Oncogene (2000) 19, 4385 ± 4395 ã 2000 Macmillan Publishers Ltd All rights reserved 0950 ± 9232/00 $15.00 www.nature.com/onc CAIR-1/BAG-3 forms an EGF-regulated ternary complex with phospholipase C-g and Hsp70/Hsc70

Howard Doong1, John Price1, Young Sook Kim1, Christopher Gasbarre1, Julie Probst1, Lance A Liotta1, Jay Blanchette1, Kathryn Rizzo1 and Elise Kohn*,1

1Molecular Signaling Section, Laboratory of Pathology, National Cancer Institute, Bethesda, Maryland, MD 20892, USA

CAIR-1/BAG-3 forms an EGF-regulated ternary com- Introduction plex with Hsp70/Hsc70 and latent phospholipase C-g (PLC-g). The expression of CAIR-1, CAI stressed-1, Understanding the crosstalk between cellular signal was induced in A2058 human melanoma cells by pathways is critical to decipher the regulation of continuous exposure to CAI, an inhibitor of nonvol- cellular physiology. Intracellular calcium concentra- tage-gated calcium in¯ux. CAIR-1 sequence is identical, tions are tightly regulated and are altered during save 2 amino acids, to BAG-3 also cloned recently as transmembrane signaling events, such as those stimu- Bis, a member of the bcl-2-associated athanogene family. lated by growth factors, and also in response to cellular We show that CAIR-1/BAG-3 binds to Hsp70/Hsc70 in injury or stress (Berridge et al., 1998; Bukau and intact cells and this binding is increased by short term Horwich, 1998; Zwick et al., 1999). We have identi®ed exposure to CAI (P50.007). CAIR-1/BAG-3 is phos- an inhibitor of calcium in¯ux in nonexcitable cells, phorylated in vivo in the absence of stimulation. Basal CAI (Felder et al., 1991; Gusovsky et al., 1993; Kohn phosphorylation is inhibited by treatment with d-erythro- et al., 1994a). Exposure to CAI causes tumor and sphingosine (d-ES), a broad inhibitor of the protein endothelial cytostasis and inhibits cell attachment, C family. CAIR-1/BAG-3 contains several PXXP migration, and angiogenesis and tumorigenesis in vitro SH3 binding domains leading to the hypothesis that it is and in vivo (Kohn et al., 1992, 1994a,b, 1995; Kohn a partner protein of phospholipase C-g. PLC-g is bound and Liotta, 1990). We hypothesized that continuous to CAIR-1/BAG-3 in unstimulated cells. It is increased exposure to CAI by altering stimulated and basal by CAI or d-ES (P=0.05) treatment, and abrogated by intracellular calcium concentrations would stress the EGF (r2=0.99); d-ES treatment blocks the EGF- tumor cells. This would then either stimulate a stress mediated dissociation. We show that CAIR-1/BAG-3 recovery response or alter regulation or expression of binds to PLC-g and Hsp70/Hsc70 through separate and proteins required for calcium activation pathways, such distinct domains. Hsp70/Hsc70 binds to the BAG as the phospholipases. Activated phospholipases C domain of BAGs-1 and -3. CAIR-1/BAG-3 from control produce inositol trisphosphate, a potent second and EGF-treated cell lysates bound selectively to the messenger inducing internal calcium release and SH3 domain of PLC-g, but not its N-SH2 or C-SH2 subsequent transmembrane calcium in¯ux (Berridge et domains. Con®rming the SH3 interaction, PLC-g was al., 1998). We developed a stress model in which tumor pulled down by CAIR-1/BAG-3 PXXP-GST fusions, but cell sublines were isolated after long term continuous GST-PXXP constructs confronted with lysates from exposure to increasing concentrations of CAI. We now EGF-treated cells did not bind PLC-g as was seen in report cloning of CAIR-1, CAI stressed-1, a protein intact cells. Hsp70/Hsc70 was brought down by the nearly identical to BAG-3 (Takayama et al., 1999) PLC-g SH3 construct equally from native and EGF- from CAI-stressed cells, and demonstrate novel regula- treated cells, but did not bind the PXXP construct under tion and function of CAIR-1/BAG-3. either