Establishing a synthetic orthogonal replication system enables accelerated evolution in E. coli.

The evolution of new function in living organisms is slow and fundamentally limited by their critical mutation rate. Here, we established a stable orthogonal replication system in Escherichia coli. The orthogonal replicon can carry diverse cargos of at least 16.5 kilobases and is not copied by host polymerases but is selectively copied by an orthogonal DNA polymerase (O-DNAP), which does not copy the genome. We designed mutant O-DNAPs that selectively increase the mutation rate of the orthogonal replicon by two to four orders of magnitude. We demonstrate the utility of our system for accelerated continuous evolution by evolving a 150-fold increase in resistance to tigecycline in 12 days. And, starting from a GFP variant, we evolved a 1000-fold increase in cellular fluorescence in 5 days.

Adding α,α-disubstituted and ß-linked monomers to the genetic code of an organism.

The genetic code of living cells has been reprogrammed to enable the site-specific incorporation of hundreds of non-canonical amino acids into proteins, and the encoded synthesis of non-canonical polymers and macrocyclic peptides and depsipeptides1-3. Current methods for engineering orthogonal aminoacyl-tRNA synthetases to acylate new monomers, as required for the expansion and reprogramming of the genetic code, rely on translational readouts and therefore require the monomers to be ribosomal substrates4-6. Orthogonal synthetases cannot be evolved to acylate orthogonal tRNAs with non-canonical monomers (ncMs) that are poor ribosomal substrates, and ribosomes cannot be evolved to polymerize ncMs that cannot be acylated onto orthogonal tRNAs-this co-dependence creates an evolutionary deadlock that has essentially restricted the scope of translation in living cells to α-L-amino acids and closely related hydroxy acids. Here we break this deadlock by developing tRNA display, which enables direct, rapid and scalable selection for orthogonal synthetases that selectively acylate their cognate orthogonal tRNAs with ncMs in Escherichia coli, independent of whether the ncMs are ribosomal substrates. Using tRNA display, we directly select orthogonal synthetases that specifically acylate their cognate orthogonal tRNA with eight non-canonical amino acids and eight ncMs, including several ß-amino acids, α,α-disubstituted-amino acids and ß-hydroxy acids. We build on these advances to demonstrate the genetically encoded, site-specific cellular incorporation of ß-amino acids and α,α-disubstituted amino acids into a protein, and thereby expand the chemical scope of the genetic code to new classes of monomers.

Continuous synthesis of E. coli genome sections and Mb-scale human DNA assembly.

Whole-genome synthesis provides a powerful approach for understanding and expanding organism function1-3. To build large genomes rapidly, scalably and in parallel, we need (1) methods for assembling megabases of DNA from shorter precursors and (2) strategies for rapidly and scalably replacing the genomic DNA of organisms with synthetic DNA. Here we develop bacterial artificial chromosome (BAC) stepwise insertion synthesis (BASIS)-a method for megabase-scale assembly of DNA in Escherichia coli episomes. We used BASIS to assemble 1.1 Mb of human DNA containing numerous exons, introns, repetitive sequences, G-quadruplexes, and long and short interspersed nuclear elements (LINEs and SINEs). BASIS provides a powerful platform for building synthetic genomes for diverse organisms. We also developed continuous genome synthesis (CGS)-a method for continuously replacing sequential 100 kb stretches of the E. coli genome with synthetic DNA; CGS minimizes crossovers1,4 between the synthetic DNA and the genome such that the output for each 100 kb replacement provides, without sequencing, the input for the next 100 kb replacement. Using CGS, we synthesized a 0.5 Mb section of the E. coli genome-a key intermediate in its total synthesis1-from five episomes in 10 days. By parallelizing CGS and combining it with rapid oligonucleotide synthesis and episome assembly5,6, along with rapid methods for compiling a single genome from strains bearing distinct synthetic genome sections1,7,8, we anticipate that it will be possible to synthesize entire E. coli genomes from functional designs in less than 2 months.

Quintuply orthogonal pyrrolysyl-tRNA synthetase/tRNAPyl pairs.

