auxotrophic strain features a non-detectable escape frequency. The described auxotrophy can additionally be easily transferred to other bacteria such as Pseudomonas aeruginosa or Acinetobacter baumanii in order to produce conditionally viable living vaccine against these hospital germs. - to the original literature

Reinkemeier et al. (2019) work on the orthogonal translation of the amber stop codon (TAG). To liberate this codon for the incorporation of a non-canonical amino acid into a protein, an orthogonal tRNA/tRNA synthetase pair is required. The tRNA/tRNA synthetase pair and a mRNA, whose amber stop codon was to be recoded, were brought into close proximity inside an artificial organelle. For this purpose, the tRNA/tRNA synthase pair and an RNA-binding domain were fused to a so-called assembler protein. Proteins used as assembler proteins may be proteins that undergo a phase separation in the cell or kinesins that enrich at the microtubule plus ends. The tRNA/tRNA synthase pair as well as an mRNA with a recognition sequence for the RNA-binding domain are enclosed in close proximity, enabling the recoding of the stop codon only in selected RNAs. - to the original literature

Swenson et al. (2019) developed a synthetic, bilingual biopolymer that unites the properties of an amino acid sequence and a nucleotide sequence in a single peptide nucleic acid (PNA) scaffold. The PNA strand thereby has amino acid-like side chains that, based on their arrangement, feature a similar aggregation or folding behaviour as a peptide. Additionally, the PNA strand however also carries a nucleotide in a specific sequence succession that can be bound by complementary DNA or RNA. As an application, the authors showed that PNA strands with hydrophobic and hydrophilic amino acid side chains come together into micelles in an aqueous milieu. The addition of RNA complementary to a nucleic acid sequence contained in the PNA strand caused the micelles to become disordered again. The PNAs should also be used to mediate a precise interaction between various target proteins and nucleic acids. - to the original literature

Simulating 500 million years of evolution with a language model (Hayes et al. 2025)

The authors present ESM3 (Evolutionary Scale Modeling version 3), a multimodal AI language model for proteins that models sequence, structure, and function in an integrated way and can generate novel proteins in a targeted manner. Unlike previous protein language models, which mostly consider only amino acid sequences, ESM3 additionally integrates three-dimensional structural information as well as functional properties within a unified model. ESM3 was trained on billions of protein sequences as well as hundreds of millions of protein structures and functional annotations.

The authors show that the model can solve complex design tasks, such as the incorporation of functional binding sites into novel protein structures, as well as the generation of proteins with low sequence and structural similarity to known proteins. Through targeted inputs (prompts), desired properties, structural motifs, or functions can be specified, upon which ESM3 generates appropriate protein sequences. As a central experimental example, the authors present the development of a novel fluorescent protein (esmGFP). This protein exhibits functional fluorescence despite sharing only about 58% sequence identity with known fluorescent proteins. The authors estimate that this sequence distance corresponds to an evolutionary timescale of over 500 million years relative to the closest known naturally occurring protein.

Megabase-scale human genome rearrangement with programmable bridge recombinases (Perry et al. 2026)

The authors developed a new class of RNA-guided recombinases (bridge recombinases) for targeted genome manipulation in human cells. In contrast to classical CRISPR systems, which recognize only a single DNA target sequence via a guide RNA, bridge recombinases can bind two different DNA sequences simultaneously using a bridge RNA (bRNA). This enables not only gene insertions but also precise excisions, inversions, and large-scale genomic rearrangements.

Starting from 72 natural bridge recombinase systems of bacterial origin, the authors identified ISCro4 from Citrobacter rodentium as particularly efficient for genome editing in human cells. For genetic optimization, both the recombinase itself and the associated bRNA were further engineered. Through a deep mutational scan of the recombinase in human cells, the mutations S30T, P54Q, and S243H were identified as activity-enhancing. The combined enhanced ISCro4 variant achieved an insertion efficiency of up to 20% and enables genomic rearrangements on the megabase scale.

As a proof-of-concept, several therapeutically relevant applications were demonstrated, including precise excision of the BCL11A enhancer (relevant for the therapy of sickle cell disease and β-thalassemia) as well as removal of pathogenic GAA repeat expansions in the FXN gene, which are causative for Friedreich’s ataxia, a rare, genetic and progressive neurological disorder.

High-fidelity human chromosome transfer and elimination (Petris et al. 2025)

The authors describe a new method that enables human chromosomes to be transferred between cells, genetically modified, and subsequently reintroduced into human cells. The long-term goal of this approach is the construction of synthetic human genomes.

As a chromosome donor, the non-transformed, immortalized human RPE1 cell line was used, in which chromosomes were labeled with fluorescent and antibiotic resistance markers using a PiggyBac system. Condensed chromosomes were then isolated from mitotic cells and transferred via transfection into murine embryonic stem cells (mESCs). These so-called assembly cells contain only a single copy of the respective human chromosome, which facilitates precise genetic modifications. Building on this system, the authors demonstrated targeted genetic engineering by modifying a defined region on chromosome 20. First, a synthetic landing pad sequence carrying fluorescent and selection markers was integrated into chromosome 20. This target region was then replaced using CRISPR/Cas9, homologous recombination, and a BAC vector with approximately 10 kb of synthetic DNA containing designed silent mutations (watermarks).

In the next step, the modified chromosomes were transferred back into human cells using an improved microcell-mediated chromosome transfer (R-MMCT) method. This process enabled the generation of targeted aneuploidies such as trisomies and tetrasomies. Subsequently, the endogenous chromosome was removed via CRISPR-mediated cleavage in the centromeric region, resulting in diploid cells that contained exclusively the synthetically modified chromosome.

Using various methods, including whole-genome sequencing and optical genome mapping, the authors demonstrated that chromosomes remained largely intact throughout the entire process. Only a small number of unwanted mutations such as SNPs (single nucleotide polymorphisms) and INDELs (insertions and deletions) and minor structural variations were observed, with frequencies comparable to those in control cells.

Ultra-high-throughput mapping of genetic design space (Rai et al. 2026)

The authors present CLASSIC (combining long- and short-range sequencing to investigate genetic complexity) as a high-throughput platform for the analysis of complex genetic circuits, whose experimentally generated datasets can subsequently be used for AI-assisted prediction of novel circuit designs.

The central set of training data was generated by constructing and barcoding 100.000 gene circuits, which are then tested within human 293T cells. These synthetic circuits followed a common functional principle: an engineered transcription factor is activated by the inducer 4-hydroxytamoxifen (4-OHT), subsequently translocates into the nucleus, and regulates the expression of a reporter gene (e.g., GFP). By systematically varying genetic components such as promoters, activation domains, DNA-binding motifs, spacing between expression units, and other regulatory elements, the authors generated large libraries of distinct circuit designs, which were then stably integrated into the 293T cells.