condition. We propose that CAIR-1/BAG-3 may BAG-3 was cloned from a screen for homologs of act as a multifunctional signaling protein linking the the bcl-2 binding protein BAG-1 (bcl-2-associated Hsp70/Hsc70 pathway with those necessary for activa- athanogene) family (Takayama et al., 1995, 1999). tion of the EGF receptor tyrosine kinase signaling BAG-1 was identi®ed in a screen for bcl-2 binding pathways. Oncogene (2000) 19, 4385 ± 4395. proteins and has been shown to inhibit apoptosis and promote tumorigenesis (Stuart et al., 1998; Takayama Keywords: calcium; EGF; phospholipase C; Hsp70/ et al., 1995, 1997). Recent homology cloning Hsc70; signal transduction identi®ed fragments of at least four other BAG family members, all of which contain a conserved C- terminal domain, the BAG domain (Takayama et al., 1999). There is little other homology between BAG-1 and BAG-3, including lack of the BAG-1L nuclear localization signal or the BAG-1 ubiquitin region. The BAG domain was shown by pull down experiments and functional assays to bind to and promote substrate release from both HSP-70 and HSC-70 (Stuart et al., 1998; Takayama et al., 1997, *Correspondence: EC Kohn, 10 Center Drive, 10/2A33, Bethesda, 1999). Direct binding of BAG-1 to bcl-2 could not be MD 28092-1500, USA Received 21 January 2000; revised 21 June 2000; accepted 11 July demonstrated upon direct confrontation, however 2000 addition of ATP markedly enhanced BAG-1/bcl-2 An EGF-regulated complex of CAIR-1/BAG-3, PLC-g and Hsc70 H Doong et al 4386 binding, suggesting the requirement for an HSP-70 from a K; 237Q from an R) from BAG-3, a member intermediate (Takayama et al., 1997). HSP-70 and its of the bcl-2-associated athanogene family of bcl-2 constitutive form, HSC-70, are molecular chaperones binding proteins (Takayama et al., 1995, 1999). activated by varied cellular stresses including heat Genomic clones from chromosome 10q25 con®rm shock, oxygen-free radicals, transition heavy metals, the CAIR-1 sequence. Northern analysis revealed a in¯ammation, ischemia, anti-neoplastics, oncogenes, twofold net increase in expression of CAIR-1/BAG-3 and proto-oncogenes (Morimoto, 1998). They bind in cells stressed by constant exposure to 10 ± 30 mM ATP through an ATPase domain and have a CAI (10 ± 30R, Figure 1b) but not in cells exposed to separate substrate binding domain (Bukau and 10 mM CAI for 424 h (Figure 1c). Domain analysis Horwich, 1998; Pilon and Schekman, 1999). The site of CAIR-1/BAG-3 reveals the previously reported of Hsp70/Hsc70 binding to BAG-1 has been de®ned BAG region (Takayama et al., 1999) and WW as the ATPase domain (Bukau and Horwich, 1998; domain, as well as seven putative protein kinase C Takayama et al., 1997); a 1 : 1 molar binding results (PKC) and ®ve putative CKII phosphorylation sites, in reduction in the refolding function of HSP-70 multiple potential tyrosine phosphorylation sites, and (Stuart et al., 1998). While other functions for BAG- a series of proline-rich PXXP repeats of the SH3 1 such as binding to and activating Raf-1 have been binding type (Pawson and Scott, 1997). No genomic reported, no link between those functions and HSP- ampli®cation of CAIR-1/BAG-3 gene in the CAI- 70 have been demonstrated. CAIR-1/BAG-3 was stressed cells was demonstrated by Southern analysis recently cloned also as bis (Lee et al., 1999), in a or ¯uorescent in situ hybridization (data not shown). protein interaction cloning procedure using bcl-2 as Broad tissue expression of CAIR-1/BAG-3 was found bait. Lee and coworkers demonstrated a weak anti- in adult and embryonic tissues with lowest expression apoptotic activity of transfected bis and further seen in hematopoietic tissues (Figure 1d). Genomic showed synergy when suboptimal doses of both bis conservation was limited to the higher eukaryotes