Refinement of the solution structure and dynamic properties of Ca²⁺-bound rat S100B

Biological context

The small (~10.5 kDa) acidic Ca²⁺-binding protein S100B belongs to the S100 protein family, a group of over 20 members that share significant sequence homology and undergo a large conformational change upon the addition of calcium. S100 proteins have no intrinsic enzymatic activity; instead, a Ca²⁺-dependent conformational change enables them to bind and modulate the activity of a diverse pool of intracellular target proteins. For S100B, these include cytoskeletal and filament associated proteins (e.g. tubulin, GFAP, tau, desmin, vimentin, CapZ, calponin, calpactin I, and caldesmon), other Ca²⁺-binding proteins (annexins II, V, VI, S100A1, S100A6, S100A11, and neurocalcin-δ), membrane associated proteins (neuromodulin, neurogranin, MARCKS, giant protein AHNAK, and IQGAP1), transcription factors and their regulators (e.g. p53, hdm4, and hdm2), and several enzymes (e.g. aldolase, phosphoglucomutase, photoreceptor guanyl cyclases, Ndr kinase, and protein kinase C) (reviewed in Wilder et al. 2006). In addition, S100B is secreted from the cell under stress conditions and targets cell surface receptors such as RAGE in the extracellular environment (reviewed in Heizmann et al. 2007).

Several S100 proteins, including S100B, are up-regulated in response to various pathologies including Alzheimer’s disease, Down’s syndrome, and a number of malignancies, and are used as disease markers and prognostic indicators. Mechanistic studies suggest that S100B and perhaps other S100 proteins contribute to these diseases. For example in cancers such as malignant melanoma, anaplastic astrocytomas, and glioblastomas, high concentrations of S100B correlate directly with very low protein levels of wild-type p53 tumor suppressor (reviewed in Salama et al. 2008). Upon further examination, it was found that S100B binds p53 in a Ca²⁺-dependent manner, significantly reducing p53 protein levels and inhibiting wild-type p53 functions (Lin et al. 2004; Rustandi et al. 2000). Likewise, siRNA inhibition of S100B production restores wild-type p53 protein levels and its activities (i.e. cell cycle regulation, apoptosis) as necessary to restore normal tumor suppression functions (Lin et al. 2004). Thus, as with other negative regulators of p53 (i.e. hdm2), efforts are currently underway to design small molecule inhibitors that specifically bind to the Ca²⁺-bound form of S100B and block the S100B-p53 interaction as a potential therapeutic strategy to treat cancers with wild-type p53 and elevated S100B (Charpentier et al. 2008). For such drug design efforts to be successful, it is important to fully understand the structural and dynamic properties of Ca²⁺-bound S100B.

While several NMR and X-ray crystal structures of Ca²⁺-S100B are available (Drohat et al. 1998; Matsumura et al. 1998; Smith and Shaw 1998; Wilder et al. 2006), there is a discrepancy with regard to the conformation of the last eight residues in its C-terminus. In the X-ray crystal structure, helix 4 is extended to include nearly the entire end of the molecule (to F88), which is different from the published NMR structures in which helix 4 ends at A83 with the remaining eight residues considered part of the C-terminal loop (loop 4). The importance of this region for binding p53 and other protein targets is well-established (Inman et al. 2002), so the structural and dynamic properties of this portion of Ca²⁺-S100B need to be clearly understood. To this end, we have refined the solution NMR structure of rat Ca²⁺-S100B by including long-range residual dipolar coupling (RDC) constraints. In addition, ¹⁵N-relaxation data for backbone amide groups are presented here. Together, these new data provide direct evidence that residues at the C-terminus of Ca²⁺-S100B (C84-E91), as well as several residues in loop 2 (termed the hinge region), are not fully structured, but rather are dynamic in solution. That these residues are flexible is likely a necessary characteristic of Ca²⁺-S100B for accommodating its interaction with a large number of protein targets.

Methods and results

Materials

Deuterated MES, D₂O, ¹⁵NH₄Cl, and ¹³C-glucose were purchased from Cambridge Isotopes Laboratories (Woburn, MA). Other materials were of the highest commercial quality and passed through Chelex-100 resin (BioRad) to remove trace metals.

