Design and Synthesis of 3-(β-d-Glucopyranosyl)-4-amino/4-guanidino Pyrazole Derivatives and Analysis of Their Glycogen Phosphorylase Inhibitory Potential

Glycogen phosphorylase (GP) is a key regulator of glucose levels and, with that, an important target for the discovery of novel treatments against type 2 diabetes. β-d-Glucopyranosyl derivatives have provided some of the most potent GP inhibitors discovered to date. In this regard, C-β-d-glucopyranosyl azole type inhibitors proved to be particularly effective, with 2- and 4-β-d-glucopyranosyl imidazoles among the most potent designed to date. His377 backbone C=O hydrogen bonding and ion–ion interactions of the protonated imidazole with Asp283 from the 280s loop, stabilizing the inactive state, were proposed as crucial to the observed potencies. Towards further exploring these features, 4-amino-3-(β-d-glucopyranosyl)-5-phenyl-1H-pyrazole (3) and 3-(β-d-glucopyranosyl)-4-guanidino-5-phenyl-1H-pyrazole (4) were designed and synthesized with the potential to exploit similar interactions. Binding assay experiments against rabbit muscle GPb revealed 3 as a moderate inhibitor (IC50 = 565 µM), but 4 displayed no inhibition at 625 µM concentration. Towards understanding the observed inhibitions, docking and post-docking molecular mechanics—generalized Born surface area (MM-GBSA) binding free energy calculations were performed, together with Monte Carlo and density functional theory (DFT) calculations on the free unbound ligands. The computations revealed that while 3 was predicted to hydrogen bond with His377 C=O in its favoured tautomeric state, the interactions with Asp283 were not direct and there were no ion–ion interactions; for 4, the most stable tautomer did not have the His377 backbone C=O interaction and while ion–ion interactions and direct hydrogen bonding with Asp283 were predicted, the conformational strain and entropy loss of the ligand in the bound state was significant. The importance of consideration of tautomeric states and ligand strain for glucose analogues in the confined space of the catalytic site with the 280s loop in the closed position was highlighted.


Introduction
Glycogen phosphorylase (GP; EC 2.4.1.1) is the rate-determining enzyme in the glycogenolysis pathway and a validated target for the development of new anti-hyperglycemic agents [1]. Diabetes is a major socio-economic burden with over 0.5 billion people living with diabetes worldwide, and with this predicted to increase by 643 million by 2030 and 783 million in 2045, the problem urgently requires more effective interventions that those currently available [2]. GP is a validated target for type 2 diabetes (T2D) and has considerable potential in this regard. A number of studies demonstrated the anti-diabetic potential of GP inhibitors in vivo [3,4]. Additionally, GP modulators have potential for treatment of other serious conditions such as myocardial and cerebral ischemias [5][6][7], and cancer [5,8]. Indeed, there was much recent interest with respect to control of glycogenolysis through GP inhibition for different cancers, that Scheme 1. C-Glucopyranosyl imidazole 1, with the important interactions in the catalytic site highlighted and pyrazole 2 type GP inhibitors, together with their efficacy against rabbit muscle GPb (rmGPb). The target compounds in this work are shown as 3 and 4.
Considering the binding features of 1, it was speculated whether modifications of hydrogen bonding potential of an azole substituent with the Asp283 sidechain and/or its 3D arrangement in forming ion-ion interactions with the carboxylate sidechain group might be favourable, while maintaining the critical hydrogen bond interaction with His377 backbone C=O. In that regard, 3 and 4-the C-4 substituted derivatives of pyrazole 2 (Scheme 1)-were considered for synthesis in this study. The pyrazole linker of 3 has a potential for both His377 C=O (as shown earlier [21]) and Asp283 side-chain hydrogen bonding, while 4 has the same hydrogen bonding potential but an additional possibility for ion-ion interactions with the Asp283 side-chain. In theory, the T state conformation of the enzyme would be favoured by stabilisation of the closed conformation of the 280s loop, blocking access of the substrate to the catalytic site. In this paper, we describe syntheses of compounds 3 and 4, as well as their inhibitory potencies against rabbit muscle GPb (rmGPb). Additionally, extensive computations on the bound and unbound states of the ligands (Monte Carlo conformational searches, DFT, docking and post-docking molecular mechanics-generalized Born surface area (MM-GBSA)) are presented, to rationalize the observed binding assay results.

