Larger active site in an ancestral hydroxynitrile lyase increases catalytically promiscuous esterase activity

Hydroxynitrile lyases (HNL's) belonging to the α/β-hydrolase-fold superfamily evolved from esterases approximately 100 million years ago. Reconstruction of an ancestral hydroxynitrile lyase in the α/β-hydrolase fold superfamily yielded a catalytically active hydroxynitrile lyase, HNL1. Several properties of HNL1 differ from the modern HNL from rubber tree (HbHNL). HNL1 favors larger substrates as compared to HbHNL, is two-fold more catalytically promiscuous for ester hydrolysis (p-nitrophenyl acetate) as compared to mandelonitrile cleavage, and resists irreversible heat inactivation to 35 °C higher than for HbHNL. We hypothesized that the x-ray crystal structure of HNL1 may reveal the molecular basis for the differences in these properties. The x-ray crystal structure solved to 1.96-Å resolution shows the expected α/β-hydrolase fold, but a 60% larger active site as compared to HbHNL. This larger active site echoes its evolution from esterases since related esterase SABP2 from tobacco also has a 38% larger active site than HbHNL. The larger active site in HNL1 likely accounts for its ability to accept larger hydroxynitrile substrates. Site-directed mutagenesis of HbHNL to expand the active site increased its promiscuous esterase activity 50-fold, consistent with the larger active site in HNL1 being the primary cause of its promiscuous esterase activity. Urea-induced unfolding of HNL1 indicates that it unfolds less completely than HbHNL (m-value = 0.63 for HNL1 vs 0.93 kcal/mol·M for HbHNL), which may account for the ability of HNL1 to better resist irreversible inactivation upon heating. The structure of HNL1 shows changes in hydrogen bond networks that may stabilize regions of the folded structure.