Sample preparation

¹⁵N-labeled S100B or ¹⁵N, ¹³C-labeled S100B was overexpressed and purified from Escherichia coli as previously described (Inman et al. 2001). While apo-S100B is dimeric in solution at millimolar concentrations, there was evidence of higher order aggregates in samples of Ca²⁺-S100B prepared at subunit concentrations above 1 mM. However, at concentrations below 600 μM, no aggregation was detected and R ₂ values for backbone ¹⁵N resonances throughout the protein did not change as a function of protein concentration. Specifically, the average trimmed mean R ₂ values at 14.4 T were 17.1 ± 1.6 s⁻¹ at 1.3 mM, 16.6 ± 1.7 s⁻¹ at 900 μM, 15.8 ± 1.9 s⁻¹ at 600 μM, and 15.7 ± 1.9 s⁻¹ at 400 μM protein concentrations, respectively. Furthermore, individual R ₂ values for all residues were indistinguishable within error (<4% different) at 600 and 400 μM S100B subunit concentration. Therefore, NMR experimental collection of both the structural and ¹⁵N-relaxation rate data were conducted at 400 μM S100B, 5 mM CaCl₂, 25 mM NaCl, 10 mM DTT, 10 mM MES, pH 6.7. All samples were degassed by vacuum and argon replacement shortly before each experiment. For partial alignment, necessary for collecting RDC data, a 2× stock of the Ca²⁺-S100B protein/buffer solution was soaked into a cylindrical 5% acrylamide gel overnight to give a final S100B concentration of 400 μM. This gel was then radially compressed (i.e. stretched) or axially compressed to obtain the aligned samples, as previously described (Wright et al. 2005, 2008).

NMR data collection

NMR spectra were collected at 37°C with a Bruker DMX600 NMR spectrometer (600.13 MHz for protons) and a Bruker AVANCE 800 NMR spectrometer (800.27 MHz for protons) equipped with four frequency channels and a triple-resonance z-axis gradient 5 mm cryoprobe. As reported previously for apo-S100B (Inman et al. 2001), ¹⁵N-relaxtion rate data (T ₁, T ₂, heteronuclear NOE) were also examined for Ca²⁺-S100B, so backbone dynamics could also be considered for the state in which this protein binds its targets. Specifically, backbone ¹⁵N T ₁ and T ₂ spectra were acquired with 32 scans per t ₁ point at both 600 and 800 MHz. A recycle delay of 3.0 s was used at both fields. T ₁ delay times of 30 (x2), 70 (x2), 130, 200, 300, 480, and 960 ms were used for data collection at 600 MHz, and 30 (x2), 70 (x2), 100, 200, 300, 400, and 500, 600, and 900 ms at 800 MHz. T ₂ delay times of 16 (x2), 24, 32 (x2), 40, 48, 56, and 64 ms were used for data collection at 600 MHz, and 10 (x2), 20 (x2), 30, 40, 50, 60, and 70 ms at 800 MHz. {¹H}-¹⁵N NOE ratios were measured with a 3-s pre-saturation period and a 2-s relaxation delay; the control experiment had an equivalent 5-s delay. Both NOE and control experiments were acquired in an interleaved fashion, with 56 scans at 600 MHz and 48 scans at 800 MHz and with 256 t ₁ points.

NMR data processing and analysis

All NMR data were processed and analyzed with NMRPipe and NMRDraw software (Delaglio et al. 1995), as previously described (Inman et al. 2001). Data was extended in the indirect dimension (t ₁) by linear prediction, using a prediction order of no more than one-third the number of complex points and zero-filled in both dimensions to double the number of points. Apodization with a 5%-shifted mixed Gaussian/exponential function was applied with weighting chosen to reproduce the natural line width values. For the relaxation series, correlations with S/N > 15 were selected and peak heights were measured by fitting Gaussian surfaces to the transformed data and taking the maximum height of the fitted surface using the program NlinLS (Delaglio et al. 1995). The jackknife procedure was used with the Levenberg-Marquardt non-linear least squares algorithm to fit peak heights, using the computer program Curvefit (A.G. Palmer, Columbia University). Peak heights for the heteronuclear NOE and reference experiments were also extracted from Gaussian fits to the frequency-domain data as previously described (Inman et al. 2001).

