Biochimica et Biophysica Acta 1844 (2014) 561–566

Contents lists available at ScienceDirect

Biochimica et Biophysica Acta journal homepage: www.elsevier.com/locate/bbapap

Searching for protein binding sites from Molecular Dynamics simulations and paramagnetic fragment-based NMR studies Andrea Bernini a, Lucia Henrici De Angelis a, Edoardo Morandi a, Ottavia Spiga a,b, Annalisa Santucci a, Michael Assfalg c, Henriette Molinari c, Serena Pillozzi d, Annarosa Arcangeli d, Neri Niccolai a,b,⁎ a

Department of Biotechnology, Chemistry and Pharmacy, University of Siena, 53100 Siena, Italy SienaBioGrafiX Srl, 53100 Siena, Italy Department of Biotechnology, University of Verona, 37134 Verona, Italy d Dipartimento di Medicina Sperimentale e Clinica, University of Florence, 50134 Florence, Italy b c

a r t i c l e

i n f o

Article history: Received 30 October 2013 Received in revised form 16 December 2013 Accepted 18 December 2013 Available online 27 December 2013 Keywords: Protein–protein interaction Transient surface pocket Dynamic drug design Paramagnetic probe Fragment-based hotspot screening

a b s t r a c t Hotspot delineation on protein surfaces represents a fundamental step for targeting protein–protein interfaces. Disruptors of protein–protein interactions can be designed provided that the sterical features of binding pockets, including the transient ones, can be defined. Molecular Dynamics, MD, simulations have been used as a reliable framework for identifying transient pocket openings on the protein surface. Accessible surface area and intramolecular H-bond involvement of protein backbone amides are proposed as descriptors for characterizing binding pocket occurrence and evolution along MD trajectories. TEMPOL induced paramagnetic perturbations on 1H–15N HSQC signals of protein backbone amides have been analyzed as a fragment-based search for surface hotspots, in order to validate MD predicted pockets. This procedure has been applied to CXCL12, a small chemokine responsible for tumor progression and proliferation. From combined analysis of MD data and paramagnetic profiles, two CXCL12 sites suitable for the binding of small molecules were identified. One of these sites is the already well characterized CXCL12 region involved in the binding to CXCR4 receptor. The other one is a transient pocket predicted by Molecular Dynamics simulations, which could not be observed from static analysis of CXCL12 PDB structures. The present results indicate how TEMPOL, instrumental in identifying this transient pocket, can be a powerful tool to delineate minor conformations which can be highly relevant in dynamic discovery of antitumoral drugs. © 2013 Elsevier B.V. All rights reserved.

1. Introduction Attractive strategies for antitumoral drug design are often based on the use of small molecules to interfere with protein–protein interactions (PPI) [1,2]. However, discovering such PPI disruptors is not an easy task, due to the fact that molecular interfaces of PPI are large and flat regions of the proteins, whereas small molecule–protein interactions occur only wherever surface pockets can be available. Furthermore, the transient nature of some protein binding hotspots adds complexity for predicting PPI disruptors on a static structural base [3]. Thus, it has been suggested that the time evolution of protein surface shape, monitored by Molecular Dynamics (MD) simulations can be a suitable framework for delineating the formation of transient pockets where small molecules can bind proteins [4,5]. The fact that protein backbone amides not involved in intramolecular H-bonds are often found in

⁎ Corresponding author at: Department of Biotechnology, Chemistry e Pharmacy, University of Siena, Via A. Moro 2, I-53100 Siena, Italy. Tel.: +39 0577234910; fax: +39 0577234903. E-mail address: [email protected] (N. Niccolai). 1570-9639/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.bbapap.2013.12.012

