A novel S100 target conformation is revealed by the solution structure of the Ca2+-S100B-TRTK-12 complex.

The Alzheimer-linked neural protein S100B is a signaling molecule shown to control the assembly of intermediate filament proteins in a calcium-sensitive manner. Upon binding calcium, a conformational change occurs in S100B exposing a hydrophobic surface for target protein interactions. The synthetic peptide TRTK-12 (TRTKIDWNKILS), derived from random bacteriophage library screening, bears sequence similarity to several intermediate filament proteins and has the highest calcium-dependent affinity of any target molecule for S100B to date (K(d) <1 microm). In this work, the three-dimensional structure of the Ca(2+)-S100B-TRTK-12 complex has been determined by NMR spectroscopy. The structure reveals an extended, contiguous hydrophobic surface is formed on Ca(2+)-S100B for target interaction. The TRTK-12 peptide adopts a coiled structure that fits into a portion of this surface, anchored at Trp(7), and interacts with multiple hydrophobic contacts in helices III and IV of Ca(2+)-S100B. This interaction is strikingly different from the alpha-helical structures found for other S100 target peptides. By using the TRTK-12 interaction as a guide, in combination with other available S100 target structures, a recognition site on helix I is identified that may act in concert with the TRTK-12-binding site from helices III and IV. This would provide a larger, more complex site to interact with full-length target proteins and would account for the promiscuity observed for S100B target protein interactions.

The S100 proteins are low molecular weight (10 -12 kDa) members of the EF-hand family of calcium-binding proteins. Many of these proteins, including several of the S100s, the muscle contractile protein troponin C, and the ubiquitous protein calmodulin act as signaling molecules by converting an influx of cellular calcium into a biological response. Calcium binding to these EF-hand proteins triggers a conformational change and allows the protein to interact with an appropriate target molecule. The S100 proteins are unique among this family because, unlike troponin C or calmodulin, they exist in solution as homo-or heterodimers. Each S100 monomer contains two helix-loop-helix calcium-binding motifs as follows: a basic N-terminal pseudo EF-hand comprising 14 residues (site I), and a canonical and acidic C-terminal EF-hand of 12 residues (site II). A central linker region joining the two EF-hands along with the extreme N and C termini of these proteins exhibit the most sequence divergence among family members and are therefore believed to provide specificity for target protein interactions. This feature, along with the dimeric state of the S100 proteins, likely indicates these calcium-signaling proteins have the distinctive ability to interact with more than one target molecule at a time. S100B, 1 a homodimer of 91-residue S100␤ monomers, is found primarily in glial cells and has been implicated in neurological diseases including Alzheimer's disease and Down's syndrome (1)(2)(3). More than 20 calcium-sensitive in vitro binding partners have been identified for S100B (4) including several cellular architecture proteins such as tubulin (5) and GFAP (6), where S100B can inhibit polymerization of these oligomeric molecules. Furthermore, S100B inhibits the phosphorylation of multiple kinase substrates including the Alzheimer protein tau (7,8) and neuromodulin (GAP-43) (9) through a calcium-sensitive interaction with the protein substrates. Consistent with the calcium-induced conformational change mechanism, a comparison of three-dimensional structures of apo-and calcium-bound S100B reveals that binding of calcium leads to the exposure of a hydrophobic surface(s) for proteinprotein interactions (10).
The 12-residue peptide TRTK-12, derived from random bacteriophage library experiments, has been used in previous studies as a model for S100B-target protein interactions. Experiments utilizing this peptide have indicated that peptides containing the consensus sequence (R/K)(L/I)(XWXXIL) bind specifically to S100B in a calcium-sensitive manner (11). Furthermore, this consensus sequence is conservatively found in the cytoskeletal proteins tubulin, desmin, vimentin, and GFAP, shown previously to interact with Ca 2ϩ -S100B (12). The TRTK-12 peptide competes with S100B target proteins including CapZ-␣ and GFAP for Ca 2ϩ -S100B binding (6,11) and has the highest affinity (K d ϳ260 nM) (13) of any known S100B target. Fluorescence studies (13) have shown that Trp 7 of TRTK-12 is a key residue for the calcium-sensitive interaction with S100B becoming buried at the protein-peptide interface. In addition, deletion of residues 85-91 from S100B leads to a Ͼ2000-fold decrease in affinity for TRTK-12 indicating the C-terminal helix is an important site of interaction.