Structure calculations

Residual dipolar coupling data (¹ D _NH and ¹ D _CαHα) were collected in order to refine the solution NMR structure of Ca²⁺-bound S100B (Drohat et al. 1998). Specifically, 2D IPAP ¹H-¹⁵N HSQC experiments were used to record ¹⁵N-H^N splittings, and 3D CT-(H)CA(CO)NH experiments without H_α decoupling during C_α acquisition in t ₂ were used to record ¹³C_α-H_α splittings. Residual dipolar coupling values (in Hz) were determined by measuring the difference in J splitting between data collected under conditions with isotropic tumbling and analogous data collected with partially aligned samples. The RDCs were incorporated into structure calculations using alignment tensors generated from data collected under two alignment conditions as previously described (Tjandra et al. 1997; Bax et al. 2001; Wright et al. 2005). ¹ D _NH and ¹ D _CαHα data from a radially compressed gel (i.e. stretched gel) were fit simultaneously to enforce a single alignment tensor (A _a  ~ 0.6 × 10⁻³) since these data were collected under the same conditions. A second ¹ D _NH data set was also collected in an axially compressed gel (A _a  ~ 0.6 × 10⁻⁴), and these data slightly improved the overall quality of the structures. The Q-factor reported for the family of structures was calculated twice and the average value is listed in Table 1 (Q = 0.19 ± 0.02; Table 1). In each calculation, 10% of the ¹D_NH and 10% of the ¹D_CαHα RDCs from the radially compressed gel were randomly removed and then used to calculate the Q-factor, as described (Bax et al. 2001; Wright et al. 2005). Most structures (>180 of 200) had no NOE or dihedral violations, and many (>75) displayed a high level of convergence (backbone rmsd <0.85). The final 20 structures were selected based on lowest energy and were of high quality based on the statistical criteria listed in Table 1. The coordinates of the refined solution structure of Ca²⁺-S100B have been deposited in the Protein Data Base (PDB identification code: 2K7O), and the T ₁, T ₂, NOE, and RDC values were deposited at the BMRB under the accession code 15923.

Table 1 NMR-derived restraints and statistics of 20 NMR structures^a

Refined solution structure of Ca²⁺-S100B

The solution NMR structure of dimeric Ca²⁺-S100B reported previously (Drohat et al. 1998) was refined by the inclusion of 246 ¹ D _NH and 90 ¹ D _CαHα residual dipolar coupling constraints (RDCs) (Fig. 1a). RDCs for residues with heteronuclear NOE values below 0.75 were not included in the structure calculations (i.e. for residues L3, E45, E46, E49, E62, H85, F87-E91). The addition of the RDC data into the structure calculations produced no NOE or dihedral angle violations and confirmed the accuracy of the original NOE-based structure (Drohat et al. 1998). As expected, the additional RDC constraints decreased the root mean squared difference (rmsd) for all ordered backbone atoms (1–88) from 1.2 to 0.8 Å in the 20 best structures, with Q-values lowering from 0.51 ± 0.05 in the original NOE-based structure to 0.19 ± 02 in the refined NMR structure (Table 1; Fig. 1b). The interhelical angles and distances were also unchanged within error upon addition of RDC data. While no differences in structure could be detected, the refined NMR structure reported here is better resolved and higher in quality than that originally published (Drohat et al. 1998).

