2022 Volume 70 Issue 4 Pages 293-299
We designed and synthesized non-peptide organic molecular ligands for integrin αvβ3. Candidate ligands featured amidino analog and carboxy groups as binding sites on either side of a spacer, which consisted of benzophenone or an analog, such as diphenyl sulfide, diphenyl sulfoxide, diphenyl sulfone, or diphenyl ether. Competitive binding assays to integrin αvβ3 with respect to [125I]echistatin were used to determine inhibitory activity of the synthetic ligands. Ligands bearing 2-aminobenzimidazoyl and glycyl groups separated by a benzophenone spacer demonstrated more potent binding than did a linear Arg-Gly-Asp (RGD) tripeptide that represents the native integrin αvβ3 binding motif. Ligands possessing 2-aminobenzimidazoyl and carboxy groups and diphenyl sulfoxide or diphenyl ether spacers inhibited binding of [125I]echistatin with IC50 values similar to that of the linear RGD tripeptide.
Integrins are transmembrane heterodimeric receptors that are associated with multiple cellular processes, including adhesion, extracellular matrix organization, survival and proliferation. At least 18 α subunits and eight β subunits have been identified in humans. The functions of a particular integrin complex are determined by the combination of α and β subunits.1–3)
Integrin αvβ3, which is mainly expressed on epithelial, melanoma, and glioblastoma cells, is known to be related to the promotion of vascularization and bone formation. This integrin is overexpressed on the surfaces of some kinds of tumor cells, and it plays crucial roles in tumor angiogenesis and metastasis by facilitating the migration of endothelial and tumor cells.4–7) Therefore, integrin αvβ3 is utilized as a molecular target for tumor imaging and drug delivery systems.8–12)
Peptides and proteins containing the Arg-Gly-Asp (RGD) motif are known to be effective ligands for integrin αvβ3.13,14) In particular, cyclic RGD peptides show high binding affinity against integrin αvβ3, and the cyclic RGD derivative cyclo(Arg-Gly-Asp-D-Phe-Val) (c(RGDfV)) has been utilized as a lead compound in drug design efforts.15,16) The addition of a reactive amino group by substituting lysine for valine in c(RGDfV) yields c(RGDfK), which can be readily conjugated to various functional moieties. For example [18F]-galacto-c(RGDfK)17) and [99mTc]-HYNIC-Ec(RGDfK)18) were developed as tracers for positron emission tomography (PET) and single photon emission computed tomography (SPECT), respectively, to facilitate tumor visualization in vivo. Modified peptides based on c(RGDfK) have also been utilized for the surface decoration of drug carriers such as liposomes and polymeric micelles.19–23) Intratumor accumulation and cellular uptake of these c(RGDfK)-modified carriers were enhanced, and drugs encapsulated into the carriers were delivered with higher efficiency and lower side effects.
While cyclic RGD peptides are proving to be powerful ligands for molecular therapy targeting integrin αvβ3, there remain some problems limiting their utilization in the clinic. For example, the native peptide sequence is quickly degraded by metabolic processes in vivo. To offset this issue, at least one D-amino acid can be added as a constituent of the cyclic RGD peptide; this modification results in a significant stabilization of cyclic RGD peptides.24) In addition, the peptides themselves can be expensive to produce. Synthetic methods leading to more efficient production of the RGD peptide have been developed and enhanced, and some peptide derivatives are commercially available; however, cost remains an important issue.
To overcome these problems, syntheses of antagonists that mimic the RGD peptide have been attempted.25,26) In the present study, we synthesized a small organic molecular ligand for integrin αvβ3 based on a modular design consisting of two integrin binding units and a linker unit. The modularity of this structure makes it possible to control the distance and angle between the binding units. The binding affinity of these ligands to integrin αvβ3 was evaluated.
Ligand candidates were designed using RGD peptides as base compounds. X-ray crystal structure analysis of integrin αvβ3 with a peptide ligand that included the RGD motif revealed that the Arg guanidino group and the Asp β-carboxy group play crucial roles in forming hydrogen and coordination bonds with residues in the αv and β3 subunits, respectively.27) The binding affinity of the cyclic RGD peptide is generally higher than that of the linear form, because the distance and orientation of the guadinino and carboxy groups are fixed by cyclization. These results suggest that the arrangement of these binding moieties is of importance in the design of efficient ligands.
