Heparin and heparan sulfate proteoglycans promote HIV-1 p17 matrix protein oligomerization: computational, biochemical and biological implications

p17 matrix protein released by HIV+ cells interacts with leukocytes heparan sulfate proteoglycans (HSPGs), CXCR1 and CXCR2 exerting different cytokine-like activities that contribute to AIDS pathogenesis. Since the bioactive form of several cytokines is represented by dimers/oligomers and oligomerization is promoted by binding to heparin or HSPGs, here we evaluated if heparin/HSPGs also promote p17 oligomerization. Heparin favours p17 dimer, trimer and tetramer assembly, in a time- and biphasic dose-dependent way. Heparin-induced p17 oligomerization is of electrostatic nature, being it prevented by NaCl, by removing negative sulfated groups of heparin and by neutralizing positive lysine residues in the p17 N-terminus. A new computational protocol has been implemented to study heparin chains up to 24-mer accommodating a p17 dimer. Molecular dynamics show that, in the presence of heparin, two p17 molecules undergo conformational modifications creating a continuous “electropositive channel” in which heparin sulfated groups interact with p17 basic amino acids, promoting its dimerization. At the cell surface, HSPGs induce p17 oligomerization, as demonstrated by using B-lymphoblastoid Namalwa cells overexpressing the HSPG Syndecan-1. Also, HSPGs on the surface of BJAB and Raji human B-lymphoblastoid cells are required to p17 to induce ERK1/2 activation, suggesting that HS-induced oligomerization plays a role in p17-induced lymphoid dysregulation during AIDS.

same region of other p17 molecules 19 , and amino acids E 42 -N 47 , Q 59 , Q 63 , that interact with the Q 69 -E 74 region of another p17 molecule 20 . On the other hand, due to the presence of "coiled coil" sequences, p17 tends to misfold 21 , behaving as an "amyloidogenic" protein that forms toxic assemblies in the brain that have been demonstrated to contribute to AIDS-associated neurodegeneration 4 .
Taken together, the capacity of p17 to oligomerize and to bind to heparin/HSPGs, along with the involvement of the GAG in the process of oligomerization of many cytokines prompted us to study the effect of heparin/ HSPGs on p17 oligomerization and biological consequences by adopting a multidisciplinary approach including bioinformatics, biochemical and cell-based models.
For what concerns bioinformatics, the docking of protein to heparin (used as a structural analog of HSPGs) and molecular dynamics (MDs) have been limited only to short (di-, tetra-or hexa-) oligosaccharides, mainly due to heparin conformational flexibility and high charge density, the weak surface complementarity of heparin/ protein interactions, the absence of well-defined binding pockets, the difficulty to define the impact of solvation/ desolvation, the large electrostatic interactions involved and the large number of torsional angles between glycosidic bonds 22,23 . Only in selected cases, 14-mer heparin oligosaccharides were used with molecular docking to study FGF 24 , VEGF 25 and CXCL8 26 dimerization. Relevant to this point, longer heparin or heparan sulfate (HS) chains are found in nature that are responsible for cytokine oligomerization. With these premises, to corroborate our experimental data, we have here modeled heparin chains up to 24-mer and performed docking and MDs with p17 dimer.

Results
Characterization of the spontaneous p17 oligomerization. p17 has been demonstrated to spontaneously self-assemble, forming trimers in solution 17 . We thus exploited SPR to calculate the kinetic parameters of spontaneous p17/p17 interaction. As shown in Fig. 1a, free p17 interacts with a p17-containing sensorchip but not with a void sensorchip (negative control). When increasing concentrations of free p17 are injected onto surface-immobilized p17, a sensorgrams overlay is obtained (Fig. 1b). Equilibrium binding data were then used to generate a saturation curve and Scatchard plot regression that allow the determination of a Kd value equal to 5.29 ± 1.5 × 10 −7 M and 6.65 ± 1.8 × 10 −7 M when using the softwares Prism GraphPad (Fig. 1c) and Origin Microcal (data not shown), respectively. SPR analysis does not allow to discriminate among dimer, trimer or higher orders p17 oligomers. We thus used WB analysis of p17 after chemical cross-link, an approach that stabilizes and makes evident the different p17 complexes: p17 self-assembly gives origin to three order complexes: a dimer (the most abundant, corresponding to 29% ± 8,2 of total protein in the sample), a trimer (12% ± 3,0) and a tetramer (that corresponds only to 3% ± 0.03), in respect to a 59% ± 7.4 of the protein that remains in its monomeric form. Urea and dithiothreitol (DTT) prevent p17 oligomerization, indicating that a proper tridimensional conformation of the protein is required for self-assembly (Fig. 1d).
