Divalent Metal Ion Complexes of S100B in the Absence and Presence of Pentamidine

Abstract

As part of an effort to inhibit S100B, structures of pentamidine (Pnt) bound to Ca²⁺-loaded and Zn²⁺,Ca²⁺-loaded S100B were determined by X-ray crystallography at 2.15 Å (R_free = 0.266) and 1.85 Å (R_free = 0.243) resolution, respectively. These data were compared to X-ray structures solved in the absence of Pnt, including Ca²⁺-loaded S100B and Zn²⁺,Ca²⁺-loaded S100B determined here (1.88 Å; R_free = 0.267). In the presence and absence of Zn²⁺, electron density corresponding to two Pnt molecules per S100B subunit was mapped for both drug-bound structures. One Pnt binding site (site 1) was adjacent to a p53 peptide binding site on S100B (±Zn²⁺), and the second Pnt molecule was mapped to the dimer interface (site 2; ±Zn²⁺) and in a pocket near residues that define the Zn²⁺ binding site on S100B. In addition, a conformational change in S100B was observed upon the addition of Zn²⁺ to Ca²⁺–S100B, which changed the conformation and orientation of Pnt bound to sites 1 and 2 of Pnt–Zn²⁺,Ca²⁺–S100B when compared to Pnt–Ca²⁺–S100B. That Pnt can adapt to this Zn²⁺-dependent conformational change was unexpected and provides a new mode for S100B inhibition by this drug. These data will be useful for developing novel inhibitors of both Ca²⁺-and Ca²⁺,Zn²⁺-bound S100B.

Introduction

There are more than 20 S100 proteins originally named because of their solubility in 100% saturated ammonium sulfate.¹ Like calmodulin and most other EF-hand-containing proteins, S100 proteins typically function as a calcium-activated switch that binds and regulates the biological function of numerous protein targets (Fig. 1).^2-5 A unique feature for several S100 proteins is that they also bind Zn²⁺ (K_d ∼ 90 nM; S100B) at a structurally conserved site that is separate from the EF-hand calcium-binding loops.^6-11 Although the biological function of Zn²⁺ binding is not completely understood, it is not considered a structural ion since S100 proteins are typically stable, folded, and active in its absence. Instead, Zn²⁺ likely plays a regulatory role by modulating the affinity of various S100 proteins for Ca²⁺ and/or protein targets, which has been observed in vitro for several S100 proteins, including S100B.^2,5,8,12-15

Fig. 1.

The “calcium switch” in S100B. Shown are ribbon diagrams of apo–S100B (NMR, PDB code: 1B4C) and Ca²⁺-bound S100B (NMR, PDB code: 1QLK; X-ray, PDB code: 1MHO) illustrating the calcium-dependent reorientation of helix 3 (yellow) in each S100B subunit, termed the “Ca²⁺ switch.” The other three helices in each subunit of the symmetric S100B homodimer are colored red and blue, respectively. The loop regions of S100B are colored gray, and the two calcium ions per subunit are cyan spheres. This calcium-dependent conformational change is required for S100B to interact with specific protein targets such as p53.

Members of the S100 protein family are distributed in a cell-specific manner,^3,4,16 including those in a large number of human cancers.^17-19 For S100B, protein levels are elevated in malignant melanoma,¹⁹ anaplastic astrocytomas,^20,21 and glioblastomas.²² In the case of malignant melanoma, high concentrations of S100B correlate directly with poor prognosis in patients,^23,24,25 and hence, it is used as a clinical marker for this and other cancers. In more recent mechanistic studies, S100B was found to bind directly to the p53 tumor suppressor protein in primary human malignant melanoma cells, reduce p53 protein levels, and inhibit wild-type p53 functions.^6,26-30 Therefore, elevated levels of S100B contribute to cancer progression by down-regulating wild-type p53 protein.³⁰ Correspondingly, p53 protein levels and its associated activities were restored in malignant melanoma when S100B expression was inhibited by siRNA (small interfering RNA).²⁹ With these siRNA^S100B results in mind, a rational drug design approach was initiated to develop small-molecule inhibitors that bind Ca²⁺–S100B, prevent S100B–p53 complex formation, and restore p53-dependent tumor suppression in cancers with wild-type p53 such as malignant melanoma.^31-33 Furthermore, engineering inhibitors of the Ca²⁺,Zn²⁺-bound form of S100B may also be important since Zn²⁺ homeostasis is disrupted and its levels are typically elevated in patients with skin cancer.^34-37

Atomic structures of S100B in the Ca²⁺- and p53 peptide-bound states together with computer-aided drug design and high-throughput screening approaches were used to identify small molecules that bound S100B.^31-33 One such compound, pentamidine (Pnt), was characterized by NMR and found to interact with aromatic residues in helix 4 and the C-terminal loop of Ca²⁺–S100B.³¹ To pursue this S100B–drug interaction further, we determined the X-ray crystal structures of Pnt bound to divalent metal ion complexes of S100B in this study. In the calcium-bound state, S100B was found to bind two Pnt molecules per S100B subunit with one molecule located near the p53 peptide binding site (termed site 1), as predicted previously.³¹ The second Pnt molecule bound a separate site (site 2) at the dimer interface of Ca²⁺–S100B and near residues of the unoccupied Zn²⁺ site. However, in the presence of Ca²⁺ and Zn²⁺, the conformation and relative orientation of both S100B-bound Pnt molecules were different from that observed in the presence of Ca²⁺ alone. Together, the two S100B–Pnt structures will be important for developing novel, high-affinity inhibitors of Ca²⁺–S100B in the absence and presence of Zn²⁺, which could have clinical benefits for treating cancers with elevated levels of S100B.

Results

The binding of S100B to p53 down-regulates tumor suppressor activity in cancer cells such as malignant melanoma^29,38; thus, a search for small molecules that bind S100B and prevent S100B–p53 complex formation was undertaken.^31,33 For one such molecule, Pnt, a model for drug binding was reported based on chemical shift perturbations, saturation transfer difference (STD) data, and several intermolecular nuclear Overhauser enhancement constraints.³¹ Here, the structure of this Pnt–Ca²⁺–S100B complex was solved using X-ray crystallography. Due to the relatively high levels of Zn²⁺ found in skin and in skin cancers such as malignant melanoma,^36,37,39 these studies were extended to also include a detailed examination of the interaction between Pnt and Zn²⁺,Ca²⁺–S100B.

Thermodynamic binding studies

As reported previously,³¹ the interaction between Pnt and S100B was confirmed to be calcium dependent [isothermal titration calorimetry (ITC): ^PntK_d=53±10 μM; fluorescence: ^PntK_d=39±3 μM; Fig. 2]. However, additional experiments showed that two Pnt molecules bound each S100B subunit with indistinguishable affinities but weaker than that reported previously for the F43W mutant of S100B (K_d∼1 μM)³¹ (Table 2; Fig. 2a). Under the same conditions, Pnt was also found to bind Zn²⁺,Ca²⁺–S100B by ITC (^PntK_d=64±15 μM; Table 2); however, when the Pnt:S100B subunit ratio approached 2:1 in these titrations, some precipitation was observed. To avoid precipitation, we also monitored the binding by titrating both Zn²⁺ and S100B (at a 1:1 ratio) into a solution of Pnt and monitoring the change in Pnt fluorescence (^PntK_d =39±5 μM; Table 2; Fig. 2b). The salt KCl was also included to curb precipitation, but this salt had only modest effects at higher Pnt:S100B ratios. Furthermore, KCl had little or effect on Pnt binding to either Ca²⁺–S100B (ITC: ^PntK_d=50±5 μM; fluorescence: ^PntK_d=44±4 μM) or Zn²⁺,Ca²⁺–S100B (ITC: ^PntK_d=35±12 μM; fluorescence: ^PntK_d=40±5 μM; Table 2). Pnt binding to each divalent metal ion state of S100B (i.e., Ca²⁺-bound, Zn²⁺,Ca^2+--bound) was enthalpically favored as judged by negative ΔH_app values measured using ITC (−4.0 to −5.0 kcal/mol; Table 2). The entropic contribution to Pnt binding could also be estimated using the dissociation constant and ΔH_app, and in each binding experiment, a thermodynamically favored TΔS resulted (+1.0 to +2.1 kcal/mol; Table 2). Such a thermodynamic profile is consistent with productive hydrophobic and hydrogen-bonding interactions upon complex formation.^40,41 When Zn²⁺ binding to Ca²⁺–S100B and to Pnt–Ca²⁺–S100B was examined, the dissociation constants (no drug: ^ZnK_d=94±17 nM; with Pnt: ^ZnK_d=134±9 nM) and the thermodynamic signature of Zn²⁺ binding (i.e., ΔH_app, TΔS) were found to be similar to that determined previously in the absence of Pnt.⁷

Fig. 2.

