S100A6 and S100A11 Are Specific Targets of the Calcium- and Zinc-binding S100B Protein in Vivo *

In solution, S100B protein is a noncovalent homodimer composed of two subunits associated in an antiparallel manner. Upon calcium binding, the conformation of S100B changes dramatically, leading to the exposure of hydrophobic residues at the surface of S100B. The residues in the C-terminal domain of S100B encompassing Phe87 and Phe88 have been implicated in interaction with target proteins. In this study, we used two-hybrid technology to identify specific S100B target proteins. Using S100B as bait, we identify S100A6 and S100A11 as specific targets for S100B. S100A1, the closest homologue of S100B, is capable of interaction with S100B but does not interact with S100A6 or S100A11. S100B, S100A6, and S100A11 isoforms are co-regulated and co-localized in astrocytoma U373 cells. Furthermore, co-immunoprecipitation experiments demonstrated that Ca2+/Zn2+stabilizes S100B-S100A6 and S100B-S100A11 heterocomplexes. Deletion of the C-terminal domain or mutation of Phe87 and Phe88 residues has no effect on S100B homodimerization and heterodimerization with S100A1 but drastically decreases interaction between S100B and S100A6 or S100A11. Our data suggest that the interaction between S100B and S100A6 or S100A11 should not be viewed as a typical S100 heterodimerization but rather as a model of interaction between S100B and target proteins.

S100 proteins are a family of low molecular weight, acidic, calcium-binding proteins that contain two EF-hand calcium binding sites. To date, 19 different proteins have been assigned to the S100 protein family (reviewed in Refs. 1 and 2). They show different degrees of similarity, ranging from 25 to 56% identity at the amino acid level, and share conserved structural features and possibly common mechanisms of action (1). Some of the S100 proteins, including S100B, S100A6, and S100A11 are not only calcium-binding proteins but also bind Zn 2ϩ with high affinity (3)(4)(5). Zn 2ϩ binding has a pronounced effect of increasing the Ca 2ϩ affinity in S100B (3) and of increasing S100B binding to target peptides (6). The S100B has attracted much interest in the past few years because, like other proteins implicated in neurodegeneration (e.g. ␤-amyloid, superoxide dismutase), its gene is located within a segment of chromosome 21, which is trisomic in Down's syndrome (DS) 1 (7). The observed overexpression of S100B in the brains of patients with DS, Alzheimer's disease (8,9), or AIDS (10) has led to the hypothesis that S100B plays a contributory role in neuropathologies associated with these diseases. Within the brain, expression of S100B is mainly restricted to glial cells. Moreover, S100B protein is up-regulated in many tumors (11,12). A functional interaction between S100B and the p53 pathway of cell growth inhibition and apoptosis has been reported, suggesting a role for the protein in cell cycle regulation (13,14). Identification of the in vivo S100B target proteins is essential for the determination of the exact contribution of the protein to cell functions. Several putative S100B target proteins have already been characterized, including cytoskeleton-associated proteins (15)(16)(17)(18)(19)(20), nuclear protein kinase (21), and nuclear transcription factors (22,23). In all cases, these interactions depend on the Ca 2ϩ -bound S100B conformation. NMR spectroscopy and x-ray crystallography have been used to determine the three-dimensional structure of apo-and holo-S100B forms (24 -30). The structure revealed that the apoprotein forms noncovalent dimers with two molecules that associate in an antiparallel manner to form a tightly packed hydrophobic core at the dimer interface involving six of the eight helices present in the protein and the C-terminal loop (Fig. 2). Upon calcium binding, the conformation of S100B changes dramatically from the compact structure in the apo-form to a more extended form. These changes lead to exposure of several residues in the C-terminal domain and the hinge region of S100B that serve for interaction with target proteins (6,27,31,32).
Most of the targets thought to be regulated by Ca 2ϩ -bound S100B have been identified by employing biochemical approaches. To our knowledge, only the Ndr kinase has been co-immunoprecipitated with S100B from whole cell extracts (21). In several cases, tissue or cellular distribution of reported targets does not correlate with that of the S100B (i.e. CapZ␣), casting doubts on the purely biochemical approaches. Here, we have screened a cDNA library of human brain using yeast two-hybrid technology to find interacting proteins for S100B in vivo. We have identified S100A6 (previously named calcyclin, CaBP, PRA, 2A9, or 5B10) and S100A11 (previously named S100C or calgizzarin) as specific S100B target proteins.

EXPERIMENTAL PROCEDURES
Materials-T4 DNA ligase and Taq polymerase were purchased from Life Technologies, Inc. The TA cloning kit was from Promega. The QuickChange site-directed mutagenesis kit was from Stratagene. Amino acids and other chemicals were from Sigma.
Primary Antibodies-Monoclonal LexA and Gal4 antibodies and rabbit polyclonal anti-GFP antibody were from CLONTECH and were used at 20 ng/ml. The mouse monoclonal anti-S100A6 (MabS100A6, S5049) was from Sigma and was used at a 1:6000 dilution in Western blots and at a 1:500 dilution in immunofluorescence. The polyclonal rabbit anti-S100B (RAbS100B, A5110) was from Dako and was used at 2.6 g/ml in Western blots and at 0.43 g/ml in immunofluorescence. The affinitypurified polyclonal guinea pig anti-S100A11 (GPAbS100A11) was prepared by "Elevage Scientifique des Dombes" (France) using a mixture of human and pig recombinant S100A11 proteins. The IgGs were purified using S100A11-Sepharose beads and were used at 10 and 0.1 g/ml in Western blots and immunofluorescence, respectively. The ascites anti-␣-tubulin (Mab␣-tub) was a gift from L. Paturle and D. Job and was diluted 1:20,000 in Western blot. S100 Recombinant Proteins-The purification protocol of human S100A1, S100A6, S100B, and S100A11 recombinant proteins has been described previously (33).
