X-Ray Crystal Structure of a 2-Amino-3,4-dihydroquinazoline 5-HT₃ Serotonin Receptor Antagonist and Related Analogs

Abstract

Certain 2-amino-3,4-dihydroquinazolines bind at 5-HT₃ serotonin receptors and act as antagonists (e.g. 6-chloro) whereas others bind with little to no affinity and lack functional activity (e.g. 8-chloro). The purpose of this investigation was to gain insight as to why this might be the case. X-Ray crystallographic studies revealed that the N–C–N distances in the examined analogs are nearly identical (1.31 – 1.34 Å), suggesting that differences in N–C–N delocalization does not account for differences in affinity/action. Homology modeling hydrophatic interactions (HINT) analysis revealed that the 6-chloro analog formed a greater number, and more favorable, interactions with the receptor, whereas the 8-chloro analog formed fewer, and unfavorable, interactions. The affinity and activity of the 6-chloro quinazoline relative to its 8-chloro counterpart are unrelated to the N–C–N delocalization pattern but might be related to specific (favorable and unfavorable) interactions of quinazoline substituents with certain receptor features as determined by HINT analysis.

Graphical Abstract

Introduction

We have previously demonstrated that 2-amino-3,4-dihydroquinazolines, such as 1, represent a novel class of 5-HT₃ serotonin receptor antagonists and that receptor affinity is aryl-substituent-dependent [1,2]. Furthermore, some of the compounds displayed antidepressant-like actions in a mouse tail-suspension test [1]. Structure-activity relationship (SAR) studies showed that each of the three nitrogen atoms of 1-related analogs is important for 5-HT₃ receptor binding [2]. That is, compound 2 binds at 5-HT₃ receptors with a K_i of 0.2 μM. Replacement of N1 or N3 by a carbon atom decreased affinity by up to 10-fold, and removal of the 2-amino group decreased affinity by at least 5-fold; replacement of both ring nitrogen atoms, to afford 6-chloro-2-aminotetralin resulted in an inactive compound (K_i >10 μM). These findings were explained on the basis of homology modeling/docking studies indicating that all three nitrogen atoms, via possible hydrogen bond formation, interact with receptor-related features.

In this report, we studied a positional isomer of 2 (i.e., 3), differing only in the position of the chloro substituent; 3 lacked affinity for 5-HT₃ receptors (K_i >10 μM; i.e., 37% inhibition at 10 μM) (data obtained from the NIMH Psychoactive Drug Screening Program; PDSP).

The C–N distance of protonated guanidines (i.e., guanidinium cations), including substituted guanidines, as determined by X-ray crystallography, is typically 1.3 Å [3]. That is, all three C–N distances are essentially identical due to extensive charge delocalization. Exceptions exist; for example, in the 2,6-dichlorophenyl-substituted arylguanidine 4, the C_guanidinium–N_terminal distances were found to be, as expected, 1.3 Å; however the C_guanidinium–N_anilinic distance was similar to a C–N single bond (i.e., 1.49 Å) [4]. The difference is that while the “guanidine” of 4 is not constrained, that of 2 (and 3) are conformationally-constrained (i.e., are 2-amino-3,4-dihydroqinazolines). The C–N distance of protonated 2-amino-3,4-dihydroquinazolines should show similar bond lengths (as represented by the delocalized structure 5) regardless of the aryl substitution pattern.

The above observations, coupled with the fact that no X-ray crystal structure of a 2-amino-3,4-dihydroquinazoline had been previously reported in detail, prompted the current investigation. It might be noted that a crystal structure of a 2-amino-3,4-dihydroquinazoline has been reported [5]. However, because this was an N1-substituted analog, the results might not be particularly relevant here. Furthermore, the structure was provided as a figure and no bond lengths were indicated.

