Anagliptin, a potent dipeptidyl peptidase IV inhibitor: its single-crystal structure and enzyme interactions

Abstract

The single-crystal structure of anagliptin, N-[2-({2-[(2S)-2-cyanopyrrolidin-1-yl]-2-oxoethyl}amino)-2-methylpropyl]-2-methylpyrazolo[1,5-a]pyrimidine-6-carboxamide, was determined. Two independent molecules were held together by intermolecular hydrogen bonds, and the absolute configuration of the 2-cyanopyrrolidine ring delivered from l-prolinamide was confirmed to be S. The interactions of anagliptin with DPP-4 were clarified by the co-crystal structure solved at 2.85 Å resolution. Based on the structure determined by X-ray crystallography, the potency and selectivity of anagliptin were discussed, and an SAR study using anagliptin derivatives was performed.

• Co-crystal structure
• DPP-4 inhibitor
• transition-state interaction
• type 2 diabetes
• X-ray crystallography

Introduction

In 1966, dipeptidyl peptidase IV (DPP-4, EC 3.4.14.5) was discovered as a new aminopeptidase that hydrolyses glycyl-dl-prolyl-β-naphthylamide1. In 1993, one of the substrates of this enzyme was reported to be glucagon-like peptide-1 (GLP-1)2. GLP-1 is a peptide stimulating insulin secretion depending on the blood glucose concentration level, and is inactivated by cleavage of two N-terminal amino-acid residues by DPP-4. Accordingly, DPP-4 inhibitors can function as diabetes medicines by minimizing the reduction in GLP-1 concentration. The effects of DPP-4 inhibition via the inhibitor val-pyrrolidine have been investigated in anesthetized pigs3, and a perspective on the use of DPP-4 inhibitors for the treatment of type 2 diabetes was reported in 19984. DPP-4 consists of 766 amino acids and belongs to the prolyloligopeptidase family of serine proteases; its crystal structure was revealed in 20035–9.

The DPP-4 crystal structure forms a homodimer, suggesting that it is an in vivo form. Each subunit consists of two domains, an α/β-hydrolase domain and an eight-bladed β-propeller domain. The active site was revealed by observation of the tetrahedral intermediate of a peptide inhibitor, diprotin A (Ile-Pro-Ile)7. The catalytic triad including catalytic residue Ser630 is located at the interface of the two domains, and the S₁ subsite is formed by the hydrophobic residues. The N-terminal amino group of the substrate is recognized in the S₂ subsite that consists of the acidic residues Glu205 and Glu206.

In a subsequent report, Engel et al. focused on whether the substrates approached the active site by opening a side or a propeller region. They found that a substrate approaching via the side opening would be in the proper orientation for docking to the active site10.

Until recently, various inhibitors of DPP-4 were developed by chemical design and synthesis, as well as by biological evaluation with consideration for hypoglycaemic activity and – as a safety concern – with consideration for selectivity against DPP-4-related proteases such as DPP8, DPP9 and fibroblast activation protein (FAP). These efforts resulted in numerous inhibitors; the eight inhibitors shown in Figure 1 are currently available for use11. Sitagliptin was the first of these inhibitors to be approved, gaining authorization in 2006 in the US and in 2007 in the EU. Vildagliptin was approved next, in 2008 in the EU. Subsequently, saxagliptin (in 2009), linagliptin (in 2011) and alogliptin (in 2013) were approved in both the US and EU, respectively. Gemigliptin was approved in Korea in 2012, and teneligliptin and anagliptin were approved in Japan in 2012. With the exceptions of gemigliptin and anagliptin, Nabeno et al. categorised all of these inhibitors into three classes according to the binding subsites of DPP-4, which tend to be characteristic of their chemical structures12. Specifically, the inhibitors binding to only the S₁ and S₂ subsites were categorised into Class 1, those binding to the S₁′ and/or S₂′ along with the S₁ and S₂ subsites were categorised into Class 2, and those binding to the S₂ extensive subsite along with the S₁ and S₂ subsites were categorised into Class 3. The S₂ extensive subsite was defined as the site beyond the S₂ subsite by Yoshida et al13; natural substrates do not interact with the subsite.

