Stress-inducible phosphoprotein 1 (STIP1) is a cellular co-chaperone, which regulates heat-shock protein 70 (Hsp70) and Hsp90 activity during client protein folding. Members of the S100 family of dimeric calcium-binding proteins have been found to inhibit Hsp association with STIP1 through binding of STIP1 tetratricopeptide repeat (TPR) domains, possibly regulating the chaperone cycle. Here, we investigated the molecular basis of S100A1 binding to STIP1. We show that three S100A1 dimers associate with one molecule of STIP1 in a calcium-dependent manner. Isothermal titration calorimetry revealed that individual STIP1 TPR domains, TPR1, TPR2A and TPR2B, bind a single S100A1 dimer with significantly different affinities and that the TPR2B domain possesses the highest affinity for S100A1. S100A1 bound each TPR domain through a common binding interface composed of α-helices III and IV of each S100A1 subunit, which is only accessible following a large conformational change in S100A1 upon calcium binding. The TPR2B-binding site for S100A1 was predominately mapped to the C-terminal α-helix of TPR2B, where it is inserted into the hydrophobic cleft of an S100A1 dimer, suggesting a novel binding mechanism. Our data present the structural basis behind STIP1 and S100A1 complex formation, and provide novel insights into TPR module-containing proteins and S100 family member complexes.
The S100 family of proteins comprises ∼25 members involved in several crucial biological processes including calcium (Ca2+) homeostasis, proliferation, differentiation, inflammation and apoptosis [1–3]. These proteins are tissue- and cell-specific homo/heterodimers composed of two EF-hand helix motifs capable of Ca2+ ion co-ordination . Typically, upon Ca2+ binding, α-helix III of each subunit undergoes a significant conformational change exposing a hydrophobic cleft, which represents the binding site to various ligands . Thus, the S100 family members are considered Ca2+-dependent regulators of several cellular pathways through multiple protein–protein interactions, which include the receptor for advanced glycation products, p53, annexin and CapZ [4,6–9]. Altered expression patterns of specific S100 proteins have been noted in several disease states including different forms of cancer, cardiomyopathy and neurodegenerative diseases, such as Alzheimer's disease (AD), allowing them to serve as clinical diagnostic biomarkers [10–13].
Calcium serves as a ubiquitous signaling molecule that regulates several neuronal activities, and deregulation of intracellular calcium has been suggested to be involved in AD pathogenesis . Ca2+-binding proteins play an important role in regulating Ca2+ homeostasis and in modulating protein–protein interactions . For instance, many S100 family members have been implicated in aberrant AD signaling [10,15–17]. S100A1, a protein abundantly expressed in skeletal muscle, cardiomyocytes and neurons, has been found to regulate amyloid precursor protein processing, tau phosphorylation and Aβ oligomer-induced neuronal cell death [10,12,18–20]. In addition, studies have revealed roles for S100A1 in the heat-shock protein response, regulating Hsp70 and Hsp90 binding to stress-inducible phosphoprotein 1 (STIP1) during client protein refolding [21,22].
STIP1 is a cellular co-chaperone that co-ordinates client protein transfer from Hsp70 to Hsp90 during protein folding of various cellular ligands, including numerous oncogenic kinases and transcription factors . STIP1 is important for the maintenance of proper protein folding and integrity, which suggests that STIP1 may influence protein aggregation, a pathological hallmark of numerous neurodegenerative diseases, including AD . In addition, secreted STIP1 transduces neuroprotective signals through the cellular prion protein (PrPC) protecting cells from various cellular stressors [25–27]. STIP1 association with PrPC inhibits neurotoxic signaling events induced by toxic Aβ oligomeric species characteristic of AD [28,29].
STIP1 is composed of three structurally related tetratricopeptide repeat (TPR1, TPR2A and TPR2B) domains and two aspartate and proline-rich DP (DP1 and DP2) domains. TPR domains comprise 34-amino acid degenerate consensus motifs arranged in tandem forming a series of antiparallel amphipathic α-helices . They are common protein–protein interaction modules involved in numerous cellular processes including cell cycle regulation, transcription and protein folding . TPR2A and TPR2B mediate interactions between the C-terminal EEVD motif of Hsp90 and through a series of contacts located in the middle domain of Hsp90 [23,30,32,33]. Hsp70 co-ordination involves binding to the TPR1 and TPR2B domains of STIP1 .
S100A1, S100A2 and S100A6 family members have been identified as STIP1 ligands, forming molecular complexes through interactions with the TPR domains of STIP1 [21,22]. The association of S100A2 and S100A6 with STIP1 has been shown to inhibit Hsp70 and Hsp90 binding in vitro and in cell culture upon Ca2+ stimulation, implicating their function in STIP1-directed protein folding. However, the molecular basis of S100 binding to STIP1 is largely unknown.