Cell Cultures-The human astrocytoma U373 MG cells and COS-7 cells were maintained in Hepes-buffered Dulbecco's modified Eagle's medium plus 10% fetal bovine serum and 1% penicillin and streptomycin in a humidified 95% O 2 , 5% CO 2 incubator.
Cell Extracts-Transformed COS-7 and human U373 MG cells were lysed in Laemmli loading buffer. The protein concentration was measured by the bicinchoninic acid method (Pierce) with bovine serum albumin as a standard.
Western Blot-Analysis of S100 protein isoforms in U373 MG cells was carried out using 11% SDS-Tris-Tricine polyacrylamide gels (34). Analysis of fusion proteins from cell or yeast extracts was performed using 12.5% SDS-Tris-glycine polyacrylamide gels (35). Proteins were visualized using an ECL kit (NEN Life Science Products).
Co-immunoprecipitation Analysis-COS-7 cells were co-transfected with a vector coding for the fusion EGFP-S100B protein and a vector coding for one of the Myc-S100 fusion proteins using the Fugene6 Kit (Roche Molecular Biochemicals). After 48 h, cells were washed twice with ice-cold Tris-buffered saline (TBS; 30 mM Tris-HCl, pH 7.4, 150 mM NaCl), scraped off, and lysed in 800 l of TBS containing 0.2% Triton X-100 (TBS-T) plus 2 mM EDTA or 0.5 mM CaCl 2 and 10 m ZnSO 4 . Cell lysates were clarified by microcentrifugation and incubated with 5 g of purified mouse anti-Myc IgG or purified mouse anti-MyoD IgG as a control in the presence of 20 l of protein G-Sepharose (Amersham Pharmacia Biotech) for 1 h at 8°C. The beads were pelleted and washed in TBS-T, and the immunoprecipitated proteins were analyzed by Western blot using a polyclonal antibody directed against GFP. For the analysis of endogenous S100B-S100A6 heterocomplex in human U373 MG cells, protein complexes were first immunoprecipitated with polyclonal purified anti-S100B IgG (Dako) or with anti-MyoD IgG in the presence of 20 l of protein A-Sepharose (Amersham Pharmacia Biotech). Immunoprecipitated proteins were analyzed by Western blot using the mouse anti-S100A6 monoclonal antibody.
Confocal Microscopy-Human U373 MG cells were grown on 10-mm glass coverslips coated with polylysine at 0.01 mg/ml. After 5 or 15 days of culture, cells were fixed for 30 min with 4% paraformaldehyde in HBS (10 mM Hepes, pH 7.4, 130 mM NaCl, 15 mM KCl, 5 mM MgCl 2 ). Cells were first permeabilized in TBS containing 0.1% Triton X-100, 2 mM CaCl 2 , 20 M ZnSO 4 , and 10% normal goat serum for 30 min and then incubated with the primary antibody for 90 min in TBS containing 2 mM CaCl 2 , 20 M ZnSO 4 , and 5% normal goat serum. The appropriated secondary antibodies conjugated with cyanin 3 (Jackson Immunoresearch Laboratories) or with Alexa488 (Molecular Probes, Inc., Eugene, OR) were incubated in the same buffer as described above for 1 h. Preparations were analyzed with a confocal scanning laser microscope LSM410 (Zeiss). Fluorescence from Alexa488 was obtained using the 488-nm excitation wavelength of the argon laser, and the emitted light was selected using a 510 -525-nm band pass filter. Fluorescence of cyanin 3 was obtained using a 543-nm excitation wavelength of the argon laser, and the emitted light was selected using a 560-nm dichroic filter and a 550-nm-long pass filter. The gain and the contrast of the photomultiplier detector were set in order to obtain optimal imaging of the two types of fluorescence while limiting the cross-talk between channels. Immunostaining controls have been performed by omitting the primary antibody.
Plasmid Constructions-Fusion proteins with LexA DNA-binding domain (LexADBD) were constructed in pLexA9, and those with the GAL4 activation-binding domain (GAL4AD) were constructed in pGAD424 (36). All cDNA coding sequences were polymerase chain reaction (PCR)-amplified with the addition of new restriction sites using the primers as indicated below. Each construct was checked by sequencing. Human wild-type and C-terminal deleted S100B were amplified by PCR using the primers 5Ј-AAGcccgggGATGTCTGAGCT-GGAGAAGG-3Ј integrating the SmaI restriction site (lowercase type) and either 5Ј-GGCTGCTTTctgcagTCACTCATGTTCAAAGAACTCG-3Ј (PstI site) for S100 wild-type or the primer 5Ј-TTCAAAGAActgcagTC-AGGCAGTAGTAACC-3Ј (PstI site and a stop codon (underlined) for C-terminal deleted S100B. PCR of human S100A1 was carried out with the primers 5Ј-ACTgaattcATGGGCTCTGAGCTGG-3Ј (EcoRI site) and 5Ј-AATGTGGCTgtcgacTCAACTGTTCTC-3Ј (SalI site). Myc-S100 fusions were constructed by inserting two fragments, a BamHI/NcoI fragment (coding for five copies of Myc epitope and recognized by the 9e10 monoclonal antibody (37)) and the NcoI/NotI fragment of human S100A1, S100A6, or S100A11 or the BspHI/NotI fragment of human S100B (33), into the BamHI/NotI sites of the vector pCDNA3.1 (Invitrogen).
The fusion of human S100B with the green fluorescence protein (GFP) was constructed by inserting the SalI/BamHI fragment of human S100B from pCDNAneoS100B (13) within pEGFP-C1 vector (CLONTECH).
Yeast Medium-Yeast medium was purchased from Difco. Yeast were grown on the YPAD-rich medium or YC-supplemented minimal medium to maintain selection for introduced plasmids or for the integrated reporter. Omitting uracil (U), tryptophan (W), or leucine (L) from medium maintains integrated His3 reporter, LexADBD fusion plasmid, or GAL4AD fusion plasmid, respectively (38).
