Glycoside Hydrolases Restrict the Side Chain Conformation of their Substrates to Gain Additional Transition State Stabilization

Abstract

Carbohydrate side chain conformation confers a significant influence on reactivity during glycosylation and anomeric bond hydrolysis due to stabilization of the oxocarbenium intermediate or oxocarbenium-like transition state. By analysis of 513 pyranoside-bound glycoside hydrolase (GH) crystal structures, we determine that most glucosidases and β-mannosidases preferentially bind their substrates in the most reactive gauche,gauche (gg) conformation, thereby maximizing stabilization of the corresponding oxocarbenium ion-like transition state during hydrolysis. α-Galactoside hydrolases mostly show a preference for the second most activating gauche,trans (gt) conformation to avoid the energy penalty that would arise from imposing the gg conformation on galacto-configured ligands. These preferences stand in stark contrast to the side chain populations observed for these sugars both in free solution and bound to non-hydrolytic proteins, where for the most part a much greater diversity of side chain conformations is observed. Analysis of sequences of GH-ligand complexes reveals that side chain restriction begins with the enzyme-substrate complex and persists through the transition state until release of the hydrolysis product, despite changes in ring conformation along the reaction coordinate. This work will inform the design of new generations of glycosidase inhibitors with restricted side chains that confer higher selectivity and/or affinity.

Graphical Abstract

Introduction

Some of the most important transformations in glycoscience involve the making or breaking of C-O bonds at the anomeric position. Thus, the glycosylation reaction, in which a glycosyl acceptor displaces a leaving group from the anomeric center with formation of a new glycosidic linkage (Scheme 1a) is at the heart of glycochemistry. In nature these transformations are accomplished by glycosyltransferase (GT) enzymes, which catalyze the displacement of a nucleotidyl diphosphate from the substrate by an acceptor alcohol (Scheme 1b) and glycoside hydrolases (GHs), or glycosidases, which cleave glycosidic bonds either by direct displacement with water as nucleophile or indirectly with a catalytic aspartate or glutamate residue as nucleophile (Schemes 1c and d).¹ GH inhibitors are exploited in a wide range of therapeutic applications ranging from antiviral activity, seen in neuraminidase and α-glucosidase inhibitors, to treatment of various genetic disorders in chaperone therapy via inhibition of lysosomal glycosidases, to diabetes mitigation with α-glucosidase and glycogen phosphorylase inhibitors.^{2, 3} Likewise, trehalase and chitinase inhibitors find application in insecticidal and fungicidal products.⁴ Given the extent of their applications, a thorough understanding of GHs is crucial and has been pursued by many groups, with the notable exception of the role of side chain conformation.

Scheme 1.

(a) General glycosylation reaction (b) Glycosyltransferase reaction (retaining or inverting) (c) Mechanism of inverting glycoside hydrolases (d) Mechanism of retaining glycoside hydrolases

In recent years considerable light has been shed on the details of both chemical and enzymatic glycosidic C-O formation and cleavage,^5–8 such that it is now evident that all four transformations presented in Scheme 1 pass through either a transient glycosyl oxocarbenium ion supported by the close proximity of a counter-ion (Figure 1a) - a carboxylate in enzymes and frequently trifluoromethanesulfonate in chemical reactions - or take place in a concerted manner via an exploded transition state that has a high degree of oxocarbenium ion character (Figure 1b).^9–11 Parallels between the chemical and enzymatic processes have been noted on several occasions and potentially cross-fertilize new developments such as those of new glycosidation reactions and new classes of glycosidase inhibitors.^12–15

Figure 1.

(a) Glycosyl oxocarbenium ion-counter ion pair (b) Concerted oxocarbenium-like transition state with an inverting glycosidase

The parallels between chemical and enzymatic glycosidic C-O formation and cleavage are not limited to gross similarities at the oxocarbenium ion level but can involve other subtle features that result in enhanced reactivity. For example, in chemical glycosylation the presence of a 2,3-O-cyclic carbonate group in the manno and rhamnopyranosyl series results in exquisite selectivity for the formation of the axial glycoside, even to the extent of overcoming the powerful equatorial directing effect of a 4,6-O-acetal.^{16, 17} This result constitutes an example of ground state destabilization whereby the fused cyclic carbonate twists imposes an approximate^{18 O}H₅ half-chair conformation on the activated donor thereby reducing the activation energy for oxocarbenium ion formation.¹⁶ In close analogy, several members of the GH92 family of α-mannosidases employ a Ca²⁺ or other bivalent metal ion to distort the substrate away from its ground state ⁴C₁ conformation and toward the transition state conformation by five-membered chelate formation with O2 and O3.¹⁹ The crucial nature of the C3-O3 bond in 4,6-O-benzylidene directed β-mannosylation^{20, 21} and the importance of the substrate O3-enzyme hydrogen bond in cleavage by a series of mannosidases provides another example of this type of subtle parallel.^22–24

