The binding selectivity of vonoprazan (TAK-438) to the gastric H⁺,K⁺−ATPase

Summary

Background

The gastric H⁺,K⁺-ATPase is the preferred target for acid suppression. Until recently, the only drugs that effectively inhibited this ATPase were the proton pump inhibitors (PPIs). PPIs are acid-activated prodrugs that require acid protection. Once acid activated, PPIs bind to cysteines of the ATPase, resulting in covalent, long-lasting inhibition. The short plasma half-life of PPIs and continual de novo synthesis of the H⁺,K⁺-ATPase result in difficulty controlling nighttime acid secretion. A new alternative to PPIs is the pyrrolo-pyridine, vonoprazan (TAK-438), a potassium-competitive acid blocker (PCAB) that does not require acid protection. In contrast to other PCABs, vonoprazan has a long duration of action, resulting in 24 hour control of acid secretion, a high pK_a of 9.37 and high affinity (K_i = 3.0 ηM).

Aim

To determine binding selectivity of vonoprazan for the gastric H⁺,K⁺-ATPase and to explain its slow dissociation.

Methods

Gastric gland and parietal cell binding of vonoprazan was determined radiometrically. Molecular modeling explained the slow dissociation of vonoprazan from the H⁺,K⁺-ATPase.

Results

Vonoprazan binds selectively to the parietal cell, independent of acid secretion. Vonoprazan binds in a luminal vestibule between the surfaces of membrane helices 4, 5 and 6. Exit of the drug to the lumen is hindered by asp137 and asn138 in the loop between TM1 and TM2, which presents an electrostatic barrier to movement of the sulfonyl group of vonoprazan. This may explain its slow dissociation from the H⁺,K⁺-ATPase and long lasting inhibition.

Conclusions

The binding model provides a template for design of novel PCABs.

Background

Inhibition of the gastric H⁺,K⁺-ATPase, the gastric proton pump, is the preferred means of reducing acid secretion for the treatment of peptic ulcer disease (PUD) and gastroesophageal reflux disease (GERD), and is required in combination with antibiotics for eradication of Helicobacter pylori. The currently available drugs that inhibit the H⁺,K⁺-ATPase are proton pump inhibitors (PPIs) such as omeprazole, esomeprazole, lansoprazole, dexlansoprazole, pantoprazole, and rabeprazole. These are acid-activated prodrugs that convert to a thiophilic sulfenic acid and bind covalently to one or more cysteines of the alpha subunit of the ATPase, inhibiting acid secretion. Only active pumps are inhibited, and not all pumps are active at any one time. These drugs are administered before meals and steady state inhibition requires three to five days of once a day dosing. PPIs have a short plasma half-life of about 90 minutes and do not completely inhibit acid secretion because the half-life of the ATPase is about 50 hours ^{1, 2}. Hence, about one third of the pumps are synthesized in 24 hours, so that even with twice a day treatment, newly synthesized pumps will be secreting acid before the next administration of a PPI. This makes control of nighttime acid secretion difficult ³.

An alternative to PPIs, the potassium-competitive acid blockers (PCABs) or acid pump antagonists (APAs), were first discovered when a group of tertiary amines targeted to prevent Ca²⁺ accumulation were shown to inhibit acid secretion in rabbit gastric glands by K⁺-competitive inhibition of the gastric H⁺,K⁺-ATPase and not regulation of intracellular Ca^{2+ 4}. In the context of H⁺,K⁺-ATPase inhibition, the terms inhibitor, blocker and antagonist are synonymous. Subsequently, an imidazo-pyridine, SCH28080, (figure 1) was synthesized, targeted against acid secretion as a possible analog of omeprazole, and shown to be an effective inhibitor of gastric acid secretion ^5–9. Investigation of its mechanism showed that it was a purely K⁺-competitive inhibitor of the gastric H⁺,K⁺-ATPase with a K_i of ~60 nM at pH 7.4 ¹⁰. However, SCH28080 is hepatotoxic and was never developed for clinical use. Its pK_a of 5.5 results in selective accumulation only in acid-secreting parietal cells, where the canalicular pH is ≪5.0, in contrast to other compartments such as lysosomes. Many compounds developed as K⁺-competitive inhibitors were designed with similar pK_a’s to ensure parietal cell selectivity. Several have entered clinical trials such as a pyrimidine, revaprazan ¹¹, and another imidazopyridine, AZD0865 (IC₅₀, 130 ηM) ^11–13. AZD0865 efficacy was equivalent to esomeprazole on once daily dosing, but rapid reversal of inhibition resulted in poor nighttime control of intra-gastric pH ¹². The same is true of revaprazan ¹¹. All of these compounds are acid stable, in contrast to the PPIs; hence, they do not require gastro-protective coating. However, they have a short duration of action and do not control nighttime acid secretion on once a day dosing due to rapid dissociation from the ATPase.