We addressed several questions. Is compound 3 inactive because delocalization of the guanidinium species differs from that suggested by 5 (and is similar to that of 4)? Is the position of chloro substitution important in delocalization of the constrained guanidine group, that is, is guanidinium delocalization in 3 and 2 similar or different, which might also explain their different activity? Hence, we set about to solve the crystal structures of 2, 3, and the des-amino analog of 2 (i.e., amidine 6) for purpose of comparison. This could provide information on the nature (i.e., length and, therefore, delocalization or lack thereof) of the C–N bond found in these protonated compounds and shed insight into their biological activities.

Note: The IUPAC numbering system was employed for structure 1. The 2-amino group will be referred to as N2. The numbering shown in the Figures were automatically generated by the program employed. Hence, they are not identical. Both numbering systems will be employed here, but the official numbering system will be followed, in the discussion, by the program-generated numbering (in parenthesis) when applicable.

Materials and Methods

General

Compounds were prepared and characterized using a combination of melting point (m.p.), proton nuclear magnetic resonance (¹H NMR), and infrared (IR) spectroscopy. Purity of 3 was determined based on elemental analysis for C, H and N performed by Atlantic Microlab Inc. (Norcross, GA) and a compound was considered pure if the values obtained were within 0.4% of theoretical values. A MEL TEMP melting point apparatus was utilized to obtain uncorrected melting points of compounds in glass-walled capillary tubes. ¹H NMR spectra were obtained using a Bruker ARX 400 MHz spectrometer. The spectra obtained were reported by indicating the position of the peaks in parts per million (ppm) downfield from tetramethylsilane (TMS), used as an internal standard, followed by the splitting pattern of the peak (s = singlet, d = doublet, t = triplet, q = quartet, dd = doublet of doublets, m = multiplet), coupling constant (J, Hz) and integration. IR spectra were determined using a Thermo Nicolet iS10 FT-IR. Reactions were monitored using thin-layer chromatography (TLC) on silica gel GHLF plates (250 μm, 2.5 × 10 cm; Analtech Inc. Newark, DE). For the purpose of biological studies, a water-soluble hydrochloride salt of 3 was prepared. Compounds 2 [1] and 6 [2] were also examined as their hydrochloride salts. Starting material, 2-amino-3-chlorobenzonitrile (7), was purchased form Sigma-Aldrich and used as provided.

Synthesis

2-Amino-8-chloro-3,4-dihydroquinazoline Hydrochloride (3)

Cyanogen bromide (3 M solution in CH₂Cl₂, 3 mL, 8.6 mmol) was added to a stirred solution of 2-amino-3-chlorobenzylamine (8) (0.90 g, 5.6 mmol) in toluene (6 mL). The stirred reaction mixture was heated at reflux for 4.5 h, allowed to cool to room temperature and stirred overnight. This was followed by removal of toluene under reduced pressure to give a yellow solid. The residue was dissolved in H₂O (60 mL) followed by addition of a saturated solution of NaHCO₃ to pH = 9 to 10. The H₂O layer was further basified by addition of NaOH (3 M, to pH 12). The solid was collected by filtration and washed with H₂O (30 mL) to afford 0.85 g (81%) of the crude free base of 3. The solid was dissolved in EtOH (20 mL) and a saturated solution of gaseous HCl in EtOH was added (pH 1–2) and the reaction mixture was allowed to stir at room temperature overnight. The solid was collected by filtration to afford 0.39 g of product which upon recrystallization from EtOH gave 0.17 g (28%) of 3 as a white solid: m.p.: 202–204 °C. ¹H NMR (DMSO-d₆) δ 4.52 (s, 2H, CH₂), 7.06–7.11 (t, J = 6 Hz, 1H, ArH), 7.15–7.18 (d, J = 12 Hz, 1H, ArH), 7.38–7.40 (d, J = 8 Hz, 1H, ArH), 8.08 (s, 2H, NH₂), 8.82 (s, 1H, NH), 10.07 (s, 1H, NH). Anal. Calcd for (C₈H₈N₃Cl·HCl) C, 44.06; H, 4.16; N, 19.27. Found: C, 44.05; H, 4.01; N, 19.15.