As determined from the examination of crystal structure databases and NMR data, the side chains of free hexopyranose sugars populate three staggered conformations called here gauche,gauche (gg), gauche,trans (gt), and trans,gauche (tg), where the first and second descriptors refer to the stereochemical relationship of the C6-O6 bond in the side chain with the C5-O5 bond and the second C5-C4 bond in the pyranose ring, respectively. The unrestricted distribution of side chain conformations between the gg, gt, and tg conformations depends mainly on the configuration at C4 in the pyranose ring: gluco- and mannopyranose-like sugars with an equatorial hydroxy group at C4 are typically found to be ~50:50 mixtures of the gg and gt conformations with very little population of the tg conformer, while galactose-like sugars with the axial C4-OH bond are usually ~15:55:30 mixtures of the gg, gt, and tg conformers (Fig 2a).^25–27 4-Deoxyglycopyranosides adopt similar side chain populations to the glucopyranosides.²⁸ Pyranosides in the N-acetyl neuraminic acid series are nearly exclusively found in the gg conformation in free solution.²⁹

Figure 2.

(a) The three staggered side chain conformations and their approximate populations in free solution for gluco- and galactopyranoses and N-acetyl neuraminic acid (b) Spatial relationships of side chain hydroxyl groups with the oxocarbenium π* orbital (c) Relative rates of spontaneous hydrolysis of conformationally-locked dinitrophenyl glycopyranosides.

Restriction of the conformation of the pyranoside exocyclic C5-C6 bond in bicyclic structures has emerged as an important tool in the control of glycosylation reactions.^30–34 Thus, it has been established that, of the three staggered conformers of the side chain (see Figure 2a), the tg conformation in which the C6-O6 and C5-O5 bonds are antiperiplanar is strongly deactivating toward the formation of oxocarbenium ions and oxocarbenium ion-like transition states. In the gg conformation on the other hand the C6-O6 bond is gauche to the C5-O5 bond and periplanar with the vacant π* orbital of any oxocarbenium ion–like species to which it is able to provide through space electrostatic stabilization. The gt conformer also enjoys a gauche relationship between C6-O6 bond and C5-O5 bonds resulting in through space electrostatic stabilization of oxocarbenium ion like species but to a lesser extent than the gg conformer as there is little overlap of the C6-O6 bond with the π* orbital (Figure 2b). Overall, the established order of reactivity in conformationally locked bicyclic glycosyl donors with reactions proceeding via intermediate oxacarbenium ions or related transition states is gg>gt>tg as illustrated in Figure 2c.

More recently, we have demonstrated that the influence of side chain conformation on reactivity and selectivity is not restricted to conformationally-locked bicyclic systems but also applies to monocyclic glycosyl donors. Thus, for example, a pseudaminic acid donor 1 and a KDO donor 2, both of which exhibit the tg conformation of their respective side chains, were found to be more stereoselective than related pseudo-diastereomeric legionaminic acid 3 and neuraminic acid 4 donors with the gg conformation (Figure 3).^35–37

Figure 3.

Ulosonic glycosyl donor side chain conformations

With the generality of side chain stabilization of glycoside hydrolysis and glycosylation transition states in chemical systems established, we asked whether Nature employs a comparable strategy as an adjunct to transition stabilization in the course of glycoside hydrolysis by GHs. We reasoned that evidence for such conformational restriction would be found in crystallographic structures of transition state inhibitors and even substrates of GHs in complex with their respective enzymes, and report that this is overwhelmingly the case. This insight should inform the design of improved, conformationally restricted glycosidase inhibitors. Indeed, Nature again shows the way with the plant-derived GH inhibitor castanospermine 5, a bicyclic derivative of 1-deoxynojirimycin 6, whose side chain is locked in the gg conformation, which displays fourfold higher binding affinity than its monocyclic counterpart (Figure 4).³⁸

Figure 4.

