Itineraries of enzymatically and non-enzymatically catalyzed substitutions at O -glycopyranosidic bonds

Several crystal structures of glycoside hydrolases in complex with a substrate analog, or of inactive mutants complexed with a substrate, reveal non-ground state carbohydrate conformations within subsite –1 of the active site. These “frozen” local minima as preferred by the enzyme represent pre-transition-state situations along the reaction itinerary. Substantiated by theoretical considerations, this leads to the proposal that substitutions on β -equatorial as well as α -axial D - O -glycopyranosidic bonds may always follow an itinerary that is predetermined by the original configuration at the anomeric center, where the formation of a transition state that is similar to a half-chair is always preceded by a conformational change away from the ground state.


Introduction
In Nature, many different carbohydrate monomers, oligomers and polymers play important structural or functional roles in a vast array of biological processes.O-glycoside hydrolases (EC 3.2.1.x)as carbohydrate-modifying or -degrading enzymes represent a structurally diverse group, where primary amino acid sequence analysis allows classification into more than 100 different families; [1][2][3][4] many being grouped into one of fourteen fold-related clans. 5Most of these enzymes need neither metals nor cofactors, yet achieve k cat values that are in the order of 1-100 s -1 , a truly stunning rate enhancement in comparison to k uncat values of 10 -15 s -1 . 6[9][10][11] With almost no exceptions, O-glycoside hydrolases possess at least two mechanistically important carboxyl functions at the junction of subsites -1/+1: 1) a semi-lateral syn-or antipositioned 12 proton donor close to the glycosidic oxygen; and 2) a nucleophilic carboxylate (retaining hydrolases) or a similar residue assisting solvent nucleophilicity (inverting hydrolases).The general but undetailed catalytic mechanisms of β-D-glycopyranoside hydrolysis by retaining (double displacement) and inverting (single displacement) enzymes have been proposed by Koshland as early as 1953 13 and are presented in Figure 1.

Figure 1.
Catalytic mechanisms of β-glycoside hydrolases according to Koshland. 13rrent understanding of glycosidase mechanisms has been reviewed thoroughly, [14][15][16][17] and recently three new general concepts have been put forward: 1) the syn/anti proton donor positioning; 12 2) the presence of a hydrophobic platform as a common feature within the -1 subsite; 18 and 3) enzymes with syn-positioned proton donors perform electrostatic transition state (TS) stabilization (known to be the most important factor for enzymatic catalysis 19,20 ) towards the substrate's ring oxygen by means of the conjugate base of their proton donor, whereas antiprotonating enzymes provide a separate electron-rich residue for this purpose. 21he considerations in this work pertain to pyranoside glycosidic bond substitutions that operate by a classical exocyclic mechanism on oxyalkyl-type leaving groups, thus not to: 1) the endocyclic mechanism 22 which has not yet been observed with glycoside hydrolases; 2) the SN isubstitution mechanism that occurs in glycosyl fluoride solvolysis 23 and in-situ anomerisations with good leaving groups, 24 and has been proposed to be operative in retaining glycosyltransferases; 25 3) catalysis by GH family 4 (GH4) glycosidases that operate by a redox and elimination mechanism; 26 and 4) GH families 18, 20 (Clan K) 27,28 and 84 29 that utilize anchimeric assistance from the substrates' N-acetyl group, resulting in a strained oxazoline-type glycosyl intermediate (chitinases/chitobiases).Based on its orientation relative to a reference atom in the Fisher projection (IUPAC nomenclature), an α/β glycosidic bond often -but not always -corresponds with an axial/equatorial substituent orientation in the ground state conformation.For the rational development of the present discussion, an α-axial or β-equatorial O-glycosidic bond of a 4 C 1 ground state D-glycopyranoside will be assumed throughout; the IUPAC-idiosyncratic sialosyl glycosides and some L-glycopyranosides are dealt with in separate paragraphs.

Inconsistencies within the current understanding of itineraries for glycopyranosidic bond substitutions
A classical glycopyranosidic bond substitution proceeds via a transient 30 oxocarbenium-like TS, with an sp 2 -hybridised geometry at both C1 and O5, allowing a considerable mutual double bond character and resulting in C5-O5-C1-C2 coplanarity.This process does not involve a discrete carbocation in a first-order reaction, but is borderline SN 1 -SN 2 .Recently published primary 13 C 31,32 as well as secondary α-deuterium 33 kinetic isotope effects indeed are consistent with SN 2type pathways.The four-atom coplanarity at the TS is obtained in four conformational situations: 1) the 4 H 3 -and its flanking 4 E and E 3 conformers; 2) the 3 H 4 -and its flanking 3 E and E 4 conformers; 3) the B 2,5 conformer; and 4) the 2,5 B conformer (Figure 2).The route from the reactant via the transient TS into the product is called the substitution itinerary, which has been lively debated up to now.

The ALPH itinerary for β-glycopyranoside substitutions
5][36] The reaction then proceeds through an ALPH-compliant β-skew → 4 H 3 -like TS → α-4 C 1 itinerary (Figure 3 from left to right).The relevance of ALPH for glycosidic bond substitutions has been controversial for a long time, but much of the criticism has ceased after a seminal article by P. Deslongchamps. 37Some objections on the general validity of ALPH have since been formulated, e.g. on a conformationally restricted 2-methoxy-1,3-dioxane 38 and on cyclic amidinium substrates, 39 but they have been re-analyzed in a recent extensive review by S. Chandrasekhar as not to invalidate ALPH. 40