Synthesis
For the preparation of the target compounds, we chose a straightforward method based on the reaction of a sugar derived 1,3-dielectrophile and hydrazine. The C-N bond of the amino substituent of the heterocycle was formed at the level of dielectrophile, since we failed to produce the necessary intermediate from the known O-perbenzoylated 3-(β-D-glucopyranosyl)-5-phenyl-1H-pyrazole [21] by C-4 nitration/nitrosation of the pyrazole.
The dielectrophilic precursor was prepared from glucosyl cyanide 5 [23] with phenacyl bromide under Blaise conditions [24] (Scheme 2). Acidic workup at low temperature gave enaminone 6, while hydrolysis of the Blaise reaction mixture at 85 °C gave enol 7, which, upon treatment with NaNO2 under acidic conditions [25], furnished oxyme 10. Scheme 1. C-Glucopyranosyl imidazole 1, with the important interactions in the catalytic site highlighted and pyrazole 2 type GP inhibitors, together with their efficacy against rabbit muscle GPb (rmGPb). The target compounds in this work are shown as 3 and 4.

Synthesis
For the preparation of the target compounds, we chose a straightforward method based on the reaction of a sugar derived 1,3-dielectrophile and hydrazine. The C-N bond of the amino substituent of the heterocycle was formed at the level of dielectrophile, since we failed to produce the necessary intermediate from the known O-perbenzoylated 3-(β-Dglucopyranosyl)-5-phenyl-1H-pyrazole [21] by C-4 nitration/nitrosation of the pyrazole.
The dielectrophilic precursor was prepared from glucosyl cyanide 5 [23] with phenacyl bromide under Blaise conditions [24] (Scheme 2). Acidic workup at low temperature gave enaminone 6, while hydrolysis of the Blaise reaction mixture at 85 • C gave enol 7, which, upon treatment with NaNO 2 under acidic conditions [25], furnished oxyme 10. Ring closure of 10 with an excess of hydrazine monohydrate resulted in 4-aminopyrazole 8 as the main product together with the non-reduced 4-nitrosopyrazole 9. The latter compound was identified based on its green colour; furthermore, the C-4 chemical shift difference of the nitrosated pyrazole 9 (157.1 ppm) compared to its non-nitrosated counterpart (101.4 ppm) [21] was 55.7 ppm, which is in good agreement with literature data [26]. Catalytic hydrogenation of the nitroso compound 9 led to the amino derivative 8, from which we obtained one of the target compounds (3) via O-deprotection under Zemplén conditions. Treatment of 8 with N,N -di-Boc-1H-pyrazole-1-carboxamidine [27] resulted in the protected guanidine derivative 11, from which the other target molecule 4 was obtained after N-Boc and O-benzoyl cleavage.

Glycogen Phosphorylase Binding Assays
The inhibitory potency of 3 and 4 was assessed using binding assay experiments against rmGPb. The determined IC 50 values are shown in Table 1, together with the previously determined K i value for benchmark compound 1. Compound 3 (IC 50 = 565 µM) was a moderate inhibitor of rmGPb and, while better than compound 2 (IC 50 = 850 µM), it was much less potent than compound 1 (K i = 0.28 µM); compound 4 was revealed as a poor inhibitor (no inhibition at 625 µM). Towards understanding the observed potencies, extensive computations on the bound and unbound states of the ligands were performed.

In-Silico Studies
The binding assay results necessitated structural studies to elucidate the nature of protein-ligand interactions leading to the observed low potencies. Computational studies have proven to be a useful tool to rationalize GP inhibitor efficiency [28,29], including previously reported glucose analogues containing heterocyclic linkers [19,30,31]. As the initial design hoped to exploit interactions with the His377 backbone C=O and strong interactions with the Asp283 sidechain carboxylate stabilizing the closed position of the 280s loop, it was important to first establish the most stable states of the free unbound ligands, prior to the protein-ligand binding calculations. For this purpose, ionization and tautomeric state stabilities of the ligands [32] were explored using Monte Carlo conformational searches supplemented by DFT post-processing minimizations (M06-2X/6-31+G*) to determine the stable unbound conformations. In a very recent benchmarking study of druglike scaffolds, the M06-2X method outperformed a range of semi-empirical and quantum mechanical methods in terms of accurate calculation of relative tautomeric energies [33]. Furthermore, the M06-2X/6-31+G* level of theory was previously successfully applied to study tautomeric states of glucose-azole type inhibitors [19]. The chemical structures of the relevant unbound states (ionization/tautomeric) of the ligands 3 and 4 are shown in Figure 1. In a study of tautomer preferences in PDB complexes, the most stable water state tautomer is predominantly the most favoured binding state tautomer, depending on the ∆G between the two tautomers [34]. Solution phase energies were calculated for the optimized DFT conformations using M06-2X/6-31+G* with solvation effects included with the SM8 water solvation model.
With respect to the protein-ligand bound states, the end point method MM-GBSA is recognised as an effective post-docking strategy for the calculation of relative binding free energies (ΔGbind) [35]. Glide-SP docking poses of 3 and 4 were post-processed using MM-GBSA using two equations.
In Equation (1), represents the total molecular mechanics energy (internal, electrostatic and van der Waals); , the solvation free energy calculated using the variable-