The present study employed nuclear magnetic resonance (NMR) spectroscopy, analytical ultracentrifugation (AUC) and isothermal titration calorimetry (ITC) to define the molecular mechanisms of S100A1 binding to STIP1. Our results reveal that each TPR domain is capable of binding to a single S100A1 dimer with varying thermodynamic parameters, with TPR2B possessing the highest affinity for S100A1. Each TPR domain associates with S100A1 through a common binding site, spanning α-helices III and IV of S100A1 in a Ca2+-dependent manner. Through NMR chemical shift mapping, we have identified the C-terminal region of TPR2B to be the S100A1-binding site.
Materials and methods
Full-length recombinant STIP1 or individual TPR domains [TPR1 (residues 1–118), TPR2A (residues 217–352) and TPR2B (residues 353–480)] were cloned into pDEST17 expression vectors (Invitrogen–ThermoFisher Scientific) containing an additional N-terminal 6xHis-tag joined by a cleavable tobacco etch virus (TEV) recognition site to facilitate tag removal. STIP1 constructs were expressed and purified as recently described . Briefly, individual constructs were transformed into Escherichia coli (E. coli) BL21 (DE3) pLysS strain and grown in standard M9 minimal medium at 37°C to an OD600 of 0.9. Expression was induced with 1 mM isopropyl β-d-1-thiogalactopyranoside (IPTG) and cultures were grown at 22°C overnight. 15N-labeled protein samples for NMR spectroscopy were expressed in standard M9 medium supplemented with 1 g/L 15N-labeled NH4Cl.
Recombinant STIP1 proteins were purified by Ni2+-affinity chromatography using Ni Sepharose 6 Fast Flow beads (GE Healthcare Life Sciences). 6xHis-tag cleavage was achieved by incubation with 6xHis-tagged TEV overnight at 22°C. TEV and 6xHis-tag were subsequently removed by additional Ni2+-affinity chromatography.
pETDuet vector encoding S100A1 (kindly provided by Dr Gary Shaw at the University of Western Ontario) was transformed into E. coli BL21(DE3)-RIL, and cultures were grown in standard M9 minimal medium or in M9 supplemented with 1 g/L 15N-labeled NH4Cl for NMR spectroscopy studies . Cultures were grown to an OD600 of 0.9, induced with 1 mM IPTG and grown overnight at 22°C. The bacterial pellet was resuspended in 50 mM Tris–HCl and 5 mM MgCl2 (pH 8.0) and lysed by French press at 10 000 psi. Cellular debris was pelleted by centrifugation at 40 000×g for 30 min. Supernatant was filtered through a 0.45 μm syringe filter (Thermo Scientific) and loaded onto a 2 ml HiTrap Q anion exchange column (GE Healthcare) equilibrated in Buffer A [25 mM Tris–HCl, 1 mM EDTA and 1 mM DTT (pH 8.0)] using an AKTA FPLC (GE Healthcare) at a flow rate of 1 ml/min. The column was washed with 10 column volumes of Buffer A and subsequently eluted by a 20-column volume linear gradient of Buffer A with 1 M NaCl. Fractions containing S100A1 were pooled and the CaCl2 concentration was adjusted to 5 mM. Fractions were incubated with phenyl-Sepharose resin (Amersham Bioscience) for 1 h at room temperature and washed with 25 mM Tris–HCl, 100 mM NaCl, 1 mM DTT and 1 mM CaCl2 (pH 8.0). Bound S100A1 was eluted with 25 mM Tris–HCl, 100 mM NaCl, 1 mM DTT and 5 mM EGTA (pH 8.0), and all proteins were dialyzed into 10 mM HEPES, 100 mM NaCl and 1 mM DTT (pH 8.0) with 5 mM CaCl2 or 5 mM EGTA for further studies.
AUC — sedimentation equilibrium
Sedimentation equilibrium-AUC experiments were performed at 20°C on STIP1 (18 μM) and S100A1 (30 μM), or STIP1 and S100A1 mixtures containing 8 and 24 μM of each protein, respectively, using a Beckman Optima XL-A Analytical Ultracentrifuge equipped with an An-60Ti rotor containing six-channel Epon-charcoal centerpieces and with a 1.2 cm path length (Biomolecular Interactions and Conformations Facility, University of Western Ontario). All protein concentrations for S100A1 correspond to dimer concentrations. Absorbance measurements were collected at 250 nm in 0.002 cm radial steps and averaged over 10 scans. Sedimentation equilibrium measurements were collected at rotor speeds of 12 000, 15 000 and 18 000 rpm (STIP1) or 20 000, 24 000 and 28 000 rpm (S100A1) or 7000 and 10 000 rpm (STIP1–S100A1 complex). Data were fit to the following global single-ideal species model equation:
where C is the concentration of protein at radius x, Co is the concentration at initial radius xo, ω is the angular velocity, is the partial specific volume of the protein, ρ is the solvent density, R is the ideal gas constant, T is the absolute temperature, Io is the baseline offset and Mobs is the fit of the molecular mass based on sedimentation data.