protein binding sites [6,7] suggests that monitoring both H-bonding and surface accessibility of these groups can yield important clues for identifying permanent and/or transient protein hotspots. Fragment-based approaches have been considered to discover small molecules inducing PPI interference [8]. A “rule of three” has been proposed for the latter fragments, defining an optimal molecular weight of less than 300 Da, less than three hydrogen-bond, HB, donors and acceptors, less than three rotatable bonds and water–octanol partition coefficient, log P, lower than three [9]. TEMPOL, a rather rigid and moderately hydrophilic paramagnetic probe with a MW = 172.24 Da, a logP = 0.53 [10] and two HB donor/acceptor groups, can be certainly included among the limited number of molecules satisfying the latter “rule of three”. In all the paramagnetic perturbation profiles so far reported, TEMPOL has been observed to approach preferentially protein active sites [11], confirming to be a very good NMR probe for fragment-based drug discovery. The observed remarkable agreement between protein surface hotspots defined for bovine RNAses from multiple solvent crystal structures [12,13] and TEMPOL perturbation profiles [14], underlines the reliability of these two experimental approaches for fragment-based delineation of protein surface binding sites.

562

A. Bernini et al. / Biochimica et Biophysica Acta 1844 (2014) 561–566

The fact that unpaired electron of TEMPOL, predominantly localized on the N-oxyl nitrogen atom, can approach a donor backbone amide hydrogen as close as 3.4 Å, allows detectable paramagnetic perturbations even from scarcely populated conformational states. Thus, in paramagnetic perturbation profiles powerful information about solution dynamics is encoded, suggesting that protein surface hotspots, including transient pockets, can be suitably delineated. TEMPOL perturbation profiles have been here generated for the chemokine CXCL12, a protein with a well characterized structure both in solution [15] and in crystal states [16]. The choice of the system under study is related to its pivotal importance in activation and regulation of the pattern of tumor growth and metastatic spread [17]. The chemokine CXCL12 and its receptor CXCR4 play a key role in the regulation of hematopoietic stem cells and cell migratory function during morphogenesis. The interaction of CXCR4 with CXCL12 plays a critical role in cancer metastasis by facilitating the homing of tumor cells to metastatic sites. CXCL12 has a direct growth and survivalpromoting effect for various cancer cells and enhances moreover tumor angiogenesis by recruiting endothelial progenitor cells to tumors. Drugs which modulate the CXCL12/CXCR4 axis are therefore promising candidates in anti-cancer therapies. Thus the delineation of transient CXCL12 binding pockets, by means of the combined use of TEMPOL perturbation and MD simulations, is a good example of a strategy to be employed for a challenging drug design. 2. Material and methods 2.1. Production of recombinant CXCL12 Experimental procedures for the expression and purification of the recombinant, 15N isotopically enriched CXCL12 are extensively described in Supporting Information. 2.2. NMR investigation NMR measurements have been performed at 300 K on 0.25 mM samples of 15N-CXCL12 in 20 mM MES buffer (pH 6.8) or 50 mM phosphate buffer (pH 5.0), always with 10% D2O. NMR spectra have been recorded by using a Bruker Avance 600 spectrometer equipped with cryoprobe. The 1H–15N HSQC diamagnetic and paramagnetic spectra were obtained with 256 increments and 16 scans over 2048 data points, a relaxation delay of 5 s and a sweep width of 7200 Hz. Data processing was performed with Bruker TopSpin software and the obtained NMR spectra have been analyzed with Sparky software [18]. Complete 1H and 15N resonance assignments were obtained from previously published spectra [19] and confirmed by 3D NOESY-HSQC. Most of the CXCL12 amide groups were identified as 1H–15N well resolved cross-peaks in the 1H–15N HSQC spectra obtained at pH 5.0 and 6.8. Signal volumes for each spectrum were measured both in the presence and in the absence of the paramagnetic probe. Such volumes named Vdi and Vpi , respectively, have been measured with an estimated error lower than 10% and their autoscaled values, υi, were obtained according to the relation: p;d

υi

¼

V p;d i X p;d ð1=nÞ Vi :

ð1Þ

n

Paramagnetic attenuations [20], Ai, were calculated from the autoscaled diamagnetic and paramagnetic peak volumes, according to the relation: p

