Handicap-Recover Evolution Leads to a Chemically Versatile, Nucleophile-Permissive Protease

Mutation of the tobacco etch virus (TEV) protease nucleophile from cysteine to serine causes an approximately ∼104-fold loss in activity. Ten rounds of directed evolution of the mutant, TEVSer, overcame the detrimental effects of nucleophile exchange to recover near-wild-type activity in the mutant TEVSerX. Rather than respecialising TEV to the new nucleophile, all the enzymes along the evolutionary trajectory also retained the ability to use the original cysteine nucleophile. Therefore the adaptive evolution of TEVSer is paralleled by a neutral trajectory for TEVCys, in which mutations that increase serine nucleophile reactivity hardly affect the reactivity of cysteine. This apparent nucleophile permissiveness explains how nucleophile switches can occur in the phylogeny of the chymotrypsin-like protease PA superfamily. Despite the changed key component of their chemical mechanisms, the evolved variants TEVSerX and TEVCysX have similar activities; this could potentially facilitate escape from adaptive conflict to enable active-site evolution.

Thomas Shafee, [a, b, c] Pietro Gatti-Lafranconi, [a] Ralph Minter, [b] andFlorian Hollfelder* [a] Mutationo ft he tobacco etch virus (TEV) protease nucleophile from cysteinet os erine causes an approximately~10 4 -fold loss in activity.T en rounds of directed evolution of the mutant, TEV Ser ,o vercamet he detrimental effectso fn ucleophile exchange to recover near-wild-typea ctivity in the mutant TEV Ser X. Rather than respecialising TEV to the new nucleophile, all the enzymesa longt he evolutionary trajectory also retained the ability to use the original cysteinen ucleophile. Therefore the adaptive evolution of TEV Ser is paralleled by an eutral trajectory for TEV Cys ,i nw hich mutations that increases erine nucleophile reactivity hardly affect the reactivity of cysteine. This apparent nucleophile permissiveness explains how nucleophile switchesc an occur in the phylogenyo ft he chymotrypsin-like proteaseP As uperfamily.D espite the changedk ey component of their chemical mechanisms, the evolvedv ariants TEV Ser Xa nd TEV Cys Xh ave similara ctivities;t his could potentially facilitate escape from adaptivec onflict to enable active-site evolution.
Enzymes achieve efficient catalysis through the precise orientation of ak ey set of active-site residues.T his arrangementi s dependent on chemical constraints, to the extent that some active-site geometries have convergently evolvedm any times. [1] Consequently,a ctive-site residues are the moste volutionarilyc onserved within enzymef amilies. Phylogenetic analysis of extended protein superfamilies suggestst hat even residues that are crucial for activity are exchanged during evolution over sufficiently long timescales. [2] The evidence for such exchanges raises the question of how ag ene codingf or an inefficient enzyme can persevere in the transition from one type of active site to another.T herei sn oe volutionary advantage for maintaining ag ene coding for ac atalytically impaired or inactive protein, thusc reatingt he scenario of "adaptive conflict". [3] We know from studies on enzymes [4] and enzyme models [5] that precise positioning is easily disturbed. Minute disturbancesd own to the picometer scale causes ubstantial rate reductions. [6] Given the delicacyo fc atalytic arrangements, it is unknown how the evolution of active sites avoidsu nfit variants.