2-Amino-3-chlorobenzylamine (8)

BH₃·THF complex (1 M in THF, 26 mL, 26.2 mmol) was added in a dropwise manner at 0 °C (ice-bath) to 2-amino-3-chlorobenzonitrile (7) (1.00 g, 6.6 mmol) under an N₂ atmosphere. The stirred reaction mixture was heated at reflux overnight, allowed to cool to room temperature, quenched by addition of HCl (6 N, to pH 2), and heated at reflux for 30 min. The reaction mixture was allowed to cool to room temperature and then cooled to 0 °C (ice-bath) and basified with NaOH (3 M, to pH 12) and extracted with CH₂Cl₂ (2 × 40 mL). The combined organic portion was washed with H₂O (100 mL), dried (MgSO₄) and evaporated under reduced pressure to afford 0.93 g (90%) of 8 as a sticky solid. IR spectroscopy indicated the absence of a characteristic cyano band at 2226.58 cm⁻¹.

X-Ray Crystallography

For the crystal structure determination, single crystals of 2, 3, and 6 were mounted on a MiTeGen Microloop and used for data collection on a XtaLAB MM007-HF system (Cu Ka radiation, λ = 1.54184 Å) coupled with a Hybrid Photon Counting detector (Dectris Eiger 4M) at room temperature. CrysAlis PRO software (Rigaku OD, 2015) [6] was used for data collection, data processing, cell refinement, and data reduction. The structures were solved with ShelXTL [7] using Intrinsic Phasing, and refined with the olex2-refine [8,9] refinement package using Gauss-Newton minimization. H atoms were determined from a difference Fourier synthesis. The final difference Fourier maps showed no peaks of chemical significance.

Compound 2.

Compound 2 was recrystallized from methanol. A single crystal with approximate dimensions of 0.215 mm × 0.258 mm × 0.324 mm was used for diffraction data collection. The crystal belonged to the triclinic P-1 space group with cell dimensions: a = 6.2965(2) Å, b = 7.4380(1) Å, c = 10.1913(3) Å, α = 90.160(2)°, β = 101.445(2)°, γ = 97.776(2), Z = 2, volume = 463.28(2) ų. Integration of the data yielded a total of 4342 reflections to a maximum θ angle of 66.94° (0.84 Å resolution), of which 1612 were independent (completeness = 97.5%, R_int = 1.45%, R_sig = 1.03%), and 1589 were greater than 2σ(F²). The final data were corrected for absorption effects using the multi-scan method (SADABS). The final anisotropic full-matrix least-squares refinement on F² with 118 variable converged at R₁ = 4.91%, for the observed data and wR2 = 14.28% for all data. The goodness-of-fit was 1.051. The largest peak in the final difference electron density synthesis was 0.62 e⁻/ų and the largest hole was −0.37 e⁻/ų. On the basis of the final model, the calculated density was 1.5632 g/cm³ and F(000), 226 e⁻.

Compound 3.

Compound 3 was recrystallized from methanol/hexane (1:1). A single crystal with approximate dimensions 0.09 mm × 0.210 mm × 0.327 mm was used for diffraction data collection. The crystal belonged to an FDD2 space group with unit cell dimension: a = 10.6349(2) Å, b = 34.6072(4) Å, c = 10.6985(1) Å, Z = 16, volume = 3937.52(9) ų. A total of 33798 reflections were measured to a maximum θ angle of 74.98° (0.80 Å resolution), of which 1990 were independent (completeness = 98.0%, R_int = 5.64%, R_sig = 1.25%) and 1981 were greater than 2σ(F²). Data were corrected for absorption effects using the multi-scan method (SADABS). The final anisotropic full-matrix least-squares refinement on F² with 118 variable converged at R₁ = 3.64%, for the observed data and wR₂ = 10.44% for all data. The goodness-of-fit was 1.046. The largest peak in the final difference electron density synthesis was 0.16 e⁻/ų and the largest hole was −0.23 e⁻/ų. On the basis of the final model, the calculated density was 1.4714 g/cm³ and F(000), 1807 e⁻.

