In a mouse model of Ras-driven liver cancer, systemic treatment with the circuits (lower row) reduced tumor burden compared to untreated controls (upper row). Much more to explore.

We compared circuits (right) with Ras inhibitors (RMC-7977 and Sotorasib) by targeting mutant cells (magenta) co-cultured with wild-type Ras cells (green). Circuits were potent and selective.

We tried many, many, many circuit designs, eventually generating ones that sensitively discriminate Ras mutant cancer cells. We also generated protease-activated proteases to amplify weak oncogenic signals.

In this paradigm, engineered proteins are introduced into cancer and healthy cells using LNP-delivered mRNA. They then sense cancer markers (such as mutant Ras) and trigger engineered cell death effectors (caspases or gasdermins) only in cancer cells. (idealized schematic)

Our new paper on “synpoptosis” circuits, led by the amazing @ShiyuXia, is now out: sciencedirect.com/science/ar… (1/3)

Excited to have lineage motifs out in the world. Hoping the concept and methods can be useful in various developmental systems as more lineage data become available. Congratulations to @tranmartink and @AskaryLab on the work, and thanks to Adara Koivula and @Wonderstrucksci for the cover image.

Finally, we performed a long-term (>24 week) Rett syndrome gene therapy experiment on Rett model mice, assaying behavioral Rett symptoms. The untreated mice showed increasing (worsening) symptom scores over time. The unregulated MeCP2 gene therapy (grey line) only modestly reduced the symptoms at earlier times. The strong circuit variant (“4x”) suppressed symptom scores much more strongly and durably (dark blue line).

In the second paper, we integrated circuits into a Rett syndrome gene therapy. We took advantage of a brain-targeting AAV capsid developed by the Gradinaru lab to specifically deliver the circuit-regulated therapeutic gene (MeCP2) to the brain. The circuits successfully reduced ectopic mRNA expression to levels similar to or lower than endogenous, as shown by single molecule analysis in brain.

We systematically engineered a toolbox of miRNA-based circuits. In these systems, cells with more gene copies transcribe more mRNA, but proportionally reduce expression per copy due to the miRNA. The circuits maintain roughly constant expression across two orders of magnitude of gene dosage, while allowing tuning of set-point over a similar range. They also function in multiple cell lines. Thus, one can deliver genes using a “messy” but convenient approach like transient transfection or AAV, and still achieve a desired expression level. We term the circuits DIMMERs (Dosage-Invariant miRNA-Mediated Expression Regulators).

The problem: Almost all gene expression methods are sensitive to unavoidable variation in gene delivery, generating broad distributions of protein expression across individual cells. This variability complicates downstream analysis in experiments. It is also a problem in gene therapy for dosage-sensitive diseases like Rett syndrome, where both too little or too much expression of the therapeutic gene causes disease. What is needed is a compact, protein-free, “plug and play” system that can compensate for gene dosage variability to achieve a defined, but tunable, set-point for transgene expression.

Delivering and expressing a gene in cells is usually a messy (heterogeneous) process. The messiness interferes with research and applications such as gene therapy. In two new papers, we introduce a toolbox of synthetic miRNA-based control circuits that enable more precise, gene dosage-invariant gene expression, and show how they improve a mouse model of Rett syndrome gene therapy: biorxiv.org/content/10.1101/… ,led by @RongrongDu123 and @MichaelJAFlynn biorxiv.org/content/10.1101/… , led by @MichaelJAFlynn and Acacia Mayfield, in collaboration with Viviana Gradinaru.

History in space. Because the method is spatial, reconstructed lineage trees (above) can be superimposed on spatially organized mouse embryonic stem cell colonies, as shown here for a mouse ES cell colony where you can see both the cell states (left) and lineage relationships (right, cf. lineage tree in previous tweet).

Reconstruction. Computational reconstruction is a challenge for all recording systems. Duncan worked with Remco Bouckaert to adapt Bayesian methods from evolutionary phylogenetics. The reconstruction incorporates phenotypic information along with barcode edits to enable accurate estimation of cell lineage trees. It also allows inference of earlier cellular states (colors on tree, for an ES cell colony). The "fuzziness" in the tree reflects the uncertainty of Bayesian reconstruction. These computational approaches should be broadly useful for other recording systems.