Castanospermine 5 and 1-deoxynojirimycin 6

Results and Discussion

To test the hypothesis that GHs restrict substrate side chain conformation we turned to the Protein Data Base (PDB) using CAZy³⁹ to search for all known structures of hexopyranoses or hexopyranose-like ligands bound to the −1 subsite of hexopyranoside hydrolases (the site of glycosidic bond cleavage, using the nomenclature put forth by Davies).⁴⁰ GH classes known to operate by alternative mechanisms (GH 4 and 109)^41–43 not involving oxocarbenium ion-like transition states were excluded from the analysis. Likewise, any structures whose −1 subsites could not be determined were not included in the study. To minimize complications arising from errors in the crystallographic data,⁴⁴ we limited the data set to structures with resolution of ≤2 Å,⁴⁵ and inspected every structure manually to exclude structures not involving binding to the active site and other obvious errors. Additionally, wherever possible,⁴⁶ structures were examined with the Privateer software,⁴⁷ using a minimum acceptable real space correlation coefficient (RSCC) of 0.8 (see SI for details).⁴⁸ Structures containing multiple copies of the active site with bound carbohydrates were counted as a single entry if all occupied sites contained the same conformation of the ligand; when differences between these sites were observed, the entry was classed as ambiguous. We organized the 513 located structures that meet these basic criteria into two Tables according to the type of ligand: transition state or apparent transition state inhibitors¹ (Table 1) and non-transition state inhibitors and simple glycosides (Table 2). All raw data are compiled in Table S1. Tables 1 and 2 are broken down into the classes of glycosidic bond hydrolyzed by the GH and record the number of structures with the ligand bound in each of the three staggered conformations of the side chain for each class. Also recorded in Tables 1 and 2 are the number of structures in each class bearing hydrogen bonds between the side chain of the ligand and the protein, as well as the number of ligands not adopting one of the three staggered conformations.

Table 1.

Bound TS inhibitor and apparent TS inhibitor side chain conformations

Enzyme Category	gg	gt	tg	Eclipsedd	Total	H-bonds to O6
α-Galactosidase	0	0	0	4	4	4
β-Galactosidasea	1	0	2	0	4	4
α-N-Acetyl Galactosaminidase	0	2	0	0	2	2
α-Glucosidase	37	0	0	0	37	36
β-Glucosidase	35	3	0	0	38	34
β-N-Acetyl Glucosaminidasea	22	1	0	0	24	24
β-Glucosidase/Galactosidaseb	2	0	0	0	2	2
β-N-Acetyl Hexosaminidasec	2	6	0	0	8	7
α-Mannosidase	11	10	0	1	22	22
β-Mannosidase	12	0	0	0	12	10
Neuraminidase	24	0	0	0	24	8
Total	146	22	2	5	177	153

^aOne crystal structure exhibits an ambiguous side chain conformation as described above

^bEnzymes with a slash exhibit both types of activity

^cThese enzymes exhibit both glucosaminidase and galactosaminidase activity

^dThese cyclohexene-derived inhibitor side chains eclipse with the ring C=C bond

Table 2.

Side Chain Conformations of GH-Bound Non-TS Inhibitors

Enzyme Category	gg	gt	tg	Eclipsed	Total	H-bonds to O6
α-Galactosidase	0	12	0	0	12	12
β-Galactosidase	4	11	24	0	39	38
α-N-Acetyl Galactosaminidase	0	3	0	0	3	3
β-N-Acetyl Galactosaminidase	0	1	0	0	1	1
α-Glucosidase	40	1	0	0	41	38
β-Glucosidasea	96	16	0	0	118	105
α-N-Acetyl Glucosaminidase	9	0	0	0	9	9
β-N-Acetyl Glucosaminidaseb	32	15	0	0	48	48
β-Glucosidase/Galactosidase	1	0	0	0	1	1
β-N-Acetyl Hexosaminidase	2	4	0	0	6	6
α-Mannosidase	14	18	0	0	32	32
β-Mannosidase	9	0	0	0	9	5
α-Mannosidase/Glucosidase	3	0	0	0	3	3
Neuraminidase	14	0	0	0	14	6
Total	224	81	24	0	336	307