Figure 1.

Structures of specific H⁺,K⁺-ATPase inhibitors, vonoprazan (A), and the [1,2α] imidazopyridine, SCH28080 (B). Vonoprazan retains high affinity at neutral pH and its amino group was therefore protonated for modeling, while its pyridine nitrogen was left unprotonated. SCH28080 binds with highest affinity at low pH and its binding was modeled in the monoprotonated form.

In contrast to the above PCABs, a pyrrolo-pyridine, 1-[5-(2-fluorpphenyl)-1-(pyridine-3-ylsulfonyl)-1H-pyrrol-3-yl]-N-methylmethanamine monofumarate, vonoprazan (TAK-438) (figure 1), is a PCAB with very slow dissociation from the H⁺,K⁺-ATPase ¹⁴, resulting in long-lasting inhibition. This compound has a high pK_a of 9.37 ¹⁵. Vonoprazan is predicted to bind to the pump independent of acid secretion and, like all PCABs, does not require gastro-protective formulation.

The gastric H⁺,K⁺-ATPase is a heterodimer comprised of an alpha and beta subunit. The alpha subunit has 10 transmembrane segments and the beta subunit a single transmembrane segment ¹⁶. The gastric ATPase has luminally accessible cysteines, the targets of the activated PPIs. The K⁺ competitive inhibitors all access the luminal domain and are selective for the H⁺,K⁺-ATPase as compared to the ubiquitous Na⁺,K⁺-ATPase. The slow dissociation of vonoprazan is a key feature of its superior clinical profile ¹⁷. Here molecular modeling was utilized to predict the binding of vonoprazan to the gastric H⁺_,K⁺ ATPase and rationalize its long duration of action as compared to other PCABs such as SCH28080, AZD0865 or revaprazan.

A previous study investigated the binding of omeprazole to resting or stimulated parietal cells in isolated rabbit gastric glands ¹⁸. As expected, binding of omeprazole to the parietal cells of rabbit gastric glands increased when acid secretion was stimulated and decreased with inhibition of secretion.

The aim of the current study was to investigate binding selectivity of vonoprazan for the gastric H⁺,K⁺-ATPase and explain its slow dissociation. Binding of [¹⁴C]-vonoprazan was studied by autoradiography after uptake in resting versus stimulated rabbit gastric glands. The high pK_a of vonoprazan contrasts with the pK_a’s of other PCABs (~ 5.0), which prevents accumulation in less acidic compartments than the secretory canaliculus of the parietal cell. In contrast to omeprazole, vonoprazan binding was independent of the secretory status of the parietal cell. Molecular modeling of the H⁺,K⁺-ATPase was used to describe the very slow dissociation of vonoprazan as compared to other PCABs such as SCH28080, AZD0865 or revaprazan. It appears that vonoprazan occupies the same luminal vestibule of the pump as SCH28080, but its sulfonyl group hinders exit from the pump.

Materials and Methods

Vonoprazan uptake into resting and stimulated rabbit gastric glands

Rabbit gastric glands were isolated as previously described ¹⁹. Briefly, male New Zealand White rabbits (Charles River Laboratories, San Diego, CA, USA) were euthanized with Fatal Plus (Vortech Pharmaceuticals, Dearborn, MI, USA), then the stomach was perfused and removed. The stomach was cut along the greater curvature and the cardiac and antral regions removed. The remaining corpus and fundus with oxyntic mucosa was washed with phosphate buffered saline and separated from the underlying muscle by scraping. The mucosa was minced using scissors, washed three times with gastric gland medium (NaCl, 140mM; MgSO₄ • 7H₂O,1.2mM; CaCl₂ • 2H₂O, 1.0mM; HEPES,10mM; d-glucose, 11.10mM; BSA, 2.0g/l) (Sigma-Aldrich, St Louis, MO, USA), then subjected to collagenase B (0.9mg/ml) (Roche Diagnostics Corporation, Indianapolis, IN, USA) digestion for 35 minutes at 37°C with rotation (200 RPM). Gastric glands were isolated from the digestion mix by washing three times at 1 × g with gastric gland medium. Isolated glands were placed in gastric gland medium containing 1 μCi [¹⁴C]-vonoprazan (Takeda Pharmaceutical Co., Osaka, JP), 10 μM histamine (Sigma-Aldrich, St. Louis, MO, USA) and 10 μM dbcAMP (Sigma-Aldrich, St. Louis, MO, USA) to stimulate acid secretion or gastric gland medium containing 1 μCi [¹⁴C]-Tak-438, 10 μM atropine (Sigma-Aldrich, St. Louis, MO, USA) and 100 μM famotidine (Sigma-Aldrich, St. Louis, MO, USA) to inhibit acid secretion. Aliquots (1 ml) were removed at 0, 5,10, 20 and 30 minutes and fixed with 2% glutaraldehyde (Ted Pella Inc., Tustin, CA, USA) for 1 hour at 37°C. Glands were pelleted by centrifugation (5 minutes at 8000 × g) and washed three times with gastric gland medium. Gland pellets were dried at 65°C overnight, weighed, and 100 μl of 1M NaOH was added to dissolve the pellet prior to scintillation counting. HiIonicFluor scintillation fluid (Perkin-Elmer Inc., Waltham, MA, USA) was added to the dissolved gland pellet and radioactivity measured using a Wallac scintillation counter (Perkin-Elmer Inc., Waltham, MA, USA). Results were expressed as DPM/mg dry weight.

