Structural insights into the effect of active-site mutation on the catalytic mechanism of carbonic anhydrase

X-ray crystallography was used to elucidate the effect of a single-site mutation on the activity of a native metalloenzyme. The subtle structural modifications around the active site of the enzyme were correlated with the retarded catalytic efficiency in terms of the mechanistic steps and their kinetics.


Introduction
Enzymes greatly enhance the catalytic rates of biochemical reactions compared with their uncatalyzed counterparts and are therefore essential to speed up biochemical processes (Jencks, 1987;Fersht, 1999;Frey & Hegeman, 2007). Enzyme active sites provide highly optimized microenvironments for their specific substrates by providing reactive groups such as nucleophiles or acids/bases that stabilize the transition state. Consequently, changes in the active-site residues can have large effects on enzyme activity. However, direct prediction of the impact of a single mutation on the activity of an enzyme remains challenging owing to the lack of precise correlations between the structure of the protein and its function at atomic resolution (Ishida, 2010). In this study, we describe the effect of a single amino-acid variation on a prototypical enzyme, human carbonic anhydrase II (CA II), by correlating its highresolution reaction-intermediate structures with the measured kinetic parameters. Human carbonic anhydrases are well suited to serve as a model system for our study because their structures and active sites are well defined, and their overall enzymatic mechanism is fairly straightforward and has been studied extensively (Krishnamurthy et al., 2008).
Human carbonic anhydrases catalyze the reversible hydration/dehydration of CO 2 /HCO 3 À (Davenport, 1984;Chegwidden et al., 2013;Frost & McKenna, 2013;Supuran & De Simone, 2015). In the CO 2hydration direction, the first step of catalysis is the conversion of CO 2 into HCO 3 À via the nucleophilic attack of a zinc-bound hydroxide. This reaction is followed by the displacement of the zinc-bound HCO 3 À by a water molecule (equation 1, where E stands for the enzyme; Silverman & Lindskog, 1988). The second step involves the transfer of a proton from the zincbound water to bulk solvent, regenerating the zinc-bound hydroxide (equation 2, where B stands for a general base: either a water or a proton-shuttling residue).
In CA II, the active-site zinc is located at the base of a 15 Å deep cleft and is tetrahedrally coordinated by three histidines (His94, His96 and His119) and a zinc-bound water (W Zn ) [ Fig. 1(a)] . The active-site cavity is further subdivided into two distinct faces consisting of hydrophilic and hydrophobic residues. The hydrophilic face (Tyr7, Asn62, His64, Asn67, Thr199 and Thr200) of the active site coordinates the hydrogen-bonded water network (W1, W2, W3a and W3b) that connects the zinc-bound water to His64, the proton-shuttling residue [ Fig. 1(b)] (Steiner et al., 1975;Tu et al., 1989;Fisher et al., 2005Fisher et al., , 2010Fisher, Maupin et al., 2007;Zheng et al., 2008). It is known that the proton-transfer process is the rate-limiting step in CA II catalysis .
The hydrophobic face (Val121, Val143, Leu198, Val207 and Trp209) is located adjacent to the zinc-bound hydroxide and is responsible for substrate binding (Liang & Lipscomb, 1990;. Leu198, Trp209 and Val121 constitute the mouth and sides of the hydrophobic pocket, while Val143 comprises the base of the hydrophobic pocket. A water molecule termed the 'deep water' (W DW ) is located at Structure of V143I CA II. (a) Overall structure of V143I-0atm: V143I CA II with no CO 2 pressurization. The active site (red box) is located at a depth of 15 Å from the surface. Note that Ile143 is located at the hydrophobic pocket in the active site. (b) Ordered water network in the hydrophilic region serving as a proton-transfer pathway. (c) Surface rendition of V143I-0atm. The entrance conduit (diameter of 7-10 Å , guided with a yellow dotted line) connects the active site to the bulk solvent outside, forming the replenishment pathway. The electron density of the entrance-conduit waters is contoured at 1.5. Hydrophobic amino acids are shaded in red, while hydrophilic amino acids are coloured white. the mouth of this pocket and forms van der Waals contacts with Leu198 and Trp209. W DW occupies the pre-catalytic association site for substrate and is displaced by one of the O atoms of CO 2 during binding (Domsic et al., 2008). The hydrophobic pocket residues are highly conserved and are known to be critical for CO 2 sequestration, although they do not directly interact with the CO 2 molecule.
