Deciphering RGDechi peptide-α₅β₁ integrin interaction mode in isolated cell membranes

2.1 Cell lines and culture conditions

Chronic myelogenous leukemia cells line (K562) (ATCC U.S.) were grown in RPMI added with heat-inactivated 10% FBS (fetal bovine serum), 2 mM glutamine, 100 U mL⁻¹ penicillin and 100 μg mL⁻¹ streptomycin (Euroclone, Italy). Human adenocarcinoma cells line (HeLa), kindly provided by Dr. Annalisa Tito (Arterra Bioscience), were grown in DMEM added with 10% fetal bovine serum (FBS), 2 mM glutamine, 100 U mL⁻¹ penicillin and 100 μg mL⁻¹ streptomycin. Cells were maintained in humidified air which contained 5% CO₂ at 37°C.

2.2 FACS analysis for α₅β₁ integrin

Cell aliquots (2.5 × 10⁵ cells, suspended in cold 1x PBS containing 0.5% BSA) were treated with primary monoclonal antibody anti-α₅β₁, anti-α_vβ₃ or anti-α_vβ₅ (Millipore MA), or isotype control (Santa Cruz Biotechnology, Germany), at a final concentration of 14 µg mL⁻¹, in a whole reaction volume of 50 µL for 45 min at 4°C. After washing, the cells were incubated with secondary antibody PE-conjugated (Thermofisher, MA), washed and analyzed. For the analysis, a flow cytometer equipped with a 488 nm argon laser (FACScan, Becton Dickinson, USA) was used. The fluorescence intensity values were attained from histogram statistic of CellQuest software.30

2.3 Cell adhesion and detachment assays

NUNC MaxiSorp 96 well plates (Dasit Sciences, Italy) were covered at 4°C overnight with fibronectin (Millipore, USA) diluted at 10 μg mL⁻¹ in PBS, pH 7.4; alternatively, coating was carried out with 10 μg mL⁻¹ anti-α₅β₁. Successively, a blocking step of 1 h at room temperature for nonspecific binding was done with 1.5% BSA in PBS. K562 cells were suspended and mixed for 30 min at 4°C to prevent receptor internalization, in Hank's balanced salt solution24 with peptides (50 μM) or anti-α₅β₁ antibody (10 μg mL⁻¹). Afterward, cells were plated on precoated plates (1.5 × 10⁴ cells/well). Nonadherent cells were delicately removed by a series of washings. After 1 h of incubation, the adherent cells were counted by crystal violet (Sigma Aldrich, Italy) assay, which correlates optical density with cell number.31 Statistical differences were calculated using the Student's test, unpaired, two-sided. All experiments were done in triplicate and repeated three times at least; a P value <0.05 was evaluated as significant.

2.4 Membrane preparation for NMR studies

K562 cellular membranes were used for NMR experiments. Cells were washed two times using PBS, then were moved into homogenization buffer, homogenized and centrifuged at 4°C for 1 h at 28 000 rpm. The homogenization buffer consisted in 20 mM Tris-HCl pH 7.4 containing 10% glycerol and protease inhibitor cocktail (Roche). The pellet was then washed and centrifuged at 4°C for 30 min at 28 000 rpm. The membranes were resuspended with 180 μL of HBSS.29

2.5 Nuclear magnetic resonance spectroscopy

NMR measurements were performed utilizing an Inova 600 MHz spectrometer (Varian Inc., Palo Alto, CA) at 298 K equipped with a cryogenic probe optimized for ¹H detection. Samples were prepared as above reported.