Isothermal titration calorimetry
Binding of STIP1 and TPR domains to S100A1 was measured using a MicroCal™ VP-ITC calorimeter or Nano-ITC (TA Instruments) at 25°C. All protein concentrations for S100A1 correspond to dimer concentrations. Twenty-five to 30 aliquots of 5 or 10 μL, respectively, of STIP1 (60 μM) or individual TPR domains (∼150 μM) were titrated to S100A1 (∼15–70 μM). The reciprocal titration of S100A1 (200 μM) titrated to STIP1 (5 μM) was collected over 30 injections of 10 μL each.
Buffer blank or ligand into buffer titrations demonstrated negligible heats of binding and heats of dilution, respectively. Each titration was replicated in duplicate. Isotherms were fit to a single-site binding model using the Origin or NanoAnalyze software provided by the manufacturer to obtain the thermodynamic parameters reported in Table 1, where Kd is the dissociation constant, n is the stoichiometry of binding, and ΔH and ΔS represent the change in enthalpy and entropy of binding, respectively.
NMR experiments were performed on a Varian Inova 600 MHz NMR spectrometer equipped with xyz-gradient triple resonance cryogenic probe at 25°C. Backbone resonance assignments for TPR2B (∼400 μM) were obtained from HNCACB and CBCA(CO)NH experiments prepared in 50 mM sodium phosphate, 50 mM NaCl and 1 mM DTT (pH 7). All spectra were processed using NMRPipe and analyzed using CARA and NMRViewJ [36–38]. Chemical shift assignments for the TPR2B domain of STIP1 have been deposited in the BioMagResBank (http://www.bmrb.wisc.edu) under accession number 26920.
Backbone amide resonance assignments for TPR1, TPR2A and S100A1 were obtained from the BioMagResBank under accession numbers 18691, 18689 and 18231, respectively [39,40]. 1H-15N heteronuclear single-quantum coherence (HSQC) spectra of 15N-labeled S100A1 (∼200 μM) were collected in the presence of each individual TPR domain (∼50 μM). All protein concentrations for S100A1 correspond to dimer concentrations.
For NMR study of the S100A1–TPR2B complex, uniformly labeled 15N, 13C, 2H or 13C, 2H labeled S100A1 and TPR2B were expressed in E. coli in deuterated minimal M9 medium. Cells initially adapted to 70% D2O were used to inoculate 100% D2O M9 growth medium and grown at 37°C to an OD600 of ∼0.7, and expression was induced with 1 mM IPTG. The culture was grown for 72 h at 22°C and proteins were purified as outlined above. 1H-15N TROSY-HSQC NMR experiments were collected for 15N, 13C, 2H-labeled S100A1 (230 μM) and TPR2B (150 μM) in the absence or presence of 13C, 2H-labeled TPR2B (125 μM) or S100A1 (50 μM), respectively.
Molecular modeling of the S100A1 and TPR2B complex
Haddock 2.2 webserver was used to model the S100A1 (PDB: 2LP3) and TPR2B (PDB: 2LNI) complex using chemical shift perturbations from the NMR experiments described above [40–42]. Non-native N-terminal residues of the TPR2B structure representing a disordered 6xHis tag (NH2-MGHHHHHHSHM) were removed for docking purposes. The lowest energy representative solution structures as defined by the authors were used for complex docking. S100A1 interacting residues were defined as those that demonstrated a decrease in intensity >1 SD from the mean intensity change of all residues upon TPR2B titration and displayed >50% surface accessibility . In addition, residues that split into two or three distinct peaks were declared to be involved in the interaction with a single protomer or both protomers, respectively. Active TPR2B residues were defined as those that showed slow exchange between the free and bound states upon the addition of S100A1 and displayed >50% surface accessibility. Passive residues were defined as surface accessible neighbors within 6.5 Å of active residues using default parameters . A total of 1000 initial complexes were calculated of which the best 200 were water refined using standard parameters. These structures were clustered based on intermolecular energies and RMSD from the lowest energy structure. The lowest energy model of the highest scored cluster was chosen as representative of the complex.
