Research ArticleGENETICS

Model-guided design of mammalian genetic programs

See allHide authors and affiliations

Science Advances  19 Feb 2021:
Vol. 7, no. 8, eabe9375
DOI: 10.1126/sciadv.abe9375
  • Fig. 1 Logical evaluation is enabled by transcriptional and posttranslational regulation.

    (A and B) Cartoons depict (A) the genetic components and (B) their arrangement to produce intended functions. In the schematics, circles are protein domains, arrows indicate splicing or regulation, yellow highlighting denotes the inputs, and the red node is the output. (C to J) A panel of logic gates was designed, simulated, and experimentally evaluated. In the electronic diagrams (teal background), lines denote splicing or regulation. Processes that have a modest effect within the dose range examined, and that because of fundamentally analog behavior do not carry out a fully digital function, are denoted by dotted lines. In the mechanistic diagrams (blue background), purple bent arrows are promoters, and black arrows indicate splicing and regulation. Yellow highlighting denotes the components for which dose is varied (in gene copies). Simulation and experimental results are presented in heatmaps that indicate how the two inputs affect reporter output (mKate2 signal in MEPTRs). Color-coding denotes the mean reporter signal from three biological replicates (bar graphs in fig. S1N and histograms in fig. S1O), scaled by the maximum value in each heatmap. Simulations in (C) are from a fit to the data, and subsequent panels (D to J) are predictions. (K) Some of the motifs that were used in gate designs confer sharp transitions in output, e.g., two inhibitors in a cascade produced ultrasensitivity. The downstream inhibitor is tagged with a PEST degron.

  • Fig. 2 Compact multi-output logic is attained through functional modularity.

    (A) A strategy for multi-output logic is proposed by using multitasking proteins that retain the functions of their constituent domains. The cartoons depict the use of multiple DNA-binding domains on a TF to regulate multiple genes, the embedding of a split intein fragment within a functioning TF to enzymatically alter its activity, and the merging of features from multiple genetic programs to enable their compact simultaneous implementation. (B to F) A panel of MIMO gates was designed, simulated, and experimentally evaluated. As an example, (C) is deconstructed to show how separate topologies containing proteins that have some domains in common and are amenable to the appending of additional domains can be compressed. In the plots, color-coding denotes the mean mKate2 and enhanced yellow fluorescent protein (EYFP) reporter signal from three biological replicates (bar graphs in fig. S2F), scaled by the maximum value in each heatmap. (G) These plots summarize the complexity of the gates that were designed and validated in Figs. 1 (red) and 2 (purple), with complexity defined based on the size and depth of the circuits in the electronic diagrams (top) or based on the numbers of genes, regulatory connections, and regulatory proteins used (bottom). The expanded toolkit of genetic parts and model-guided approach were successful for building circuits spanning a range of attributes, which suggests that this design process could be executed reliably for many future objectives.

  • Fig. 3 Analog behaviors are constructed by using TFs that play multiple roles.

    Reconstitutable TFs are conducive to (A to C) ultrasensitivity and (D to I) bandpass concentration filtering. Ultrasensitivity was increased by removing an inhibitor while producing an activator (C). For bandpass concentration filtering, a tight upper threshold was achieved with additional regulation in which moderate FKBP-ZF reconstitutes RaZFa and high FKBP-ZF inhibits the reporter and VP16-FRB (G). Simulations in (A) and (D) are fits to data, and other panels are predictions. Prediction plots indicate how output expression varies with the dose (nanograms of plasmid) of the component highlighted in yellow. Sets of simulated responses are for varying the component highlighted in red-to-blue gradation; each set contextualizes the “slice” that the tested response occupies in higher-dimensional component dose space. Each experiment plot corresponds to the simulation with the dark line. The ZF1/2x6-C promoter has six partially overlapping ZF1 and ZF2 sites. Dimethyl sulfoxide (DMSO) is the vehicle for rapamycin, which is used here as an environmental species (not an input). RaZFa simulations correspond to rapamycin treatment. Each experiment plot has either a no treatment series or both a rapamycin and a DMSO series. The mean and SEM of EYFP reporter signal are from three biological replicates (bar graphs in fig. S3, D to L). Axes are linearly scaled from 0 to 1 ng (dotted lines) and log-scaled above 1 ng. Linearly scaled versions of these plots are in fig. S3N.

  • Fig. 4 Sensors can be linked to genetic programs to make signaling cascades.

    MESA and COMET technologies can be combined to construct functional biosensors, and upstream biosensor output is well-matched to the requirements for downstream promoter input. (A and B) ABA-ZF2a and Rapa-MESA-ZF6a each exhibit ligand-inducible signaling (P = 2 × 10−3 and P = 1 × 10−3, respectively, one-tailed Welch’s unpaired t test). Ethanol is the vehicle for both ligands. For MESA, the TC contains an FRB ectodomain and intracellular COMET TF, and the PC contains an FKBP ectodomain and intracellular tobacco etch virus protease (TEVp). Each receptor chain contains a fibroblast growth factor receptor 4 (FGFR4) transmembrane domain. (C to E) Validated sensors were applied to implement multi-input sensing. AND logic was selected as a design goal, and four synonymous topologies—those that are intended to achieve the same goal through different mechanisms—were proposed and evaluated. For each input type (two columns for upstream ZFa or ligand sensing) and topology (four rows), reporter signal with two inputs differed from that with either or no input [P < 2 × 10−16 in each case, three-factor analysis of variance (ANOVA) and Tukey’s honest significant difference (HSD) test], indicating successful AND gate outcomes. Topologies 2 to 4 displayed negligible background signal (comparable to the signal with only the reporter present, ~101 to 102 MEPTRs; fig. S4H), despite involving multilayer signaling that can be a potential source of leak. The (ZF2/ZF6)x3 promoter has three pairs of alternating ZF2 and ZF6 sites. Bar graphs represent the mean, SEM, and values of mKate2 reporter signal from three biological replicates (depicted as dots; near-zero values are below the log-scaled y-axis lower limit). The numbers above bar pairs are the fold difference, and a fold difference of ∞ indicates that the denominator signal is less than or equal to zero.

Supplementary Materials

Stay Connected to Science Advances

Navigate This Article