The binding of Pnt to Ca²⁺–S100B and Zn²⁺, Ca²⁺–S100B. (a) Representative ITC data illustrating the binding of Pnt to Ca²⁺–S100B (■; n=2). Conditions included 10 mM Tes, pH 7.2, 15 mM NaCl, and 10 mM CaCl₂, at 37 °C. Shown also is a control experiment illustrating no Pnt binding to S100B in the absence of calcium (red boxes). (b) Representative titration of Zn²⁺–S100B (one Zn²⁺ ion per S100B subunit) with Ca²⁺ (10 mM) into a solution of Pnt and Ca²⁺ (10 mM) as monitored by changes in Pnt fluorescence (n=1.7). Conditions included 10 mM Tes, pH 7.2, 15 mM NaCl, 100 mM KCl, and 10 mM CaCl₂, at 37 °C.

Table 2.

Dissociation constants (K_d) and enthalpy (ΔH_app) and entropy (TΔS) for Pnt and Zn²⁺ binding to S100B

	Kd (μM), ITC	ΔHapp (kcal/mol)	TΔS (kcal/mol)	Kd (μM), fluorescence
Pnt
Ca2+-S100B (low salt)a,b	53±10c	−4.0±0.2	2.1±0.2d	39±3
Ca2+-S100B (high salt)a,b	50±5c	−4.6±0.4	1.5±0.4d	44±4
Zn2+,Ca2+-S100B (low salt)e,f	64±15	−5.0±0.1	1.0±0.2d	39±5
Zn2+,Ca2+-S100B (high salt)b,e	35±12c	−4.3±0.5	1.6±0.5d	40±5
Zn2+
Pnt-Ca2+-S100B (low salt)g	0.13±0.09	−7.6±0.5	2.1±0.4d	-
Ca2+-S100B (low salt)h	0.09±0.02h	−6.2±0.1h	3.8±0.1d,h	-

^aConditions included 10 mM TES, pH 7.2, 15 mM NaCl, and 10 mM CaCl₂, at 37 °C. For the high-salt conditions, 100 mM KCl.

^bA single dissociation constant (K_d) is reported even though two sites for Pnt binding were detected by ITC for these complexes (n = 1.9±0.2). As expected, the fluorescence data also are best fit with a two-site model (n_measured = 1.8±0.2) for both the Pnt-Ca²⁺-S100B and Pnt-Zn²⁺,Ca²⁺-S100B complex. That n_measured used to fit the fluorescence data is less than or equal to n as determined from ITC is also consistent with having two noninteracting Pnt sites. Thus, the simplest model that fits the data under these conditions is that two independent Pnt binding sites exist with indistinguishable affinities.

^cThe c values range from 3 to 5 for the ITC experiments, indicating that the binding data are appropriate for evaluating K_d values.

^dAn estimate for the entropic contribution to binding (TΔS) was calculated using the dissociation constant (K_d) and the apparent molar enthalpy of association (ΔH_app) for Pnt binding.

^eConditions included 10 mM TES, pH 7.2,15 mM NaCl, 10 mM CaCl₂, and 100–500 μM ZnCl₂, at 37 °C. For the high-salt conditions, 100 mM KCl was added. For the fluorescence titrations in the presence of Zn²⁺, the Zn²⁺-to-S100B ratio did not exceed 1:1 to avoid precipitation (see Materials and Methods).

^fThe ITC data for Zn²⁺,Ca²⁺-S100B under low-salt conditions was evaluated (K_d∼64±15), but due to large amounts of precipitation at high Pnt levels, the stoichiometry was not reliably determined, and this dissociation constant is only considered an estimate. At lower protein concentrations, the fluorescence data were more reliable (K_d = 39±5; n_measured = 1.8±0.1).

^gConditions included 10 mM TES, pH 7.2,15 mM NaCl, 10 mM CaCl₂, and 100–400 μM Pnt, at 37 °C with c values ranging from 700 to 1000.

^hConditions included 10 mM TES, pH 7.2, 15 mM NaCl, and 10 mM CaCl₂, at 37 °C, and the value is from Wilder et al.⁸

In summary, the thermodynamic data presented here indicate that there are two independent Pnt binding sites with indistinguishable affinities in the Pnt–Ca²⁺–S100B and Pnt–Zn²⁺,Ca²⁺–S100B drug complexes (Table 2). We also demonstrate that Zn²⁺ had no detectable effect on Pnt binding to Ca²⁺–S100B, whereas Pnt bound to Ca²⁺–S100B slightly decreased the affinity of the protein for Zn²⁺ (<1.3-fold; Table 2).

NMR studies characterizing Pnt binding to S100B

STD experiments and chemical shift perturbations were measured by NMR to further characterize the binding of Pnt to Zn²⁺,Ca²⁺–S100B in solution; these results were then compared to similar data recorded previously for Pnt binding to Ca²⁺–S100B (Fig. 3).³¹ Only small differences were observed in a comparison of STD data for Pnt bound in the Pnt–Ca²⁺–S100B and Pnt–Zn²⁺,Ca²⁺–S100B complexes, indicating that the H2/H5, H3/H6, and α protons of Pnt were saturated by resonances of S100B in the drug binding site of both Ca²⁺–S100B and Zn²⁺,Ca²⁺–S100B (Fig. 3). That little or no STD was observed for the other protons of Pnt (β and γ) indicated that these protons likely face solvent rather than the binding pocket in both S100B–drug complexes (Fig. 3).⁴²

Fig. 3.

The binding of Pnt to Ca²⁺–S100B and Zn²⁺,Ca²⁺–S100B as monitored by NMR spectroscopy. (a) Structure of Pnt together with the nomenclature used for protons of Pnt. (b and c) STD spectra for Pnt binding to (b) Ca²⁺–S100B and to (c) Zn²⁺,Ca²⁺–S100B. (d and e) Perturbations in ¹H and ¹⁵N chemical shift values for residues in S100B upon Pnt binding to (d) Ca²⁺–S100B and to (e) Zn²⁺,Ca²⁺–S100B. (f) Differences in chemical shift perturbations when the titrations illustrated in (d) and (e) were compared (i.e., |Δδ^{Ca–S100B + Pnt}−Δδ^{Zn,Ca–S100B + Pnt}|).

Titrations of Pnt into Ca²⁺- and Ca²⁺,Zn²⁺-bound S100B were compared next by monitoring backbone ¹H–¹⁵N correlations in 2D ¹H–¹⁵N heteronuclear single quantum coherence (HSQC) NMR experiments (Fig. 3). In a titration of Pnt into Ca²⁺–S100B, the ¹H–¹⁵N HSQC correlations of S100B shifted (>30 Hz; Fig. 3) for several residues in helix 1 (A9, I11, D12, and F14), loop 2 (L40, S41, L44, and E46), and helix 4 (Q71, M74, A75, V77, S78, V80, A83, H85, E89, and H90), as previously observed.³¹ In a similar titration with Zn²⁺,Ca²⁺–S100B plus Pnt, the ¹H–¹⁵N HSQC correlations of S100B shifted (>30 Hz; Fig. 3) for residues in helix 1 (I11, V13, F14, and H15), loop 2 (L40, S41, H42, L44, and E45), and helix 4 (M74, V77, S78, M79, T82, C84, and E91). When results from these two titrations were compared under identical buffer conditions (i.e., as a ΔΔδ; Fig. 3), Pnt binding did not perturb ¹H–¹⁵N correlations for the same residues in Ca²⁺-bound and Zn²⁺,Ca²⁺-bound S100B, nor did several of these perturbations occur to the same extent. These results indicated that Pnt bound to Ca²⁺–S100B differently (i.e., different orientation) and/or affected the conformation of S100B differently than when Pnt bound to Zn²⁺,Ca²⁺–S100B.

The X-ray structure of Zn²⁺,Ca²⁺–S100B

The structural consequences of Zn²⁺ binding to Ca²⁺–S100B was important to characterize because of its potential effect on binding specific protein targets such as p53 and/or small-molecule inhibitors such as Pnt.^7,8 With this in mind, the 1.88-Å-resolution X-ray crystal structure for Zn²⁺,Ca²⁺–S100B was solved and compared to that of Ca²⁺–S100B^43,44 (Fig. 5). The final asymmetric unit consisted of 90 residues (Met0 to Glu89), 2 calcium ions, 1 zinc ion, and 51 water molecules. The biologically significant model is a dimer composed of the asymmetric unit and a crystallographic symmetry mate. Nearly all of the residues of Zn²⁺,Ca²⁺–S100B were in the most favorable region of the Ramachandran plot (95.2%) with the remaining residues (4.8%) falling into the allowed region (Table 1; Fig. 4a).