In the present study, candidate ligands were classified as having three components: units A through C. Two components were units for binding to integrin. These components include a guanidino or modified guanidine group (unit A) and a carboxy group (unit B), respectively. These two units were connected via a linker component termed unit C, which controls the distance and orientation of the two binding functionalities (Fig. 1).
Benzophenone, which has a relatively rigid conformation, was selected to serve as the basis for unit C. To control the bond angle of the two benzene rings for fine-tuning of the distance between the two integrin-binding functionalities, benzophenone derivatives, such as diphenyl sulfide, diphenyl sulfoxide, diphenyl sulfone, and diphenyl ether, were also employed as unit C in various constructions. Guanidine equivalents, including 2-aminobenzimidazole, urea, and thiourea moieties, which can form hydrogen bonds with integrin carboxylates, were also selected for unit A to simplify the synthetic process.
We endeavored to create small molecule ligands that matched the orientations of key atoms in the RGD peptides as closely as possible. Before synthesis of designed ligands, then, an in silico conformational study was undertaken using the OPLS2005 force field in MacroModel (version 9.1; Schrödinger, Inc.; New York, U.S.A.). The predicted structures and the distances between the carbon atoms in the guanidino group or analogues and the carboxy groups are shown in Supplementary Figures S1–S15 and Table 1. As unit A can be in U-shaped (syn) or zig-zag (anti) conformations, we performed calculations with both conformational isomers. The distances in the calculated structures, between 12.5 and 14.3 Å, if the better conformers are picked up, do not perfectly match the analogous distance in the RGD crystal structure, because X-ray crystallography has demonstrated that the distance between the carbon atoms in integrin-binding guanidino and carboxy groups of the RGD peptide is approximately 13.7 Å.28) However, the flexibility of atoms around the carboxy group seems to adapt the ligand to the binding sites in integrin αvβ3, because the resulting hydrogen bonds and coordination bonds have been found to be strong.
 |
a The optimized structures were calculated with MacroModel 9.1 using OPLS2005 force field. b syn: U-shaped-conformation of unit A, and anti: zig-zag-conformation of unit A. See Supplementary Figs. S1–S15. c Ligand 7 has two carboxy groups.
The distances between C-4 and C-4′ in the two benzene rings of unit C are also shown in Table 1. The distances in ligands containing benzophenone (1–8), thioether (9 and 12), diphenyl sulfoxide (10 and 13), diphenyl sulfone (11 and 14), and diphenyl ether (15) are 7.5, 6.7, 6.9, 7.2, and 7.2 Å, respectively. Thus, the distances of key functionalities can be regulated by changing of the identity of the spacer.
We synthesized target molecules 1 through 5, which all possess an amidine moiety or an analog of amidine in unit A. Molecules with urea (1) and thiourea (2) moieties were synthesized as shown in Chart 1. An ester moiety was introduced by alkylation of benzophenone A129) and haloacetate, then bromination of A3 followed by amination of A4 afforded the amine A5, which was converted to urea A6 and thiourea A7. Finally, hydrolysis of esters A6 and A7 gave the target molecules 1 and 2, respectively.
Reagents and conditions: (i) ethyl bromoacetate or tert-butyl chloroacetate, K2CO3; (ii) NBS, benzoyl peroxide; (iii) NaN3; (iv) PPh3, H2O; (v) BuNCO or BuNCS, (vi) H+.
Synthesis of ligand 3, bearing amidino and carboxy groups, began with bromination of A2 (Chart 2). After amination of bromide A8, guanylation of amine A9 was carried out in the presence of Mukaiyama reagent.30,31) Treatment of A10 with sulfuric acid produced ligand 3.
Reagents and conditions: (i) NBS, benzoyl peroxide; (ii) NaN3; (iii) PPh3, H2O; (iv) 2-chloro-1-methylpyridinium iodide, Et3N, 1,3-bis(tert-butoxycarbonyl)thiourea; (v) H+.
Regioselective induction of the 2-aminobenzimidazole moiety was completed (Chart 3); reductive amination of aldehyde A11, synthesized from A8, and 2-aminobenzimidazole and alkylation of A8 and 2-aminobenzimidazole afforded the corresponding amines A12 and A13, which were converted into carboxylic acids 4 and 5, respectively, under basic conditions.
Reagents and conditions: (i) hexamethylenetetramine; (ii) AcOH, H2O; (iii) 2-aminobenzimidazole; (iv) NaBH(OAc)3; (v) NaOH then H+; (vi) 2-aminobenzimidazole, K2CO3.