Effect of heparin on p17 oligomerization. As already mentioned, by binding to proteins, heparin/ HSPGs favour their oligomerization. Since heparin binds p17, we evaluated if its affects p17 oligomerization. Heparin modulates p17 oligomerization in a dose-dependent, biphasic way: at concentrations between 0.0001 and 0.001 μg/ml, it increases p17 oligomerization while, at higher concentrations (0.01-1,000 μg/ml) it causes an inhibitory effect (Fig. 2). When the formation of specific oligomers was considered, the stronger promoting effect is exerted by heparin on trimer and tetramer formation (4.2 and 3.4 fold increase), while dimer formation is increased only 2 times (Fig. 2c).
Heparin-dependent p17 oligomerization is time-dependent and relatively slow: to exert its full effect, heparin must be incubated with p17 for at least 60 min. (Fig. 3a). Also, heparin-dependent p17 oligomerization is ionic strength-dependent, being inhibited by NaCl (Fig. 3b).
Heparin binds to p17 via its SO 3 − , which interact with the positive lysine residues of the N-heparin-binding domain (HBD) of the protein 10 . To evaluate if a direct p17/heparin interaction is required to p17 oligomerization, we exploited selectively 6-O-, 2-O-and N-desulfated heparins, already demonstrated to be unable to bind p17 10 : at variance with unmodified heparin, the three desulfated heparins do not promote p17 oligomerization, indicating that all the SO 3 − of heparin are required to induce p17 oligomerization (Fig. 4a). To characterize the features of p17 necessary for its heparin-dependent oligomerization, we used N-ter K → A p17, a mutant obtained by the substitution of lysine of the N-HBD with alanine (Fig. S1). N-ter K → A p17 shows a conserved capacity to undergo spontaneous oligomerization, indicating that the N-HBD is not involved in this process. Actually, for N-ter K → A p17 the spontaneous, basal oligomerization in the absence of heparin is even increased, possibly due to the neutralization of positive charges that, in wt p17, can limit self-interaction by electrostatic repulsion. However, heparin fails to promote N-ter K → A p17 oligomerization (Fig. 4b), in agreement with its incapacity to bind the protein 10 .
In conclusion, a direct interaction between the SO 3 − of heparin and the positive lysine residues of p17 are required to induce protein oligomerization.
Computational studies of heparin-induced p17 oligomerization. As already mentioned, computational docking of heparin (mostly used as a structural analogue of HSGPs) to its binding-protein still represents a challenge. Accordingly, the p17/heparin complex has been resolved only with a protein monomer and a 6-mer heparin 10 , prompting us to conceive a new computational approach to study the interaction of longer heparin chains with p17 dimers (Fig. S2).
Blind docking simulations with standard 4-mer heparin probes and wt p17 dimer show that the tetrasaccharides always take position in the N-HBDs of the two monomers. On the other hand, using a N-ter K → A p17 dimer, the www.nature.com/scientificreports www.nature.com/scientificreports/ 4-mer heparin probes dock mainly in the C-ter basic domain. We thus used a NC-ter K → A p17 dimer to predict alternative heparin-binding pose, allowing the definition of two alternative shorter and longer heparin paths.
In the shorter path, the succession of 4-mer heparin probes starts from the N-HBD of p17 monomer A and goes directly to the N-HBD of p17 monomer B (Fig. 5a). Also in the longer path the succession of 4-mer heparin probes starts from the N-HBD of monomer A but it makes then contact with H4, C-ter basic domain and H3 of monomer A and with H3, C-ter basic domain, H4 and finally N-HBD of monomer B (Fig. 5b). The sliding window and incremental docking methods were used to identify and join the best binding poses of each 4-mer heparin in the traced paths (Fig. 5c), predicting two heparin chain models: the 15-mer heparin, modelled on the shorter path, and the 24-mer heparin modelled on the longer path (Fig. 6).