The integrated circuit variants were subsequently functionally analyzed. For this purpose, CLASSIC combined two sequencing modalities: long-read sequencing (Nanopore) was used to determine the complete DNA composition of each construct together with its associated barcode, enabling an unambiguous mapping between genetic composition and barcode identity. The cells were then sorted according to reporter expression (fluorescence intensity), and short-read sequencing (Illumina) was used to quantify only the barcode sequences within the different expression bins. This enabled reconstruction of which genetic circuit designs were associated with specific expression levels.

The resulting datasets were subsequently used to train machine-learning models capable of predicting the behaviour of previously untested gene circuits. The study demonstrated that these models could reliably infer complex gene-expression patterns and identify functionally optimized circuit designs, even across extremely large genetic design spaces containing up to billions of possible combinations.

Overall, the study presents a scalable, data-driven framework for synthetic biology that substantially accelerates the design–build–test–learn cycle and enables a more systematic and predictable engineering of complex genetic systems.

A generalized platform for artificial intelligence-powered autonomous enzyme engineering (Singh et al. 2025)

The authors developed a platform for autonomous enzyme engineering that integrates machine learning methods, large language models (LLMs), and biofoundry-enabled laboratory automation. The aim of the study is to engineer proteins with desired properties more efficiently, with minimal human intervention and without the need for domain-specific expertise. The system requires only the amino acid sequence of a protein and a quantifiable target function (fitness function), such as enzyme activity, to evaluate whether a protein variant exhibits higher or lower fitness or activity relative to other variants.

The platform combines AI-guided evaluation and prioritization of protein variants with the automated biofoundry platform (iBioFAB, Illinois Biological Foundry for Advanced Biomanufacturing), a robotics-based laboratory for synthetic biology that autonomously generates, expresses, and characterizes mutants. Importantly, no de novo protein design is performed; instead, existing enzymes are iteratively optimized through targeted amino acid substitutions (mutations). For the initial evolutionary round, a first mutational library is designed using protein language models such as ESM-2 and EVmutation. Experimental results are then used to train a supervised machine learning model, which predicts improved protein variants in subsequent cycles.

To demonstrate the approach, the authors optimized two different enzymes: the halide methyltransferase from the plant Arabidopsis thaliana (AtHMT) and a phytase from the bacterium Yersinia mollaretii (YmPhytase). Over four rounds of evolution within four weeks, the platform achieved a 16-fold increase in ethyltransferase activity for AtHMT and a 26-fold increase in phytase activity at neutral pH. In each case, fewer than 500 variants were experimentally screened.

The study highlights the potential of autonomous experimentation platforms to significantly enhance efficiency and scalability in protein engineering and biotechnology research.

Establishing a synthetic orthogonal replication system enables accelerated evolution in E. coli (Tian et al. 2024)

The authors established a synthetic DNA replication system in living Escherichia coli bacteria, that works independently of the host cell`s natural replication process. It is based on four genes essential for in vivo-replication of the bacteriophage PRD1 genome combined into a single replication operon controlled by an IPTG-inducible promoter and not copied by host polymerases. The operon-bearing E. coli were able to replicate a linear replicon, in which a gfp and a kanamycin resistance gene were expressed under control of constitutive promoters flanked by bacteriophage ITR sequences. The efficiency of linear replicon expression was improved by adding expression of a nuclease inhibitor, therefore protecting linear double-stranded DNA from degradation. A replicon of 16.5 kb was successfully replicated. Under selection pressure replication was stable for over 300 generations. When three orthogonal replicons were maintained simultaneously stability lasted over 100 generations. By adding an inducible dCas9 to repress the IPTG-responsive promoter the system allows regulation of the copy number and therefore evolutionary dynamics. Specific mutations in orthogonal DNA polymerases increase the mutation rates of the replicon 100 to 10,000-fold without affecting the mutation rate of the genome. The system therefore provides a platform for accelerated continuous mutational evolution that is simple, stable and scalable.

Oriented triplex DNA as a synthetic receptor for transmembrane signal transduction (Chen et al. 2024)

The authors generated a synthetic receptor based on DNA. The transmembrane receptor is cholesterol-anchored across a lipid bilayer membrane and spans the membrane twice, with a third DNA strand facing outwards. A complementary helper sequence ensures the outward orientation. Induced by lowering the pH, conformational changes lead to a transition of the third strand from the outer to the inner membrane leaflet. This results in the receptor now spanning the membrane three times, starting transmembrane signal transduction. This work presents a versatile design for synthetic receptors characterized by modularity, programmability, and controllability.

Sequence modeling and design from molecular to genome scale with Evo (Nguyen et al. 2024)

The authors trained an artificial intelligence (AI) model, named Evo, on millions of bacterial and phage genomes to enable predictions about DNA, RNA and protein functionalities as well as generation of complex systems based on CRISPR-Cas and transposable elements. Evo generated DNA sequences with plausible genomic architecture more than 1 megabase in length. The output was experimentally validated giving the first examples of protein-RNA and protein-DNA codesign via AI. Evo learns how small changes in DNA sequence affect the fitness of the whole organism. In conjunction with new techniques for large-scale genome modification, Evo expands the application possibilities in biotechnology and biological design to the scale of whole genomes.

Cell-type-directed design of synthetic enhancers (Taskiran et al. 2024)

The authors designed enhancers (short DNA sequences that facilitate cell-type specific gene expression) with the help of already validated deep learning models. Three sequence design strategies have been used. The first strategy was a nucleotide-by-nucleotide sequence in silico evolution. It started from 500 bp random sequences and introduced mutations iteratively to target specific brain cells of the fruit fly or human cancer cells. To predict scores for the mutations, a deep learning model was used. The enhancers were then tested in transgenic lines. Additionally, it was tested if an enhancer that is active in a single cell type could be altered to become active in a second cell type as well as if an enhancer active in several cell types can be altered to become active in a single cell type. In a second strategy a combination of known activator motifs was used to design a cell-type-specific enhancer. In the third strategy generative adversarial networks have been used to generate functional and specific enhancers. In conclusion, this proof-of-concept study shows that enhancer design strategies are adaptable to different biological systems and even other species, including human.