The reverse ALPH itinerary for α-glycopyranoside substitutions
In the case of α-axial glycopyranosidic bond substitutions, the leaving group already has an ALPH-compliant set-up when the carbohydrate ring is in the ground state chair conformation.The Principle of Microscopic Reversibility (PMR) has often been invoked to suggest that here the itinerary is the reverse of that for a β-equatorial glycopyranosidic bond substitution, i.e. an ALPH-α-4 C 1 → 4 H 3 -like TS → β-skew pathway (Figure 3, from right to left).That "α-glycosides must hydrolyze via their ground state conformation" has even been explicitly stated by Deslongchamps. 34However, the PMR applies to phenomena occurring within the forward versus reverse path of a single process. 41In other words, it can be invoked for the reverse path of a given glycosidic bond substitution (e.g. on a given β-glucoside), but it cannot be used to predict the reaction behavior of another glycoside (e.g. a given α-mannoside) that has its own PMR for its reverse path.In fact, this generalizing but subtle principle has often been misapplied for deducing or disproving the mechanism of a reaction, and this has been extensively reviewed. 42If the PMR would be a mechanistic axiom for glycopyranosidic bond substitutions, then the substitution of an α-axial glycopyranoside would always progress through the ALPH-α-4 C 1 → 4 H 3 -like TS → β-skew pathway, but this would rule out a 3 H 4 -like TS, which occurs in the case of α-inverting GH47 enzymes (see further).Thus, the reverse Deslongchamps pathwayalthough allowed by the PMR -is not always operative.Later in this work we will propose that conformationally unrestricted α-glycosides may very well prefer an alternative substitution route that has a lower activation energy toward its 3 H 4 -like TS (and that has its own PMR), instead of the reverse Deslongchamps pathway with its 4 H 3 -like TS.

Itineraries involving a boat-TS
A possible alternative is a skew → boat-TS → skew glycopyranoside substitution pathway.This route implies that ALPH is not operative, but instead that the ring oxygen's free electron pair positioned cis versus the leaving group participates in the partial double bond towards C1 at the transient boat TS (Synperiplanar Lone Pair Hypothesis, SLPH) (Figure 4). 43Such a substitution is electronically equivalent to a syn-elimination, as has been suggested previously. 16,17However, syn-eliminations on a cyclohexane ring are 7000 times slower than anti-eliminations. 44On the other hand, synperiplanar effects have recently been shown to be operative in acid hydrolyses of conformationally restricted acetals. 45,46he synthesis of isopropyl and p-nitrophenyl α-and β-D-glucopyranosides 1-4 that are locked into the B 2,5 via a C2-C5 oxymethylene bridge has recently been published (Figure 5). 47heir acid-catalyzed hydrolysis is obligatorily SLPH-compliant and proceeds at remarkably similar rates to those of the corresponding unconstrained glucosides, 48 which indicates that such an itinerary may be used by enzymes as well.However, the B 2,5 is a relatively high-energy conformation due to an unfavorable C2-C5 flagpole positioning 49 as well as eclipsed substituents at the C3-C4 bond.The locked-boat compounds do not have a flagpole problem, but the authors caution that at least part of the observed reactivity must derive from a higher ground state energy in these strained bicyclo-[2,2,2] systems.The cellobiose analog 5 (with the B 2,5 -locked glucoside at the nonreducing end) was also synthesized and was assayed against glycoside hydrolases from 11 different families.It is only a very weak inhibitor for two β-retaining glycanases (from families 3 and 7), and its catalyzed hydrolysis was never observed. 48But inspection of the known crystal structures of complexed hexopyranoside hydrolases invariably shows the crucial -1 subsite to be remarkably spacious, so the unnatural oxymethylene bridge should be easily accommodated.Moreover, the C2-hydroxyl group is now replaced by an ether bridge, which removes a possible hydrogen bond interaction with the enzyme.This can only influence K ass and be of relatively minor importance for TS stabilization, which is mainly achieved through electrostatic compensation by the enzyme of the change in charges 19,20 occurring at the O5 ring oxygen. 21Thus, if a B 2,5 -TS belongs to the substitution itinerary within the tested β-glycoside hydrolases (from families 1, 3, 6, 7 and 18), one should expect the locked-boat disaccharide 5 to be a substrate or a relatively good inhibitor -which is never the case.Boat conformations have nevertheless been proposed as transition states within the mechanism of several glycoside hydrolases on the basis of observed conformers within the liganded complexes, most notably within GH26 50 and 11. 51 This has been recently reviewed. 16,17urther in this work we will show that each of those conformers can just as well be interpreted as indicators of 4 H 3 or 3 H 4 -like transition states.

A reinterpretation of kinetic isotope effects on aryl-α-D-glucopyranosides
In an effort to disprove the ALPH hypothesis, Hosie and Sinnott undertook an extensive study on primary and secondary deuterium isotope effects for the hydrolysis of aryl-α-D-glucopyranosides catalyzed by a S. cerevisiae retaining α-glucosidase. 52They came to the following conclusions: (1) After formation of the initial enzyme-substrate complex, a kinetically discrete noncovalent event must precede the actual bond-breaking process, which indeed excludes the reverse Deslongchamps-type ALPH-α-4 C 1 → 4 H 3 -like TS → β-skew pathway (Figure 3, from right to left), since it comprises a direct ground state → TS conversion.
(2) At the transient TS, the C2-H bond must be near-perpendicular to the planar oxocarbeniumtype system.This was interpreted as to be only compatible with the occurrence of the 2,5 B-TS, in spite of its severe C2-H to C5-hydroxymethyl flagpole interaction as well as the 1,3-syn-diaxial positioning of the C5-hydroxymethyl group versus the incoming nucleophile.Although a passage through the 3 H 4 -TS was not considered as an alternative possibility, their experimental results perfectly agree with this: the preceding discrete non-covalent event would then be a conformational change of the ground state α-4 C 1 into an ALPH-compliant α- whereas a near-perpendicular C2-H bond versus the C1-O5 bond is present in the 3 S 1 pre-TS as well as in the structurally nearby 3 H 4 -TS conformation (Figure 6).