Next, we modified unit B. Nitrobenzophenone B132) was used as the starting material for synthesis of ligand 6, which is N-analog of 4 (Chart 4). Dibromination and hydrolysis gave aldehyde B2, which was protected as bisacetal B3. Reduction of the nitro group, accompanied by hydrolysis of the ketal, followed by alkylation of B4, afforded aminoester B5, then conversion of aldehyde B6 to ligand 6 was completed by the same procedure as described above (from aldehyde A11 to amino acid 4). Diacid 7 was prepared from aldehyde B9, which was prepared from monoester B5 by alkylation and hydrolysis of acetal B8, then reductive amination of B9 and saponification gave the diacid 7.
Reagents and conditions: (i) NBS (3.5 equiv.), benzoyl peroxide; (ii) H2O, DMSO; (iii) ethylene glycol, TsOH; (iv) Fe, NH4Cl; (v) isopropyl bromoacetate, K2CO3; (vi) H+; (vii) 2-aminobenzimidazole; (viii) NaBH(OAc)3; (ix) NaOH then H+; (x) TfOCH2CO2iPr, K2CO3.
Preparation of C-analogue 8 is shown in Chart 5. Reductive amination of B1133) as described, and protection of the corresponding amine B12 gave B13. The C-analog moiety was constructed via the Heck reaction, and reduction of B14 and hydrolysis of the corresponding ester B15 afforded the acid 8.
Reagents and conditions: (i) 2-aminobenzimidazole; (ii) NaBH(OAc)3; (iii) Boc2O, DMAP; (iv) ethyl acrylate, Pd(OAc)2, PPh3, Et3N; (v) Pd-C, H2; (vi) NaOH then H+.
Preparation of sulfide 9, sulfoxide 10, and sulfone 11 commenced with the Cu-mediated coupling reaction of iodotoluene and thiophenol developed by Basu et al.34) (Chart 6). Alkylation of phenol C1 and haloacetates gave the corresponding esters (C2 and C3), which were transformed into bromides (C4 and C5). We used Gabriel amination of C5 to obtain amine C7 via C6, and amine C7 was treated with isocyanate to give urea C8. Oxidation of sulfide C8 using sodium bromate in the presence of sodium bromide35) afforded the corresponding sulfoxide C9, and transformation from C8 into sulfone C10 was carried out under boric acid/hydrogen peroxide oxidation conditions.36) Finally, deprotection of esters C8, C9, and C10 under acidic conditions gave acids 9, 10, and 11, respectively, which establish urea and carboxylic acid moieties within unit B.
Reagents and conditions: (i) ethyl bromoacetate or tert-butyl chloroacetate, K2CO3; (ii) NBS, benzoyl peroxide; (iii) potassium phthalimide; (iv) NH2NH2; (v) BuNCO; (vi) NaBr, NaBrO3, H+; (vii) B(OH)3, H2O2; (viii) H+.
Ligands 12 through 14, bearing aminobenzoimidazole moieties and carboxy groups, were prepared (Chart 7). Reaction of bromide C4 and hexamethylenetetramine followed by treatment of acid gave aldehyde C13, which was converted by reductive amination to give C15. Treatment of C15 with Boc2O selectively produced monocarbamate C17. In the NOESY spectrum of C17 (Supplementary materials, nuclear Overhauser effect spectroscopy (NOESY) spectrum of C17), the cross peaks between NH and benzylic proton signals were used to determine the structure of C17. Oxidation of sulfide C17 afforded the sulfone C18; however, treatment of C17 with NaBrO3/NaBr promoted oxidation of sulfide and bromination reaction of the aromatic ring to give an undesired bromosulfoxide. Therefore, starting sulfide C2 was transformed into sulfoxide C11, then Wohl–Ziegler reaction of C11 followed by formylation and reductive amination of C14 and 2-aminobenzimidazole gave the corresponding sulfoxide C16. Finally, synthesis of 12, 13, and 14 were completed by deprotection of C15, C16 and C18, respectively.
Reagents and conditions: (i) NaBr, NaBrO3, HCl; (ii) NBS, benzoyl peroxide; (iii) hexamethylenetetramine; (iv) AcOH, H2O; (v) 2-aminobenzimidazole; (vi) NaBH(OAc)3; (vii) Boc2O, Et3N, (viii) B(OH)3, H2O2; (ix) NaOH then H+.
Similarly, diphenyl ether analogue 15 was prepared from D237) (Chart 8). In a similar manner, alkylation of D2 and reductive amination of D3 followed by hydrolysis of ester D4 gave ligand 15.