^aSix crystal structures exhibit an ambiguous side chain conformation

^bOne crystal structure exhibits an ambiguous side chain conformation

We begin by examination of the conformations of bound transition state and apparent transition state inhibitors (Table 1) as it is at this level that the GHs would be expected to derive maximum benefit from restricting the side chain to the gg conformation. Inspection of Table 1 reveals that, with the exception of the galactosidases, hexosaminidases, and α-mannosidases to which we return later, GHs predominantly bind TS and apparent TS inhibitors with their side chain in the gg-conformation by means of hydrogen bonding. The limitation of the side chains of GH bound TS and apparent TS inhibitors to the gg conformation is best appreciated by comparison with the distribution of the side chain between the gg, gt and tg conformations in free solution. In principle this distribution can be readily computed from the ³J_H5,H6R and ³J_H5,H6S coupling constants in the solution phase ¹H NMR spectra of the free ligands with the aid of limiting coupling constants for each of the three conformations,²⁰ provided that assigned NMR spectra of sufficient quality are available. Unfortunately, literature NMR spectra for many of the ligands could not be used in this manner because of the absence of assignments. This method also cannot be applied to TS inhibitors lacking a proton at C5 because of further substitution or the presence of a sp²-hybridized carbon at that position. Nevertheless, we located partially assigned ¹H NMR spectra for 28 ligands and computed the distribution of their side chain conformations using the most recent⁴⁹ limiting coupling constants. Because the diastereotopic H6pro-R and H6pro-S resonances were not distinguished in these spectra, two population distributions are possible in each case. Table S2 and Figure S1 record the two computed side chain distributions analyzed in this manner, together with the conformation when bound to each particular GH.

In some cases, very clear differences between bound side chain conformation and solution phase ratios were observed. For example, the side chain of D-galacto-1,5-hydroximinolactam (7), a true transition state inhibitor according to LFER studies carried out by Withers and coworkers⁵⁰ is restricted to the gg conformation when bound to the two GHs it inhibits, but is only 11% or 14% gg in solution (Scheme 2).

Scheme 2.

Restriction of side chain conformation of hydroximinolactam 7 by GH1 enzymes

The glucoimidazole series (Scheme 3) is another example of enzymatic side chain restriction. The potent glucoimidazoles 8-13 are bound by GHs exclusively in the gg conformation, but are only 71–78% gg in solution (Scheme 3). Conversely, the less potent glucoimidazole 14, lacking the extra substitution on the imidazole ring, is observed in the gt conformation in two crystal structures, despite being 86% gg in free solution. In the crystal structure of glucoimidazole bound to D. melanogaster golgi α-mannosidase II (PDB ID 3D4Z), the observed gt conformation is consistent with the side chain preferences seen in GH family 38 mannosidases, to which we return below. Likewise, in the case of T. xylolyticum GH 116 β-glucosidase (5BX4), the unexpected gt conformation may be the result of the catalytic aspartic acid being positioned above the ring rather than syn or anti to the glycosidic oxygen, as is more typically the case.^{51, 52}

Scheme 3.

Enzymatic restriction of glucoimidazole side chain conformations

Similar observations were made with ligands that, while mimicking the charges observed in the transition state, are not true TS inhibitors because of the absence of sp²-hybrized atoms approximating the conformation of an oxocarbenium ion. Thus, the side chains of isofagomine 15 and its glucosyl derivatives 16 and 20 are calculated to be 37–47% gg in solution, but are held along with derivatives 17–19 in the gg conformation by hydrogen bonding in all but one of the 15 bound structures (Scheme 4). The isofagomine ring in the gt conformation is bound as a ring-flipped ¹C₄ inverse chair, wherein the gg conformation would unfavorably place the C6-OH directly above the ring.

Scheme 4.

Enzymatic restriction of isofagomine side chain conformations

Overall, analysis of the 28 examples for which both crystal structures of bound TS and apparent TS inhibitors and sufficiently detailed NMR spectra of the free ligands are available supports the hypothesis that GHs use hydrogen bonding to limit the conformation of the substrate side chain predominantly to the gg conformation. Kifunensine (21), though not a TS inhibitor, is an interesting case. It takes up an inverted chair conformation in free solution,^{53, 54} which places the side chain in a pseudoaxial position. This considerably destabilizes the gg conformation, which would place the side chain over the center of the pyranose ring. Consequently, it is an approximately 75:25 mixture of the gt and tg conformations in solution (Scheme 5). In two of the three bound structures (PDB ID 1FO3 and 5NE5), kifunensine retains this inverted chair conformation and the side chain is restricted to the gt conformation. In the third structure (PDB ID 1PS3), it is bound as a related ¹B₄ boat conformation, for which the gg conformer would also place O6 over the pyranose ring. The preference for the gt conformation in bound structures of kifunensine is consequently the result of steric destabilization of the gg conformation. The inherent preference for the gt-conformation in the free ligand complements the apparent preference of the α-mannosidase families it inhibits and that bind their substrates exclusively in the gt conformation as discussed below.