The α-inverting GH47 enzymes use an alternative ALPH-compliant substitution itinerary comprising a 3 H 4 -like TS
Several all-α-4 C 1 oligosaccharide Michaelis complexes have been observed in structures of αglycoside hydrolases, e.g. the GH13 Bacillus circulans cyclodextrin glycosyltransferase inactive mutant in complex with maltononaose (pdb entry 1cxk), 53 and these appear to indicate that a reverse Deslongchamps-type ALPH-α-4 C 1 → 4 H 3 -like TS → β-skew pathway is operative.Deduction of mechanism itineraries based on complexed enzyme structures should however be conducted cautiously, since these will always contain a local energy minimum, which not necessarily represents the productive substrate conformation.This is a fortiori so for the 4 C 1 ground state, which by its own nature will have a tendency to be a local minimum within the -1 subsite of glycoside hydrolases.
A recently published crystal structure of the human GH47 Class 1 α-1,2 mannosidase in complex with a thiodisaccharide substrate analog (pdb entry 1x9d) reveals the α-3 S 1 conformation in the -1 subsite which, taking into account the close shape similarity, suggests a 3 H 4 -like TS. 54 Furthermore, a detailed docking study on the analogous Saccharomyces cerevisiae enzyme with its disaccharide substrate in different glycon conformations indicates that the binding pathway passes through the α-3 S 1 just before reaching the TS, which is likely the 3 E. 55 Clearly, this enzyme family uses an itinerary towards the TS that does not proceed via a direct substitution on the ALPH-compliant α-4 C 1 , but that involves a prior conformational reorganization into the α-3 S 1 , which also is ALPH-compliant.The authors of the 3D structure have realized that the initial conformation of the end-product very likely is the β-1 C 4 inverted chair, since when the Principle of Least Motion (PLM) 56 is applied to the conversion of the 3 S 1 into the 3 H 4 (or into the closely similar 3 E), only a slight atomic displacement further in the same overall direction leads to a collapse into the 1 C 4 .Corroborating this, 54 kifunensine as well as 1deoxymannojirimycin, both very potent aza-inhibitors of GH47 enzymes, adopt the 1 C 4 within the -1 subsite of the human α-1,2 mannosidase, even though the 4 C 1 is the ground state of 1deoxymannojirimycin. 57 In summary, the conformational itinerary for the GH47-catalyzed hydrolysis of the α-1,2-D-mannosidic bond apparently proceeds through an alternative ALPH-compliant route (Figure 7): ground state α-4 C 1 → enter the skew-boat pseudorotational series and then adopt an ALPHcompliant conformation → α-3 S 1 (or closely similar; now the substitution can start) → 3 H 4 -TS (or closely similar such as the 3 E) → β-1 C 4 (end of substitution).Only thereafter will the product return to the ground state β-4 C 1 .

Figure 7.
The alternative ALPH-type substitution itinerary for α-D-glycopyranosides.The O5 lone pair that is trans to the leaving group is indicated in black.A preceding conformational change into an ALPH-compliant 3 S 1 -like pre-TS is followed by a transient 3 H 4 -like TS and initially leads to the β-1 C 4 product.The latter converts to the ground state β-4 C 1 in an independent process.L = leaving group; Nu = nucleophile.

Towards a mechanistic simplification of O-glycopyranoside substitution itineraries
One could argue that a direct substitution on a ground state 4 C 1 α-D-mannopyranoside is severely disfavored because of steric hindrance from the axial C2-hydroxyl group towards the incoming nucleophile, and therefore the alternative ALPH trajectory (Figure 7) is exceptionally preferred over the reverse Deslongchamps-type α-4 C 1 → 4 H 3 -like TS → β-skew pathway (Figure 3, from right to left).However, the itinerary used by GH47 passes the same 3 S 1 pre-TS and subsequent 3 H 4 -TS conformations as those that can be deduced for aryl-α-D-glucopyranosides when reinterpreting the isotope effect data from Hosie and Sinnott 52 (Figure 6), which is even more remarkable since glucopyranosides have no steric clash between their ground state equatorial C2-OH and the axial nucleophile.This hints to the intriguing possibility that many, if not all, classical α-glycopyranosidic bond substitutions may follow the alternative ALPH-compliant αskew → 3 H 4 -like TS → β-1 C 4 itinerary (Figure 7).And concomitantly, since a substitution itinerary that passes through a SLPH-compliant boat TS is unlikely for conformationally unrestricted compounds (as outlined above), all classical β-glycopyranosidic bond substitutions may follow the Deslongchamps-type ALPH-β-skew → 4 H 3 -like TS → α-4 C 1 itinerary (Figure 3).The α-and β-O-glycopyranoside substitution itineraries may then have a remarkable mechanistic similarity: they may both be always preceded by a conformational change into an ALPHcompliant skew conformation, pass through a half-chair TS, and end in a chair conformation.

ISSN 1424-6376
Page 99 © ARKAT And glycoside hydrolases that operate with the classical exocyclic mechanism might also exclusively use these two itineraries, and this suggests a reassessment of the currently known complexed structures, which is the further focus of this work.But first the most important question has to be addressed: could there be an intrinsic advantage for these two itineraries?
5. Basic principles in organic chemistry yield arguments for two inherently preferred substitution itineraries:

Deslongchamps-type for βglycopyranosides, alternative ALPH-type for α-glycopyranosides
The above question can only be addressed by focusing on the TS, since it is the passage through the lowest-energy TS that will dictate the preferred itinerary and its necessary conformational changes, not the other way round (S.Chandrasekhar, personal communication).So, why may a skew → half-chair TS → chair substitution be the pathway with the lowest activation energy towards the TS, for both classical β-and α-glycopyranoside substitutions?
For most D-glycopyranosides, the 4 C 1 is the ground state and the 1 C 4 is the second-lowest energy conformation, 58 since only in a chair do all ring substituents occupy staggered positions.Then a combination of the Bell-Evans-Polanyi (BEP) principle [59][60][61][62][63] with the Hammond postulate [64][65][66] suggests an inherent lowest TS pathway.This is shown in the simplified overall energy profile in Figure 8, idealized for a β-glycopyranoside substitution with an approximately 50/50 equilibrium of reactants and products (the analysis is analogous for reactions with an equilibrium more to the right of the reaction coordinate).Since this is a comparison of two possible itineraries for a carbon substitution on the same molecule (thus with the same reacting substituents), one expects a normal Hammond behavior for the respective TS positioning on the reaction coordinate. 63If a β-glycopyranoside substitution would proceed via a SLPH-compliant β-skew → boat TS → α-skew itinerary, the energy profile would be roughly symmetrical (dashed line), since skews are often of comparable steric energy.In contrast, an ALPH-compliant β-skew → halfchair TS → α-chair itinerary (full line), ending in the ground state conformation, which is of considerably lower energy than any skew (up to 6 kcal/mol depending on the sugar type 58 ), would result in a TS that is more reactant-like due to the Hammond postulate (the TS shifts to the left in the reaction coordinate).But more importantly, the BEP principle predicts that the energy of activation ∆E ‡ (S→C) for a skew → half-chair TS → chair itinerary should be lower compared to ∆E ‡ (S→S) in the related skew → boat TS → skew itinerary.The energetic advantage ∆∆E ‡ from a BEP-Hammond effect cannot be more than a few kcal/mol.6][47][48] The dual effect (advantage half-chair TS/disadvantage boat TS) may then become substantial in comparison to the ±30 kcal/mol estimate for the enthalpy of activation of spontaneous glycoside hydrolysis. 6 reverse-ALPH chair → half-chair TS → skew pathway (Figure 8, full line from right to left) would result in an activation barrier that is increased by the energy difference ∆E between a skew and a chair conformation.For many, if not all, O-glycopyranosides this may be large enough so that an alternative skew → half-chair TS → chair itinerary is preferred over a reversetype pathway.
The energy profile for a glycopyranosidic bond substitution is however not simply twodimensional as in Figure 8, but is a multidimensional landscape, wherein the actual TS is the lowest saddle point within a high-energy mountainous ridge-barrier that represents all possible TS conformations, including many that are energetically implausible.The above considerations then indicate that the lowest-energy saddle point will always be a half-chair TS.In its vicinity will be a minor energy well representing the local minimum energy conformer that precedes the TS; the Hammond postulate indicates that this conformer also structurally resembles the TS.This would be a skew-boat that is ALPH-compliant, and such conformation can be trapped in glycoside hydrolases with e.g.Driguez' thiooligosaccharide analogs. 67One does not expect that a pre-TS conformation is a strict boat, since this contains an unfavorable flagpole interaction as well as Pitzer tensions from eclipsed substituents.The BEP principle predicts that the newly formed product should fall into the deepest nearby energy minimum right after the half-chair TS.For the substitution of β-D-glycopyranosides into their α-products, this end-conformation often is the α-4 C 1 ground state, overall resulting in the ALPH-β-skew → 4 H 3 -like TS → α-4 C 1 itinerary (Figure 3) in full accordance with Deslongchamps' theory. 37n the case of α-D-glycopyranosides substituting into their β-products, the analogous lowest-energy route in Figure 8 represents the alternative ALPH-type α-skew → 3 H 4 -like TS → β-1 C 4 itinerary (Figure 7), thus initially ending in the β-inverted chair (as in GH47).But some βskews are of comparable energy -and sometimes even of slightly lower energy -than the β-1 C 4 , 58 often mainly due to the presence in the latter of an unfavorable syn-diaxial interaction from the β-C1-glycosidic bond versus the D-C5-hydroxymethyl group.However, and this holds for substitutions on α-as well as β-glycopyranosides, syn-diaxial interactions are partly relieved in a C3-C4 half-chair TS.Moreover, at the stage of the transient TS, the bonds to the leaving and incoming groups are semi-formed and therefore much longer, hence syn-diaxial interactions towards the C1 substituents are not yet present.And, the newly formed product at the start of the energy descent, very shortly after the TS, still resembles it (Hammond postulate).So, neither the TS, nor the species shortly thereafter endure the steric repulsions that will be encountered at the end of the substitution process.Hence, the downwards energetic pathway will be in the direction of a chair anyway, since this is the initial steepest descent that provides maximal initial ring strain release, irrespective of whether the end result will be a 4 C 1 or 1 C 4 .Any lower energy conformation may then only be reached by crossing another barrier -but by then the actual substitution is already over.This suggests that substitutions of α-D-O-glycopyranosides into their β-products may very well, if not always, proceed via the alternative ALPH-type α-skew → 3 H 4like TS → β-1 C 4 itinerary, and only thereafter will the product return into the β-4 C 1 ground state.