To date, only three structures are available for S100-target peptide complexes. In each case the target peptide adopts a 2.5 turn ␣-helical structure that interacts via two distinct modes with the S100 protein. The structures of human S100A10 and S100A11 in complex with peptides from the binding regions of annexin II and I, respectively, are nearly identical (14,15) with the annexin peptides bridging the two S100 monomers through contacts in the linker region and C terminus of one monomer and the N terminus of the other monomer. Surprisingly, this similarity of interaction occurs despite little sequence similarity between the annexin peptides. In contrast, the interaction of rat S100B with a 23-residue peptide from the tumor suppressor protein p53 shows each S100␤ monomer binds to a single peptide through interactions with helix III and a portion of helix IV (16). This orients the ␣-helical p53 peptide about 90°f rom that found for the annexin peptides with respect to their S100 partners. These structural variations of the S100A10, S100A11, and S100B complexes indicate that recognition differences in the protein and the target must exist for S100 proteins. In an effort to clarify these interactions, we present the three-dimensional solution structure of Ca 2ϩ -S100B in complex with the TRTK-12 peptide. The amino acid sequence of TRTK-12 does not correspond to a natural sequence for any known S100B target. However, its unusually high affinity for S100B may indicate that it contains structural determinants that remain to be uncovered for recognition of an S100 binding partner. The binding interaction of TRTK-12 is unexpected with the peptide adopting an extended and reversed orientation compared with the p53 interaction. Furthermore, there is no interaction with helix I of S100B as found in the S100A11/ S100A10 annexin structures thus providing a novel third mode of recognition for an S100-target protein complex.
Structure Determination-Approximate interproton distances in S100B were determined using 15 N-filtered NOESY (26)( mix ϭ 150 ms) and simultaneous 13 C/ 15 N-separated three-dimensional NOESY-HSQC (27)( mix ϭ 100 ms) experiments. NOEs derived from the 15 N-filtered NOESY were calibrated based on known d HN␣ and sequential d ␣HN distances. The 13 C/ 15 N-separated three-dimensional NOESY-HSQC was calibrated to geminal protons or protons on adjacent carbon atoms. In cases where direct calibration was not possible, the maximum distance of 6.0 Å was used. and angles for Ca 2ϩ -S100B and the bound TRTK-12 peptide were incorporated where greater than 9 of 10 predictions fell within the same region of the Ramachandran plot based on the TALOS program (28). Minimum standard deviations of Ϯ20°and Ϯ30°w ere used for and , respectively. angle restraints were confirmed with 3 J NH-H␣ coupling constants derived from an HNHA experiment (29). Hydrogen bond distance restraints of r NH-O ϭ 1.4 -2.7 Å and r N-O ϭ 2.4 -3.7 Å were implemented based on the identification of helical regions in initial structures. NOEs between the two S100␤ monomers were unambiguously assigned using a mixed 13 C/ 12 C S100B dimer and a 13 C F 1 -edited F 3 -filtered NOESY-HMQC experiment (30) which detects only NOEs between a proton attached to 13 C and a proton attached to 12 C. These NOEs were calibrated in the same fashion as cross-peaks from the 13 C/ 15 N-separated three-dimensional NOESY-HSQC. Secondary structure of the TRTK-12 peptide was determined using a two-dimensional wetNOESY experiment and 15 N, 2 H S100B protein.
