Heavy Enzymes and the Rational Redesign of Protein Catalysts

Abstract An unsolved mystery in biology concerns the link between enzyme catalysis and protein motions. Comparison between isotopically labelled “heavy” dihydrofolate reductases and their natural‐abundance counterparts has suggested that the coupling of protein motions to the chemistry of the catalysed reaction is minimised in the case of hydride transfer. In alcohol dehydrogenases, unnatural, bulky substrates that induce additional electrostatic rearrangements of the active site enhance coupled motions. This finding could provide a new route to engineering enzymes with altered substrate specificity, because amino acid residues responsible for dynamic coupling with a given substrate present as hotspots for mutagenesis. Detailed understanding of the biophysics of enzyme catalysis based on insights gained from analysis of “heavy” enzymes might eventually allow routine engineering of enzymes to catalyse reactions of choice.


Introduction
Enzymes are naturallyo ccurring protein catalysts that control the chemistry of life. [1] Their catalytic power is exploited by industry for,i nter alia, waste management andt he production of food, pharmaceuticals, textilesa nd fine chemicals. [2] Because they are able to catalyse difficult synthetic reactions without the need for extreme temperature, high pressure or toxic chemicals andr equire simplified downstream processing (due to fewer unwanted side products), enzymes are increasingly becoming viewed as sustainable alternatives to synthetic catalysts, offering reductions in energy costs and in environmental damage. [2,3] However,t he selectivity of enzymes is often for substrates that are not optimal for the needs of industry,amajor limitation that holds back their widespread use. [4] This problem can be addressedb yt he redesign of naturale nzymes,e ither throughr ational design or throughd irected evolution. Though some key achievements have been seen, redesigning enzymes is not as imple task. [4c, 5] Arecent reviewofe nzyme engineering showedt hat only % 5% of modified enzymesi naliterature sample (60 enzymes produced by directed evolution and 15 computationally designed/redesignede nzymes) gave more than 10 4 -fold increases in catalytic efficiency (k cat /K M )o ver the native proteins. [4c] To understand the cause of this failure and to become bettere ngineers of enzymes, it is essential further to improveo ur understanding of the principles by which natural enzymes operate.
In ac onventionalu nderstanding of enzymology,t he catalytic power of enzymes comesfrom their ability to stabilise transition states through binding interactions, thus loweringr eaction activation energies. [6] However,a ntibodies that possess binding sites that are complementary to transition states either fail to catalyser eactions or display significantly lower efficiency than the naturale nzymes. [7] There are many possible explanations for this failure, including poor design of the hapten. [7] It has also been noted that antibodies lack the residues required in order to participate in catalysis:f or acid-base protons huffling, for example. [7b] Also, unlike moste nzymes, catalytic antibodies lack the ability to mediate conformational changes for the binding and release of substrates and products. [7] To explain these observations, controversial enzymology models have been developed. Notably,i th as been suggested that the catalytic power of enzymes is mediated through dynamic motions in the protein. [8] This has been at opic of intense debate, and some researchers have disputedt he need to invokes uch "promoting motions" andh ave proposed that transition state theory alone can explain the catalytic power of enzymes. [6b, e, 9]

Heavy Enzymes
The observation that isotopically labellede nzymes sometimes show reducedr ates of catalysis was first made in 1969. [10] In 2011, Schramm and co-workersp ioneered the utilisation of such isotopically labellede nzymes to probe the possible roles of protein motions in catalysis. [11] ln this method, all non-An unsolvedm ystery in biologyc oncerns the link between enzyme catalysis and protein motions. Comparison between isotopically labelled "heavy" dihydrofolate reductases and their natural-abundance counterparts hass uggested that the coupling of protein motionst ot he chemistry of the catalysed reaction is minimised in the case of hydride transfer.I na lcohol dehydrogenases, unnatural, bulky substrates that induce additional electrostatic rearrangements of the active site enhance coupledm otions. This findingc ould provide an ew route to engineering enzymes with altered substrate specificity,b ecause amino acid residues responsible for dynamic coupling with ag iven substrate present as hotspots for mutagenesis. Detailed understanding of the biophysics of enzymec atalysis based on insights gained from analysis of "heavy" enzymes might eventually allow routine engineering of enzymes to catalyse reactions of choice. exchangeable carbon,n itrogen and hydrogen atoms in an enzyme are replaced with their heavy counterparts-15 N, 13 C and 2 H-to generate am odification with an increased mass and slower motions. The increased atomic mass alters the vibrational frequencies but accordingt ot he Born-Oppenheimer approximation leaves the potential energy surface( PES) unaltered. The rate of the chemical step is measured, and the ratio of the rate constantsf or the light enzymet ot hose for its heavy counterpart gives an enzyme kinetic isotopice ffect (KIE). If as ignificant fraction of protein atoms has been isotopically substituted, an enzyme KIE of unity implies no significant coupling of dynamics to the chemical step, whereas aK IE above or below unity is taken to imply significant coupling of protein motionstoc atalysis.