Demarcation of preferred substitution itineraries within the glycopyranoside conformational space
In Figure 9, a map is presented that is a mechanism-based adaptation of Dowd's Mercator projection for the conformational space of aldopyranosides, 58 which itself is an adaptation of the Cremer-Pople puckering sphere for six-membered rings. 68Herein, the pseudorotational series of the subsequent ideal skew and boat conformations is spread over the equator, all having a Θ orientational puckering of 90° with the Φ orientational puckering varying between 0° and 360°.
The ideal 1 C 4 and 4 C 1 conformations reside respectively at the North Pole (Θ = 180°) and at the South Pole (Θ = 0°), whereas the two pseudorotational series of the subsequent ideal half-chair and envelope conformations are near the Tropic of Cancer and Tropic of Capricorn, respectively.At the skew-boat pseudorotational series and at both chair conformations, the value above the conformational descriptors represents the ideal torsional angle between the glycosidic bond of an α-glycopyranoside versus the trans-positioned free electron pair of the ring oxygen, and the value below represents the same for a β-glycopyranoside.It is obvious that the ALPH requirement is fulfilled when this value reaches 150-180°.Such Mercator projection (Figure 9) should be preferred over the much used 15,17 conformational wheel, which is in essence a simplified azimuthal projection of the Cremer-Pople sphere.In order to cover the complete conformational space, one needs in fact two azimuthal projections, i.e. one from each Pole.Moreover, these do not take into account the relative positioning of envelopes within the conformational series near both tropics.
Within the general setting of the Mercator projection, a classical β-as well as α-Dglycopyranosidic bond substitution are both suggested to be preceded by a conformational change, away from the ground state 4 C 1 at the South Pole (Θ = 0°), into an ALPH-compliant conformation belonging to the skew-boat pseudorotational series at the Equator (Θ = 90°).From there, the PLM indicates that the reaction itinerary should follow a geodesic along the Cremer-Pople sphere, whereas the BEP principle indicates a route in the direction of a chair conformation; thus only a small variation in Φ should be involved in the process.The ultimate substitution is a fast process 30   Consequently, this results in the following two preferred general itineraries: (1) For β-D-glycopyranosides: Deslongchamps-type.These comply with the ALPH requirement when Θ = 90° at Φ = 150° to 270°, comprising the 2 S O , 1 S 3 and 1 S 5 skews as possible pre-TS local minima (boat conformations are seldom local minima 58 ).TS candidates should have a nearby value of Φ, and should be coplanar at C5-O5-C1-C2, which is only the case with the E 3 , 4 H 3 and 4 E conformations (indicated with the Cross of Lorraine in Figure 9) at the Φ = 180° to 240° range.The reaction itinerary is then a fast passage from one of the ALPH-skews over the lowest saddle point in the high-energy mountains near the Tropic of Capricorn into the α-4 C 1 , thus with Θ = 90 → 0° and with minimal variation of Φ.Note that the opposite conformations with Φ between 180° and 240° that are near the Tropic of Cancer do not have coplanarity at C5-O5-C1-C2, and therefore impose an insurmountable energy barrier ridge for a Θ = 90 → 180° itinerary towards the α-1 C 4 .
(2) For α-D-glycopyranosides: Alternative ALPH-type.These comply with the ALPH requirement when Θ = 90° at Φ = 330° to 90°, comprising the O S 2 , 3 S 1 and 5 S 1 skews as possible ARKAT pre-TS local minima.TS candidates are then only the C5-O5-C1-C2 coplanar 3 E, 3 H 4 and E 4 conformations (indicated with the Cross of Lorraine in Figure 9) at the Φ = 0° to 60° range.The reaction itinerary now passes over the Tropic of Cancer into the β-1 C 4 , thus with Θ = 90 → 180° and again with minimal variation of Φ.Note that in this case the opposite conformations with Φ between 0° and 60° that are near the Tropic of Capricorn do not have coplanarity at C5-O5-C1-C2, and therefore impose an insurmountable energy barrier ridge for a Θ = 90 → 0° itinerary directly towards the ground state β-4 C 1 .The latter will eventually be reached via an independent process, long after the substitution has occurred.
In Figure 10, these two general itineraries are shown together with the respective Θ and Φ values in relation to the shape of the conformers.If one does not take the ring substituents into account, both ultimate skew → TS → chair substitutions are in essence ring-plane-mirrored situations of each other.Each is subjected to relatively small atomic movements, which is in full accordance with the PLM.Indeed, the positions at C2-C3-C4-C5 vary only minimally, whereas large repositionings have occurred before the actual substitution, i.e. at the respective conformational change into an ALPH-compliant pre-TS skew.

Several glycopyranoside hydrolase structures reveal ligands in a pre-TS skew conformation
Since enzymes have evolved to preferentially stabilize transition states, 69,70 one may assume that the active site of an enzyme has also evolved to allow the conformational changes that the substrate has to undergo in order to be able to reach the TS.In the case of glycopyranoside hydrolases, the possible substitution itineraries contain non-ground state local minimum conformations that belong to the pseudorotational series and have a chance of being trapped by the enzyme, e.g. when using substrate-analogous inhibitors.
Table 1 is a compilation of known skew-or boat-complexed glycoside hydrolase structures, for which the carbohydrate moiety in subsite -1 does not contain a double bond (as e.g. in acarbose) since this results in a ground state ligand having an imposed half-chair conformation, whereas the focus is on non-ground state local minima that are preferred by the enzyme.It is reasonable to assume that these represent pre-TS situations, since an enzyme would prefer to bind a high-energy conformer that is en route to the TS, as a means to decrease the activation barrier by avoiding the low-energy ground state.In the further case-by-case analysis, we will show that even those glycosyl-enzyme intermediates that have been observed as (skew-) boat conformers can as well be interpreted as pre-TS local minima.
The observed pre-TS conformers allow us to infer the subsequent TS conformations for the respective glycosylation, deglycosylation or hydrolysis, and these are indicated in Table 1.For the first entry, the density at subsite -1 of the 4-nitrophenyl-1-thio-β-glucopyranoside Michaelis complex with the β-retaining GH1 maize 1,4-glucosidase (pdb entry 1e1f) was interpreted as indicative of the β-1,4 B but with a problematic sp 2 geometry at C1; 71 the observed C6-C5-(O5)-C1-O1 W-angle of -15° instead suggests the presence of the β-2 S O and is as such included.