Mutually orthogonal aminoacyl transfer RNA synthetase/transfer RNA pairs provide a foundation for encoding non-canonical amino acids into proteins, and encoded non-canonical polymer and macrocycle synthesis. Here we discover quintuply orthogonal pyrrolysyl-tRNA synthetase (PylRS)/pyrrolysyl-tRNA (tRNAPyl) pairs. We discover empirical sequence identity thresholds for mutual orthogonality and use these for agglomerative clustering of PylRS and tRNAPyl sequences; this defines numerous sequence clusters, spanning five classes of PylRS/tRNAPyl pairs (the existing classes +N, A and B, and newly defined classes C and S). Most of the PylRS clusters belong to classes that were unexplored for orthogonal pair generation. By testing pairs from distinct clusters and classes, and pyrrolysyl-tRNAs with unusual structures, we resolve 80% of the pairwise specificities required to make quintuply orthogonal PylRS/tRNAPyl pairs; we control the remaining specificities by engineering and directed evolution. Overall, we create 924 mutually orthogonal PylRS/tRNAPyl pairs, 1,324 triply orthogonal pairs, 128 quadruply orthogonal pairs and 8 quintuply orthogonal pairs. These advances may provide a key foundation for encoded polymer synthesis.

Astrocytic control of extracellular GABA drives circadian timekeeping in the suprachiasmatic nucleus.

The hypothalamic suprachiasmatic nucleus (SCN) is the master mammalian circadian clock. Its cell-autonomous timing mechanism, a transcriptional/translational feedback loop (TTFL), drives daily peaks of neuronal electrical activity, which in turn control circadian behavior. Intercellular signals, mediated by neuropeptides, synchronize and amplify TTFL and electrical rhythms across the circuit. SCN neurons are GABAergic, but the role of GABA in circuit-level timekeeping is unclear. How can a GABAergic circuit sustain circadian cycles of electrical activity, when such increased neuronal firing should become inhibitory to the network? To explore this paradox, we show that SCN slices expressing the GABA sensor iGABASnFR demonstrate a circadian oscillation of extracellular GABA ([GABA]e) that, counterintuitively, runs in antiphase to neuronal activity, with a prolonged peak in circadian night and a pronounced trough in circadian day. Resolving this unexpected relationship, we found that [GABA]e is regulated by GABA transporters (GATs), with uptake peaking during circadian day, hence the daytime trough and nighttime peak. This uptake is mediated by the astrocytically expressed transporter GAT3 (Slc6a11), expression of which is circadian-regulated, being elevated in daytime. Clearance of [GABA]e in circadian day facilitates neuronal firing and is necessary for circadian release of the neuropeptide vasoactive intestinal peptide, a critical regulator of TTFL and circuit-level rhythmicity. Finally, we show that genetic complementation of the astrocytic TTFL alone, in otherwise clockless SCN, is sufficient to drive [GABA]e rhythms and control network timekeeping. Thus, astrocytic clocks maintain the SCN circadian clockwork by temporally controlling GABAergic inhibition of SCN neurons.

Cell-of-origin-specific proteomics of extracellular vesicles.

The ability to assign cellular origin to low-abundance secreted factors in extracellular vesicles (EVs) would greatly facilitate the analysis of paracrine-mediated signaling. Here, we report a method, named selective isolation of extracellular vesicles (SIEVE), which uses cell type-specific proteome labeling via stochastic orthogonal recoding of translation (SORT) to install bioorthogonal reactive groups into the proteins derived from the cells targeted for labeling. We establish the native purification of intact EVs from a target cell, via a bioorthogonal tetrazine ligation, leading to copurification of the largely unlabeled EV proteome from the same cell. SIEVE enables capture of EV proteins at levels comparable with those obtained by antibody-based methods, which capture all EVs regardless of cellular origin, and at levels 20× higher than direct capture of SORT-labeled proteins. Using proteomic analysis, we analyze nonlabeled cargo proteins of EVs and show that the enhanced sensitivity of SIEVE allows for unbiased and comprehensive analysis of EV proteins from subpopulations of cells as well as for cell-specific EV proteomics in complex coculture systems. SIEVE can be applied with high efficiency in a diverse range of existing model systems for cell-cell communication and has direct applications for cell-of-origin EV analysis and for protein biomarker discovery.

Genetically programmed cell-based synthesis of non-natural peptide and depsipeptide macrocycles.

