All oligonucleotide sequences used in LexA_Pa amplification, cloning, and cassette mutagenesis were purchased from Integrated DNA Technologies (IDT) and are available upon request. The lexA gene was amplified via polymerase chain reaction with LexA_Pa genomic primers from the P. aeruginosa PA01 strain and cloned into the pET41 expression vector engineered with an N-terminal His tag and C-terminal Streptavidin tag.⁴⁰ The S125A active site mutant was generated by QuikChange site-directed mutagenesis. The enzyme was heterologously expressed in E. coli, followed by one-step purification using the N-terminal tag. The purified product was predominantly full-length (∼26 kDa), with trace amounts of self-cleavage that occurs during the course of purification.

For the efficient generation of saturation and positional scanning mutagenesis variants, mutations introducing the unique restriction sites AgeI and Eagl were engineered at nucleotide positions 255 and 282, respectively, in the lexA coding sequence using QuikChange site-directed mutagenesis. An additional mutation (BamHI) at nucleotide 267 and a TAG stop were included within the cassette region, as mechanisms for selection against wild-type sequences from the parent vector. This construct served as the parent cloning vector for cassette-based mutagenesis. Briefly, the LexA_Pa cloning vector was digested with AgeI, BamHI, and EagI for 2 h at 37 °C and treated with calf intestinal phosphatase. Individual oligonucleotides containing a degenerate codon (NNS) at each position from position 85 to 94 were ordered from IDT and annealed to their complement with a standard annealing protocol. Annealed oligonucleotides containing sticky ends complementary to AgeI and EagI cleavage sites were phosphorylated with T4 polynucleotide kinase for 1 h at 37 °C. The oligonucleotide cassettes were ligated into the digested cloning vector using T4 DNA ligase. The ligation product was then transformed into New England Biolabs (NEB) 10-β competent cells using standard transformation procedures. One-tenth of the transformation mixture was plated to estimate the library size, and the remaining portion was grown overnight in Luria-Bertani broth. The isolated plasmids from this culture constituted the saturation mutagenesis library at each position (I85X, G86X, etc.). Successful incorporation of the degenerate NNS codon was verified by Sanger sequencing of the library.

The same cassette-based mutagenesis strategy was adopted to generate the individual point mutations in residues spanning positions P5 to P2′ of the LexA_Pa cleavage region (residues 86–92). For each of the seven positions, 20 forward and reverse oligonucleotides encoding each amino acid mutation were constructed and ordered from IDT. Oligonucleotides containing two mutant codons were used for the LexA_Pa A89C/I94C double mutant. Oligonucleotide pairs were annealed, phosphorylated, and ligated into the digested parent vector, as described above. After transformation, individual colonies were selected and mutant plasmids were sequenced to confirm the proper insertion of the mutation cassettes.

Expression plasmids were transformed into BL21(DE3)-pLysS E. coli cells for heterologous expression. For the saturation mutant library, the liquid culture after transformation was used as the starter overnight culture. For point mutants, the transformation was plated and an individual colony selected for the starter overnight culture. The following day, 35 mL Luria-Bertani broth cultures were inoculated with 1 mL of overnight culture and grown at 37 °C until the OD₆₀₀ reached ∼0.6. The cultures were subsequently induced with 1 mM isopropyl β-d-1-thiogalactopyranoside, shifted to 30 °C, and grown for 4 h. The culture was centrifuged at 4000 rpm for 20 min at 4 °C and the pellet stored at −80 °C until it was purified. Thawed cell pellets were lysed per protocol with Bugbuster Mastermix (Novagen). The soluble supernatant was incubated with 100 μL of reconstituted HisPur resin (Pierce) for batch binding at 4 °C for 1 h in 10 mL polyprep columns (Bio-Rad). The flow-through was discarded and the resin subsequently washed three times with 10 resin volumes of wash buffer [25 mM sodium phosphate (pH 7.0), 150 mM NaCl, and 30 mM imidazole]. Following the wash, bound protein was eluted from the resin with 3 resin volumes of elution buffer [25 mM sodium phosphate (pH 7.0), 150 mM NaCl, and 200 mM imidazole]. The eluted proteins were then dialyzed into 25 mM Tris-HCl (pH 7.0), 150 mM NaCl, and 10% glycerol. The same protocol was used for the wild-type enzyme, saturation mutant libraries, and individual positional scanning mutants. For the 140 positional scanning mutants (20 amino acids × 7 positions), the cohorts of 20 mutants at each position were processed in parallel to provide an internal (wild-type) control.