Fig. 5.

Changes in side-chain positioning upon the binding of Zn²⁺ to Ca²⁺–S100B. (a) Residues on helix 4 (blue) and helix 1′ (red) of Ca²⁺–S100B prior to binding Zn²⁺. (b) Residues on helix 4 (blue) and helix 1′ (red) of Zn²⁺,Ca²⁺–S100B that changed in position upon binding Zn²⁺. (c) Average RMSDs for the position of side-chain atoms when the Ca²⁺–S100B and Zn²⁺,Ca²⁺–S100B structures are compared. Those residues with average RMSD values greater than 1 are labeled on the bar graph.

Table 1.

Diffraction and refinement statistics

	Zn2+,Ca2+-S100B	Pnt-Ca2+-S100B	Pnt-Zn2+,Ca2+-S100B
Diffraction statistics
Space group	C2221	P41212	P41212
Cell dimensions: a, b, c (Å)	34.6, 90.5, 58.3	63.6, 63.6, 48.2	62.9, 62.9, 50.1
Cell angles: α, β, γ (°)	90, 90, 90	90, 90, 90	90, 90, 90
Resolution (Å)	45.22–1.88 (1.93–1.88)	44.95–2.15 (2.20–2.15)	44.46–1.85 (1.90–1.85)
No. of unique reflections	7120 (502)	5502 (390)	8456 (560)
Completeness (%)	96.45 (91.90)	99.98 (99.76)	98.84 (93.20)
Rsyma	0.056 (0.319)	0.081 (0.667)	0.150 (0.489)
Average I/σ	23.4 (4.3)	32.3 (4.2)	20.4 (4.2)
Multiplicity	6.6 (5.1)	12 (12.1)	12.9 (8.9)
Refinement statistics
Rcrysb(%)	22.1 (31.1)	21.1 (25.2)	19.4 (21.2)
Rfreeb(%)	26.8 (46.7)	26.6 (26.7)	24.3 (22.9)
Protein atoms	727	729	737
Water molecules	51	35	82
Non-hydrogen atoms	768	814	858
Mean B values (Å2)
Overall	49.42	35.87	40.16
Protein atoms	50.19	35.05	39.01
Water molecules	51.34	35.59	45.94
Ca2+ typical EF hand	47.00	31.72	35.76
Ca2+ S100B EF hand	46.71	35.58	35.63
Zn2+	50.11		43.17
Pnt site 1		53.14c	49.85d
Pnt site 2		43.01c	67.52d
RMSD
Bond length (Å)	0.015	0.008	0.015
Bond angles (Å)	1.543	1.135	1.484
Ramachandran plot (%)e
Most favored	95.2	95.2	96.5
Additionally allowed	4.8	4.8	3.5
PDB identification	3CR2	3CR4	3CR5

Numbers in parentheses represent the last outer shell.

^aR_sym= Σ_hklΣ_i(|I_hkl,i|-|〈I_hkl〉|)/Σ_hklΣ_iI_hkl,i, for a reflection hkl whose weighted mean over i observations is 〈I〉s.

^bR_crys and R_free = Σ_hkl||F _o|-k|F_c||/Σ_hkl||F_o|, where |F_o| is the observed structure factor amplitude and |F_c| is the calculated structure factor amplitude for the working and test sets, respectively.

^cMean B values (Ų) for Pnt in sites 1 and 2 are similar to the mean B value (38.74) for side-chain atoms of S100B involved in Pnt binding (Ile11, Asp12, His42, Cys84, Phe87, and Phe88).

^dMean B values (Ų) for Pnt in sites 1 and 2 are similar to the mean B value (45.29) for side-chain atoms of S100B involved in the Pnt binding (Val8, Ile11, Phe43, Cys84, Phe87, and Phe88).

^eFor Zn²⁺,Ca²⁺-S100B and Pnt-Ca²⁺-S100B, the calculations had 80 residues in the most favored region and 4 residues in additionally allowed regions. For Pnt-Zn²⁺,Ca²⁺-S100B, the calculations had 82 residues in the most favored region and 3 residues in additionally allowed regions.

Fig. 4.

Calcium and zinc ion coordination in the X-ray structure of Zn²⁺,Ca²⁺–S100B. (a) X-ray crystal structure of Zn²⁺,Ca²⁺–S100B shown in stereo (walleye mode) as a ribbon diagram. The locations of two Ca²⁺ ions per subunit (cyan spheres) and one Zn²⁺ ion per subunit (orange spheres) are illustrated, and subunits of the symmetric S100B homodimer are colored red and blue, respectively. (b) Stereo view of the pseudo-EF-hand (Ser18, Glu21, Asp23, Lys26, and Glu31) and the canonical EF-hand (Asp61, Asp63, Asp65, Glu67, and Glu72) calcium-binding sites of S100B together with the 2mF_o–DF_c electron density map (1.0 σ). (c) Zn²⁺ coordination in Zn²⁺,Ca²⁺–S100B is illustrated together with the 2mF_o–DF_c electron density map (1.0 σ) defining the positions of His15′/His25′ from one subunit of S100B and His85/Glu89 from the other subunit. (d) Position of Zn²⁺ coordinating residues of Zn²⁺,Ca²⁺–S100B as predicted previously from the NMR solution structure and site-directed mutagenesis.⁷,⁸

The global fold of Zn²⁺,Ca²⁺–S100B was similar to that reported previously for Ca²⁺–S100B,^43,44 the NMR structure of Zn²⁺,Ca²⁺–S100B,⁸ and several other Ca²⁺-loaded S100 proteins.^2,5 Specifically, each subunit of Zn²⁺,Ca²⁺–S100B contained four helices (helix 1, S1–G19; helix 2, K28–L40; helix 3, E49–D61; helix 4, D69–F88) and one small antiparallel β-sheet (strand 1, K26–K28; strand 2, D69–E67) with the dimer interface aligned as a symmetric X-type four-helix bundle comprising helices 1, 1′ and 4, 4′, respectively (Fig. 4a).

As with Ca²⁺–S100B,^43,44 each S100B subunit in Zn²⁺,Ca²⁺–S100B has two Ca²⁺-loaded helix–loop–helix EF hands including an “S100-type” or “pseudo” EF hand comprising helices 1 and 2 and loop 1 and a “typical” EF hand with 12 residues contributed by helices 3 and 4 and loop 3 (Fig. 4). The residues that coordinate calcium in Zn²⁺,Ca²⁺–S100B were found to be the same as those reported previously for Ca²⁺–S100B⁴⁴ with minimal differences (<0.3 Å) in coordinate distances (Table 3; Fig. 4). Specifically, calcium coordination in the S100 hand of Zn²⁺,Ca²⁺–S100B was from four backbone carbonyl oxygen atoms including Ser18 (X position), Glu21 (Y position), Asp23 (Z position), and Lys26 (−Y position); a water molecule (−X position); and the two side-chain carboxylate oxygen atoms of Glu31 (−Z position; bidentate ligand). In the typical EF hand, calcium was coordinated by side-chain carboxylate oxygen atoms of Asp61 (X-position), Asp63 (Y-position), and Asp65 (Z position); the backbone carbonyl of Glu67 (−Y position); a water molecule (−X position); and two side-chain carboxylate oxygen atoms of Glu72 (−Z position; bidentate ligand) (Fig. 4b). As discussed previously,⁴⁵ a side-chain carboxylate oxygen atom of Glu67 formed a hydrogen bond with a water molecule coorinated to calcium at the −X position of the pseudo-EF hand in Zn²⁺,Ca²⁺–S100B (Fig. 4). Importantly, the orientation of helix 3 with respect to helix 4 was found to be in the “open” position in the structures of both Ca²⁺–S100B (^X-rayΩ_3/4 =109.9; ^NMRΩ_3/4 = 106 ± 4) and Zn²⁺,Ca²⁺-loaded S100B (^X-rayΩ_3/4 = 109.4; ^NMRΩ_3/4 = 118 ± 2), as necessary for binding protein targets such as p53, hdm4, and CapZ (Figs. 1 and 4; Table 4).^8,30,43,44 A slight difference (∼9°) in the relative orientation of helices 3 and 4 was observed in the X-ray structure as compared to the NMR structure of Zn²⁺,Ca²⁺–S100B.⁸ Such a difference may reflect subtle structural differences in solution versus the crystalline state since residual dipolar couplings measured for backbone amide–proton correlations were in better agreement with the NMR family of structures (Q-factor<0.3) than with the X-ray structure (Q-factor>0.4).⁸ Nonetheless, both the NMR and X-ray structures of Ca²⁺–S100B and Zn²⁺,Ca²⁺–S100B should be useful for engineering novel inhibitors of S100B–target protein interactions.