A split ribozyme that links detection of a native RNA to orthogonal protein outputs (Gambill et al. 2023)
The authors developed a strategy for the detection of specific RNA molecules through the use of ribozymes. In this plug-and-play strategy, called Ribozyme-ENabled Detection of RNA (RENDR), a cellular RNA input activates a splicing reaction, which in turn generates an mRNA that codes for any reporter protein (e. g. green fluorescent protein, GFP). The strategy is based on a ribozyme that has been synthetically cleaved into two non-functional fragments, to each of which RNA guide sequences destined to interact with the RNA input are attached. In the presence of the RNA input sequence, the two transcribed ribozyme fragments are colocalised and form a functional ribozyme complex, which splices the target mRNA leading to the expression of the functional reporter protein.
A crucial step in the development of the RENDR platform was the identification of functional split sites in the structure of the splicing ribozyme of Tetrahymena thermophila. For this purpose, a high-throughput laboratory evolution approach was used in which in vitro transposon mutagenesis was coupled with fluorescence-activated cell sorting. Different split sites were identified that enable high sensitivity and dynamics in RNA recognition.
In addition, the authors also developed further RENDR variants that allow for the facile exchange of different genes encoding protein outputs (colorimetric, gaseous, and regulatory outputs) and demonstrate the transferability to different Gram-negative bacteria (Escherichia coli, Shewanella oneidensis and Vibrio natriegens). Finally, the production of pigment-producing enzymes in the presence of antibiotic resistance genes in E. coli was demonstrated as a proof of application, providing a cost-effective and simple approach for the detection of antibiotic-resistant microbes.

A polycistronic system for multiplexed and precalibrated expression of multigene pathways in fungi (Yue et al. 2023)
The authors developed a strategy for the assembly of multigene pathways with the required expression level of each gene referred to as HACKing (Highly efficient and Accessible system by CracKing genes into the genome). The strategy is based on the integration of a 9 bp nucleotide sequence (short intergenic sequence 6 (IGG6)) that enables efficient polycistronic gene expression in yeasts and filamentous fungi, coupled with multiplexed CRISPR/Cas9-based genome editing. Results suggest that IGG6 mediates the re-initiation of translation, allowing separate genes to be joined into a functional operon. With HACKing, the expression level of individual enzymes can be pre-calibrated by linking their translation to that of host proteins whose abundance has been pre-determined under the desired fermentation conditions. The HACKing strategy was validated with the construction of biosynthetic pathways of endogenous or heterologous terpenoid products in Saccharomyces cerevisiae. Due to its predictability, simplicity, scalability and speed, HACKing can be utilised in fungal metabolic pathway engineering for valuable metabolites.

CRAPS: Chromosomal-Repair-assisted Pathway shuffling in yeast (Dykstra et al. 2023)
A fundamental challenge of metabolic engineering involves assembling and screening vast combinations of orthologous enzymes across a multistep biochemical pathway. Current pathway assembly workflows involve combining genetic parts ex vivo and assembling one pathway configuration per tube or well. Here, the authors present CRAPS (Chromosomal-Repair-Assisted Pathway Shuffling), an in vivo pathway engineering technique that enables the self-assembly of one pathway configuration per cell in a one-step transformation. CRAPS leverages the yeast chromosomal repair (CR) pathway and utilizes a pool of inactive, chromosomally integrated orthologous gene variants corresponding to a target multistep pathway. The CRAPS pathway consists of gene-less expression cassettes, each possessing a unique synthetic Cas9 target site. The gene variants possess partial homology to the flanking promoter and terminator elements and are inactive prior to CR repair. Supplying gRNAs to the CRAPS host introduces a double strand break at each expression locus, activating the CR pathway and placing one gene ortholog under transcriptional control of its designated promoter. Thus, one gene variant per pathway step is expressed, resulting in a unique pathway configuration in each cell and thus colony. The authors deployed CRAPS to build more than 1000 theoretical combinations of a four-step carotenoid biosynthesis network. Sampling the CRAPS pathway space yielded strains with distinct colour phenotypes and carotenoid product profiles.

A multiplexed bacterial two-hybrid for rapid characterization of protein–protein interactions and iterative protein design (Boldridge et al. 2023)
The authors developed a method for identifying and designing orthogonal coiled-coil proteins. This strategy involves the development of a next-generation bacterial two-hybrid (NGB2H) method, a significantly modified version of the Bordetella pertussis adenylate cyclase two-hybrid system, which enables multiplex characterization of protein-protein interactions (PPIs). The system uses a barcode approach and allows comprehensive analysis without the need to test all possible pairings manually. Ultimately, large quantities of orthogonal synthetic coiled-coils were designed, created, and tested. The large dataset was used to train a more accurate coiled-coil scoring algorithm (iCipa). This was done iteratively, increasing the size of the synthetically designed libraries from 256 interactions to more than 18,000 interactions. Based on this, the authors have identified what they believe to be the largest set of orthogonal coiled-coils to date, with fifteen on-target interactions. The approach offers a powerful tool for designing orthogonal PPIs.

Illuminating protein space with a programmable generative model (Ingraham et al. 2023)
Tthe authors present Chroma, a generative model for proteins and protein complexes, capable of directly generating new protein structures and sequences. Chroma is programmable and can create proteins with a variety of user-defined properties, such as the distance and contact between amino acid residues, specific protein regions and structures, as well as other features defined by classifiers. Chroma can generate proteins with arbitrary and complex shapes and has even begun to demonstrate the ability to accept descriptions of desired properties as free text. It is capable of generating extremely large proteins and protein complexes (with more than 3,000 amino acid residues) on a commodity graphics processor in a few minutes.
The authors selected and experimentally tested some of the proteins generated by Chroma. The experimental validation of 310 proteins showed that Chroma has learned a sufficiently accurate distribution, so that the designed proteins are expressed, fold correctly, have favourable biophysical properties, and often match the intended structures. Additionally, the crystal structures of two designed proteins were determined, showing an atomic-level match with the Chroma samples.

4-bit adhesion logic enables universal multicellular interface patterning (Kim et al. 2022) the authors have created multicellular interface patterns with swarming bacteria that harbour a cell-cell adhesion logic. They engineered Escherichia coli colonies that express synthetic adhesins derived from nanobodies and their complementary antigens. The bacterial colonies were then seeded a few millimetres apart on soft agar and allowed to grow. The authors observed the formation of interfaces where matching nanobody and antigen met and developed a model of bacterial swarming. The model calculations such as influencing geometric interface properties by tuning seeding conditions were confirmed experimentally. By combining four adhesins (four bits) in two adhesion pairs, any arbitrary tessellation and straight-interface pattern in 2D could be created with the aid of an algorithm. This bacterial adhesion logic could be used, for example, to create human-readable molecular diagnostic displays.