In-Silico Studies
The binding assay results necessitated structural studies to elucidate the nature of protein-ligand interactions leading to the observed low potencies. Computational studies have proven to be a useful tool to rationalize GP inhibitor efficiency [28,29], including previously reported glucose analogues containing heterocyclic linkers [19,30,31]. As the initial design hoped to exploit interactions with the His377 backbone C=O and strong interactions with the Asp283 sidechain carboxylate stabilizing the closed position of the 280s loop, it was important to first establish the most stable states of the free unbound ligands, prior to the protein-ligand binding calculations. For this purpose, ionization and tautomeric state stabilities of the ligands [32] were explored using Monte Carlo conformational searches supplemented by DFT post-processing minimizations (M06-2X/6-31+G*) to determine the stable unbound conformations. In a very recent benchmarking study of druglike scaffolds, the M06-2X method outperformed a range of semi-empirical and quantum mechanical methods in terms of accurate calculation of relative tautomeric energies [33]. Furthermore, the M06-2X/6-31+G* level of theory was previously successfully applied to study tautomeric states of glucose-azole type inhibitors [19]. The chemical structures of the relevant unbound states (ionization/tautomeric) of the ligands 3 and 4 are shown in Figure 1. In a study of tautomer preferences in PDB complexes, the most stable water state tautomer is predominantly the most favoured binding state tautomer, depending on the ∆G between the two tautomers [34]. Solution phase energies were calculated for the optimized DFT conformations using M06-2X/6-31+G* with solvation effects included with the SM8 water solvation model.
With respect to the protein-ligand bound states, the end point method MM-GBSA is recognised as an effective post-docking strategy for the calculation of relative binding free energies (ΔGbind) [35]. Glide-SP docking poses of 3 and 4 were post-processed using MM-GBSA using two equations.
In Equation (1), represents the total molecular mechanics energy (internal, electrostatic and van der Waals); , the solvation free energy calculated using the variable-

In-Silico Studies
The binding assay results necessitated structural studies to elucidate the nature of protein-ligand interactions leading to the observed low potencies. Computational studies have proven to be a useful tool to rationalize GP inhibitor efficiency [28,29], including previously reported glucose analogues containing heterocyclic linkers [19,30,31]. As the initial design hoped to exploit interactions with the His377 backbone C=O and strong interactions with the Asp283 sidechain carboxylate stabilizing the closed position of the 280s loop, it was important to first establish the most stable states of the free unbound ligands, prior to the protein-ligand binding calculations. For this purpose, ionization and tautomeric state stabilities of the ligands [32] were explored using Monte Carlo conformational searches supplemented by DFT post-processing minimizations (M06-2X/6-31+G*) to determine the stable unbound conformations. In a very recent benchmarking study of druglike scaffolds, the M06-2X method outperformed a range of semi-empirical and quantum mechanical methods in terms of accurate calculation of relative tautomeric energies [33]. Furthermore, the M06-2X/6-31+G* level of theory was previously successfully applied to study tautomeric states of glucose-azole type inhibitors [19]. The chemical structures of the relevant unbound states (ionization/tautomeric) of the ligands 3 and 4 are shown in Figure 1. In a study of tautomer preferences in PDB complexes, the most stable water state tautomer is predominantly the most favoured binding state tautomer, depending on the ∆G between the two tautomers [34]. Solution phase energies were calculated for the optimized DFT conformations using M06-2X/6-31+G* with solvation effects included with the SM8 water solvation model.
With respect to the protein-ligand bound states, the end point method MM-GBSA is recognised as an effective post-docking strategy for the calculation of relative binding free energies (ΔGbind) [35]. Glide-SP docking poses of 3 and 4 were post-processed using MM-GBSA using two equations.
In Equation (1), represents the total molecular mechanics energy (internal, electrostatic and van der Waals); , the solvation free energy calculated using the variable-