To qualitatively screen the cleavage ability of LexA_Pa variants, purified protein was mixed in a 1:1 ratio with 2× cleavage buffer (100 mM Tris-Glycine-CAPS and 300 mM NaCl) at pH 7.2 or 10.6. Reaction mixtures were incubated at room temperature for 16 h. Cleavage was quenched by adding 2× Laemmli sample buffer to the reaction mixture and by denaturation at 95 °C for 10 min. The extent of LexA_Pa cleavage was visualized by running reaction samples on 15% sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) gels and Coomassie staining. Mutants that displayed cleavage were selected for further quantitative analysis. Coomassie-stained gels were imaged on a Typhoon Imager using red laser excitation at 633 nm and no filters. The fraction of cleaved protein was calculated by dividing the density of the LexA_Pa cleavage products by the sum of the density of all LexA_Pa full-length and cleaved components, using Quantity One (Bio-Rad). Linear detection of Coomassie-stained products was verified by analysis of standards.

RecA of E. coli was purchased from NEB. Purified LexA_Pa was incubated with an excess of activated RecA*, generated by premixing ∼100 μg/mL RecA, 900 μM ATPγS, and 10 μM ssDNA.⁴¹ Reactions without RecA were run parallel as a negative control. Reaction mixtures were incubated at room temperature for 16 h. The cleavage products were analyzed and quantified as described for alkali-mediated cleavage.

For quantitative kinetic analysis, 25 μL of purified protein was mixed with 25 μL of 2× cleavage buffer (pH 10.6) and the mixture incubated at 37 °C for 2 h. At the given time points, 5 μL of the reaction mixture was removed and the reaction rapidly quenched in 2× Laemmli sample buffer. The extent of cleavage over time was visualized on Coomassie-stained 15% SDS–PAGE gels. For restrictive positions, time points were collected in triplicate, while reactions were run in duplicate for the permissive positions. Coomassie-stained gels were imaged and quantified as detailed above.

The density of the protein bands was quantified using Quantity One (Bio-Rad). For each time point, the fraction of uncleaved LexA_Pa was calculated as described above and plotted versus time. Rate plots were then fit to the first-order exponential decay equation A = A₀e^–kt using Prism to obtain k, the observed rate of cleavage. To calculate the specificity profile, we used the enoLOGOS web tool to generate a normalized sequence LOGO of the P5–P2′ sequence preferences.⁴² The cleavage rate constants, k, at each position were used as the relative scaling factors, with an unknown weight type and the frequency method for calculating the height of the symbol stacks.

To assist with the analysis of our biochemical data, we generated a structural model of the LexA_Pa CTD. We entered the amino acid sequence of the LexA_Pa CTD (residues 81–204) into Modeler 2.0,⁴³ with the crystal structure of activated LexA from E. coli as the homology template (PDB entry 1JHE).¹²

LC–MS/MS was performed at the Proteomics & Systems Biology Core Facility at the University of Pennsylvania Perelman School of Medicine. Purified LexA_Pa A89C/I94C samples were either reduced or not reduced with 1 M dithiothreitol and treated with 300 mM iodoacetamide (molecular mass of 57 Da). Samples were subsequently digested with trypsin and subjected to LC–MS/MS. Peptide fragments were identified and analyzed with PEAKS.⁴⁴

The LexA protein is strongly conserved across all families of bacteria and shares a high level of sequence similarity.⁹⁻¹⁶ Given the importance of drug resistance in the opportunistic pathogen P. aeruginosa, we focused our efforts on studying the LexA of P. aeruginosa (LexA_Pa). A sequence alignment of LexA_Pa with LexA of E. coli (LexA_Ec) reveals a 64% sequence identity, with notable features such as the serine-lysine catalytic dyad and the internal cleavage loop conserved (Figure S1A of the Supporting Information). This strong sequence identity provided the basis for structural modeling of the LexA_Pa CTD to help interpret our biochemical assays (Figure S2 of the Supporting Information). We modeled LexA_Pa based on the structure of the CTD of LexA_Ec in the active conformation (PDB entry 1JHE), with the high degree of sequence homology resulting in a root-mean-square deviation of 1.2 Å between the known and modeled structures.