Table 3.

Divalent metal ion and water coordinate distances

	Ca2+-S100Ba	Zn2+,Ca2+-S100B	Pnt-Ca2+-S100B	Pnt-Zn2+,Ca2+-S100B
Zn2+
His15 (Nε2)		2.0		2.0
His25 (Nε2)		1.9		2.2
His85 (Nε2)		2.0		2.0
Glu89 (Oε1 or Oε2)		2.0		1.9
Ca2+, pseudo-EF hand
Ser18 (C=O)	2.0	2.3	2.3	2.3
Glu21 (C=O)	2.3	2.6	2.3	2.3
Asp23 (C=O)	2.6	2.5	2.3	2.3
Lys26 (C=O)	2.3	2.3	2.4	2.3
Glu31 (Oε1,Oε2)	2.4, 2.6	2.3, 2.5	2.5, 2.6	2.5, 2.6
H2Ob	2.2	2.5	2.5	2.3
Ca2+, canonical EF hand
Asp61 (OΔ1 or Oδ2)	2.3	2.3	2.3	2.4
Asp63 (Oδ1 or Oδ2)	2.6	2.4	2.4	2.5
Asp65 (Oδ1 or Oδ2)	2.3	2.6	2.4	2.4
Glu67 (C=O)	2.4	2.4	2.4	2.4
Glu72 (Oε1,Oε2)	2.3, 2.5	2.4, 2.7	2.5, 2.6	2.3, 2.5
H2Oc	2.4	2.3	2.3	2.4
Ca2+-bound H2O
Asp65 (Oδ1 or Oδ2)c	2.8	2.4	2.7	2.6
Glu67 (Oε1 or Oε2)b	2.6	2.9	2.8	2.7

All distances are in angstroms and are from nitrogen, oxygen, or Ca²⁺. Distances between the two Ca²⁺ ions for the Ca²⁺-S100B, Zn²⁺,Ca²⁺-S100B, Pnt-Ca²⁺-S100B, and Pnt-Zn²⁺,Ca²⁺-S100B structures are 11.5, 11.5, 11.4, and 11.4 Å, respectively.

^aCalculated using PDB entry 1MHO.

^bH₂O ligands to Ca²⁺ from the pseudo-EF hand.

^cH₂O ligands to Ca²⁺ from the canonical EF hand.

Table 4.

Interhelical angles and distances

	Interhelical angle (°)								Interhelical distance (Å)
	I to II	I to III	I to IV	II to III	II to IV	III to IV	I to I′	IV to IV′	I to I′	IV to IV′
Ca2+-S100Ba,b	140	−121	125	99	−32	110	−152	150	13.7	10.0
Zn2+,Ca2+-S100Ba	142	−120	126	97	−32	110	−151	148	13.7	10.2
Pnt-Ca2+-S100Ba	139	−121	125	99	−32	109	−152	146	13.6	10.1
Pnt-Zn2+,Ca2+-S100Ba	139	−122	125	98	−32	109	−151	146	13.6	10.0

Interhelical angles were calculated using INTERHLX software (K. Yap, University of Toronto) except for I to I′ and IV to IV′ where the interhelical angles and distances were determined using IHA v1.4 (S.M. Gagne, SMG Software Inc.).

^aThe residues defining each helix of S100B were as follows: helix I, residues 2–18; helix II, residues 29–40; helix III, residues 50–62; and helix IV, residues 70–82.

^bCalculated using PDB entry 1MHO.

Zn²⁺ coordination in Zn²⁺,Ca²⁺–S100B

Consistent with site-directed mutagenesis and an NMR structure reported previously,^7,8 the X-ray structure of Zn²⁺,Ca²⁺–S100B has two Zn²⁺ ions per dimer coordinated in a symmetric manner by His15/His25 of one S100B subunit and His85/Glu89 of the other subunit (Fig. 4)^†. The coordination of Zn²⁺ involves a carboxylate oxygen of Glu89 and nitrogen atoms from the three His residues arranged as a slightly distorted tetrahedral with metal–ligand distances ranging from 1.94 to 2.02 Å and coordinate angles (i.e., O–Zn²⁺–N and N–Zn²⁺–N) ranging from 100° to 115° (Fig. 4; Table 3). This coordination geometry is observed in numerous other Zn²⁺ binding proteins including several S100 proteins.^46,47 For the side chain of Glu89, the Zn²⁺ ion is located in the syn position with the second carboxylate oxygen atom at a distance of 2.74 Å from Zn²⁺. This second oxygen atom of the Glu89 side chain also contributes somewhat to metal ion coordination as a fifth ligand.

Several residues near the Zn²⁺ binding site (Val13, His15, Gln16, His25, Phe87, and Phe88) were found to have different side-chain orientations (>1.0 Å RMSD) in Zn²⁺,Ca²⁺–S100B when compared to Ca²⁺–S100B (Fig. 5). The one other side chain positioned differently was Gln71, which forms a hydrogen bond to Thr82 in the Ca²⁺-bound structure, but in Zn²⁺,Ca²⁺–S100B, it is on the surface of the protein and distant from Thr82. Another difference in the two structures is that the side chain for one of the residues that coordinates Zn²⁺, Glu89, could not be traced in the electron density map of Ca²⁺–S100B⁴⁴; however, this side chain was readily observed when Zn²⁺ was bound (Figs. 4 and 5). Together, these data indicated that the movement and/or stabilization of residues at or nearby the Zn²⁺ binding site was required for Zn²⁺ to bind Ca²⁺-loaded S100B (Fig. 5).

Structures of Pnt bound to Ca²⁺–S100B and Zn²⁺,Ca²⁺–S100B

The X-ray structures of Pnt bound to Ca²⁺–S100B and Zn²⁺,Ca²⁺–S100B were solved at 2.15 and 1.85 Å resolution, respectively (Fig. 6; Table 1). The asymmetric unit for the Pnt–Ca²⁺–S100B structure included 90 residues (Met0 to Glu89), 2 calcium ions, 2 Pnt molecules, and 35 water molecules (Fig. 6; Table 1). It was evident from visual inspection of the 2mF_o–DF_c and mF_o–DF_c electron density maps that the only way to accurately model both Pnt molecules (site 1 and site 2) was to place them in two symmetrical and overlapping orientations with each site having partial occupancies (site 1=0.5, site 2=0.5). The occupancies and temperature factors for the two Pnt orientations at site 1 and site 2 were constrained to be identical owing to the fact that the Pnt molecules reside on a crystallographic 2-fold axis. Thus, the interactions between Pnt and each S100B dimer were identical across the 2-fold axis. Furthermore, the electron density for Pnt bound in sites 1 and 2 was well defined (Fig. 6) with reasonable mean B values (Ų) for bound Pnt molecules (site 1=53.14, site 2=43.01) that were comparable to those of side-chain atoms defining the drug binding sites [Table 1; mean B value (Ų) for side-chain atoms of Asp12, His42, Cys84, Phe87, and Phe88=38.74].

Fig. 6.

The X-ray structures of Pnt–Ca²⁺–S100B and Pnt–Zn²⁺,Ca²⁺–S100B. (a) The 2mF_o–DF_c electron density maps (1.0 σ) for two Pnt molecules bound to Ca²⁺–S100B (left: site 1, right: site 2). Symmetrical and overlapping orientations are shown for Pnt in each site (in orange and cyan) because of their locations at the crystallographic 2-fold axis. (b) Ribbon and surface diagram of Pnt–Ca²⁺–S100B illustrating the location of Pnt (Pnt, orange; nitrogen, blue; oxygen, red) bound to sites 1 and 2 and Ca²⁺ (cyan). (c) Residues of S100B that are within 4 Å of Pnt bound to site 1 in the Pnt–Ca²⁺–S100B complex. (d) Residues of S100B that are within 4 Å of Pnt bound to site 2 in the Pnt–Ca²⁺–S100B complex. (e) The 2mF_o–DF_c electron density maps (1.0 σ) for two Pnt molecules bound to Zn²⁺,Ca²⁺–S100B (left: site 1, right: site 2). Symmetrical and overlapping orientations are shown for Pnt in each site (in orange and cyan) because of their locations at the crystallographic 2-fold axis. (f) Ribbon and surface diagram of Pnt–Zn²⁺,Ca²⁺–S100B illustrating the location of Pnt (Pnt, orange; nitrogen, blue; oxygen, red) bound to sites 1 and 2, Ca²⁺ (cyan), and Zn²⁺ (orange). (g) Residues of S100B that are within 4 Å of Pnt bound to site 1 in the Pnt–Zn²⁺,Ca²⁺–S100B complex. (h) Residues of S100B that are within 4 Å of Pnt bound to site 2 in the Pnt–Zn²⁺,Ca²⁺–S100B complex.