Dihedral restraints for the peptide were included using the same stipulations as for Ca 2ϩ -S100B. Intermolecular NOEs between S100B and TRTK-12 were unambiguously identified using the same 13 C F 1 -edited F 3 -filtered NOESY-HMQC experiment (30) described above with 13 C/ 15 N-labeled S100B protein and unlabeled TRTK-12 as well as a simultaneously 13 C/ 15 N-separated three-dimensional NOESY-HSQC using 15 N-labeled S100B and specifically labeled TRTK-12. Initial structures were calculated using the hybrid distance geometry and simulated annealing protocol in the Crystallography and NMR System program, version 1.1. The two monomeric subunits of S100B were constrained to be identical using the non-crystallographic symmetry definition and a force constant of 10 kcal/mol. The same constraint was applied for the two molecules of TRTK-12.

RESULTS
Binding of the TRTK-12 peptide has been monitored previously by NMR and fluorescence spectroscopy and found to have a K d ϭ 260 nM (13). In this work, formation of the complex was measured using a 15 N-labeled S100B sample and monitoring the change in resonances as a function of added TRTK-12 peptide. Complex formation was complete at a ratio of 2:1 TRTK-12:S100B indicating that two TRTK-12 peptides bind to each S100B dimer protein. The resulting 1 H-15 N HSQC spectrum of this complex showed a single set of resonances in the NMR spectra for most residues (Fig. 1) indicating the arrangement of the Ca 2ϩ -S100B-TRTK-12 complex dimer is symmetric. Exceptions were noted for residues Ser 1 , Leu 3 , Lys 5 , Ser 41 , and His 42 which have duplicate peaks resulting from the presence of formyl and desformyl N-terminal methionine S100B species as observed previously for apo-and Ca 2ϩ -S100B (31). Several residues in the N-terminal calcium-binding loop have weak (Ser 18 , Glu 21 , His 25 , Lys 26 , Leu 27 , and Lys 28 ) or absent (Gly 22 and Asp 23 ) amide correlations in the 1 H-15 N-HSQC likely due to exchange with the H 2 O solvent. Gly 66 shows a substantial downfield 1 H shift characteristic of calcium binding to the C-terminal calcium loop. A comparison of this spectrum with that for Ca 2ϩ -S100B alone indicated that several residues including Ile 47 , Val 53 , Ala 78 , Ala 83 , and Cys 84 had shifted upon TRTK-12 binding.
To identify the specific interactions between TRTK-12 and human Ca 2ϩ -S100B, the solution structure of the complex (24 kDa) was determined using 2072 experimental distance restraints in combination with 536 dihedral and 136 hydrogen bond restraints. This generated a family of 17 well defined structures (Fig. 2) based on the superposition of the 8 helices in the dimer (root mean square deviation 0.63 Ϯ 0.07) ( Table I). The TRTK-12 peptide was well defined between residues 4 and 12 forming a coil structure that folds back on itself. As observed in the rat S100B-p53 structure (16), each S100B monomer binds one TRTK-12 molecule.
Structure of Ca 2ϩ -S100B in the Complex-Overall, the struc- Correlations of side chain amide groups are indicated by horizontal lines. Duplicate peaks are observed for residues at the extreme N terminus (S1, L3, and K5) and in the linker region (S41 and H42) due to the presence of both formyl and desformyl N-terminal methionine S100B forms.
FIG. 2. Structure of the Ca 2؉ -S100B-TRTK-12 complex. The N and C termini are labeled for one S100␤ monomer and TRTK-12, and helices are indicated for S100␤. A, stereo view of the backbone superposition of the final ensemble of 17 NMR derived structures of the complex. S100␤ monomers are shown in magenta and blue, and the TRTK-12 peptide molecules are shown in light blue. B, ribbon structure of the complex. Each monomer consists of four ␣-helices (helix I, blue; II, magenta; III, green; and IV, yellow) and two anti-parallel ␤-strands (orange, blue). TRTK-12 is shown in dark blue.