Yeast Mating Assay-The interactions were tested by yeast mating essentially as described by Vojtek et al. (38). Briefly, the constructs in pLexA9 vector and in pGAD10 or pGAD424 were used to transform the Saccharomyces cerevisiae L40 (MATa, trp1, leu, his3, LYS::lexA-HIS3, URA3::lexA-LacZ) and AMR70 (MAT␣, trp1, leu2, his3, URA3::lexA-LacZ) yeast strains, respectively. The L40 and AMR70 transformants were selected on YC-UW and YC-UL medium, respectively. A single colony of each transformant was grown in 5 ml of the appropriate YC liquid medium for 24 h at 30°C with shaking at 200 rpm. Afterward, 1 l of each of the L40 transformed with LexADBD fusion plasmid was dropped to a YPDA plate and then overlaid with 1 l of each of the AMR70 transformed with GAL4AD fusion plasmid. The plate was incubated overnight at 30°C, and the colonies were transferred to a sterile velvet and then to a YC-UWLH plate. The growth of positive diploid colonies was analyzed after incubation for 2 days at 30°C.
Yeast Two-hybrid Screen-Large scale yeast transformations and two-hybrid screens were carried out using L40 yeast strain essentially as described (38,39). Briefly, pLexADBD-S100B bait construct was transformed singly into L40, and cells were grown on YC-UW medium for 2 days. A single positive transformant colony was grown in 10 ml of liquid YC-UW medium overnight. The culture was diluted into 100 ml of YC-UW medium for another 24 h and was then diluted to an A 600 of 0.4 in YPAD medium. A large scale transformation was performed using a human brain cDNA library constructed in pGAD10 plasmid (Matchmaker). Primary transformants were analyzed on YC-UWLH medium plates. A ␤-galactosidase assay was performed by a filter X-gal test, using the chromogenetic substrate X-gal (38). Library plasmids expressing LEU2 from positive transformants were selected using HB101 Escherichia coli, which requires leucine supplementation for growth. Each cDNA was retested in a mating assay against pLexAlamin, pLexA-S100A1, and the original bait construct using AMR70 yeast strain as described above.
Yeast Extracts-The pLexA9 constructs and the pGAD constructs were transformed singly into the L40 or AMR70 strains of yeast, respectively. The pLexA9 and pGAD positive transformants were grown for 18 h in 5 ml of YC-UW or YC-UL liquid medium, respectively. At A 600 between 0.5 and 0.8, yeast cells were pelleted at 1200 ϫ g and were rinsed once in TBS. Cultures were resuspended in lysis buffer composed of 30 mM Tris-HCl, pH 7.5, 2 mM EDTA, 1% Triton X-100, 2% SDS, and 50 mM dithiothreitol and then sonicated for 20 s and boiled for 5 min. The lysis buffer volume was adjusted to the A 600 of the culture. For 4 ml of pelleted culture with an A 600 of 0.8, 250 l of lysis buffer was added. The same volume of protein sample (20 l) was analyzed by Western blot.
Quantitative ␤-Galactosidase Assay-The L40 yeast cells were cotransformed with two-hybrid vector encoding LexADBD and GAL4AD fusion proteins. The double transformant was grown in 5 ml of YC-UWL medium until the A 600 was 0.5-0.8. Then 1.5 ml of culture was centrifuged, washed once with 500 l of Z-buffer (100 mM NaPO 4 , pH 7.0, 10 mM KCl, and 60 mM MgSO 4 ), and resuspended in 300 l of Z-buffer. 100 l of resuspended yeast was lysed by three successive liquid nitrogen freezing/37°C thawing cycles. The ␤-galactosidase activity was assayed using O-nitrophenyl-␤-D-galactopyranoside as substrate according to the manufacturer's instructions (CLONTECH). The specific ␤-galactosidase activity was expressed in units according to Miller (40,41). Calculations were performed following the CLONTECH protocol. Activity was determined in triplicate on each colony, and at least five separate colonies were assayed for each condition. Statistical analysis of the raw data was performed by analysis of variance followed by Student's t test. The values are considered as significant when p is Ͻ0.05.

RESULTS
The S100A11 and S100A6 Interact with S100B in the Yeast Two-hybrid System-The yeast two-hybrid system has been developed to monitor interactions between proteins within the confines of a yeast cellular environment (42). We used this method to identify putative S100B target proteins. A summary of the large scale human S100B two-hybrid screen we performed is reported in Table I. Full-length human S100B cDNA was fused to the DNA-binding domain of the transcription factor LexA (LexADBD) to serve as the "bait." A yeast expression cDNA library from human brain constructed in pGAD10 and fused to the activation domain of GAL4 (Gal4AD) was screened for interaction with the bait as detailed under "Experimental Procedures." Colonies that grew on yeast medium YC-UWLH and were blue when assayed by a filter X-GAL test were selected. Fifty colonies were obtained out of 4 ϫ 10 6 primary transformants. Library plasmids from positive transformants were isolated and retransformed in AMR70 yeast strain. Interaction specificity was retested on YC-UWLH medium and filter X-gal test after mating assays using L40 yeast strain transformed with either the original bait LexADBD-S100B, the LexADBD-S100A1, or an unrelated fusion protein LexADBD-lamin. Out of 50 library plasmids retested, 18 exclu-sively interacted with LexADBD-S100B fusion protein, 23 interacted with both LexADBD-S100B and LexADBD-S100A1 fusion proteins, and seven did not interact at all. Only two library plasmids retested gave a positive result with Lex-ADBD-lamin fusion protein and were considered as false positives.
We next determined the nucleotide sequence of the 18 positive clones that specifically interacted with LexADBD-S100B fusion protein. They all encoded proteins of the S100 family. Fifteen corresponded to the full-length coding sequence of S100A11, and three cDNAs corresponded to the full-length coding sequence of S100A6. We also sequenced 11 of 23 clones that interacted with both S100B and S100A1. They all corresponded to S100B (Table I, bottom). These preliminary results suggest that S100A11 and S100A6 are specific partners of S100B.