Between the hydrophilic and hydrophobic sides of the active site, a cluster of ordered waters has recently been identified located near the active-site entrance, termed the entrance-conduit (EC) waters [Fig. 1(c)] (Kim et al., 2018). This ordered water ensemble connects the active site to the external bulk solvent, creating a pathway where water, substrate and product can interchange and interact with bulk solvent. These EC waters are believed to be involved in water replenishment during catalysis, displacing the zinc-bound bicarbonate and restoring the proton-transfer water network.
In order to elucidate the catalytic details of the active site of CA II, several mutational studies have been performed. Most of these variants have been focused at the zinc ion-binding site (Alexander et al., 1993;Kiefer et al., 1993;Ippolito & Christianson, 1994;Lesburg & Christianson, 1995;Huang et al., 1996;Lesburg et al., 1997), the hydrophilic side (protontransfer pathway; Behravan et al., 1990;Krebs, Ippolito et al., 1993;Xue et al., 1993;Ippolito et al., 1995;Huang et al., 2002;Tu et al., 2002;Fisher et al., 2005;Zheng et al., 2008;Turkoglu et al., 2012;Mikulski et al., 2013;Aggarwal et al., 2014) and the hydrophobic pocket (CO 2 -binding site; Alexander et al., 1991;Fierke et al., 1991;Krebs, Rana et al., 1993;West et al., 2012;Nair & Christianson, 1993). In the hydrophobic pocket, a series of mutational studies have been performed targeting the Val121, Val143 and Leu198 residues. It was found that CA catalysis is severely compromised (an $10 4 -10 5 -fold decrease) when the deep water W DW is displaced by replacement of the relevant amino acid by one with a larger side chain (for example, replacement of Val143 by Phe or Tyr), leading to a substantial blockage of CO 2 binding. On the other hand, some of the point mutations, such as Val121 to Ala, Val143 to Ile and Leu198 to Glu, do not directly displace W DW and the CA catalysis is only moderately compromised (a fewfold to a 20-fold decrease). Understanding these moderate effects is most challenging as the overall structures and active sites show little deviation when compared with native CA II. It is expected that delicate perturbations are introduced at the level of intermediate structures during the moderately modified CA catalysis.
In this study, we investigate one of the most challenging cases and describe the subtle effects brought on by a Val143 to Ile (V143I) mutation. As shown in Table 1, the V143I variant shows an approximately tenfold decrease in k cat /K m , while k cat remains almost the same as that for native CA II. Structural analysis was performed by comparing the catalytic intermediate states of native and V143I CA II, which were obtained by cryocooling protein crystals under four different CO 2 pressures [ranging from 0 (no CO 2 pressurization) to 15 atm]. The intermediate states are henceforth referred to as native-0atm (PDB entry 6km3), native-7atm (PDB entry 6km4), native-13atm (PDB entry 6km5) and native-15atm (PDB entry 6km6) and as V143I-0atm (PDB entry 6klz), V143I-7atm (PDB entry 6km0), V143I-13atm (PDB entry 6km1) and V143I-15atm (PDB entry 6km2), respectively. Based upon the structural modifications, we successfully estimated the alterations in the reaction rate constants and the corresponding free-energy profiles in the CA II enzymatic mechanism. This study systematically reveals how a single point mutation influences an enzyme's catalytic pathway at the atomic level, leading to an estimation of the kinetics governing its individual mechanistic steps.