In-Silico Studies
The binding assay results necessitated structural studies to elucidate the nature of protein-ligand interactions leading to the observed low potencies. Computational studies have proven to be a useful tool to rationalize GP inhibitor efficiency [28,29], including previously reported glucose analogues containing heterocyclic linkers [19,30,31]. As the initial design hoped to exploit interactions with the His377 backbone C=O and strong interactions with the Asp283 sidechain carboxylate stabilizing the closed position of the 280s loop, it was important to first establish the most stable states of the free unbound ligands, prior to the protein-ligand binding calculations. For this purpose, ionization and tautomeric state stabilities of the ligands [32] were explored using Monte Carlo conformational searches supplemented by DFT post-processing minimizations (M06-2X/6-31+G*) to determine the stable unbound conformations. In a very recent benchmarking study of druglike scaffolds, the M06-2X method outperformed a range of semi-empirical and quantum mechanical methods in terms of accurate calculation of relative tautomeric energies [33]. Furthermore, the M06-2X/6-31+G* level of theory was previously successfully applied to study tautomeric states of glucose-azole type inhibitors [19]. The chemical structures of the relevant unbound states (ionization/tautomeric) of the ligands 3 and 4 are shown in Figure 1. In a study of tautomer preferences in PDB complexes, the most stable water state tautomer is predominantly the most favoured binding state tautomer, depending on the ∆G between the two tautomers [34]. Solution phase energies were calculated for the optimized DFT conformations using M06-2X/6-31+G* with solvation effects included with the SM8 water solvation model.
With respect to the protein-ligand bound states, the end point method MM-GBSA is recognised as an effective post-docking strategy for the calculation of relative binding free energies (ΔGbind) [35]. Glide-SP docking poses of 3 and 4 were post-processed using MM-GBSA using two equations.
In Equation (1), represents the total molecular mechanics energy (internal, electrostatic and van der Waals); , the solvation free energy calculated using the variable-no inhibition at 625 µM a Atoms numbered for discussion in text.