Multiple S100A1 dimers bind to STIP1 with different affinities
Previous studies have suggested that the S100 family of proteins binds to STIP1 through its structurally related TPR domains; however, the biochemical details of S100 binding to STIP1 are poorly understood [22,44]. We have used ITC to measure the binding affinities and characterize the thermodynamic details of the STIP1–S100A1 complex.
S100A1 bound to full-length STIP1 with high affinity in the presence of Ca2+. No binding was observed when Ca2+ was sequestered by EGTA chelation, suggesting that STIP1–S100A1 complex formation is Ca2+-dependent (Figure 1A,B). The change in enthalpy (ΔH) upon STIP1 titration to S100A1 is highly endothermic (Figure 1A), indicating that complex formation is entropically driven and probably involving many hydrophobic contacts (Figure 1A). Consistent with this hypothesis, Ca2+ binding to the S100 family of proteins results in structural rearrangement of the EF-hand α-helical motifs resulting in the exposure of a hydrophobic interface involved in numerous S100-family ligand interactions . The stoichiometry of binding of STIP1 to S100A1 deviated dramatically from a single-site binding model, suggesting that multiple S100A1 dimers bind a single molecule of STIP1. This agrees with previous studies, demonstrating that the S100 family members S100A2 and S100A6 bind to multiple individual TPR domains (TPR1, TPR2A and TPR2B) of STIP1 . However, only a single inflection point is seen in the binding isotherm and it was not possible to distinguish the different binding equilibria. The reciprocal ITC experiment in which S100A1 was titrated into STIP1 was also performed (Supplementary Figure S1). The ITC isotherm obtained is uniphasic with a single inflection point. The stoichiometry of binding obtained by fitting the data to a single-site binding model deviates significantly from 1, again suggesting that multiple S100A1 dimers bind a single molecule of STIP1.
STIP1 binds multiple S100A1 dimers in a Ca2+-dependent manner.
Notably, a recent study by Brautigam  demonstrated that due to the complexity of the multisite non-sequential binding model, determination of thermodynamic parameters of three-site binding from a uniphasic isotherm with a single inflection is not reliable. Owing to the potential complexity of STIP1 binding to S100A1, measurements were conducted on individual TPR (TPR1, TPR2A and TPR2B) domains of STIP1 to obtain greater insights into complex formation. Binding affinities between the individual TPR domains and S100A1 differed significantly, an ∼15-fold range between the strongest and weakest interaction (Figure 2A–C and Table 1). Owing to the relatively weak binding of the TPR2A domain compared with the other TPR domains (TPR1 and TPR2B), the enthalpy and binding affinity could not be accurately determined for this interaction (Kd ∼ 15 μM). TPR2B bound to S100A1 with the highest affinity (Kd = 0.76 ± 0.01 μM), ∼6-fold greater than TPR1 (Kd = 4.4 ± 0.1 μM). The enthalpies of binding to each TPR domain were endothermic, in agreement with the values seen for full-length STIP1, and confirm that S100A1 complex formation to each TPR domain is entropically driven. Interestingly, the TΔS of TPR1–S100A1 binding is significantly larger than that of TPR2B–S100A1. This suggests that the amount of apolar surface areas of TPR2B and TPR1 buried are either significantly different or other factors, such as changes in side-chain and backbone dynamics of S100A1, and TPR domains also make significant contributions to the entropy changes.
The TPR domains of STIP1 bind a single S100A1 dimer with various affinities.
|TPR domain||n1||Kd (10−6 M)2||ΔH2 (kcal/mol)||TΔS2 (kcal/mol)||ΔG2 (kcal/mol)|
|TPR1||0.95||4.37 ± 0.1||55.4 ± 1.0||62.6||−7.2 ± 1.0|
|TPR2B||0.96||0.76 ± 0.01||20.8 ± 0.3||29.1||−8.3 ± 0.3|
|TPR domain||n1||Kd (10−6 M)2||ΔH2 (kcal/mol)||TΔS2 (kcal/mol)||ΔG2 (kcal/mol)|
|TPR1||0.95||4.37 ± 0.1||55.4 ± 1.0||62.6||−7.2 ± 1.0|
|TPR2B||0.96||0.76 ± 0.01||20.8 ± 0.3||29.1||−8.3 ± 0.3|
Binding stoichiometry of dimeric S100A1 and the TPR domains of STIP1.
Kd is the dissociation constant, and ΔH, ΔS and ΔG are the changes in enthalpy, entropy and Gibbs-free energy upon binding at T = 298.15 K, respectively.