Yeast Two-hybrid Analysis of S100 Protein Interactions-The in vitro interaction between S100B and S100A1 as homodimers and heterodimers has been documented (43). Our results from the large scale two-hybrid screen suggest that in vivo S100B also interacts with S100A11 and S100A6. To further characterize these interactions, we next compared the specificity and strength of interaction between individual S100 species in yeast. The cDNA of S100A11 and S100A6 were subcloned into pLexA plasmid in frame with LexADBD. Each LexADBD and Gal4AD fusion constructs were used to transform L40 and AMR70 yeast cells, respectively. Beforehand, we determined whether LexADB-and Gal4AD-S100 fusion proteins were adequately and equally expressed in yeast. Western blot analysis was performed on yeast extracts using antibodies to LexADBD and Gal4AD proteins (Fig. 1A). All fusion proteins were displayed the expected molecular mass on SDS-polyacrylamide gel electrophoresis. LexADBD-S100 fusion proteins were all expressed to similar levels, whereas the expression levels of Gal4AD-S100A11 and Gal4AD-S100A6 were slightly higher than those of Gal4AD-S100B and Gal4AD-S100A1. Each L40 single colony expressing a LexADBD-S100 fusion protein was mated to each AMR70 colony expressing a Gal4AD-S100 fusion protein as detailed under "Experimental Procedures." Interactions were scored by the growth of diploid yeast on YC-UWLH medium lacking histidine (Fig. 1B). The results first show that each individual S100 isoform interacts with itself, confirming the ability of these proteins to form homodimeric complexes, and, as reported, S100B and S100A1 also interact together to TABLE I Summary of human S100B two-hybrid screen Top, in the two-hybrid system, interactions are measured by read-out from two transcriptional reporters. Activation of one reporter results in the production of histidine, allowing yeast to grow in the absence of this amino acid. Activation of the other reporter construct results in ␤-galactosidase activity that is measured by hydrolysis of an X-gal substrate that causes yeast to turn blue within 30 min. Steps of the two-hybrid screen are outlined to demonstrate screen efficiency. LexADBD-S100B was used as a bait to screen a human brain library constructed in pGAD10 and expressing 1.2 ϫ 10 6 independent clones. The transformation efficiency is shown as number of transformants. Histidine-positive transformants growing on media lacking histidine within 48 h to 3 days were tested for ␤-galactosidase activity. Bottom library plasmids were isolated and used to transform AMR70 yeast strain. L40 containing LexADBD-S100B, -S100A1, or -lamin fusion proteins were mated to AMR70 transformed yeast as detailed at the bottom. Proteins found to interact with S100A1 and/or with S100B are indicated as isolated and sequenced clones. The number of sequenced clones is indicated in parentheses.
Bait: human S100B cDNA library from human brain Independent clones

Number of cotransformants screened
Histidine and ␤galactosidase positive clones LexADBD-S100B Gal4AD-XcDNA 1.2 ϫ 10 6 4.0 ϫ 10 6 50 Positive clones in AMR70 Mating assay with L40 containing LexADBD fusion proteins Isolated and sequenced clones S100B S100A1 Lamin form the S100a isoform. The major observation is the confirmation of the isoform-specific interaction between S100A6 or S100A11 and S100B. No interaction was detected between S100A11-S100A6 or S100A1-S100A6 and S100A1-S100A11. A quantitative analysis of the interactions between S100 proteins was then performed. L40 yeast cells were co-transfected with LexADBD-and Gal4AD-S100 fusion constructs, and the ␤-galactosidase activity was measured as detailed under "Experimental Procedures" (Fig. 1C). The ␤-galactosidase assay experiments conducted to compare homodimerization revealed high ␤-galactosidase activities associated with S100B or S100A1 homodimerization compared with S100A6 or S100A11. These observations suggest that S100B and S100A1 are more prone to associate as homodimers than S100A6 or S100A11. Evaluation of interactions of S100A1, S100A6, and S100A11 isoforms with S100B isoform showed that ␤-galactosidase activities are not significantly different from those resulting from their respective homodimerization. These data indicate that interactions of S100A1, S100A6, and S100A11 with S100B occur with the same avidity as their homodimerization. In contrast, ␤-galactosidase activity assay revealed that S100B-S100B and S100B-S100A1 interactions are significantly stronger than S100B-S100A11 or S100B-S100A6. This suggests that the homodimerization of S100B and the heterodimerization of S100B with S100A1 are preferred to heterodimerization of S100B with S100A6 or S100A11.