We next cloned, expressed, and purified LexA from reference strain PA01. In accordance with prior biochemical assays on LexA from other species, we demonstrated that our tagged LexA_Pa is proficient in both alkali- and RecA*-mediated autoproteolysis (Figure S3 of the Supporting Information). Recombinant LexA_Pa displayed cleavage kinetics similar to that of tagged or untagged LexA from other species (see below).^19,33−45

Prior biochemical studies have revealed that the LexA protease has poor cleavage activity in trans.(19,26) In its wild-type catalytic form, both the full-length LexA protein and the isolated LexA protease domain are unable to efficiently cleave the target substrate of a second LexA enzyme.¹⁹ Extensive mutagenesis of the enzyme’s active site and the target sequence can allow for trace detectable activity in trans; however, it is clear that the native substrate of LexA is itself.¹⁹ We explored the possibility that full-length LexA or the CTD alone may cleave short peptides that mimic the internal cleavage loop; however, we did not observe cleavage in fluorescence- or LC–MS-based cleavage assays (data not shown). For these reasons, conventional peptide array-, genetic library-, and mass spectrometry-based methods for assessing protease specificity could not be readily translated to characterizing LexA_Pa specificity.⁴⁶⁻⁵⁰ We therefore decided to directly assess LexA’s specificity by exhaustive mutagenesis of the cleavage loop within the enzyme and assessing (in cis) autoproteolysis activity. Given that this approach would require extensive mutagenesis, we implemented an efficient cassette mutagenesis strategy for introducing variations into the internal substrate of LexA (Figure S4 of the Supporting Information).⁵¹⁻⁵³ The specificity determinants of LexA_Pa could be delineated broadly at first by saturation mutagenesis (the introduction of a degenerate codon) and then in detail by positional scanning mutagenesis (the generation of individual point mutants to each of the 20 amino acids).

For saturation mutagenesis, 10 duplexed oligonucleotide cassettes that contain a degenerate NNS codon at each position from Ile85 to Ile94 (e.g., I85X) were produced. The degenerate NNS codon encodes all 20 potential amino acid variants and one potential stop codon (TAG). The calculated depth of the library at each position was >1000-fold, and the presence of a degenerate NNS codon at each position was verified by sequencing. Although this library is not proportionally represented for the various amino acid substitutions, we reasoned that the general patterns of restrictive and permissive positions could be gleaned from analysis of a positionally diversified cohort of variants.

For each position, we expressed and purified the mutant protein library in cohort, which showed expression characteristics and solubility similar to those of wild-type LexA_Pa. Each cohort was then incubated overnight at pH 7.2 or 10.6, and the fraction of cleavage was assessed under these conditions (Figure 2). The positions clustered into two general groups based upon comparison to wild-type LexA_Pa: positions P5 (G86), P3 (V88), P2 (A89), P1 (A90), and P1′ (G91) appear to be restrictive to mutation, displaying small changes in the fraction of cleaved protein (<30%), while the remaining positions were more permissive to the introduction of a degenerate codon, exhibiting changes in cleavage levels closer to that of the wild-type enzyme despite being highly diversified (Figure 2). We repeated the analysis in the presence or absence of RecA* and found the same overall patterns of tolerant and restrictive positions with this alternative stimulus for autoproteolysis (Figure S5 of the Supporting Information). Notably, as RecA*-mediated cleavage analysis was used as a purely qualitative measure, we used RecA from E. coli in our assays (74% identical and 87% similar to that of P. aeruginosa) because of its availability and the observation that LexA_Pa cleaves as efficiently as LexA_Ec under RecA* stimulation.