Figure 1. Launched DPP-4 inhibitors.

Anagliptin14 is one of the DPP-4 inhibitors launched in Japan. In this paper we report its crystal structure and its interactions with DPP-4 as determined by X-ray crystallography, and discuss its potency and selectivity.

Materials and methods

Compounds

Anagliptin14, sitagliptin15, vildagliptin16 and alogliptin17 were prepared by known methods, respectively.

Crystallographic study of anagliptin

A colourless single crystal (0.3 × 0.3 × 0.08 mm³) was obtained by standing at room temperature after heating to dissolve in ethyl acetate. The data were collected by the ω scan method with 0.3° in scan width and 10 s in exposure time at room temperature, using a SMART 1000 two-dimensional X-ray diffractometer (Bruker; Mo Kα, 50 kV, 40 mA) with a 0.5-mmϕ collimator. The structure was solved by the direct method using the SHELXS-97 program (Göttingen, Germany), and refined by the full-matrix least-squares method using the SHELXL-97 program. The crystal structure data has been deposited to the Cambridge Crystallographic Data Centre (CCDC 1003533).

Expression and purification of DPP-4

Human DPP-4 (hDPP-4) was expressed in the Silkworm–Baculovirus System produced by ProCube (Sysmex Corporation; http://procube.sysmex.co.jp/eng/). The hDPP-4 gene encoding Ser39-Pro766 was amplified by PCR from the synthesised full-length hDPP-4 gene template with optimised codon usage for Bombyx mori. The PCR was run using a forward primer containing a BglII site and a cleavage site for HRV 3C protease (5′-GGAAGATCTCTGGAAGTTCTGTTCCAGGGGCCCTCAAGGAAGACTTATACATT-3′) and a reverse primer containing an EcoRV site (5′-GGCGATATCTTATGGTAAACTGAAACATTG-3′). The PCR product was inserted between BglII and EcoRV in the multiple cloning site of the pM24 vector (Sysmex Corporation) based on the vector pUC19 for B. mori nucleopolyhedrovirus (BmNPV). The vector was designed to express a fusion protein containing a DDDDK-tag and HRV3C protease recognition site at the N-terminal fragment of hDPP-4. This transfer vector was co-transfected with BmNPV DNA into a B. mori-derived cell line (BmN). After 7 days of incubation, the recombinant baculovirus was injected into the body of silkworm pupae. The pupae were harvested 6 days after infection, frozen and homogenised with PBS buffer (pH 7.4) containing a protease inhibitor cocktail. The homogenised solution was centrifuged to separate the supernatant and precipitate using an ultracentrifuge (100 000 g) for 1 h at 4 °C. The supernatant containing DDDDK-tag fusion hDPP-4 was applied to DDDDK-tagged Protein PURIFICATION GEL (MBL Co., Ltd.), and the fusion protein was eluted with DDDDK peptide containing PBS buffer (pH 7.4). The eluted sample was treated with HRV 3C protease (Wako Pure Chemical Industries, Ltd.) for 16 h at 4 °C to remove the N-terminal DDDDK-tag from the recombinant fusion protein and then applied to a Glutathione Sepharose 4B column (GE Healthcare) to remove the HRV 3C protease. The flow-through fraction was collected and rapidly purified by TOYOSCREEN SuperQ (TOSOH Corporation), which is an ion exchange column with a linear gradient of 0–300 mM NaCl in Tris–HCl buffer (pH 8.0). Finally, hDDP4 was further purified by gel filtration chromatography using a Superdex 200 column (GE Healthcare) with 20 mM Tris–HCl (pH 8.0) buffer containing 100 mM NaCl. The purified hDPP-4 was concentrated using an Apollo spin concentrator (Orbital Biosciences, LLC) to 10 mg/mL for the crystallisation sample.