NMR binding studies of RGDechi with isolated K562 cell membranes were performed, according to the methods reported in Farina et al.29 Specifically, isolated K562 cell membranes from 80 × 10⁶ cells were resuspended in 180 μL of HBSS buffer (pH 7.4) and 10% ²H₂O in 3 mm tube. ¹H standard and Saturation Transfer Difference (STD) NMR spectra were acquired as reference. A 300 nmol of peptide was initially dissolved in 20 μL of HBSS buffer and then added to the membrane suspension. STD spectra were acquired with 10 000 scans with on-resonance irradiation at 0.2 ppm to selectively saturate protein resonances and off-resonance irradiation at 30 ppm for reference spectra. A total saturation time of 2 s has been obtained using a train of 40 Gaussian shaped pulses of 50 ms with 1 ms delay between pulses. STD spectra were achieved after internal subtraction of the saturated spectrum from the reference spectrum by phase cycling with a 7191.66 Hz spectral width, 1.0 s relaxation delay, 8 k data points for acquisition, and 16k for transformation. The STD effect was defined as the percentage of the STD factor relative to the maximum, set to 100%. In the case of the HN Arg2 and Arg11 sidechains, considering the solvent exchange contributions, the STD values were calculated by taking as reference the STD profile of the HN Lys1 backbone. 2D [¹H-¹H] TOCSY (TOtal Correlation SpectroscopY) and NOESY (Nuclear Overhauser SpectroscopY) spectra of RGDechi in the presence of isolated membranes were acquired similarly to those of the peptide alone with a mixing time of 70 and 250 ms, respectively.29 To test for the specific interaction between α₅β₁ integrin and RGDechi, the same experiments were performed in the presence of anti-α₅β₁ antibody (10-fold excess to the α₅β₁ integrin). Moreover, since for STD experiments using living cells it was suggested to place the off-resonance pulse at ±300 ppm,32 we performed as control a STD experiment using two irradiation frequencies at 30 and 300 ppm. The spectrum difference did not show signals from membranes, but only from low molecular weight impurities (i.e., glycerol, acetonitrile), confirming that 30 ppm represents a suitable off-resonance frequency for the membrane preparations (Figure S1, Supporting Information).

Data were processed by means of the software VNMRJ 1.1.D (Varian Inc.). Analysis of 1D spectra were performed utilizing ACD/NMR Processor 12.0 (www.acdlabs.com). 2D TOCSY and NOESY spectra were examined by means of CARA (Computer Aided Resonance Assignment) software.33

2.6 Molecular dynamics simulations

Molecular dynamics (MD) simulations were carried out using GROMACS34 software. The Amber99SB35 force field was utilized with explicit TIP3P water.36 The theoretical model of the α_vβ₃/RGDechi complex reported in a previous manuscript21 was used to obtain the starting 3D structure for the RGDechi peptide. Energy minimization was performed using 100 steps of steepest descent to avoid any kind of bad contacts generated while solvating the system. All simulations utilized a 2 fs inner step and were completed in NVT ensemble utilizing a Nose-Hoover thermostat after equilibration to constant box volume in the NPT ensemble. All simulations were run at 298 K. Electrostatic and van der Waals interactions were measured by means of partial mesh Ewald algorithm; the short-range neighbor list cut-off, short-range electrostatic cut-off and short-range van der Waals cut-off were fixed at 1 nm. The time-averaged NOE distance restraints were imposed with a force constant of 6 000 kJ mol⁻¹ nm⁻² and was utilized a 20 ps memory relaxation time.37, 38 Numerous simulations were run and the ensemble attained from the 10 ns MD simulation was considered to get the reference structure. In this case, the snapshots were saved every 100 ps in order to have 100 conformers total in the conformational ensemble. The obtained structures were clustered by using the tools implemented in UCSF Chimera39 and for each cluster the representative structure has been selected. The comparison of the representative structures has been performed by using UCSF Chimera.39 All molecular dynamics simulation models were evaluated by using the PyMOL40 and MolMol41 software.