Scheme 5.

Bound and unbound side chain conformations of kifunensine

Seminal crystallographic studies by Davies and coworkers, supported by computational analyses by the Rovira laboratory, have revealed the conformational itineraries followed by the pyranose ring of several GH substrates through the sequence of binding, traversal of the transition state, and exit.⁵⁵ Inspection of these studies suggested that restriction of the side chain to the gg conformation is not limited to the transition state but begins on initial binding and continues until release, despite the considerable changes in the conformation of the pyranoside ring itself throughout the course of the hydrolysis: the B. xylanisolvens GH99 endo-α-mannanase serves to illustrate this point (Figure 5).⁵⁵

Figure 5.

X-ray crystal structures of B. xylanisolvens GH99 endo-α-mannanase with (a) a tetrasaccharide substrate analog (PDB ID 6FWG), (b) a disaccharide transition state analog (PDB ID 6FWJ), and (c) a disaccharide hydrolysis product (PDB ID 6FWP)

X-ray studies on termite GH1 β-glucosidase by Wang and coworkers further illustrate the notion that restriction of side chain conformation begins at the level of the enzyme-substrate complex, and continues through the transition state up to the release of the hydrolyzed material.⁵⁶ In this sequence focused on the termite GH1 β-glucosidase the gg conformation is observed with a cellobiose substrate analog, with a deoxynojirimycin transition state mimic, and with a Hepes-glucose conjugate generated by the enzyme in situ, in spite of the evident changes in ring conformation (Figure 6). The observation that the restriction of side chain conformation begins with substrate binding is consistent with the observation by Namchuk and Withers⁵⁷ in their study of the Agrobacterium faecalis β-glucosidase that H-bonding interactions with O6 are heavily involved with substrate binding, unlike those with O3 and O4 which mainly impact the transition state. It appears that many GH’s have evolved to pre-organize the enzyme-ligand complex by hydrogen bonding to bind the ligand side chain in a conformation that affords additional enthalpic stabilization to the transition state complex, thereby primarily influencing k_cat/K_m: the activation enthalpy is effectively reduced by both ground state destabilization and transition state stabilization.

Figure 6.

X-ray crystal structures of termite GH1 β-glucosidase with (a) cellobiose (PDB ID 3VIK), (b) 1-deoxynojirimycin (PDB ID 3VIG), and (c) a glucose-Hepes conjugate (PDB ID 3VIO)

Broader inspection of the universe of GH bound non-TS inhibitors and even substrates and or substrate mimics (Table 2) supports the idea that the gg conformation is generally favored by GHs. As these ligands more closely approximate standard pyranosides, or are actual pyranosides, detailed inspection of the side chain conformations in solution is not strictly necessary and the well-established solution-phase side population distributions for pyranosides (Figure 2) suffice. Taking β-glucopyranose 22 as an example, it is an approximately 50:50 gg:gt mixture in solution, but is bound in the gg conformation in 13 of the 14 examples (93%) located in our search (Scheme 6).

Scheme 6.

Enzymatic restriction of the β-D-glucopyranose side chain conformation

When the ensemble of Tables 1 and 2 is considered (Table 3), it is evident that in the universe of glucosidases and β-mannosidases there is a very clear preference for ligands and/or substrates to be bound with the gg conformation: 84% for the β-glucosidases, 75% for β-glucosaminidases, 99% for α-glucosidases, and 100% for α-glucosaminidases and β-mannosidases. Neuraminidases similarly bind ligands and substrates exclusively in the gg conformation; however, given the high preference for the gg conformation of neuraminic acids in free solution and in X-ray crystal structures, the neuraminidases can be considered to better reflect lectins for higher carbon sugars which have evolved to bind their ligands with the ground state conformation of the side chain.⁴⁵

Table 3.