Crystallographic study of the DPP-4 complex with anagliptin

The complex was prepared by adding a 10-fold molar excess of anagliptin in a buffer (20 mM Tris–HCl, 0.1 M NaCl, pH 8.0) to the enzyme solution (20 mM Tris–HCl, 0.1 M NaCl, pH 8.0, concentration 9.6 mg/mL). The crystals of the complex were grown from a solution containing 18% PEG400, 15% PEG 1000 and 0.15 M Na/K phosphate (pH 6.5) at 293 K by the sitting-drop vapour-diffusion method. These crystallisation trials were conducted using the LIGHT method (laser irradiated growth technique)18 to produce high-quality crystals. Before data collection, the crystal was cryo-cooled using a mixture of 25% glycerol, 18% PEG400, 15% PEG1000 and 0.15 M Na/K phosphate (pH 6.5). X-ray diffraction data were collected on BL41XU at SPring-8 (Harima, Japan) at 100 K using a MX225HE (Rayonix, Evanston, IL) detector.

The structure of the DPP-4–anagliptin complex was solved by the molecular replacement method with the program MOLREP19 using the structure of DPP-4–saxagliptin complex20 refined to 2.85 Å resolution (PDB code 3BJM) as a model structure. The rigid body refinement was performed with the program REFMAC21. The refinements for the range of 47.7–2.85 Å resolution were performed with the program PrimeX, version 2.2 (Schrödinger, LLC, New York, NY). The coordinates of the structure were deposited in the Protein Data Base under code 3WQH.

Measurements of in vitro activities

Inhibition of DPP-4 activity was determined by measuring the rate of hydrolysis of a surrogate substrate, Gly-Pro-4-methylcoumarin-7-amide (Gly-Pro-MCA; Peptide Institute, Osaka, Japan). Ten microliters of various concentrations of the test compound solutions and 45 μL of human recombinant DPP-4 (R&D Systems, Minneapolis, MN) with an assay buffer (25 mM HEPES, 140 mM NaC1, 0.1 mg/mL BSA, pH 7.8) were added to a black 96-well microplate and mixed. After a 30-min preincubation at room temperature, the enzymatic reaction was initiated by the addition of 45 μL of 0.4 mM Gly-Pro-MCA and allowed to react for 20 min. The reaction was stopped by the addition of 100 μL of 25% acetic acid solution, and the fluorescence intensity was measured at an excitation wavelength of 390 nm and a fluorescence wavelength of 460 nm. As standards, the fluorescence intensity of 7-amino-4-methylcoumarin (AMC) solutions plus 25% acetic acid solution was measured. The percent inhibition relative to the control without inhibitor was calculated and the IC₅₀ values were determined by a nonlinear regression model.

The levels of inhibition of the DPP8 and DPP9 activities were determined by the degradation of the fluorescent substrate Gly-Pro-MCA. Human DPP8 and DPP9 were expressed in baculovirus-infected Sf9 insect cells and purified using his-tagged protein purification resins. Inhibition of the DPP8 and DPP9 activities was determined as described above. Ten microliters of the test compounds and 50 μL of 1.0 mM Gly-Pro-MCA in buffer solution (50 mM HEPES, 0.1 mg/mL BSA, pH 8.0) were added to black 96-well microplates, and the reaction was initiated by the addition of 40 μL of the diluted enzyme solution for 20 min at room temperature. The reaction was stopped by the addition of 25% acetic acid solution and the fluorescence intensity was measured.

Inhibition of FAP activity was determined by the degradation of the fluorescent substrate Ala-Pro-7-amino-3-trifluoromethylcoumarine (Ala-Pro-AFC; Bachem AG, Switzerland). Human FAP was expressed in baculovirus-infected Sf9 insect cells and purified using his-tagged protein purification resins. The inhibition of FAP activities was measured as described above. Ten microliters of the test compound solutions and 50 μL of 1.0 mM Ala-Pro-AFC in buffer solution (50 mM Tris hydrochloride–150 mM sodium chloride, 0.1 mg/mL BSA, pH 8.0) were added to black 96-well microplates. The reaction was initiated by the addition of 40 μL of the diluted enzyme solution for 60 min at room temperature. The fluorescence intensity was measured at an excitation wavelength of 400 nm and a fluorescence wavelength of 505 nm. To establish the standards, various concentrations of AFC solutions were added to a 96-well black microplate, and the fluorescence intensities were measured.