The C-terminal Domain of S100B Is Involved in the Interaction with S100A6 and S100A11 Isoforms-Considering the difference of interaction of S100B with S100A1 and S100A6 or S100A11, we sought to determine the domain on S100B that might be involved in specific interaction with S100A6 and S100A11. Crystallographic studies indicate that all members of the S100 family have similar core structure and similar modes of dimerization (26,44,45). The C-terminal domain of the S100B, which may contribute to the dimer interface (26,44), is essential for the stabilization of the protein dimer in solution. Alkylation of Cys 84 within the C-terminal domain destabilizes the quaternary protein structure, allowing subunit exchange between protein dimers (46). We therefore investigated the function of the C-terminal domain of S100B in homo-and heterodimeric interactions with S100A1, S100A6, and S100A11. Using PCR, we introduced a premature stop codon at Cys 84 position, resulting in a truncated S100B (S100B⌬Ct) (Fig. 2). A mating assay on YC-UWLH medium showed that deletion of the C terminus on S100B has no significant effect on S100B dimerization or on interaction with S100A1 (Fig. 3B,   FIG. 1. Interactions between S100B and S100A6, S100A1, or S100A11 isoforms were analyzed by two-hybrid method. A, Western blot analysis of expression of LexADBD-S100 (LexADBD fusion) and Gal4AD-S100 (Gal4AD fusion) fusion proteins. The yeast cell extracts were performed as detailed under "Experimental Procedures" from L40 and AMR70 yeast strains transformed with LexADBD and Gal4AD fusion constructs, respectively. Equal volumes of sample (20 l) were loaded on SDS-Tris-Tricine-11% polyacrylamide gels. The proteins from L40 and AMR70 yeast samples were transferred to nitrocellulose membrane and immunoblotted with anti-LexA (lower panel) and anti-GAL4 (upper panel) monoclonal antibodies, respectively. The LexADBD and the Gal4AD proteins (lane pLexA/pGAD424) were expressed from yeast transformed with pLex9 and pGAD424 plasmids, respectively. Molecular mass markers are in kDa on the left. B, activation of histidine reporter was monitored by mating assay. Each L40 colony containing a LexADBD construct (LexADBD fusion) was mated to each of the AMR70 containing a GAL4AD construct (Gal4AD fusion) as described under "Experimental Procedures." The homo-and heterointeractions were scored by growth on YC-UWLH medium lacking histidine. L40 cells expressing LexADBD-lamin fusion protein and the AMR70 cells transformed with pGAD424 plasmid coding for Gal4AD protein were used as control. C, the strength of interactions of LexADBD-S100 with Gal4AD-S100 fusion proteins was quantified by ␤-galactosidase activity assay. All combinations of LexADBD and Gal4AD fusion constructs were co-transformed in L40 yeast strain. The colonies of each double transformant were grown in 5 ml of YC-UWL medium. The ␤-galactosidase activities were assayed in triplicate using O-nitrophenyl-␤-D-galactopyranoside as substrate. One ␤-galactosidase unit is defined as the amount of enzyme that hydrolyzes 1 mol of O-nitrophenyl-␤-Dgalactopyranoside/min/cell (40,41) following the calculation described in the CLONTECH protocol. Each bar represents the average of at least five separated colonies, each measured in triplicate. The number of separated colonies is indicated below in parentheses. Statistical analysis of S100B-S100B interaction versus S100B-S100A11 and S100B-S100A6 Ϯ S.E. is indicated (**, p Յ 0.002). Triangles indicate ␤-galactosidase activity below 1.5 units that corresponds to background obtained with the negative control LexADBD-lamin.
upper and lower panels, respectively). In contrast, S100B⌬Ct fused to LexADBD or Gal4AD proteins totally lost its capacity to interact with S100A11 or S100A6 fusion constructs. No ␤-galactosidase activity could be measured over background when yeast cells were co-transfected with LexADBD-S100B⌬Ct and either Gal4AD-S100A11 or Gal4AD-S100A6 fusion proteins (Fig. 3C). Note that the expression level of S100B⌬Ct fusion proteins is lower than wild type S100B fusion proteins especially for Gal4AD-S100B⌬Ct (Fig. 3A). However, the loss of interaction observed with S100A11 and S100A6 could not be explained by the lower expression of LexADBD-S100B⌬Ct, since the activities measured with Gal4AD-S100B or Gal4AD-S100A1 fusion constructs were unchanged (Fig. 3C).
We also generated site-specific amino acid mutations within the C terminus by replacement of Phe 87 and Phe 88 by Ala (S100BF87A, S100BF88A) (Fig. 2). Both S100BF87A and S100BF88A mutants significantly decreased the yeast growth on YC-UWLH medium and ␤-galactosidase activity when constructs were transfected in yeast cells with S100A6 (Fig. 3, B and C). S100A11 still interacted with S100BF87A, but interaction with S100BF88A mutant was significantly decreased.
We next investigated if the C terminus of S100B is sufficient for interaction with S100A6 and S100A11. S100B/A1Cter and S100A1/BCter chimeric proteins were generated by exchange of their respective C terminus (Fig. 2). The interactions of the chimeric proteins with S100A6 and S100A11 were compared in yeast using mating assay by growth on YC-UWLH medium lacking histidine (Fig. 4). The results showed that S100B/ A1Cter fused either with LexADBD or Gal4AD still interacts with S100A6 or S100A11. Moreover, the presence of the C-FIG. 2. Schematic representation of S100 monomer. A, the S100 monomer consists of two EF-hand motifs (EF1 and EF2), which are the structural motifs for calcium binding. The hinge region, which connects the two EF hand motif and the terminal hand of helix IV (Cter), is implicated in the interaction of Ca 2ϩ -bound S100B with target proteins. B, the primary structure of the C-terminal end of S100B constructs, chimeric S100B/A1Cter, and S100A1/BCter are represented. The substitutions of Phe 87 and Phe 88 in Ala residues are in boldface type.
FIG. 3. The C-terminal domain of S100B is crucial for the interaction with S100A11 and S100A6. A, expression of LexADBD-S100 (LexADBD fusion) and Gal4AD-S100 (Gal4AD fusion) fusion proteins was analyzed by Western blot. Yeast cell extracts were performed as detailed under "Experimental Procedures" from L40 and AMR70 yeast strains transformed with LexADBD and Gal4AD fusion constructs, respectively. Equal volumes of sample (20 l) were loaded on 11% SDS-Tris-Tricine polyacrylamide gels. The proteins from L40 and AMR70 yeast samples were transferred to nitrocellulose membrane and immunoblotted with LexA (lower panel) and GAL4 (upper panel) monoclonal antibodies, respectively. The LexADBD and the Gal4AD proteins were expressed from yeast transformed with pLex9 and pGAD424 plasmids, respectively (lane pLex9/pGAD424). Molecular mass markers are in kDa on the left. B, activation of histidine reporter was monitored by mating assay. Each L40 colony containing a LexA DNA binding domain protein construct (LexADBD fusion) was mated to each of the AMR70 containing a GAL4 activation domain construct (Gal4AD fusion) as described under "Experimental Procedures." The interactions were scored by growth on YC-UWLH medium lacking histidine. L40 cells expressing LexADBD-lamin fusion protein and the AMR70 cells transformed with pGAD424 plasmid were used as control. C, interactions of LexADBD-S100 with Gal4AD-S100 fusion proteins were quantified by ␤-galactosidase activity. All combinations of LexADBD and Gal4AD fusion constructs were co-transformed in L40 yeast strain. The separated colonies of each double transformant were grown in 5 ml of YC-UWL medium. Each bar represents the average of five separated colonies, each measured in triplicate. Statistical analysis of S100A11-S100B interaction versus S100A11-S100BF88A Ϯ S.E. is indicated (*, p Ͻ 0.01). Triangles indicate ␤-galactosidase activity below 1.5 units that corresponds to background obtained with the negative control LexADBD-lamin.