Compound 6.

A small needle crystal was obtained for compound 6 when recrystallized from methanol, with approximate dimensions of 0.032 mm × 0.067 mm × 0.109 mm. The crystal belonged to a P-1 space group with final cell constants: a = 5.7857(3) Å, b = 7.8573(5) Å, c = 10.1875(7) Å, α = 92.088(5)°, β = 104.168(6)°, γ = 99.188(5)°, Z = 2, volume = 441.89(5) ų. Integration of the data yielded a total of 7121 reflections to a maximum θ angle of 66.58° (0.84 Å resolution), of which 1508 were independent (completeness = 96.4%, R_int = 3.35%, R_sig = 2.24%) and 1368 were greater than 2σ(F²). Data were corrected for absorption effects using the multi-scan method (SADABS). The final anisotropic full-matrix least-squares refinement on F² with 109 variable converged at R₁ = 4.78%, for the observed data and wR₂ = 14.59% for all data. The goodness-of-fit was 1.110. The largest peak in the final difference electron density synthesis was 0.44 e⁻/ų and the largest hole was −0.37 e⁻/ų. On the basis of the final model, the calculated density was 1.526 g/cm³ and F(000), 208 e⁻.

Crystallographic data of compounds 2, 3 and 6 have been deposited at the Cambridge Crystallographic Data Centre (CCDC) and allocated the deposition numbers CCDC 1893635, 1892730 and 1905075, respectively. These data can be obtained free of charge from the Cambridge Crystallographic Data Centre via www.ccdc.cam.ac.uk/data-request/cif.

3D Modeling

Preparation of the mouse 5-HT₃A (m5-HT₃A) receptor crystal structure models and ligands was performed as previously described [1,2]. In short we generated models of the extracellular ligand binding domains (LBDs) of two adjacent subunits of the homopentameric m5-HT₃A receptor consisting of the A (primary) and E (complementary) subunits from the m5-HT₃A receptor crystal structure (PDB ID: 4PIR).

The X-ray structures of ligands 2 and 3 obtained in the current studies (please see X-ray crystallography experimentals above) were employed in our modeling studies. Docking studies of the ligands into the orthosteric binding site with C-loop opening of 15–16 Å were conducted using GOLD Suite v5.5 (Cambridge Crystallographic Data Center, Cambridge, UK) [10] as previously reported [2]. A binding site for docking was defined within a 20 Å distance of W183 and 10 different non-diverse solutions were generated for each molecule to identify their common binding modes. A common binding mode was identified based on the GOLD scores and their predicted interactions with binding site residues. The ligands were merged with the receptor and the complexes were energy-minimized in SYBYL-X 2.1.1 (Certara USA, Inc., Princeton, NJ) under the same conditions described above for the unbound ligands. HINT [11] analysis was conducted on each complex using SYBYL 8.1(Certara USA, Inc., Princeton, NJ) and those that produced the highest scores, while maintaining expected binding site interactions, were selected. The PyMOL Molecular Graphics System, Version 1.7.4 (Schrödinger, LLC) was used to generate images.

Results and discussion

Although 3 has been previously reported [5], neither its synthesis nor physicochemical properties were described. 2-Amino-8-chloro-3,4-dihydroquinazoline hydrochloride (3) was obtained in a cyclization reaction of 2-amino-3-chlorobenzylamine (8) with cyanogen bromide as shown in Scheme 1.

Scheme 1.

^a Synthesis of 3.

^aReagents and conditions: a. BH₃·THF, reflux; b. CNBr, toluene, reflux.

The structure of 3 was confirmed by IR, ¹H NMR, and elemental analysis for C, H, N (see Materials and Methods). Crystallographic data and structural refinement details for all three crystals are given in Table 1.