Dihydrofolate Reductase
Intensive studies on enzyme KIEs have been performed on dihydrofolate reductase (DHFR), including combined experimental and computationala pproaches to investigate different homologues andv ariants. [12h, j, 14a, 16] DHFR catalyses hydride transfer from NADPH to tetrahydrofolate (THF,Scheme 1). [17] Because of the small mass of hydrogen, this reaction has a significant tunnelling component and is therefore particularly sensitivet oc hanges in protein dynamics. The temperature dependence of the heavy enzyme KIE on hydride transfer has been investigated,b yu sing pre-steady state kinetics, for DHFRsf rom organisms that have adapted to live at different temperatures. [12h-j, 16b] At physiological pH, DHFR from Escherichia coli (EcDHFR) shows an enzyme KIE on the hydride transfer rate of 0.93;i tr ises to 1.18 as the temperature is increased from 10 to 40 8C. [12j] The DHFR from the psychrophilic Moritella profunda (MpDHFR) shows an enzymeK IE that rises from 1.07 at 5 8Cto1.45 at 30 8C. [12i] The DHFR from the thermophile Geobacillus (formerly Bacillus) stearothermophilus shows the reverse trend, with the KIE falling from 1.65 at 5 8Ct o1 .09 at 45 8C. [12h] In each case, the KIE approaches unity close to the physiological temperature of the hosto rganism;t his thus strongly suggests that dynamic coupling mightb ea taminimum around physiological temperature ( Figure 2). [12h-j, 16b] Computational studies illustrate that the observed KIEs in DHFR are not due to hindered "promoting motions" but rather to increased re- Figure 1. Strategies for the production and analysis of heavy enzymes. A) to D) show A) the production of natural-abundance enzyme,B)whole-enzyme isotopelabelling, C) single-residue isotopelabelling, and D) segmental isotope labelling by production of two peptides, only one of which is labelled with heavy isotopes,that are ligated together and refolded. E) How heavy enzymes are analysed by kineticsa nd computational analysis.   [12h-j, 14a] The increased mass has reducedt he frequencies associated with protein motions, leading to ap ossible delay in the reorganisationo ft he active site environmenti nr esponse to the fast changes taking place in the chemical system during barrierc rossing. Incomplete environmental relaxation can induce ab arrier recrossing event in which the chemical system is unable to progress to the product state and has to return towards the substrate state. Increased recrossing events can also be interpreted as a consequence of an effective "friction" acting on the reaction coordinate, as discussed later.
This hypothesis was supported by the heavy enzyme study of the conformationally restricted EcDHFR mutant N23PP/ S148A, which shows an inefficientp rocess of electrostaticp reorganisation and an increasei nf ast-timescale dynamics in the active site. [16a] These findings are consistent with those made in computational analysiso ft he unrelated enzymes HIV-1 PR and PNP. [18] To understand the precise origin of these enzyme KIEs further,i ti sn ecessary to identify the residues andr egions of the protein responsible for the observed effects. In the case of EcDHFR,n ative chemical ligationw as used to constructh ybrid isotopomers in which either the Nterminus-including the flexible M20 loop-was labelleda nd the remainder of the protein was left with natural-abundance isotopes, or vice versa. [14a] Labelling of the M20 loop impacted the steady-statek inetics, in whichp hysicals teps are rate-limiting,b ut an enzyme KIE of unity was observed under pre-steady state conditions, under which hydride transfer is rate-limiting. Labelling the whole protein with the exceptiono ft he M20 loop restored the full enzyme KIE observed in the case of the fully labelled enzyme. This demonstrates that the origin of the enzymeK IE is not in the first 28 residues of the protein.T od iscover the microscopic origin of dynamic effects, it will be necessary to extend the work to label the FG and GH loops selectively.