In the complexes formed by p17 dimer with the 15-and 24-mer heparins, H-bonds are mainly formed with SO 3 − or COO − . In the 15-mer heparin complex H-bonds occur mainly within the N-HBD domains and α-helix 2 of both monomers ( Fig. 6a and Table S1), while in the 24-mer heparin complex they occur in all the helices of both monomers and in the C-HBDs ( Fig. 6b and Table S2).
MDs were performed to evaluate the stability, conformational drift and refine the two heparin/p17 dimer complexes. Root mean square fluctuation (RMSF) analysis of the 15-mer heparin/p17 dimer complex shows that the C-termini of both monomers are the most variable regions (fluctuation up to 5 Å), in agreement with relaxation and amide proton exchange studies 27 . Three other flexible regions span from 7-14, 20-30 and 60-75 (Fig. S3a). The 15-mer heparin backbone is stable, while its SO 3 − are endowed with a higher fluctuation due to solvent exposure. In particular, monosaccharides from 10 to 14 exhibit the highest fluctuation (up to 5 Å), due to their interaction with N-HBD of p17 monomer B (Fig. S3b).
Root mean square deviation (RMSD) of the p17 C-α globular domain is stable during the simulations (average fluctuation = 0.5 Å) while the C-termini of both the monomers, due to solvent exposure, reach stability only after 25 and 45 ns. Heparin remains stable during the simulation (average fluctuation = 2 Å). It also reaches a second stability plateau after 28 ns due to the fitting upon binding of the 8 th monosaccharide (Fig. S3c).
Electrostatic surface analysis of docking model shows that 15-mer heparin sets up an articulate network of interactions inside the cationic N-HBDs of monomer A and B, between which is interposed a negative electrostatic surface potential generated by the exposure of COO − of E 40 , by the N-SO 3 − and 3-OH − of the 10 th and 8 th   Table S1). However, MDs show that, after 20 ns, these two H-bonds are lost due to solvent exposure of N-SO 3 − and 3-OH − , while a new H-bond network is formed between the COO − of E 40 and the guanidinium groups of R 39 and R 43 and between the N-SO 3 − of the 10 th Glc and the guanidinium group of R 39 ( Fig. 6a and Table S1). This new arrangement induced by the heparin conformational drift leads to the formation of a continuous positive channel in the p17 dimer structure (Fig. 6a).
H-bond analysis shows that the 15-mer heparin makes an articulate network of H-bonds with the p17 dimer (Table S1). Worth of mention, K 30 interacts with 2-O-SO 3 − and 3-OH _ of the 3 rd IdoA and with 6-O-SO 3 − of the 4 th Glc with an average persistency of 65%, while H 33 interacts with 6-O-SO 3 − of the 4 th Glc with an average persistency of 50%. Interestingly, the H-bond persistency of the last two residues together is higher than 80%, indicating a synergic effect mainly due to the groups of the the 3 rd IdoA. Furthermore, R 39 and R 43 of monomer B interact simultaneously with E 40 and heparin (persistency = 50% for both), further stabilizing the interaction network between the protein and the sugar.
RMSF analysis of the 24-mer heparin/p17 dimer complex shows that the C-terminus of p17 monomer B displays a lower fluctuation in respect to that in the 15-mer heparin/p17 dimer complex, due to a better H-bond network and a lower exposure to solvent, in particular in the second half of the heparin chain (monomers 12-24) ( Fig. S3d and Table S2). As already highlighted in the 15-mer heparin/p17 dimer complex (Fig. S3e), a higher fluctuation can be instead appreciated for monomer A binding to the first half of the heparin chain (monomers 1-12) for which a greater fitting needs to be induced to generate the positive channel in the protein. More into the details, the first four monosaccharides of heparin, that interact with the N-terminus of monomer A, are endowed with the higher fluctuation, with their SO 3 − showing an average fluctuation up to 2 Å, likely due to their solvation. RMSD analysis demonstrates that the p17 C-α globular domains in the dimers are stable during the simulation, with an average fluctuation = 1 Å. This stability is shared also by the C-terminus of monomer B, while the C-terminus of monomer A and heparin reach stability only after 35 ns (Fig. S3f).