The direct genetically encoded cell-based synthesis of non-natural peptide and depsipeptide macrocycles is an outstanding challenge. Here we programme the encoded synthesis of 25 diverse non-natural macrocyclic peptides, each containing two non-canonical amino acids, in Syn61Δ3-derived cells; these cells contain a synthetic Escherichia coli genome in which the annotated occurrences of two sense codons and a stop codon, and the cognate transfer RNAs (tRNAs) and release factor that normally decode these codons, have been removed. We further demonstrate that pyrrolysyl-tRNA synthetase/tRNA pairs from distinct classes can be engineered to direct the co-translational incorporation of diverse alpha hydroxy acids, with both aliphatic and aromatic side chains. We define 49 engineered mutually orthogonal pairs that recognize distinct non-canonical amino acids or alpha hydroxy acids and decode distinct codons. Finally, we combine our advances to programme Syn61Δ3-derived cells for the encoded synthesis of 12 diverse non-natural depsipeptide macrocycles, which contain two non-canonical side chains and either one or two ester bonds.

Refactored genetic codes enable bidirectional genetic isolation.

The near-universal genetic code defines the correspondence between codons in genes and amino acids in proteins. We refactored the structure of the genetic code in Escherichia coli and created orthogonal genetic codes that restrict the escape of synthetic genetic information into natural life. We developed orthogonal and mutually orthogonal horizontal gene transfer systems, which permit the transfer of genetic information between organisms that use the same genetic code but restrict the transfer of genetic information between organisms that use different genetic codes. Moreover, we showed that locking refactored codes into synthetic organisms completely blocks invasion by mobile genetic elements, including viruses, which carry their own translation factors and successfully invade organisms with canonical and compressed genetic codes.

Bacterial divisome protein FtsA forms curved antiparallel double filaments when binding to FtsN.

During bacterial cell division, filaments of tubulin-like FtsZ form the Z-ring, which is the cytoplasmic scaffold for divisome assembly. In Escherichia coli, the actin homologue FtsA anchors the Z-ring to the membrane and recruits divisome components, including bitopic FtsN. FtsN regulates the periplasmic peptidoglycan synthase FtsWI. To characterize how FtsA regulates FtsN, we applied electron microscopy to show that E. coli FtsA forms antiparallel double filaments on lipid monolayers when bound to the cytoplasmic tail of FtsN. Using X-ray crystallography, we demonstrate that Vibrio maritimus FtsA crystallizes as an equivalent double filament. We identified an FtsA-FtsN interaction site in the IA-IC interdomain cleft of FtsA using X-ray crystallography and confirmed that FtsA forms double filaments in vivo by site-specific cysteine cross-linking. FtsA-FtsN double filaments reconstituted in or on liposomes prefer negative Gaussian curvature, like those of MreB, the actin-like protein of the elongasome. We propose that curved antiparallel FtsA double filaments together with treadmilling FtsZ filaments organize septal peptidoglycan synthesis in the division plane.

Cryptochrome 1 as a state variable of the circadian clockwork of the suprachiasmatic nucleus: Evidence from translational switching.

The suprachiasmatic nucleus (SCN) of the hypothalamus is the principal clock driving circadian rhythms of physiology and behavior that adapt mammals to environmental cycles. Disruption of SCN-dependent rhythms compromises health, and so understanding SCN time keeping will inform management of diseases associated with modern lifestyles. SCN time keeping is a self-sustaining transcriptional/translational delayed feedback loop (TTFL), whereby negative regulators inhibit their own transcription. Formally, the SCN clock is viewed as a limit-cycle oscillator, the simplest being a trajectory of successive phases that progresses through two-dimensional space defined by two state variables mapped along their respective axes. The TTFL motif is readily compatible with limit-cycle models, and in Neurospora and Drosophila the negative regulators Frequency (FRQ) and Period (Per) have been identified as state variables of their respective TTFLs. The identity of state variables of the SCN oscillator is, however, less clear. Experimental identification of state variables requires reversible and temporally specific control over their abundance. Translational switching (ts) provides this, the expression of a protein of interest relying on the provision of a noncanonical amino acid. We show that the negative regulator Cryptochrome 1 (CRY1) fulfills criteria defining a state variable: ts-CRY1 dose-dependently and reversibly suppresses the baseline, amplitude, and period of SCN rhythms, and its acute withdrawal releases the TTFL to oscillate from a defined phase. Its effect also depends on its temporal pattern of expression, although constitutive ts-CRY1 sustained (albeit less stable) oscillations. We conclude that CRY1 has properties of a state variable, but may operate among several state variables within a multidimensional limit cycle.