Fig. 1

Refined structure of rat Ca²⁺-S100B using residual dipolar couplings. a Plot of calculated vs. experimental RDC for D _NH and D _CαHα, showing that the data fit the structure well. b Overlay of the 20 best Ca²⁺-S100B structures in stereo view. Residues in helices are colored red in subunit 1 and blue in subunit 2. Side chains of ordered secondary structure are colored in green. All other residues in both subunits are colored in gray

¹⁵N relaxation rate data for Ca²⁺-S100B

It was observed that B-factor values in the X-ray crystal structure were elevated in several regions of Ca²⁺-S100B (Matsumura et al. 1998), perhaps as a result of internal motion. Specifically, residues in helix 1 (K5, H15, Q16), the pseudo EF-hand (E21, K24, H25), loop 2 (i.e. the hinge region) (E45, E46, K48, and E49), the canonical EF-hand (S62, D63, G64), and in the C-terminus (E86, F87, F88) displayed B-factors that are more than 1.5 times higher than the standard deviation of the 10% trimmed mean (Fig. 3a). It was also observed that the largest differences observed between RDCs predicted by the X-ray structure and those measured by NMR occurred for most of these same residues. Furthermore, when these regions with elevated B-factors were not included in the Q-factor calculation, the quality factor for the X-ray structure improved from Q = 0.45 to Q = 0.35, providing additional evidence that these regions may exhibit internal dynamics. Therefore, NMR ¹⁵N-relaxation data (R ₁, R ₂, and {¹H}-¹⁵N NOE) were collected at 14.4 and 18.8 Tesla to more directly determine whether these and/or other regions of Ca²⁺-S100B are mobile on one or more timescales.

Of the 91 backbone resonances in Ca²⁺-S100B, ¹⁵N-relaxation rate data were evaluated for 79 residues with the remaining 12 residues not analyzed either due to spectral overlap and/or because they were missing or very weak because of exchange broadening. The 10% trimmed mean values of the R ₁ (1/T ₁) values for the 79 observable correlations were 0.894 s⁻¹ at 800 MHz and 1.275 s⁻¹ at 600 MHz (Fig. 2a). For R ₂, the trimmed means were 19.67 s⁻¹ for 800 MHz and 15.32 s⁻¹ for 600 MHz (Fig. 2b), and the NOE trimmed mean values were 0.835 for 800 MHz and 0.796 for 600 MHz (Fig. 2c). In addition to residues that were missing or visibly exchange broadened (residues S1, K5, G22, D23, K24, H42, F43, I47, K48, Q50, V80, T81, T82, H85, E86, and F87), several other residues (D12, H15, L44, A83, and C84) were also identified as having R_ex from a plot of residual dipolar couplings ¹ D _NH versus T ₁/T ₂ ratio (Fig. 2d) (de Alba et al. 1999). Thus, residues located in helix 1 (K5, D12, H15), the pseudo EF-hand (G22, D23, K24), the hinge region (H42, F43, L44, I47, K48, Q50) and the C-terminal tail of S100B (V80, T81, T82, A83, C84, H85, E86, and F87) may be mobile on the chemical shift timescale, perhaps due to motion in the C-terminal loop (residues 84–91) as observed previously in apo-S100B (Inman et al. 2001) (Fig. 3b). In addition, several residues in the hinge region (E45, E46, E49) and the C-terminal loop (loop 4) of Ca²⁺-bound S100B (H85, F87-E91) exhibited {¹H}-¹⁵N NOE values that were below 0.75 at two field strengths indicating that these regions of the protein are also mobile on fast timescales (ns-ps; Fig. 3c).

Fig. 2

¹⁵N-relaxation rate data for Ca²⁺-S100B. a Longitudinal relaxation rate R ₁, b transverse relaxation rate R ₂, and c {¹H}-¹⁵N heteronuclear NOE (η + 1) data are given for nitrogen Larmor frequencies of 81 MHz (solid diamonds) and 61 MHz (open diamonds). d Residual dipolar couplings, D _NH, plotted against ¹⁵N T ₁/T ₂ ratios collected at 61 MHz; data included in the linear correlation are displayed as filled circles (●), and those considered outliers (>2× the standard deviation to the right of the best-fit line) as open circles (○)