Electrostatic analysis of docking model shows that some SO 3 − of 24-mer heparin take contact with the N-ter basic motif of both the p17 monomers and with the C-ter of monomer B (Fig. 6b and Table S2) while some other remain instead exposed to the solvent. However, MDs shows that these free SO 3 − , by inducing molecular drifts, succeed in taking contact with basic residues of the C-HBDs of both the p17 monomers. Thus, the heparin-induced conformational drift of p17 monomers creates sandwich-like structures that lead to the formation of the continuous positive channel, already described in literature 27 , that prevents solvent exposition of bound heparin ( Fig. 6b and Table S2). The results of the H-bonds analysis of the 24-mer heparin/p17 dimer complex are reported in Table S2. Considering the whole simulation, residues A 115 and Q 116 of the C-terminus of monomer A, that exhibit low H-bond persistency with heparin (55 and 29%), increase their persistency after 35 ns to 99% and 85%, due to binding-induced stabilization of both heparin and p17 C-termini.
In conclusion, the new computational protocol here adopted discloses the possibility that heparin chains can connect two p17 monomers with different binding modes depending on their lengths and that, during interaction, heparin induces dynamic conformational drift of the p17 dimer that facilitate the interaction between the GAG and the protein monomers, stabilizing the dimer and likely promoting further oligomerization.
Heparan sulfate-dependent p17 oligomerization at the lymphoblastoid cell surface. Heparin, that is produced and released by mast cells in the blood, is generally used experimentally as a structural analogue of HSPGs expressed on the surface of almost all eukaryotic cells, were they act as receptors for many proteins. Thus, once characterized the capacity of heparin to promote p17 oligomerization at a biochemical and computational level, we wondered if this capacity was shared by HSPGs associated to lymphoblastoid cells. We thus used  www.nature.com/scientificreports www.nature.com/scientificreports/ the Namalwa cell model already exploited to study the HSPGs with HIV-1 Tat/HSPGs interaction 12 : p17 effectively oligomerizes at the surface of SYN-Ncs, originating those same high orders oligomers characterized by WB in the cell-free system. p17 oligomerization at the cell surface specifically depends on Syndecan-1, since it does not occur on HSPGs-deficient EV-Ncs (Fig. 7). Interestingly, an additional 30 kDa band is visible in both SYN-and EV-Ncs, possibly representative of p17 interaction with an unknown protein present at the surface of both clones.
We finally investigated if p17 binding to HSPGs and the consequent p17 oligomerization at the lymphoblastoid cell surface were functionally relevant. Since SO 3 − of heparin are required for its p17-oligomerizing capacity, we treated Raji and BJAB cells with chlorate, that inhibits sulfation of HS and that has been already used to study HSPGs functions in lymphoid cells 12 . Chlorate-treated cells were evaluated for their sensitivity to p17 stimulation by evaluating extracellular-regulated kinase 1/2 (ERK 1/2 ) phosphorylation, a second messenger involved in p17-dependent lymphoid cell activation 6,17 : p17 induces a dose-dependent, bell-shaped activation of ERK 1/2 in Raji cells that is abolished by chlorate-treatment (Fig. 8a,b). The same results were obtained with BJAB cells (Fig. 8c). Taken together, these results indicate that, at the surface of lymphoblastoid cells, the binding of p17 to HSPGs and possibly the consequent p17 oligomerization are required to activate signal transduction.

Discussion
Protein oligomerization is involved in a variety of biological processes. It can occur in the form of "simple" dimers or higher order oligomers of cytokines, chemokines and growth factors that favour receptor clustering at the cell surface and transduction of the signal that, in turn, trigger cytoskeletal rearrangements, cell movement and other cellular responses including proliferation and survival involved in physiological and pathological processes such as immune responses, angiogenesis and cancerogenesis 14,28-31 . Differently, some "amyloidogenic" proteins are endowed with the propensity to change conformation from an α-helix to β-sheet, tending to misfold and to form aspecific large aggregates responsible of the pathogenesis of many human diseases, including neurodegenerative pathologies such as Alzheimer and Parkinson' disease 32 .
p17 undergoes trimerization during viral assembly in the proximity of the inner cell surface 33 , where it reaches the millimolar range of concentration 34 . Compatible with this high concentration, here we found that p17 spontaneously self-interacts with a relatively low affinity (Kd ≅ 600 nM). p17 oligomerization is prevented by DTT and urea, indicating that this process requires a proper 3D conformation of p17 and suggesting that, in our experimental conditions, it is correctly folded. Accordingly, circular dichroism spectroscopy revealed that our p17 is released by HIV-1-infected cells in the extracellular environment, where it acts as a cytokine, with its bioavailability and activity regulated by heparin or HSPGs 10 . At variance with the high p17 concentrations that can be found inside HIV-infected cells, released p17 can be found in body fluids at the nanomolar range of concentrations 3,4 , suggesting that, in this environment it likely does not oligomerize. Importantly, extracellular protein oligomerization of many physiological or pathological cytokines is often regulated by free heparin and cell-surface HSPGs 13,14,28 .