Docking simulation

The initial ligand structure was prepared from the single-crystal structure of anagliptin determined in this paper. All ligands for docking simulation were provided from anagliptin analogues using the LigPrep software package, version 2.8 (Schrödinger, LLC, New York, NY, 2013). These ligand-calculation conditions adopted conventional values. The molecular-docking simulation of compounds against the DPP-4 crystal structure, PDB code 3QWH, was performed by Glide22.

Results and discussion

Crystal structure of anagliptin

The crystal structure of anagliptin was solved and refined by single-crystal X-ray analysis. Anagliptin has a single asymmetric carbon on its pyrrolidine moiety. Since the pyrrolidine moiety is prepared from l-prolinamide without racemization, the stereochemistry must be 2S, like that of l-prolinamide. The structure having a 2S configuration was therefore refined; the final R index was 0.0454 and the absolute structure parameter χ23 was 0.0. Consequently, it was clarified that the configuration was 2S (space group: P3₂). For comparison, a 2R configuration (space group: P3₁) was modelled and refined. In this case the final R index was 0.0459 and the χ parameter was 1.1; these parameters showed that the 2R configuration was not correct. The crystallographic data and refinement are shown in Table 1 (CCDC 1003533).

Table 1. Crystallographic data and refinement of anagliptin.

Formula	C19H25N7O2
Formula weight	383.45
Crystal system	Trigonal
Space group	P32
Unit cell parameters	a = 10.7727(3) Å, α = 90°
	b = 10.7727(3) Å, β = 90°
	c = 30.027(2) Å, γ = 120°
Temperature of data collection	Room temperature
Z, calculated density	6, 1.266 g/cm^3
F(000)	1224
Theta range for data collection	2.03–28.27°
Index ranges	−9 ≤ h ≤ 14, −14 ≤ k ≤ 12,  −39 ≤ l ≤ 39
Reflections collected/unique	18953/9266
Goodness-of-fit on F^2	0.879
Final R indices	R = 0.0454, wR = 0.1005
Absolute structure parameter	χ = 0.0(12)

The solved structure consisted of two independent molecules which were formed by intermolecular hydrogen bonds; the ORTEP diagram is shown in Figure 2. The ethylenediamine moieties and carbonyl groups linked to the pyrrolidine ring in the pair of molecules were contributed as proton donors and acceptors, respectively, and the pair of A and B molecules formed a crystalline lattice by helical π–π stacking24 of the pyrazolopyrimidine rings (Figure 3).

Figure 2. ORTEP diagram of the A and B molecules.

Figure 3. Interactions of the A and B molecules in single-crystals of anagliptin (left). The schematic diagram shows the distances of intermolecular hydrogen bonds and π/π stacking of A and B molecules (right).

Co-crystal structure of DPP-4 with anagliptin

The recombinant human DPP-4 expressed in the silkworm baculovirus system was purified by ion exchange column chromatography and gel filtration. The single crystals were grown from the complex with anagliptin by the sitting-drop vapour-diffusion method. The X-ray diffraction data were collected on BL41XU at SPring-8 at 100 K. The structure was solved by the molecular replacement method and refined. The crystallographic data and refinement data are shown in Table 2 (PDB code 3WQH). The 2F_obs − F_calc electron density map (σ = 1.1) around anagliptin after refinement is represented in Figure 4.

Figure 4. 2F_obs − F_calc electron density map (σ = 1.1, magenta) around anagliptin (green) after refinement.

Table 2. Crystallographic data and refinement of the complex of DPP-4 with anagliptin.