On β-glycoside hydrolases
The glycosylation step of retaining β-glycoside hydrolases yields a covalent α-glycosyl-enzyme intermediate, which is subsequently hydrolyzed in the deglycosylation step (Figure 1).Several structures of such glycosyl-enzyme intermediates have been solved, many of them as Witherstype 2-deoxy-2-fluoro derivatives. 72With notable exceptions, which will be discussed in the following sections, many show a ground state α-4 C 1 and therefore ALPH-compliant glycon conformation.This may suggest that a reverse Deslongchamps-type α-4 C 1 → 4 H 3 -like TS → βskew deglycosylation pathway is operative, but might just as well be interpreted as the end situation of a Deslongchamps-type β-skew → 4 H 3 -like TS → α-4 C 1 glycosylation itinerary (Figure 3).
As noted earlier, one should be very cautious when deducing an enzyme's reaction itinerary solely based on a ground state conformation in a complexed enzyme structure.Using the PMR for comparing a glycosylation to a deglycosylation is again inappropriate since these are two separate reaction steps; in other words, the PMR does not make it obligatory for the deglycosylation step to follow a reverse-like itinerary of the one that formed the glycosyl enzyme intermediate.This even accounts for a glycosylation versus a subsequent transglycosylation: although here the return reaction for the glycosylation appears to be the same as a subsequent transglycosylation, the PMR only indicates that a reverse Deslongchamps-type return reaction (or transglycosylation) is possible, but it does not exclude the possibility that a lower-energy alternative ALPH-type itinerary that has its own PMR is operative.

GH26 (Clan A)
With the β-retaining GH26 mannanase from Pseudomonas cellulosa (Man26A), the crystal structure of the proton donor mutant in Michaelis complex with 2,4-dinitrophenyl 2-deoxy-2fluoro-β-mannotrioside (pdb entry 1gvy) shows the glycon in the -1 subsite clearly in an ALPHcompliant β-1 S 5 , whereas that in the corresponding 2-deoxy-2-fluoro-mannotriosyl-enzyme intermediate (pdb entry 1gw1) adopts an ALPH-compliant α-O S 2 . 50Since these two conformations flank the B 2,5 within the skew-boat pseudorotational series, it was suggested that the TS for formation of this glycosyl-enzyme intermediate has a B 2,5 conformation.However, as argued above, a B 2,5 as TS in a mannopyranoside is not very likely because of an unfavorable C2-C5 flagpole positioning as well as eclipsed substituents at the C3-C4 bond, and because it should react according to the SLPH instead of ALPH.On the other hand, both observed β-1 S 5 and α-O S 2 conformations are each fully consistent with the proposed general ALPH-skew → half-chair-like TS → chair pathways as indicated in Figures 9 and 10, even though the glycosylenzyme intermediate contains an acetal-ester bond: (1) The glycosylation step of GH26.The observed β-1 S 5 at Φ = 270° in the Michaelis complex is the pre-TS local minimum for a Deslongchamps-type β-1 S 5 → 4 E-TS → α-4 C 1 glycosylation itinerary (Figure 3).The enzyme apparently favors a 1 S 5 β-mannopyranoside that has an equatorial C2-OH, providing the least steric hindrance for the axial nucleophilic attack on the anomeric center.A reaction itinerary starting from a β-1 S 5 with the least atomic motions (minimal ∆Φ) then indicates the passage through the 4 E-TS (Φ = 240°) and a subsequent fast descent into the α-4 C 1 energy minimum (Θ = 90 → 0°).The latter is, however, not the starting conformation for the subsequent deglycosylation step.
(2) The deglycosylation step of GH26.The observed α-O S 2 at Φ = 330° of the glycosyl-enzyme intermediate has arisen from a conformational change away from the α-4 C 1 , which was the end of the glycosylation step, back into the skew-boat pseudorotational series.Here the enzyme has favored the local-minimum O S 2 α-mannopyranosyl ester with again an equatorial C2-OH, which does not hinder the axial attack on C1 of the incoming water nucleophile.This then suggests that an alternative ALPH-type α-O S 2 skew → 3 H 4 -like TS → β-1 C 4 deglycosylation itinerary is operative (Figure 7).However, the D-mannopyranoside α-O S 2 conformation has a 1,3-syn-diaxial interaction between C3-OH and the incoming nucleophile that is partly relieved in the nearby 3 S 1 conformation at Φ = 30°, making this a better pre-TS candidate.And starting from a pre-TS α-O S 2 or α-3 S 1 , the PLM (thus with minimal ∆Φ) suggests a passage through the 3 E (Φ = 0°) or 3 H 4 (Φ = 30°) TS, followed by a fast descent into the nearby β-1 C 4 energy minimum (Θ = 90 → 180°).The return to the ground state β-4 C 1 is then a subsequent independent process.It appears at first sight unlikely that such large conformational changes within the -1 subsite are still possible with a molecule that is covalently attached, and/or that an inverted chair can be temporarily accommodated within this crucial subsite.However, as indicated above, the known crystal structures of complexed hexopyranoside hydrolases always show a remarkably ARKAT spacious -1 subsite, especially at the region where the C5-hydroxymethyl group of the substrate resides.