In-Silico Studies
The binding assay results necessitated structural studies to elucidate the nature of protein-ligand interactions leading to the observed low potencies. Computational studies have proven to be a useful tool to rationalize GP inhibitor efficiency [28,29], including previously reported glucose analogues containing heterocyclic linkers [19,30,31]. As the initial design hoped to exploit interactions with the His377 backbone C=O and strong interactions with the Asp283 sidechain carboxylate stabilizing the closed position of the 280s loop, it was important to first establish the most stable states of the free unbound ligands, prior to the protein-ligand binding calculations. For this purpose, ionization and tautomeric state stabilities of the ligands [32] were explored using Monte Carlo conformational searches supplemented by DFT post-processing minimizations (M06-2X/6-31+G*) to determine the stable unbound conformations. In a very recent benchmarking study of drug-like scaffolds, the M06-2X method outperformed a range of semi-empirical and quantum mechanical methods in terms of accurate calculation of relative tautomeric energies [33]. Furthermore, the M06-2X/6-31+G* level of theory was previously successfully applied to study tautomeric states of glucose-azole type inhibitors [19]. The chemical structures of the relevant unbound states (ionization/tautomeric) of the ligands 3 and 4 are shown in Figure 1. In a study of tautomer preferences in PDB complexes, the most stable water state tautomer is predominantly the most favoured binding state tautomer, depending on the ∆G between the two tautomers [34]. Solution phase energies were calculated for the optimized DFT conformations using M06-2X/6-31+G* with solvation effects included with the SM8 water solvation model.
(protein was treated as rigid throughout), as well as an estimate for the loss of ligand vibrational, rotational and translational (VRT) entropy on binding, a corrected ∆ was calculated by Equation (2) as follows: Benchmark ligand 1 was also recalculated [19] and included for comparative purposes.  Table 2. The unbound state calculations for 3 revealed, as expected, that the -NH2 azole ring substituent would not be protonated using Jaguar pKa (predicted pKa for the protonated -NH3 + state was 4.69), as was also predicted by LigPrep [36]. There are two neutral tautomeric states t1 and t2 for compound 3 ( Figure 1) and the calculated key dihedral angle ω (C 1 H-C 1 -C 2 -N 3 ) for the lowest energy solution phase unbound conformation of each was ω = 72.3° and −107.2° (Table 2), respectively, from the DFT calculations. The same conformations were also the most stable in the gas phase. The most stable solution phase tautomer was predicted as t1 (ω = 72.3°), but just by 0.1 kcal/mol. These most stable solution phase tautomeric conformations are shown in Figure 2A, where we can also see that intra-molecular hydrogen bonds stabilize the conformations and also, that the geometry around the -N(7)H2 substituent is not co-planar with its heterocycle. In line with expectations, the t1 tautomer is the preferred binding tautomer from MM-GBSA calculations (Table 3), with a ΔGbind value of −40.9 kcal/mol (compared to −23.4 kcal/mol for t2). The ΔGbind value is, therefore, significantly less favourable than that of benchmark ligand 1 (−53.1 kcal/mol), which binds in the protonated state [19]. The predicted binding modes of both compounds 1 and 3 are shown in Figure 3, (A) and (B), respectively. Compound 1 (ω = −163.7°) exploits favourable hydrogen-bonding with His377 backbone C=O and ion-ion interactions with the Asp283 sidechain carboxylate; the bound state is consistent with its solved crystallographic complex (PDB code: 5JTT),with a ligand RMSD for heavy atoms of 0.152 Å. Compound 3 also adopts a conformation (ω = −173.0°) to have an With respect to the protein-ligand bound states, the end point method MM-GBSA is recognised as an effective post-docking strategy for the calculation of relative binding free energies (∆G bind ) [35]. Glide-SP docking poses of 3 and 4 were post-processed using MM-GBSA using two equations.
In Equation (1), E MM represents the total molecular mechanics energy (internal, electrostatic and van der Waals); G solv , the solvation free energy calculated using the variabledielectric generalized Born solvation model. As MM-GBSA is an endpoint method, it considers the differences ∆ between the bound and unbound states of the complex, calculated with the OPLS3e forcefield, yielding a ∆G bind (NS) in which strain/reorganisation effects on binding are neglected (NS = no strain). To further include both the ligand strain energy (protein was treated as rigid throughout), as well as an estimate for the loss of ligand vibrational, rotational and translational (VRT) entropy on binding, a corrected ∆G bind was calculated by Equation (2) as follows: Benchmark ligand 1 was also recalculated [19] and included for comparative purposes. Analysis of 3. Results of the DFT calculations for predicted conformations of compound 3 are shown in Table 2. The unbound state calculations for 3 revealed, as expected, that the -NH 2 azole ring substituent would not be protonated using Jaguar pK a (predicted pK a for the protonated -NH3 + state was 4.69), as was also predicted by LigPrep [36]. There are two neutral tautomeric states t1 and t2 for compound 3 ( Figure 1) and the calculated key dihedral angle ω (C 1 H-C 1 -C 2 -N 3 ) for the lowest energy solution phase unbound conformation of each was ω = 72.3 • and −107.2 • (Table 2), respectively, from the DFT calculations. The same conformations were also the most stable in the gas phase. The most stable solution phase tautomer was predicted as t1 (ω = 72.3 • ), but just by 0.1 kcal/mol. These most stable solution phase tautomeric conformations are shown in Figure 2A, where we can also see that intra-molecular hydrogen bonds stabilize the conformations and also, that the geometry around the -N(7)H 2 substituent is not co-planar with its heterocycle. In line with expectations, the t1 tautomer is the preferred binding tautomer from MM-GBSA calculations (Table 3), with a ∆G bind value of −40.9 kcal/mol (compared to −23.4 kcal/mol for t2). The ∆G bind value is, therefore, significantly less favourable than that of benchmark ligand 1 (−53.1 kcal/mol), which binds in the protonated state [19]. The predicted binding modes of both compounds 1 and 3 are shown in Figure 3, (A) and (B), respectively. Compound 1 (ω = −163.7 • ) exploits favourable hydrogen-bonding with His377 backbone C=O and ion-ion interactions with the Asp283 sidechain carboxylate; the bound state is consistent with its solved crystallographic complex (PDB code: 5JTT),with a ligand RMSD for heavy atoms of 0.152 Å. Compound 3 also adopts a conformation (ω = −173.0 • ) to have an N(3)H to His377 backbone C=O hydrogen bond. However, for 3, there are no ion-ion interactions with Asp283 and the -N(7)H 2 substituent cannot form direct hydrogen bonds with the carboxylate side-chain, although water-bridging interactions may be possible. Analysing the breakdown of contributions to ∆G bind (Table 3), it is these less favourable contacts as well as potential contributions from competing tautomeric states (the unbound state tautomeric energy differences are predicted as low, as mentioned just above) that is the likely source of the much lower potency of 3 compared to 1; the ∆G bind (NS) value is less favourable by 14.3 kcal/mol, while the strain energy and entropy contributions are more similar.     contacts as well as potential contributions from competing tautomeric states (the unbound state tautomeric energy differences are predicted as low, as mentioned just above) that is the likely source of the much lower potency of 3 compared to 1; the ∆ ( ) value is less favourable by 14.3 kcal/mol, while the strain energy and entropy contributions are more similar.   Figure 1). The solution phase relative energies of the two tautomers of each ligand are listed, with gas phase relative energies in parentheses for comparison.  (1) and (2). c c.f.  kcal/mol higher in energy (or 8.3 kcal/mol in the gas phase with ω = 131.7°). These most stable solution phase conformations are shown in Figure 2B. The most stable tautomer t2 has a strong network of intra-molecular hydrogen bonds stabilizing the structure, but for a conformation and tautomeric state that is not consistent with forming desired hydrogen bonding with His377 C=O and ion -ion interactions with Asp283 sidechain carboxylate on binding to GP. As mentioned, in a study of ligand tautomeric preferences in PDB complexes, the most stable solution phase tautomer (in this case t2) is predominantly the binding state tautomer (depending on the ∆G between the two tautomers) [34]. In agreement with this study, and of particular significance, we observed for other β-D-glucopryranosylazole inhibitors, considering the binding and MM-GBSA ΔGbind values of the most stable (solution phase) free state tautomer gave better agreement with experiment [19]. The ΔGbind value of 4 (t2) is −28.2 kcal/mol (Table 3), much less favourable than that of 3 (t1; ΔGbind = −40.9 kcal/mol), which is consistent with the experimental binding assay results ( Table  1). The binding of t2 is shown in Figure 3C, where the ligand is seen to adopt the expected binding conformation (ω = −177.1°) that does allow for strong ion-ion and hydrogen bonding interactions with Asp283 sidechain carboxylate. However, there is no hydrogen bond with His377 C=O for this tautomer (∆ ( ) = −67.9 kcal/mol). Additionally, the guanidino ligand substituent due to its steric bulk also extends somewhat towards the positively charged side-chain of Lys574, forming unfavourable ion-ion interactions with this group. Correspondingly, the strain energy effects on ΔGbind are considerably more (~ +18 kcal/mol) compared to 3 (~ +8 kcal/mol) or 1 (~ +10 kcal/mol). The strain from Prime MM-GBSA calculations is estimated from a (local) minimization of bound ligand conformation. Likewise, the ligand entropy cost (− ∆ ) on binding for 4 is ~4 kcal/mol more than both 1 and 3, all of which is consistent with its poor observed binding potential. kcal/mol higher in energy (or 8.3 kcal/mol in the gas phase with ω = 131.7°). These most stable solution phase conformations are shown in Figure 2B. The most stable tautomer t2 has a strong network of intra-molecular hydrogen bonds stabilizing the structure, but for a conformation and tautomeric state that is not consistent with forming desired hydrogen bonding with His377 C=O and ion -ion interactions with Asp283 sidechain carboxylate on binding to GP. As mentioned, in a study of ligand tautomeric preferences in PDB complexes, the most stable solution phase tautomer (in this case t2) is predominantly the binding state tautomer (depending on the ∆G between the two tautomers) [34]. In agreement with this study, and of particular significance, we observed for other β-D-glucopryranosylazole inhibitors, considering the binding and MM-GBSA ΔGbind values of the most stable (solution phase) free state tautomer gave better agreement with experiment [19]. The ΔGbind value of 4 (t2) is −28.2 kcal/mol (Table 3), much less favourable than that of 3 (t1; ΔGbind = −40.9 kcal/mol), which is consistent with the experimental binding assay results ( Table  1). The binding of t2 is shown in Figure 3C, where the ligand is seen to adopt the expected binding conformation (ω = −177.1°) that does allow for strong ion-ion and hydrogen bonding interactions with Asp283 sidechain carboxylate. However, there is no hydrogen bond with His377 C=O for this tautomer (∆ ( ) = −67.9 kcal/mol). Additionally, the guanidino ligand substituent due to its steric bulk also extends somewhat towards the positively charged side-chain of Lys574, forming unfavourable ion-ion interactions with this group. Correspondingly, the strain energy effects on ΔGbind are considerably more (~ +18 kcal/mol) compared to 3 (~ +8 kcal/mol) or 1 (~ +10 kcal/mol). The strain from Prime MM-GBSA calculations is estimated from a (local) minimization of bound ligand conformation. Likewise, the ligand entropy cost (− ∆ ) on binding for 4 is ~4 kcal/mol more than both 1 and 3, all of which is consistent with its poor observed binding potential.