by and bcl-2 were transfected, simultaneously. They zoo blot (not shown, Bios, New Haven, CT, USA). localized the site of interaction of bis to the BH1 Immunoprecipitation from A2058 cells followed by domain of bcl-2. immunoblot with anti-peptide antibodies to CAIR-1/ Phospholipase C-g (PLC-g) regulates cytosolic free BAG-3 recognized a 74 kDa protein which was calcium concentration in response to transmembrane increased in quantity in 20 mM CAI-stressed cells signal transduction, and has been linked to malig- (Figure 1e). No similar induction of CAIR-1/BAG-3 nant transformation, invasive potential, and angio- protein was observed in wild type A2058 or MDA-435 genesis (Chang et al., 1997; Khoshyomn et al., 1999; human breast cancer cells upon exposure to 10 mM Smith et al., 1998; Turner et al., 1996, 1997; Yang CAI for up to 24 h (Figure 1f), consistent with the et al., 1998). Several investigators have demonstrated lack of increased gene expression with acute CAI transforming capacity of both full length PLC-g and exposure. the isolated SH2-SH2-SH3 moiety (Bar-Sagi et al., 1993; Chang et al., 1997; Schlessinger, 1994; Smith CAIR-1/BAG-3 is phosphorylated in vivo et al., 1998). Its link to calcium homeostasis, transmembrane signal transduction, and transforming CAIR-1/BAG-3 has not been demonstrated pre- potential made it a putative target protein in the viously to be a phosphoprotein. Domain analysis calcium in¯ux stress model. We demonstrate that of CAIR-1/BAG-3 indicated the presence of poten- CAIR-1, which is increased in expression in CAI- tial PKC and CKII phosphorylation sites, as well as treated cells, is a partner protein for PLC-g,and putative tyrosine phosphorylation sites. In vivo forms a ternary complex with HSP-70. This provides phosphorylation demonstrated that CAIR-1/BAG-3 a previously unknown link between transmembrane is phosphorylated in unstimulated wild type A2058 signaling and the HSP-70 system. cells and unstimulated MDA-435 human breast cancer cells (Figure 2). No change in in vivo phosphorylation of CAIR-1/BAG-3 was found after cell treatment with PMA (100 ng/ml) for 30 min to Results stimulate endogenous PKC or 24 h to down-regulate PMA-sensitive classical and novel PKC isotypes Identification and cloning of CAIR-1/BAG-3 (Figure 2a) (Hata et al., 1993; Nishizuka, 1988). A2058 human melanoma cells were exposed chroni- Incubation with a broad spectrum inhibitor of all cally to continuing and escalating quantities of the classes of PKC isotypes, d-ES (10 mg/ml for 2 h), calcium entry blocker, CAI. A subline previously markedly reduced CAIR-1/BAG-3 incorporation of shown to be sensitive in vitro and in vivo to the anti- 32P-o-phosphate without e€ecting total cellular invasive and anti-tumorigenic activity of CAI was CAIR-1/BAG-3 (Figure 2b). No tyrosine phosphor- used as the control for subtractive hybridization. A ylation of CAIR-1 was demonstrated in unstimulated chemical cross-link solution hybridization protocol A2058 or MDA-435 human breast cancer cells. EGF was used as described to minimize loss of low copy treatment of MDA-435 cells, which did not e€ect number transcripts (Hampson et al., 1992) and yielded the quantity of CAIR-1, resulted in tyrosine 35 non-overlapping clones. A clone containing the 3' phosphorylation of CAIR-1/BAG-3 (Figure 2c). 1200 bp was upregulated by approximately threefold CAIR-1/BAG-3 remained restricted to the cytosol and was chosen for further analysis. CAIR-1, CAI- upon EGF exposure. A small but statistically stressed-1, was cloned as a 2.6 kb transcript encoding signi®cant e€ect of CAI (10 mM, 2 h) was seen on a 74 kDa protein (Figure 1a), found by GenBank CAIR-1/BAG-3 basal phosphorylation in MDA-435 BLAST analysis to di€er by 2 amino acids (227Q cells (87+4% of control, P50.05), along with the