Table 1.

Crystallographic data for compounds 2, 3 and 6.

Compound	2	3	6
Formula	C8H9Cl2N3	C8H9Cl2N3	C8H8Cl2N2
Formula weight	218.09	218.09	203.06
T (K)	295.5(1)	295.6 (3)	296.5(4)
λ (Å)	1.54184	1.54184	1.54184
Crystal system	Triclinic	Orthorhombic	Triclinic
Space group	P-1	FDD2	P-1
a (Å)	6.2965(2)	10.6349(2)	5.7857(3)
b (Å)	7.4380(1)	34.6072(4)	7.8573(5)
c (Å)	10.1913(3)	10.6985(1)	10.1875(7)
α (°)	90.160(2)	90	92.088(5)
β (°)	101.445(2)	90	104.168(6)
γ (°)	97.776(2)	90	99.188(5)
Volume (Å3)	463.28(2)	3937.52(9)	441.89(5)
Z	2	16	2
ρcalc (g/cm3)	1.5632	1.4714	1.526
μ (mm−1)	5.925	5.577	6.133
F (000)	226	1807	208
Crystal size (mm3)	0.324 × 0.258 × 0.215	0.327 × 0.210 × 0.090	0.109 × 0.067 × 0.032
Theta range for data collection (°)	4.43–66.94	5.11–74.98	4.49–66.58
Index ranges	−7 ≤ h ≤ 7	−12 ≤ h ≤ 12	−6 ≤ h ≤ 6
	−8 ≤ k ≤ 7	−43 ≤ k ≤ 43	−9 ≤ k ≤ 9
	−12 ≤ l ≤ 12	−13 ≤ l ≤ 13	−12 ≤ 1 ≤ 12
Reflections collected	4342	33798	7121
Independent reflections	1612 [Rint = 0.0145]	1990 [Rint = 0.0564]	1508 [Rint = 0.0335]
Data completeness (%)	97.5	98.0	96.4
Absorption correction	Multi-scan	Multi-scan	Multi-scan
Data / restraints / parameters	1612 / 0 /118	1990 / 1 /118	1508 / 0 /109
Goodness-of-fit on F2	1.051	1.046	1.110
Final R indices [I > 2σ(I)]	R1 = 0.0491	R1 = 0.0364	R1 = 0.0478
	wR2 = 0.1425	wR2 = 0.1043	wR2 = 0.1432
R indices (all data)	R1 = 0.0494	R1 = 0.0365	R1 = 0.0499
	wR2 = 0.1428	wR2 = 0.1044	wR2 = 0.1459
Largest diff. peak and hole (e. Å−3)	0.62 and −0.37	0.16 and −0.23	0.44 and −0.37

We have determined the crystal structures of two 2-amino-3,4-dihydroquinazolines (i.e., 2 and 3) and the related structure 6; each with a chloride anion. The crystal structure of compound 2 with atom labeling and polymeric structure are shown in Fig. 1. The C–C (arene ring) distances are in the range of 1.38–1.39 Å. The guanidine moiety is protonated and all three C–N distance are within the range of 1.31–1.34 Å due to extensive charge delocalization (Table 2). An interaction between the protonated guanidine group and the chloride anion also involves four hydrogen bonds, with three of the interactions leading to the formation of a polymeric structure along the b axis. Furthermore, the N2H (N13-H13b) ⋯ Cl1ⁱⁱ [D⋯A = 3.2885(1) Å] interactions have a contribution in the formation of a stable structure (Fig. 1). The molecular packing is shown in Fig. S1. Hydrogen-bond geometry is shown in Table 3.

Figure 1.

Top: Molecular structure of compound 2. Thermal ellipsoids are drawn at the 40% probability level. Bottom: Polymeric H-bonding geometry. Hydrogen bonds are drawn as dotted lines (view 100).

Table 2.

Selected bond distances and angles (Å, °) in the compounds 2, 3, and 6.