The asymmetric unit for Pnt–Zn²⁺,Ca²⁺–S100B had 91 residues (Met0 to His90), 2 calcium ions, 1 zinc ion, 2 Pnt molecules, and 82 water molecules (Fig. 6; Table 1). As for Pnt–Ca²⁺–S100B, the 2mF_o–DF_c and mF_o–DF_c electron density maps for Pnt bound to Zn²⁺,Ca²⁺–S100B were also on a crystallographic 2-fold axis and modeled with partial occupancies (site 1=0.3, site 2=0.5). The electron density observed for Pnt bound in sites 1 and 2 of Zn²⁺,Ca²⁺–S100B was also well defined (Fig. 6) with mean B values (Ų) for the bound drug molecules (site 1=67.52, site 2=49.85) that were only slightly higher than side-chain atoms of residues interacting with the drug [Table 1; mean B value (Ų) for side-chain atoms of Val8, Ile11, Phe43, Cys84, and Phe87=45.29]. As with the structure of Zn²⁺,Ca²⁺–S100B, nearly all of the residues in Pnt–Ca²⁺–S100B (95.2%) and Pnt–Zn²⁺,Ca²⁺–S100B (96.5%) were in the most favorable region of the Ramachandran plot, with the remaining 4.8% and 3.5% of the residues in the allowed region, respectively (Table 1).

Description of the Pnt–Ca²⁺–S100B complex

The global fold, interhelical angles, and calcium-ion coordination of S100B in the Pnt–Ca²⁺–S100B complex were very similar to that reported previously for Ca²⁺–S100B^43,44 (Fig. 6; Tables 3-5). However, specific side chains in Ca²⁺–S100B, including those in helix 1 (Glu2, Lys5, Asp12, Val13, and Gln16), loop 2 (Glu45, Glu49, and Gln50), helix 3 (Glu51 and Lys55), and helix 4 (Gln71, Cys84, Glu86, Phe87, and Phe88), significantly changed position (RMSD>1.0 Å) upon the addition of Pnt (Fig. 7)^‡. These changes in S100B structure correspond well with the interactions of two Pnt molecules with S100B observed in the 2mF_o–DF_c and mF_o–DF_c electron density maps for Pnt–Ca²⁺–S100B (arbitrarily termed sites 1 and 2). One Pnt binding site was adjacent to a p53 peptide binding site described previously (site 1; Fig. 6).^27,31 The most notable S100B–Pnt interactions in site 1 resulted from hydrophobic stacking of the benzamidine rings of the drug with aromatic side-chain moieties of the protein including Phe88 from the C-terminal loop to ring 1 of Pnt and His42 from the “hinge” region to ring 2 (Fig. 6). For this interaction to occur, the side chain of Phe88 reoriented upon binding Pnt (Fig. 7). The backbone carbonyls of both His42 and Phe43 also lined this Pnt binding site of Ca²⁺–S100B (<4.0 Å). Importantly, the binding of Pnt to site 1 is consistent with solution NMR data, including an unambiguous intermolecular nuclear Overhauser enhancement between protons of Pnt and Phe88 of S100B.³¹

Table 5.

EF-hand angles determined using the vector geometry mapping (VGM) method

	N-terminal coordinate of second helix			Θ (°)	Φ (°)	Ω (°)
Pseudo-EF hand
Ca2+-S100Ba,b	8.2	−0.2	−6.8	50	87	109
Zn2+,Ca2+-S100Ba	8.2	0.0	−6.4	50	87	108
Pnt-Ca2+-S100Ba	8.3	−0.1	−6.5	51	91	107
Pnt-Zn2+,Ca2+-S100Ba	8.2	0.0	−6.7	51	90	105
Typical EF hand
Ca2+-S100Ba,b	8.9	0.3	−6.6	74	102	75
Zn2+,Ca2+-S100Ba	8.8	−0.2	−6.8	76	104	75
Pnt-Ca2+-S100Ba	8.8	0.4	−6.5	74	104	69
Pnt-Zn2+,Ca2+-S100Ba	8.8	0.3	−6.5	75	105	70

VGM angles were calculated by way of VGM software (K. Yap, University of Toronto) using the helices and structures indicated.

^aThe residues used for the VGM calculations were as follows: helix I, residues 10–17; helix II, residues 29–39; helix III, residues 53–60; and helix IV, residues 70–80.

^bCalculated using PDB entry 1MHO.

Fig. 7.

Changes in side-chain positioning upon the binding of Pnt to Ca²⁺–S100B or to Zn²⁺,Ca²⁺–S100B. (a) Average RMSDs for the position of side-chain atoms when the X-ray structures of Ca²⁺–S100B and Pnt–Ca²⁺–S100B are compared. Those residues that have an average RMSD value greater than 1 are labeled, and residues marked with an asterisk (*) are involved with the crystallographic lattice. (b) Residues on helix 4 (blue) and helix 1′ (red) of Ca²⁺–S100B in the absence and presence of Pnt (orange) illustrating the change in side-chain positioning upon drug binding. (c) Average RMSDs for the position of side-chain atoms when the X-ray structures of Zn²⁺,Ca²⁺–S100B and Pnt–Zn²⁺,Ca²⁺–S100B are compared. Those residues that have an average RMSD value greater than 1 are labeled, and residues marked with an asterisk (*) are involved with the crystallographic lattice. (d) Residues on helix 4 (blue) and helix 1′ (red) of Zn²⁺,Ca²⁺–S100B in the absence and presence of Pnt (orange) illustrating the change in side-chain positioning upon drug binding.

The second Pnt binding site (site 2) of Pnt–Ca²⁺–S100B was located at the dimer interface in a pocket adjacent to residues that coordinate Zn²⁺ (Fig. 6). This Pnt binding site was best defined by a bidentate hydrogen bond from the two carboxylate oxygen atoms of Asp12 to protons of the −NH₂ moieties of one benzamidine ring (2.8 and 3.1 Å; Fig. 6). The second benzamidine ring of Pnt was found to be stacked with the aromatic side chain of Phe87 in helix 4′ of the other subunit of S100B. In addition, a potential weak hydrogen bond between the backbone carbonyl of His85 and an −NH₂ moiety of the second benzamidine ring of Pnt was identified (3.3 Å; Fig. 6). As found for site 1, side-chain moieties that interact with the drug in site 2 (Asp12 and Phe87) rotated into position upon binding Pnt (Fig. 7). Likewise, side-chain atoms of Ile11 and Cys84 and the backbone carbonyl of Cys84 were also proximal to Pnt in site 2 (<4.0 Å), which may contribute to binding.

The Pnt–Zn²⁺,Ca²⁺–S100B complex

When the X-ray structures of Zn²⁺,Ca²⁺–S100B and Pnt–Zn²⁺,Ca²⁺–S100B were compared (^mainchainRMSD=0.52), the global fold, interhelical angles, and calciumion coordination were nearly identical (Figs. 4 and 6; Tables 3-5). The two Pnt binding sites on Pnt–Zn²⁺,Ca²⁺–S100B were evident in the 2mF_o–DF_c and mF_o−DF_c electron density maps of Pnt–Zn²⁺,Ca²⁺–S100B, and the drug binding sites in this complex were located in positions analogous to sites 1 and 2 in Pnt–Ca²⁺–S100B (Fig. 6). As expected, several side chains changed position when Pnt bound to Zn²⁺,Ca²⁺–S100B (RMSDN>1.0 Å) including residues in helix 1 (Val8, Asp12, Val13, and Gln16), loop 2 (Lys48), helix 3 (Val53 and Glu58), and helix 4 (Met74, Met79, Cys84, Glu86, and Phe88)^‡. However, the residues that changed position in Zn²⁺,Ca²⁺–S100B upon the addition of Pnt (Fig. 7d) and/or the magnitude of such changes were notably different when compared to the changes observed in side-chain positioning between Pnt–Ca²⁺–S100B and Ca²⁺–S100B (Fig. 7b and d). Furthermore, the conformation and orientation of Pnt in both binding sites of Pnt–Zn²⁺,Ca²⁺–S100B were different when compared to the X-ray structure of Pnt–Ca²⁺–S100B (Figs. 6 and 7). These results were in agreement with the observation that Zn²⁺ binding to Ca²⁺–S100B reoriented several residues in helices 1 and 4 that interacted with Pnt binding in the absence of Zn²⁺ (Figs. 5 and 7). That Pnt could adapt to this Zn²⁺-induced conformational change and still bind Zn²⁺,Ca²⁺–S100B was a novel and unanticipated finding. Supporting this result, small differences were also found in solution when STD data were compared for Pnt bound to Ca²⁺–S100B versus Zn²⁺,Ca²⁺–S100B (<2-fold changes; relative to the largest STD for the α proton; Fig. 3). While the differences in the STD data are not easily quantified,⁴⁸ they are consistent with having different Pnt orientations as was observed when the X-ray structures for the two drug complexes are compared.