These studies on DHFR have progressed our knowledge of protein dynamics andsuggest that the catalytic rate reductions observedi nh eavy enzymes under non-physiological conditions should not be taken as evidencet hat protein motions drive the chemical step (i.e.,p romoting motions). [12h-j, 16, 19] Instead, heavy enzymes present more trajectories that recross the transition state dividings urface, because the ability of the heavy protein to adapt to changes in the chemical system is less than that of the light enzyme. In other words, the isotopic substitution reduced protein motions, giving greater friction between the protein and the chemical system as it advanced along the reaction coordinate. The friction concept is ac onvenientw ay to express the effect of enzymatic degrees of freedom on ag iven reactionc oordinate, by viewing it as an effective friction acting against the advancing of the system past the coordinate.I no ur treatment, for enzymes acting on their naturals ubstrates under physiological conditions, this effect can be seen as as mall perturbation of the equilibrium description assumedi nt ransition state theory. [20] In nature, an enzyme's active site is preorganised to work with specifics ubstrates at ag iven temperature. Thus, under these conditions, barrierc rossinge ssentially involves the degrees of freedom of the chemical system and, for the purposes of calculating activation free energies and rate constants,p rotein motions can be considered in equilibrium with them. Slow motionsp recede fast motions on the way to the transition state. The actives ite reorganisationn eeded to accommodate the charged istribution of the chemical step takes place on ad ifferent timescale from the transition state crossing. Hence, protein motions will have their greatest impactb efore or after the chemical step. [16a, 20] This viewpoint conflicts with theories developed from work on other,u nrelated enzymes. [11, 12c-g, 13a] According to these authors, enzymatic degrees of freedomw ould be an integrative part of the barrier-crossing event and are therefore important to determine the properties of the transition state. It has furthermore been suggested that DHFR is unique and that conclusions from this enzyme should not be extended to other systems. [21] In response to this criticism, the insights gained from DHFR weref urthert ested in another system:a lcohol dehydrogenase (ADH). [12k] Alcohol Dehydrogenase According to the theories developed in the DHFRw ork, dynamic coupling is increased under non-physiological conditions requiring greater reorganisation of the active site during the chemical step of the catalytic process. If this is correct,i t therefore followst hat dynamic coupling should also be increased when unnatural substrates are used, because the active site architecturei sn ot optimised for the corresponding chemicalt ransformations. [12k] To test this proposal,t he promiscuous zinc-dependent alcohold ehydrogenasef rom G. stearothermophilus (BsADH) was used. [12k] BsADHi sathermostable tetramer that catalyses the NAD + -dependent interconversion of alcohols and aldehydes (Scheme 2). [22] It has been used as a biocatalyst in the generation of cinnamyl alcohol, av aluable compound used for fragrance, food flavouring and synthesis of pharmaceuticals. [23] Like DHFR, this enzymec atalyses as imple hydride transfer reaction. The heavy enzymeK IE for BsADH was measured for a range of substrates from well-tolerated "good" substrates to poorly tolerated "bad" substrates. [12k] "Good" substrates were small nonconjugated molecules with values of k cat ranging from 2t o8s À1 ( Figure 3A). "Bad" substrates were bulky and highly conjugated, with k cat values below 2s À1 .B ecause hydride transfer is partially rate-limiting, the heavy enzyme KIE measurements were based on k cat . [24] No dynamic coupling was observedf or any of the substrates at physiological temperature (40 8C), at which enzyme KIEs were around unity.H owever, at lower temperature (20 8C), the enzyme KIEs rose with inverse correlation to k cat ( Figure 3B). The absence of dynamic coupling at physiological temperature is consistent with earlier work on DHFR and shows that dynamics do not contributet o the reaction under physiological conditions. [12h-j, 16, 19] The KIEs observed at lower temperature correlated with k cat and confirmed the hypothesis that unnatural, bulky substrates require assistance from protein dynamics to produce al arger reorganisation of the active site. [12k] As ar esult,s lightly greaterp rotein friction is generated on the chemical system along the evolution of the reaction coordinate. This meansm ore protein movements( femtosecondm ass-dependent protein motions) that can be coupled to the crossing of the transition state dividing surface, an event that occurs on at imescale of the same order of magnitude as thep rotein motions (femtoseconds).