Reagents and conditions: (i) pyridinium chloride; (ii) ethyl bromoacetate, K2CO3; (iii) 2-aminobenzimidazole; (iv) NaBH(OAc)3; (v) NaOH then H+.
A competitive binding assay was employed to determine the affinity of the ligand candidates against recombinant human integrin αvβ3. Various kinds of ligands for integrin αvβ3 have been reported.38) Among them, Echistatin, which belongs to the disintegrin family, is a 49-residue protein containing an RGD motif and binds to integrin αvβ3 with high affinity in an irreversible manner.39) Therefore, binding affinities of the ligands were examined by determining ability to compete for binding with 125I-radiolabeled echistatin.
Results of the binding assay are summarized in Table 2. IC50 values of the RGD tripeptide and unlabeled echistatin used as controls were 6.03 ± 0.83 × 10−5 and 4.43 ± 0.44 × 10−10 M, respectively. The inhibition by the two ligand candidates 6 and 7 was more potent than that of the RGD tripeptide; the IC50 values for compounds 6 and 7 were 7.66 ± 4.62 × 10−6 and 2.73 ± 1.22 × 10−5 M, respectively.
| Ligand | IC50 (×10−6 M) | Ligand | IC50 (×10−6 M) |
|---|---|---|---|
| 1 | n.d. | 10 | n.d. |
| 2 | n.d. | 11 | n.d. |
| 3 | n.d. | 12 | >400 ± 323 |
| 4 | n.d. | 13 | >159 ± 120 |
| 5 | n.d. | 14 | >207 ± 88.1 |
| 6 | 7.66 ± 4.62 | 15 | >172 ± 100 |
| 7 | 27.3 ± 12.2 | RGD peptide | 60.3 ± 8.28 |
| 8 | 106 ± 33.6 | Echistatin | 0.000443 ± 0.0000442 |
| 9 | n.d. |
IC50 is mean ± standard deviation (S.D.) of three repeated experiments. Those written as n.d. shows the average value was over 1*10−3 M. IC50 values of compounds 12–15 are relatively high, and lack of accuracy. Table 2 shows the minimum expected IC50 value, and therefore, the numbers are marked with “>.”
Ligands 6 and 7, which show low IC50 values in the inhibition of binding of [125I]-echistatin to integrin αvβ3, possess a 2-aminobenzimidazole moiety instead of guanidine group in unit A (Table 2). Furthermore, all compounds showing inhibitory activity have the 2-aminobenzimidazole moiety in unit A. Fabbrizzi et al. reported that π-stacking interactions between Arg mimetics in the area corresponding to unit A of ligands and the Tyr178 residue of integrin β3, which is located at the RGD peptide binding domain, are important for ligand recognition.40) Therefore, we suspect that π-stacking interaction between the benzimidazole moiety and hydroxyphenyl group in Tyr178 residue supports the binding to integrin β3. While isomeric molecule 5 also contains the 2-aminobenzimidazole moiety, it did not show inhibitory activity.
The IC50 values of ligands bearing urea, i.e., molecules 9 (X = S), 10 (X = SO), and 11 (X = SO2), were larger than those of 12, 13, and 14, which bear aminobenzoimidazole instead of urea in unit A. These results suggest that the weak hydrogen-bonding ability of urea and thiourea in comparison to aminobenzoimidazole results in lower integrin-binding affinities.
Ligands 4, 12, 13, 14, and 15 were designed to have different unit C moieties in order to control the bond angle of the two benzene rings. Here, ligands 12 (X = S), 13 (X = SO), 14 (X = SO2), and 15 (X = O) showed inhibition activities, but ligand 12 showed less inhibitory activity, and ligand 4 did not show any detectable inhibitory activity. According to the simulations of these compounds, the distances between carbon atoms in aminoimidazoyl and carboxy groups in 4 (X = C=O) were not a suitable value (13.2 Å); therefore, it could be predicted that 4 would not be in a favorable conformation for binding to integrin αvβ3. Sulfoxide 13 showed the best inhibition activity among these ligands; therefore, diphenyl sulfoxide might be a suitable spacer among urea-based ligands. The distances determined in the molecular simulation suggested that sulfone 14 would be the optimal spacer (13.69 Å), but in practice IC50 value of 14 was larger than that of RGD peptide. The orientations of the amidino and carboxy functionalities in calculated structure of 14 might not fit completely within the binding pocket. From the molecular simulation, ligand 15 (X = O) has better spacer, but its activity was close to that of 13.