Introduction
Divergent evolution creates superfamilies of enzymes, which share the same protein fold, but differ in substrate specificity or in the type of catalytic activities. The focus of this paper is understanding how evolution creates new catalytic activity during divergent evolution. This question is of interest to evolutionary biologists and also to protein engineers seeking to introduce and optimize new catalytic activities in proteins. Divergent evolution of enzymes to create new catalytic activity is thought to involve intermediate catalytically promiscuous enzymes [1,2]. Catalytic promiscuity is the ability of enzymes to catalyze additional, chemically distinct reactions besides their primary reaction [3]. Duplication of the genes for these promiscuous enzymes followed by optimization of the promiscuous catalytic activity driven by increased organismal fitness is believed to give rise to enzymes with new primary activities. Support for this notion includes the observation that differing catalytic activities within a superfamily share mechanistic features or transition states and that enzymes within a superfamily often show promiscuous activities that correspond to the primary activities of other enzymes in the superfamily [4,5]. Characterization of resurrected likely ancestral enzymes supports the hypothesis that new functions evolved from ancestors with multiple functions. Reconstructed ancestral enzymes have shown substrate promiscuity [6] and also catalytic promiscuity [7,8].
Previous work identified three molecular mechanisms that promote catalytic promiscuity in enzymes. First, catalytic promiscuity may have less to do with the enzyme and more with the two reactions being compared. The two reactions may involve similar transition states so the interactions that stabilize the primary reaction also stabilize the promiscuous reaction. In such cases, almost all enzymes catalyzing these reaction types will show promiscuity. For example, the primary function of proteases is amide hydrolysis, but almost all proteases similarly catalyze ester hydrolysis because both reactions involve similar transition states. Second, a catalytically promiscuous enzyme may change its conformation thereby temporarily creating a different enzyme active site structure with different catalytic abilities. For example, a lactonase with promiscuous phosphotriesterase activity catalyzes lactone hydrolysis via a closed con-formation, but phosphate triester hydrolysis via an open conformation [9,10]. The third mechanism for promiscuity is a larger active site with multiple possibilities for interactions between enzyme and transition state. The enzyme active site remains the same, but substrates adopt different orientations within it. For example, phosphonate monoester hydrolase catalyzes promiscuous hydrolysis of sulfate monoesters and other analogs but contains a rigid active site.
The varied substrates presumably adopt different orientations within the active site. Catalytic promiscuity correlates with larger active sites and larger polar solvent-accessible surface area [11,12]. A large active site accommodates a broader range of substrates and allows them to bind in multiple conformations, while a large polar surface allows multiple alternative electrostatic interactions that can stabilize the transition state.
A catalytically promiscuous enzyme may simultaneously use all three of these mechanisms, for example, the lactonase mentioned above also known as paraoxonase I [9]. The enzyme catalyzed hydrolysis of both lactones and phosphate triesters. The first mechanism applies since both hydrolyses have similar negatively charged intermediates and lactonase/phosphotriesterase catalytic promiscuity is common among these enzymes. The lactone hydrolysis intermediates are tetrahedral and the phosphate triester hydrolysis intermediates are pentavalent, but they are nevertheless similar. The second mechanism occurs when different active site conformations enable the two distinct reactions. Finally, the rich catalytic network within the active site enables the third mechanism by promoting multiple reaction pathways by using subsets of active site residues or using them for different roles. In the paraoxonase I case, three amino acid residues (E53, H115, D269) near the active site calcium stabilize the attack of water on the lactone, while only two of these residues (E53, D269) stabilize the attack of water on the phosphotriester.
A reconstructed ancestral hydroxynitrile lyase, HNL1, from the α/β-hydrolase-fold superfamily is the focus of this paper [8]. Most enzymes in the α/β-hydrolase-fold superfamily are esterases, which catalyze the hydrolysis of esters, but this superfamily also includes lyases, which catalyze the cleavage of hydroxynitriles and the corresponding reverse addition reaction [13]. HNL1 primarily catalyzes the cleavage of hydroxynitriles, but can also catalyze promiscuous ester hydrolysis. Both reactions involve nucleophilic attack on carbonyl compounds with tetrahedral transition states with partial negative charge on the oxygens. However, the differences in the transition states (acyl enzyme formation versus no acyl enzyme formation, hydrophobic versus polar leaving group) leads to a strong trade-off between esterase and hydroxynitrile lyase catalytic activities. Indeed, it is rare that an esterase catalyzes hydroxynitrile cleavage and vice versa. Thus, the mechanistic basis for the catalytically promiscuous esterase activity of HNL1 remains unclear.
No x-ray crystal structure of a catalytically promiscuous ancestral enzyme has previously been reported, but we expect the mechanistic basis of catalytic promiscuity to be similar to that in modern catalytically promiscuous enzymes. Here we report the x-ray crystal structure of HNL1 and compare the structures and properties of HNL1 and HbHNL (hydroxynitrile lyase from rubber tree, Hevea brasiliensis). The larger and more flexible active site of HNL1 as compared to HbHNL allows the ester substrate to adopt the alternative orientation needed for reaction. HNL1, like several other reconstructed ancestral enzymes [14][15][16][17][18], is more thermostable than its modern descendants as measured by melting temperatures. In addition, the denaturation of HNL1 likely only partially unfolds the protein, which allows it to resist the aggregation that causes irreversible inactivation.

General
Chemicals were bought from commercial suppliers and used without further purification.
Racemic mandelonitrile (Sigma-Aldrich, St. Louis, MO) was aliquoted in 10 mL portions and stored at −18 °C. Protein concentrations were determined from the absorbance at 280 nm using calculated extinction coefficients from the ProtParam computational web tool [19]. Protein gels were run using sodium dodecyl sulfate polyacrylamide gradient gels (NuPage 4−12% Bis-Tris gel from Invitrogen) using the BenchMark protein ladder (Invitrogen, 5 μL/lane) as a stan-dard. Nickel affinity chromatography resin (NiNTA, Qiagen) was used to purify HNL1 containing a 6-His tag expressed in Escherichia coli. This chromatography resin was regenerated according to Qiagen protocol. 1 H NMR spectra were run at 400 MHz in deuterochloroform.