Data collection
Radiation source	SPring-8 BL41XU
Space group	P212121
Cell dimensions (Å)	a = 65.16, b = 68.99,  c = 419.60
Resolution range (Å)	49.11–2.85
No. of observed reflections	312 285
No. of unique reflections	45 509
Completeness (%)	99.9 (100.0)*
I/σ (I)	12.1 (2.8)*
Rmerge (%)†	14.9 (78.0)*
CC1/2‡	0.996 (0.902)*
Refinement
Resolution range (Å)	47.67–2.85
No. of amino acid residues (Chains A/B)	728/728
No. of atoms
Protein	11 926
Inhibitor and sugar	239
Water	104
No. of unique reflections	45 508
No. of reflections used for calculating Rfree	2275
Rcryst (%)¶	22.7
Rfree (%)§	28.8
Average B-factor (Å)
Protein molecules	52.9
Water molecules	44.0
Rmsd∥
Bond length (Å)	0.0120
Bond angles (°)	1.85
Ramachandran plot#
Most favored (%)	92.1
Additionally allowed (%)	7.4

*Values in parentheses are for the highest-resolution shells (2.92–2.85 Å).

†R_merge = ∑_hkl∑_i |I_i(hkl) − <I(hkl)>|/∑_hkl∑_i I_i(hkl), where I_i(hkl) is the ith observed intensity of reflection hkl and <I(hkl)> is the average intensity over symmetry-equivalent measurements. Values in parentheses are for the highest resolution shell.

‡Reference Karplus and Diederichs25.

¶R_cryst = ∑||F_o| − |F_c||/∑|F_o| calculated from 95% of the data and used during refinement.

§R_free = ∑||F_o| − |F_c||/∑|F_o| calculated from 5% of the data and used during refinement.

∥Rmsd: root-mean-square deviation.

#http://molprobity.biochem.duke.edu/ 26.

The solved DPP-4 structure complexed with anagliptin consisted of two independent protein molecules in an asymmetric unit, one anagliptin molecule bound to each protein molecule. The three interaction subsites of anagliptin with DPP-4 were revealed; Figure 5 shows the interactions with the B molecule of two proteins. (1) The first site was the S₁ subsite, in which the cyano group had a dipole interaction. The distance between the cyano carbon atom and the Oγ of Ser630 (catalytic residue) was 2.8 Å, and the distance between the cyano nitrogen atom and the Oη of Tyr547 (proton donor) was 3.2 Å. (2) The second site was the S₂ subsite, in which the secondary amine formed double salt bridges. The distances of 3.0 and 3.3 Å between the nitrogen atom of the secondary amine and each Oɛ of Glu205 and Glu206 were appropriate distances for salt-bridge formation. (3) The third site was the S₂ extensive subsite13, in which the carbonyl group linked to the pyrazolopyrimidine ring formed a hydrogen bond with the Nη of Arg358, and the pyrazolopyrimidine ring underwent a π–π stacking interaction with the phenyl ring of Phe357. The distance of 3.4 Å between the oxygen atom of the carbonyl group and the Nη of Arg358 was within a hydrogen bond. The distance of 4.3 Å between the centres of each six-membered ring of the heterocycle and the phenyl ring of Phe357 was within an energetically-favoured range for π–π stacking24.

Figure 5. Interactions of anagliptin in the active site of DPP-4. The distances of their interactions (dotted lines) are represented.

As described above, the interacting subsites of anagliptin with DPP-4 were the S₁, S₂ and S₂ extensive subsites. Anagliptin should thus be included in class 3 according to the categorization by Nabeno et al12. The covalent imidate intermediates with the Oγ of Ser630 in the S₁ subsite are observed in the X-ray analyses of the inhibitors having a cyano group, vildagliptin27 and saxagliptin20.

However, the cyano group of anagliptin underwent dipole interactions in the S₁ subsite. These interactions did not lead to covalent binding with the Oγ of Ser630, but could have resulted in a transition state in the formation of the covalent imidated intermediate. Based on the study of ¹⁴C-labelled anagliptin, the major metabolite of anagliptin is known to be a hydrolysis product of the cyano group28. The metabolic mechanism of anagliptin therefore likely proceeds through the S₁ subsite via an imidate intermediate.