Analysis of 4.
Results of the DFT calculations for predicted conformations of compound 4 are shown in Table 4. In the case of 4, the azole ring C6 substituent has a resonating +1 charge on N9/N10 atoms and hence, the potential to form strong ion -ion interactions with the Asp283 side-chain carboxylate, as was observed for 1. As with 3, compound 4 can exist in two tautomeric states t1 and t2 (Figure 1). DFT calculations on the different conformations of t1 and t2 revealed tautomer t2 with a dihedral angle (ω = −97.7 • ) as the most favoured tautomeric state and conformation in solution phase (also in the gas phase). In comparison, the lowest energy t1 solution phase conformation (ω = 161.4 • ) is 2.2 kcal/mol higher in energy (or 8.3 kcal/mol in the gas phase with ω = 131.7 • ). These most stable solution phase conformations are shown in Figure 2B. The most stable tautomer t2 has a strong network of intra-molecular hydrogen bonds stabilizing the structure, but for a conformation and tautomeric state that is not consistent with forming desired hydrogen bonding with His377 C=O and ion -ion interactions with Asp283 sidechain carboxylate on binding to GP. As mentioned, in a study of ligand tautomeric preferences in PDB complexes, the most stable solution phase tautomer (in this case t2) is predominantly the binding state tautomer (depending on the ∆G between the two tautomers) [34]. In agreement with this study, and of particular significance, we observed for other β-D-glucopryranosyl-azole inhibitors, considering the binding and MM-GBSA ∆G bind values of the most stable (solution phase) free state tautomer gave better agreement with experiment [19]. The ∆G bind value of 4 (t2) is −28.2 kcal/mol (Table 3), much less favourable than that of 3 (t1; ∆G bind = −40.9 kcal/mol), which is consistent with the experimental binding assay results ( Table 1). The binding of t2 is shown in Figure 3C, where the ligand is seen to adopt the expected binding conformation (ω = −177.1 • ) that does allow for strong ion-ion and hydrogen bonding interactions with Asp283 sidechain carboxylate. However, there is no hydrogen bond with His377 C=O for this tautomer (∆G bind (NS) = −67.9 kcal/mol). Additionally, the guanidino ligand substituent due to its steric bulk also extends somewhat towards the positively charged side-chain of Lys574, forming unfavourable ion-ion interactions with this group. Correspondingly, the strain energy effects on ∆G bind are considerably more (~+18 kcal/mol) compared to 3 (~+8 kcal/mol) or 1 (~+10 kcal/mol). The strain from Prime MM-GBSA calculations is estimated from a (local) minimization of bound ligand conformation. Likewise, the ligand entropy cost ( −T∆S MM ) on binding for 4 is~4 kcal/mol more than both 1 and 3, all of which is consistent with its poor observed binding potential.