terminal domain of S100B in the chimeric S100A1/BCter protein is not sufficient to allow the protein to interact with S100A6 or S100A11. Together, the data indicate that the S100B interaction with S100A6 or S100A11 involves the C terminus but that the specificity of interaction requires another domain on the S100B molecule.
The S100B Forms Heterocomplexes with S100A6 and S100A11 in Vivo-To confirm interaction between S100B with the other S100 protein species in vivo, we compared heterocomplex formation between S100 proteins in COS-7 cells (Fig. 5). COS cells were co-transfected with EGFP-S100B and either with Myc-S100B, Myc-S100A1, Myc-S100A6, or Myc-S100A11 fusion proteins. Each Myc-S100 fusion isoform was immunoprecipitated using anti-Myc antibody and protein G-Sepharose, and co-immunoprecipitation of EGFP-S100B was then analyzed by Western blot. The results showed that EGFP-S100B is capable of associating with Myc-S100B with nearly equal efficiency as with Myc-S100A1. The interactions are calcium-independent (Fig. 5, upper  panel). In contrast, the heterocomplex EGFP-S100B⅐Myc-S100A6 and EGFP-S100B⅐Myc-S100A11 are stabilized in the presence of Ca 2ϩ and Zn 2ϩ (Fig. 5, lower panel). The amount of EGFP-S100B co-immunoprecipitated with Myc-S100B or Myc-S100A1 fusion proteins is much higher than with Myc-S100A6 or Myc-S100A11. These observations suggest that S100B is more prone to homodimerize or to heterodimerize with S100A1 and confirm the yeast two-hybrid data.
The S100B, S100A11, and S100A6 Isoforms Are Co-expressed and Co-localized in U373 MG Astrocytoma Cells-To address the question of the physiological relevance of the interactions between S100B and S100A6 or S100A11, we next analyzed S100 protein expression and localization in the human astroglioma cell line U373 MG (47). Western blot analysis using specific antibodies directed against S100B, S100A6, and S100A11 revealed that in this human tumor cell line the three S100 species are co-expressed (Fig. 6A), while S100A1 is not expressed in these cells (data not shown). A strong induction of S100B is observed when cells reach confluence. This observation corroborates a previous study showing that, in the rat glioma C6 cells, S100B synthesis also correlates with cell contact inhibition of growth (13). As noted for S100B, a drastic increase in S100A6 and S100A11 content occurs when U373 MG cells reach confluence. The expression of S100A6 and S100A11 was also seen to depend on cell contact. A recent work reports elevated S100A11 expression in growth arrested fibroblast at confluence (48). The overexpression of S100 proteins in confluent glial cells is not caused by from the age of the cultures. S100 protein expression decreases if cells are trypsined and grown again at low cell density (data not shown). ␣-tubulin staining demonstrated that equivalent amounts of protein were present in each lane (Fig. 6A, bottom panel). Western blot analysis also showed that the antibodies used are highly specific for each individual S100 isoform. These highly specific antibodies were used to compare the subcellular localization of S100B with S100A6 or S100A11 by indirect double immunofluorescence analysis of exponentially growing (Fig. 7A) or confluent U373 MG cells (Fig. 7B). In the exponential growth phase, the intensity and the subcellular localization of the S100B immunoreactivity were heterogeneous. Some cells were totally devoid of S100B immunoreactivity. Only few cells showed high S100B immunofluorescence signal, which partitioned between cytoplasm and nucleus (Fig. 7A, b and e, and  arrowheads). The immunostaining signals for S100A6 (Fig. 7A, c) and S100A11 (Fig. 7A, f) were more homogenous and also partitioned between cytoplasm and nucleus (Fig. 7A, d, arrowheads). Double immunofluorescence analysis showed that S100B co-localizes with S100A6 and S100A11 in cytoplasm at the membrane ruffling level of isolated cells (Fig. 7A, a, b, c,  and d, arrows). 4. The C-terminal domain of S100B is not required for the specificity of interaction with S100A6 or S100A11 isoforms. Activation of the histidine reporter was monitored by mating assay. Each L40 colony containing a LexA DNA binding domain protein construct (LexADBD fusion) was mated to each of the AMR70 containing a GAL4 activation domain construct (Gal4AD fusion) as described under "Experimental Procedures." The interactions were scored by growth on YC-UWLH medium lacking histidine. A, mating assay using S100B/A1Cter and S100A1/BCter chimeric proteins fused in frame with LexADBD. B, mating assay using S100B/A1Cter and S100A1/BCter chimeric proteins fused in frame with Gal4AD. L40 cells expressing LexADBD-lamin fusion protein and the AMR70 cells transformed with pGAD424 plasmid were used as control.