The qualitative overview from saturation mutagenesis set the stage for detailed quantitative analysis. Here, we focused on the residues of positions P5–P2′, spanning the most restrictive sites we identified by saturation mutagenesis. Furthermore, the LexA_Pa homology model suggests that these residues form the most extensive interactions with the surrounding active site pocket (Figure S2 of the Supporting Information).

In our positional scanning mutagenesis approach, for each position of interest, we individually mutated the residues to each of the 20 potential amino acids (G86A, G86S, G86T, etc.) using cassette mutagenesis. For each position, all 20 variants were expressed, purified in parallel and subjected to both alkali- and RecA*-mediated cleavage overnight to identify specific variants that permitted autoproteolysis (Figure S6 of the Supporting Information). Together, the individual point mutants in this assay validated the patterns observed with saturation mutagenesis. The full cleavage profiles for each position are provided in Figure S6 of the Supporting Information, with select examples of a restrictive and tolerant position shown in Figure 3. Restrictive positions, such as P1 (A90), tolerate only a few mutations, while permissive positions, such as P4 (R87), tolerated all amino acids, although to varying degrees (Figure 3). Both alkali- and RecA*-mediated cleavage screens yielded similar cleavage results, with the exception of position P5, where RecA*-mediated cleavage was more restrictive than alkali-mediated cleavage (Figure S6A of the Supporting Information). Overall, the cleavage behavior of select restrictive mutations in LexA_Pa (G86V, G86D, V88M, V88E, A89V, A90D, and A90T) agrees qualitatively with the results of a prior study of LexA from E. coli that identified slow-cleaving variants via limited mutagenesis and genetic screening.⁵⁴ One notable exception is the aspartate mutation at the P1′ glycine, which completely abrogates alkali-mediated and RecA*-mediated cleavage in LexA_Pa (Figure S6F of the Supporting Information) but allows limited levels of cleavage in LexA from E. coli.⁵⁴

Individual mutants deemed cleavage-proficient under overnight cleavage conditions were next subjected to detailed kinetic analysis. We performed the cleavage reaction under alkali-induced conditions to allow for direct comparison of the various mutants. Alkali-mediated cleavage, while nonphysiological, has the distinctive advantage of being a unimolecular first-order reaction, allowing us to probe the intrinsic substrate preferences of LexA_Pa without confounders such as potential alterations in RecA* interactions (Table 1). Employing our method, we determined the rate of autoproteolysis of all cleavable mutants. Representative data from one highly restrictive (P1) position and one tolerant position (P4) are presented in panels C and D of Figure 3, with kinetic plots for all cleavable variants shown in Figure S7 of the Supporting Information. Notably, the rate of cleavage of the wild-type enzyme at each position differed slightly (range of 53–144 × 10^–5 s^–1) as a result of batch-to-batch variability in expression and purification. Given that each cohort of 20 mutants at a position was prepared in parallel, the rate of each mutant was scaled to that of the wild type to obtain the relative amino acid preferences, with the comprehensive analysis summarized (Figure 4A).

The positions upstream of the cleavage site show an alternating pattern of being restrictive or permissive to variation, with restrictive positions favoring the wild-type residue over all other amino acids. At position P5 (G86), we see a modest tolerance for amino acids with either a small or flexible side chain, such as A or S, while larger side chains reduce the level of LexA_Pa self-cleavage (Figure 4B). Position P4 (R87), which appears solvent-exposed in the LexA_Pa model, shows tolerance to all variants, despite being conserved as a basic residue across species (Figure 1C). Indeed, certain acidic or large hydrophobic variants such as R87E and R87W appear to cleave at rates higher than that of the wild-type enzyme. Position P3 (V88) is restrictive, with the wild-type residue preferred 5-fold over the next best variant, V88I. Three of the four best variants are the β-branched amino acids, suggesting the possibility of active site engagement with the branched side chains. Other tolerated variants are hydrophobic and predominantly small amino acids. Position P2 (A89), exposed in the active conformation of LexA_Pa, tolerates all amino acids, although to different extents. Variants with the smallest side chains (A89G, wild type with A89, and A89S) are most readily cleaved, while the level of cleavage generally decreases as a function of increasing size.