The Pnt molecule in site 1 of Pnt–Zn²⁺,Ca²⁺–S100B was relatively extended with little or no bending or kink (Figs. 6 and 7) unlike Pnt bound to site 1 of Pnt–Ca²⁺–S100B. Furthermore, the S100B–drug interaction in site 1 involved only one of the two benzamidine rings. This interaction was dominated by π electron stacking⁴⁹ in which one benzamidine ring of Pnt was sandwiched in between two aromatic side chains of Zn²⁺,Ca²⁺–S100B (i.e., Phe43 and Phe87; Figs. 6 and 7). Backbone carbonyl oxygen atoms lining the site (<4.0 Å) include those from residues His42, Phe43, and Ala83, which may also contribute to drug binding. In site 2, the Pnt molecule was relatively kinked when compared to drug bound to the analogous site in Pnt–Ca²⁺–S100B. In addition, Pnt no longer formed a bidentate hydrogen bond with the carboxylate oxygen atoms of Asp12. Rather, a new hydrophobic ring stacking interaction was observed between the benzamidine ring and Phe88, which was rotated into this site (from site 1) upon binding Zn²⁺. In addition, a potential weak hydrogen bond between Cys84 and an −NH₂ moiety of the benzamidine ring of Pnt was identified (3.2 Å; Fig. 6). Side-chain atoms from other hydrophobic residues are also near Pnt in site 2 of Zn²⁺,Ca²⁺–S100B including those of Val8 and Ile11. Lastly, the differences observed in the conformation and orientation of Pnt in sites 1 and 2 in the two X-ray structures were consistent with solution NMR studies, which showed differences in chemical shift perturbations when titrations of Pnt into Ca²⁺–S100B versus Zn²⁺,Ca²⁺–S100B were compared (i.e., as a ΔΔδ; Fig. 4).

Discussion

S100B functions as a calcium-activated “switch” (Fig. 1) that undergoes a conformational change as necessary to bind and inhibit the tumor suppressor protein p53.^2,30 In addition, S100B binds Zn²⁺ (one per subunit) at a site separate from the two EF-hand calcium-binding loops. It has been proposed that Zn²⁺ binding might be important for the biological activity of S100B.^7,8,10 Because solution NMR studies did not precisely define the coordination geometry of Zn²⁺ bound to S100B, we solved the crystal structure of Zn²⁺,Ca²⁺–S100B (Fig. 4).^7,8 We also obtained details regarding a protein conformational change that was induced by Zn²⁺ binding to Ca²⁺-loaded S100B (Fig. 5), which could be important to consider when inhibiting S100B–target protein interactions. Thus, when the X-ray structures of Pnt–Ca²⁺–S100B and Pnt–Zn²⁺,Ca²⁺–S100B were compared (Figs. 6 and 7), we found that two drug molecules (per S100B subunit) bound sites that were similarly located in the two structures; however, the conformation and orientation of Pnt bound in these two sites were very different (±Zn²⁺). Structural data such as these can now be used to design/synthesize higher-affinity inhibitors that target specific divalent metal ion-bound states of S100B as might be necessary for melanoma where Ca²⁺ and Zn²⁺ levels are unregulated.

Zn²⁺ binding to S100 proteins

It was predicted that the Zn²⁺ binding site in S100B would be composed of residues from both subunits (subunit 1: His15/His25, subunit 2: His85′/Glu89′).^7,8 This coordination scheme was confirmed by our X-ray crystallographic studies (Fig. 4). When the Zn²⁺ site of S100B [Fig. 4; Protein Data Bank (PDB) code: 3CR2] was compared to other S100 proteins (using S100B numbering), a structurally conserved Zn²⁺/Cu²⁺ binding site, distinct from the two EF-hand Ca²⁺ binding sites, was also observed for S100A7 (Zn²⁺: His17/Asp24/His86′/His90′; PDB code: 2PSR)⁴⁶ and S100A12 (Cu²⁺: His15/Asp25/His85′/His89′; PDB code: 1ODB)^50,51 (Fig. 8). As found here for S100B, four of the five ligands (one side-chain carboxylate oxygen; three His nitrogen atoms) from S100A7 and S100A12 coordinate divalent metal ion in a slightly distorted tetrahedral geometry with similar metal ion–ligand distances (∼2 Å). The fifth ligand is at a longer metal ion–ligand distance (>2.5 Å) and is provided by the second side-chain carboxylate oxygen atom of Glu89 in subunit 2 of S100B. In contrast, for S100A7 and S100A12, this more weakly coordinating oxygen ligand was from an aspartate side chain in subunit 1 (S100A7: Asp24, S100A12: Asp25; Fig. 8).

Fig. 8.

> A comparison of Zn²⁺ or Cu²⁺ coordination in S100B (Zn²⁺), S100A7 (Zn²⁺), and S100A12 (Cu²⁺). Shown is an overlay of the Zn²⁺/Cu²⁺ligands for S100B (green, PDB code: 3CR2), S100A7⁴⁶ (cyan, PDB code: 2PSR), and S100A12⁵⁰ (violet, PDB code: 1ODB) as determined by X-ray crystallography.

Several other S100 proteins also bind Zn²⁺ including S100A1,⁵² S100A2,⁵³ S100A3,⁵⁴ S100A5,⁵⁵ S100A6,^56,57 and S100P (low affinity),⁵⁸ but these proteins likely coordinate Zn²⁺ differently than observed here for S100B, since there is little sequence homology for the Zn²⁺-liganding residues when compared to S100B, S100A7, or S100A12.⁴⁷ For S100A8/A9, it was predicted that the heterodimer coordinates Zn²⁺ in a manner similar to that of S100B, S100A7, and S100A12,⁵⁹ and as found here for S100B, Zn²⁺ binding induced a conformational change in the S100A8/A9 heterodimer complex that was distinct from the Ca²⁺-dependent change.⁵⁹ Nonetheless, it still remains necessary to directly determine the high-resolution structures for each of these other Zn²⁺ binding S100 proteins to accurately characterize their coordination geometry and structural changes that occur upon binding Zn²⁺.

The effects of Zn²⁺ on Ca²⁺ and target protein binding to S100B

The biological significance of Zn²⁺ and/or Cu²⁺ binding has not yet been fully elucidated for any S100 protein despite the fact that several of these proteins contain structurally homologous Zn²⁺/Cu²⁺ binding sites (Fig. 8). It was proposed that antimicrobial activity associated with S100A7 and S100A8/A9 proteins resulted from sequestering free Zn²⁺ and inhibiting Zn²⁺-activated enzymes in vivo.^59-61 However, the ability of S100 proteins to compete for free Zn²⁺ is unlikely due to the relatively high dissociation constant of Zn²⁺ from S100 proteins (i.e., ∼100 nM; Table 2) as compared with most Zn²⁺-activated enzymes.^62,63

It is more likely that Zn²⁺ binding contributes to S100 function by modulating their Ca²⁺ and/or target protein interactions as was observed in vitro for several S100 proteins.⁴⁷ For example, the affinity of S100B for Ca²⁺ is increased by nearly an order of magnitude in the presence of Zn²⁺.^9,10 One possibility for this observation was that Zn2+ binding caused a conformational change such that Ca²⁺ ions in the typical and pseudo-EF hands were more optimally coordinated. Such an explanation was ruled out here since no change in calcium coordination in S100B was observed in either EF-hand domain when the crystal structures of Ca²⁺–S100B and Zn²⁺,Ca²⁺–S100B were compared (Fig. 5; Tables 3-5). Another possibility is that perhaps Zn²⁺ binding to apo–S100B contributes to increased Ca²⁺ binding; however, this and several other alternate mechanistic explanation(s) for how Zn²⁺ binding increases Ca²⁺ binding affinity must now be explored (i.e., effects on K_on, conformational equilibriums, protein dynamics, etc.).