STIP1 binds to three S100A1 dimers
Studies have indicated additional S100 protein complex with STIP1 through TPR–motif interactions in each domain through pull-down experiments . These include S100A2 and S100A6, but not S100B, S100A4, S100A10, S100A11, S100A12 or S100A13 . Thus, interaction with STIP1 is not a property shared by all S100 family members. As well, S100 family members may differ in their stoichiometry of binding to STIP1. S100A2 and S100A6 dimers have been suggested to bind STIP1 monomer in a 4:1 and 2:1 molar ratio, respectively . Since our ITC results did not fit well to a single-site model, the stoichiometry of binding of S100A1 to STIP1 could not be determined. Thus, sedimentation equilibrium-AUC experiments were conducted to gain additional insights into the stoichiometry of binding of S100A1 to full-length STIP1. S100A1 exhibited an Mobs value of 24.2 ± 0.2 kDa, consistent with the fact that S100A1 exists as a homodimer in solution of a calculated molecular mass of 21.1 kDa (Figure 3A) . STIP1 presented an Mobs value of 55.3 ± 0.9 kDa, reasonably close to the predicted molecular mass of 62.6 kDa and consistent with current monomeric models of STIP1 function in solution during Hsp70 and Hsp90 client protein transfer (Figure 3B). AUC of STIP1 in the presence of S100A1 revealed the formation of a larger molecular mass species with an Mobs value of 126.1 ± 1.2 kDa (Figure 3C). These results are in agreement with a single S100A1 dimer binding to each of the three TPR domains of STIP1, which would have an inferred molecular mass of 125.9 kDa. We are aware that with the concentrations of STI1P and the S100A1 dimer used in the AUC experiments (8 and 24 μM, respectively), the three binding sites on STIP1 would not be completely saturated. Therefore, a mixture of molecular species (i.e. free S100A1, STIP1 binds to one S100A1, STIP1 binds to two S100A1 and STIP1 binds to three S100A1) could exist in equilibrium. However, due to the low rotor speed used in the AUC experiment (for the detection of high molecular species) and the exponential equation used for the data fitting (Eqn 1), the calculated molecular mass would be biased towards the high molecular mass species (i.e. STI1P in complex with three S100A1).
A single STIP1 molecule binds three S100A1 dimers simultaneously.
S100A1 shares a common interface with each TPR domain
We then used NMR spectroscopy to map the locations involved in binding between S100A1 and individual TPR domains. Titration of 15N-labeled S100A1 with increasing concentrations of each TPR domain (TPR1, TPR2A and TPR2B) resulted in a global reduction in peak intensity (Supplementary Figure S2). The loss in signal is probably the result of signal broadening through enhanced relaxation, the consequence of increased molecular mass of the S100A1–TPR complex (37.4 kDa) in relation to the unbound S100A1 spectrum (21.1 kDa). Interestingly, a significantly greater loss in peak intensity was observed in residues localized to a continuous region spanning α-helices III and IV of S100A1 (Supplementary Figure S2A–C). We speculated that these residues are involved in interacting with the TPR domains. Upon binding TPRs, the signals of these residues were split into the free- and bound-state forms, leading to a greater loss in their peak intensities. Residues of S100A1 showing the largest peak intensity attenuations upon the addition of each TPR domain were mapped to an identical region, suggesting that they share a common binding site on S100A1. However, the reduction in intensity of some S100A1 signals upon binding TPRs can also be due to the changes in protein dynamics in addition to the direct interactions with the TPR-binding partner.
1H-15N TROSY-HSQC NMR experiments were then performed on uniformly deuterated S100A1 and TPR2B in an attempt to observe the bound-state peaks of S100A1, which were not observed in the regular HSQC experiment due to the large molecular size of the bound-state complex. Studies focused on the TPR2B domain, which possesses the highest binding affinity for S100A1 among the TPR domains of STIP1 (Table 1). 1H-15N TROSY-HSQC of S100A1 upon the addition of TPR2B resulted in a global loss in signal intensity, with residues spanning α-helices III and IV of S100A1 experiencing greater attenuations (Figure 4A,B), consistent with experiments conducted on the non-deuterated proteins. In addition, two distinct phenomena were observed in selected peaks in the S100A1 spectra. For a few residues, the initial single cross-peak corresponding to the unbound S100A1 dimer splits into three distinct peaks (the original resonance from the free-state S100A1 and a couple of new peaks from the two protomers of the bound-state S100A1 homodimer) with unique chemical shifts upon titration of TPR2B (Figure 4C, bottom). These results indicate that the exchange between the free and bound states is slow on the NMR timescale and are consistent with a single TPR2B molecule interrupting the symmetry of the S100A1 dimer. Identical residues on each S100A1 protomer experience different chemical environments through interactions with unique regions of TPR2B, thus becoming structurally non-equivalent in the bound conformation. These results agree with an asymmetric mode of binding of one TPR2B monomer to an S100A1 dimer, reflected by our ITC measurements (Figure 2). These changes were concentrated to residues close to the center of the S100A1 hydrophobic groove in close proximity to the dimeric interface of S100A1. Alternatively, a large number of residues were observed to split into two distinct peaks (the original resonance and a new peak from the free- and bound-state of S100A1, respectively) upon TPR2B binding (Figure 4C, top). The majority of these spectral changes were mapped to residues of α-helix III, which, with helix IV, form the binding pocket of the S100A1 dimer. The appearance of a single NMR peak in the bound conformation agrees with them resulting from TPR2B contacts with a single protomer of S100A1 or concerted conformational changes in the dimer. The localization of these two different splitting patterns suggests a model where a single TPR2B molecule binds asymmetrically across the hydrophobic binding pocket of S100A1, predominately forming contacts with a single protomer and accounting for the appearance of single peak species in the bound conformation. However, S100A1 residues in the center of the cleft and dimer interface would be in contact with non-equivalent regions of TPR2B, resulting in two peaks with distinct resonances.