Compound 2 (Fig. 1)
C7 – N6	1.340(3)	C5 – C10	1.388(3)
C7 – N8	1.309(3)	C10 – C9	1.501(3)
C7 – N13	1.324(3)	C9 – N8	1.454(3)
N6 – C5	1.399(3)
N3 – C7 – N6	117.3(2)	C5 – C10 – C9	120.15(19)
C7 – N6 – C5	122.19(18)	C10 – C9 – N8	111.77(18)
N6 – C5 – C10	119.43(18)	C9 – N8 – C7	125.42(18)
N6 – C7 – N8	120.4(2)	N8 – C7 – N13	122.36(19)
Compound 3 (Fig. 2)
C4 – N12	1.324(3)	C6 – C7	1.497(4)
C4 – N5	1.318(4)	C7 – C2	1.395(3)
C4 – N3	1.340(3)	C2 – N3	1.396(3)
N5 – C6	1.437(4)
N12 – C4 – N5	120.8(2)	N5 – C6 – C7	112.7(3)
N12 – C4 – N3	118.7(2)	C6 – C7 – C2	120.0(2)
N3 – C4 – N5	120.5(2)	C7 – C2 – N3	119.2(2)
C4 – N5 – C6	125.2(2)	C2 – N3 – C4	122.29(2)
Compound 6 (Fig. 3)
C7 – N8	1.291(3)	C10 – C5	1.386(3)
N8 – C9	1.458(3)	C5 – N6	1.412(3)
C9 – C10	1.509(3)	N6 – C7	1.315(3)
N6 – C7 – N8	122.9(2)	C9 – C10 – C5	120.9(1)
C7 – N8 – C9	125.1(1)	C10 – C5 – N6	118.8(1)
N8 – C9 – C10	110.5(1)	C5 – N6 – C7	120.9(1)

Table 3.

Hydrogen-bond geometry (Å, °) in the compounds 2, 3, and 6.

Compound	D – H ⋯ A	D – H	H ⋯ A	D ⋯A	D – H ⋯ A
	N8 - H8 ⋯ Cl1	0.86	2.3251(1)	3.176 6(1)	170.5328(3)
2	N6 - H6 ⋯ Cl1i	0.86	2.4129(1)	3.2209(1)	156.7366(8)
	N13 – 13a ⋯ Cl1i	0.86	2.6358(1)	3.3945(1)	147.7917(10)
	N13 – 13b ⋯ Cl1ii	0.86	2.5705(1)	3.2885(1)	141.6411(13)
	N5 - H5 ⋯ Cl13	0.86	2.484(3)	3.275(3)	153.27(6)
3	N3 - H3 ⋯ Cl13iii	0.86	2.460(2)	3.239(2)	151.03(5)
	N12 - H12b ⋯ Cl13	0.86	2.424(3)	3.215(3)	153.20(6)
	N12 - H12a ⋯ Cl13iii	0.86	2.366(3)	3.160(3)	153.73(6)
6	N6 - H6 ⋯ Cl1	0.86	2.2946(1)	3.1231(2)	162.7737(18)
	N8 - H8 ⋯ Cl1iv	0.86	2.2921(1)	3.1226(2)	162.3797(18)

Symmetry Codes:

⁽ⁱ⁾+x, 1+y, +z;

⁽ⁱⁱ⁾3-x, 1-y, 2-z;

⁽ⁱⁱⁱ⁾1/2 −x, 1-y, −1/2+z;

^(iv)+x, −1+y, +z.

The crystal structure of compound 3 with atom labeling and polymeric structure is shown in Fig. 2. All the C-C (1.37–1.39 Å) distances in the arene ring are within typical range. As with compound 2, the three C–N bonds in the guanidinium moiety are identical (1.32–1.34 Å), suggesting bond delocalization. The difference between the structures of compounds 2 and 3 is the position of the interacting chlorine atoms. In compound 3, the 8-position chlorine atom is involved in a weak intramolecular H-bond interaction with the NH group of the guanidinium moiety N1H (N3–H3) ⋯ Cl11 = 2.583(2) Å. Also, as with compound 2, each chloride anion made four intermolecular hydrogen bond interactions with the guanidinium moiety to generate a polymeric chain along the c axis (Fig. 2 and Table 3). Selected geometric parameters of the guanidine cation are shown in Table 2. The packing diagram of the crystal structure of 3 is shown in Figure S2.