FIG. 5. S100B forms homodimeric and heterodimeric complexes with S100A1, S100A6, and S100A11 isoforms in COS-7 cells. Growing COS-7 cells were co-transfected with EGFP-S100B and Myc-S100 fusion plasmids for 24 h. Whole cell extracts were used for immunoprecipitation using either the anti-Myc antibody (MAbMyc) or the anti-MyoD antibody as a control. Immunoprecipitations were performed in the presence of either 2 mM EDTA (lanes Ϫ) or 0.5 mM CaCl 2 and 10 M ZnSO 4 (lanes ϩ) as detailed under "Experimental Procedures." Co-immunoprecipitated EGFP-S100B fusion proteins were analyzed by Western blot using an anti-GFP antibody. This antibody recognizes EGFP-S100B fusion protein in a crude extract of transformed COS-7 cells (lane EGFP-S100B). IP MycS100B, IP MycS100A1, IP MycS100A11, and IP MycS100B indicate the Myc-S100 isoform immunoprecipitated.
When U373 MG cells reach confluence, a general increase in immunofluorescence signal was observed for S100B, S100A6, and S100A11. Confocal microscopy analysis showed residual and diffuse nuclear staining and strong cytoplasmic immunoreactivity. Upon overlapping of the same field of U373 astrocytoma stained for S100B (red pixels) and S100A11 or S100A6 (green pixels), the resulting yellow color clearly demonstrated that S100 isoforms are mostly co-localized in the cytoplasm of quiescent cells (Fig. 7B, c and f).
We confirmed that S100B is capable of interacting physically with S100A6 in confluent U373 astrocytoma cells by co-immunoprecipitation assay (Fig. 6B). S100B was immunoprecipitated from cytosolic U373 cell extract, and the immunoprecipitates were analyzed by Western blot using anti-S100A6 antibodies. Significant amounts of S100A6 were found associated with the immunoprecipitated S100B. The amount of S100A6 recovered in immunoprecipitated S100B was higher in the presence of calcium and zinc than in the presence of EDTA (Fig. 6B, upper panel). This result suggests that the heterocomplex is stabilized in the presence of divalent ions. Experiments investigating S100B-S100A11 heterocomplex formation in U373 cells failed because in these cells, solubility of S100A11 is considerably reduced compared with S100A6. Preliminary results suggest that S100A11 is strongly associated with the membrane fraction. Recent work has shown that S100C is associated with early endosomal membrane, and this interaction is mediated through an interaction with annexin I in the presence of calcium (49). DISCUSSION Identification of the physiological S100B target proteins and characterization of their mode of interaction with S100B are essential to determine the exact contribution of the S100B-dependent signaling pathway to cell functions. In this study, we used the two-hybrid system to identify specific S100B target proteins. A large screen of a cDNA library from human brain using human S100B fused with the LexA DNA binding domain as bait allowed us to identify S100A6 and S100A11 as putative specific in vivo targets of S100B. The specificity of those interactions was further demonstrated by the observations that S100A1, which presents 56% identity at the amino acid level, is capable of interaction with S100B but does not interact with S100A6 or S100A11. Furthermore, no cross-interaction has been observed between S100A6 and S100A11. The interaction between S100B and S100A6 or S100A11 was confirmed by co-immunoprecipitation experiments using COS-7 cells and MG U373 astroglioma cells.
The first question raised by the isoform-specific interaction between S100B and S100A6 or S100A11 in yeast concerns the molecular mechanisms that govern such interactions. Comparison of ␤-galactosidase activities for qualitative evaluation of interactions (50) revealed differences in the strength of homoand heterointeraction between S100 species, suggesting that the mode of interaction between S100 species is not equivalent. The high avidity of interaction between S100B and S100A1 would correspond to the conventional S100 dimerization model revealed by the solution structure of apo-S100B, with helix I and helix IV flanking the calcium binding sites I and II as the major contributors to the dimer interface (24 -27, 29, 44). As expected, deletion of the extreme C terminus had no significant effect on S100B homodimerization and heterodimerization with S100A1. The apparent equivalent affinity for S100B homodimerization or heterodimerization with S100A1 is consistent with previous in vitro data showing that, after acid-dependent dissociation of the S100B and S100A1 homodimers, the proteins reassociate as homodimer (S100A1, S100B) or heterodimer (S100a) with the same probability (46). An equivalent affinity for S100B homodimerization or heterodimerization with S100A1 is also supported by co-immunoprecipitation experiments using COS-7 cells transfected with EGF-S100B and Myc-S100 fusion proteins. An equivalent amount of EGF-S100B protein is co-immunoprecipitated with Myc-S100B or Myc-S100A1 fusion proteins.
The avidity of interaction between S100B and S100A6 or S100A11 is much less pronounced than for the interaction between S100B with itself or with S100A1. Differences were confirmed by deletion or mutation within the C-terminal domain of S100B. Deletion of the C terminus on S100B or mutation of either Phe 87 or Phe 88 to Ala was found sufficient to abrogate interaction with S100A6 and S100A11. The structural change of S100B in the presence of Ca 2ϩ involves the exposure FIG. 6. Expression of S100B, S100A6, and S100B is co-regulated in astroglioma U373 MG cells. A, Western blot analysis of ␣-tubulin, S100B, S100A11, and S100A6 expression in whole cell extract from U373 MG culture cells at different confluence stages. Cells were lysed in Laemmli loading buffer, and total proteins (25 g) were immunoblotted using specific antibodies directed against ␣-tubulin (MAb␣-tub), S100B (RAbS100B), S100A11 (GPAbS100A11), and S100A6 (MAbS100A6). Days of culture and level of confluence are indicated above each lane. Recombinant S100B, S100A6, and S100A11 protein were used for reference. B, calcium and zinc stabilize the endogenous heterocomplex S100B-S100A6 in astroglioma U373 MG cells. Confluent U373 cells were lysed in 0.2% Triton X-100-TBS buffer in the presence of 2 mM EDTA (lanes Ϫ) or in the presence of 0.5 mM CaCl 2 and 10 M ZnSO 4 (lanes ϩ). The S100B was immunoprecipitated using a rabbit polyclonal antibody (RAbS100B) as described under "Experimental Procedures." A mouse monoclonal antibody against MyoD (MAbMyoD) was used as a control. Immunoprecipitated S100B and co-immunoprecipitated S100A6 were immunoblotted using an anti-S100B (RAbS100B) or an anti-S100A6 (MAbS100A6) antibody, respectively.