Serine and cysteine proteases are textbook examples of enzymese mploying covalent, nucleophilic catalysis ( Figure 1A) [7] that leads to substantial rate accelerationo fadifficult reaction (with ah alf-life of % 500 years). [8] The sophisticated interplay of the multiple active-site residues involved,i ncluding for example, the charger elay system of the catalytic triad, [7c] suggests that anyd eviation from such ah ighly efficient arrangementi s likely to be penalised. [9] However,p hylogenetica nalysiso fp roteasess uggests that nucleophile exchanges do occur during evolution. The PA clan of chymotrypsin-like proteases [10] encompasses both serine and cysteinep roteases evolved from a hypothetical common ancestor. [11] Constructing ap hylogeny of this clan of proteases ( Figure 1B)s hows that-withinahighly conserved structure-nucleophile switches musth ave occurredb yd ivergent evolution at least once:c ellular proteases use aserine nucleophile, but both cysteine and serineprotease families are found in viruses. [12] When the active-site nucleophiles of serine or cysteinep rotease are interconverted, the single atomic change typically leads to a > 10 4 -fold reduction in k cat /K M . [9a, 14] Although both thiol and hydroxy groups can act as nucleophilic catalysts, their positioning is likely to be suboptimala fter mutation due to structurald ifferences between oxygen ands ulfur:o xygen's smaller atomic radius (by~0.4 ) [14a] and the formationo f shorter bonds (decreasing d C-X and d X-H by~1.3-fold each) would be expectedt od isturb the precise nucleophile positioning. In addition there are reactivity differences:s ulfur is softer, and its different pK a (4-5 units lower for RSH compared to ROH) provides al arger fraction of the activef orm of the nucleophile at physiological pH;t his explains why the reactivity of the hydroxy side chain of serine that replaces cysteine would be compromised. [14d] These considerations of chemical reactivity and structure raise the question of how such nucleophile transitions have occurredi np roteases, despite the enzymei nactivation typically associated with mutating ak ey active-site residue. Handicaprecover experiments can be used to find if any mutationsc an epistatically offset ak nown deleterious mutation or the replacemento fanative cofactor. [15] Here we use this approach to demonstrate ascenariothat could satisfythe fitness requirementso fp rotein evolution by mutating the crucial nucleophile of tobaccoe tch virus cysteinep rotease (TEV Cys )t os erine (TEV Ser ), recovering activity by directed evolution (DE) and measuring trade-offsinr esponse to nucleophile switches. The nucleophile mutation compromised the centrepiece of the catalytic mechanism and consequently had am uch greater effect on activity than in previoush andicap-recover experiments. [15] AT EV Ser mutant (C151S mutation) was constructed, and the substitution of the cysteine thiol nucleophile by a serine alcohol was found to reduce activity by four orders of magnitude (Figure 2a nd Figure S4 in the Supporting Information). The effect of this deliberately introduced handicap was then quantified by measuring the reactionk inetics (by monitoring the hydrolysis of the C-Y FRET-pair substrate, [16] Figure S2). Conversion of the catalytic nucleophile from sulfur to oxygen resulted in biphasic kinetics (Figures 2A and S3, Ta ble S2); this is consistent with the formation and breakdown of an acyl-enzyme intermediate (i.e.,afast first step followed by as lower,r ate-limiting step, Figure S5). The second-order rate constants k obs1 2 and k obs2 2 of TEV Ser were found to be 80 and 20 000 times lower,r espectively,t han the measured secondorder rate constant of TEV Cys (representing k cat /K M ).
In order to investigate how the enzymec an compensate for the handicapo fu sing an on-native nucleophile and altered reaction chemistry,t en rounds of DE for activity recovery were performed (numbered TEV Ser to TEV Ser X). Each round of DE consisted of error-prone PCR (1.3 AE 0.4 amino acid mutations per gene), activity screening of 350 enzyme variants by destruction of FRET in cell lysate ( Figure S2), ands election of the single best variant.A ny S151C revertants were discarded to force evolution to follow af orwardp athway.M easurement of turnover rates in cells during screening reflects enzyme fitness as the product of both chemicalr eactivity and catalyst concentra-tion (determined by biophysical properties, such as folding and solubility). Thes ameF RET-pair substrate was used for both in vivo screening and in vitro kinetics to describe the enzyme-substrate interactions that were relevant for the selection (and avoid unique effects of the recognition of for example, small-peptide substrates with differentreactivity and affinity).