The automated Galaxy-SynBioCAD pipeline for synthetic biology design and engineering (Herisson et al. 2022) the authors introduce the open access Galaxy-SynBioCAD portal, a set of tools for synthetic biology, metabolic engineering and industrial biotechnology. The tools and workflows currently shared on the Galaxy-SynBioCAD portal enable the users to design different metabolic routes and build libraries of strains for the production of a compound of interest. To predict if a given pathway is a valid one or not, a pathway scoring is performed via machine learning using a training set of pathways extracted from literature and pathways validated by pathway engineering experts. The portal is a growing community effort where developers can add new tools and users can evaluate the tools´ performing design for their specific projects. As proof of concept a library of 88 E. coli lycopene-producing strains was designed and engineered. Within this library three enzymes from Pantoea ananas (CrtE, CrtB and CrtI) were assembled in varying gene order in an operon with six different ribosome-binding-sites and two different promoters.

basicsynbio and the BASIC SEVA collection: software and vectors for an established DNA assembly method (Haines et al. 2022) the authors present basicsynbio, an open-source software, that allows users to easily and robustly design a large repertoire of assemblies, enabling applications in synthetic biology and the life sciences. basicsynbio is based on a standardized DNA assembly method developed in 2015 by Storch et al. called Biopart Assembly Standard for Idempotent Cloning (BASIC) DNA assembly. This method utilizes modular parts and linkers as functional units and can assemble up to 14 of these units per round with > 90% accuracy. The method was successfully applied to several areas of research including combinatorial pathway engineering, synthetic operon and small non-coding RNA circuit design, combinatorial guide RNA expression for gene editing, ribosome binding site tuning and fusion protein engineering. The basicsynbio design software is accessible as web app or as python package. Users can access commonly used parts and linkers, robustly design new parts, linkers, and assemblies while exporting sequence data. The exported data can easily be parsed into custom workflows, enabling the automation of BASIC DNA assembly on further liquid-handling platforms and the generation of instructions for manual workflows.
To demonstrate the functionality of basicsynbio, 30 vectors were assembled using modules from the Standard European Vector Architecture (SEVA) database. Each vector contains a specific combination of antibiotic resistance marker and origin of replication.

Cross-kingdom expression of synthetic genetic elements promotes discovery of metabolites in the human microbiome (Patel et al. 2022) the authors developed a computational and experimental strategy for the redesign, expression, mobilization and characterization of multigene biological pathways in prokaryotes and eukaryotes. Studying biosynthetic gene clusters (BGCs) in their native context has been hampered by the inability to grow microorganisms outside of their native environments. When cultivation is possible, many BGCs are silenced under laboratory conditions. The new strategy allows for the decoupling of BGCs from native levels of regulation, thereby enabling the discovery of a variety of natural products and biosynthetic pathways.
As a first step in this approach yet uncharacterized BGCs are computationally redesigned into synthetic genetic elements (SGEs) and functionalized for expression across diverse hosts. The computer-aided design of SGEs includes the redesign of each coding sequence by implementing N-terminal codon bias, which creates versatile 5´ hybrid untranslated regions (UTRs), and screening to avoid internal start and termination signals. Furthermore, hybrid eukaryotic and prokaryotic regulatory elements are integrated for cross-kingdom expression. This includes the design of a library of synthetic yeast promoters and an inducible T7 RNA polymerase expression circuit. Finally, chromosomal-integrated landing pads for SGE mobilization were developed for the stable transfer of SGEs across diverse hosts.
To functionally apply this system for natural product discovery the authors targeted an uncharacterized BGC that had previously been computationally predicted from the genome of Lactobacillus iners. A new class of nucleotide metabolites, called tyrocitabines, was discovered.

Three-dimensional structure-guided evolution of a ribosome with tethered subunits (Kim et al. 2022) the authors present a directed evolution approach that they used to improve previously designed tethered ribosome systems (Ribo-T) (Carlson et al., 2019). In these tethered ribosome systems, the 16S and 23S rRNAs are joined together to form a single chimeric molecule. The single-subunit ribosomes are able to incorporate unnatural amino acids and synthesize protein sequences that are inaccessible to the natural ribosome. So far the potential of Ribo-T systems was limited by their low activity. Directed evolution is a possible way to improve Ribo-T. However, the complexity of macromolecular machines such as the ribosome complicates genetic manipulation. Macromolecular machines have complex tertiary structures with functionally important residues that are far apart in primary sequence but proximal in three-dimensional space. The authors present Evolink (evolution and linkage), a next-generation library construction approach, which brings together separated regions of interest for a continuous next generation sequencing (NGS) read in a three-step process (PCR, ligation and a second PCR) by using basic and low-cost molecular biology methods. In this work two regions of interest within Ribo-T version 2 (Ribo-T v2) were targeted by directed evolution using a degenerate library of 5-15 nt. Afterwards the regions were connected to enable NGS readout. Central to the efforts was the iterative application of design-build-test-analyse (DBTA) cycles, where multiple libraries can be tested, each library building on results and analysis of previous ones. Finally, Ribo-T v3 was developed with a 58 % increased activity in orthogonal protein translation and a 97 % improvement in doubling times in E. coli SQ171 cells compared to the previously developed tethered ribosome (Ribo-T v2).
In general, the Evolink approach may enable enhanced engineering of macromolecular machines for new and improved functions for synthetic biology.

A versatile active learning workflow for optimization of genetic and metabolic networks (Pandi et al. 2022) the authors present METIS (Machine-learning guided Experimental Trials for Improvement of Systems), a modular and versatile active machine learning workflow for data-driven optimization of biological targets with minimal experiments. METIS was developed to overcome the gap between machine learning algorithms and their application for biological systems. Therefore, METIS was created for experimentalists with no experience in programming. METIS runs on Google Colab, a free online platform to write and execute Python codes.
The versatility of METIS was proven on various biological systems. For instance, by identifying a major bottleneck the activity of a recently reported cell-free in vitro gene circuit (LacI-based multi-level controller; Greco et al., 2021) was improved by two orders of magnitude. Also an in vitro transcription-translation system of Escherichia coli (Borkowski et al., 2020) was optimized showing a ten-fold increase in protein production. Finally, METIS was used to improve a complex metabolic network, the so-called crotonyl-CoA/ethylmalonyl-CoA/hydroxybutyryl-CoA (CETCH) cycle (Schwander et al., 2016), a CO₂ fixation cycle comprising 17 different enzymes plus 10 cofactors, which is more efficient than natural photosynthesis. For the CETCH cycle 1025 different conditions were explored and 1000 experimental conditions performed. This resulted in a ten-fold improved productivity, representing the most efficient CO₂-fixing in vitro system described to date.
Overall, METIS has the ability for the optimization of various complex biological networks with minimal experimental efforts, providing multiple opportunities for the study and engineering of different biological systems in the future.

Machine learning-aided engineering of hydrolases for PET depolymerization (Lu et al. 2022) the authors engineered a robust and active polyethylene terephthalate (PET) hydrolase (PETase), an enzyme capable of depolymerization of PET into monomers that can be used in a closed-loop PET recycling process to resynthesize PET. The enzyme was called FAST-PETase (functional, active, stable and tolerant PETase) and designed using a structure-based, machine learning algorithm. FAST-PETase contains five mutations (N233K, R224Q, S121E, D186H and R280A) compared to Ideonella sakaiensis wild-type PETase and shows superior PET-hydrolytic activity between 30 and 50 °C and a range of pH levels. The new enzyme could provide a possible method for the industrial enzymatic degradation of plastic waste.