Conclusions
The search for potent GP inhibitors acting at the catalytic site has focussed on glucose analogues, with C-β-D-glucopyranosyl azole type inhibitors proving to be among the most successful. Two new analogues of this type, compounds 3 and 4, were synthesized that, in theory, had the potential to exploit key interactions with key binding site residues His377 and Asp283, but only 3 demonstrated a moderate potency. Extensive computations were performed on the free/unbound (Monte Carlo, DFT) and bound state protein-ligand complexes (docking, MM-GBSA) and revealed tautomeric state preferences and ligand strain/reorganization energies as key reasons. We observed that taking the ∆G bind value for the most stable (unbound) state tautomer produced results more in line with the experiment, consistent with a previous work on related analogues [19]; however, this is likely to be system dependent and sensitive to relative tautomeric state stabilities (comparing bound and unbound) [34]. Although some experimental techniques such as neutron scattering can sometimes be employed to determine ligand-bound state tautomeric states, routine X-ray crystallography structures of protein-ligand complexes will not show the H-atom positions [37][38][39]. The importance of careful consideration of ligand tautomeric/ionization state preferences in structure-based inhibitor design using computation was, therefore, highlighted, as well as consideration of tautomeric state conformational preferences (bound versus unbound) that limits the reorganization/strain energy on binding. This information can be exploited in further studies of this type targeting GP, but also other drug targets where ionization/tautomerism of ligand designs plays an important role.