Cloning, gene expression, and protein purification
Ancestral enzyme reconstruction Ancestral enzyme reconstruction of HNL1 [8,23] started with a Maximum Likelihood phylogenetic tree built using the software tool RAxML [24]. The tree contained 1,285 α/β hydrolase-fold enzyme sequences between 30% and 99% identical and was used to infer ancestral sequences using the same software. This clade included plant hydroxynitrile lyases (HNLs) from Hevea brasiliensis and Manihot esculenta, as well as plant esterases from Nicotiana tabacum and Rauvolfia serpentine (UniProtKB: P52704, P52705, Q6RYA0, Q9SE93 respectively). HNL1 is the reconstruction at the node that is the last common ancestor of the HNLs from Hevea brasiliensis (HbHNL), Manihot esculenta (MeHNL), and Baliospermum montanum (BmHNL). The gene for the ancestral enzyme HNL1 was synthesized by GenScript (Piscataway, NJ) and subcloned into a pET21a(+) vector at NdeI and XhoI restriction sites resulting in an upstream T7 promoter and lac operator and a C-terminal six His-tag. Fidelity of cloning and gene synthesis was confirmed by DNA sequencing of the gene (ACGT, Wheeling, IL). Genes for the modern HNLs, HbHNL and MeHNL were cloned and expressed in the same way.

Site-directed mutagenesis
The HNL1-esterase variant was created by three single-amino-acid substitutions in the gene for HNL1 added stepwise (Thr11Gly, Lys236Gly, Glu79His) in the order listed using the QuickChange II method (Agilent Technologies) according to the manufacturer's instructions.
The residues at three positions in HNL and HbHNL (121,178 and 146) were interchanged by site-directed mutagenesis stepwise in the order listed using the Q5 site-directed mutagenesis method (New England BioLabs, Inc.) according the manufacturer's instructions. S1 Table lists the primer pairs used for the polymerase chain reaction step of this site-directed mutagenesis.
A typical procedure for the Q5 method is as follows. The polymerase chain reaction (total volume 25 µL) used typically 15-20 ng template DNA, 0.5 µM each of forward and reverse primers, 12.5 µL of 2X Q5 Hot Start High-Fidelity Master Mix and nuclease free water. The temperature program was: initial denaturation at 98 °C for 2 minutes, followed by 25 cycles of denaturation at 98 °C for 10 seconds, annealing at a temperature between the melting temperatures of the forward and reverse primers for 20 seconds, and extension at 72 °C for 3.5 minutes. The final step concluded with a 10-min extension at 72 °C followed by cooling to 4 °C for storage. 1 µL-2.5 µL of PCR reaction was visualized on a 1% agarose gel to confirm the presence of linear PCR product. For the Q5 site-directed mutagenesis method the PCR product was then treated with the kinase-ligase-DpnI enzyme mix in a 10 µL reaction containing 1 µL of KLD mix, 5 µL of KLD Buffer, 1-2 µL of PCR product and 2-3 µL of nuclease free water to remove the template DNA as well as to re-circularize the linear PCR product for more efficient transformation. E. coli DH5α cells were transformed with the mutated plasmids and allowed to grow overnight at 37 °C. The plasmid DNA was isolated via miniprep (Qiagen) and sequenced prior to initiating the next mutation.
Protein expression and purification HNL1 and its variants expressed well in E. coli BL21 and purification by Ni-affinity chromatography yielded high purity protein. Lysogeny broth media containing ampicillin (100 μg/ mL, LB-amp, 5 mL) was inoculated with a single bacterial colony from an agar plate and incubated in an orbital shaker at 37 °C and 200 rpm for 15 h to create a pre-culture. A 1-L baffled flask containing terrific broth (TB)-amp media (250 mL) was inoculated with this pre-culture.
This culture was incubated at 37 °C and 250 rpm for 3-4 h until the absorbance at 600 nm

Enzyme assays
Enzyme activity was monitored at ambient temperature in triplicate for 10 min using a microplate reader. The rates were corrected for any spontaneous reaction.