Reading. The 4 dinucleotide edit states can be read out through multiple rounds of hybridization and imaging (seqFISH from @LongCai_Lab), and by taking advantage of the “Zombie” T7 in situ amplification approach developed by @AskaryLab in our previous work. This scheme also allows edits to be read out alongside endogenous transcripts so that one can recover lineage history, cell state, and spatial organization, all in the same cells.

Writing. baseMEMOIR introduces a dinucleotide base editing approach that uses CRISPR A-to-G base editors (@davidrliu). Each target site has an “AA” dinucleotide that can be edited to terminal AG, GA, or GG end states. The use of multiple terminal end states is important because it prevents information being effectively erased (by everything being edited to the same end state) after long durations of editing. By genomically integrating a set of uniquely barcoded dinucleotide target sites, we constructed a mouse embryonic stem cell line with 792 bits of recordable, addressable memory, 50 times more information storage than in the previous intMEMOIR approach.

Just posted a preprint on our new spatial genomic recording system, baseMEMOIR. biorxiv.org/content/10.1101/… Motivation: Cells divide, differentiate, and migrate to form exquisitely organized tissues. Reconstructing the dynamic histories of individual cells, including their lineage relationships and ancestral states, is essential for understanding how intrinsic and extrinsic signals generate tissues during development and in regenerative medicine. Engineered genomic recording systems can reconstruct cell lineage histories from endpoint measurements. However, existing methods either require sequencing (disrupting spatial organization) or have been limited in memory size and scalability. To address this need, @ChadlyDuncan, Kirsten Frieda, and others in the lab created a high memory capacity image-readable recording system termed baseMEMOIR.

This work shows that mammalian circuits can shuttle, identifies minimal circuit sufficient for shuttling, and — using hybrid interspecies circuits — suggests that the ability to shuttle is an evolutionarily conserved developmental module.

Remarkably, a 3 protein circuit—Chordin, the BMP-1 (Tolloid) protease, and Twsg (twisted gastrulation)—is sufficient to shuttle BMP ligands, as can be seen by the displaced gradient forming in these movies. Spooky patterning at a distance.

Two decades ago @AvigdorEldar & @BarkaiLab introduced concept of shuttling, in which protein circuits push morphogens away from their source, or smush them into narrower regions. We have now reconstituted shuttling in mammalian cell culture: science.org/doi/10.1126/scia….

Engineered exporters exhibited high specificity for target RNA. Complementary designs used either viral components or protein nanocages (developed by Neil King, Wesley Sundquist, and David Baker.) Both can be effective.

RNA is informant and agent. Controlled RNA export from cells could enable cell-cell delivery of functional and non-destructive tracking of cell dynamics. Our new paper, led by brilliant @FelixHorns, engineers RNA export for both capabilities: authors.elsevier.com/sd/arti…

“Messy” many-to-many protein networks control signaling, transcription, adhesion, etc. Is this a bug or a feature? In this new perspective, we argue for feature, critical for multicellularity. With brilliant @HeidiKlumpe, @jgojalvo, & @yaronantebi. authors.elsevier.com/a/1hHu6…

Hard to believe it’s almost 2 decades since I was served, and later consumed, this molecularly precise model of a gram negative bacterium. The model captures 3 features of the cell: - high protein density, but low molecule number - carbohydrate surface - cheese in periplasm

Computation shows that stable single morphogen patterns can be initiated by transient perturbations, and made robust to noise through additional intracellular feedbacks or by operating on growing tissues.

Our new 2021 Lab t-shirt is finally here, better late than never. A re-do, representing the seven states of the synthetic multistable MultiFate circuit science.org/doi/10.1126/scie…

Special thanks also to @sysbiosig for excellent preview of the papers: doi.org/10.1016/j.cels.2022.… and to @maayanvisuals for the cover image, which represents the tangled wiring of promiscuous protein interaction systems.