The S₂ extensive subsite contributes to the affinity and selectivity as described in the characterizations of sitagliptin15 and teneligliptin29. The selectivity against DPP-4 homologues in particular has been reported to be important for safety11^,30. Rummey et al. reported the homology models of DPP8 and DPP9 based on the DPP-4 X-ray structure, and suggested that the selectivity of sitagliptin was based on the absence of key interactions in DPP8/931. In their models, the three residues of DPP-4 (Phe357, R358 and Ser209), which had the key interactions with sitagliptin, were not retained in either DPP8 or DPP9, and the corresponding residues in DPP8 and DPP9 lost the interactions with sitagliptin. Since anagliptin also exhibited interactions with Phe357 and Arg358 at the S₂ extensive subsite, its selectivity can be accounted for in the same way.

Potency and selectivity of DPP-4 inhibitors

The in vitro activities of DPP-4 inhibitors against DPP-4 and its homologues DPP8, DPP9 and FAP were measured. Their selectivities were calculated from the IC₅₀ values as shown in Table 3. The results showed that the IC₅₀ values of vildagliptin and anagliptin were lower than those of sitagliptin and alogliptin. We considered that this difference was likely attributable to the different binding modes in the S₁ subsite. Vildagliptin and anagliptin, both of which can form a covalent bond as described in the previous section, might bind more slowly than sitagliptin and alogliptin, which undergo hydrophobic interactions at the subsite. Presumably, since the incubation time was relatively long by this method, the IC₅₀ values of vildagliptin and anagliptin were probably lowered due to their binding modes.

Table 3. In vitro IC₅₀ values (nM) of DPP-4 inhibitors against enzymes.

	IC50 values (nM)*
Enzymes	Anagliptin	Sitagliptin	Vildagliptin	Alogliptin
DPP-4	3.3	14.8	2.5	9.4
DPP8	84 700	84 800	6400	>500 000
DPP9	56 100	299 400	1100	>500 000
FAP	72 700	>500 000	54 600	>500 000
Selectivity†	>16 000	>5700	>440	>53 000

*The values are the mean of three measurements.

†Selectivity is the lowest IC₅₀ value of the other enzymes tested over that of DPP-4.

With regard to the selectivity, anagliptin showed high selectivities against DPP8, DPP9 and FAP (Table 3). The high selectivities were probably attributable to the interactions with the Phe357 and Arg358 in the S₂ subsite as discussed in the previous section.

SAR of anagliptin derivatives

Some of the anagliptin derivatives are shown in Table 4 along with their inhibitory activities14. The monosubstitution at the 7 position (R³) of the heterocycle tended to decrease the inhibitory activity (compounds 2 and 3), and disubstitution at the 5 and 7 positions (R² and R³) significantly decreased the activity (compounds 4 and 5). In this X-ray study, the heterocycle and the linked carbonyl group of anagliptin were coplanar as shown in Figure 5. These findings suggested that the substituent introductions at the ortho position of the heterocycle caused steric hindrance or distortion of the planarity, and thereby reduced the effects of the hydrogen bond of Arg358 and π–π stacking of Phe357 to decrease the IC₅₀ values.

Table 4. In vitro IC₅₀ values (nM) of and GlideScores of anagliptin derivatives.

Compounds	R^1	R^2	R^3	IC50 values (nM)*	GlideScores†
1 (Anagliptin·HCl)	Me	H	H	3.8	−7.56
2	Me	H	Me	13	−8.11
3	tBu	H	Me	15	−7.76
4	tBu	Me	Me	80	−6.84
5	Me	Me	Me	110	−6.24

*Data are excerpted from Kato et al14.

†GlideScore = 0.998 × log(IC₅₀) − 8.664, r = 0.801.

On the other hand, replacement of the methyl group by the tert-butyl group at the 2-position (R¹) had no influence on the inhibitory activity, based on the comparisons between compounds 2 and 3 and between compounds 4 and 5 (Table 4). It was considered that the extra space around R¹ allowed introduction of the bulky substituent.