Figure 2.

Top: The molecular structure of 3, showing atom-labeling scheme. Displacement ellipsoids are drawn at the 40% probability level. Bottom: Detail of solid state inter- and intra-molecular H-bonding.

Unlike, compounds 2 and 3, compound 6 lacks the 2-position −NH₂ group. The molecular structure of 6 and the polymeric chain created by H-bonds along the b axis are shown in Fig. 3. In this structure, because of the absence of the −NH₂ group, only two hydrogen-bond interactions were observed with the chloride anions. Note that both nitrogen atoms show protonation, suggesting delocalization of the N–C–N bond. Clearly, the structures of compounds 2 and 3 suggest more packing when compared to compound 6. Intermolecular hydrogen bonds N1H (N6–H6) ⋯ Cl1 (2.2946 (1) Å and N3H (N8–H8) ⋯ Cl1^iv (2.2921(1) Å (symmetry code: +x, −1+y, +z) generated a polymeric chain along the b axis (Fig. 3). The other geometric parameters are similar to compounds 2 and 3 (see Table 3). Packing of the molecules in the crystal structure of 6 is shown in Figure S3.

Figure 3.

Top: Molecular drawing of compound 6. Non-H atoms are drawn as 30% probability displacement ellipsoids. Bottom: Intermolecular hydrogen bonds along b axis are shown as green dotted lines.

The X-ray structures showed similar delocalization of the constrained guanidine groups and thus could not offer a definite explanation of the difference in binding affinity of the positional isomers 2 and 3 at the m5-HT₃A receptor. Hence, we conducted 3D molecular modeling studies to identify possible binding modes and interactions of 2 and 3 at the atomic level. The crystal structures (determined in the current study) of compounds 2 and 3 were docked to a model of the m5-HT₃A receptor crystal structure (PDB: 4PIR) previously reported by us [2]. Note that the X-ray structure and the SYBYL-X2.1.1-built structure of 2 employed in our reported modeling studies were nearly identical (RMSD = 0.031). The docking studies indicated that the three guanidinium nitrogen atoms of 2 and 3 mimic interactions with E236, T181, and S182 as previously reported for 2 [2]. That is, the N1 and N2 nitrogen atoms formed hydrogen bonds with the side chain carboxylate oxygen atom of E236. The N2 nitrogen atom also participated in a hydrogen bond interaction with the side chain hydroxyl group of T181. The N3 nitrogen atom formed a hydrogen bond with the backbone carbonyl oxygen of S182. Because 2 and 3 displayed the same specific guanidinium nitrogen atom interactions, we surmised that the position of the chloro group and its interactions might better explain the lack of affinity of compound 3 (Fig. 4 and S4). The 6-Cl group of 2 seems to participate in hydrophobic interactions with I228, I71, and I207 as well as in an acid/base (halogen bond) interaction with R92 (Fig. S4 and S5). The 8-Cl group of 3 lacked all of these interactions except the I207 hydrophobic interaction (Fig. S4 and S6). Inspection of the receptor-compound complex revealed that the 8-Cl group in 3 faced the bulk solvent compared to the 6-Cl group in 2. Moreover, there appeared to be a possible steric clash between the chloro substituent of compound 3 and F226 and W90.

Figure 4.

Putative binding mode of dihydroquinazoline 2 and 3 (capped sticks rendering; 2, orange carbon atoms; 3, magenta carbon atoms) in the orthosteric binding site of the 5-HT₃ receptor (capped sticks rendering; primary subunit, pink carbon atoms; complementary subunit, cyan carbon atoms). Hydrophobic surface (I228, I71, I207) occludes 6-Cl of 2.