In Klumpe et al, we systematically analyzed pairs of BMP signaling to cells w/different BMP receptor combos. Different ligands can act equivalently with respect to one receptor context but quite differently with respect to another... doi.org/10.1016/j.cels.2022.…

In Su et al, “Ligand-receptor promiscuity enables cellular addressing” we show how promiscuous architectures counter-intuitively enable ligand combinations to selectively activate (“address”) specific cell types or cell type combinations: doi.org/10.1016/j.cels.2022.…

Many intercellular communication systems (BMP, Notch, Wnt, FGF, etc) use sets of promiscuously interacting ligand and receptor variants rather than seemingly simpler one-to-one architectures. Why? Our 2 papers grappling w/this question just out in @CellSystemsCP

In ~month-long movies of cell cultures, the paradoxical design maintained control significantly longer than the simpler non-paradoxical design. Thus, the naturally inspired principle of paradoxical regulation can extend mutational robustness in a synthetic system.

Borrowing this natural design principle, we created Paradaux, a syn circuit that secretes auxin and responds to it by both inhibiting growth and inducing cell death. (Bonus: each arm of the circuit can be independently toggled with small molecules.)

In 2017, @omer_karin and @UriAlonWeizmann noticed that cytokines sometimes stimulate both proliferation and death of same cells, and showed that this “paradoxical” design makes pop. control robust to signal sensing mutations system.

Auxin sending+receiving together allow quorum sensing, which is useful by itself, and, when coupled to proliferation, allows pop. control. But…prob #2: circuits of this type are automatically susceptible to “cheater” mutants that lose the ability to receive signals.

1st question: what signal to use? Auxin is ideal: orthogonal to mammalian pathways & often used to degrade engineered target proteins. We identified enzymes from other species that enable aux biosynthesis from Trp in mammalian cells, producing a private comm system.

Brief backstory: In 2004, here at @Caltech, @lingchongyou and @FrancesArnold, described a synthetic circuit that allowed bacteria to limit their own population size: nature.com/articles/nature02…

Multistability is great subject for lab t-shirt: Here, each of the 7 states lounges in its own stable valley. The three TFs (colored red, green, and blue, as in the experiment) combinatorially establish the states. Inspired by Waddington’s 2nd most famous drawing (1957)…

This illustration by @Wonderstrucksci, inspired by Waddington’s second most famous drawing, shows with remarkable directness how MultiFate’s 3 transcription factors, interacting in combinations, produce a multistable epigenetic landscape.

Another excellent pair of talks this Wednesday: @priscaliberali and Chris Walsh!

The new building, from freakishly sterile fantasy to teeming fertile reality (Image: @jeriscience)

Next Hindsight 2020 event: Wednesday March 17 -- will have two fantastic speakers: Ellen Rothenberg and Jan Huisken (@huiskenlab). Should be excellent…

The 2020 lab t-shirt…(admittedly a bit late this time)

Stability of these 7 states can be vividly visualized in time-lapse movies. Here is movie of cells with the 3 TF circuit. Individual cells in each of 7 states grow into colonies retaining initial state. TFs are shown in RGB, so each of the 7 states is a distinct color.

MultiFate scales. Plugging a 3rd transcription factor into the 2 TF system generates 7 states (septastability), with no re-engineering of the existing circuit (because inhibition occurs at the protein level.) All 7 stable for >18d. # of states grows exp w/num of TFs.

MultiFate allows de- and re-stabilization of states. Preinstalled control knobs in TFs allow one to specifically destabilize the ‘double positive’ state by reducing TF half-life, causing bifurcation and irreversible cell transitions, analogous to natural differentiation.

To make MultiFate, we engineered a set of dimer-dependent TFs. The minimal MultiFate system with two of these TFs successfully generates three states, each stable for >18 days, nicely matching model prediction.

MultiFate is a naturally-inspired design in which transcription factors homodimerize to positively autoregulate their own expression, and heterodimerize to cross-inhibit each other. Here’s a simple example, in which 2 transcription factors generate tristability.

The Hindsight 2020 developmental recording seminar series continues next week with exciting talks from Detlev Arendt (@wormduty) and Celina Juliano: register here: allen-lineage.org/portfolio-… — Next Wednesday Feb. 17

Work of many brilliant scientists: @HeidiKlumpe, @ChristinaJSu, @yaronantebi, @murugan_chicago, James Linton, and others.

Function: Ligand combos are message “addressing” systems that can selectively activate different cell type(s) based on receptor expression. Outperforms 1-1 signaling arch. Works b/c ligand-receptor interactions “compute” complex responses.(Theory paper) biorxiv.org/content/10.1101/…