Oncogene An EGF-regulated complex of CAIR-1/BAG-3, PLC-g and Hsc70 H Doong et al 4387 A

D

Figure 1 CAIR-1/BAG-3 sequence and expression. (a) CAIR-1/BAG-3 sequence. Putative functional domains are marked: BAG domain (dotted), WW region (dashed), target tyrosine phosphorylation sites (starred), and 4 PXXP regions (bold underline). The amino acids di€ering from BAG-3 are boxed. GenBank accession number: AF071218. (b) Induction of CAIR-1/BAG-3 by constant CAI stress. Northern analysis of RNA from A2058 cells chronically stressed in 10 ± 30 mM (10R ± 30R) CAI were probed for CAIR- 1/BAG-3 under stringent conditions, exposed to ®lm and assessed by densitometry using a GAPdH control (n=6). (c) Lack of gene induction by short exposure to CAI. Cells were treated with CAI 10 mM for up to 24 h and analysed for CAIR-1/BAG-3 expression by Northern analysis with GAPdH used for a loading control. (d) Tissue expression. A commercial multiple tissue Northern blot was probed for CAIR-1/BAG-3 expression with GAPdH control. (e) CAIR-1/BAG-3 protein is increased in CAI conditioned cells. Whole cell bu€er A lysates of A2058 cells were analysed by immunoblot with anti-CAIR-1/BAG-3 Ab2. Pep lane demonstrates blockade of immunodetection by preincubation of Ab2 with cognate peptide 2. No e€ect was seen with peptide 8 competition of Ab2 (not shown). WT=control A2058 cells; 20R=20 mM CAI conditioned A2058 cells. (f) No change in CAIR-1/BAG-3 protein in A2058 or MDA-435 cells with short exposure to CAI. Cells were exposed to 10 mM CAI for the indicated times and lysed for immunoblot expected reduction in basal phosphorylation in vivo binding of CAIR-1/BAG-3 to Hsp70/Hsc70 and response to d-ES (22+6% of control, P50.0008). further show its regulation by CAI treatment. An An increase in in vivo incorporation of 32P-o- antibody recognizing both molecular forms HSP-70 phosphate could be detected after 1 min exposure and HSC-70 was used for the CAIR-1/BAG-3 studies. to EGF (100 ng/ml; P50.005, n=3; Figure 2d). CAIR-1/BAG-3 co-immunoprecipitated Hsp70/Hsc70 under unstimulated conditions from MDA-435 cells (Figure 3). Treatment with 10 mM CAI for 90 min, a In vivo binding of CAIR-1/BAG-3 to Hsp70/Hsc70 is time that does not alter expression of either CAIR-1/ increased by CAI exposure BAG-3 or Hsp70/Hsc70, resulted in a 207% increase in BAG-1 and BAG-3 have been shown through ex vivo bound Hsp70/Hsc70 to CAIR-1/BAG-3 (P50.007, binding experiments to bind to HSP-70 (BAG-1 only) n=5). Nor did d-ES exposure signi®cantly change and HSC-70 (both) through the newly described BAG Hsp70/Hsc70 ± CAIR-1/BAG-3 complex formation domain (Takayama et al., 1999). We demonstrate in (Figure 3b).

Oncogene An EGF-regulated complex of CAIR-1/BAG-3, PLC-g and Hsc70 H Doong et al 4388

Figure 2 CAIR-1/BAG-3 is a phosphoprotein. (a) PMA-independent in vivo phosphorylation of CAIR-1/BAG-3. Unstimulated A2058 cells were phosphate-starved then incubated with 32P-o-phosphate in 0.1% DMSO vehicle or PMA 100 mg/ml for either 30 min or 24 h. Cells were lysed in bu€er A then subjected to immunoprecipitation as described in Materials and methods. (b) d-ES inhibits in vivo phosphorylation of CAIR-1/BAG-3. Unstimulated A2058 cells were treated with d-ES (10 mg/ml) for the ®nal 2 h of the 32P-o-phosphate incubation, then lysed and analysed. The bottom panel shows immunoblot for CAIR-1/BAG-3 quantity during d-ES exposure. (c) CAIR-1/BAG-3 is tyrosine phosphorylated upon EGF simulation. MDA-435 cells were exposed to EGF 50 ng/ ml for 10 min and then fractionated into cytosol and membrane components. Each fraction was subjected to immunoprecipitation with anti-phosphotyrosine, 4G10, and immunoblotted for CAIR-1/BAG-3. (d) CAI slightly but signi®cantly decreases and EGF increases in vivo phosphorylation of CAIR-1/BAG-3 in MDA-435 cells. Cells were prepared and immunoprecipitated as described. EGF does not induce CAIR-1 protein within a 5 min exposure (data not shown). Lane 1: DMSO control; 2: CAI 10 mM, 2 h; 3: d- ES 10 mg/ml, 2 h; 4: EGF 100 ng/ml, 1 min; 5: EGF 100 ng/ml, 5 min

association of PLC-b with CAIR-1/BAG-3 was found CAIR-1/BAG-3 is a binding protein for latent PLC-g by coimmunoprecipitation in MDA-435 cells under PLC-g was chosen as a putative CAIR-1/BAG-3 standard or high stringency RIPA lysate conditions binding protein because of its potential involvement (Figure 4c), indicating a selectivity of CAIR-1/BAG-3 in the response to CAI stress, because its activation has for PLC-g. been shown to be CAI-sensitive (Gusovsky et al., 1993), and because it contains an SH3 domain to EGF stimulation results in rapid dissociation of PLC-g which the CAIR-1/BAG-3 PXXP regions could target. from CAIR-1/BAG-3 Unstimulated A2058 and MDA-435 cells were used initially to determine if CAIR-1 interacted with PLC-g We hypothesized that phosphorylation of CAIR-1/ independently of growth factor receptor induced BAG-3 would regulate its binding to PLC-g. EGF interactions. Untreated cells were lysed under high treatment of MDA-435 cells induced a dose-dependent stringency conditions (Bu€er A adjusted to 300 mM reduction of the binding of PLC-g to CAIR-1/BAG-3 NaCl with added 0.1% Triton X-100 and 0.1% SDS) (Figure 5a). The EGF dose response for this 2 and subjected to immunoprecipitation with anti-CAIR- dissociation is log-linear (r =0.99) with an EC50 of 1/BAG-3 Ab 2 or Ab 8. Immunocomplexes were 55 ng/ml. EGF induces a rapid loss of PLC-g binding detected by immunoblot for PLC-g (Figure 4a). Similar from CAIR-1/BAG-3, with almost all loss of binding results were seen after precipitation with anti-PLC-g within the ®rst minute (Figure 5b). The potential antibody and blotting with Ab2 or Ab8; approximately tyrosine phosphorylation sites in the BAG domain 3 ± 5% of total cellular PLC-g is bound to CAIR-1/ may regulate CAIR-1 Hsp70/Hsc70 complex forma- BAG-3 under basal conditions (data not shown). No tion. However, no signi®cant e€ect of EGF treatment