Division and regrowth of phase-separated giant unilamellar vesicles (Dreher et al. 2021) the authors established a fully controlled division mechanism for giant unilamellar vesicles (GUV). The authors used phase separation to construct GUVs with a lipid disordered and a lipid ordered phase. When the osmolarity in the surrounding medium was increased, the GUV´s volume decreased by water efflux and two smaller vesicles were formed. The technique was also used to construct light-inducible vesicle division. To this end, bis-(5-carboxymethoxy-2-nitrobenzyl)-ether (CMNB)-caged fluorescein was added to the GUV containing medium. Upon illumination this initially non-fluorescent compound is split up into three components, so that osmolarity is increased and the GUVs divide. The authors also studied vesicle growth through fusion of lipid disordered and lipid ordered vesicles and showed a fusion with a rather low frequency. The fusion frequency can be increased by using CaCl2 or zipper-like DNA-based mimics of SNARE proteins that bring the vesicles into close proximity to each other and drive fusion. Taken together, this method enables a controllable GUV-division into two second-generation compartments and represents another step towards synthetic cell division.

Manufacture of multi-layered artificial cell membranes through sequential bilayer deposition on emulsion templates (Ip et al. 2021) the authors developed an emulsion-based method to generate multi-layered giant vesicles. To generate these multi-layered vesicles, water-in-oil droplets were used as a template and sequentially surrounded with lipid monolayers. The innermost monolayer droplet was formed by incubating aqueous emulsion droplets in an oil solution. Additional monolayers were added by driving the initial droplet in a centrifuge through several monolayer-stabilized oil/water interfaces with alternating hydrophobic/hydrophilic outside orientation. The generated multi(bi)layer vesicles ranged from 0.5 – 50 µm in radius and were mechanically stable. Successful layering was controlled by adding different fluorescent lipids (Rh-PE, NBD-PE, Cy-5-PE) to each double layer. In addition, sodium dithionite, a membrane impermeable NBD quencher, was added to the vesicle medium. Due to lack in quenching the authors could demonstrate that the inner membranes, containing NBD-PE, are protected by the additionally applied outer membranes, thus verifying a successful and stable layering. The membrane mechanics, like bending rigidity, were tested by thermal fluctuation of the vesicles, detected via phase-contrast microscopy. As a result, the increased number of layers led to an increase in bending rigidity, confirming that the multilayers are mechanically coupled. This method may offer a potential solution for incorporating a broad range of functional modules into synthetic vesicles.

Li et al. (2020) established a scalable magnetic assembly of cell-mimicking giant unilamellar vesicle (GUVs) colonies. In order to develop a magnetic field-based colony formation of GUVs, a stainless steel-mesh with a microwell pattern in a paramagnetic medium was loaded with GUVs and placed between two face-to-face magnets. The GUVs were magnetically gathered to form colonies and furthermore exhibited a higher stability against osmotic imbalances compared to suspensions with individual GUVs. To mimic the structural complexity of biological tissue with different cell types, two types of GUVs with different compositions were generated and magnetically manipulated. Depending on the direction of the magnetic field, the different GUVs could be sorted separately in different directions, forming layers or a serial distribution within the microwell. In a proof of principle experiment intended to mimic a spatialized biochemical reaction among organelles, two different GUVs were generated and magnetically separated. GUVs in colony 1 possessed pores and were formed in the presence of glucose oxidase, while GUVs in colony 2 were formed together with horseradish peroxidase. When glucose was added to the medium, it entered GUVs of colony 1 through the pores, H₂O₂ was produced and diffused into the GUVs of colony 2, where the horseradish peroxidase catalyzed the production of a red fluorophore. Since the red fluorophore could only be detected in colony 2, the system’s ability to rudimentarily mimic the capacity of natural tissue in compartmentalization and spatialization of biochemical reactions was shown. This work could enable the creation and study of tissue like structures and represents an important step towards creating synthetic tissues. - to the original literature

Hamashima et al. (2019) extended a sequencing method based on the work of Kimoto et al. (2013) for precise localisation of non-natural base D_s: (7-(2-thienyl)-imidazo-[4,5-b]pyridine) in aptamer-candidates from ExSELEX-libraries (genetic alphabet expansion for systematic evolution of ligands by exponential enrichment). Two different replacement PCRs were used for this in order to exchange non-natural bases with natural bases (N), with the goal of studying the generated products by means of deep sequencing. To be able to draw conclusions on the actual position of Ds, an encyclopedia was created that contained the frequency and composition of the conversion in a specific sequence context (NNNDsNNN). By comparing the current data with the encyclopedia data the position of the Ds could thus be deduced. - to the original literature

Koch et al. (2019) merged DNA molecules with functional materials to generate stable data storage without loss of information (DNA-of-things storage architecture). For this, the DNA molecules were in a capsule o silica nanoparticles embedded in order to prevent a decomposition of the DNA in the production process of the storage. In a proof of principle experiment, DNA encapsulated in silica (SPED) is embedded in polycaprolactone (PCL), a biologically degradable thermoplastic polyester, and printed in the shape of a rabbit by means of 3D-printing. A part of the rabbit was then randomly removed and the DNA located within it was extracted, amplified and decoded. The file stored in the part of the rabbit was able to be reproduced exactly and served as a template to print the F1-generation. Also, no data loss was observed in the following generations up to F5. Following experiments in which the SPED video data was contained and embedded in the most varied material verified the stability of the SPED. A system has thus been created that could be used e.g. for the storage of sensitive data across human generations. - to the original literature

Lee et al. (2019) - The authors extended the knowledge about flexizymes, which can be used to bind novel chemical monomers to tRNAs. The tRNAs bring the monomers to ribosomes, so that these can be incorporated into peptides as non natural amino acids. Lee et al. for the first time produced defined design rules for the novel monomers, so that these can be bound to tRNAs without an effortful trial-and-error process. A plurality of monomers can thus be assembled into peptides hybrid products through ribosomes. As an application of this reprogramming of the genetic code, the synthesis of novel (biological) materials, e.g. high-performance materials, or novel medicines is envisioned. - to the original literature

Zhang et al. (2019) developed a method to secure the data in DNA-data storage, which decouples from the "readout" of the DNA by means of sequencing. The origami cryptography used in the work is based on the self assembly of biomolecules. For the sender to encrypt the DNA-data storage a long scaffold DNA strand is loaded with message strands that contain a biomolecule such as biotin and that differ, depending on the information to be coded, in number and position on the scaffold. Decoding the message by the recipient takes place by means of hundreds of message-specific scaffoldstrands that fold the scaffold into a specific DNA-based nanostructure. This nanostructure is similar to a QR code of square form, but with a pattern similar to Braille writing, comprising numerous points. The orientation of the square that is important for reading the data is assured through a marker DNA strand. This is provided during the encryption by the sender to the messagestrands and is located at a determined position on the scaffold. The data read-out then takes place microscopically with the addition of streptavidin, whereby a Braille point corresponds to a binary number and, in text messages, represents either a letter or its position. For the first time, QR code-like DNA storage has been created that is based on a biomolecule and is cryptographically flexible. - to the original literature