Structure of TRTK-12 in the Complex-Our previous studies (33) have shown that the TRTK-12 peptide is unstructured in the absence of Ca 2ϩ -S100B. Consistent with this, the circular dichroism spectrum of TRTK-12 (Fig. 3) showed an ellipticity minimum at Ͻ200 nm indicative of a random coiled structure. Upon binding, TRTK-12 assumes a coiled conformation that folds back upon itself (Fig. 2). The structure determination of TRTK-12 was aided by using a peptide having uniform 13 Clabeling at Ile 5 and Ile 10 and the N-terminal acetyl CH 3 positions thus providing 13 C markers throughout the peptide length. The coiled structure of TRTK-12 was supported by definitive NOEs between the side chains of residues Trp 7 -Leu 11 , and peptide spectra were characterized by an absence of ␣-helical NOEs and ␣-proton (and ␣C for Ile 5 and Ile 10 ) chemical shifts not consistent with ␣-helical structure. These findings are supported by CD spectra of Ca 2ϩ -S100B (Fig. 3) that show ellipticity minima at 208 and 222 nm typical of an ␣-helical protein. Upon addition of 1 eq of TRTK-12 per S100␤ monomer (Fig. 3). little difference in this spectrum occurs indicating the TRTK-12 peptide does not assume an ␣-helical conformation. Furthermore, the 208 / 222 ratio for the complex appears nearly identical to that of Ca 2ϩ -S100B, indicating little change in helix-helix interactions occurs.
Ca 2ϩ -S100B-TRTK-12 Interactions-The interface between Ca 2ϩ -S100B and TRTK-12 is defined by several residues in FIG. 3. Far-UV circular dichroism spectra of Ca 2؉ -S100B in the presence of the TRTK-12 peptide. The spectra show TRTK-12 (q), Ca 2ϩ -S100B (OE), and Ca 2ϩ -S100B in the presence of TRTK-12 (ࡗ). The complex was formed using a 1:2 ratio of Ca 2ϩ -S100B (50 M) and TRTK-12 (100 M) consistent with the binding of one TRTK-12 peptide for each S100␤ monomer. helix III (Val 52 , Lys 55 , Val 56 , Thr 59 , and Leu 60 ) and helix IV (Phe 76 , Ala 80 , Ala 83 , and Cys 84 ). In particular, 13 C-labeled Ile 5 and Ile 10 revealed definitive interactions to residues Thr 82 in helix IV and Val 52 and Val 56 in helix III, respectively (Fig. 4). Furthermore, residues Lys 9 , Ile 10 , and Leu 11 in the C terminus of TRTK-12 have multiple contacts in helix III, and Thr 1 , Thr 3 , and Lys 4 are in close proximity to Ala 83 in helix IV. These interactions position the N terminus of the TRTK-12 peptide near helix IV and the C terminus near helix III resulting in a peptide orientation opposite to that observed for the S100B-p53 complex. Although 13 C labeling of the N-terminal acetyl group in TRTK-12 was used, no observable peptide-protein crosspeaks resulting from this label were observed in 13 C-edited NOE spectra, indicating the extreme N terminus of TRTK-12 is exposed to solvent and does not interact with residues in helix I as the annexin peptides do in the S100A10 and S100A11 structures. A large hydrophobic cavity exists on Ca 2ϩ -S100B where TRTK-12 is bound (Fig. 5). This observation is consistent with previous results showing that calcium binding to human S100B results in exposure of a hydrophobic surface (10). Trp 7 of the peptide is the anchoring residue nestled in a hydrophobic pocket including residues Val 56 , Thr 59 , Phe 76 , and Val 80 of S100B. This environment results in the observed blue shift and an increase in fluorescence intensity for Trp 7 upon protein binding (33). Comparison of the exposed surface area between Ca 2ϩ -S100B alone and in complex with TRTK-12 shows the side chains of Val 56 , Thr 59 , and Ala 83 decrease their surface exposure by 86, 96, and 99%, respectively, upon TRTK-12 binding. Similarly, residues Trp 7 and Ile 10 of TRTK-12 have more than 84 and 90%, respectively, of their surface area buried in the complex. Recent fluorescence studies of the S100B-TRTK-12 interaction have indicated that deletion of the Cterminal 7 residues in S100B results in a drastic decrease in TRTK-12 binding affinity (13). This was attributed to a loss in ␣-helical structure near Ala 83 in the truncated S100B protein.