Ai ¼ 2−

υi

υdi

ð2Þ

Combined 15N and 1H chemical shift perturbations (CSP) [21] were computed as: h i 2 2 0:5 CSP ¼ ð5ΔδHN Þ þ ðΔδN Þ

ð3Þ

where ΔδHN and δN are the changes in chemical shifts for 1H and 15N from backbone amides, respectively. 2.3. Molecular Dynamics simulation CXCL12 solution structure with PDB ID: 2KEE [15] has been used as a reference for atom depth calculations, see Supporting Information for details, and as a starting structure for 1 μs MD simulations performed in explicit solvent. GROMACS package [22] with the AMBER force field [23] was used for the MD run of CXCL12 solvated structure in a triclinic box of equilibrated TIP3P water molecules. The initial shortest distance between the protein and the box boundaries was set to 1.0 nm and chloride ions were added to achieve global electric neutrality. Afterwards, the energy of the system was minimized with 900 steps of conjugate gradients. To achieve equilibration before the MD simulation, the system was subjected to a short (20 ps) MD run where protein atoms were restrained to their position and only solvent molecules were allowed to move. The protein–water system was simulated in the NPT ensemble at constant temperature (300 K) and pressure (1 atm). A weak coupling to external heat and pressure baths was applied, with relaxation times of 0.1 ps and 0.5 ps, respectively. All covalent bonds were constrained using the LINCS algorithm and non-bonded interactions were computed using the PME method [24] with a grid spacing of 0.12 nm for electrostatic contribution and with a 0.9 nm cut-off for the van der Waals contribution. An integration time step of 2 fs was used and trajectory snapshots were saved every 1.0 ps. Cut-off values of 0.35 nm for donor–acceptor distance and of 30° for hydrogen-donor–acceptor have been considered for hydrogen bond, HB, formation. For each backbone amide, the fraction of the MD trajectory where criteria for HB occurrence are fulfilled, is defined as HB occupancy [25,26]. To account for local conformational rearrangements due to acid and neutral pH effects on H25 side chain, positively charged and neutral imidazole rings were inserted respectively in the starting 2KEE derived structures for two different 1 μs MD simulations. 2.4. Molecular docking simulation A MD snapshot, taken at picosecond 796,000, was representative of the transient pocket opening at K24–H25 position, see Video 1. The corresponding structure was used as target for docking of 3-(3-(2naphthoyl)thioureido)benzoic acid (ZINC: 310454) [27] in Autodock Vina [28]. 3. Results 1 H–15N HSQC spectra of 15N-CXCL12 were recorded at pH 5.0 and 6.8 in the absence and in the presence of 10 mM TEMPOL, see Fig. 1. The pH choice was suggested by previous investigations on CXCL12 interactions with heparin [29], CXCR4 [30] and heparan sulfate [31]. The observed paramagnetic perturbations and chemical shift changes of 1H–15N signals upon additions of TEMPOL have been measured and reported respectively in Fig. 2 and in Fig S1 of Supporting Information. From a comparison of the paramagnetic attenuations, Ai, obtained at pH 5.0 and 6.8, Ai values above the standard deviation σ = 0.4, are observed for K27, I28, Q48, V49 and C50 amide signals. High Ai values are also observed for CXCL12 backbone NH signals of F13, K24, A40 and N45 at pH 5.0 and for C11 amide signal at pH 6.8. These findings suggest different solution dynamics at the two used pHs, as confirmed also by the different TEMPOL induced CSP profiles shown in Fig. S1. The latter CSPs are always below 0.1, in spite of the high concentration

A. Bernini et al. / Biochimica et Biophysica Acta 1844 (2014) 561–566

563

Fig. 1. Zoomed regions of 1H–15N HSQC NMR spectra obtained for 0.25 mM aqueous solutions of 15N-CXCL12 (a) in phosphate buffer (50 mM) at pH 5.0 and (b) in MES buffer (20 mM) at pH 6.8. Spectra, recorded at 300 K in the absence and in the presence of 10 mM TEMPOL, are shown as overlaid contour plots respectively colored in green and red.