2.7 Molecular docking protocol

The molecular docking studies were carried out using the software AutoDock 4.0.42 The optimized Molecular docking protocol involved several steps: (1) Preparation of α₅β₁ receptor and RGDechi peptide structures: ligand and integrin coordinates files were initially prepared to include the required information for the docking calculation (atom types, spatial charges, polar hydrogen atoms and torsional degrees of freedom). In particular, for the receptor the X-ray structure of the complex between the α₅β₁ integrin and the RGD peptide (PDB ID code: 3VI4) was utilized as reference structure; whereas the representative structure of the main cluster selected from the 10 ns MD ensemble was used for RGDechi. Polar hydrogens were added to the reference structure by means of the software REDUCE.43 and the RGDechi rotatable bonds were selected automatically. (2) Performing the Autogrid calculation: the tool Autodock needs pre-calculated grid maps obtained by AutoGrid. The grid map is defined as a 3D lattice of regularly spaced points, surrounding (either entirely or partly) and centered on some region of interest of the receptor. The grid size was chosen to be 64 × 64 × 64 with grid space of 0.658 Å. (3) Performing the docking calculation by using Autodock: a docking file containing the input parameters for the docking calculation was generated by the AutoDockTools. In our molecular docking protocol the Lamarckian genetic algorithm (LGA) search method was chosen and the minimized ligand was randomly located within the grid box. The docking process was started with the following parameters: torsion step 58, torsional degrees of freedom 3, number of energy evaluations 2 500 000 and run number of 100. Finally, the final representative structure of each cluster was minimized by using the UCSF Chimera39 tools and validated by using PROCHECK44and MolProbity45 (Supporting Information Table S1).

The relative STD factors (relative NOE) were calculated for each RGDechi residue of the analyzed cluster pose as a sum of intermolecular peptide-receptor 1/r6 distances. In particular, as previously described,44 we determined Ʃ(1/r6) for each residue of the RGDechi peptide, where r is the inter-proton distance between each RGDechi nonexchangeable proton and surrounding receptor protons within 6.5 Å. The structures were clustered evaluating the binding free energy of the complex at the end of the docking simulation. The structures were displayed and analyzed by means of the software PyMOL40 and MolMol.41

3.1 Expression of α₅β₁ integrin on K562 and HeLa cells

The expression levels of α₅β₁ integrin on membrane surface of K562 and HeLa cells were evaluated by means of flow cytometric analysis utilizing anti-α₅β₁ monoclonal antibody. K562 cells resulted to express α₅β₁ at high levels_, and α_vβ₃ or α_vβ₅ at very low levels (Figure S2, Supporting Information), and HeLa cells did not express α₅β₁ (Figure S3, Supporting Information). These results indicate that K562 cells represent a good model to study α₅β₁ interaction.

3.2 RGDechi effect on cell adhesion

The effect of RGDechi (synthesized as previously reported24) on the adhesion of K562 cells seeded onto fibronectin (extracellular matrix recognizing a range of different integrins, including α_vβ₃ and α₅β₁ receptors) or, alternatively, on α₅β₁ antibody was studied in the specific cell adhesion buffer as reported in a previous manuscript.21

As shown in Figure 2A, RGDechi decreases the α₅β₁ mediated adhesion of K562 cells plated onto fibronectin having similar effect of cRGDfK, a α_vβ₃/α_vβ₅/α₅β₁ antagonist used as positive control12, 46, 47 and the adhesion inhibition by RGDechi occurs in a concentration-dependent mode (Figure 2B). Incubation with scrambled peptide (Ac-KPGRGHNDPDPGhCitDeMHAT-OH),24 utilized as negative control, does not decrease K562 adhesion onto matrix protein or specific antibody. To highlight the binding to α₅β₁ integrin, a further adhesion assay was carried out onto anti-α₅β₁ antibody coated plate (Figure 2C). Data obtained indicate that RGDechi decreases the adhesion of K562 with an activity slightly but significantly lower than cRGDfK. More, the adhesion assay performed on HeLa cells not expressing α₅β₁ (Supporting Information Figure S3) and α_vβ₃ and likely expressing other fibronectin binding integrins clearly confirms the binding of peptide to α₅β₁ (Supporting Information Figure S4).