During the rounds of evolution, no mutations in residues that make direct contacts with the triad (or are within ar adius of 4)r esulted from experimental selections. Conversely,n ine of the 13 point mutations accumulated 4-8 f rom the catalytic triad, in the second shell of residue interactions (Figure3). In vitro kinetics of purified variants showedthat the process of directed evolution recoveredp roteolytic activity by an improvement in both k obs1 2 (2 10 3 -fold) and k obs2 2 (3 10 3 -fold) and also changed the burst amplitude (Figure 2A,T ables S2 and S3), with diminishing improvements in later rounds. The accumulation of mutations around the enzyme active site (seconds hell) that increased catalysis reduced soluble expression sixfold; however,i nr ounds Va nd VI, surface mutations( W130C and E194D)a nd aC -terminal truncation (due to af rameshift) were selected that improved both solubility and activity ( Figure 2C).
In addition to the adaptive trajectory of TEV Ser ( Figure 2B, front), the identified adaptive mutationsw ere examined in the parental background, by reverting the nucleophile to the original cysteinea te ach step of the evolutionary trajectory (Figure 2B,b ack). The kinetics of these TEV Cys variants could be fitted to am onophasicm odel with good correlation coefficients. Rather than respecialising the active site to use serine, as is typical of directed-evolution experiments,t he 14 mutations accumulated by DE proved nearly neutralt oa ctivity with Figure 1. A) Simplified catalytic mechanism of cysteine and serine proteases (Nu = S, e.g.,i nTEV Cys ;o rN u= O, e.g.,i nTEV Ser ). [7] In the enzyme's catalytic triad (black) aspartate aligns and polarisesh istidine, whichr educes the nucleophile's pK a and positionsi tfor reaction. The activatednucleophile attackst he carbonyl of the peptideb ond of the substrate (red), atetrahedral intermediate is stabilised by backboneamide hydrogens ("oxyanionh ole"), and the Cterminus of the substrate is ejected, leavingacovalenta dductbehind. This acyl-enzyme intermediate (lowerp anel) is hydrolysed by histidine-activated water to release the Nterminus of the substrate (aided by proton transfer to histidine), and this regenerates the free enzyme. B) The phylogeny of the PA clan (MEROPS classification) [11] of proteases with nucleophile and possible nucleophile-switching events indicated. Proteases with knowns tructures were aligned to TEV Cys by structural comparison using DALI, [13] sequences of unknownstructure were addedb ys equence alignment with BLASTp (see Ta ble S1 and FigureS1f or details).
ChemBioChem 2015, 16,1866 -1869 www.chembiochem.org the original cysteinenucleophile (i.e.,did not trade-off). Whereas the k obs1 2 and k obs2 2 of TEV Ser Xa re 1000-fold improved over those of TEV Ser ,t he nucleophile revertants retained activity within fourfoldo fT EV Cys (Figures 2B and S6). The evolutionary trajectory therefore results in twin enzymes, differing only in their nucleophile (TEV Ser X$TEV Cys X) and with only as mall, 2.3fold differenceinactivity upon nucleophile exchange.