The positions immediately flanking the cleavage site show highly restrictive patterns, while preferences are again relaxed in the downstream positions. At position P1 (A90), only the five smallest amino acids are tolerated: the wild-type A90 is preferred over the next best variant A90G by 4-fold, while cleavage is detectable with A90C, A90S, and A90T, suggesting significant active site constraints at this position. Position P1′ (G91) is situated at the β-turn of the LexA cleavage region in its active conformation. It is the most restrictive position among those examined, with only a serine mutation retaining any detectable self-cleavage activity (∼50-fold slower than the WT). Position P2′ (A92) is a solvent-exposed residue and situated at the end of the β-turn. The tolerance of this position to mutation is in line with its poor conservation across species isoforms of LexA (Figure 1C).

The kinetic data determined at each position allow for the construction of the overall sequence preference heat map for LexA_Pa (Figure 4A). The calculated cleavage rates were used to build the position-specific scoring matrix, which was then graphically converted into a sequence LOGO depiction that encapsulates the substrate specificity profile for LexA_Pa (Figure 4B). As evidenced in the sequence LOGO, positions tolerant to mutation have high entropy or variability, while restrictive positions show low variability.

In our kinetic assays of the LexA_Pa mutants, we observed that certain mutants interestingly exhibited a cleavage rate higher than that of the wild-type enzyme (“hyperactive” mutants). While some hyperactive mutants of LexA have previously been discovered through genetic screens,^55,56 the molecular basis for hyperactive self-cleavage has not been well-investigated. In our mutagenesis, A92P shows a notable cleavage rate ∼4-fold higher than that of the wild-type enzyme and undergoes autoproteolysis even under nonstimulatory conditions (neutral pH, no RecA*) (Table 1 and Figure S6G of the Supporting Information). The transition between the cleavage-incompetent and cleavage-competent forms of LexA is associated with a β-turn at the cleavage site (A90-G91).¹² Because position A92 is located at the end of the β-turn in the LexA cleavage loop, we hypothesized that a mutation to proline stabilizes the β-turn and could thereby promote the active conformation of the protein.^57,58 To test this hypothesis, we introduced cysteine mutations at positions A89 and I94, two residues that are within disulfide bonding distance in the active conformation (Figure S2 of the Supporting Information). We then confirmed by mass spectrometry that the variant formed an intramolecular disulfide bond across the β-turn (Figure S8 of the Supporting Information). While the individual point mutants have rates of cleavage slower than (A89C) or comparable to (I94C) that of the wild type, the A89C/I94C double mutant is enhanced, in line with the A92P cleavage rate (Figure 5). While the mechanism cannot be definitively assessed, the A92P and A89C/I94C LexA variants point to the importance of secondary structure formation in the cleavage region, with potential implications for inhibitor design.

We have identified three positions in the LexA cleavage region that are restrictive to extensive mutation. Interestingly, although exhaustive mutagenesis was not performed, point mutations at these positions were also identified to be important for the self-cleavage of the related λ CI repressor, and the positions are well conserved in LexA isoforms across species (Figure 1C).²² For LexA_Pa, positions P3 and P1 favor the wild-type residues and show tolerance toward β-branched or small amino acids, respectively. For both positions, analysis of the crystal structure shows extensive van der Waals contacts between the side chains and hydrophobic pockets in the LexA active site (Figure 6B). Because increasing or decreasing the size of the amino acid at these positions reduces cleavage efficiency (Table 1), we conclude that positions P3 and P1 determine specificity through active site contacts.

Position P1′ is the most restrictive position, tolerating only the wild-type glycine and a serine mutation. Position P1′ is located at the end of the β-hairpin turn and undergoes a dramatic bond rotation during the transition between active and inactive conformations.¹² Only glycine can readily adopt and interconvert between these unusual Ramachandran angles, thus explaining the strong preference for the wild-type residue at that position. Larger amino acids likely cause steric clash with the surrounding protein or are hindered in making the necessary bond rotations. Interestingly, trace cleavage was also detectable in the G91S mutant (Figure S6F of the Supporting Information). We speculate that mutation to a serine likely allows for the formation of hydrogen bonds between the side-chain hydroxyl group and peptide backbone that can stabilize the turn, compensating for the decreased level of rotational freedom.