Very recently, Kapp et al. reported the IC₅₀ values, calculated by competitive ELISA assay, of cRGDfK peptide for different integrins subtypes including α_vβ₃ (IC₅₀ = 2.25 ± 0.34 nM) and α₅β₁ (IC₅₀ = 141 ± 15 nM).12 On the basis of these data and considering that RGDechi inhibits the adhesion of cell seeded onto α_vβ₃ antibody plates similarly to cRGDfK (Figure 7C in Farina et al.29), we can reasonably conclude that RGDechi significantly binds to α₅β₁ but with an affinity lower than that to α_vβ₃ integrin.

3.3 NMR binding studies of RGDechi to α₅β₁ integrin using K562 membrane preparations

In the last years, we have characterized the structures and functions of peptides recognizing membrane receptors either in vitro and on cellular membranes.29, 48, 49 Recently, we have also investigated the use of isolated cellular membranes and provided significant advantages in the structural study of the RGDechi interaction with α_vβ₃ integrin.29 On this basis, to gain structural insights into the binding of RGDechi to α₅β₁ integrin, STD and trNOESY experiments were performed using K562 membrane preparations.

Interestingly, STD spectrum indicates that in presence of isolated K562 membranes RGDechi receives saturation transfer (Figure 3A). This effect, negligible in the absence of cell membranes (Figure S5, Supporting Information), indicates the binding of the peptide to the K562 membrane cells and, thus, to the α₅β₁ integrin. To further confirm that the STD effects observed are ascribed to the specific binding between RGDechi and α₅β_1, STD experiments were repeated in the presence of the anti-α₅β₁ antibody. Interestingly, a strong intensity reduction in the STD spectrum was observed, according to the presence of a 10 folds excess of the antibody with high affinity for α₅β₁ and a large excess of the peptide with lower affinity (Figure S6, Supporting Information). This result confirms the specificity of the effects observed above. Relative STD values were calculated for amide, aromatic and aliphatic protons and pictured as a function of the amino-acid residue (Figure 3B). STD effects from K562 membranes higher than average and higher than average plus one standard deviation (SD) were classified, respectively, as medium or strong. In the amide/aromatic region, the Lys1, Gly3, Asp4, Gly16 backbone amides and Arg2 and Arg11 side chain amide protons show strong STD effects (Figure 3B, top). Medium STD signals were assigned to Arg2 overlapped to Gly10, Asp8, Asn12, His14, and hCit15 H_Ns. Strong STD effects in the aliphatic region were ascribed to the Arg2 H_β2 and H_β3, DGlu5 H_α and H_β3, and Pro13 and Pro17 H_β protons (Figure 3B, bottom). Medium STD effects were observed for Arg2 H_γ, Asp4 H_β2 and H_β3, Arg11 H_α and H_γ, hCit15 H_γ, H_δ and H_ɛ, and Pro17 H_γ. These data outline a RGDechi surface interacting with α₅β₁ that shares a certain similarity to that of RGDechi with α_vβ₃ previously reported.29 Indeed, the side chains of Arg2, Asp4, and DGlu5 of the RGD cycle and that of Arg11, Pro13, hCit15, and Pro17 of the C-terminal echistatin moiety seem to directly contact both integrins, though with different modality. The most significant difference was observed for the H_N Lys1, which was not influenced by α_vβ₃ binding, while it appears here affected by α₅β₁ binding. To obtain information on the α₅β₁-bound conformation of RGDechi, trNOESY spectrum of the peptide in the presence of K562 cell membranes was acquired and compared to that obtained in presence of HeLa membranes.29 Interestingly, several new peaks appear and the intensities of many others increase in the presence of K562 cell membranes (Figure 4A,B). trNOEs were identified and mapped on the primary sequence of the peptide (Figure 4C). In particular, Lys1, Arg2 and Asp4 of the cyclic portion, together with the Arg11, Asn12, His14, and hCit15 of the C-terminal segment exhibited the highest number of trNOEs. Moreover, similarly to WM266 membranes, strong crosspeaks connecting Asp8, Asn12, and Gly16 Hα to the δ protons of Pro9, Pro13, and Pro17, respectively, were observed, indicating a trans-arrangement for all three bonds in the presence of K562 membranes.