Oncogene An EGF-regulated complex of CAIR-1/BAG-3, PLC-g and Hsc70 H Doong et al 4389

Figure 3 CAIR-1/BAG-3 binds Hsp70/Hsc70. Anti-Hsp70/ Hsc70, recognizing both the constitutive and inducible forms of Hsp70, was used for these studies. MDA-435 cell Bu€er A lysates from control, CAI (10 mM CAI, 90 min), or d-ES (10 mg/ml, 2 h) were precleared and then immunoprecipitated with Ab2 and immunoblotted with anti-Hsp70/Hsc70. (a) CAI increases CAIR- 1/BAG-3 binding with Hsp70/Hsc70. Lane 1: Untreated cell lysate control; 2: CAI; 3: lysate Western control. (b) No e€ect of d-ES on CAIR-1/BAG-3 binding of Hsp70/Hsc70. Lane 1: control rabbit IgG; 2: control lysate; 3: d-ES on this complex was observed (n=4, Figure 5c). Since d-ES alters the basal phosphorylation status of CAIR- 1/BAG-3, we tested its ability to alter binding to PLC- g. Cell treatment with d-ES caused an approximately 150% increase in PLC-g binding to CAIR-1/BAG-3 in unstimulated cells (P50.05) and abrogated the statistically signi®cant (P50.05) EGF-mediated dis- sociation of PLC-g binding to CAIR-1/BAG-3 Figure 4 CAIR-1 is a PLC-g binding protein. Whole cell lysates (P=0.49 control vs dES/EGF, Figure 6a). Treatment were obtained using high stringency Bu€er B. Lysates from of MDA-435 cells with d-ES did not e€ect the unstimulated A2058 or MDA-435 cells were immunoprecipitated quantity of EGF receptor protein or the tyrosine with Ab2 or Ab8 as described in Materials and methods and probed for PLC-g.(a) A2058 cells. Lane 1: PLC-g immunoblot phosphorylation status of the EGF receptor under the control; lanes 2 and 3: control IgG or Ab8 immunoprecipitation conditions used for the co-immunoprecipitation stu- probed for PLC-g.(b) MDA-435 cells. Immunoprecipitates with dies. CAI (10 mM) incubation for 90 min also control rabbit IgG (lane 1) or Ab8 (lane 2) were probed for PLC- increased binding of PLC-g to CAIR-1/BAG-3 with- g. Similar results were seen with immunoprecipitation for PLC-g and immunoblot with CAIR-1 antibodies Ab2 (not shown). (c) out altering quantity of either protein (P50.05, n=4, CAIR-1 does not coimmunoprecipitate PLC-b. Lysates from Figure 6b). MDA-435 cells were immunoprecipitated as indicated and then probed for PLC-b (upper panel) or PLC-g (lower panel) in sequentially stripped blots (n57) CAIR-1/BAG-3 binds to PLC-g and Hsp70/Hsc70 through separate and distinct domains CAIR-1/BAG-3 binds to the PLC-g SH3 domain al., 1995, 1997, 1999). No whole cell in vivo co- PLC-g can interact with partner proteins through precipitation of PLC-g and Hsp70/Hsc70 was ob- multiple domains, including its two SH2 domains and served (data not shown). Reprobing of the PLC-g- an SH3 domain (Koch et al., 1991; Pawson and SH3 domain binding blot for Hsp70/Hsc70 demon- Scott, 1997). CAIR-1/BAG-3 was pulled down from strated that the PLC-g-SH3 construct pulled down unstimulated MDA-435 cell lysates by the GST-PLC- Hsp70/Hsc70 from unstimulated and EGF-treated cell g-SH3 domain but not the PLC-g-N-SH2 or C-SH2 lysates equally (Figure 7c). domains (Figure 7a). Treatment with EGF of the cells donating CAIR-1/BAG-3 did not alter CAIR-1/ PLC-g binds to the CAIR-1/BAG-3 PXXP domain BAG-3 binding to the PLC-g-SH3 domain, in Pull down experiments were done using the PXXP4 contrast to what was observed with whole cell region of CAIR-1/BAG-3, a region containing four coimmunoprecipitation studies where EGF treatment di€erent PXXP repeats (Figure 1) of SH3 binding reduced binding of CAIR-1/BAG-3 to PLC-g in type (Koch et al., 1991; Pawson and Scott, 1997). whole cell lysates (Figure 7b). No binding to SH2 This region does not include the BAG domain. This domain constructs was seen after EGF stimulation. was done to determine if this region is the PLC-g Hsp70/Hsc70 has been reported to bind to BAG-1 binding site on CAIR-1/BAG-3 and thus the and BAG-3 through the BAG domain (Takayama et potential mechanism through which PLC-g associates