Zhang et al. (2019) - In the field of Synthetic Biology in recent years a considerable quantity of genetic data has been created that, in turn, could be used for the construction of genetic circuits. These data are accessible to the user through various platforms. Sequence-based data, comprising known natural sequences, can be queried via databases with the Basic Local Alignment Search Tool (BLAST). More complex synthetic circuits and designs are in repositories like SynBioHub and are shared publically. For this, SynBioHub uses large datasets of various platforms such as International Genetically Engineered Machine (iGEM). For standardized data exchange the data are coded in the Synthetic Biology Open Language (SBOL). Because the data sets of the various platforms that SynBioHub feeds differ qualitatively and, in part, are deposited translated unstructured in SBOL, search questions and genetic designs are often imprecise. For this reason Zhang et al. implemented organisational technology to sort information from the world wide web in SynBioHub. The goal of the work was to make the search results for search questions on usable circuits and their individual components more precise. Methods were combined with one another for this, establishing the priority of the popularity of the circuit elements. (PageRank) and simplifying the weighting of arising duplicates. (clustering with data infrastructure). The program developed in this work, SBOLExplorer, is implemented as backend service in SynBioHub. The user can therefore compose the circuit graphically as usual via the application SBOL-Designer in SynBioHub and automatically obtains the sorted information for a search question. - to the original literature

Bourgeois et al. (2018) developed a CRISPR-Cas9-integration system that uses a series of synthetic DNA landing pads to integrate a number of copies of a DNA into the genome of S. cerevisiae. The pads comprise a specific gRNA-sequences with flanking recombinant regions and function as a zone for multi-copies gene transfer. Varying pads can thereby be present once, twice, three or four times in the yeast genome. It is thereby possible to integrate up to four copies of a gene precisely into the genome of the yeast in only one reaction. In a proof of principle experiment an enzyme (norcoclaurine synthase) with low catalytic activity was selected and was used for the synthesis of benzylisoquinoline alkaloids. By integrating the four copies of the enzyme gene into the landing pads the production of the intermediate molecule ((S)-norcoclaurine) for the synthesis of alkaloids could be doubled. - to the original literature

Atomically accurate de novo design of antibodies with RFdiffusion (Bennett et al. 2026) and De novo designed proteins neutralize lethal snake venom toxins (Vázquez Torres et al. 2025)

The authors of both publications employed customized versions of RoseTTAFold Diffusion (RFdiffusion; Watson et al., 2023), a generative AI model for protein design, to develop novel therapeutic binding proteins targeting medically relevant structures.

In the first study, RFdiffusion was used to design de novo antibody fragments, including VHHs (variable domains of heavy-chain-only antibodies) and scFvs (single-chain variable fragments), against user-defined epitopes. The approach was applied to a range of viral, bacterial, and tumor-associated targets, including targets from SARS-CoV-2, Influenza A, RSV, Clostridioides difficile toxin B, and a peptide–MHC complex. Following experimental validation and optimization of binding affinity, several highly specific antibodies with low-nanomolar binding affinities were obtained. Cryo-electron microscopy analyses confirmed a high degree of agreement between the computationally predicted and experimentally determined structures.

In the second study, the approach was extended to the design of entirely novel protein-based antitoxins targeting snake venom toxins. The focus was on neurotoxins and cytotoxins from the three-finger toxin (3FTx) family, a group of protein toxins responsible for severe envenomation. The designed antitoxins exhibited high stability, nanomolar binding affinities, and near-atomic agreement with the computational design models. In cell-based assays, they effectively neutralized the toxic activity of the target toxins. In mouse models, the neurotoxin-binding antitoxins in particular provided effective protection against otherwise lethal envenomation.

Together, these studies demonstrate that generative AI-based methods can enable the targeted development of antibodies and novel protein therapeutics. As a result, development times and costs could potentially be reduced compared with conventional approaches such as animal immunization or extensive screening campaigns.

Atom-level machine learning of protein-glycan interactions and cross-chiral recognition in glycobiology (Carpenter et al. 2025) 

The authors developed MCNet, a machine learning model for predicting protein–glycan interactions that is based on atom-level structural representations and can predict protein–glycan binding with high accuracy. In contrast to conventional approaches that represent glycans through monosaccharide building blocks, MCNet employs a complete atom-level representation including stereochemical information. As a result, the model can not only accurately predict known glycan–protein interactions but also generate predictions for previously unobserved structures, such as rare or synthetic enantiomeric (mirror-image) glycans.

The model integrates various existing experimental datasets by converting the results obtained from different measurement methods onto a common scale, thereby making them comparable.

Experimental validation using glycan arrays and lectin arrays confirmed numerous predictions, including unexpected interactions between L-glucose and classical glycan-binding proteins, such as fucose-binding lectins. Overall, the study demonstrates that atom-level machine learning models can provide new insights into the role of chirality and molecular recognition in glycobiology while extending the predictive capabilities of existing approaches.

Fabrication of cytotoxic mirror image nanopores (Firzan et al. 2025)

The authors developed the first fully functional mirror-image nanopores composed of D-amino acids (DpPorA), based on a natural bacterial porin (PorACj) from Corynebacterium jeikeium as the template structure. These pores form stable membrane channels and are structural and functional mirror images of their L-counterparts. By replacing one aspartate (D) and one glutamate (E) residue with alanine, the authors generated the DpPorA DE variant, which exhibited enhanced conductance and anion selectivity. The pores enabled single-molecule detection of various biomolecules, including peptides and alpha-synuclein. Experiments in artificial membranes and molecular dynamics simulations confirmed their structural stability and transport function. In artificial vesicles, the peptides formed large, flexible pores that allowed the transport of molecules up to approximately 3 kDa. 

In addition to the studies conducted in artificial vesicles, cell culture experiments were also performed. DpPorA DE exhibited a selective cytotoxic effect upon treatment of breast cancer cells, whereas no significant effect was observed in normal mammary epithelial cells. This effect is attributed to the more negatively charged membrane of breast cancer cells, which promotes the binding of the cationic DpPorA DE peptides. Following membrane binding, the peptides can insert into the cell membrane and form pore-like structures, leading to membrane destabilization and reduced cell viability.

The study highlights the potential of mirror-image nanopores for biosensing applications and the development of novel cancer therapeutics.