Results
Our X-ray studies (methods are reported in the supporting information) revealed that the overall protein backbones (tertiary structures) of the native and V143I CA II structures were very similar, with C -C r.m.s.d. values of less than 0.14 Å (Supplementary Tables S1 and S2). However, careful structural analysis successfully established subtle but clear changes in the active site (CO 2 -binding site, proton-transfer pathway and water-replenishment pathway; EC waters). The key bound water molecules in native and V143I CA II are listed in Supplementary Tables S3 and S4. , the water molecules (W Zn / W DW /W1) around the zinc ion are initially well ordered at 0 atm CO 2 pressure. At higher CO 2 pressures of 7-15 atm, electron density for the CO 2 molecule becomes apparent, displacing the deep water (W DW ). Concurrently, two intermediate waters W I and W I 0 emerge near Thr200, while W1 disappears at higher CO 2 pressures. The distance between W I and W2 is $4.7 Å , suggesting that the hydrogen-bonded water network that facilitates proton transfer is disrupted when the CO 2 -binding site is fully occupied.

CO 2 -binding site around the zinc ion
On the other hand, the active site of V143I CA II shows noteworthy modifications. The most striking difference is that HCO 3 À is stabilized and is observable at 0 atm CO 2 pressure with an estimated occupancy of $20%  Table 1 Steady-state kinetic parameters for CO 2 hydration by native and V143I CA II.
absence of experimentally introduced CO 2 , it is likely that the captured HCO 3 À is converted from CO 2 absorbed into the crystal from ambient air. As the CO 2 pressure increases, the HCO 3 À occupancy increases to 54% at 13 atm and subsequently decreases slightly to 48% at 15 atm, with increased CO 2 occupancy [Figs. 2( f )-2(h)]. It is likely that the observed decrease in the HCO 3 À occupancy at 15 atm is owing to steric hindrance from the bound CO 2 molecule. When superimposed with the previously reported coordinates of HCO 3 À bound to native CA II (PDB entry 2vvb; Sjö blom et al., 2009), the HCO 3 À position observed in V143I CA II shows noticeable deviations. The HCO 3 À molecule is tilted by 35 with respect to the plane containing W Zn , CO 2 and HCO 3 À in native CA II (Supplementary Fig. S1). In addition, the central C atoms of the two superimposed HCO 3 À molecules in native and V143I CA II are separated by 0.5 Å .
Unlike the HCO 3 À molecule, the CO 2 molecule is less stable in V143I CA II. For example, CO 2 shows almost full occupancy at 7 atm in native CA II, while no CO 2 is visible at 7 atm in V143I CA II, which instead shows the appearance of W DW [Figs. 2(b) and 2( f )]. The CO 2 molecule appears at higher pressures (13 and 15 atm) with a decreased occupancy of $50% [Figs. 2(g) and 2(h)]. The bound CO 2 is tilted by $6 and is situated closer to the zinc ion by 0.34 Å compared with that in native CA II ( Supplementary Fig.  S1). The distance between the end carbon (C 1 ) of Ile143 and the CO 2 molecule is only 3.0-3.2 Å , and this steric disruption seems to affect the critical interactions between the CO 2 molecule and the   À and W DW were determined in V143I-0atm and V143I-7atm, and partial occupancies of HCO 3 À and CO 2 were determined at the higher CO 2 pressures (see the supporting information). The inset (red box) in V143I-0atm shows the difference map (F o À F c contoured at 3.0; green) when the HCO 3 À molecule is not included in the structure refinement. hydrophobic pocket, thereby destabilizing it in the active site. It is worth noting that in contrast to native CA II, the bound CO 2 molecule in V143I CA II is distorted from the plane defined by the bound HCO 3 À molecule, and this seems to be detrimental to the efficient conversion of CO 2 to HCO 3 À .  Fig. 3 shows the proton-transfer pathway including the water network (W1/W2/W3a/W3b) and His64. In native CA II, the water network is initially well ordered at 0 atm CO 2 pressure [ Fig. 3(a)]. As the CO 2 pressure increases, W1 disappears and intermediate waters (

Proton-transfer pathway