Unlike the stringent determinants, positions P5 and P2 tolerate a wide range of mutations yet display a preference for a particular subset of amino acids. At position P5, the wild-type residue, glycine, is preferred over a larger amino acid, such as alanine or serine. Similar to position P1′, in the determined structures of LexA, the glycine at position P5 undergoes a significant rotation during the transition from the inactive state to the active state, thus explaining the presence of a flexible residue that can serve as a molecular hinge.¹²

Position P2 displays an intermediate phenotype, as well: although all variants are tolerated, amino acids that exceed the size of serine decrease the rate of self-cleavage. Structurally, the side chain of position P2 forms contacts with the side chain of position P4′ (I94) located on the opposite side of the loop (Figure S2 of the Supporting Information). Increasing the size of the amino acid likely reduces the rate of cleavage due to steric clash.

In addition to deciphering specificity determinants, our study has also identified two highly permissive positions that flank the LexA scissile bond. Positions P4 and P2′ can be mutated to the whole amino acid spectrum without abrogating LexA cleavage, although changes in the rate of cleavage are observed. The LexA_Pa model structure shows both positions to be relatively solvent-exposed, forming few specific contacts with the surrounding active site. Interestingly, both positions also harbor mutations that enhance the level of self-cleavage compared to that of the wild-type enzyme. At position P4 (R87), mutations removing a basic residue appear to enhance self-cleavage (Table 1), despite the conservation of the basic residue in LexA across species (Figure 1C). We speculate that these rate-enhancing mutations may result from the removal of a potential repulsive interaction with an adjacent (and well-conserved) positively charged residue (R120) in the active conformation. Position P2′, in contrast to position P4, is highly variable across species (Figure 1C), which correlates well with its tolerance to mutagenesis in our biochemical study. At position P2′, introduction of a proline mutation enhances self-cleavage, likely by stabilizing the β-turn around the scissile bond and thereby promoting the cleavage-competent state of LexA. This model is supported by our analysis of the disulfide cross-linked A89C/I94C mutant, which shows cleavage kinetics similar to those of the A92P mutant (Figure 5B).

Our results expand the list of previously identified mutations that enhance LexA cleavage.⁵⁵ In a prior study, mutations located at positions P5′ and P8′ in E. coli LexA, distant from the catalytic site, were shown to stabilize the active conformation of the entire LexA cleavage loop.^12,55 Our results for A92P and A89C/I94C mutants suggest that stabilizing the β-turn around the scissile bond with rigid cyclic structures similarly enhances cleavage and could offer novel avenues for inhibitor design.

The nature of our experimental design, which focused on mutations in the cleavage loop and kinetic analysis of autoproteolysis, requires that the enzyme and its substrate change simultaneously. Additionally, given that we evaluated single-point mutants, the impact of combinatorial modifications in the substrate could not be readily assessed from our approach. Despite these limitations, the relative independence of the cleavage loop and the active site residues make it possible to integrate across our biochemical analysis and structural modeling to provide a detailed picture of LexA’s specificity. From our analysis, we postulate that the nature of the specificity determinants of LexA can be classified under two broad categories: those permissive for substrate recognition and those permissive for the conformational change in self-cleavage (Figure 6B). Much like conventional proteases, LexA has substrate recognition pockets that determine its selectivity, most evident at positions P3 (V88) and P1 (A90) (Figure 6B). As a key difference from conventional proteases, however, the LexA cleavage region must internally rearrange to adopt a proper cleavage-competent conformation for proteolysis to occur. Positions P5 (G85) and P1′ (G91) are highly dynamic positions, implying that conformational flexibility at either end of the cleavage region appears to be important for autoproteolysis. In support of the importance of dynamic rearrangements in LexA, stabilizing the β-turn across the scissile bond promotes self-cleavage at a level above that of the wild type. We postulate that these requirements for both proper sequence and conformation serve as a mechanism of self-recognition and limit off-target proteolytic activity. Our findings help rationalize the low cleavage activity of the LexA protease in trans and offer insights into rational probe design.¹⁹