Measuring catalytic promiscuity
To quantitatively compare the catalytic promiscuity of different enzymes, we define below a measure of promiscuity based on comparison of two substrates in analogy to substrate specificity. This focus on specific substrates facilitates identification of specific molecular features that favor one reaction over the other but suffers the disadvantage that some of these differences will be substrate-specific instead of catalytic-reaction specific. Other researchers have defined catalytic promiscuity indices based on the number of reactions catalyzed [27], but these measures do not account for changes in the relative rates of these reactions. Other promiscuity measures only apply to substrate promiscuity [28] or are based only on structure, not catalytic properties [29].
Selectivity of an enzyme is the ratio of its catalytic efficiency for a fast reaction over a slow reaction, eq 1. In most cases, this measure is used to compare two substrates. For example, enantioselectivity measures the ability of an enzyme to discriminate between the two enantiomers. A selectivity of 1 corresponds to a non-selective reaction, while large values correspond to high selectivity. (1) Promiscuity is the opposite of selectivity; it is the ability to efficiently catalyze multiple reactions. We define promiscuity for any pair of comparison reactions, P, as the inverse of selectivity, eq. 2.
A promiscuity of 1 corresponds to equal catalytic efficiency of both reactions. Values less than one correspond to lower promiscuity, while values more than one correspond to a switch in the primary reaction. As for selectivity, promiscuity is defined by specifying a pair of reactions. Substrate promiscuity is defined by a pair of substrates, while reaction promiscuity is defined by a pair of reactions where the substrate must be specified as well. We used the following approximation to convert differences in Tm to differences in Gibbs energies of unfolding. A comparison of pairs of homologous proteins showed that the melting temperatures, Tm, were 31.5 °C higher and the Gibbs energies of unfolding 8.7 kcal/mol higher for the proteins from thermophiles as compared to the protein from mesophiles [30]. From this comparison, we can derive a rule of thumb that stabilizing a protein by 1 kcal/mol will increase the melting temperature by ~3.6 °C.
where R is the gas constant and T is the temperature. A plot of this free energy versus urea concentration yields a straight line of the form in eq 5, where m is the slope and represents the sensitivity to denaturation by urea. ΔGunfold,H2O is the yaxis intercept and represents the free energy of unfolding in the absence of denaturant [31]. The linear model in the statistical software R [25] was used to fit the variation of ΔGunfold with urea concentration to a line.

Crystallization and x-ray data collection
The Before data collection, crystals were transferred to a cryosolution consisting of a well solution with 5% glycerol for 5-10 s and flash-cooled in liquid nitrogen. The x-ray diffraction data were collected at 100 K on beamline 23 ID-B (Eiger 16M detector) at Advanced Photon Source at Argonne National Labs, Lemont, IL.

Structure determination
The data sets were integrated with XDS [32] and scaled using Aimless from the CCP4 suite [33,34]. The structure was determined by molecular replacement with Phaser MR [35] using the coordinates of HbHNL with pdb id 1yb6 [36], having 81% sequence identity to HNL1.
Search models were derived from these coordinates by removing side chains and loop regions that differ using structure-based sequence alignment. Molecular replacement using this truncated model gave one solution that was clearly above the background, with scores: RFZ=3.3 PAK=0 LLG=5261 TFZ==29.2. Four molecules were found in the asymmetric unit. A full structural model could be built into the electron density obtained from these phases.
Phase improvement was made with REFMAC5 [37]. Model building was done with Coot [38]. Residues 2-260 were modeled in all HNL1 chains, with some also including a portion of the C-terminal linker to the 6xHis tag. The 6xHis tag and methionine 1 were unable to be mod-eled in any of the chains. The final refinement run with REFMAC5 included TLS refinement [39].
MolProbity [40] was used to analyze sterics and geometry during the refinement process.