Labeling of resting and stimulated parietal cells by [¹⁴C]-vonoprazan

Gastric glands were isolated from the oxyntic region of the rabbit stomach as described above ¹⁹. Isolated gastric glands were divided into two groups. One group (acid-stimulated) was treated with 10 μM histamine and 10 μM 3-Isobutyl-1-methylxanthine (IBMX, Sigma-Aldrich, St. Louis, MO, USA) and the other group (resting) with 100 μM cimetidine (Sigma-Aldrich, St. Louis, MO, USA ) and 10 μM atropine. [¹⁴C]-labeled vonoprazan was added to both conditions. Aliquots were removed at 0 and 30 minutes, briefly centrifuged to pellet the gastric glands, and the supernate removed by aspiration. O.C.T. compound (Sakura Finetek, USA, Inc., Torrance, CA, USA), was added to the gland pellet and frozen on dry ice. Five μm sections of the frozen glands were obtained using a cryo-microtome and collected on glass slides. The slides were covered with Ilford L4 emulsion (Polysciences Inc., Warrington, PA, USA) and stored at 4°C for 30 days. The slides were developed according to the manufacturer’s instructions. The slides were stained with hematoxylin and eosin. After development and staining, digital dark-field photomicrographs were obtained to quantify the silver grains and bright-field photomicrographs were collected to identify parietal cells. Silver grains corresponding to the location of [¹⁴C]-labeled vonoprazan over parietal cells were quantified using NIH Image J software.

Modeling of vonoprazan binding to the H⁺,K⁺ ATPase

Modeling

Modeling procedures were performed with Discover and InsightII molecular modeling software (Laboratory for Molecular Simulation, Texas A&M University, College Station, TX, USA). Minimizations were executed to an average absolute derivative less than 0.0001 kcal (mole Å⁻¹) with the cvff force field, a dielectric constant of 15, and a non-bonded cutoff of 25 Å, as previously described ²⁰. The structures of SCH28080, and (figure 1) were optimized in vacuo by using the semi-empirical AM1 method within the Ampac/Mopac module of the InsightII software. Connolly surface areas were calculated with a probe radius of 1.4 Å. AutoDockVina (Molecular Graphics Lab, The Scripps Research Institute, La Jolla, CA, USA) was downloaded and implemented within the Microsoft Windows environment ²¹ using an exhaustiveness of 35 and a search box with dimensions 28x26x28 Å centered on the side chain oxygen of tyr799.

Biochemical Procedures

Mutagenesis, clonal selection, membrane isolation, and ATPase assays utilizing NH₄+ as the activating ligand were performed by the methods previously described ^{10, 20}. LogD calculations were made by using the ACD/Labs Percepta Predictors software module (ACD/Labs, Toronto, CAN). Images were prepared using UCSF Chimera software (Resource for Biocomputing, Visualization, and Informatics, University of California, San Francisco, USA).

Results

Vonoprazan binds equally to resting and stimulated rabbit oxyntic gastric glands

In contrast to [¹⁴C]-omeprazole, an acid activated inhibitor of the gastric H⁺,K⁺-ATPase, where labeling was greater in stimulated than in resting rabbit gastric glands ¹⁸, labeling by [¹⁴C]-vonoprazan of rabbit gastric glands was independent of the secretory status (figure 2).

Figure 2.

Vonoprazan binding to resting and stimulated rabbit oxyntic gastric glands. Labeling by [¹⁴C]-vonoprazan was independent of the secretory status of gastric oxyntic glands, indicating that inhibition of acid secretion by vonoprazan was immediate since there was no delay due to requirement for acid secretion. Acid secretion was stimulated using histamine and dbcAMP (red) and inhibited by atropine and famotidine (blue).