. The W2 water, which transfers a proton from W1 to His64, shows an alternative position denoted W2 0 . Considering the steric hindrance between His64 in and W2 0 , it seems that the presence of W2 0 pushes His64 towards the 'out' conformation (away from the water network). Indeed, His64 shows a net movement from the 'in' to the 'out' conformation as W2 0 becomes prominent at higher CO 2 pressures. W3a shows little change, but W3b shows an alternative water position, W3b 0 , and is found to interact with one of the entrance waters W EC1 (and its alternative position W 0 EC1 ). Similarly, in V143I CA II W2 shows the same alternative position W2 0 , and His64 shows the same 'in' to 'out' flip, with similar occupancies as observed in native CA II with increasing CO 2 pressure [Figs. 3(e)-3(h)]. W3b and W EC1 also show the same alternative positions, although their electron densities are slightly weaker. As the distance between W2 and the N atom (N 1 ) of His64 in is relatively long (3.3 Å ), efficient proton transfer seems to depend on the dynamical motions of W2/W2 0 and His64 in /His64 out . These motions are quite similar in native and V143I CA II and therefore proton transfer is not significantly impacted in the variant. Fig. 4 shows the water-replenishment pathway consisting of the ordered EC waters (W EC1 /W EC2 /W EC3 /W EC4 /W EC5 ). In native CA II, the five W EC waters are well ordered at 0 atm CO 2 pressure [ Fig. 4(a) and HCO 3 À cannot coexist in tandem. Indeed, W 0000 EC2 is only visible in V143I CA II at 0 atm, when the HCO 3 À occupancy is low, and disappears as the HCO 3 À occupancy increases at higher CO 2 pressures [Figs. 4(e)-4(h)]. It seems that the relationship between W 0000 EC2 and HCO 3 À is analogous to the relationship between W DW and CO 2 . In native CA II, the hydrophobic cavity produces an electrostatic environment in which the deep water (W DW ) can be locally stabilized around the zinc ion and then replaced with one of the O atoms of CO 2 upon CO 2 binding. In V143I CA II, the altered hydrophobic cavity owing to the V143I mutation produces a slightly different electrostatic environment in which an additional water position (W 0000 EC2 ) can be locally stabilized around the zinc ion and subsequently replaced with one of the O atoms in HCO 3 À during CA II catalysis. The release of the W 0000 EC2 molecule upon HCO 3 À binding seems to reduce the entropic cost of the process, analogous to the release of W DW upon CO 2 binding in native CA II. Thus, it is likely that the altered multiple conformations of W EC2 allow HCO 3 À to bind more easily and firmly in a tilted configuration at the active site.

Discussion
In this study, we have successfully identified the 'fine' structural changes in the CA II intermediates induced by a singleresidue mutation at the active site. The V143I mutation in CA II produces steric hindrance and induces subtle changes in the electrostatic environment of the active site. The resulting effects on the CA II intermediates can be summarized as follows: (i) the dynamical motions and the allowed configurations of CO 2 are slightly restricted and the binding affinity of HCO 3 À is increased with a distorted configuration and (ii) The water-replenishment pathway including entrance-conduit waters (cyan) and intermediate waters ( constants (k 1 , k À1 , k 2 and k 3 ) and the corresponding freeenergy profiles during the CO 2 -hydration reaction of CA II.
Firstly, the V143I mutation restricts the configurational freedom of the CO 2 molecule within the hydrophobic pocket of CA II. Previous mutational studies on hydrophobic pocket residues (Val121 and Val143) suggest that the hydrophobic pocket of native CA II is involved in 'ushering' the CO 2 molecule to the zinc hydroxide in the active site, but does not directly interact with the CO 2 molecule to hold it in a specific orientation. Therefore, it is most likely that the CO 2 molecule in aqueous solution is guided to the hydrophobic pocket of CA II but retains some degrees of freedom with regard to the acceptable configurations (positions and orientations) around the zinc hydroxide that eventually facilitate its rapid conversion into HCO 3 À . In the V143I variant, although the hydrophobicity of the active-site pocket is increased, the added methyl group appears to sterically restrict the number of favourable configurations and the dynamical motions accessible to the CO 2 molecule within the cavity. This estimation is supported by the crystallographic observation that the CO 2 molecule is distorted towards the zinc ion, tilted by $6 and is destabilized with lower occupancies in the V143I variant. Consequently, interconversion from CO 2 to HCO 3 À becomes less efficient, leading to a reduced k 1 (k 1 V143I < k 1 native ). The lower k 1 value also implies an enhanced activation-energy barrier for the step [Arrhenius relationship: k 1 (T) = Aexp(ÀE 1 /RT), where A is a pre-exponential constant, R is the molar gas constant, T is the absolute temperature and E 1 is the activation energy] (Fig. 5).