Presumably this would considerably alter the arrangement of residues Val 80 , Ala 83 , and Cys 84 resulting in a significantly reduced affinity.

DISCUSSION
Examination of the structure of Ca 2ϩ -S100B-TRTK-12 and comparison to the existing structures of S100A10-annexin II, S100A11-annexin I, and S100B-p53 allows a distinction to be made between target recognition by these S100 proteins. Fur- FIG. 5. Electrostatic potential surface showing the TRTK-12-binding site on Ca 2؉ -S100B. Negative potential is indicated in blue, and positive potential is shown in red for Ca 2ϩ -S100B. A, location of the two TRTK-12 molecules (green) on opposite sides of Ca 2ϩ -S100B. The TRTK-12 peptide fits into the hydrophobic cleft generated by helices III and IV of each monomer. Note that on either side of the TRTK-12-binding site a significant uncharged region exists. B, environment of the anchoring Trp 7 residue from TRTK-12 (green) showing interactions with S100B residues (blue) in helix III (Val 56 and Thr 59 ) and IV (Phe 76 and Val 80 ).
thermore, the TRTK-12 sequence was identified from random bacteriophage peptide selection and therefore may contain unique binding features in comparison to the natural sequences for the annexin and p53 peptides. In the current structure, the TRTK-12 peptide lies in a hydrophobic cleft formed between helices III and IV maximizing the interaction of the hydrophobic residues Ile 5 , Trp 7 , Ile 10 , and Leu 11 with residues in both helices of S100B. Nearly all the interaction is hydrophobic in nature reaffirming earlier work that the S100B-TRTK-12 interaction is tighter at increased ionic strength (13). The interaction is distinct from that of the rat S100B-p53 peptide (16), likely due to the sequence differences of the interacting peptides. The target sequence for p53 (SRH-KKLMFKT) bears little similarity to TRTK-12 containing a highly positively charged C terminus that orients near the highly acidic N terminus of helix III (EEIKEEQE). In contrast, the charged residues in TRTK-12 are more dispersed throughout its sequence and most are occupied through potential hydrogen-bonding interactions. Nevertheless, some similarities between the target peptide molecules exist. The anchoring residues for the p53 peptide are Leu 383 and Phe 385 , spaced only one residue apart, as are Ile 5 and Trp 7 in TRTK-12. The distinguishing feature of these two residues in each peptide is their positions with respect to S100B. In the current work Trp 7 is positioned in an analogous region to Leu 383 in p53, whereas the side chain of Ile 5 is located nearby that of Phe 385 in p53, albeit somewhat shifted due to a different orientation. The reverse nature of these two interactions together with the differential charge balance in the peptides likely results in the reversal of orientation of TRTK-12 to p53 upon binding to S100B.
The regions of interaction of TRTK-12 and p53 differ significantly from that of both annexin I and II with S100A11 and S100A10, respectively. Both of the annexin peptides have interactions with several residues in the N-terminal helix I of these S100 proteins (Fig. 6A). An examination of the sequences and interactions reveals three key residues in the annexin peptides (Val 3 , Glu 5 , and Leu 7 ) that have identical interactions with the N-terminal residues Glu 5 , Met 8 , Glu 9 , and Met 12 in S100A10 (Glu 9 , Ile 12 , Glu 13 , and Ile 16 in S100A11). This pattern in the target peptide, where two hydrophobes are separated by three residues, including a central acidic residue, is not found in either the TRTK-12 or p53 peptides. Furthermore, the side chain carboxylate of Glu 9 in S100A10 (Glu 13 in S100A11) has conserved hydrogen bonds to the Thr 2 and Val 3 amides in annexin II (Met 2 and Val 3 , in S100A11). In S100B the position of Glu 9 (Glu 13 ) is held by a hydrophobe (Val 8 ) removing its side chain hydrogen bonding ability. Interestingly, this position is among the least conserved within the S100 protein family. This sequence divergence in both the S100B and S100A10/S100A11 proteins and in the peptide motif is likely responsible for the different target peptide recognition sites between the proteins. Consistent with this, previous experiments have shown that TRTK-12 shows no observable interaction with S100A11 (33) when monitored by Trp 7 fluorescence.