Synthetic organizer cells guide development via spatial and biochemical instructions (Yamada et al. 2025)

The differentiation of pluripotent stem cells in vitro lacks morphogenesis, i.e. the formation of three-dimensional structures, and pattern formation, as well as the spatial organization of distinct cell types, as observed during native embryonic development, where these processes are guided by a natural signaling system composed of morphogen gradients. A native-like development can be guided by extra-embryonic cells that self-organize around progenitor cells. The authors programmed such a synthetic organizer cell line from fibroblasts that self-assemble around mouse embryonic stem cells (mESCs) in a spatially defined manner. 

They created a GFP-expressing mESC line whose surface GFP could then be recognized by a set of recently developed synthetic cell adhesion molecules (synCAMs). It was shown that organizer cells could form nodes or shells on/around the mESCs depending on the expressed synCAM. The organizer cells were then engineered to express the morphogen WNT3A (Wingless-related integration site 3) or its antagonist DKK1 (Dickkopf-1). 

Furthermore, small molecule-inducible promoters allowed dynamic control, and suicide switches were programmed to stop organizer signaling when necessary. WNT3A plays a role in symmetry breaking during mammalian gastrulation and it could be shown that providing the morphogen from an asymmetric node induced symmetry breaking and created embryoid elongation. When different organizers were used in combination (for example, a WNT3A node and a DKK1 shell) the WNT3A morphogen gradient in the embryoid was generated and different cell lineages within the same embryoid were created. The synthetic organizers can be used to study developmental processes and eventually be applied to generate tissues or organs for clinical applications. 

Construction of a Mirror-Image RNA Nanostructure for Enhanced Biostability and Drug Delivery Efficiency (Zhang et al. 2025)

The authors developed a novel nanostructure based on mirror-image RNA (L-RNA) designed for stable and targeted drug delivery. The central element is the L-RNA three-way junction (L-3WJ), composed of three L-RNA oligonucleotides that self-assemble into a stable, three-branched structure via complementary base pairing. Due to its L-configuration, this platform is highly resistant to enzymatic degradation and therefore significantly more stable than natural D-RNA.

Various functional components are modularly attached to this RNA scaffold. An siRNA targeting the anti-apoptotic gene MCL1 is incorporated through specific base-pair hybridization to defined sequences within the L-3WJ structure, enabling targeted gene silencing after uptake in target cells. In addition, an L-DNA double helix is attached as a carrier module for the chemotherapeutic agent doxorubicin (DOX). DOX is loaded via intercalation into the L-DNA. For targeted cellular uptake into tumor cells, the nanoparticle is further functionalized with folic acid to enable receptor-mediated uptake into folate receptor-positive tumor cells.

In cell experiments, nanoparticles functionalized with anti-MCL1 siRNA, DOX and folic acid showed selective uptake in folate receptor-positive breast cancer cells (MDA-MB-468), whereas no uptake was observed in folate receptor-negative cells (HEK-293T). At the same time, siRNA-mediated inhibition of MCL1 resulted in a marked reduction of MCL1 protein expression. The combined administration of siRNA and DOX produced a synergistic antiproliferative effect, leading to stronger inhibition of tumor growth compared to the individual components. Comparative studies further demonstrated reduced nonspecific toxicity of the L-RNA-based nanoparticles compared to D-RNA-based systems.

While certain aspects of intracellular drug release and processing still require further investigation, the results overall indicate strong functional performance of the platform. Overall, the study demonstrates that L-RNA-based nanostructures represent a promising platform for a stable and flexible system for the delivery of therapeutics, which can be applied in cancer therapies, among other areas. 

Incompatibility in cell adhesion constitutes a barrier to interspecies chimerism (Ballard et al. 2024)

Blastocyst complementation is a possibility to address the shortage of donor organs for transplant-seeking persons. For this purpose, pluripotent stem cells (PSCs) are used to complement an organogenesis-disabled animal host-embryo with the objective of producing donor organs within the host. The technique has its limitations when applied across distantly related species such as mice and humans due to xenogeneic barriers. One of these barriers seems to be cell adhesion incompatibility. During development, epiblast cells are held together by strong anchoring junctions and the authors hypothesized that mismatched cell adhesion molecules might hinder the integration of human PSCs. To improve cell adhesion, they used nanobodies, which are single-domain fragments derived from camelid antibodies to achieve interspecies adhesion in vitro. Human PSCs were engineered to express a nanobody against GFP on their cell surface and were shown to bind mouse embryonic stem cells. In vivo, the nanobody-expressing human induced PSCs were injected into GFP-expressing mouse blastocytes and transferred into surrogate mothers. The resulting embryos were shown to contain more chimeric cells than control embryos. This technique could be used to advance interspecies organogenesis.

An electrogenetic interface to program mammalian gene expression by direct current (Huang et al. 2023)

The authors developed an interface to link wearable electronic devices to medical interventions. The interface, called direct current-actuated regulation technology (DART), enables programming of transgene expression in human cells. Via direct current non-toxic levels of reactive oxygen species are generated that interact with a biosensor, a protein called KEAP1. KEAP1 releases the transcription factor NRF2 that translocates into the cell nucleus and binds to a synthetic promoter, resulting in NRF2-mediated reversible expression of a gene of interest. The system was tested in vivo. A battery pack providing direct current, wired via a manual ON/OFF power switch to two customized acupuncture needles located at an implantation site on the back of type 1 diabetic mice was used. The implantation site contained subcutaneously microencapsulated engineered human cells that express native KEAP1 and NRF2 and NRF2-mediated expression of the insulin gene. In type 1 diabetic mice, DART decreased high blood sugar levels by stimulating insulin expression. The link of analogue biological systems with digital electronic devices facilitated by DART promises versatile applications for gene- and cell-based therapies, real-time dosing, and remote monitoring of patients through medical personnel.

Leveraging DNA-Encoded Cell-Mimics for Environment-Adaptive Transmembrane Channel Release-Induced Cell Death (Li et al. 2024)

The authors established an artificial DNA-encoded T cell mimic model (ARTC). It imitates the function of cytotoxic T lymphocytes, important players in the immune systems response to tumors, by releasing the protein perforin, which causes membrane pores disrupting the homeostasis of target cells. ARTC is comprised of DNA-based spherical particles containing a double-strand DNA complex (IG) with three functional regions; the m* region for bridging, the i-motif DNA segment for response to changes in the pH and LG4, a guanine-rich DNA strand. Structural changes in IG as a response to a mildly acidic environment leads to the release of LG4 that exits the DNA sphere and anchors into the target cell membrane, forming a potassium ion channel. The following efflux of K+ disrupts the cell homeostasis, eventually triggering apoptosis of the target cell. Cell mimetics such as this are to be further developed to enable precise and controlled treatments to restore cell function.