Our biochemical results are particularly informative when we compare LexA_Pa to members of the larger LexA/signal peptidase superfamily, many of which mediate bacterial stress responses (Figure S1B of the Supporting Information). This superfamily includes LexA homologues such as Bacillus subtilis DinR and Vibrio cholerae SetR, the phage λ CI repressor, and UmuD and MucA, self-cleaving enzymes that function in the SOS-linked translesion DNA synthesis.^{11,13,18,19,22,36,61,62} In these enzymes, position P1′ is universally conserved as a glycine (Figure S1 of the Supporting Information), suggesting that the flexibility of position P1′ is likely critical to the mechanism. The stringent residues involved in side-chain contacts are also strongly conserved, with a β-branched residue at position P3 and a small residue at position P1 conserved across comparators (Figure S1B of the Supporting Information). Notably, the position S3 and S1 recognition pockets are entirely conserved between LexA_Pa and LexA_Ec (Figure S1A of the Supporting Information), suggesting a common mechanism of side-chain recognition in forms of LexA from different species. For the intermediate determinants, the position P5 glycine offers the most revealing comparison to other self-cleaving enzymes. We speculate that this residue is a critical “hinge” for the conformational change that allows self-cleavage within a LexA monomer. Interestingly, position P5 is not conserved in UmuD where cleavage of one monomer can occur in the other monomer’s active site.⁹ Overall, comparing the superfamily to other common serine proteases, we note that LexA favors small amino acids in the positions flanking the scissile bond, most similar to the elastase family of serine proteases.^24,25 However, the unique self-cleaving mechanism and its conformational requirements are distinguishing features that appear to be well conserved across the LexA/signal peptidase superfamily.

In addition to revealing insights into the mechanism of self-cleavage, several aspects of our substrate specificity studies can potentially help direct future efforts to develop small molecule probes of the SOS pathway. First, while LexA does not have proficient activity with peptides in trans, it does have detectable activity. Peptides or peptidomimetics that incorporate preferred features, such as the side chains of positions P3 and P1 or β-turn-stabilizing structures, may be exploited to enhance binding. Even weak binding peptides or peptidomimetics could be converted into more potent probes by the incorporation of mechanism-based covalent inhibitor warheads into such molecules. Second, our discovery of tolerance at positions P4 and P2′ could be exploited in inhibitor discovery. The introduction of fluorescent reporters at these positions could aid in screening for inhibitors of self-cleavage or, given the highly dynamic nature of position P2′, potential allosteric modulators of the conformational change. Finally, our discovery of a range of hypocleavable LexA variants can potentially help further validate LexA’s viability as a therapeutic target. Given that LexA inhibitors would be unlikely to fully recapitulate the LexA catalytic mutant, these variants can be used to reveal the amount of inhibition that will be necessary to synergize with current antibiotics or to slow acquired antibiotic resistance.

We hypothesize that the identification of rate-enhancing mutations by prior genetic studies and our biochemical studies has wider implications for bacterial mutagenesis and evolution.^55,56 The fact that mutations can enhance cleavage suggests the possibility that the rate of LexA autoproteolysis may be finely tuned to be fast enough to facilitate robust SOS responses but slow enough to prevent aberrant SOS pathway activation. This characteristic has been thought to be functionally important in different members of the LexA superfamily. As an example, temperate phage repressors such as the phage λ CI repressor self-cleave more slowly than LexA during SOS-inducing treatments, thereby inducing prophage formation only with extensive DNA damage.¹⁹ Similarly, mutagenesis proteins, such as UmuD, self-cleave at rates slower than that of LexA, which could promote translesion DNA synthesis later in the SOS response, only after higher-fidelity repair mechanisms have failed.¹⁸ Given that self-cleavage can be either slowed or enhanced by mutations, we hypothesize that the LexA cleavage rate has been selected for to allow for proper activation of the SOS pathway and that LexA may serve as a rheostat for evolution under stress. Our study offers the possibility of modulating the cleavage rates of LexA across a large range to assess the impact on bacterial mutation and survival directly under varying degrees of stress.