Comparison of properties of HNL1 and modern HNLs
To test the hypothesis that ancestral enzymes are more catalytically promiscuous that their modern descendants, we previously resurrected a putative ancestral hydroxynitrile lyase, HNL1, within the α/β-hydrolase fold superfamily [8]. This ancestral enzyme separates a small group of hydroxynitrile lyases from the vast number of esterases in this superfamily. We compared the substrate specificity and catalytic promiscuity of this ancestral HNL1 to three of its modern descendants whose x-ray crystal structures have been solved: HNLs from rubber tree   Table. Catalytic promiscuity for hydroxynitrile lyases in this paper refers to their ability to catalyze ester hydrolysis. The comparison reactions are cleavage of mandelonitrile for the hydroxynitrile lyase activity and hydrolysis of pNPAc for the esterase activity. Both reactions can be measured colorimetrically. Both substrates contain one aromatic ring and have similar sizes (133 and 181 g/mol, respectively ), but neither is a natural substrate. The natural substrate for Hb-HNL is acetone cyanohydrin [44], while the natural substrate for SABP2 is methyl salicylate [45]. We define this catalytic promiscuity as the ratio of the catalytic efficiency for the hydrolysis of pNPAc over the catalytic efficiency of cleavage of mandelonitrile, eq. 7.
The two assays require slightly different pH, but are otherwise similar. Detecting the hydrolysis of pNPAc requires a pH above 7.2 so that the released the p-nitrophenol ionizes to the yellow p-nitrophenoxide. Cleavage of mandelonitrile requires a lower pH of 6.0 to minimize the spontaneous cleavage of mandelonitrile.
The ancestral enzyme HNL1 is approximately twice as promiscuous for esterase activity as compared to HbHNL, Table 1. The catalytic efficiency (kcat/KM) of HNL1 is 2.5-fold higher than for HbHNL for mandelonitrile cleavage, but is 4.5-fold higher for pNPAc hydrolysis. Thus, the promiscuity of HNL1 is two-fold higher that the promiscuity of HbHNL. In both cases the pro- miscuous activity is ≥1000-fold slower than the cleavage of mandelonitrile. The site-directed mutagenesis data in Table 1 are described later in this paper. Other support for the promiscuity of HNL1 is its ability to catalyze the hydrolysis of naphthyl acetate esters, while no activity was detected with HbHNL, S2 Urea-induced unfolding measurements confirm that HNL1 is more stable than HbHNL and suggest that unfolded HNL1 retains more folded structure than HbHNL,  S5 Fig panel B), which indicates the sensitivity of the protein to urea and depends on the solvent accessible surface area exposed upon unfolding [49], is significantly higher for Hb-HNL (0.93±0.02 kcal/mol·M) than for HNL1 (0.61±0.03 kcal/mol·M). The m-value for HbHNL is typical, but the m-value for HNL1 is unusually low. The x-ray structure below shows that the folded structures of both proteins are similar, so the lower m-value for HNL1 suggest that it exposes less solvent accessible surface area upon unfolding. The low sensitivity of HNL1 to unfolding by urea suggests that the unfolded form of HNL1 retains more folded structure than the unfolded form of HbHNL. This retention of structure may account for the ability of HNL1 to refold and avoid aggregation upon heat unfolding. Robic et al. [50] found that ribonuclease H from thermophiles retained partial structure upon unfolding, but homologs from mesophiles unfolded more completely. Thermostable reconstructed ancestral ribonucleases also retained partially folded structure upon unfolding [51]. The Gibbs energy of unfolding for HNL1 is estimated to be 0.7 kcal/mol higher than for Hb-HNL. This increased stability is consistent with the 9 °C higher melting temperature of HNL1.
When the slopes of the urea concentration versus unfolding Gibbs energy (m-values in Table 3) are similar, one can compare the extrapolated Gibbs energies of unfolding in pure water. This comparison incorrectly suggests that HNL1 is less stable than HbHNL: ΔG°H2O = 2.69±0.5 kcal/ mol for HbHNL and 2.3±0.1 for HNL1. However, when these slopes differ, as they do in this case, the unfolding processes of the two proteins differs, so this comparison is misleading and suggests that HNL1 is less stable. To compare stability when the slopes differ, Pace and Scholtz [31] recommend comparing the urea concentrations at half unfolding. To convert these to Gibbs energies, one multiplies the difference in the half-unfolding-urea concentrations (0.9 M in this case) by the average of the slopes (0.72 kcal/mol·M), which yields 0.7 kcal/mol.

X-ray structure of HNL1
We next solved the x-ray structure of HNL1 to identify structure differences that could ex-   The labels mark serine 80 and histidine 235 from the catalytic triad and the catalytically essential lysine 236.
The asymmetric unit contains a dimer of dimers. Chains A and B form one dimer, Fig 4; chains C and D form the second dimer, and the interaction of these two dimers appears to result from crystal packing. Modern hydroxynitrile lyases form dimers both in crystals and in solution [52][53][54][55][56]. For HNL1, the calculated binding between chains A and B is strong (-16.8 kcal/ mol) and statistically significant (p-value = 0.013) according to the PISA assessment of macromolecular interfaces [57]. The calculated binding between the dimers is weak and not statistically significant confirming that this interaction is the result of crystallization. Therefore, the native arrangement of HNL1 is as the dimer like the modern hydroxynitrile lyases discussed here.  Taken together, the structure and properties of HNL1 are consistent with an ancestor of modern hydroxynitrile lyases.