TPR2B interacts with the cleft formed by the EF hand of S100A1.
S100A1 binds to the C-terminal α-helix of TPR2B
1H-15N TROSY-HSQC spectra of deuterated TPR2B in complex with the S100A1 dimer at a stoichiometric ratio of 1 : 0.3 resulted in the appearance of multiple additional peaks (Figure 5B–E), which were absent from or poorly resolved in the non-deuterated HSQC spectra of the complex (data not shown). Chemical shift mapping indicated the majority of novel signals visible in the bound state localized to a continuous stretch spanning α-helix 7 of TPR2B (K118, A120, D122, G123, Q125, R126, M129 and Y132; Figure 5A). Several additional changes were mapped to α-helix 6 and the interconnecting loop adjoining α-helices 6 and 7 of TPR2B. The identity of a few new peaks seen only in the bound state of TPR2B could not be determined with great confidence due to the large chemical shift differences between the free- and bound-state signals and spectral crowding in those regions. Thus, they are not reflected in the chemical shift mapping of the TPR2B-binding site. NMR spectra of equimolar ratios of TPR2B and the S100A1 dimer resulted in significant peak broadening of the majority of TPR2B signals, possibly due to the potential aggregation of the sample at the high concentrations necessary for NMR studies. Attempts at deletion of the C-terminal of TPR2B resulted in the protein becoming insoluble and prevented further experimentation (data not shown). Nevertheless, these results suggest that the C-terminal of TPR2B is the primary region contributing to complex formation with the hydrophobic groove of S100A1.
The C-terminal of TPR2B interacts with S100A1.
Molecular model of the S100A1–TPR2B complex
Chemical shift perturbations observed in the S100A1 and TPR2B NMR spectra were used to generate a model of the complex using the HADDOCK webserver . In the model, α-helix 7 of TPR2B extends across the hydrophobic pocket of S100A1, which is composed of α-helices III and IV and the interconnecting loop of S100A1 (Figure 6A,B). The majority of S100A1 NMR resonances that split into two distinct peaks upon the addition of TPR2B clustered to α-helix III, suggesting that TPR2B predominately interacts with a single protomer of S100A1. The C-terminal α-helix 7 of TPR2B projects into the binding cleft of S100A1, in close proximity to the interconnecting loop of the EF hand of each protomer. Residues in these regions exhibited the appearance of two additional HSQC peaks in the TPR2B-bound form of S100A1 (Figure 4) and are predicted to make contacts with a single TPR2B molecule. Thus, TPR2B may partially occlude binding of a second TPR2B molecule to the unbound protomer of S100A1 and result in an asymmetric arrangement indicated by the 1 : 1 stoichiometry observed in our ITC experiments.
Representative model of the S100A1 and TPR2B complex.
The studies presented here provide molecular details of the interactions between S100A1 and STIP1. ITC thermograms indicated a high-affinity interaction between S100A1 and full-length STIP1. The stoichiometry of binding (n) differed dramatically from a single one-site model; however, only a single inflection point was evident in the isotherms, suggesting multiple binding sites with similar thermodynamic properties. The possibility of multiple binding sites complicated fitting the isotherm to a single binding model; thus, we investigated binding to individual TPR domains. Previous studies have demonstrated that each individual TPR domain of STIP1 is capable of binding related S100 family members (S100A2 and S100A6) . However, these studies were limited to affinity pull-down experiments and the thermodynamic details of S100 binding to each TPR domain and S100 proteins are lacking. Though similar in structure, S100A2 and S100A6 dimers have been shown to bind STIP1 at stoichiometric ratios of 4:1 and 2:1 by pull-down experiments, respectively . Steric hindrance or conformational rearrangements in the STIP1 modular structure may alter the binding capacity of STIP1 for S100 proteins, even though individually each TPR domain possesses the capacity to bind S100 proteins.