Vonoprazan labels both active and inactive H⁺,K⁺-ATPase

Autoradiographic analysis of [¹⁴C]-vonoprazan labeling of resting and stimulated gastric parietal cells found no difference in labeling. Labeling was independent of the secretory state, showing that vonoprazan does not require the acid activation required for PPIs ¹⁸ (figure 3). Figure 4A and 4B are representative resting and stimulated parietal cells, respectively, stained with hematoxylin and eosin (H & E) before autoradiography. The parietal cell resting morphology is defined by numerous cytoplasmic, H⁺,K⁺-ATPase containing tubulo-vesicles (figure 4A), while the stimulated parietal cell has a markedly expanded H⁺,K⁺-ATPase-containing canalicular membrane (figure 4B, blue arrows). Figures 4C and 4D are photomicrographs of vonoprazan labeled and H & E stained cells. Figures 4E and 4F are darkfield images showing that the number of silver grains decorating the cytoplasm of acid-inhibited and acid-stimulated parietal cells are the same, indicating vonoprazan labels both active and inactive H⁺,K⁺-ATPase. The selective labeling of the parietal cell shows that it is the high affinity binding of vonoprazan that determines its binding selectivity and not the pH of the canalicular compartment.

Figure 3.

Vonoprazan labeling of resting and stimulated rabbit parietal cells. Autoradiographic analysis of [¹⁴C]-vonoprazan labeling of resting and stimulated gastric parietal cells found no significant difference in labeling (Ttest, p=0.21). Digital darkfield photomicrographs were obtained to quantify the silver grains decorating parietal cells. Silver grains corresponding to the location of [¹⁴C]-vonoprazan were quantified using NIH Image J software. Results were reported as mean pixel densities of labeled cells, n=9 for resting cells, n=11 for stimulated cells.

Figure 4.

Autoradiographic analysis of [¹⁴C]-vonoprazan labeling of resting and stimulated gastric parietal cells. Figure 4A and 4B are representative resting and stimulated parietal cells, respectively, stained with H & E before autoradiography. The secretory canaliculus of the stimulated cell is highlighted by blue arrows. Figures 4C and 4D are photomicrographs of vonoprazan labeled and H & E stained. Figures 4E and 4F are darkfield images showing that the number of silver grains decorating the cytoplasm of acid inhibited and acid stimulated parietal cells are equal, indicating that vonoprazan labels both active and inactive H⁺,K⁺-ATPase.

Modeling of inhibitor binding to the E₂ conformation of the H⁺,K⁺-ATPase

The purpose of molecular modeling studies of the H⁺,K⁺-ATPase and PCAB binding is to better understand the clinical functionality and benefits of these compounds. Previously, a homology model of the gastric H⁺,K⁺-ATPase, generated from the E₂ • thapsigargin conformation (E₂[TG], pdb.1iwo) of the SERCA-ATPase, was used to construct a binding site model for the K⁺ competitive [1,2 α] imidazopyridine inhibitor, SCH28080 ²². Recently, a new homology model of the gastric H⁺,K⁺-ATPase determined by cryo-EM has been reported ²³. The structure, 4ux2.pdb, was generated by modifying the molecular structure of the Na⁺,K⁺-ATPase pdb.3a3y, to fit a low resolution electron density map (7 Å) based on cryo-EM of the H⁺,K⁺-ATPase in the presence of SCH28080. Unfortunately, the resolution of the map was insufficient to identify the molecular structure of the bound inhibitor.

Improved binding site models for both SCH28080 and vonoprazan are presented, using the recent gastric ATPase structure (4ux2.pdb) as a template and the methods described previously ²⁰. Briefly, the structure of SCH28080 was optimized in vacuo for binding to the new model, then placed within the luminal vestibule of the alpha subunit formed by transmembrane helices 1, 2, 4, 5, and 6, arranged to maximize contacts with side chains whose mutations are known to affected the affinity for SCH28080 ^{10, 20}, oriented to account for the known structure activity relationships (SAR) of the [1,2 α] imidazopyridines ²⁴, and optimized by energy minimization (figure 5A). The phenyl group of SCH28080 had been shown to bind between TM1 and TM2 by photoaffinity labeling ⁷ and this observation was used to orient the molecular structure of SCH28080 within the model prior to energy minimization.

Figure 5.