Because of possible intramolecular hydrogen bond interactions between the close proximity of the 8-Cl group and the N1 nitrogen atom, we speculated that this would decrease the ability of N1 to form hydrogen bonds with the protein residues, specifically E236. All of these differences in binding site interactions might contribute to the lack of binding affinity of compound 3. These results were supported by an overall negative total HINT [11] score of 3 (−63) compared to the total HINT score for 2 (148). HINT, named for Hydropathic INTeractions, analysis is a method based on LogPo/w that scores and ranks all non-covalent interactions (hydrophobic and polar) in ligand-protein complexes by giving positive scores for favorable (hydrogen-bonding, acid-base, and hydrophobic) interactions and negative scores for unfavorable (acid-acid, base-base, and hydrophobic-polar) interactions [11]. HINT score analysis showed numerous differences in the interactions of the chloro groups in 2 and 3 with the m5-HT₃A receptor. For example: the lower total hydrophobic contribution score of compound 3 (163) as compared to compound 2 (251) and the clashes with F226 (both hydrophobic and hydrophobic/polar) for 3 (−106) vs 2 (−28) might enlighten our understanding of the poor binding affinity for 3. The relatively low total HINT score for 2 can be explained by the fact that there are numerous energetic contributions that are factored into the free energy calculation of a receptor-ligand complex that are not accounted for by HINT. These contributions should theoretically result in a more favorable interaction between a receptor and a ligand; however, because HINT only accounts for localized binding site interactions of the ligand, other conformational changes within other regions of the receptor remain unaccounted for (resulting in the low total HINT score). Such contributions include van der Waals, electrostatic, and conformational entropy energy values. That being said, it is still useful to compare the total HINT scores of similar compounds such as 2 and 3 because both of their guanidinium nitrogen atoms participate in similar hydrogen-bond interactions with the receptor and so their scaffolds are approximately aligned within the binding site. It appears that our 3D model of the 3-protein complex is in agreement with the biological data.

Conclusions

The crystal structures of all three compounds showed that the guanidinium/amidinium moiety is involved in hydrogen bond interactions with a chloride anion. Furthermore, the structures revealed nearly identical C–N bond distances suggesting delocalization. In particular, the delocalization found for 3 was different than that reported for 4. It is unlikely, then, that differences in delocalization explain the difference in the 5-HT₃ receptor affinity of 2 and 3. However, the hydrogen bonding in 3 might be geometry related. That is, the N1H of 3, due to possible steric hindrance introduced by the 8-chloro group, might be unable to form a hydrogen bond with a necessary receptor feature. Hence, compound 2 with four potential hydrogen bond interactions between the guanidine moiety and the receptor, should bind with higher affinity than 3, consistent with the biological data. Additionally, the 6-chloro group of 2 formed productive hydrophobic and acid/base interactions with the 5-HT₃ receptor. That is, the 6-chloro group of 2 formed more and favorable interactions with the receptor, whereas, the 8-chloro group of 3 showed fewer and unfavorable interactions.

Compound 6 (K_i = 1.1 μM), the desamino analog of 2, binds at 5-HT₃ receptors with lower affinity than 2. As previously demonstrated, all three nitrogen atoms of 2 are required for optimal receptor affinity. The reduced affinity of 6 cannot be ascribed to differences in N–C–N delocalization and must, of consequence, be attributed to the lack of the 2-amino group and the inherent decrease in the number of productive hydrogen bonds it makes with the receptor.

Highlights.

1. Certain quinazoline analogs bind at 5-HT₃ serotonin receptors; certain other do not.

2. X-Ray structures were solved to determine if delocalization was involved in their binding.

3. 3D Modeling suggests that a 6-Cl dihydroquinazoline analog binds in a positive fashion whereas the 8-Cl analog produces steric clashes.