Substrate-binding site differences
The substrate-binding site is defined as those residues that may directly contact the substrate  Another size comparison is the distance between the catalytic Ser80 and Trp128, Table 4.
The substrate binds between these two residues, so the distance between them limits substrate size and orientation. By this distance measure, the substrate-binding site of HNL1 is also larger (8.2 Å) than that of HbHNL (6.9 Å) and MeHNL (7.2 Å). The binding sites in BmHNL and esterase SABP2 are nearly as large as HNL1. Trp128 in BmHNL is in a similar position as in HNL1 [58], despite lower overall sequence similarity (68% identity) than to HbHNL (81%) and MeHNL (77%). Like HNL1, BmHNL accepts large aromatic substrates [53], but BmHNL's natural substrate is unknown. The position of the tryptophan is also similar in the related esterase SABP2 from N. tabacum [59]. Esterase PNAE differs from SABP2 since the smaller amino acid methionine replaces Trp128. This replacement creates a substrate binding region even larger than observed in HNL1.  show increased flexibility as compared to HbHNL. The most relevant to catalytic promiscuity are three regions with increased flexibility that shape the active site. These regions are from slightly more than one to over three standard deviations more flexible than the corresponding residues in HbHNL. This higher flexibility may create alternative conformations that enable promiscuous ester hydrolysis. For HNL1, four of the incorrect orientations were sideways orientations within the pocket that likely could rotate into a correct orientation for catalysis. This ability to explore new substrate binding orientations in HNL1 due to the larger active site may contribute to its higher promiscuous esterase activity as compared to HbHNL.

Structure differences related to thermostability
The numbers of intra-protein interactions in HNL1 and HbHNL are similar and do not suggest that their stabilities would significantly differ, S4 Table. The total solvent accessible surface area of HNL1 is 1.2% larger than HbHNL. Since the structure of HNL1 contains 1.5% more amino acids residues in each monomer (261 vs. 257 for HbHNL), this difference suggests similar compactness for both structures. On one hand, HNL1 shows slightly more hydrophobic interactions: a 2.3% higher number of non-polar contacts and 1.5% less non-polar surface area at the surface. On the other hand, HbHNL shows slightly more electrostatic interactions: a 1.8% higher number of oppositely charged atoms within 7 Å of each other (111 vs 109) and a 2.1% higher number of hydrogen bonds (143 vs. 140).

Although the numbers of interactions are similar, an examination of individual interactions
shows at least one hydrogen bond network that is stronger in HNL1 than in HbHNL, Fig 7. In HbHNL, this hydrogen bond network incorporates bridging water molecules, while in HNL1, the hydrogen bond network involved direct hydrogen bonds between protein atoms. The direct interaction is expected to be more favorable since it does not require the immobilization of water molecules.