(A) Luminal view of the surface (blue) of the vestibule formed by TM1 (blue), TM2 (light blue), TM3 (green), TM4 (light green), TM5 (yellow), and TM6 (orange). A slot-like cavity at the bottom of the vestibule encloses bound SCH28080 (stick, carbons and surface in orange). The color scheme for the transmembrane helices is maintained in all figures. Methyl (red arrow) and cyanomethyl substituents (orange arrow) of the imidazopyridine ring are buried next to TM4 and TM5, respectively, while the outer edge of the imidazopyridine ring (blue arrow) and oxymethylene bridge (green arrow) are exposed to the solvent. TM3 and the remainder of the membrane domain including TM7, TM8, TM9, TM10, and the beta subunit are omitted for clarity in all figures. TM4 is labeled on both sides of its nonhelical segment near the middle of the membrane. (B) Lowest energy binding site conformation for SCH28080 (stick, carbons in cyan) found by AutoDockVina compared to the one based on biochemical data and energy minimization (stick, carbons in orange). tyr799 and cys813 (ball and stick, carbons in cyan) are shown for reference. Positions of permissible substituent addition on the imidazopyridine ring (left green star) and of the diol bridge (right green star) are shown between aromatic rings in the imidazopyridines. Substituents are size-restricted at the buried edge of the imidazopyridine ring (maximum size is an ethyl group in the R3 position, lower red star and methyl in R2, upper red star).

The results show a luminal-oriented vestibule within the H⁺,K⁺-ATPase roughly in the shape of a funnel with the wide end open to the lumen and a slot at the bottom into which the inhibitor fits (viewed from the lumen in figure 5A). The cyanomethyl group of SCH28080 (orange arrow) fits tightly next to TM5, the methyl group (red arrow) inserts next to TM4, the imidazo-pyridine points toward the TM5/TM6 loop with its plane roughly perpendicular to the membrane, and the phenyl moiety is between TM1 and TM4 with the plane of the ring facing TM2. Analysis of the Connolly surface, or the surface inaccessible to a potential solvent, of unbound SCH28080 (313.7 Ų) compared to the inhibitor in the binding site (58.0 Ų) shows that most of the bound surface (255.7 Ų, 81.5%) is inaccessible to the solvent. The edge of the imidazopyridine ring (blue arrow) and the oxymethylene bridge between the aromatic rings are exposed (green arrow). This explains the prior laboratory finding that modifications of amino acids in these sites that do not result in loss of inhibition of SCH28080 ²⁴ and confirm that these sites are not clinically relevant for inhibitor function.

Among the key results reported earlier, using studies involving point mutations of specific amino acids in transmembrane segments of the H⁺,K⁺-ATPase, were the complete loss of inhibition in the ala335cys mutant and 60-fold loss of affinity in the tyr799ser mutant ^{20, 25}. In contrast, there was no loss of inhibition for tyr799phe, a mutation which, unlike the change to serine, maintains the phenol group found in tyrosine. These results show the importance of the phenyl group of tyr799 (TM5) and the methyl group of ala335 (TM4) for binding of SCH28080. These side chains are spaced closely together on one side of the vestibule and apparently present a surface essential for SCH28080 interaction.

The predicted conformation for SCH28080 and its orientation in the vestibule was tested by using an unbiased ligand-docking algorithm, AutoDockVina ²¹. The best binding pose for SCH28080 found by AutoDockVina (figure 5B, stick form with carbons in blue) closely aligned to that predicted from the biochemical data using point mutations of specific amino acids (stick, carbons in orange). The binding site model is also consistent with the effects of modification of the parent structure that either gives loss (2 and 3 positions on the imidazopyridine ring, red stars) or retention of inhibition (green stars) in drugs such as soraprazan ²⁶. Using molecular modeling of SCH28080 binding in combination with mutations of specific amino acids shown to have an effect in laboratory studies allows for use of these mutations in modeling of the structurally similar vonoprazan, for the purpose of understanding its clinical utility.

In vacuo optimization of vonoprazan was found to give a surface envelope similar to that of SCH28080. The possibility that the two compounds share binding site amino acid residues was tested by analyzing the kinetics of inhibition for vonoprazan with wild type H⁺,K⁺-ATPase and mutants ala335cys, try799ser, and tyr799phe (Table 1). Kinetic analysis of the wild type found the K_i for vonoprazan to be 3.0 ± 0.3 ηM. The dramatic decrease in inhibitor efficacy for the try799ser mutant (>100 fold) compared to only 2-fold for tyr799phe shows the importance of the phenyl side chain for binding vonoprazan and that the hydroxyl group of tyr799 is not essential. Similarly, vonoprazan inhibition was undetectable for the ala335cys mutant (Table 1). These results support the importance of a binding surface for vonoprazan formed by tyr799 and ala335 as found for SCH28080. The optimized structure of vonoprazan was therefore oriented to match the bound surface envelope of SCH28080 then merged to the protein and the complex energy minimized to an average absolute derivative less than 0.0001 kcal (mole Å⁻¹). During the minimization the beta subunit and the backbone of the alpha subunit were kept fixed except for TM1 to TM6. The backbone atoms of TM1 through TM6 (but not their luminal loops which were unrestrained) were instead tethered to their original positions to allow minimal deviation from the 4ux2.pdb coordinates. Tethering was accomplished by adding an energy displacement term to the cvff force field (spring constant 50 kcal/Å). All side chains and vonoprazan were unrestrained during the minimization to optimize their orientations. The tethered backbone atoms were displaced during optimization with an average RMS deviation of 0.4 Å, showing only a very small deviation from the template was required to obtain a favorable fit to the binding site. These modifications, based on the known effect of specific mutations of the H⁺,K⁺-ATPase on SCH28080 binding and software based predictions of bioenergetically feasible conformational changes in the transporter allow for prediction of the binding site and orientation of vonoprazan, which was then used for further studies.