By forcingT EV into al ocal fitness valley (TEV Ser )a nd experimentally evolving for activity recovery,w em apped out an uphill trajectory (TEV Ser !TEV Ser X) that lies parallel in sequence space to an early flat, neutralt rajectory of mutantsw ith constant Cys nucleophile (TEV Cys !TEV Cys X; Figure 2B). Therefore, despite deliberatelye volvingT EV Cys by using af itness valley, the nucleophile-permissive TEV Ser Xc an also be accessed from TEV Cys by nearly neutralm utationsw ithout anyl arge drops in activity.T he most closely relatedn atural serine protease only retains1 5% sequence identityt oT EV Cys .H owever,t he mutant TEV Ser Xs hows 92.4 %i dentity ( Table S2, Figure 3), thus suggesting that only af ew mutations are necessary to accommodate an ucleophile switch. The > 1000-fold improvementt og enerate an ucleophile generalist with only 13 mutationse xplains how divergente volutiono fc ore catalytic machinery can occur within evolutionarily superfamilies such as the PA clan (Figure 1B). It also emphasises the powero fd irected evolution to find solutions for the challenge of retuning chemical reactivity. [17] The challenge of nucleophile permissiveness is conceptually similart ot hat of catalytic promiscuity (the ability of an enzymet oa ccept different substrates): how can an enzyme make and breakb onds different from those it has evolved for? [18] The evolution of promiscuousa ctivitiesh as been previously observed to pass through catalytic generalists, which were able to promote an ew reaction, whileretaining some activity on their original substrate. Generalist enzymes are proposed to perform an important role in the evolution of new functions by being particularly evolvable. [19] Althoughe nzyme promiscuity towards different substrates is well documented, [18,20] the ability to use different residues for nucleophilic, covalentc atalysis represents an alternative kind of chemical versatilityi nt he core catalytic machinery. [21] Figure 2. Activity recovery of the directed evolution lineagefrom TEV Ser to TEV Ser Xa nd the corresponding nucleophile revertants, in which the original Cys nucleophile was restored. A) Kinetic traceso fe nzymea ctivityf or the wild type-type enzyme (TEV Cys ), the nucleophile mutant (TEV Ser ), the evolved enzyme from the tenth round (TEV Ser X) and the evolved mutantw ith ar everted nucleophile( TEV Cys X). Conditions: [E] = 1 mm,[ S] = 1 mm,p H8,258C. B) Developmento ft he activity of purified TEV mutantsa safunction of the roundso fd irected evolution.Activities are plotteda st he second orderrate constants k obs2 2 of the ten TEV Ser variants and k cat /K M of the ten TEV Cys nucleophile revertants. Solid arrows indicatethe DE route (starting with the deleterious TEV Cys !TEV Ser mutation,then activity recovery in ten rounds from TEV Ser to TEV Ser X) to arrive at the neutral twin enzymes TEV Ser Xa nd TEV Cys X, capable of usinge ither nucleophile. The dotted arrow indicates an alternative, nearly neutral TEV Cys !TEV Cys Xr oute. Conditions:[ E] = 1-8 mm,[S] = 1 mm,p H8,2 58C. Standardd eviations of four repeats were below 15 % (Figure S5). C) Soluble expression of evolved variants and TEV Cys back-mutants. Error bars representstandard deviationoft wo biological repeats. Even though the lynchpin of catalysis in the protease active site-then ucleophile-was mutated, the large activity drop was readily recoverable by evolution. Quantitatively,b oth the handicapi ntroduced and the extent of the recovery exceed those previously observed by approximately two orders of magnitude. [15] Specifically,a120-fold reduction triggered by cofactor exchange was followed by a7 0-foldr ecovery in as tudy by Miller et al. [15a] and a4 00-fold reduction caused by mutation of ac onserved residue was followed by a4 0-fold recovery observed by Wellner et al., [15b] compared with a20000-fold reduction and 3000-fold recovery in this work.
What is more, the similar rates of TEV Cys Xa nd TEV Ser X( 310 3 vs. 710 3 min À1 m À1 )r epresent ar aree xample of catalytic promiscuitya th igh, wild-type levels (in contrast to promiscuous, yet low-activity catalytic generalists). [22] Paradoxically,anucleophile mutation would be predicted to be more difficult to recover from, when compared to evolution to accommodate promiscuous bindingo fm ultiple substrates, as two different types of bonds (OÀCv s. SÀC) are being formed and cleaved. However,o ur data suggest that the differences in nucleophile reactivity and structure can be readily accommodated by TEV protease with apparently little trade-off between rates for different nucleophiles. The unexpected nucleophile tolerance suggestst hat chemically versatile intermediates such as TEV Ser/Cys Xe xist that could facilitate the phylogenetically observed switch between protease clans that differ in their nucleophiles prior to specialisation. The protein framework of TEV Ser/Cys Xa llows two nucleophilest oe xecute their function with good efficiency and constitutes am olecular solution to escape from adaptive conflict.