Fig. 3

Ribbon diagrams of Ca²⁺-S100B illustrating elevated B-factors from the X-ray crystal structure and ¹⁵N-relaxation rate data relevant to its dynamic properties on slow and fast timescales. a Residues with B-factors significantly larger than the average B-factor are mapped (in blue) onto the ribbon diagram for the X-ray structure of Ca²⁺-S100B (PDB accession number 1MHO). b Residues for which chemical exchange is detected are mapped (in green) onto the refined solution NMR structure of Ca²⁺-S100B. c Residues that have low {¹H}-¹⁵N NOE values at 600 MHz, indicative of fast timescale motions, are mapped (in red) onto the refined solution NMR structure

In summary, helix 1 and all four loops in S100B likely have motion on fast and/or slow timescales as judged by NMR and supported by B-factors from X-ray crystallography. As was observed previously for apo-S100B (Inman et al. 2001), residues in C-terminal loop of Ca²⁺-bound S100B are mobile on fast and perhaps slow timescales as judged by low heteronuclear NOE values (H85, F87-E91), exchange broadening effects (A83-F87), and elevated B-factors or no density in the X-ray crystal structure (E86-E91). Similarly loop 2, termed the ‘hinge region,’ experiences fast-timescale dynamics for several residues (E45, E46, E49) in addition to having exchange broadening (E45, E46, K48, E49) and elevated B-factors (E45, E46, K48, E49). It is noteworthy to point out that these same two regions of S100B (the hinge and the C-terminal loop) make up a large portion of the target protein binding site on S100B, so it is possible that the dynamic properties in these regions are functionally important since they may provide a means for S100B to interact with its diverse set of targets.

In addition to loops 2 and 4, three NH correlations for residues in helix 1 are exchange broadened (K5, D12, and H15) with L3 exhibiting a low heteronuclear NOE value (<0.75) and K5 having an elevated B-factor in the X-ray crystal structure; however, no other helix in S100B shows such features. While it is possible that these exchange broadening effects could result from movements of helix 1 on the chemical shift timescale, the extremely low S100B dimer dissociation constant suggests that this helix is quite stable; it is more likely that such exchange broadening results from the C-terminal loop approaching helix 1 as was observed in apo-S100B (Inman et al. 2001). Similarly, three residues in the loop of the pseudo-EF-hand have elevated B-factors (G22-K24) and/or exhibit exchange broadening effects by NMR (E21, K24, H25), which may also be the result of its proximity to the C-terminal loop. Lastly, several residues in the typical EF-hand (S62-G64) have elevated B-factors in the X-ray structure, and one residue (N62) exhibits a low heteronuclear NOE value. This result was somewhat surprising since N62 is in between two Ca²⁺-liganding residues (D61 and D63). However, previous studies have shown that the Ca²⁺-binding affinity of S100B is rather weak in comparison to other EF-hand proteins, but increases upon binding protein targets (Markowitz et al. 2005). These issues will make it interesting to determine whether the dynamic properties observed here for Ca²⁺-S100B are eliminated upon binding a protein target since such an effect could provide a mechanistic explanation for how protein target binding can increase the affinity that S100B has for calcium.

Discussion and conclusions

The solution NMR structure of Ca²⁺-S100B was refined by the inclusion of residual dipolar coupling data, which resulted in a higher quality ensemble of structures with lower backbone rmsd values (Table 1; Fig. 1). The comparison of this structure to that of S100B in the apo-state is important for fully understanding the Ca²⁺-dependent conformational change in S100B that is necessary for it to bind its protein targets. As discovered for several S100 proteins so far, the rotation and conformational change of helix 3, brought about by calcium binding, is accompanied by breaking and forming numerous hydrophobic contacts in S100B (Wilder et al. 2006; Drohat et al. 1998; Smith and Shaw 1998). In the apo-form (Drohat et al. 1999), helix 3 and helix 4 are held together by a large hydrophobic strip of residues, consisting primarily of V53, V56, and L60 in helix 3, and F76, M79, V80, and A84 in helix 4. In contrast to the well-packed helix 3–4 interface, helices 2 and 3 do not pack together well in apo-S100B, primarily because of the position of the bulky M57, along with the presence of a charged residue (K33) in the middle of an otherwise hydrophobic area. Upon the addition of calcium, the large conformational change of helix 3 allows for several residues in helix 3 (V56 and L60) to pack against helix 4 (F76 and M79), forming a smaller hydrophobic region. However, other residues, including V52, V80, and A83, no longer contribute to the helix 3–4 helix packing. On the opposite face of helix 3, the conformational change alters the residues of helix 3 that are located at the helix 2–3 interface; M57 is rotated to the periphery of this interface, and V57 and V53 exhibit much more contact with helix 2. Complementing this movement, the slight reorientation of helix 2, brought about by Glu-31 coordinating with calcium, presents a more hydrophobic side of helix 2 to helix 3; L32 and I36 are rotated into the helix interface, and importantly K33 is moved slightly out of this region. This allows for more substantial hydrophobic contact between the helices 2 and 3, without the destabilizing effect of a charged residue in this region. The presence of these hydrophobic interactions between helices 2 and 3 in the Ca²⁺-bound state, most notably those between V53 and L32, is probably necessary to stabilize the open conformation of Ca²⁺-S100B.