We therefore employed the flexible ligand-docking program Glide22 to assess the influences of the substituents at R¹ based on the co-crystal structure of human DPP-4 with anagliptin determined in this study. The extra precision mode (XP mode) of Glide was adopted in this simulation, and the compound-inhibitory activities were estimated using the GlideScore value. The docking simulation showed that the best binding model of anagliptin in the active site of DPP-4 presented an RMSD of 1.1 Å relative to the X-ray structure determined in this study. Using these docking conditions, we obtained a moderate correlation coefficient of r = 0.801 between the log(IC₅₀) values of the five compounds 1–5 and their GlideScores, as shown in Table 4. Virtual compounds 6–9 having polar and non-polar substituents were then evaluated under these docking conditions, as shown in Table 5. Interestingly, both virtual compounds 8 and 9 having a polar group at R¹ were hydrogen-bonded to the Oη of Tyr585 as shown in Figure 6; these compounds had the better GlideScores, −8.55 and −8.59. In addition, this simulation suggested that T-shaped interactions occurred between compounds 8 and 9 and the phenyl ring of Phe357, respectively, although anagliptin showed a slipped parallel interaction with the Phe357. This simulation therefore indicated that compounds 8 and 9 simultaneously obtained their interaction energies for T-shaped interaction and hydrogen bonding. In their quantum chemical study of the benzene–dimer interaction, Tsuzuki et al. reported that the interaction energy of the slipped parallel interaction was approximately equal to that of the T-shaped interaction24. Ninković et al. also recently reported that a large number of phenyl–phenyl contacts in proteins had tilt angles above 60°, corresponding to the T-shaped interaction in the aromatic–aromatic interaction study at a tilt angle range from 0° (parallel) to 90° (T-shaped) by using PDB data32. With regard to the hydrogen bonding with the Tyr585, Ahn et al. showed that a carboxyl substituent on the homopiperazine derivative appeared to interact with the Oη of Tyr585 in the co-crystal structure of DPP-4, and the acid moiety could regulate CYPs activities33.

Figure 6. Co-crystal structure of anagliptin (red) determined by X-ray crystallography, and virtual compounds 8. (green) and 9. (blue) by the docking simulation.

Table 5. GlideScores of virtual compounds by docking simulations*.

Virtual compounds	R^1	R^2	R^3	GlideScores
6	Et	H	H	−8.10
7	nPr	H	H	−8.43
8	CO2H	H	H	−8.55
9	NH2	H	H	−8.59

*The positions of R¹, R² and R³ are the same as in Table 4.

As a result of the SAR study, we therefore concluded that the introductions of polar substituents at R¹ were likely to be an effective approach for further development of the derivatives.

Conclusion

The single-crystal X-ray analysis of anagliptin was performed. Anagliptin is not hygroscopic, but is very soluble in water. This favourable property for a pharmaceutical ingredient is probably attributable to the unique crystal structure. The intermolecular hydrogen bonds of the hydrophilic moieties and stackings of hydrophobic bicyclic-hetero rings could prevent contact with moisture, accounting for the non-hygroscopicity of the crystal, while the hydrophilic moieties could promote the solvation with water once the lattice breaks up due to contact with excess water.

The three interaction subsites of anagliptin with DPP-4 were revealed by the co-crystal structure analysis. Interestingly, a dipole interaction of the cyano group was observed in the S₁ subsite in contrast to the covalent bonds observed in the analyses of vildagliptin27 and saxagliptin20. This novel non-covalent bond interaction of the cyano group was considered to be a transition state before formation of a covalent imidate, which was supported by the facts that both anagliptin and vildagliptin were highly potent in this analysis, and a major metabolite of anagliptin was a hydrolysed compound of the cyano group.

We concluded that the high selectivity of anagliptin was attributable to the utilization of the S₂ extensive site as in the cases of sitagliptin31 and teneligliptin29.

Further, the SAR study of anagliptin derivatives reasonably accounted for the influence of the substituent on the heterocycle, which interacted at the S₂ extensive site. The docking simulation results suggested a direction for further optimization.

Thus, the chemical structure of anagliptin was characterised and shown to have favourable properties for medicinal application. Continued study of the DPP-4 inhibitors, including anagliptin, is expected to underscore their efficacy for the treatment of type 2 diabetes.

Acknowledgements

We are grateful to Dr. Shigeru Sugiyama (Associate Professor, Osaka University, Japan) for useful suggestions and discussion.

Declaration of interest

The authors report no declarations of interest.