Oncogene An EGF-regulated complex of CAIR-1/BAG-3, PLC-g and Hsc70 H Doong et al 4390

Figure 5 EGF causes reduction of PLC-g binding to CAIR-1/BAG-3. MDA-435 cells, treated as indicated were lysed in bu€er B and subjected to immunoprecipitation with Ab2 and probed for PLC-g as described. (a) EGF dose-dependent reduction in PLC-g ± CAIR-1/BAG-3 binding. MDA-435 cells were treated with increasing doses of EGF for 10 min then analysed by coimmunoprecipitation studies. Blots were analysed by densitometer and results plotted. A log linear relationship is shown (r2=0.99). A representative gel of control rabbit IgG (7) and EGF (0 ± 200 ng/ml) treatment is shown. (b) Time course of release of PLC-g by CAIR-1/BAG-3. Cells were treated with EGF (100 ng/ml) from 0 ± 30 min and subjected to coimmunoprecipitation of CAIR-1/BAG-3 and PLC-g and for direct PLC-g immunoblot alone. (c) No signi®cant e€ect of EGF on Hsp70/Hsc70 binding to CAIR-1/BAG-3. Cells were exposed to EGF, 100 ng/ml for the indicated times, lysed, and subjected to immunoprecipitation with Ab2. The upper blot was probed for PLC-g and the lower blot for Hsp70/Hsc70. Controls include the rabbit IgG immunoprecipitation control (7, ®rst lane) and a simple Western blot (7, last lane)

with Hsp70/Hsc70. A second construct containing the EGF-treated lysates to the CAIR-1/BAG-3 PXXP4 three C-terminal PXXP repeats was also generated domain was found. Further, no Hsp70/Hsc70 could and tested. Both constructs bound PLC-g from be detected in the GST-PXXP4 construct experiments unstimulated MDA-435 cell lysates (Figure 8a) (Figure 8b). These data indicate a novel role for demonstrating that the PLC-g binding site on CAIR-1/BAG-3 as a binding partner of latent PLC-g, CAIR-1/BAG-3 is a functional PXXP SH3 binding that PLC-g binding to the CAIR-1/BAG-3 SH3 site. In contrast to the PLC-g-SH3 experiments and binding domain is regulated by EGF treatment, and consistent with the whole cell immunoprecipitation that PLC-g and Hsp70/Hsc70 bind CAIR-1/BAG-3 experiments, no or minimal binding of PLC-g from through di€erent domains.

Oncogene An EGF-regulated complex of CAIR-1/BAG-3, PLC-g and Hsc70 H Doong et al 4391

Figure 6 Regulation of PLC-g binding to CAIR-1/BAG-3. (a) d-ES increases PLC-g - CAIR-1/BAG-3 complex formation. Cells were treated with d-ES (10 mg/ml for 2 h) with or without EGF (200 ng/ml) added for the ®nal 10 min prior to lysis. Densitometer analysis was done for quantitation (d-ES: P50.05; EGF release: P50.05; control vs dES/EGF: P=0.49). (b) CAI increases PLC-g complex formation with CAIR-1/BAG-3. MDA-435 cells were incubated with CAI (10 mM) for 90 min prior to lysis and immunoprecipitation. No e€ect of CAI on CAIR-1 quantity or PLC-g quantity was seen (data not shown)