GH11 (Clan C)
A crystal structure of the β-retaining GH11 xylanase from Bacillus circulans reveals the covalent 2-deoxy-2-fluoro-xylobiosyl-enzyme intermediate in the α-2,5 B that is slightly distorted towards the 5 S 1 (pdb entry 1bvv). 73The same glycosyl-enzyme intermediate on the GH11 Bacillus agaradhaerens xylanase (pdb entry 1h4g) even apparently shows a strict α-2,5 B in spite of the C3-C4 eclipsed positioning and the C2-C5 flagpole, whereas the xylotetraose Michaelis complex with the nucleophile G94A mutant (pdb entry 1h4h) rather seems to adopt a conformation that is between the β-2,5 B and 5 S 1 for the xylosyl moiety within subsite -1. 51This has led these authors to suggest a start-4 Such pathway would be different from the reaction itinerary within the GH12 glucanases where the 2,5 B-TS of their glucoside glycon would place the C5-hydroxymethyl group in axial position, thereby imposing a severe flagpole interaction with the C2-hydrogen atom; however, both families belonging to the same GH Clan C (thus having a common ancestor 5 ) could tentatively be expected to operate with similar itineraries.But the GH12 Humicola grisea endoglucanase cellopentaose Michaelis complex (pdb entry 1w2u) reveals the ALPH-β-1 S 3 within the -1 subsite 74 whereas the Streptomyces lividans endoglucanase's 2-deoxy-2-fluorocellotriosyl-enzyme intermediate (pdb entry 2nlr) clearly shows its glycon in the α-4 C 1 , 75 both in accordance with respectively the pre-TS and the end-situation of a Deslongchamps-type β-1 S 3 → 4 H 3 -TS → α-4 C 1 glycosylation itinerary (Figure 3).The 2,5 B is at Φ = 120° in the Θ = 90° skew-boat pseudorotational series, flanking the 5 S 1 and 2 S O at Φ = 90° and 150°, respectively (Figure 9).In the absence of a C5-hydroxymethyl group, interconversion between these skews is relatively unhindered, and as such a 2,5 B Dxylopyranoside may even become a local minimum within an enzyme, as apparently is observed in GH11.As outlined above, the two ALPH pathways (Figures 9 and 10) may very well be operative, instead of an itinerary passing through a SLPH-compliant 2,5 B-TS.Both again start with a prior conformational change from the 4 C 1 into the skew-boat pseudorotational series: (1) The glycosylation step of GH11.Slight atomic movements of the observed β-2,5 B in the Michaelis complex can easily yield the ALPH-compliant β-2 S O at Φ = 150°, enabling a Deslongchamps-type β-2 S O skew → E 3 -TS → α-4 C 1 glycosylation itinerary (Θ = 90 → 0° with minimal ∆Φ).The β-2 S O is now preferred since this has an equatorial C2-OH yielding the least steric hindrance for the incoming nucleophile towards C1 (the above β-mannopyranoside uses the β-1 S 5 for the same reason).The end conformation is then the α-4 C 1 , as observed in the GH12 S. lividans glycosyl-enzyme intermediate, 75 but that is not the productive conformation for the deglycosylation step.
With a half-chair-like TS in both itineraries, the orthogonal ring oxygen lone pair (forming a partial double bond towards C1) can also interact with the phenolic oxygen of a GH11 invariant tyrosine; this interaction has been proposed to be essential for electrostatic TS stabilisation. 21,51,73And proximality of the C2 hydroxyl group towards the nucleophile is also obtained; such interaction may be important in different β-glucoside hydrolases. 16,76The apparent strict xylan preference by GH11 enzymes may then not be achieved through an exotic reaction itinerary but through a simple steric incompatibility towards cellulose.Indeed, especially the +1 subsite appears to be too narrow to accept a glucose entity, whereas it has been noticed in the original references 51,73 that a conserved leucine or valine in the -1 subsite is in Van der Waals contact with the C5-methylene of the substrate, blocking the accommodation of a C5hydroxymethyl group.

On α-glycoside hydrolases
The formation of the glycosyl-enzyme intermediate by α-retaining glycoside hydrolases, as well as the direct hydrolysis catalyzed by α-inverting glycoside hydrolases, are both proposed to proceed through itineraries that start with a conformational change into the skew-boat pseudorotational series, followed by an alternative ALPH-type α-skew → 3 H 4 -like TS → β-1 C 4 substitution (Figure 7).This can be supported by kinetic isotope effects on the enzymatic hydrolysis of aryl-α-D-glucopyranosides, as noted earlier.For α-inverting enzymes, this initially yields the inverted chair conformation for the liberated glycoside, and a return to the ground state will be an independent process.For α-retaining enzymes, the alternative ALPH-type glycosylation step initially results in formation of a β-axial glycosyl-enzyme intermediate with the 1 C 4 inverted chair conformation.A return to the ground state β-4 C 1 may occur, but this results in a β-equatorial anomeric substituent that cannot be a productive species for hydrolysis in the deglycosylation step.A direct substitution on the 1 C 4 conformer would be via the reverse of an alternative ALPH-type itinerary (Figure 7, from right to left), leading initially to a product with a local minimum skew conformation.But a re-entry into the pseudorotational series, followed by a Deslongchamps-type β-skew → 4 H 3 -like TS → α-4 C 1 deglycosylation pathway (Figure 3), leads directly to the ground state product and thus is expected to be the preferred deglycosylation itinerary.
As with the β-glycoside hydrolase structures and with the same caution for interpretation, many liganded α-glycoside hydrolase structures show ground state chair conformers within the -1 subsite.But the observed non-ground state conformers may very well represent pre-TS situations that are indicators for the reaction itineraries that are operative.

α-Retaining glycoside hydrolases
The GH38 Drosophila melanogaster Golgi α-mannosidase II's 5-fluoro-gulosyl-enzyme intermediate (pdb entry 1qwn) reveals the pre-TS ALPH-β-1 S 5 (Φ = 270°) that has an equatorial C2-OH. 85This was again interpreted as indicative of a SPLH-compliant B 2,5 -TS, but can just as well be interpreted as indicative of a Deslongchamps-type β-1 S 5 → 4 E-TS → α-4 C 1 pathway for the deglycosylation (Figure 3).The same enzyme in complex with 5-thio-D-mannopyranosyl amine and with the corresponding benzyl amidinium bromide (pdb entries 1r33 and 1r34, respectively) in both cases also reveal the 1,4 B (Φ = 240°) that appears to be slightly distorted towards the 1 S 5 . 86These α-substituted amino compounds, however, show an equatorial anomeric substituent and therefore do not mimic a productive pre-TS Michaelis complex, which is expected to have an axial ALPH-α-skew conformation with Φ between 330° and 90°.Such conformation is probably disfavored due to a steric clash of the ligand's large sulfur atom with the nearby Arg228 residue, and the enzyme may then have favored a non-ground state conformation that reflects the pre-TS of the deglycosylation step.
In the GH33 Trypanosoma cruzi trans-sialidase Michaelis complex (pdb entry 1s0i), 89 the conformation of the α-sialosyl moiety appears to be a strict ALPH-α-B 2,5 and indicates a subsequent E 5 -TS (both at Φ = 60°) into the β-2 C 5 as itinerary for the glycosylation.With the aldopyranoside numbering this coincides with the α-B 1,4 followed by the E 4 -TS into the β-1 C 4 , thus similar to the alternative ALPH-type pathway for a classical α-axial glycopyranosidic bond substitution.The same sialidase's 2-fluoro-sialosyl-enzyme intermediate with the Tyr342 nucleophile (pdb entry 1s0k) 89 shows the β-2 C 5 , apparently in accordance with the end of the glycosylation step.