Table 1.

Kinetic parameters for mutants of the H⁺,K⁺ ATPase affecting the K_i for vonoprazan

Location	Mutant	Ki [vonoprazan] nM	Type of inhibition	Vmaxa μmol/mg/hr	Km,app [NH 4+] mM +SE
	Wtb	3.0 ± 0.3	competitive	88.0 ± 6.0	2.3 ± 0.1
TM4	A335Cc	> 500	none	99	1.5 ± 0.2
TM5	Y799F	7.0 ± 2.0	competitive	109	0.9 ± 0.2
TM5	Y799Sd	410	competitive	11	2.1 ± o.2

^aV_max was normalized to protein expression. Standard errors for the V_max and K_i calculated from assay data were proportional to those given for the K_m,app.

^bDifferent standard preparations of porcine H⁺,K⁺-ATPase vary with respect to V_max.

^cKinetic analyses over a range of ammonia concentrations resulted in no inhibition by vonoprazan for the A335C mutant. The K_i was estimated to be greater than 500 nM for this mutant based on the absence of inhibition by 25 nM vonoprazan in the presence of 1 mM NH₄⁺.

^dAsano et. al. reported a 60–80% lower affinity for the alanine and serine mutants of Y799²⁵.

A surface rendering (Figure 6A) of bound vonoprazan shows the inhibitor is largely occluded by the H⁺,K⁺-ATPase once it is bound, with the methyl amino group next to TM5, the pyridine moiety between TM1 and TM4, and the fluorophenyl pointing toward the vesicle opening near TM6. Connolly surface analysis of inaccessible surface area of vonoprazan showed the bound inhibitor has 89.2% of its total surface area (299.6 out of 336.0 Å2) inaccessible in the bulk solvent. Therefore, vonoprazan is predicted to bury a larger surface area and a larger percentage of its surface than SCH28080. Clinically, this finding could at least partially explain why dissociation of vonoprazan from the H⁺,K⁺-ATPase is slower than SCH28080.

Figure 6.

(A) Surfaces of the luminal vestibule (blue) and vonoprazan (stick, carbons in orange). The inhibitor is proposed to fit in the slot at the bottom of the vestibule as suggested for SCH28080 with the methylamino group (red arrow) next to TM5, the sulfonyl group (yellow arrow) facing TM2 near the start of TM6, and the fluorophenyl ring between TM5 and TM6 (helices colored as in figure 2). Only the outer edge of the fluorophenyl ring (center) is exposed to the solvent. (B) Proposed binding site for vonoprazan (surface and stick with carbons in orange) with residues whose mutation affect binding (ball and stick) in cyan. The inhibitor fits in the loop between TM5 and TM6 (light orange and dark orange ribbons, respectively) on one end and TM1 (blue ribbon) and TM4 (green ribbon) on the other (TM3 omitted). Mutation of tyr799 or ala335 (ball and stick) on one side of the bound inhibitor severely affect binding. Methoxy substitution in the 6 position of the pyridine ring (red circle) has been reported to give retention of high affinity H⁺,K⁺-ATPase inhibition while replacement of fluorine in the 2 position of the phenyl ring (blue circle) with methoxy gives a loss of more than 100-fold in affinity. (C) A nearly identical pose for vonoprazan (stick, carbons in blue) docking is found by an unbiased scoring algorithm (AutoDockVina). As found for the energy minimized conformation in 6B, the inhibitor is oriented in the binding space with the methylamino group in next to an expanded turn in TM5 produced by pro798 and hydrogen bonding to the backbone carbonyl of glu795 (blue arrow). The sulfonyl oxygens face the sulfhydryl hydrogen from cys813. Asp137 and asn138 present a likely electrostatic barrier to the sulfonyl oxygens (red arrow) of the inhibitor for movement toward the open end of the vestibule and exit to the lumen (green arrow).