3.4 Sampling RGDechi conformational flexibility by NMR and MD simulations

To elucidate the details of the α₅β₁ integrin recognition mechanism by the RGDechi peptide, we performed a series of Molecular Dynamics simulations (MD) studies. To achieve a rigorous structural investigation of this interaction we applied a combined approach29, 49, 50 that includes in the Molecular Docking protocol data obtained by MD and Nuclear Magnetic Resonance (NMR) techniques. This strategy allows to take into account in the description of the binding mechanism not only the structural properties, but also the conformational heterogeneity of the RGDechi peptide. In fact, on one side NMR and MD are two complementary tools providing a detailed description on the atomic level of the structural features, of the dynamic events that occur in proteins/peptides on the picosecond and millisecond timescale51-54 and of the protein folding mechanisms.55 On the other side, the Molecular Docking represents a powerful computational tool in structural biology and computer-aided drug design to predict at the atomic level the preferred orientation of a molecule to a second one in protein–protein and protein–ligand complexes.56 At first, we performed a series of MD simulations of varied length, as reported in the experimental section, using as structural restraints the intra-molecular NOEs observed in the trNOESY NMR experiment for the RGDechi peptide in the presence of the K562 cell membranes. Successively, in order to describe the conformational peculiarities (i.e., backbone flexibility) of the RGDechi, we analyzed the trajectories of the time-averaged restrained 10 ns MD ensemble. The root mean square fluctuations (RMSF) profile of the 10 ns MD simulation (Supporting Information Figure S7A), in agreement with the absence of long range intra-molecule NOEs (Figure 4A,B), indicates that the RGDechi does not adopt in solution any well-defined secondary structure with high backbone conformational flexibility on the picoseconds time-scale (Supporting Information Figure S7A,B). Conformational families are believed to be a more realistic representation of the structure when dynamics or motions are extensive.57-60 For this reason, to detect the structural diversity induced by the conformational heterogeneity of the RGDechi peptide, we extracted the conformationally related subfamilies of structures using the program NMRCLUST.57 Therefore, we clustered the 10ns MD ensemble on the backbone heavy atoms of residues 1–19 (Supporting Information Figure S7B) obtaining 17 subfamilies of structures (Supporting Information Figures S7B, S8A). Among these, the first four clusters contain more than 70 of the 100 structures, whereas each of the other classes is poorly populated. Then, we restricted the conformational analysis on the more populated clusters (1–4) (Supporting Information Figure S8B) by evaluating the structural differences between the representative structures of each cluster. Interestingly, despite to the high conformational flexibility, the representative structure of cluster 1 shows a similar contacts map to the other representative clusters (Supporting Information Figure S8B). In particular, as reflected by the root mean square distributions (RMSD_bb1-19= 0.767 Å) (Supporting Information Figure S7B), the representative structures of the two most populated MD clusters (cluster 1 and cluster 2) are characterized by similar conformational proprieties. Altogether the data indicate that the representative structure of the MD cluster 1 represents a reasonable description of the RGDechi conformational heterogeneity observed within and between the four investigated MD clusters. Therefore, in order to explore the structural details of the RGDechi-α₅β₁ interaction, considering the contribution of the molecular flexibility in the recognition mechanism, we used as reference structure the representative conformer of the most populated MD cluster (cluster 1) to perform a molecular docking investigation.