Here, by exploiting chemical cross-linking, already used to study FGF2 oligomerization 36 , we found that, beside the already known trimer, free p17 spontaneously forms dimers and tetramers and that p17 oligomerization is promoted by heparin in a ionic strength-dependent way, with 1.0 M NaCl necessary to achieve a complete complex dissociation. Interestingly, the spontaneous p17 trimers are instead almost completely disrupted by 0.5 M NaCl 17 , suggesting that heparin not only increases p17 oligomerization, but stabilizes it by increasing its affinity, making it possible to occur also at the cell surface and in the extracellular environment, where p17 concentrations are significantly lower 3,4 .  www.nature.com/scientificreports www.nature.com/scientificreports/ Heparin-induced p17 oligomerization occurs only in the presence of low GAG concentrations. It is tentative to hypothesize that a molar excess of p17 favors its multiple binding to adjacent sites on a single heparin chain. Differently, in the presence of higher concentrations of heparin, the large availability of binding sites on separate GAG chains in the presence of a limited amount of protein likely lowers the chances of p17 molecules to be close enough to self-interact. Interestingly, a similar behavior was already observed for heparin-induced FGF2 dimerization 37 .
As already mentioned in introduction, the C-terminal portion 19 and amino acids E 42 -N 47 , Q 59 , Q 63 , and Q 69 -E 74 20 of p17 mediate its oligomerization. Our MDs predict a heparin-induced drift of p17 monomers that leads to the resetting of globular domains and loss of interaction only between A 45 of monomer A and S 72 of monomer B that, in turn, induces the alignment of both the α-helix 2. Thus, heparin could bring together multiple copies of p17 through the relatively high affinity p17/heparin interaction (Kd ≅100 nM) 10 , leaving exposed the p17 self-assembly regions on adjacent proteins as to facilitate the low affinity p17/p17 interaction (Kd ≅ 600 nM). Also, it is possible that the formation of high order oligomers (trimer and tetramers) can be triggered by the initial assembly of the p17 dimer induced by heparin. In effect, oligomerization is often the outcome of a cooperative interaction process, during which a first binding makes easier the following ones 13 . Thus, the action of heparin may consist in providing the appropriate scaffold to facilitate and stabilize the relatively weak p17/p17 interaction, as already observed for different other factors and cytokines, including FGF2 38 , HGF 39 and platelet-derived growth factor 40 .
Beside heparin, also HSPGs of lymphoblastoid cell surface bind p17 favoring its oligomerization and are required to induce p17-dependent ERK 1/2 phosphorylation, suggesting the possibility that, in vivo, p17 binding to HSPG and the following oligomerization at the cell surface are required for a full lymphoid cell activation, concurring to pathological effects exerted by p17 during AIDS progression.
The capacity of heparin to promote the formation of p17 tetramers and even examers 18 infers the possibility that it could favor the formation of even higher order p17 aggregates. In effect, after prolonged exposures, in our WB analyses very high molecular weight bands are detectable both in solution and at the cell surface (data not shown). On the other hand, at very high concentration (4.0 μM, that is 40 times higher than that used in our experimental conditions) p17 generates amyloidogenic assemblies 4 . Thus, it is tentative to hypothesize that, at low concentrations and with the help of heparin/HSPGs, p17 forms specific oligomers that are functional to the activation of specific p17 receptors and hence to the cytokine-like activity of the protein. At variance, in peculiar sites of accumulation, p17 reaches high concentration undergoing misfolding and aspecific aggregation that contribute to AIDS-associated neurodegeneration 4 . We can not rule out however the possibility that heparin plays a role also in p17 amyloidogenic oligomerization, as already demonstrated for other GAGs in various amyloid diseases 41 . To properly investigate this possibility, suitable techniques such as analytical ultra-centrifugation approach or atomic force and transmission electron microscopy 4 should be employed.