Site-directed mutagenesis
To test whether the enlarged binding site causes the increased promiscuity, we used sitedirected mutagenesis to add or remove the three residues associated with the larger binding pocket, Table 1. First, we replaced these three residues in HbHNL with the corresponding ones in HNL1 to expand its binding site (HbHNL + binding, Table 2; Leu121Phe Leu146Met Leu178Phe). This expansion in HbHNL increased the esterase activity 125 fold. This replacement also increased HNL activity 3 fold, so the promiscuity increased 50 fold. This promiscuity is 20-fold higher than for the ancestral enzyme HNL1. This large increase in promiscuity, caused by the large increase in esterase activity, is consistent with the notion that the larger active site increases promiscuous esterase activity. In a complementary experiment, we contracted the active site of HNL1 by replacing the three residues with the corresponding residues in HbHNL (HNL1 w/o binding). The esterase activity of HNL1 w/o binding decreased 15 fold as compared to HNL1 and the HNL activity decreased 2 fold. The promiscuity of HNL1 w/o binding decreased 7 fold as compared to HNL1, mainly due to the decrease in esterase activity.
This change is also consistent with the notion that reducing the size of the binding site would decrease esterase activity.
Expanding the binding site was not as effective in introducing esterase activity as the introduction of catalytic residues associated with esterase activity. In previous work, we switched the HNL activity of HbHNL to esterase activity by replacing the catalytic residues associated with HNL activity in HbHNL with the corresponding residues in esterases (Thr11Gly, Lys236Gly, Glu79His; [46]. This replacement decreased HNL catalytic efficiency 200 fold and increased esterase activity 85 fold. The major activity switched from HNL activity to esterase activity, but the catalytic efficiency was low compared to natural esterases. Here, we tested this same replacement in HNL1 (Thr11Gly Lys236Gly Glu79His = esterase catalysis), hypothesizing that the ancestral enzyme would be more malleable and result in a more efficient esterase. The HNL catalytic efficiency decreased 10-fold while the esterase catalytic efficiency increased 22 fold.
This HNL1+esterase catalysis variant was similarly effective as an esterase, but retained more HNL activity. The esterase activities of HbHNL+esterase catalysis and HNL1+esterase catalysis were ~10-fold lower than that for a related esterase (SABP2).
An attempt to increase the esterase activity of HbHNL+esterase catalysis by expanding its binding site was unsuccessful. The resulting enzyme (HbHNL+esterase catalysis+binding) showed no detectable esterase activity and a drop in HNL activity compared to HbHNL+esterase catalysis, Table 1. This lack of success indicates incompatibility between the two sets of substitutions, which individually increased esterase activity.

Discussion
The three previously identified mechanisms for catalytic promiscuity are 1) similar transition states, 2) different active site conformations, and 3) rich catalytic network within the active site.
The transitions states and mechanisms for ester hydrolysis and cyanohydrin cleavage differ significantly. There is some evidence for differences in active site conformations since the x-ray structure suggests increased flexibility of the residues in the substrate-binding site of HNL1 as compared to HbHNL. These differences are likely small local changes, not large shifts in secondary structure elements. The rich catalytic network within the active site is likely the most important for catalytic promiscuity of HNL's. The active site serine can act as a nucleophile for ester hydrolysis and or as a base for cyanohydrin cleavage, Fig 9. But exploiting this rich catalytic network also requires different orientations of the substrate. The three-fold higher catalytic promiscuity of ancestral hydroxynitrile lyase HNL1 and compared to modern HbHNL is likely due to a larger substrate binding site in HNL1 compared to HbHNL. This larger substrate-binding site allows the promiscuous substrate, pNPAc, to orient in a catalytically productive manner. The comparison substrates, mandelonitrile and pNPAc, both contain aryl groups, but the catalytically productive orientation of these aryl groups differs, Fig 8. If one divides the active site into an acyl-binding region and an alcohol-binding region, then the productive orientation of pNPAc places the aryl group in the alcohol-binding region, but the productive orientation of mandelonitrile places the aryl group in the acyl-binding region. This differing requirement for the substrate orientation creates the expectation that a large active site that could accommodate both orientations would lead to increased catalytic promiscuity. The x-ray structure shows a larger active site in HNL1 as compared to HbHNL, so this difference can contribute to the 3 fold higher promiscuity of HNL1 for ester hydrolysis. Higher catalytic promiscuity in the alkaline phosphatase family was also associated with a larger active site [11].   [62,63]. Global temperatures during this period were only ~4 °C warmer than today [64]. Thermostability may also be a byproduct of stability to other environmental stresses such as high levels of oxidative stress or radiation [65].
A second reason for high thermostability of resurrected ancestral proteins is reconstruction bias toward stabilizing substitutions [65][66][67]. Stabilizing residues are more likely to be conserved in modern descendants, and thus the identity of these residues is available to correctly infer the ancestral residue, while destabilizing residues are more likely to be lost in the descendants and therefore are more likely to be missed while inferring the ancestral sequence. As proteins diverge, destabilizing residues are often mutated to stabilizing residues to offset the effects of mutations that enhance other properties [68]. This tendency means that any destabilizing residues that were present in an ancestral protein are unlikely to be conserved in modern proteins and thus cannot be correctly predicted in a reconstructed ancestor. Stabilizing residues present in an ancestor are likely to be correctly inferred more frequently. This tendency leads to a reconstructed ancestral protein like HNL1 that may be more stable than the actual ancestor.