of the paramagnetic probe, [TEMPOL] / [CXCL12] = 40, required to obtain sizeable signal attenuations. The extent of these TEMPOL induced CSPs, compared with the ones measured for a series of CXCL12 ligands [19], is consistent with a very weak TEMPOL–protein interaction. The fact that 1H–15N signals experiencing the highest paramagnetic perturbations, are among the ones exhibiting the largest CSPs, see Fig. 3, indicates that the latter paramagnetic and diamagnetic effects are similarly driven by dynamics in solution. It is worth noting that at pH 5.0, enhanced TEMPOL accessibility is observed for a protein region, which is centered on K24 amide group and buried in all the available PDB structures (see Fig. S2 from

Supporting Information). The origin of this anomalous result involving a CXCL12 moiety far from the well established CXCR4 receptor binding site [15], requires a more detailed interpretative framework for analyzing the present NMR data. Thus, structural dynamics have been investigated by performing 1 μs MD simulations in explicit water. Positively charged and neutral imidazole rings have been inserted in the CXCL12 reference structure, accounting for the predominant H25 species respectively at pH 5.0 and 6.8 [15] and two independent MD trajectories have been obtained. From both MD simulations large root mean square fluctuations, rmsf, have been calculated for the most flexible regions, i.e. the CXCL12 amino- and carboxy-termini. This

Fig. 2. Plots of signal attenuation (Ai) showing the paramagnetic perturbation profiles obtained for CXCL12 at pH 5.0 (panel a) and at pH 6.8 (panel c). Ai values are plotted as dispersion around the average, and the two limiting values of 2 and 0 respectively correspond to fully broadened or not perturbed resonances. An asterisk marks those residues where severe overlap or line broadening in the diamagnetic spectrum prevented Ai determinations. Panel b reports the HB occupancy of backbone amides (Q, gray bars) and the root mean square fluctuations of the backbone (rmsf, solid line) calculated from a 1 μs MD simulation, using the CXCL12 PDB: 2KEE as starting structure.

564

A. Bernini et al. / Biochimica et Biophysica Acta 1844 (2014) 561–566

Fig. 3. CXCL12 backbone is colored according to amide group attenuations, from red (Ai = 2) to blue (Ai = 0), as obtained at pH 5.0 (a, c) and pH 6.8 (b, d). Gray is used for residues whose data are not available. In a) and b) residues relevant for the binding of CXCR4 are labeled. Structures in c) and d) are snapshots taken respectively at 796,000 and 98,000 ps from the 1 μs MD trajectory. Open (c) and closed (d) conformations of H25 side chain (colored in magenta) are shown. The intrinsically disordered N-terminal segment 1–5 is omitted from view for clarity.