The interaction of Trp 7 in TRTK-12 likely contributes to the tightest binding for any S100B target to date. It is also interesting that replacement of Phe 385 in the p53 peptide with a Trp residue increases its affinity for rat S100B by about 5-fold (34). In addition, important similarities exist between the environments of Trp 7 (TRTK-12) and Trp 11 (annexin I) even though the orientation of the two target peptides is different with respect to the protein helices in each complex. A comparison of the anchoring Trp 7 of TRTK-12 reveals side chain interactions (Ͻ6.5 Å) with Ile 47 , Val 52 , and Val 56 in a majority of the NMR structures. The analogous interactions, involving Gln 52 , Val 57 , and Met 61 exist for S100A11 (Fig. 6B). Despite this similarity, the differences of the peptide orientations for TRTK-12 and annexin I result in a significant hydrophobic surface for each protein that remains exposed, even in the presence of the bound peptide ( Fig. 5A and Fig. 6C). This may occur simply because the target peptides presently used are too short or that the interaction can be modulated through a second region of interaction. As a result, it has been suggested recently (35) that S100 proteins may bind full-length targets using more than one site. The similarity of the locations for Trp 7 (TRTK-12) and Trp 11 (annexin I) would provide a unique bridging residue to encompass a larger, contiguous binding region that includes an ␣-helical annexin type interaction, utilizing the VEL residues found in annexin I and II, and an extended TRTK-12 type interaction based on hydrophobic interactions anchored by Trp 7 (Fig. 6C). It is interesting that the recent structure of the calmodulin-calmodulin kinase kinase peptide complex displays exactly these features, where the N-terminal portion of the peptide forms an ␣-helical structure and the C-terminal region is more extended (36), folding back on the helix in addition to having important protein contacts. As in calmodulin, this larger interaction site would also provide a rationale for the promiscuity of S100B target interactions, including cellular architecture proteins such as tubulin, vimentin, desmin, GFAP, and caldesmon, which contain the TRTK-12 consensus motif (12). It remains to be seen whether the natural sequences for the intermedi-FIG. 6. Interactions of S100 proteins with annexin I, annexin II, and TRTK-12. A, aligned sequences of helices I (top) for S100B, S100A11, and S100A10 along with observed interactions to their respective target peptides (O). The annexin I and II peptides were aligned based on the structurally conserved interactions to the N-terminal residues of S100A11 and S10010, respectively (green), and involving the identical residues (Val, Glu, and Leu) in the annexin peptides (dark shading). The conserved interaction involving Glu 9 in S100A10 (Glu 13 in S100A11) is unlikely in S100B where a valine residue occupies this position (blue). B, the aligned sequences of helix III and the C-terminal portion of the linker are shown together along with the common interactions to Trp 7 (pink) in the TRTK-12 peptide and Trp11 in the annexin I peptide (O). Interactions found between S100B and TRTK-12, but not in annexin I, are also shown (---). C, proposed full-length target site for S100B. The accessible surface for Ca 2ϩ -S100B was calculated based on the complex with TRTK-12 in the absence of the peptide. Residues where the side chain interactions with the TRTK-12 peptide (green) and those where a proposed site exists based on the S100A11-annexin I structure (yellow) are shown based on the observations in A. Residues that have common contacts to the bridging Trp residue, as shown in B, are shown in red. A comparison of the structures reveals that the hydrophobic surface used by S100B for recognition of TRTK-12 (green) is largely exposed in the S100A11-annexin I structure and vice versa. ate filament proteins interact in a similar manner as TRTK-12 or via the ␣-helical mode found for either annexin or p53 peptides.
Coordinates-Chemical shift assignments for S100B and TRTK-12 in the complex have been deposited in the Bio-MagResBank (accession number 5377).