A Dueling-Competent Signal-Sensing Module Guides Precise Delivery of Cargo Proteins into Target Cells by Engineered Pseudomonas aeruginosa (Wu et al. 2023) The authors developed a highly selective platform for protein delivery, called DUEC (dueling competent), by utilising the tit-for-tat/dueling response of the H1-T6SS (type VI secretion system) in Pseudomonas aeruginosa. P. aeruginosa can sense and remember the exact location of physical contact with a neighbouring bacterium and mount a retaliatory response, known as sister cell dueling or tit-for-tat, against heterologous attackers. This unique response in P. aeruginosa is mediated by T6SS, which secretes toxic effectors that kill competitors on direct contact. The authors used DUEC cells, which are P. aeruginosa cells lacking the toxic effectors but are capable of producing H1-T6SS and engaging in dueling. The further developed DUEC cells were able to deliver a nuclease as cargo in a discriminatory manner into the cytosol of T6SS+ but not T6SS- Vibrio cholerae cells to selectively kill provocative cells in a mixed community. Cre recombinase was also used as cargo, demonstrating that DUEC cells are not only a prototypical physical contact sensing and delivery platform, but can also be coupled with recombination-based circuits that have the potential for complex tasks in mixed microbial communities.

A DNA turbine powered by a transmembrane potential across a nanopore (Shi et al. 2024)
The authors generated bottom-up designed nanoscale DNA origami turbines that operate autonomously in physical conditions, converting energy from naturally abundant electrochemical potentials into mechanical work. The turbine contains a central axle with three blades arranged in a chiral configuration, either left or right-handed, and has a height of 24 or 27 nm. It can drive a long DNA bundle as a hydrodynamic load by sustained rotary motion of up to 10 revolutions s−1. The rotational direction of the DNA turbines can be controlled by the ionic strength of the buffer that might be due to a change in the electrophoretic anisotropy ratio with salt concentration.

Periplasmic biomineralization for semi-artificial photosynthesis (Lin et al. 2023)
Semiconductor-based biointerfaces are typically established either on the surface of the plasma membrane or within the cytoplasm. Here, the authors have discovered that semiconductor nanocluster precipitates within the periplasmic space of Gram-negative bacteria can efficiently solar-drive chemical production in Escherichia coli. They created nanostructured “exoskeletons” and established semiconductor-based interfaces in the periplasma of E. coli. Specifically, the formation of semiconductor nanoclusters of CdS, one of the most studied optically active materials, was mediated by a H2S-producing nongenetic pathway. Furthermore, they showed that these in situ produced semiconductor nanoclusters could elevate adenosine triphosphate (ATP) levels and enhance malate production under light condition in E. coli. The authors showed that this process of semi-artificial photosynthesis can be applied to the construction of a continuous bioreactor based on living materials for multi-element conversion. The authors suggest that by harnessing the power of biomineralization, and extend it to other bacterial cells, the periplasmic biosynthesis has immense potential for constructing semiconductor-based biohybrids that can be applied in environmental remediation, living bioreactor fabrication, and semi-artificial photosynthesis for bioproduction and various sustainable applications.

Sub-1.4 cm3 capsule for detecting labile inflammatory biomarkers in situ (Inda-Webb et al. 2023)
Transient molecules in the gastrointestinal tract are key signals and mediators of inflammation. Owing to their highly reactive nature and extremely short lifetime in the body these molecules are difficult to detect. The authors developed a miniaturized wireless bio-electronic pill, a device that integrates genetically engineered probiotic biosensors (living bacteria) that respond to inflammation-associated molecules like nitric oxide, hydrogen peroxide, tetrathionate, and thiosulfate by producing luminescence, combined with a custom-designed photodetector and readout chip to remotely track these molecules in the gastrointestinal tract. This is realised by transcription of a recombinase in response to the signal biomarker that in turn leads to expression of a luminescence gene in a quantitative manner. Low-power electronic readout circuits integrated into the device convert the light emitted by the encapsulated bacteria to a wireless signal that can be transmitted to a smartphone. All the components were integrated into a sub-1.4 cm3 capsule, compatible to ingestion. The authors demonstrated in vivo biosensor monitoring in the gastrointestinal tract of mice and pigs. With this device, diseases such as inflammatory bowel disease could be diagnosed earlier as currently possible, and disease progression could be more accurately tracked.

Cell-free biosynthesis combined with deep learning accelerates de novo-development of antimicrobial peptides (Pandi et al. 2023) bioactive peptides are key molecules in health and medicine with deep learning holding a big promise for their discovery and design. Here, the authors established a high throughput and low-cost approach to identify and validate bioactive peptide candidates within less than 24 h. They used deep learning to design thousands of antimicrobial peptides (AMPs) de novo. Using computational methods, they prioritized 500 candidates that they produced and screened with their newly established cell-free protein synthesis (CFPS) pipeline that rapidly and inexpensively produces AMPs directly from DNA templates. The method takes place in a 384-well setting and the new AMPs are directly tested on 20 μl bacterial cell cultures, followed by OD₆₀₀ measurement. They identified 30 functional AMPs, which they characterized further through molecular dynamics simulations, antimicrobial activity and toxicity. Notably, six de novo-AMPs feature broad-spectrum activity against multidrug-resistant pathogens and do not develop bacterial resistance.

A synthetic transcription platform for programmable gene expression in mammalian cells (Chen et al. 2022) the authors built a synthetic transcription system consisting of CRISPR-based transcriptional activators and guideRNAs (gRNAs). The parts are interchangeable and have been tested both in transiently transfected cells and as stable chromosomally integrated constructs in mammalian cells. Transcription of a gene is controlled by the presence of a gRNA and 2 to 16 complementary binding sites inserted as operators upstream of the gene. A deactivated CRISPR protein fused to an activation domain such as VP16 initiates the transcription. By using different gRNAs, CRISPR transcriptional activators and copy numbers of the gRNA binding sequence as well as through inserting genetic elements to enhance gene expression such as synthetic introns the authors developed promoters with a >1000-fold range of gene expression. The system has been used for the production of a human antibody in CHO cells, in which the genes for the heavy and light chain were each placed under the control of gRNA binding sequences. This approach yielded a high antibody expression and could be applied to produce industrially relevant antibody titres in a controllable and scalable fashion.

Independent control of amplitude and period in a synthetic oscillator with modified repressilator (Zhang et al. 2022) the authors created an oscillator circuit that can be controlled by two inducers in Escherichia coli, one regulating the period of expression, the other the amplitude. They also developed a model for the oscillator that forecasts its regulation. To accomplish the independent control of amplitude and period the circuit was constructed from two plasmids. One contains the repressilator elements and the other the amplitude regulation system. Cells containing the system were placed into a microfluidic system allowing to test eight different concentrations of the inducers at once. Different amplitudes and periods were shown as a function of the relevant inducer concentration. This repressilator could be used, for example, for pulsatile drug delivery.