Importantly, ligands 6, 7, and 8 have the same benzophenone spacer as ligand 4 and inhibited integrin binding with high efficiency. These results might be due to the favoring of a planar and rigid conformation by the benzophenone-type spacer. Ligand 4 was found to be poorly soluble, and it was thought to be stacked or recrystallized during the binding assay and thus could not enter the binding domain of integrin αvβ3. In the case of ligands 6 and 7, however, the dicarboxylic acid and glycil groups in unit B improved water solubility. In consideration of these results, we suggest that a rigid benzophenone is the favored spacer for the ligand. Furthermore, ligand 8, a carbon analogue of 4 in unit B, could bind integrin αvβ3, indicating that the corresponding binding pocket is sterically unhindered.
In this study, we designed non-peptide organic molecular ligands that featured analogs of amidino group (unit A) and carboxy group (unit B) as binding sites for integrin αvβ3, and these groups were separated by spacers (unit C). Molecular modeling suggested that the distances between the two binding functionalities matched optimal distances as calculated from structural analyses of RGD-based ligands. We synthesized 15 non-peptide ligands, and competitive integrin αvβ3 binding assays with respect to [125I]echistatin were used to determine the inhibitory activities of the synthetic ligands.
2-Aminobenzimidazoyl group in unit A is effective for the inhibition activity, and glycil group in unit B is improved the activity in comparison with the corresponding C- and O-analogues. Benzophenone and related moieties in unit C effectively work as spacers for regulation the distances between the two binding sites. Finally, IC50 values of two ligands were better than that of RGD peptide.
As the non-peptide ligands, which have been synthesized in this study, consist of completely new spacers for integrin αvβ3 binding, stabilization of the ligands against hydrolysis in vivo could be improved, because the ligands do not have any amide and ester moieties. Furthermore, high-affinity and selective binding ability to integrin αvβ3 are next subjects for development of powerful ligands for molecular therapy targeting integrin αvβ3.
Unless otherwise noted, reagents and solvents were used as received from commercial sources. RGD tripeptide and echistatin were purchased from Wako (Japan) and Sigma (U.S.A.), respectively. [125I]-Echistatin was from PerkinElmer, Inc. (U.S.A.), and recombinant human integrin αvβ3 was from R&D Systems, Inc. (U.S.A.).
Syntheses and CalculationsThe details of syntheses of all compounds are described in Supplementary Materials. All new compounds were characterized by 1H- and 13C-NMR, IR spectroscopy, and mass spectrometry. The optimized structures of newly designed ligands were calculated with MacroModel 9.1 (Schrödinger, Inc., New York, U.S.A.) using the OPLS2005 force field.
Binding AssayThe binding assay was performed as follows based on a previously reported method.31) A buffer solution (coating buffer) containing 20 mM Tris, 150 mM NaCl, 2 mM CaCl2, 1 mM MgCl2 and 1 mM MnCl2 was prepared and adjusted to pH 7.4 with HCl. Blocking buffer containing 50 mM Tris, 100 mM NaCl, 2 mM CaCl2, 1 mM MgCl2, 1 mM MnCl2 and 1% bovine serum albumin (BSA) was also prepared, and the pH was adjusted to 7.4 with HCl.
Human recombinant integrin αvβ3 was dissolved in coating buffer at a concentration of 500 ng/mL, and the solution (100 µL) was added into wells of 96-well plates (Thermo Fisher Scientific, U.S.A.). The plates were incubated at 4 °C overnight, and the solution was removed. Blocking buffer (200 µL) was added to the well. After incubation at room temperature for 2 h, the wells were washed two times with blocking buffer.
Purified ligand candidate, RGD peptide and echistatin were dissolved in the blocking buffer at concentrations of 0.5 nM to 5000 µM. To the 96-well plate, blocking buffer (110 µL), the ligand solution (40 µL), and 0.2 nM [125I]-echistatin solution (50 µL) were added, and the mixture was incubated at room temperature for 3 h. After removal of the solution, the well was washed three times with blocking buffer. The 96 well plates were separated, and radioactivity in each well was measured by a γ-counter (PerkinElmer, Inc., Wizard).
On the assay, experiment of n = 3 was repeated three times. From the data obtained from each experiment, IC50 value was calculated using GraphPad Prism 6 (GraphPad Software, Inc., U.S.A.).
Part of this research was financially supported by Life Science Innovation Center, University of Fukui.
The authors declare no conflict of interest.
This article contains supplementary materials.