As mentioned in the introduction, computational models for heparin/protein interaction have been so far restricted to docking studies with short GAG chains and aimed at the identification of HBDs in proteins. Here, we have set up a new "step by step" protocol ( Fig. S2) by which it has been possible to identify two heparin paths and then create two computational models of a p17 dimer in complex with 15-and 24-mer heparins whose validity is confirmed by their good agreement with the experimental data. Indeed, docking and MDs predict 15-mer and 24-mer heparin to bind to p17 dimer through an articulate network of interactions occurring mainly between the negative SO 3 − of heparin and the positive amino acids of the proteins. Accordingly, these interactions have been experimentally demonstrated to mediate heparin/p17 interaction and p17 oligomerization by experiments with NaCl, desulfated heparin and the p17 K → A mutant.
MDs studies with the heparin chains models disclosed the capacity of heparin to interact with a p17 dimer with different binding modes depending on its length. Also, heparin can induce a "fitting-upon-binding" on the p17 dimer that causes the exposition of positively charged residues on the globular domains and the warping of both the C-termini to create a sandwich-like structure around heparin. Finally, RMSF analysis showed that when complexed with the 24-mer heparin, the p17 dimer complex displays a lower fluctuation in respect to when complexed with 15-mer heparin, suggesting that the longer is the heparin chain, the stronger is its stabilizing effect on the p17 dimer. MDs also show that, during the binding process, heparin fits together with the protein, simulating an annular-like structure already described in literature 4 .
Beside p17, this computational protocol could be useful to understand at an atomic level the role of heparin in the processes of oligomerization of a variety of cytokines, chemokines and growth factors, and even in the formation of ternary complexes among heparin, heparin-binding proteins and their receptors.

Material and Methods
computational studies. Models. p17 and the heparin tetrasaccharide (4-mer heparin) were modelled as described 18 . A 4-mer heparin probe with the deletion of the H atom of the hydroxyl groups at position 4 in the first IdoA and of the hydroxyl group at position 1 of the Glc (4-mer heparin modified probe) was prepared and used in docking simulations to promote the 1 → 4 glycosidic linkage. Two p17 mutants were modelled by Alanine Scanning Method in which the lysine residues of both the N-and C-terminal basic domains were replaced with alanine (N-ter K → A p17 and NC-ter K → A p17).
Heparin path identification. Blind docking simulations were performed by ClusPro web-server 42 using 4-mer heparin and N-ter K → A or NC-ter K → A p17 dimers to identify heparin-binding regions on p17 other than those in the N-and C-ter basic domains. The identified heparin probe sites were then filtered by best score, cluster size, visual inspection and positioned in a dimer of wild type (wt) p17 to finally achieve an alignment and hence a traced heparin path. www.nature.com/scientificreports www.nature.com/scientificreports/ Incremental docking and heparin modelling. The 4-mer heparin modified probe was used in local docking simulation along the traced heparin path in the wt p17 dimer by Autodock 4.2 43 . The "sliding window method" was set up to create a sequence of overlapping sliding grids, each covering a whole 4-mer heparin and the last saccharide unit of the previous one. Local docking poses were filtered for free energy of binding, clusters size and correct orientation. The aligned 4-mer heparin modified probes were joined by 1 → 4 glycosidic linkages using Pymol 44 . Gasteiger-Hückel charges were assigned to the sugar and then minimized using steepest descendent and conjugate gradient methods by Chimera 45 , obtaining heparin chains of increasing length.
Molecular dynamic simulation (MDs). Amber14 package 46 was used for MDs of p17 dimer in complex with 15-and 24-mer heparins. MDs were carried out using ff99SB force field parameters for protein and GLYCAMo6 for heparins. Each complex was neutralized by adding Ca 2+ restrained away from the protein 47 and solvated with TIP3P water model.
Energy minimization was carried out with the non-bonded cut-off of 8 Å through the following steps: (i) p17/ heparin complexes and counter ions were restrained by a harmonic potential of 5 kcal/mol × Å 2 , while water molecules were relaxed using 2,500 cycles of steepest descent and conjugate gradient methods; (ii) counter ions and hydrogens were relaxed using 5,000 cycles of steepest descent and conjugate gradient methods and restrained by a harmonic potentials of 3 kcal/mol × Å 2 and then of 1 kcal/mol × Å 2 ; (iii) the system was relaxed using 5,000 cycles of steepest descent and conjugate gradient methods without any restraint; iv) the system was heated from 0.1 K to 100.0 K in NVT (constant volume) and from 100.0 to 300.0 K in NPT (1.0 atm constant pressure). The two generated complexes were simulated in a periodic boundary conditions using the Langevin algorithm at 300.0 K. During heating and simulations the Ca 2+ ions were restrained at 500 kcal/mol × Å 2 .