finding can explain the absence of strong paramagnetic effects in the latter CXCL12 moieties, as high mobility of protein backbone reduces lifetimes of loose intermolecular adducts. MD data have been used to monitor the stability of intramolecular hydrogen bond, HB, networks [25,26] and the complete profiles of HB occurrences are shown in Fig. 2. As previously described [32–34], a good correlation between surface accessibility of protein HB donors and TEMPOL induced paramagnetic effects is observed for CXCL12. Backbone amides exhibiting Ai values higher than 1.6, indeed, have accessible surface areas, ASA, larger than 0 and, at the same time, low HB occupancies. This is the case of amide groups from residues F13, I28, N45, Q48 and C50 at pH 5.0 and from residues C11, I28, Q48 and C50 at pH 6.8. All these backbone amides contribute to the CXCL12 binding to the CXCR4 receptor, with the exception of I28 which is involved in the chemokine dimer formation [35]. In this respect, the similarity of TEMPOL perturbations on I28 backbone amide observed at pH 5.0 and 6.8 proves that the protein is monomeric in our experimental conditions. This is in agreement with a previous study [19], which showed the monomeric species to be prevalent in nonionic buffers like MES at pH 6.8, but also in phosphate buffer at lower pH and low protein concentration as in our sample condition, as inferred by the upfield/downfield 15N chemical shift of K24/K27 signal pair (Fig. 1a). 4. Discussion CXCL12 surface region involved in the interaction with CXCR4 has been structurally characterized in detail. Backbone amide groups of C11, F13-V18, L29, T31, V39-A40, and Q47-I51 have been found primarily involved in the protein-receptor complex formation [35]. Structural features of CXCL12 dimer, the actual chemokine bioactive form, have been also delineated showing that residues L26–L29 (β-strand 1) from both subunits form an antiparallel β-sheet, stabilized by intermolecular HB of L26 and I28 backbone amides. Paramagnetic attenuation profile of the dimer interface is limited by the lack of well resolved signals which can be diagnostic of dimer formation. In any case, the large I28 Ai values, similarly observed at pH 5.0 and 6.8, prove that the protein is always predominantly monomeric in our experimental conditions. The strong paramagnetic perturbations, unambiguously delineated at pH 5.0 for K24, K27 and A40 backbone amides, reveal an additional surface hotspot, even though the available PDB structures suggest that these amide groups are all deeply inserted inside the protein and involved in intramolecular HB, see Fig. S2 of Supporting Information. Thus, the presence of transient pockets accounting for the observed unexpected PREs has been considered. The occurrence of such transient pockets on the chemokine surface has been explored along MD trajectories by analyzing the surface exposure evolution of HB-free backbone amides. MD predicted transient pocket disclosure, see Fig. 3c/d, can explain TEMPOL access

to K24, K27 and A40 backbone amides. A relevant role in the formation of this transient pocket can be attributed to the imidazole ring of H25, whose pKa = 5.7 has been determined [15]. Indeed, at the operative pH of 5.0, more than 80% of the H25 imidazole ring is protonated, inducing electrostatic repulsion with the positively charged side chains of the nearby R41 and K43 residues. The latter situation favors partial opening of a surface pocket, see Figs. 3c and 4, and, consequently, the accessibility of TEMPOL to K24, K27 and A40 backbone amides. At pH 6.8 more than 90% of H25 side chains are in neutral form and, therefore, the small fraction of open pockets can be efficiently selected only by those molecules exhibiting significant affinity for that CXCL12 surface region. This can be the case of 3-(3-(2-naphthoyl)thioureido) benzoic acid (ZINC: 310454) whose CXCL12 affinity has been measured, Kd is 64 ± 15 μM [21], and 1H/15N CSP values larger than 0.2 ppm have been observed for K24 and K27 NHs. The latter selection seems not to be performed by TEMPOL, due to the weak interaction with the chemokine, as already discussed. Overall, paramagnetic attenuations and MD results highlight the H25-driven switch off of the surface hotspot accessibility. Combined inspection of ASA's and HB occupancies of K24 and K27 backbone amides along the two 1 μs MD trajectories, predicts that this transient surface pocket opens four and five times with lifetimes of the open conformations ranging from 4 to 140 ns (see Fig. 5). The fact that TEMPOL perturbs also the backbone amide signal of A40, which does not experience any chemical shift change in the presence of 310454, can be ascribed to the smaller size of the paramagnetic probe, which more easily can approach the bottom of this part of the pocket

Fig. 4. Close-up of the transient pocket opened at picosecond 796,000. Surface is colored as atom Depth index (blue = deeply inserted, red = highly exposed); residues involved in the pocket formation are labeled.