ITC binding experiments of each TPR domain titrated to S100A1 revealed significantly different binding properties. Binding affinities differed ∼20-fold between the tightest bindings, TPR2B (Kd = 0.76 ± 0.01 μM), and the weakest, TPR2A (Kd estimated to be ∼15 µM). However, we cannot rule out potential co-operativity in binding in the context of the full-length protein, which may enhance association between S100A1 dimers and the TPR domains of STIP1. TPR domains are believed to function as ancient protein–protein interaction modules and are often seen in tandem across multiple TPR-containing proteins . It is therefore not uncommon for a single ligand to interface with multiple TPR domains, as is seen with the prototypical and best characterized STIP1-binding partners, Hsp70 and Hsp90 [33,34,50]. Interestingly, in the case of Hsp70 and Hsp90 co-ordination, while each TPR domain is capable of recognizing C-terminal region ‘EEVD’ motifs of Hsp70 and Hsp90, the individual TPR domains take on different roles to regulate binding properties of the full-length molecule . TPR1 and TPR2B show preference for Hsp70 binding, whereas Hsp90 interacts with TPR2A–TPR2B. In addition, it appears that Hsp70 and Hsp90 are not capable of binding each TPR domain concurrently due to steric hindrance or conformational changes in full-length STIP1 . The binding model for S100A1 differs in that each TPR domain is occupied by an S100A1 dimer.
S100A1 and STIP1 migrated as a dimer or monomer, respectively, in their native states. There has been some disagreement as to the native state of STIP1 in solution, as studies have suggested monomeric and dimeric forms [33,51–53]. The most detailed and recent model of STIP1 suggests a large elongated structure in the free state, while attributing previous dimeric models to be the result of an atypical elongated structure and large hydrodynamic radius of the full-length protein [51,54]. In the presence of S100A1, STIP1 sediments as a larger species with a molecular mass consistent with three S100A1 dimers binding to a single STIP1 monomer.
S100 interactions with chaperone and co-chaperone proteins vary on their dependence on Ca2+ for protein binding, suggesting differing modes of interaction, though structural details involving these interactions are scarce. Therefore, we used NMR chemical shift perturbation mapping to determine the binding interfaces between S100A1 and the individual TPR domains (TPR1, TPR2A and TPR2B) of STIP1. The results revealed a significant decrease in intensity or the appearance of new peaks for residues localized to α-helices III and IV of EF-hand motifs. Each S100A1 subunit is composed of a 14-residue N and 12-residue C-terminal calcium-binding loop . The latter is situated between helices III and IV. NMR solution structures of the apo- and Ca2+-bound forms indicate Ca2+ binding which triggers a large conformational change resulting in the reorientation of α-helix III by ∼100° and exposing a large hydrophobic region in α-helix IV . The Ca2+-dependent binding of STIP1 presented by ITC and the NMR signal changes in α-helices III and IV upon TPR binding agree with the interaction localizing to an interface spanning α-helices III and IV and provide the rationale for the interaction being Ca2+-dependent. The mapped hydrophobic interface on S100A1 for TPR binding is consistent with a variety of S100 family member ligands that associate in a Ca2+-dependent manner . Interestingly, the stoichiometry of complex formation determined by ITC and AUC suggests that a single TPR domain binds a single S100A1 homodimer. This differs from many S100-target peptide complexes solved to date, which display symmetric binding [7,55,56]. However, these studies involved short peptides corresponding to ligand-binding regions, which may not be representative of full-length proteins, where alternate structural elements may occlude binding to both S100 monomers [57,58]. Previously reported asymmetrical complexes between S100A10 and AHNAK or S100A4 and nonmuscle myosin IIA [46–48] indicate that this mode of interaction is shared among S100 family members and is ligand-specific [48,59].