Combined Side Chain Conformations of Bound Ligands

Enzyme Category	gg	gt	tg	Eclipsedd	Total	H-bonds to O6
α-Galactosidase	0	12	0	4	16	16
β-Galactosidasea	5	11	26	0	43	42
α-N-Acetyl Galactosaminidase	0	5	0	0	5	5
β-N-Acetyl Galactosaminidase	0	1	0	0	1	1
α-Glucosidase	77	1	0	0	78	74
β-Glucosidaseb	131	19	0	0	156	139
α-N-Acetyl Glucosaminidase	9	0	0	0	9	9
β-N-Acetyl Glucosaminidasec	54	16	0	0	72	72
β-Glucosidase/Galactosidase	3	0	0	0	3	3
β-N-Acetyl Hexosaminidase	4	10	0	0	14	13
α-Mannosidase	25	28	0	1	54	54
β-Mannosidase	21	0	0	0	21	15
α-Mannosidase/Glucosidase	3	0	0	0	3	3
Neuraminidase	38	0	0	0	38	14
Total	370	103	26	5	513	460

^aOne crystal structure exhibits an ambiguous side chain conformation

^bSix crystal structures exhibit an ambiguous side chain conformation

^cTwo crystal structures exhibit an ambiguous side chain conformation

^dIn these cyclohexene-derived inhibitors the side chains eclipse the ring C=C bond

Turning to the exceptions to the rule already apparent in Table 1 and continuing through Table 2, these belong to four classes: the α-mannosidases, the α- and β-galactosidases, and the 2-acetamido-2-deoxy-aminidases, be they α- or β-galactosaminidases or the more broadly accommodating hexosaminidases. It is informative here to subdivide the structures according to CAZy families, with Table S3 showing the breakdown of preferred side chain conformations for each family and Table S4 showing the families encompassed by each category of enzyme. Thus, for the α-mannosidases all 25 GHs binding the ligand with the gg conformation of the side chain belong to families 63, 92, 99, and 125, while the 28 examples preferring the gt conformation belong to families 38 and 47. Thus, the overall 25:28 gg:gt distribution, which at first sight appears to simply reflect the solution phase distribution for hexopyranosides with an equatorial C-O bond at the 4-position, is the result of GH class specific changes. Perhaps separate courses of evolution have resulted in distinctive preferences between these families for either of the two predominant conformers of mannose in solution. Likewise, two families of β-glucosidases, 81 and 116, restrict their substrates in the less reactive gt conformation, accounting for 12 of the 19 total gt-bound β-glucosidase crystal structures. The α-galactosidases enforce the gt conformation in 12 of the 16 examples located; the remaining four are acarbose-type cyclohexene rings, in which the side chain is bound in a gt-like conformation that eclipses the C6-O6 bond with the alkene maximizing the H-bond with the enzyme despite the significant allylic strain engendered. However, each eclipsed structure can be considered as akin to a distorted gt conformation (Figure S2). These results clearly suggest that these GHs benefit from binding the side chain of their substrates in a conformation that stabilizes the transition state for hydrolysis; albeit the gt conformation only provides intermediate stabilization. Apparently, the energy penalty for binding the substrate side chain in the higher energy gg conformation is not sufficiently offset by the extra transition state stabilization this would afford.

Like the α-mannosidases, the β-galactosidases are best viewed when broken down into subgroups. The few (5) β-galactosidases observed with the ligand side chain in the gg conformation belong to GH families 16 and 43, whereas the 11 β-galactosidases binding the ligand with its side chain in the second most TS stabilizing conformation, and the one most populated in free solution, belong to GH families 42, 59, and 98. The majority (26) of β-galactosidases in the data set bind their ligands with the side chain held in the tg conformation and belong to CAZy families 2, 35, and 86. Though it is possible that these galactosidases have evolved to bind one of the two conformations prevalent in solution, with a larger number of solved structures favoring the tg conformation, it remains tempting to suggest that these GHs have evolved to bind the deactivating tg conformation as means of control in the hydrolysis of what is one of the more reactive classes of pyranosidic bond, thereby also increasing specificity of the reaction. The N-acetyl galactosaminidases, whether α- or β-, show a clear preference for binding ligands with the second most activating gt conformation of the side chain, again avoiding the energy penalty that would be required to bind the more activating but higher energy gg conformer. The more broadly accommodating β-hexosaminidases also share this preference, though not to the same degree (71%).

Finally, to address the alternative hypothesis that proteins have generally evolved to bind carbohydrate ligands primarily in the gg conformation, we analyzed an unbiased set of previously validated crystal structures of hexopyranosides bound to carbohydrate-binding proteins assembled by Kiessling and coworkers in a study of carbohydrate-aromatic interactions (Table S5).⁴⁸ Additionally, as an internal control, we examined the side chain conformations of hexose units at other subsites in the GH data. In the latter case, Privateer was used to screen and validate the data, which are compiled in Table S1, and summarized in Table 4 along with the GH −1 subsite set for comparison.