Future Outlook for Protein Engineering
Our growing understanding of protein dynamics could help to engineer better enzymes in the future.T heoreticiansh avep roposed that introducing an ew "promotingm otion" would improve the activity of aromatic amine dehydrogenase butt his has never been tested experimentally. [25] Similarly,w ork on PNP has shown that the efficiency of barrier crossingi nah eavy enzyme can be modified by mutations that enhance promot-ing vibrations. [15] Experimentally,t his resulted in the inversion of the enzymeK IE from an ormalK IE of 1.31 to an inverse KIE of 0.75. [15] However,t he mutant enzyme was less catalytically efficient than the wild type. An alternative or complementary approachf or engineering dynamics emerges from the study of 16,19] In these enzymest here are no "promoting motions" but rather protein dynamics are involved in active site reorganisation under non-physiological conditions or when poorly tolerated substrates are used. Identification and mutationo fr esidues responsible for dynamic effects in BsADHc ould therefore provide ar oute towards the rational reengineering of this enzyme for unnatural substrates. It is thus now necessary to locate the region of BsADH and the particular aminoa cid residues responsible for dynamic effects in order to test the hypothesis that these residues are hotspots for re-engineering of the enzyme's substrate profile. An umber of experimental and computational techniques for identifying residues responsible for dynamic effects exist;t hey include labellingo fasingle amino acid (a techniquep reviously applied to PNP) [13] or the production of hybrid isotopomers in which one particular loop or domain of ap rotein is isotopically labelled. [14] Such hybridsc an be constructed by using av ariety of techniques including chemical ligation, in which ap eptide containing aC -terminal thioester can be ligated to ap eptide containing af ree N-terminalc ysteiner esidue. This technique has been applied to EcDHFR. [14] Other approaches involve the use of peptide ligases [26] or protein trans splicing with split inteins. [27] Theb est approachf or ap articular protein has to be experimentally determined and often requires time-consuming optimisation of the ligationa nd subsequent refolding of the ligated chain. For the technique to becomep ractical as ar outine methodf or enzymeengineering, furthera dvances need to be made in protein ligation and refolding technologiest o enablee asy construction of hybrid isotopomers. An alternative, complementary approach is to use computer simulations to predict the residues responsible for friction along the reaction coordinate. [18] It mayb eq uestioned whethert he gains in catalytic efficiency from such engineering will be large, given the small enzymeK IEs observed. Although the enzymeK IEs appear small, there is as ignificant kineticd ifferenceb etween "good" and "bad" substrates. Hence, mutation of residues that hinder the progression of the reaction can translate into am easurable increasei nc atalytic turnover.N evertheless, because only a small number of enzymes have to date been studied by the heavy enzyme methodology,i ti sc urrently unclear which enzymes obey the rule of minimised dynamic coupling under physiological conditions. Hence investigations of protein dynamics over ab road range of enzyme families with different chemistries and cofactors are essential.

Summary
Heavy-isotope labelling of proteins combined with detailed computational work is au seful tool for studying the contribution of protein motions to the catalytic step. Studies on DHFR variants have shown that dynamic effects are only significant under nonphysiological conditions that require reorganisation of the active site. [12h-j, 16, 19] This leads to the hypothesis that dynamic coupling should also be increased when unnatural substrates are used because the active site architecture is not optimised for the corresponding chemical transformations. [12k] By this argument, dynamic coupling indicates the extentt ow hich an active site is suited to ap articulars ubstrate. This was confirmed through heavy enzyme studies on an alcohol dehydrogenasef rom G. stearothermophilus. Amino acid residues responsiblef or dynamic effectsm ight be useful as targets for mutagenesis to create an active site optimally suited for a designed substrate. Unlike directed evolution, which requires a high-throughputs creen to assayalarge number of variants, such ar ational approach based on insight into dynamic effects only requires as mall number of mutantstob eanalysed. [28]