The binding of Zn²⁺ to Ca²⁺–S100B has also been reported to increase the affinity of several protein targets for S100B including CapZ,¹³ the AHNAK protein,¹⁵ and the tau protein.¹⁴ Likewise, target protein interactions involving S100A6 and S100A11 were of higher affinity when these S100 proteins were bound to Zn²⁺.¹² For S100A1, Zn²⁺ binding provided more than a 30-fold enhancement in binding the protein target twitchen kinase⁵² and raised the question of whether Zn²⁺ binding modulates S100 protein function as part of a more general mechanism in vivo for this family of proteins.^4,47 Close examination of conformational change that Zn²⁺ induced in Ca²⁺–S100B (Fig. 5) together with the availability of structural data for several target peptide complexes of S100B can provide some insight for how this might occur. For example, the TRTK-12 peptide binds Zn²⁺,Ca²⁺–S100B an order of magnitude more tightly than Ca²⁺–S100B.¹³ In this S100B–peptide complex, several residues that interact directly with TRTK-12 including Phe87 and Phe88,⁶⁴ were found to be reoriented when Zn²⁺ bound to Ca²⁺–S100B (Fig. 5). Perhaps this Zn²⁺-induced conformational change in Ca²⁺–S100B at or nearby the TRTK-12 binding site introduced additional and/or more energetically favorable interaction(s) with the TRTK-12 peptide, although such an explanation needs to be verified by directly examining the structure of Zn²⁺,Ca²⁺–S100B bound to a target protein/peptide.

The structures of Pnt bound to Ca²⁺–-S100B (±Zn²⁺)

As part of an effort to inhibit S100B, the Food and Drug Administration-approved drug Pnt was characterized by NMR and found to interact with aromatic residues in helix 4 and the C-terminal loop of Ca²⁺-loaded S100B.³¹ To pursue this S100B–drug interaction further, we determined the X-ray crystal structures of Pnt bound to Ca²⁺- and Zn²⁺, Ca²⁺-bound S100B in this study. In both complexes, it was discovered that S100B bound two Pnt molecules per S100B subunit with one molecule located adjacent to the p53 peptide binding site (termed site 1), as predicted previously.³¹ The second Pnt molecule was found to bind to a separate site (site 2), adjacent to a well-characterized Zn²⁺ binding site on S100B.⁸ Pnt bound to site 2 was found to interact with residues in helix 1 of one subunit and helix 4′ of the second subunit but had no direct interaction with Zn²⁺. Nonetheless, a change in the relative orientation of the drug in both sites 1 and 2 was observed when Zn²⁺ was bound (Fig. 6).

Next, the dependence of calcium on Pnt binding to S100B was examined in detail by comparing the three-dimensional structures of apo–S100B, Ca2+–S100B, Pnt–Ca²⁺–S100B, and Pnt–Zn²⁺,Ca2+–S100B (Fig. 9). As observed previously,⁴³ the relative orientation of helix 3, helix 4, and the hinge region (loop 2) is quite different in the apo- and Ca²⁺-bound states of S100B (Figs. 1 and 9), which could explain the calcium-dependent binding of Pnt to sites 1 and 2. Furthermore, in the absence of calcium, a large portion of both Pnt sites 1 and 2 was occluded by several side chains of S100B including Glu46 and His90 (Fig. 9). As a result, two residues (His42 and Phe88) that approach each other in Ca²⁺–S100B (8 Å) and then ring stack with Pnt in site 1 of the Pnt–Ca²⁺–S100B complex are completely separated in the apo state (>16 Å; Fig. 9). In site 2, additional changes are observed upon the addition of calcium. Specifically, in the apo structure, His90 is in a location that occludes much of site 2, and Phe87, another important residue for binding Pnt in site 2, is ring stacked with Phe88 and unavailable to interact with the drug. Upon binding Ca , the position of His90 becomes uncertain, likely due to dynamic properties of the protein, and a drug-binding pocket appears, which includes Asp12, a residue that forms a bidentate hydrogen bond with the amidine moiety of Pnt. Ring stacking between Phe87 and Phe88 is also disrupted when Pnt binds Ca²⁺–S100B such that Phe88 interacts with Pnt in site 1 and Phe87 interacts with Pnt in site 2. While additional protein conformational changes occur upon the addition of Zn²⁺, Pnt molecules are remarkably adaptive and still bind both sites 1 and 2 with an affinity similar to that when only Ca²⁺ was present (Figs. 6 and 9; Table 2).

Fig. 9.

The calcium dependence of Pnt binding to S100B. Shown are ribbon/surface diagrams of apo–S100B (NMR, PDB code: 1B4C; top left), Ca²⁺–S100B (X-ray, PDB code: 1MHO; top right), Pnt–Ca²⁺–S100B (X-ray, PDB code: 3CR4; bottom right), and Pnt–Zn²⁺,Ca²⁺–S100B (X-ray, PDB code: 3CR5; bottom left). A hydrogen bond between His90 and Glu46 (red) that occludes both Pnt binding sites (residues in sites 1 and 2 are colored green) is observed in apo–S100B (top left). Upon binding Ca²⁺ (top right), helix 3 reorients (see Fig. 1) and Glu46 is rotated out to the surface of the protein and His90 becomes dynamic and unobservable. Residues involved in Pnt binding two sites on S100B (highlighted in green) start to become localized as necessary for formation of the Pnt–Ca²⁺–S100B complex (lower right). Upon binding Zn²⁺,a second conformational change occurs, and surprisingly, Pnt is still able to bind both sites 1 and 2 (see also Figs. 6 and 7).

Summary

Overall, a comparison of the X-ray crystal structures of Ca²⁺–S100B and Zn²⁺,Ca²⁺–S100B showed for the first time that Zn²⁺ binding had no discernable effect on Ca²⁺ ion coordination in either the pseudo- or the typical EF-hand calcium-binding domains. Although the binding of Zn²⁺ did induce a conformational change in Ca²⁺–S100B that affected the protein's structure in regions where target proteins bind (i.e., for residues at or nearby Phe87 and Phe88 in helix 4). These differences in S100B structure, in turn, affected the conformation and orientation of two Pnt molecules bound to analogous sites on Ca²⁺–S100B and Zn²⁺,Ca²⁺–S100B. Together, the structural studies presented here will likely impact the design of new S100B inhibitors, which could have therapeutic value for the treatment of cancers with elevated S100B such as malignant melanoma.

Materials and Methods

Materials

All chemicals and reagents were American Chemical Society grade or higher and typically purchased from Sigma-Aldrich unless otherwise indicated. ¹⁵NH₄Cl was purchased from Cambridge Isotope Laboratories (Andover, MA). All buffers were passed through Chelex-100 resin to remove trace metals prior to use.

Bacterial expression and purification of the wild-type S100B

Recombinant S100B protein (rat, bovine) was expressed in Escherichia coli [HMS174(DE3) strain] and purified as previously described.⁴³ Yields of S100B protein were typically 20–30 mg of purified protein per liter of bacterial culture. For NMR experiments, S100B protein was prepared using defined media that included ¹⁵N-labeled NH₄Cl as the only nitrogen source.⁴³

NMR spectroscopy

Purified ¹⁵N-labeled S100B was dialyzed against 0.1 mM 2-{[2-hydroxy-1,1-bis(hydroxymethyl)ethyl] amino}ethanesulfonic acid (Tes), pH 7.2, and 0.05 mM DTT; lyophilized; and hydrated in a small aliquot of ddH₂O (deionized and doubly distilled H₂O) and stored at −80 or −20 °C. The Ca²⁺-loaded S100B–Pnt NMR sample was prepared in a manner similar to that previously described³¹ and contained 0.5 mM S100B subunit concentration, 0.5–0.75 mM Pnt, 0.34 mM NaN₃, 15 mM NaCl, 0–5% d₆-dimethyl sulfoxide, 10–15 mM CaCl₂, 10% D₂O, and 10 mM Tes buffer and adjusted to pH 7.2 with HCl. As described previously, 0.7 mM ZnAc was added to Ca²⁺–S100B, and the resonance assignments determined previously for Zn²⁺,Ca²⁺–S100B were confirmed.⁷ These chemical shift assignments were then used as a starting point during titrations with Pnt to assign the 2D ¹H–¹⁵N HSQC spectra of S100B in the Pnt–Zn²⁺, Ca²⁺–S100B complex.⁶⁵

Group epitope mapping via STD NMR was completed for the Pnt–Zn²⁺,Ca²⁺–S100B complex in a manner similar to that described previously for Pnt–Ca²⁺–S100B.⁶⁶ Specifically, during the 2-s presaturation pulse, the onresonance irradiation of the protein was performed at a chemical shift of −0.4 ppm and the off-resonance irradiation was applied at 30 ppm, where no protein signals were present. As a control, the STD experiments were collected in the absence of a T_1ρ filter, and as expected, the 1D spectrum of holo–S100B was fully restored. The final sample contained either 60 μM ZnAc or no ZnAc and 50 μM S100B, 500 μM Pnt, 10 mM CaCl₂, 5% d₆-dimethyl sulfoxide, 99.98% D₂O, 15 mM NaCl, 100 mM KCl, 0.34 mM NaN₃, and 10 mM Tris-d₁₁, pH 7.2. The STD data were collected at 25 °C to achieve more efficient saturation of the protein. Proton assignments of Pnt were confirmed in the above conditions in the absence and presence of S100B using 2D total correlated spectroscopy experiments.⁶⁷