to solvent of the C terminus encompassing Phe 87 and Phe 88 (24,25,27,30). The C terminus, and specifically Phe 87 and Phe 88 , is implicated in interaction of S100B with target proteins (1). It is therefore possible that S100B interaction with S100A6 and S100A11 has similarity with interaction between S100B and target protein. Using chimeric S100B/A1Cter and S100A1/ BCter proteins, we also showed that the C terminus of S100B is not sufficient for specific interaction with S100A6 and S100A11. Multiple residues distributed throughout on helix II, helix III, and the hinge region of S100B, not involved at the dimer interface, are also implicated in interaction with target peptide (32). These residues most likely also contribute to interaction between S100B and S100A6 or S100A11. Altogether, these data suggest that the specific interaction between S100B and S100A6 or S100A11 should not be viewed as typical heterodimerization but rather as a model of interaction between S100B and target proteins. This model is also supported by the lack of interaction between S100A1 and S100A6 or S100A11. The specificity of interaction of S100B versus S100A1 with target protein has recently been reported in vitro (23). The stabilization of the S100B-S100A6 or S100B-S100A11 heterocomplexes by Ca 2ϩ /Zn 2ϩ (Figs. 5 and 6B; see Ref. 51) also supports such a model.
At this stage, some questions remain as to why any other proteins than S100 are not detected as S100B targets in the yeast two-hybrid system. It is noteworthy that several attempts to demonstrate interaction between S100B and two putative calcium-dependent target proteins, p53 and CapZ␣ (18,23), in the yeast two-hybrid system also proved unsuccessful (data not shown). Other studies also demonstrated the limitation of two-hybrid methodology in identifying S100 target proteins other than S100 isoforms (51)(52)(53). Two hypotheses to explain these observations can be suggested. It is possible that in the yeast two-hybrid, strict Ca 2ϩ -dependent interactions cannot be monitored. This could be due to tight calcium homeostasis regulation in yeast. Such a hypothesis is compatible with the fact that using calmodulin as bait, only calciumindependent targets have been detected (39,54,55). Calciumindependent interactions between S100B and S100A6 or S100A11 have been observed in co-immunoprecipitation exper-FIG. 7. The S100B co-localizes with S100A11 and S100A6 in astroglioma U373 MG cells. U373 MG cells cultures were fixed and double-stained as described under "Experimental Procedures" using the same primary antibodies as in Fig. 6A, conventional microscopy analysis of growing U373 cells after 5 days of culture shows that S100B (b and e), S100A11 (a), and S100A6 (d) are co-localized in cytoplasm and membrane ruffling (arrowheads) of U373 cells. The arrows indicate cells with strong nuclear staining. The same microscopic fields are shown in a, b, c, and in d, e, and f. The nuclei are stained using Hoechst reagent (c and f). Bar, 25 m. B, confocal microscopy analysis of quiescent U373 cells after 15 days in culture shows that S100B (b and e), S100A11 (a), and S100A6 (d) are essentially co-localized in the cytoplasm of U373 cells. The same microscopic fields are shown in a-c, and in d-f. Fields c and f correspond to the superimposition of a (green) and b (red) and of d (green) and e (red), respectively. Bar, 6 m.
iments, and Ca 2ϩ only strengthens the interactions (Figs. 5 and 6B). Interaction between S100A4 and S100A1 could also be detected in the absence of calcium (53). A second possible explanation is that the native conformation of S100B is affected when fused with LexA DNA binding domain, preventing interaction with calcium-dependent target proteins.
In vitro, purified S100B and S100A1 proteins exist as homodimers but do not spontaneously associate as heterodimers, unless the proteins are first denatured or the quaternary structure is destabilized upon calcium-binding (46). In contrast, when expressed in yeast, S100B and S100A1 associate as heterodimers. The same phenomenon has been reported with S100A8 and S100A9 (52). The dynamic equilibrium that seems to exist between monomeric and dimeric forms in living cells indicates that the biologically active form of S100 proteins is probably not restricted to homodimers. The S100 proteins are also capable of interaction with each other (present study, and see Refs. 51 and 53). The physical interaction between S100B and S100A6 or S100A11 might also have physiological implications. The co-expression of S100 species in the same cells should not be viewed as strictly coincidental. In normal brain, a co-localization of S100B with S100A6 and S100C has been reported in astrocytes, ependymal cells, and Schwann cells (56,57). We have confirmed the co-expression of S100B, S100A6, and S100A11 in primary cultures of rat brain astrocyte by reverse transcription-PCR and Western blot analysis (data not shown). It is also significant that misregulation of S100B, S100A6, and S100A11 expression is also often observed in tumors, including astrocytic tumors (58) and melanoma (11). We have shown in this study that in the human astrocytoma cell line MG U373, S100B, S100A6, and S100A11 are also co-expressed. In exponentially growing cultures, the three S100 protein species co-localize within membrane ruffling-like structures. Significant S100 immunoreactivity is also present in the cell nuclei (Fig. 7). When MG U373 cells reach confluence, the S100 protein expression is drastically up-regulated in a concerted fashion, and the proteins accumulate within the cell cytoplasm. The co-expression of the S100B, S100A6, and S100A11 in the same cells, together with the ability of the individual proteins to interact with each other, suggests that they might interfere with the function of their related proteins. Although the exact functions of the individual S100 proteins are still unknown, there are suspicions that they might be involved in different cell regulation pathways. As an example, the calcium-dependent interaction of S100B with the p53 protein is specific and has not been observed with S100A6 (23). We hypothesize that interaction between S100A6 or S100A11 and S100B could modulate binding of S100B to target proteins. Hence, S100A6 or S100A11 could behave as regulator of the S100B-dependent signaling pathway.