3.5 Molecular docking studies of the α₅β₁/RGDechi complex

Molecular docking calculations were performed, as reported in the experimental section, by using as reference structure for the RGDechi peptide the representative conformation of the most populated cluster (cluster 1) of the 10ns MD ensemble. We used the coordinates of the α₅β₁ integrin in complex with the RGD motif (PDB ID: 3VI4) for the receptor. Following the optimized docking protocol,29 100 solutions were generated that, based on the binding energy values, were successively sorted into clusters (Supporting Information Figure S9). This procedure selected four most populated clusters, characterized by different binding energy and showing substantial structural differences in terms of RGD motif position and orientation with respect to the α₅β₁ active site (Figure 5A–D). On one hand, in the first two clusters (cluster 3 and 7) composed of fourteen and nine conformers, respectively, the recognition mechanism of the α₅β₁ integrin by the RGDechi peptide is principally modulated by the RGD motif, though not completely embedded inside the RGD binding cavity, with the C-terminal echistatin residues that participate in the complex formation (Figure 5A,B). On the other hand, in the clusters 29 (5 conformers) and 17 (8 conformers) the RGD region is inserted into the binding pocket delimited by the two integrin subunits with the C-terminal part that strongly contributes in the stabilization of the interaction (Figure 5C,D). In both cases the RGD motif, despite being located in the conventional binding site, presents significant structural differences in terms of position and orientation with respect to that observed in the α_vβ₃/RGDechi complex suggesting that the recognition of the α₅β₁ integrin may occur with a different mechanism. Then, in order to correctly identify the model able to describe the structural details of the α₅β₁ recognition mechanism by RGDechi peptide, the obtained docking models were investigated by calculating, as reported in the experimental section, relative STD factors for each RGDechi residue of the analyzed representative structure as a sum of peptide-receptor intermolecular proton distances (relative NOE) and comparing the results with the experimental STD effects. (Figure 6A–D). In the reference model of cluster 3 (Figure 6A), although the RGD motif is particularly exposed to the solvent, the amide backbone of the Lys1 (strong effect) is directly involved in the interaction whereas in disagreement with the experimental STD data the relative intermolecular NOE profile indicates that side chains of the Arg2, Asp4, DGlu5 as well as the residues (Pro13, hCit15) located in the C-terminal region appear to be less involved in the binding. Instead, for the model of the cluster 7 (Figure 6B), the back-calculated relative NOE profile, results to be more in line with the experimental STD data with respect to the representative model of the cluster 3 but it still presents significant differences. In particular, the Lys1 and the residues of the RGD motif as well as the C-terminal echistatin moiety appear to have a marginal role in the α₅β₁/RGDechi complex stabilization. On the other hand, the representative models of the cluster 29 (Figure 6C) and cluster 17 (Figure 6D), as mentioned above, suggest that the recognition process of the α₅β₁ is principally mediated by the RGD motif that is situated inside the conventional binding pocket. In the case of the cluster 29, the calculated relative NOEs are in good agreement with the observed STD data. In particular, the data indicate that the interaction is mainly modulated by the residues surrounding the RGD region (Lys1, Arg2, Gly3, Asp4) and it is further stabilized by the residues of the C-terminal part (Arg11, hCit15, Pro17). Notably, in the case of the glycine the calculation of the relative NOE profile is strongly influenced by the nature of the side-chain that has a single hydrogen atom . On the contrary, the back-calculated relative NOE profile for the model of the cluster 17, contradicting the experimental STD data, indicate a lower contribution of the RGD region than the C-terminal part in the recognition process of the integrin. Overall, the data analysis indicates that the representative model of cluster 29 agrees more with the experimental data with respect to the other representative conformers and it has been therefore chosen as the reference structure of the α₅β₁/RGDechi complex (see the Supporting Information for the coordinate file a5b1RGDechicomplex.pdb).