Secondly, the V143I mutation induces a slightly different electrostatic environment in the hydrophobic cavity, thereby altering the location and dynamics of the W EC2 water. It seems that one of these altered W EC2 waters (W 0000 EC2 ) increases the binding affinity of the HCO 3 À molecule in the active site. This stronger binding of HCO 3 À suggests that the free energy of the enzyme-product (EZnHCO 3 À ) complex is lowered in the V143I variant (Fig. 5). In conjunction with the larger activation energy for the k 1 reaction, it can be deduced that the activation energy for the reverse reaction k À1 is increased even further, leading to a reduced k À1 value (k À1 V143I < k À1 native ). On the other hand, the alterations of the W EC2 water in the V143I variant make the intermediate waters (W I and W I 0 ) slightly less stable, therefore possibly slowing down the replenishment of the active-site water network and HCO 3 À dissociation. This results in a reduced k 2 value (k 2 V143I < k 2 native ). Thirdly, the V143I mutation seems to have little effect on the kinetics of the proton-transfer reaction. It is observed that W1 is more stabilized in the V143I variant intermediates. It should be noted that the intermolecular proton transfer can occur via the fully established water network W Zn !W1! W2!His64. This implies that if a mutation induces the destabilization of W1, proton transfer could be significantly perturbed. However, stabilization of W1 as in the V143I variant does not necessarily suggest a faster proton-transfer process. Rather, it is the dynamical motions of W2 and His64 that have a more critical influence on the proton-transfer rate. Our results suggest that both W2 and His64 show very similar dynamical motions. Taken together, it seems that the overall proton-transfer rate k 3 is not significantly altered in the V143I mutant (k 3 V143I ' k 3 native ). The interplay between the modified reaction rate constants, as discussed above, now allows us to determine the effect of the V143I mutation on the measured kinetic parameters (k cat and k cat /K m ; Krebs, Ippolito et al., 1993). The steady-state kinetic parameters for the CO 2 -hydration reaction are listed in Table 1. The k cat value shows little change, but the second-order rate constant k cat /K m shows an approximately tenfold decrease in the V143I variant. The parameter k cat contains rate constants from the initial enzyme-substrate complex through the remaining steps, including proton transfer. Therefore, for the proposed mechanistic scheme (equations 1 and 2), k cat can effectively be represented as k cat = k 2 k 3 /(k 2 + k 3 ). On the other hand, the ratio k cat /K m contains rate constants for the initial association of the substrate CO 2 through the dissociation of the product HCO 3 À . Hence, k cat /K m only contains rate constants from equation 1 (and not equation 2) and is represented as k cat /K m = k 1 k 2 /(k À1 + k 2 ).
It is known that in native CA II the HCO 3 À -dissociation process (k 2 ) is much faster than the reverse interconversion from HCO 3 À to CO 2 (k À1 ) and the proton-transfer rate (k 3 ), i.e. k 2 native ) k À1 native , k 2 native ) k 3 native , with k 2 native ' 10k À1 native , k 2 native ' 15k 3 native and k À1 native ' 1.5k 3 native (Table 1). Combining this information with the insights gleaned from our study (k 1 V143I < k 1 native , k À1 V143I < k À1 native , k 2 V143I < k 2 native and k 3 V143I ' k 3 native ), we can estimate the effect of the V143I point mutation on the observed kinetic parameter k cat in the following way. In the native state [k cat ] native = k 2 native k 3 native /(k 2 native + k 3 native ) ' k 3 native (using k 2 native ) k 3 native ), indicating that the turnover rate in the native is almost the same as the proton-transfer rate and that the proton-transfer process is the rate-limiting step. In the Estimated free-energy profiles for the CO 2 -hydration reaction catalyzed by CA II. The energy states of native CA II (black) are from a previous study (Behravan et al., 1990 . This result indicates that HCO 3 À dissociation (k 2 ) remains faster than proton transfer (k 3 ) in the V143I variant and that the proton-transfer process is still the rate-limiting step in the V143I variant. The negligible reduction in k 2 V143I suggests that its activation energy is not significantly increased (Fig. 5).