Discovery and Genetic Code Expansion of a Polyethylene Terephthalate (PET) Hydrolase from the Human Saliva Metagenome for the Degradation and Bio-Functionalization of PET.

We report a bioinformatic workflow and subsequent discovery of a new polyethylene terephthalate (PET) hydrolase, which we named MG8, from the human saliva metagenome. MG8 has robust PET plastic degradation activities under different temperature and salinity conditions, outperforming several naturally occurring and engineered hydrolases in degrading PET. Moreover, we genetically encoded 2,3-diaminopropionic acid (DAP) in place of the catalytic serine residue of MG8, thereby converting a PET hydrolase into a covalent binder for bio-functionalization of PET. We show that MG8(DAP), in conjunction with a split green fluorescent protein system, can be used to attach protein cargos to PET as well as other polyester plastics. The discovery of a highly active PET hydrolase from the human metagenome-currently an underexplored resource for industrial enzyme discovery-as well as the repurposing of such an enzyme into a plastic functionalization tool, should facilitate ongoing efforts to degrade and maximize reusability of PET.

RBBP6 activates the pre-mRNA 3' end processing machinery in humans.

3' end processing of most human mRNAs is carried out by the cleavage and polyadenylation specificity factor (CPSF; CPF in yeast). Endonucleolytic cleavage of the nascent pre-mRNA defines the 3' end of the mature transcript, which is important for mRNA localization, translation, and stability. Cleavage must therefore be tightly regulated. Here, we reconstituted specific and efficient 3' endonuclease activity of human CPSF with purified proteins. This required the seven-subunit CPSF as well as three additional protein factors: cleavage stimulatory factor (CStF), cleavage factor IIm (CFIIm), and, importantly, the multidomain protein RBBP6. Unlike its yeast homolog Mpe1, which is a stable subunit of CPF, RBBP6 does not copurify with CPSF and is recruited in an RNA-dependent manner. Sequence and mutational analyses suggest that RBBP6 interacts with the WDR33 and CPSF73 subunits of CPSF. Thus, it is likely that the role of RBBP6 is conserved from yeast to humans. Overall, our data are consistent with CPSF endonuclease activation and site-specific pre-mRNA cleavage being highly controlled to maintain fidelity in mRNA processing.

Mechanism-based traps enable protease and hydrolase substrate discovery.

Hydrolase enzymes, including proteases, are encoded by 2-3% of the genes in the human genome and 14% of these enzymes are active drug targets1. However, the activities and substrate specificities of many proteases-especially those embedded in membranes-and other hydrolases remain unknown. Here we report a strategy for creating mechanism-based, light-activated protease and hydrolase substrate traps in complex mixtures and live mammalian cells. The traps capture substrates of hydrolases, which normally use a serine or cysteine nucleophile. Replacing the catalytic nucleophile with genetically encoded 2,3-diaminopropionic acid allows the first step reaction to form an acyl-enzyme intermediate in which a substrate fragment is covalently linked to the enzyme through a stable amide bond2; this enables stringent purification and identification of substrates. We identify new substrates for proteases, including an intramembrane mammalian rhomboid protease RHBDL4 (refs. 3,4). We demonstrate that RHBDL4 can shed luminal fragments of endoplasmic reticulum-resident type I transmembrane proteins to the extracellular space, as well as promoting non-canonical secretion of endogenous soluble endoplasmic reticulum-resident chaperones. We also discover that the putative serine hydrolase retinoblastoma binding protein 9 (ref. 5) is an aminopeptidase with a preference for removing aromatic amino acids in human cells. Our results exemplify a powerful paradigm for identifying the substrates and activities of hydrolase enzymes.

Cryptochrome proteins regulate the circadian intracellular behavior and localization of PER2 in mouse suprachiasmatic nucleus neurons.