A. Bernini et al. / Biochimica et Biophysica Acta 1844 (2014) 561–566

565

Fig. 5. Analysis of MD trajectories in terms of ASA and HB occupancy for selected backbone amides. Results for the simulation of the neutral (pH 6.8) and the protonated (pH 5.0) form of H25 are reported in panels a) and b), respectively. The amide from K24 shows a well defined pattern of gained accessibility with loss of H-bonding (highlighted by boxes), connected with the opening of the transient pocket as inferred by MD snapshot analysis (see Fig. 3c). This allows for four and five openings to be detected along the 1 μs simulation in a) and b) respectively. Opening time ranges from 21 to 98 ns in a) and from 2 to 141 ns in b). A similar pattern can be recognized for H25 in a), while in b) the same amide shows a broader accessibility and a looser HB, to be ascribed to its sidechain being pulled apart (see Fig. 3c) by electrostatic repulsion between the charged imidazole ring and the nearby R41 and K43 residues. By comparing the accessibility gain of b) in respect to a), it is apparent how the behavior of protonated H25 favors the opening of the transient pocket.

where A40 amide group is located. The opening of this transient pocket in the CXCL12 surface is best detected from paramagnetic effects at lower pH where a major population of protonated H25 side chains triggers pocket formation. The occurrence of this pocket, predicted by MD simulations, but invisible in all the different PDB structures, is experimentally verified. As shown in Fig. 6, docking simulations of the 310454–CXCL12 interaction involving the latter transient pocket yield adduct structures which are consistent with reported chemical shift perturbations [21].

promising for delineating transient binding sites of PPI disruptors, preventing, in the present case, CXCL12 dimer formation. Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.bbapap.2013.12.012. Acknowledgements This work has received financial support from the Istituto Toscano Tumori. Graphical abstract credit: SienaBioGrafiX Srl. MA thanks FIRB, Futuro in Ricerca, grant no. RBFR08R7OU. AB thanks Miss Caterina Bernini for assistance in figure development.

5. Conclusions References Investigating the time evolution of surface exposure and availability for HB formation of protein backbone amides along MD trajectory is here proposed as a general computational procedure for predicting transient pockets of the protein surface where small molecules can bind. Moreover, TEMPOL, a small molecule satisfying the “rule of three” for fragment-based drug discovery, represents a universal molecular probe for targeting PPI site which can validate MD predicted binding sites. Indeed, transient HB donors from protein backbone amides can be accepted by the N-oxyl group of the paramagnet, yielding very strong dipolar interactions, even in the presence of intermolecular interactions scarcely represented in solution. A CXCL12 surface transient pocket site, indeed, has been delineated by using the present MD-assisted paramagnetic fragment-based NMR procedure. This dynamic drug discovery procedure seems to be very

Fig. 6. 310454 (sticks) docked to a transient pocket of CXCL12 (surface). The naphthyl moiety of 310454 is inserted in the A40 region of the pocket, forming a cation-π bond with K27, while the carboxyl group is surrounded by the positive charges of R41and K43. The protein surface is colored in orange for those residues exhibiting large chemical shift perturbations upon binding of the ligand, data from reference [21], showing a good agreement with expected ring current shifts.