S100A1 and S100A6, two other S100 proteins for which both Ca²⁺-loaded and apo structures are available, also exhibit poor helix packing between helices 2 and 3 in the apo form of the molecule. As is the case in S100B, this unfavorable packing is brought about by interactions between the bulky methionine at residue 57 and the charged lysine at position 33 (S100B numbering). Interestingly, the rotation of Lys-33 out of the helix 2–3 interface upon the addition of calcium is conserved among S100A1, S100A6, and S100B. Additionally, all three proteins retain a conserved leucine at residue 33 and a conserved valine at residue 53, and demonstrate similar sidechain orientations in the presence of calcium. Such helix packing in both the apo and the Ca²⁺-bound form of S100 proteins is thus probably a common feature within this family of Ca²⁺-binding proteins and could explain why little if any internal dynamics are observed for helical residues of Ca²⁺-S100B.

The X-ray and the newly refined NMR structures of Ca²⁺-S100B are nearly indistinguishable with the only difference being eight residues in the extreme C-terminus (residues C84-E91); the NMR shows this region to be largely unstructured while the X-ray structure shows this area to be helical. Examination of ¹⁵N-relaxation rates resolved this discrepancy, since these data demonstrated that the C-terminus of Ca²⁺-S100B (residues A83 to E91) is mobile on fast timescales (ps-ns) in solution and several residues in this region also exhibit exchange broadening effects consistent with motion on slow timescales (μs-ms). However, it is important to note that the presence of target peptide induces the same helix 4 extension in solution as was reported for crystallized Ca²⁺-S100B (Inman et al. 2002; Matsumura et al. 1998; Rustandi et al. 2000); therefore, the extended helix 4 observed in the crystal structure may be an effect of crystal contacts that mimic target protein binding.

The refined NMR structure of Ca²⁺-bound S100B and the accompanying ¹⁵N-relaxation rate data provide the foundation for more complete analyses of the dynamic properties of the protein in this state as well as for making detailed comparisons of S100B in the apo-, Ca²⁺-bound, and target-bound forms. In summary, the results presented here provide evidence that residues at the C-terminus of Ca²⁺-S100B (A83-E91), as well as several residues in loop 2 (termed the hinge region), are dynamic in solution. The unstructured, flexible regions we have observed for the hinge (loop 2) and loop 4 are likely to be important for accommodating the binding of wide variety of proteins to Ca²⁺-S100B. Because of the high degree of structural and sequence homology, similar dynamic regions may be present in other S100 proteins. In fact, the Ca²⁺-bound form of both S100A1 and S100A6 demonstrate increased rmsd in the hinge region and the C-terminus, presumably due to fewer NOE correlations perhaps caused by a higher degree of residue mobility (Wright et al. 2005; Mäler et al. 2002). In order to assess the interplay between structure, dynamics and target binding, it is important to extend NMR studies of Ca²⁺-S100B to also include target peptide(s) complexes as well as to other S100 proteins to determine whether the features observed here for Ca²⁺-S100B are common throughout the S100 protein family.