The ∼20,000 cells of the suprachiasmatic nucleus (SCN), the master circadian clock of the mammalian brain, coordinate subordinate cellular clocks across the organism, driving adaptive daily rhythms of physiology and behavior. The canonical model for SCN timekeeping pivots around transcriptional/translational feedback loops (TTFL) whereby PERIOD (PER) and CRYPTOCHROME (CRY) clock proteins associate and translocate to the nucleus to inhibit their own expression. The fundamental individual and interactive behaviors of PER and CRY in the SCN cellular environment and the mechanisms that regulate them are poorly understood. We therefore used confocal imaging to explore the behavior of endogenous PER2 in the SCN of PER2::Venus reporter mice, transduced with viral vectors expressing various forms of CRY1 and CRY2. In contrast to nuclear localization in wild-type SCN, in the absence of CRY proteins, PER2 was predominantly cytoplasmic and more mobile, as measured by fluorescence recovery after photobleaching. Virally expressed CRY1 or CRY2 relocalized PER2 to the nucleus, initiated SCN circadian rhythms, and determined their period. We used translational switching to control CRY1 cellular abundance and found that low levels of CRY1 resulted in minimal relocalization of PER2, but yet, remarkably, were sufficient to initiate and maintain circadian rhythmicity. Importantly, the C-terminal tail was necessary for CRY1 to localize PER2 to the nucleus and to initiate SCN rhythms. In CRY1-null SCN, CRY1Δtail opposed PER2 nuclear localization and correspondingly shortened SCN period. Through manipulation of CRY proteins, we have obtained insights into the spatiotemporal behaviors of PER and CRY sitting at the heart of the TTFL molecular mechanism.

Cryo-EM structure of MukBEF reveals DNA loop entrapment at chromosomal unloading sites.

The ring-like structural maintenance of chromosomes (SMC) complex MukBEF folds the genome of Escherichia coli and related bacteria into large loops, presumably by active DNA loop extrusion. MukBEF activity within the replication terminus macrodomain is suppressed by the sequence-specific unloader MatP. Here, we present the complete atomic structure of MukBEF in complex with MatP and DNA as determined by electron cryomicroscopy (cryo-EM). The complex binds two distinct DNA double helices corresponding to the arms of a plectonemic loop. MatP-bound DNA threads through the MukBEF ring, while the second DNA is clamped by the kleisin MukF, MukE, and the MukB ATPase heads. Combinatorial cysteine cross-linking confirms this topology of DNA loop entrapment in vivo. Our findings illuminate how a class of near-ubiquitous DNA organizers with important roles in genome maintenance interacts with the bacterial chromosome.

Precise optical control of gene expression in C elegans using improved genetic code expansion and Cre recombinase.

Synthetic strategies for optically controlling gene expression may enable the precise spatiotemporal control of genes in any combination of cells that cannot be targeted with specific promoters. We develop an improved genetic code expansion system in Caenorhabditis elegans and use it to create a photoactivatable Cre recombinase. We laser-activate Cre in single neurons within a bilaterally symmetric pair to selectively switch on expression of a loxP-controlled optogenetic channel in the targeted neuron. We use the system to dissect, in freely moving animals, the individual contributions of the mechanosensory neurons PLML/PLMR to the C. elegans touch response circuit, revealing distinct and synergistic roles for these neurons. We thus demonstrate how genetic code expansion and optical targeting can be combined to break the symmetry of neuron pairs and dissect behavioural outputs of individual neurons that cannot be genetically targeted.

Animal behaviour and movement emerges from the stimulation of nerve cells that are connected together like a circuit. Researchers use various tools to investigate these neural networks in model organisms such as roundworms, fruit flies and zebrafish. The trick is to activate some nerve cells, but not others, so as to isolate their specific role within the neural circuit. One way to do this is to switch genes on or off in individual cells as a way to control their neuronal activity. This can be achieved by building a photocaged version of the enzyme Cre recombinase which is designed to target specific genes. The modified Cre recombinase contains an amino acid (the building blocks of proteins) that inactivates the enzyme. When the cell is illuminated with UV light, a part of the amino acid gets removed allowing Cre recombinase to turn on its target gene. However, cells do not naturally produce these photocaged amino acids. To overcome this, researchers can use a technology called genetic code expansion which provides cells with the tools they need to build proteins containing these synthetic amino acids. Although this technique has been used in live animals, its application has been limited due to the small amount of proteins it produces. Davis et al. therefore set out to improve the efficiency of genetic code expansion so that it can be used to study single nerve cells in freely moving roundworms. In the new system, named LaserTAC, individual cells are targeted with UV light that 'uncages' the Cre recombinase enzyme so it can switch on a gene for a protein that controls neuronal activity. Davis et al. used this approach to stimulate a pair of neurons sensitive to touch to see how this impacted the roundworm's behaviour. This revealed that individual neurons within this pair contribute to the touch response in different ways. However, input from both neurons is required to produce a robust reaction. These findings show that the LaserTAC system can be used to manipulate gene activity in single cells, such as neurons, using light. It allows researchers to precisely control in which cells and when a given gene is switched on or off. Also, with the improved efficiency of the genetic code expansion, this technology could be used to modify proteins other than Cre recombinase and be applied to other artificial amino acids that have been developed in recent years.