3.6 The α₅β₁ integrin recognition mechanism of RGDechi

The structural model of the α₅β₁/RGDechi complex, as obtained by using experimental and computational data, indicates that the recognition mechanism of the α₅β₁ integrin by the RGDechi peptide is mainly modulated by the Lys1, by the residues of the RGD cycle (Arg2, Gly3, Asp4, D-Glu5) interacting with the α₅ subunit and it is stabilized by the C-terminal region of the peptide (Figure 7A–D) (Table 1). Notably, all residues of the RGDechi showing STD and trNOEs effects are involved in the α₅β₁ integrin binding. In particular, the residue Lys1 mainly interacts with the β₁ subunit (Gln191, Tyr133) (Figure 7A); whereas the Arg2 (Figure 7B), makes contact with the β₁-Pro186 and Cys187. The other residues within the RGD region (Gly3, Asp4, and D-Glu5) contribute to stabilize the interaction (Figure 7B). As reported in Figure 7B, the side chain carbonyl oxygen of the Asp7 (Figure 7C) (Table 1) interacts with the residues β₁-Lys182 (Figure 7C). The interaction between RGDechi and α₅β₁ integrin, as observed in the α_vβ₃/RGDechi complex, is further stabilized by the residues of the C-terminal echistatin moiety (Arg11-Thr19) that contact both the α₅ and β₁ subunits. In particular, the Arg11 and the side chain carbonyl oxygen of the α₅-Glu126 are hydrogen bonded (Figure 7C) whereas the hCit15, located in the proximity of the Asp4 of the RGD cycle is found to establish interactions only with the α₅ subunit (Figure 7D). Instead, the Pro13 (Figure 7C) is buried into a hydrophobic pocket situated on the surface of the α₅ subunit and formed by the residues Ala159, Trp157. Additionally, the His14 interacts with the residue Trp157 of the α₅ subunit. (Figure 7D). The proline 17 is deeply embedded into a hydrophobic cleft principally formed by the residues β₁–Pro186 β₁-Leu225 (Figure 7D). Finally, the residue Thr19 slightly contributes to stabilize this interaction making contacts with the α₅-Ser129 and α₅-Asp130 (Figure 7D).

3.7 Molecular basis driving the α_vβ₃ and α₅β₁ integrins recognition by RGDechi

The availability of selective subtype integrin ligands is crucial for developing of effective targeted systems in drug delivery, diagnosis and tumor imaging applications.

We have thus compared the here obtained structural model of RGDechi in complex with α₅β₁ with that previously determined for α_vβ₃/RGDechi complex29 to derive the molecular determinants of the two interactions. As shown in Figure 8A,B, the RGDechi orientation and position in the α₅β₁ binding site is rather different with respect to that observed in α_vβ_3. In particular, the RGD region of the RGDechi peptide, despite being located in the conventional RGD binding site, presents a substantial displacement of ∼10 Å in the direction of the β₁-subunit and a rotation of ∼90° around the X-axis with respect to that observed in the case of α_vβ₃ (Figure 8A,B). This profound structural difference clearly depends on the amino acid composition of the two integrins subunits. In fact, the two negative charged Asp residues (α_v-Asp148, α_v-Asp150) (Figure 8C) (Table 1) that have a crucial role in the stabilization of the α_vβ₃/RGDechi complex by interacting with the residues of the RGD motif are replaced by two hydrophobic residues (α₅-Trp157 and α₅-Ala159) (Figure 8C) in the α₅-subunit. Because of these significant structural differences within the recognition site, when RGDechi binds to α₅β₁ integrin Arg2 interacts with the residues of the β₁-subunit (Table 1) driving the RGD motif in a different position. Moreover, α₅β₁ recognition mechanism of RGDechi is also influenced by the structural differences of the β₁-SDL with respect to β₃-SDL (Figure 8D). Indeed, as previously described, the interaction between α_vβ₃ and RGDechi is primarily modulated by hCit15, which is projected toward the β₃-Arg216 and the β₃-Asp179 (Table 1) (located within the SDL loop) and by Asp4 of the RGD portion that recognizes the β₃-Arg214, Arg216 (Table 1) and SDL Tyr166 (Figure 8D). In the case of α₅β₁ integrin, β₃-Tyr166, Asp179, Arg214, and Arg216, are replaced by β₁-Lys182, Pro186, Gln191, and Leu225, thus driving the C-terminal echistatin moiety to bind α₅β₁ in a different way to that observed for the α_vβ₃ integrin. Particularly, h-Cit15, which solely contacts with the residues α₅-Asp227 (Table 1), appears to play a reduced role in stabilizing the binding, whereas His14, Pro17, and Thr19 are involved in an extended number of interactions with the both subunits (Table 1). Some of them are located in the β1-SDL loop (Lys182 and Pro186) and others at the α₅ β₁ interface, giving rise to a recognition mechanism which, though preserving a degree of similarity with the one used to bind α_vβ_3, nonetheless contains a significant level of peculiarity.