[1] R.K. Davis, S. Chellappan, Disrupting the Rb–Raf-1 interaction: a potential therapeutic target for cancer, Drug News Perspect. 21 (2008) 331–335. [2] H. Hondermarck, C. Tastet, I. El Yazidi-Belkoura, R.A. Toillon, X. Le Bourhis, Proteomics of breast cancer: the quest for markers and therapeutic targets, J. Proteome Res. 7 (2008) 1403–1411. [3] J.A. Wells, C.L. McClendon, Reaching for high-hanging fruit in drug discovery at protein–protein interfaces, Nature 450 (2007) 1001–1009. [4] S. Eyrisch, V. Helms, Transient pockets on protein surfaces involved in protein–protein interaction, J. Med. Chem. 50 (2007) 3457–3464. [5] S. Eyrisch, V. Helms, What induces pocket openings on protein surface patches involved in protein–protein interactions? J. Comput. Aided Mol. Des. 23 (2009) 73–86. [6] A. De Simone, G.G. Dodson, C.S. Verma, A. Zagari, F. Fraternali, Prion and water: tight and dynamical hydration sites have a key role in structural stability, Proc. Natl. Acad. Sci. U. S. A. 102 (2005) 7535–7540. [7] A. De Simone, R. Spadaccini, P.A. Temussi, F. Fraternali, Toward the understanding of MNEI sweetness from hydration map surfaces, Biophys. J. 90 (2006) 3052–3061. [8] A. Winter, A.P. Higueruelo, M. Marsh, A. Sigurdardottir, W.R. Pitt, T.L. Blundell, Biophysical and computational fragment-based approaches to targeting protein– protein interactions: applications in structure-guided drug discovery, Q. Rev. Biophys. (2012) 1–44. [9] M. Congreve, R. Carr, C. Murray, H. Jhoti, A ‘rule of three’ for fragment-based lead discovery? Drug Discov. Today 8 (2003) 876–877. [10] S. Thaler, M. Fiedorowicz, P. Grieb, Z. Wypych, N. Knap, T. Borowik, K. Zawada, J. Kaminski, M. Wozniak, R. Rejdak, E. Zrenner, F. Schuettauf, Neuroprotective effects of tempol acyl esters against retinal ganglion cell death in a rat partial optic nerve crush model, Acta Ophthalmol. 89 (2011) e555–e560. [11] A. Bernini, V. Venditti, O. Spiga, N. Niccolai, Probing protein surface accessibility with solvent and paramagnetic molecules, Prog. Nucl. Magn. Reson. Spectrosc. 54 (2009) 278–289. [12] M. Dechene, G. Wink, M. Smith, P. Swartz, C. Mattos, Multiple solvent crystal structures of ribonuclease A: an assessment of the method, Proteins 76 (2009) 861–881. [13] C. Mattos, C.R. Bellamacina, E. Peisach, A. Pereira, D. Vitkup, G.A. Petsko, D. Ringe, Multiple solvent crystal structures: probing binding sites, plasticity and hydration, J. Mol. Biol. 357 (2006) 1471–1482. [14] R. Spadaccini, C. Ercole, M.A. Gentile, D. Sanfelice, R. Boelens, R. Wechselberger, G. Batta, A. Bernini, N. Niccolai, D. Picone, NMR studies on structure and dynamics of the monomeric derivative of BS-RNase: new insights for 3D domain swapping, PLoS One 7 (2012) e29076. [15] C.T. Veldkamp, J.J. Ziarek, J. Su, H. Basnet, R. Lennertz, J.J. Weiner, F.C. Peterson, J.E. Baker, B.F. Volkman, Monomeric structure of the cardioprotective chemokine SDF-1/CXCL12, Protein Sci. 18 (2009) 1359–1369. [16] E.K. Ryu, T.G. Kim, T.H. Kwon, I.D. Jung, D. Ryu, Y.M. Park, J. Kim, K.H. Ahn, C. Ban, Crystal structure of recombinant human stromal cell-derived factor-1alpha, Proteins 67 (2007) 1193–1197. [17] B.A. Teicher, S.P. Fricker, CXCL12 (SDF-1)/CXCR4 pathway in cancer, Clin. Cancer Res. 16 (2010) 2927–2931. [18] T.D. Goddard, D.G. Kneller, SPARKY 3, University of California, San Francisco. 2008. [19] C.T. Veldkamp, F.C. Peterson, A.J. Pelzek, B.F. Volkman, The monomer-dimer equilibrium of stromal cell-derived factor-1 (CXCL 12) is altered by pH, phosphate, sulfate, and heparin, Protein Sci. 14 (2005) 1071–1081.