A 68-codon genetic code to incorporate four distinct non-canonical amino acids enabled by automated orthogonal mRNA design.

Orthogonal (O) ribosome-mediated translation of O-mRNAs enables the incorporation of up to three distinct non-canonical amino acids (ncAAs) into proteins in Escherichia coli (E. coli). However, the general and efficient incorporation of multiple distinct ncAAs by O-ribosomes requires scalable strategies for both creating efficiently and specifically translated O-mRNAs, and the compact expression of multiple O-aminoacyl-tRNA synthetase (O-aaRS)/O-tRNA pairs. We automate the discovery of O-mRNAs that lead to up to 40 times more protein, and are up to 50-fold more orthogonal, than previous O-mRNAs; protein yields from our O-mRNAs match or exceed those from wild-type mRNAs. These advances enable a 33-fold increase in yield for incorporating three distinct ncAAs. We automate the creation of operons for O-tRNA genes, and develop operons for O-aaRS genes. Combining our advances creates a 68-codon, 24-amino-acid genetic code to efficiently incorporate four distinct ncAAs into a single protein in response to four distinct quadruplet codons.

Sense codon reassignment enables viral resistance and encoded polymer synthesis.

It is widely hypothesized that removing cellular transfer RNAs (tRNAs)-making their cognate codons unreadable-might create a genetic firewall to viral infection and enable sense codon reassignment. However, it has been impossible to test these hypotheses. In this work, following synonymous codon compression and laboratory evolution in Escherichia coli, we deleted the tRNAs and release factor 1, which normally decode two sense codons and a stop codon; the resulting cells could not read the canonical genetic code and were completely resistant to a cocktail of viruses. We reassigned these codons to enable the efficient synthesis of proteins containing three distinct noncanonical amino acids. Notably, we demonstrate the facile reprogramming of our cells for the encoded translation of diverse noncanonical heteropolymers and macrocycles.

Creating custom synthetic genomes in Escherichia coli with REXER and GENESIS.

We previously developed REXER (Replicon EXcision Enhanced Recombination); this method enables the replacement of >100 kb of the Escherichia coli genome with synthetic DNA in a single step and allows the rapid identification of non-viable or otherwise problematic sequences with nucleotide resolution. Iterative repetition of REXER (GENESIS, GENomE Stepwise Interchange Synthesis) enables stepwise replacement of longer contiguous sections of genomic DNA with synthetic DNA, and even the replacement of the entire E. coli genome with synthetic DNA. Here we detail protocols for REXER and GENESIS. A standard REXER protocol typically takes 7-10 days to complete. Our description encompasses (i) synthetic DNA design, (ii) assembly of synthetic DNA constructs, (iii) utilization of CRISPR-Cas9 coupled to lambda-red recombination and positive/negative selection to enable the high-fidelity replacement of genomic DNA with synthetic DNA (or insertion of synthetic DNA), (iv) evaluation of the success of the integration and replacement and (v) identification of non-tolerated synthetic DNA sequences with nucleotide resolution. This protocol provides a set of precise genome engineering methods to create custom synthetic E. coli genomes.

Selective CRAF Inhibition Elicits Transactivation.

Discovering molecules that regulate closely related protein isoforms is challenging, and in many cases the consequences of isoform-specific pharmacological regulation remains unknown. RAF isoforms are commonly mutated oncogenes that serve as effector kinases in MAP kinase signaling. BRAF/CRAF heterodimers are believed to be the primary RAF signaling species, and many RAF inhibitors lead to a "paradoxical activation" of RAF kinase activity through transactivation of the CRAF protomer; this leads to resistance mechanisms and secondary tumors. It has been hypothesized that CRAF-selective inhibition might bypass paradoxical activation, but no CRAF-selective inhibitor has been reported and the consequences of pharmacologically inhibiting CRAF have remained unknown. Here, we use bio-orthogonal ligand tethering (BOLT) to selectively target inhibitors to CRAF. Our results suggest that selective CRAF inhibition promotes paradoxical activation and exemplify how BOLT may be used to triage potential targets for drug discovery before any target-selective small molecules are known.