Discussion CAIR-1/BAG-3 is a protein containing multiple protein interaction motifs and phosphorylation sites We demonstrate that CAIR-1/BAG-3 forms a ternary (Figure 1), in addition to a conserved BAG domain complex between latent PLC-g and Hsp70/Hsc70. This (Takayama et al., 1999). We have focused on the is the ®rst demonstration of a regulated crosstalk PXXP region, a putative SH3 binding domain, and between Hsp70/Hsc70 and the EGFR-PLC-g signaling have demonstrated that this region binds the latent pathway. The lack of binding of Hsp70/Hsc70 to the form of PLC-g. There is a rapid reduction in PLC-g CAIR-1/BAG-3 PXXP PLC-g binding domain suggests binding to CAIR-1/BAG-3 seen after cellular EGF that PLC-g and Hsp70/Hsc70 interact with CAIR-1/ treatment. Release of PLC-g from CAIR-1/BAG-3 is BAG-3 at distinct, non-overlapping locations on the rapid, nearly complete by 1 min. The kinetics of PLC-g CAIR-1/BAG-3 molecule. Partner protein binding of phosphorylation as previously reported (Wahl et al., both CAIR-1/BAG-3 with PLC-g and CAIR-1/BAG-3 1988; Yang et al., 1994, 1998) and con®rmed by us in with Hsp70/Hsc70 is increased by CAI, the nonvoltage this model system shows the peak of phosphorylation gated calcium in¯ux inhibitor used to stress cells for occurring between 1 and 2 min. Maintenance of PLC- cloning of CAIR-1/BAG-3; d-ES, the PKC inhibitor, g ± CAIR-1/BAG-3 binding is seen when a wild type increased CAIR-1/BAG-3 binding to PLC-g but not to PLC-g SH3 construct is used but a marked reduction Hsp70/Hsc70. Release of PLC-g from CAIR-1/BAG-3 in partner protein complex is observed when the is stimulated by EGF treatment but abrogated by CAIR-1/BAG-3 PXXP region is presented with lysates exposure to d-ES. CAIR-1/BAG-3 binds in vivo and in from EGF-treated cells. This implicates a change(s) in vitro to PLC-g from lysates from unstimulated cells but the conformation of the PLC-g SH3 domain is not from EGF treated cells whereas wild type PLC-g necessary for the process of CAIR-1/BAG-3 release SH3 pulls CAIR-1/BAG-3 from both unstimulated and of PLC-g upon EGF stimulation. EGF-treated cells. This suggests that the PLC-g SH3 The BAG domain of CAIR-1/BAG-3 contains two domain is a target site in the EGF receptor signaling tyrosine and three threonine residues that are not pathway and argues against tyrosine phosphorylation conserved from the BAG domain of BAG-1 (Takaya- of CAIR-1/BAG-3 as playing a role in the release of ma et al., 1999). We demonstrate endogenous binding PLC-g after EGF treatment. These data are consistent of CAIR-1/BAG-3 to Hsp70/Hsc70 and show an with the concept that CAIR-1/BAG-3 is a stimulus to increase in complex formation when cells are stressed bring Hsp70/Hsc70 into proximity of PLC-g to allow for 90 min by exposure to 10 mM CAI. Previous studies or facilitate the change in the PLC-g SH3 domain and have shown a BAG-3 interaction with Hsp70/Hsc70 subsequent release from CAIR-1/BAG-3 (Figure 9). using BAG-GST pull down experiments, from BAG-3

Oncogene An EGF-regulated complex of CAIR-1/BAG-3, PLC-g and Hsc70 H Doong et al 4392

Figure 7 CAIR-1/BAG-3 binds to the PLC-g SH3 domain and coprecipitates Hsp70/Hsc70. Bu€er A MDA-435 cell lysates were used for the following studies. (a) CAIR-1/BAG-3 binds to the PLC-g SH3 domain selectively. Whole cell lysates were incubated with GST constructs containing the N- and C-SH2 domain or the SH3 domain of PLC-g and probed for CAIR-1/BAG-3 with Ab2. (b, c) No e€ect of EGF on CAIR-1/BAG-3 or Hsp70/Hsc70 binding to the PLC-g SH3 domain. Bu€er A lysates from cells treated with EGF (100 ng/ml, 1 min) were subjected to pull down experiments as described and probed for CAIR-1/BAG-3 (b) or Hsp70/ Hsc70 (c)