Equilibration (3 ns) and simulation were validated using the physical observables parameters of the system confirming that the complexes obeyed the NPT ensemble. Electrostatic interactions were calculated using the Particle Mesh Ewald method. A cut-off of 8.0 Å was applied to van der Waals forces. Integration time step was set to 1.0 fs during equilibration and 2.0 fs during simulation. MDs were performed over 50 ns using the pmemd CUDA program of the Amber14 package and a server Tesla K20 Graphical Processing Unit.
Surface plasmon resonance (SpR) binding assay. SPR measurements were performed on a BIAcore X 100 (GE-Healthcare, Milwaukee, WI). p17 (10 μg/ml in sodium acetate buffer 10 mM pH 4.8) was immobilized onto one of the two flow cells of a CM5 sensorchip [2,500 resonance units (RU), ≅ 0.15 fmoles/mm2] by amine-coupling method as for manufacturer's instructions. The second flow cell was left void and used for blank subtraction. For binding analysis, increasing concentrations of p17 in 10 mM HEPES pH 7.4, 150 mM NaCl, 3 mM EDTA, and 0.005% surfactant P20 (HBS-EP) were injected over the p17-or void surface for 2 min. and then washed until dissociation. After each run, the sensorchip was regenerated by injection of HBS-EP containing 2.0 M NaCl. Association and dissociation rates (kon and koff) and dissociation constant (Kd) were calculated from blank-subtracted sensorgrams overlay using the non linear fitting (single site model) software package BIAevaluation 3.2. Also, Kd was calculated by being fitted with the proper form of Scatchard's equation for the plot of the bound RU at equilibrium versus the ligand concentration in solution using two different softwares (Prism GraphPad and Origin Microcal). p17 oligomerization Western blot (WB) assay. wt or N-ter K → A p17 (2 μg/ml in 20 μl) were incubated for 2 h at 37 °C in PBS alone or with Urea, DTT, unmodified or selectively desulfated heparins or NaCl. Protein oligomers were chemically cross-linked by a 2 h incubation at 37 °C with 1 mM BS3. Then, samples were incubated for 5 min. at 90 °C in reducing sample buffer and run onto a SDS 12% PAGE. Proteins were transferred onto a PVDF membrane (Millipore, Bedford, MA) that was then incubated for 18 h at 4 °C with anti-p17 antibody [1:1000 in Tris-HCl 10 mM pH 7.5, NaCl 150 mM, 0.1% TWEEN 20 (TTBS)]. Antigen-antibody complexes were detected by a 1 h incubation at 4 °C with peroxidase-coupled goat anti-mouse IgG (1:5000 in TTBS) and the ECL System (Santa Cruz Biotechnology). The 17, 34, 51 and 68 kDa bands, corresponding to p17 monomer and higher-order oligomers, were quantified by computerized image analysis using Image LabTM software (Bio-Rad). cell cultures. Burkitt lymphoma-derived Namalwa cell clones genetically deficient for HSPGs expression were transfected with the empty vector or with a vector containing syndecan-1 cDNA (EV-Ncs and SYN-Ncs) and cultured as described 49 . BJAB and Raji human B-lymphoblastoid cells were from American Type Culture Collection and were cultured in RPMI-1640 medium containing 10% fetal calf serum (FCS), 100 U/mL penicillin, 50 mg/ml streptomycin, 1 mM L-glutamine.
Oligomerization of p17 at the surface of lymphoblastoid cells. Aliquots (1,5 × 10 6 cells) of EV-or SYN-Ncs were incubated for 2 h at 37 °C in culture medium with p17 (1 μg/ml) without serum. Then, chemical cross-linker BS3 (1 mM) was added and incubated for 30 min. at room temperature. At the end of incubation, cells were washed with PBS to remove unbound p17 and with 0.5 M NaCl in PBS to remove proteins weakly