Synthetic Methods
Thin-layer chromatography was carried out on aluminium sheets coated with silica gel 60 F 254 . TLC plates were inspected under UV light and developed by spraying with a staining reagent (5% of cc. H 2 SO 4 and 1% of p-anisaldehyde in EtOH) followed by heating. Column chromatography was performed on silica gel 60 (63-200 µm). Optical rotations were measured using a Perkin Elmer 241 polarimeter. 1 H and 13 C NMR spectra (supplementary materials) were recorded using Bruker DRX 360 or Bruker DRX 400 spectrometers with TMS ( 1 H spectra in CDCl 3 ) or the residual solvent peak ( 1 H spectra in CD 3 OD, 13 C spectra in CDCl 3 and CD 3 OD) as the internal standard. Mass spectra were recorded using a Bruker maXis II UHR ESI-TOF MS spectrometer. Anhydrous THF was distilled from sodium benzophenone ketyl and then, stored over sodium wires. Anhydrous MeOH was prepared by distillation over Mg turnings and iodine. Anhydrous CHCl 3 was dried by distillation from P 4 O 10 , and was then stored over 4Å molecular sieves.
Pyrazole tautomerization results in signal broadening; therefore, the 13 C peaks of the heterocycle (and, in one case, the anomeric carbon of the sugar) cannot be identified in the carbon spectrum. In these cases, HRMS confirms the presence of the pyrazole moiety in the molecules.

Synthetic Methods
Thin-layer chromatography was carried out on aluminium sheets coated with silica gel 60 F254. TLC plates were inspected under UV light and developed by spraying with a staining reagent (5% of cc. H2SO4 and 1% of p-anisaldehyde in EtOH) followed by heating. Column chromatography was performed on silica gel 60 (63-200 µm). Optical rotations were measured using a Perkin Elmer 241 polarimeter. 1 H and 13 C NMR spectra (supplementary materials) were recorded using Bruker DRX 360 or Bruker DRX 400 spectrometers with TMS ( 1 H spectra in CDCl3) or the residual solvent peak ( 1 H spectra in CD3OD, 13 C spectra in CDCl3 and CD3OD) as the internal standard. Mass spectra were recorded using a Bruker maXis II UHR ESI-TOF MS spectrometer. Anhydrous THF was distilled from sodium benzophenone ketyl and then, stored over sodium wires. Anhydrous MeOH was prepared by distillation over Mg turnings and iodine. Anhydrous CHCl3 was dried by distillation from P4O10, and was then stored over 4Å molecular sieves.
Pyrazole tautomerization results in signal broadening; therefore, the 13 C peaks of the heterocycle (and, in one case, the anomeric carbon of the sugar) cannot be identified in the carbon spectrum. In these cases, HRMS confirms the presence of the pyrazole moiety in the molecules.
Prior to use, zinc powder was activated as described in the literature [40]. On a sintered glass funnel, zinc powder (100 mesh) was washed sequentially with 10% w/w aqueous HCl, distilled water, ethanol and diethyl ether and dried in a desiccator over P 2 O 5 .
In a flame dried three-neck round bottom flask, activated Zn powder (2 equiv., 1.65 mmol, 108 mg) and Me 3 SiCl (0.03 equiv., 0.03 mmol, 3 µL) were refluxed in anhydrous THF (4 mL) under argon atmosphere for 25 min. To this boiling suspension, the solution of cyanide (5, 1 equiv., 0.83 mmol, 500 mg) and phenacyl bromide (1.5 equiv., 1.24 mmol, 247 mg) in anhydrous THF (4 mL) was added dropwise in 45 min and the reflux was maintained for another 45 min. After cooling down to room temperature, the solution and the insoluble materials were separated by decantation and the residual solid was washed with THF (3 mL). The combined THF solutions were cooled down to 0 • C, 10% w/w aqueous HCl (4 mL) was added and the solution was stirred for 20 min at this temperature. Water (20 mL) was added and the mixture was extracted with DCM (3 × 20 mL). The combined organic layers were washed with saturated NaHCO 3 , dried over MgSO 4 , filtered and the solvent was removed. The resulting crude product was purified by column chromatography (eluent: hexane/EtOAc 2:1) to yield 223 mg (37%) colourless syrup.

Determination of Inhibitory Constants (K i ) for Glycogen Phosphorylase
Enzyme activity was assayed into the direction of glycogen synthesis as previously presented [22]. Kinetic data were collected using the muscle phosphorylase b (dephosphorylated, GPb) isoform. Kinetic data for the inhibition of GPb by the compounds were obtained in the presence of 10 µg/mL enzyme, varying concentrations of α-D-glucose-1-phosphate (4-40 mM), constant concentration (1%) of glycogen and AMP (1 mM). Enzymatic activities were presented in the form of a double-reciprocal plot (Lineweaver-Burk). The plots were analysed by a non-linear data analysis program. The inhibitor constants (K i ) were determined by secondary plots, replotting the slopes from the Lineweaver-Burk plot against the inhibitor concentrations. The means of standard errors for all calculated kinetic parameters averaged to less than 10% [48,49].