566

A. Bernini et al. / Biochimica et Biophysica Acta 1844 (2014) 561–566

[20] N. Niccolai, O. Spiga, A. Bernini, M. Scarselli, A. Ciutti, I. Fiaschi, S. Chiellini, H. Molinari, P.A. Temussi, NMR studies of protein hydration and TEMPOL accessibility, J. Mol. Biol. 332 (2003) 437–447. [21] C.T. Veldkamp, J.J. Ziarek, F.C. Peterson, Y. Chen, B.F. Volkman, Targeting SDF-1/CXCL12 with a ligand that prevents activation of CXCR4 through structure-based drug design, J. Am. Chem. Soc. 132 (2010) 7242–7243. [22] H.J.C. Berendsen, D.V.D. Spoel, R.V. Drunen, GROMACS: a message-passing parallel molecular dynamics implementation, Comput. Phys. Commun. 91 (2012) 43–56. [23] E.J. Sorin, V.S. Pande, Exploring the helix–coil transition via all-atom equilibrium ensemble simulations, Biophys. J. 88 (2005) 2472–2493. [24] T. Darden, L. Perera, L. Li, L. Pedersen, New tricks for modelers from the crystallography toolkit: the particle mesh Ewald algorithm and its use in nucleic acid simulations, Structure 7 (1999) R55–R60. [25] R.B. Sessions, N. Gibbs, C.E. Dempsey, Hydrogen bonding in helical polypeptides from molecular dynamics simulations and amide hydrogen exchange analysis: alamethicin and melittin in methanol, Biophys. J. 74 (1998) 138–152. [26] S.G. Dastidar, C. Mukhopadhyay, Structure, dynamics, and energetics of water at the surface of a small globular protein: a molecular dynamics simulation, Phys. Rev. E Stat. Nonlinear Soft Matter Phys. 68 (2003) 021921. [27] J.J. Irwin, B.K. Shoichet, ZINC—a free database of commercially available compounds for virtual screening, J. Chem. Inf. Model. 45 (2005) 177–182. [28] O. Trott, A.J. Olson, Software news and update AutoDock Vina: improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading, J. Comput. Chem. 31 (2010) 455–461.

[29] J.W. Murphy, Y. Cho, A. Sachpatzidis, C. Fan, M.E. Hodsdon, E. Lolis, Structural and functional basis of CXCL12 (stromal cell-derived factor-1 alpha) binding to heparin, J. Biol. Chem. 282 (2007) 10018–10027. [30] Y. Kofuku, C. Yoshiura, T. Ueda, H. Terasawa, T. Hirai, S. Tominaga, M. Hirose, Y. Maeda, H. Takahashi, Y. Terashima, K. Matsushima, I. Shimada, Structural basis of the interaction between chemokine stromal cell-derived factor-1/CXCL12 and its G-protein-coupled receptor CXCR4, J. Biol. Chem. 284 (2009) 35240–35250. [31] C. Laguri, R. Sadir, P. Rueda, F. Baleux, P. Gans, F. Arenzana-Seisdedos, H. Lortat-Jacob, The novel CXCL12gamma isoform encodes an unstructured cationic domain which regulates bioactivity and interaction with both glycosaminoglycans and CXCR4, PLoS One 2 (2007) e1110. [32] A. Bernini, O. Spiga, V. Venditti, F. Prischi, M. Botta, G. Croce, A.P.L. Tong, W.T. Wong, N. Niccolai, The use of a ditopic Gd(III) paramagnetic probe for investigating alpha-bungarotoxin surface accessibility, J. Inorg. Biochem. 112 (2012) 25–31. [33] A. Bernini, O. Spiga, R. Consonni, I. Arosio, P. Fusi, S. Cirri, A. Guagliardi, N. Niccolai, Hydration studies on the archaeal protein Sso7d using NMR measurements and MD simulations, BMC Struct. Biol. 11 (2011). [34] A. Bernini, O. Spiga, V. Venditti, F. Prischi, L. Bracci, A.P.L. Tong, W.T. Wong, N. Niccolai, NMR studies of lysozyme surface accessibility by using different paramagnetic relaxation probes, J. Am. Chem. Soc. 128 (2006) 9290–9291. [35] C.T. Veldkamp, C. Seibert, F.C. Peterson, N.B. De la Cruz, J.C. Haugner III, H. Basnet, T.P. Sakmar, B.F. Volkman, Structural basis of CXCR4 sulfotyrosine recognition by the chemokine SDF-1/CXCL12, Sci. Signal. 1 (2008) ra4.