HSQC NMR data were collected at 37 °C with a Bruker Avance III 600 (600.13 MHz for protons) or with an Avance 800 US2 (800.27 MHz for protons) NMRspectrometer both equipped with pulsed-field gradients, four frequency channels, and triple-resonance, z-axis gradient cryogenic probes.⁶⁵ Data were processed with NMRPipe,⁶⁸ and proton chemical shifts were reported with respect to the H₂O or HDO signal taken as 4.658 ppm relative to external trimethylsilylpropionic acid (0.0 ppm). The ¹⁵N chemical shifts were indirectly referenced as previously described using the following ratio of the zero-point frequency: 0.10132905 for ¹⁵N to ¹H.^69-71

Thermodynamic binding studies

The calcium-dependent interaction between S100B and Pnt was monitored in titrations of S100B into Pnt (100 μM) and by measuring increases in Pnt fluorescence intensity at 345 nm (λ_ex=262 and 299 nm). The fluorescence data were collected on a Varian Cary Eclipse fluorescence spectrophotometer with the temperature maintained at 37 °C using a circulating constant-temperature bath. All measurements were performed in quartz cuvettes with buffer containing 10 mM Tes, pH 7.2, 15 mM NaCl, 0–100 mM KCl, 0–10 mM CaCl₂, and 0–500 μM ZnAc. Binding of Pnt to Ca²⁺–S100B and Zn²⁺,Ca²⁺–S100B was also examined via ITC with a VP-ITC titration microcalorimeter (MicroCal, Inc., Northampton, MA), as described previously.⁷ For ITC, all S100B and Pnt solutions were degassed under vacuum for 5 min and equilibrated at 37 °C prior to use. The reference cell contained ddH₂O, and the sample cell (1.4 mL) contained 10–50 mM Tes at pH 7.2, 15 mM NaCl, 0–100 mM KCl, 10 mM CaCl₂, 0–100 μM ZnCl₂, and 0.1 mM S100B. Upon equilibration, a 2.5- to 5-mM Pnt solution prepared in the same buffer as above (without S100B) was injected in 5-μL aliquots with a 300-s interval between each injection. In all titrations containing Zn²⁺, the Zn²⁺-to-S100B subunit ratio did not exceed 1:1 to avoid precipitation. The resulting titration curves were corrected for the protein-free buffer control and analyzed using the Origin for ITC software supplied by MicroCal, Inc.

Protein crystallization

Bovine S100B protein was dialyzed into buffer (0.1 mM Tes, pH 7.2, and 0.05 mM DTT), lyophilized, dissolved in ddH₂O to 80–100 mg/mL (∼8–10 mM subunit concentration), and stored frozen. Diffraction quality crystals for the Zn²⁺,Ca²⁺–S100B complex were obtained by sitting drop vapor diffusion at 22 °C by mixing 2 μL of S100B protein (40 mg/mL S100B, 7.5 mM CaCl₂, 3.8 mM ZnAc₂, and 20 mM cacodylate buffer, pH 7.2) with 2 μL of reservoir solution [7.5 mM CaCl₂, 0.1 M cacodylate buffer, pH 6.5, and 28% polyethylene glycol (PEG) 3350] and equilibrating for 2–3 days. After crystals formed, they were cryoprotected in a harvest solution (7.5 mM CaCl₂, 3.8 mM ZnAc₂, 0.1 M cacodylate buffer, pH 6.5, 30% PEG3350, and 5% glycerol) for 30–60 s and then flashcooled in liquid nitrogen. The Pnt–Ca²⁺–S100B and Pnt–Ca²⁺,Zn²⁺–S100B complexes were crystallized in a similar manner using complex-specific protein buffers (Pnt–Ca²⁺–S100B: 40 mg/mL S100B, 7.5 mM CaCl₂, 3.8 mM Pnt, and 20 mM cacodylate buffer, pH 7.2; Pnt–Ca²⁺,Zn²⁺–S100B: 40 mg/mL S100B, 7.5 mM CaCl₂, 3.8 mM ZnAc₂, 3.8 mM Pnt, and 20 mM cacodylate buffer, pH 7.2), reservoir solutions (Pnt–Ca²⁺–S100B: 7.5 mM CaCl₂, 0.1 M cacodylate buffer, pH 7.8, and 26% PEG3350; Pnt–Ca²⁺,Zn²⁺–S100B: 7.5mMCaCl₂, 0.1Mcacodylate buffer, pH 7.6, and 22% PEG3350), and harvest solutions (Pnt–Ca²⁺–S100B: 7.5 mM CaCl₂, 3.8 mM Pnt, 0.1 M cacodylate buffer, pH 7.8, 28% PEG3350, and 5% glycerol; Pnt–Ca²⁺,Zn²⁺–S100B: 7.5 mM CaCl₂, 3.8 mM ZnAc₂, 3.8 mM Pnt, 0.1 M cacodylate buffer, pH 7.6, 24% PEG3350, and 5% glycerol). Space groups and unit cell parameters are given in Table 1. Each of the above crystal forms had one S100B subunit in the asymmetric unit.

X-ray data collection, model building, and refinement

X-ray data for the Zn²⁺,Ca²⁺–S100B and Pnt–Zn²⁺,Ca²⁺–S100B crystals were collected remotely at the 9-1 beamline (Stanford Synchrotron Radiation Laboratory, Menlo Park, CA) using an ADSC Quantum-315R CCD detector (Area Detector Systems Corporation, Poway, CA). X-ray data for Pnt–Ca²⁺–S100B were collected at 100 K using an in-house X-ray generator (MSC Micromax 7; Rigaku Texas, USA) and a Raxis-4++ image plate detector (Rigaku Texas). The reflection intensities were integrated and scaled with the HKL2000 suite of computer programs.⁷² The crystals of Zn²⁺,Ca²⁺–S100B, Pnt–Ca²⁺–S100B, and Pnt–Zn²⁺,Ca²⁺–S100B diffracted to 1.88, 2.15, and 1.85 Å resolution, respectively. Preliminary phases were obtained via molecular replacement techniques using the structure of Ca²⁺-bound S100B (PDB code: 1MHO;⁴⁴) as a search model and the computer program Molrep and/or Phaser from the CCP4 program suite.⁷³ Model building and refinement of S100B were completed using REFMAC5 and COOT.^74,75 The locations of the two Pnt molecules and several water molecules were determined by visual inspection of 2mF_o–DF_c and mF_o–DF_c electron density maps with the program COOT.⁷⁵ It was evident from visual inspection of the 2mF_o–DF_c and mF_o–DF_c electron density maps that the only way to accurately model the bound Pnt in site 1 was to place the molecule in two overlapping orientations, each with partial occupancy. We determined the final occupancy for Pnt orientation in the Ca²⁺–S100B and Zn²⁺,Ca²⁺–S100B of site 1 to be 0.5 and 0.3, giving a total occupancy for Pnt at that site of 1.0 and 0.6, respectively. The occupancy was set manually and the value was chosen because higher or lower values resulted in positive or negative difference density around the drug. The occupancies and temperature factors for the two Pnt orientations at site 1 were constrained to be identical owing to the fact that the Pnt resides on a crystallographic 2-fold axis. Modeling of Pnt at site 2 was similar to that at site 1, with two molecules having overlapping orientations with occupancies of 0.5, resulting in a total occupancy of 1.0 for Pnt in site 2. As it turned out, the occupancies in site 1 varied while those in site 2 were found to be similar for both Pnt–S100B complex structures (i.e., Pnt–Ca²⁺–S100B and Pnt–Zn²⁺,Ca²⁺–S100B), whereas the temperature factors for the Pnt molecules at those sites varied between the two complexes (Table 1). The stereochemistry was checked with the programs WHATCHECK and PROCHEK,^76,77 and interhelix angles and distances were calculated using the programs INTERHLX,⁷⁸ VGM,⁷⁸ and IHA.⁷⁹ The quaternary structure and accessible surface areas were analyzed using the PISA server^§. Figures were generated with the program PyMOL^∥.

Abbreviations used

Pnt

pentamidine

ITC

isothermal titration calorimetry

PEG

polyethylene glycol

HSQC

heteronuclear single quantum coherence

STD

saturation transfer difference

Tes

2-{[2-hydroxy-1,1-bis (hydroxymethyl)ethyl]amino}ethanesulfonic acid

ddH₂O

deionized and doubly distilled H₂O

PDB

Protein Data Bank

VGM

vector geometry mapping