Analysis of the detailed vonoprazan binding site interactions show the pyrrole and fluorophenyl rings stacked against the side chain of tyr799 while ala335 fits in the curvature between the pyrrole and pyridine rings in a manner analogous to the imidazopyridine ring of SCH28080 (figure. 6B). This docking mode accounts for the severe effects of mutating these two residues. The opposite surface of the fluorophenyl ring faces asp137, and asn138 (figure 6B, black arrows), leu141, and cys813 with the sulfur of cys813 oriented toward the two sulfonyl oxygens within distances of 3.05 Å and 4.40 Å, respectively. The edges of the pyrrole and fluorophenyl rings are enclosed by ile816 and the TM5/TM6 loop from leu809 to cys813. The effect is to pin the pyrrole and fluorophenyl rings tightly in the space between tyr799 and ala335 on one side and asp137, asn138, leu141, and cys813 on the other. The lack of free movement and close proximity to cys813 suggests the specificity of vonoprazan for the gastric H⁺,K⁺-ATPase is partially due to the replacement of cysteine at this position with the beta-branched side chain of threonine in the Na⁺,K⁺-ATPase. The methyl moiety fits in the space between ala335 and ala339 with the amino group facing TM5 and hydrogen bonded to the backbone carbonyl of glu795 (figure 6C, blue arrow). Space for the buried methylamino group between TM4 and TM5 (figure 6B, red text) is given by an expanded turn in TM5 before pro798 (figure 6C), which also makes the backbone carbonyl of glu795 available for hydrogen bonding. One of the K⁺ binding sites is in the vicinity of glu795 ²⁷. The orientation of the pyridine ring closely matches that of the phenyl ring of bound SCH28080 and aligns between cys120 (TM1), leu141 (TM2), and met334 (TM4), the latter of which is substituted with isoleucine in the Na⁺,K⁺-ATPase and could also play a role in inhibitor selectivity. There are differences in the two ion pumps, however, that imply that more subtle changes in conformation could affect binding that are not likely to be a function of the residues in direct contact with the inhibitor. Allosteric effects on SCH28080 affinity originating at the ion site have been shown by mutation of lys791 to serine, the amino acid substituted in the Na⁺,K⁺-ATPase. A 20-fold decrease in affinity was observed for this mutant, suggesting an indirect effect on the structure of the vestibule. A similar mechanism could affect vonoprazan binding or exit from the vestibule. These modeling studies may at least partially explain the specificity of vonoprazan for the H⁺,K⁺-ATPase and the reversibility of inhibitor binding, which have significant implications for the clinical use of this drug.

The predicted position and orientation for vonoprazan was tested with AutoDockVina as done for SCH28080. All of the unfixed bonds of the inhibitor were allowed free rotation during the docking analysis. A very similar position was found by AutoDockVina as the best binding pose for vonoprazan, although the pyridine ring was rotated approximately 180° (figure 6C, stick, carbons in light blue). This finding supports the orientation and placement of vonoprazan derived from both the experimental results and the similarity of its surface envelop compared to that of SCH28080. In summary, this molecular model is clinically relevant because it provides a structural explanation for the specificity, long duration of action, and slow reversibility of inhibitor binding, all factors significant for understanding of the mechanism of action of a drug with promise for improved clinical efficacy.

Discussion

The burden of acid-related diseases is clinically significant worldwide, and inhibition of acid secretion is the major unifying therapy for these conditions ²⁸. While PPIs have gained global acceptance for their role in treating acid-related diseases, there are limitations to their use. PPIs require specific timing of doses with meals, take several days to reach full effect due to the need to reach a steady state of pump turnover and inhibition, are not ideal for preventing nighttime acid breakthrough, and have inconsistent effects based on individual patients’ CYP2C19 status ^{29, 30}. The PCABs have been studied extensively as a potential alternative to overcome the limitations of available PPIs. PCABs provide fast onset of action, do not require specific timing with food, and are not metabolized by CYP2C19, allowing for less variability between patients ^{30, 31}. The specific benefits of vonoprazan compared to other drugs in this class include slow dissociation from the H⁺K⁺-ATPase, providing longer duration of action, and lack of the hepatotoxicity, due to absence of the imidazopyridine ring, that has limited the clinical application of other PCABs ^{12, 15, 32, 33}. Understanding the mechanism of binding of vonoprazan to the H⁺K⁺-ATPase provides novel insight into its clinical efficacy and a framework for continued development of improved compounds for treatment of acid-related disease.