GH90 Tailspike protein
In GH90 α-inverting Salmonella phage P-22 tailspike protein complexes with O-antigen fragments (pdb entries 1tyx, 1tyu and 1yw), the terminal L-rhamnose atom densities at subsite -1 have been interpreted as being non-ALPH α-5 S 1 conformations. 90However, these are not -1/+1 subsite-spanning product complexes for which one expects to find the lower energy ground state not a local minimum (the ALPH-α-O S 2 is, at Φ = 330°) and has an estimated steric burden of 7 kcal/mol versus the ground state α-4 C 1 at Θ = 0°. 58A docking study on this enzyme also indicates that the terminal L-rhamnose ground state α-1 C 4 should largely be preferred over the α-5 S 1 .The same study inter alia suggests that the assigned L-rhamnose C2-hydroxyl oxygen could have actually been the oxygen from a nearby water molecule; the opposite assignment indeed would fit the more realistic α-1 C 4 conformation. 91The predicted itinerary (Figures 9 and 10 but with L-glycopyranoside nomenclature: inverted numbering at the conformational descriptors) is then an alternative ALPH-type α-2 S O (Φ = 330°; C2-OH equatorial) → E 3 -TS (Φ = 0°) → β-4 C 1 (Θ = 180°) glycoside hydrolysis pathway.

Conclusions
From a recapitulation of current inconsistencies within the literature, as well as considerations of basic principles in organic chemistry, it appears that pyranoside glycosidic bond substitutions that operate by a classical exocyclic mechanism on oxyalkyl-type leaving groups proceed via two inherently preferred itineraries: 1) β-equatorial D-glycopyranosides are expected to follow a Deslongchamps-type β-skew → 4 H 3 -like TS → α-4 C 1 itinerary (Figure 3); and 2) α-axial Dglycopyranosides may very well not react by a direct substitution on the ground state in a reverse Deslongchamps pathway, but instead follow an alternative ALPH-type α-skew → 3 H 4 -like TS → β-1 C 4 itinerary, with an independent return to the ground state β-4 C 1 occurring thereafter (Figure 7).Both substitutions are in essence ring-plane mirrored situations of each other, in which the route to the transient TS is preceded by a conformational change into an ALPH-compliant conformation belonging to the skew-boat pseudorotational series.The route demarcations within the conformational space are presented in Figures 9 and 10.
A reassessment of the currently known 3D structures of glycoside hydrolases that are ligand-complexed with non-ground state carbohydrate conformers indicates that these respective itineraries are operative in enzymes as well.Even deglycosylations or transglycosylations of glycosyl-enzyme intermediates appear to proceed in this way, and not through a reverse-type itinerary of the preceding glycosylation step.Arguments are given against the possible occurrence of a boat TS within a SLPH-compliant itinerary, whereas the alternative ALPHpathway (Figure 7) can be supported by experimental evidence when revisiting results from a study 52 on kinetic isotope effects observed with enzymatic hydrolyses of aryl α-D-glucosides (Figure 6).
The route demarcations suggest a general mechanistic strategy for protection of the glycosyl-enzyme intermediate by transglycosidases: these may lock their intermediate in a nonproductive conformation for as long as the aglycon subsites are still occupied by water molecules solely, but once these are properly replaced by the carbohydrate-nucleophile, a minor amino acid residue reorientation (induced fit) may favor the productive pre-TS conformation.This explains the equatorial anomeric positioning of the β-4 C 1 4-deoxymaltotriosyl-enzyme

Figure 3 .
Figure 3.The Deslongchamps-type substitution itinerary for β-D-glycopyranosides.The O5 lone pair that is trans to the leaving group is indicated in black.A preceding conformational change into an ALPH-compliant 1 S 3 -like pre-TS is followed by a transient 4 H 3 -like TS and leads to the α-4 C 1 product.A 3 H 4 -like TS does not belong to this itinerary.L = leaving group; Nu = nucleophile.

Figure 4 .
Figure 4. Comparison of the regular B 2,5 and 2,5 B conformers with the corresponding boat-TS conformations shows a synperiplanar topology of the leaving group versus an O5 lone pair (indicated in black).Only the situations for β-D-glycopyranosides are shown; the topology is analogous for α-D-glycopyranosides. L = leaving group; Nu = nucleophile.

Figure 6 .
Figure 6.A 2,5 B-TS of an α-D-glycopyranoside has a near-perpendicular positioning of the C2-H bond versus the C1-O5 bond, but the same relative positioning occurs in the 3 S 1 pre-TS and its overall-similar 3 H 4 -TS.
that passes through the lowest saddle point in the high-energy mountain ridges of TS conformation possibilities near the Tropic of Capricorn or Tropic of Cancer and ends in a chair at Θ = 0° or 180° for either a β-or an α-original glycosidic bond configuration.

Figure 9 .
Figure 9. Mercator projection map of the Cremer-Pople puckering sphere (adapted from Dowd 58 ), showing the locations of characteristic conformers.This is a mechanism-based adaptation for D-glycopyranosides.Pre-TS conformations at Θ = 90° are ALPH-compliant, thus with a torsional angle between the glycosidic bond versus the trans-oxygen free electron pair of at least 150°, indicated above (for α-D-glycopyranosides) or below (for β-D-glycopyranosides) the conformational descriptors.Preferred transition states are near a pre-TS (PLM) and are indicated with the Cross of Lorraine.The arrows indicate alternative ALPH-type (left) and Deslongchamps-type (center) substitution itineraries.