The second-order rate constant k cat /K m shows the following relationship in the native state: [k cat /K m ] native = k 1 native k 2 native / (k À1 native + k 2 native ) ' k 1 native (using k 2 native ) k À1 native ). This indicates that [k cat /K m ] native mostly reflects the CO 2 -binding step and its interconversion to HCO 3 À . On the other hand, in the V143I variant both the dissociation of HCO 3 À (k 2 ) and the reverse interconversion (k À1 ) from HCO 3 À to CO 2 are retarded (k 2 V143I < k 2 native and k À1 V143I < k À1 native ), leaving the relation k 2 V143I ) k À1 V143I unresolved. However, the relation k 2 V143I ) k 3 V143I ' k 3 native obtained from the k cat analysis above suggests that k 2 V143I is still an order of magnitude larger than k 3 native , while k À1 V143I < k À1 native ' 1.5k 3 native suggests that k À1 V143I is comparable to or less than k 3 native , thereby ensuring that the relation k 2 V143I ) k À1 V143I remains valid. Thus, we can estimate the k cat /K m in the V143I variant in the following manner: . This estimation indicates that the reduced [k cat /K m ] V143I is mostly owing to the decrease in k 1 , reflecting the slower CO 2 binding and interconversion to HCO 3 À in the V143I variant. It should be noted that the reduction in the k À1 value has little effect on [k cat /K m ] V143I in the direction of CO 2 hydration, as long as the reduction in k 2 is small enough to keep the relation k 2 V143I ) k À1 V143I valid. However, it is expected that the reduction in the k À1 value would have significant consequences for the HCO 3 À -dehydration direction. Finally, considering that k cat ' k 3 and [k cat /K m ] ' k 1 in both native and V143I CA II, the Michaelis constant K m is expressed in the following way: K m native = k 3 native /k 1 native and K m V143I = k 3 V143I /k 1 V143I . Consequently, K m V143I > K m native can be obtained using the estimated relationships k 1 V143I < k 1 native and k 3 V143I ' k 3 native . The relationship shows that the substrate concentration needed to reach half of the maximum reaction velocity is larger in V143I CA II mainly owing to the slower CO 2 binding and interconversion to HCO 3 À . Although our study was performed for a single point mutation within the hydrophobic pocket (V143I), our approach and interpretations can be extended to arbitrary mutations in CA II. Considering the forward CO 2 -hydration direction, the mutation can first perturb substrate funnelling into the hydrophobic pocket via steric hindrance, thereby limiting the configurations that allow its efficient conversion into product. This influence is directly reflected in k cat /K m , but not in k cat . Secondly, the mutation can structurally distort the proton-transfer pathway by perturbing the water network or its associated stabilizing residues, and this effect is directly reflected in k cat but not in k cat /K m . Thirdly, the mutation can alter the product-dissociation process via direct steric hindrance or perturbations in the water-replenishment pathway. This influence can be intricate and is reflected both in k cat and k cat /K m . On the other hand, the mutation can affect the reverse reaction, the interconversion from product to substrate and substrate dissociation, but this has little influence on either k cat or k cat /K m in the hydration direction, as long as the reverse interconversion process is much slower than the product-dissociation process.

Conclusion
We systematically studied the effect of a single-residue mutation on the CA II catalytic pathway at atomic resolution. We have successfully captured the high-resolution intermediate states of the V143I variant and shown clearly that the single point mutation induces noticeable changes in substrate and product binding at the active site and in the waterreplenishment pathway, but has little effect on the protontransfer pathway. The structural information was then utilized to estimate the reaction rate constants and the free-energy profiles during the catalytic cycle, unravelling the effect of the point mutation on the altered kinetic parameters. We believe that the detailed and systematic approach in our CA II study can be extended to identify the specific roles of target aminoacid residues in many other biologically important enzymes. We also anticipate that our detailed descriptions could serve as a reference point for future theoretical and computational studies that may lead to an advanced understanding of enzyme mechanisms at the quantum-chemistry level.