As shown here for both gastric glands and parietal cells, vonoprazan binding is independent of acid secretion and is quite evident after 30 minutes. This is in contrast to omeprazole binding, that is increased by 50% in acid secreting parietal cells when compared to non-secreting parietal cells ¹⁸. In addition, it is clear that the uptake seen in gastric glands is due to binding to the H⁺,K⁺-ATPase and not due to pH dependent accumulation in the active secretory canaliculus. Accumulation and acid activation are required for the action of the PPIs but are not required for the action of vonoprazan. Onset of inhibition of acid secretion is therefore effective for both active and resting pumps and does not change with repeated dosing ¹⁷. Further, since vonoprazan is acid stable, no gastro-protective coating is required. Human testing has shown that once daily dosing with 40 mg is sufficient to maintain intragastric pH > 5.0 over 24 hours ¹⁷, a level of inhibition not achieved by b.i.d. dosing with any PPI ³⁴.

A binding site orientation for vonoprazan has been proposed previously that differs significantly from the one predicted here ³⁵. In the reported binding simulation, an initial homology structure for the H⁺,K⁺-ATPase was based on the Na⁺,K⁺-ATPase in the E₂P state, pdb.2zxe, then used for molecular dynamics in the presence of various inhibitors including vonoprazan. Each of the final structures showed a large RMS deviation (~3.0 Å) in the overall backbone coordinates compared to the initial structure and an even greater deviation in the luminal vestibule, whose confirmation was further dependent on the inhibitor used for the simulation. The final structure for protonated vonoprazan showed only peripheral interaction of ala335 and tyr799 with the distal end of the pyridine ring, thus not accounting for the dramatic effect of mutating these residues.

The effects of various additions or substitutions on the pyridine and fluorophenyl rings have been reported ³². The results given by these modifications can be interpreted by this binding model. In particular, substitution by methoxy in the 6 position of the pyridine ring (figure 6B, red circle) gives retention of inhibition, and there is space to tolerate this modification in the proposed binding mode. In contrast, there is more than a 100-fold loss of affinity given by replacement of the 2-fuoro group with 2-methoxy, and the model suggests the reason is likely steric conflict with L809 (figure 6B, blue circle).

SCH28080 shows 20-fold lower apparent binding affinity than vonoprazan and a much more rapid rate of dissociation ^{8, 15}. The binding affinity predicted by AutoDockVina for vonoprazan was only 2-fold greater than for SCH28080, and the four-fold difference in the calculated logD values for SCH28080 and vonoprazan (2.69 and 0.56, respectively) also do not support the difference in hydrophobicity as an explanation for the higher affinity of vonoprazan. The occluded appearance of the bound inhibitors (figures 5 and 6) suggests a possible mechanism in which exit from the binding site is more hindered for vonoprazan. A higher total surface area and a higher percentage of vonoprazan is buried compared to SCH28080, showing that vonoprazan fills the vestibular space more efficiently that SCH28080. In addition, its tight curvature around ala335 (figure 6B), hydrogen bonding of its methylamino group to the backbone of TM5, and energetically unfavorable passage of the sulfonyl group across the electrostatic barrier presented by asp137 and asn138 (figures 6b and 6c) all contribute to the slower rate of dissociation measured for vonoprazan compared to SCH28080 and account for the higher affinity. The presence of the sulfonyl group suggests that ala137 and asn138 present a more significant barrier to dissociation of vonoprazan as compared to SCH28080. This model will be useful in designing other K⁺ competitive inhibitors of the gastric ATPase and predicting selectivity as compared to the Na⁺,K⁺ or SERCA ATPases.

Development of a more useful and accurate model for binding of vonoprazan is timely since the medication is coming into clinical use ³⁶. Safety in humans was confirmed by phase I studies conducted in Japan and the United Kingdom, looking both at single dose and consecutive dose regimens, with minimal adverse effects noted and lack of sensitivity to CYP2C19 polymorphism ^{17, 37}. In line with the data shown here regarding binding and dissociation of vonoprazan, onset of action was seen within 24 hours, pH >4 was sustained over the treatment period in most patients including overnight, and tolerance to the drug was not seen ^{17, 37}. Vonoprazan provided more rapid and sustained inhibition of acid secretion when compared directly with two different PPIs in a Japanese population with rapid PPI metabolizer genotype ³⁸. Vonoprazan was also shown to be clinically efficacious and non-inferior to PPIs for the treatment of erosive esophagitis, a condition that requires sustained acid suppression for healing ^{39, 40}. The acid-independent binding of vonoprazan and the slow dissociation as shown here by labeling of rabbit gastric glands and molecular modeling demonstrate the mechanisms behind these promising clinical results. The relative efficacy of similar PCABs suggests that, provided safety concerns are met, PCAB treatment of gastric acid related diseases might replace the use of PPIs.

Abbreviations

TAK-438

1-[5-(2-fluorophenyl)-1-(pyridin-3-ylsulfonyl)-1H-pyrrol-3-yl]-N-methylmethanamine monofumarate

SCH28080

3-cyanomethyl-2-methyl-8-(phenylmethoxy)imidazo[1,2α] pyridine