Table 4.

Comparison of −1 Subsite-, Alternate Subsite-, and Lectin-Bound Staggered Side Chain Conformations^a

Sugar	Position	gg	gt	tg
α-Galactosides	GH −1 Subsite	-	12	-
	Lectin	3	10	-
	GH Alt Sites	-	2	8
β-Galactosides	GH −1 Subsite	5	11	26
	Lectin	4	27	-
	GH Alt Sites	2	17	4
α-N-Acetyl galactosaminides	GH −1 Subsite	-	5	-
	Lectin	3	5	-
	GH Alt Sites	-	7	-
β-N-Acetyl galactosaminides	GH −1 Subsite	-	1	-
	Lectin	1	8	1
	GH Alt Sites	-	-	-
α-Glucosides	GH −1 Subsite	77	1	-
	Lectin	7	8	-
	GH Alt Sites	48	46	2
β-Glucosides	GH −1 Subsite	131	19	-
	Lectin	8	6	1
	GH Alt Sites	92	73	6
β-N-Acetyl glucosaminides	GH −1 Subsite	62	16	-
	Lectin	12	8	-
	GH Alt Sites	22	41	-
α-Mannosides	GH −1 Subsite	25	28	-
	Lectin	18	5	-
	GH Alt Sites	6	23	-
β-Mannosides	GH −1 Subsite	21	-	-
	Lectin	3	4	-
	GH Alt Sites	10	7	-

^a)Data organized by natural substrate in the GH −1 case and by the bound ligand in the lectin and GH alternate site cases. Bold text denotes significantly higher selectivity in the GH −1 site compared to the lectin, and GH alternate binding sites.

Overall, while Table 4 clearly shows that different classes of lectin, and even of GH alternate subsites, bind different types of pyranosides with different and in some cases clear preferences for particular side chain conformation, these preferences do not on the whole match those seen at the GH active sites. For example, Table 4 clearly shows that the distinct preference for the gg conformation among the various glucosidases and β-mannosidases at the GH −1 subsites is not paralleled at other sites in the GH’s or in the lectins, which interestingly trend towards the corresponding free solution conformation ratios (with the exception of GH alternate subsites on β-N-acetyl glucosaminidases, which trend towards the gt conformation). In the α-galactose series, there is a definite preference among the lectins for the gt conformation, but not to the extent seen at the GH −1 site. While lectins express a distinct preference for the gg conformation of α-mannosides and the GH alternate subsites predominantly favor the gt conformation, neither correspond to the population distribution seen at the GH −1 sites.

Conclusion

Glucosidases and β-mannosidases all show a clear preference for binding their ligands with the side chain held in the gg conformation that far exceeds the equilibrium population of this conformation in free solution. We suggest that this preference has evolved to permit these GHs to pre-organize the enzyme-ligand complex in order to benefit from the additional enthalpic transition state stabilization the gg conformation affords. The neuraminidases also show a very high preference for the gg conformation of the ligand side chain, but as this is also by far the most common conformation in solution, no conclusions can be drawn. α-Galactosidases and α- and β-N-acetyl galactosidases all show a strong preference for the moderately transition state stabilizing gt conformation that exceeds its equilibrium solution phase population: only a limited number of β-galactosidases enforce the higher energy gg conformation of the side chain of the bound ligand, presumably because the energy penalty this imposes outweighs any stabilization of the transition state it would afford. α-Mannosidases exhibit a family-dependent restriction of the side chain conformation: with families 38 and 47 enforcing the gt and all others the gg conformation. The −1 subsite-bound side chain populations show distinctly higher biases compared to populations observed in lectins and non-hydrolytic sites (both of which predominantly approach the free solution populations), indicating that transition state stabilization is likely the driving force behind the −1 side chain restrictions. Armed with this knowledge, it should clearly be possible to design new generations of GH inhibitors that build on this knowledge base by restricting side chain conformation in such a way as to gain increased affinity and/or selectivity. Viewed from another perspective, this new insight should aid in the design and evolution of new and improved glycosidases for the degradation of biomass. We anticipate that similar patterns may emerge for the glycosyltransferases and furanosidases and intend to report on them in due course.