Chemical shift mapping of the S100A1-binding site on TPR2B indicates that the C-terminal of TPR2B mediates complex formation, with the majority of chemical shift perturbations observed in residues of α-helix 7 of TPR2B. This result deviates from the traditional TPR-binding mechanism where the ligand is cradled within a C-shaped groove on the concave surface of the domain [30,34]. Crystal structures of TPR domains of STIP1 in complex with their respective Hsp70 and Hsp90 C-terminal ‘EEVD’ peptide ligands reveal that most protein–peptide contacts are formed by α-helices 1–5 of each TPR domain with the C-terminal α-helices making little to no contact. Notably, the critical carboxylate clamp residues involved in these interactions [(K429 and R433) of TPR2B in human] found in the N-terminal of TPR2B appear to not be involved in TPR2B binding to S100A1 . This agrees with the observations of carboxylate clamp mutations in each of the TPR domains having little to no effect on S100A2 and S100A6 family members binding to STIP1 . Additionally, carboxylate clamp mutations in the TPR domain of PP5 (protein phosphatase 5) had no effect on S100 family member binding, suggesting that this property may be shared among TPR–S100 interactions . This traditional binding mode between TPR domains and their ligand is prohibited due to the steric clashing generated by α-helices III of S100A1 subunits and the N- and C-terminal α-helices of each TPR domain unless significant conformational rearrangements were to occur in the bound state of the complex. The lack of chemical shift perturbations across the C-shaped groove of TPR2B suggests an alternate binding mechanism to S100A1 compared with the traditional TPR-binding surface, where protein–protein interactions are mediated through the C-terminal helices which extend out forming the inner wall of the groove and are accessible binding sites.
Molecular docking simulations of the TPR2B–S100A1 complex suggest that TPR2B binds asymmetrically to the S100A1 dimer, with one of the protomer largely unoccupied. Similar observations have been made in complexes involving S100A4 and short peptide fragments of nonmuscle myosin IIa, which adopt an α-helical structure and share a similar binding site to TPR2B on their respective S100-binding partners [46–48]. The short length of the α-7 helix of TPR2B (∼18 residues) may mimic the binding mode shared by these fragments.
In the TPR2B–S100A1 model, the outer face of the α-helix 7 of TPR2B is inserted into the hydrophobic cleft formed by the S100A1 dimer, with the residues of the carboxylate clamp remaining undisturbed. This agrees with their insignificant contribution to S100 binding. Interestingly, this arrangement leaves the TPR2B ligand-binding groove accessible for binding to the Hsp70 C-terminal peptide . Thus, S100 inhibition of Hsp interactions may not result from direct occlusion of the Hsp C-terminal ‘EEVD’ peptide-binding groove, but through alternate mechanisms. Additional regions have been documented to be involved in TPR2A binding to Hsp90 outside the TPR-binding cleft . S100 binding may interfere with these additional contacts, leaving the binding cleft uninhibited.
Numerous protein–protein interactions have been reported involving the S100 family and TPR domain-containing proteins, many of which are members of the cellular co-chaperone network. These include STIP1, Tom70, FKBP52, FKBP38, CyP40 (cyclophilin 40), CHIP (C-terminus of Hsc70-interacting protein) and PP5 [21,22,44,61,62]. S100 association with these proteins inhibits their interactions with their respective ligands. However, the structural features, which mediate complex formation, have been poorly understood. The structural conservation between TPR domains suggests that they may bind the S100 family of proteins through a common mechanism as the TPR2B domain of STIP1, where the extended C-terminal helices form the basis of the interaction. Thus, the S100 proteins may regulate the chaperone cycle at multiple nodes through binding their respective TPR domains.
In summary, we found that the TPR domains of the co-chaperone STIP1 bind the hydrophobic cleft of S100A1 in a Ca2+-dependent manner, with TPR2B possessing the highest affinity. The C-terminal α-helices of TPR2B mediate this interaction and present a novel binding mechanism for S100 proteins, which may be shared among other TPR domain-containing proteins. Our studies reveal the molecular details of STIP1 interactions with S100A1 and its TPR domains, which may influence STIP1 function upon S100A1 deregulation in multiple disease states.
- E. coli
heteronuclear single-quantum coherence
isothermal titration calorimetry
nuclear magnetic resonance
cellular prion protein
stress-inducible phosphoprotein 1
tobacco etch virus
A.M. performed the experiments and analyzed the data. M.A.M.P. and W.-Y.C. assisted in experimental design and discussion of results. A.M., V.F.P., M.A.M.P. and W.-Y.C. wrote the manuscript.
This work was supported by Canadian Institutes of Health Research (CIHR) [Operating Grants MOP 136930, 126000 and 89919] and the Natural Sciences and Engineering Research Council (NSERC) [Discovery Grant RGPIN 06372-2014].
We thank the Biomolecular NMR Facility and Bimolecular Interaction and Conformation Facility (University of Western Ontario) for their assistance and use of equipment. We also thank Anne Brickenden for her technical expertise and Dr Gary Shaw (University of Western Ontario) for helpful discussions.
The Authors declare that there are no competing interests associated with the manuscript.