expressed in intact cells, and using a *24 kDa unstimulated cells is present so that Hsp70/Hsc70 may recombinant BAG-3 fragment (Takayama et al., assist in an EGF-stimulated PLC-g SH3 conforma- 1999). These studies demonstrated that the 24 kDa tional change resulting in release of PLC-g from CAIR- BAG-3 fragment had an attenuated binding to Hsc70 1/BAG-3. in a BIA-Core assay and that Hsc70 dissociated more A binding protein for latent PLC-g has not been rapidly from the BAG-3 fragment BIA-Core chip than recognized previously. The function of activated did the puri®ed BAG-1 or BAG-2 molecules. Our PLC-g is to hydrolyze phosphatidylinositol 4, 5- demonstration of endogenous binding of CAIR-1/ bisphosphate to inositol 1, 4, 5-trisphosphate and BAG-3 to Hsp70/Hsc70 con®rms that the interaction diacylglycerol, mediators of release of intracellular of CAIR-1/BAG-3 occurs in intact cells. The increase calcium and calcium in¯ux, and activation of protein seen after CAI exposure indicates that this protein kinase C, respectively. We ®nd that abrogation of interaction may be important to the cellular response calcium in¯ux with CAI and of PKC activity with to stress. It is possible that the ternary complex of d-erythrosphingosine augments CAIR-1/BAG-3 ± CAIR-1/BAG-3, PLC-g, and Hsp70/Hsc70 found in PLC-g complex formation and that d-erythrosphin-

Oncogene An EGF-regulated complex of CAIR-1/BAG-3, PLC-g and Hsc70 H Doong et al 4393

Figure 8 PLC-g binds to the PXXP domain of CAIR-1/BAG-3. Bu€er A whole cell lysates were used for CAIR-1/BAG-3 PXXP pull down experiments. Constructs containing the PXXP3 or PXXP4 region of CAIR-1/BAG-3 were subjected to binding with unstimulated or EGF-treated (100 ng/ml, 1 min) whole cell lysates. Complexes were probed for PLC-g (a) or Hsp70/Hsc70 (b)

a feedback mechanism to limit the availability of PLC-g for activation. That d-erythrosphingosine also regulates in vivo phosphorylation of CAIR-1/BAG-3 suggests that there may be an internal regulatory site in CAIR-1/BAG-3. Studies have shown that Ser/ Thr phosphorylation in the vicinity of an SH3 binding domain can regulate that binding (Chen et al., 1997; Hu and Bowtell, 1996; Pawson and Scott, 1997). Published examples show that SOS phosphor- ylation prevented or reduced its binding to Grb2, yielding the hypothesis that phosphorylation of Sos1 alters recognition or anity of the Grb2 N-SH3 domain. Our observations indicate that a phosphor- ylation event may increase binding in an opposite paradigm to that presented for the Grb2/SOS interaction. Xenobiotics and anti-cancer therapeutics have been well documented as cell stress inducers. We demon- strate that continuous long term exposure to CAI, an inhibitor of nonvoltage-gated calcium in¯ux and an anti-cancer agent in clinical trials, stimulates an increase in expression of CAIR-1/BAG-3, a binding partner of both latent PLC-g and Hsp70/Hsc70. The demonstration of a ternary complex and an obligate Figure 9 Model of proposed CAIR-1/BAG-3 protein binding change in the PLC-g SH3 domain required for release interactions of PLC-g from CAIR-1/BAG-3 upon growth factor stimulation implicates Hsp70/Hsc70 in the release process. Hsp70/Hsc70 is not released from CAIR-1/ gosine can overcome the release of PLC-g stimulated BAG-3 by growth factor treatment. These data suggest by EGF. These regulatory pathways may thus have that regulation of activation of PLC-g may begin with

Oncogene An EGF-regulated complex of CAIR-1/BAG-3, PLC-g and Hsc70 H Doong et al 4394 its docking site and not with growth factor receptor was tested by co-immunoprecipitation. Bu€er B lysates were activation and phosphorylation. pre-cleared with protein A/G beads incubated with control rabbit IgG, after which the supernatant was immunoprecipi- tated either with control antibody (puri®ed rabbit IgG) or Materials and methods anti-CAIR-1 Ab2 or Ab8. Immunocomplexes were captured, washed ®ve times with bu€er B, and assayed as above. Where Materials indicated, cell monolayers were exposed to CAI (10 mM for cDNA synthesis kits and protein A agarose beads were 90 min), d-erythrosphingosine (d-ES, 10 mg/ml for 2 h) or obtained from Gibco/BRL (Gaithersburg, MD, USA). The EGF prior to lysis. Unless otherwise indicated, results shown labeling kit was from Amersham (Arlington Heights, IL, are representative of n53 replicate experiments. USA) and all radionuclides from NEN-Dupont (Boston, MA, USA). Precast polyacrylamide gels were from Novex In vivo phosphorylation (San Diego, CA, USA). The multiple tissue Northern blot was from Clontech (Palo Alto, CA, USA). Anti-phosphotyr- Cells were grown to subcon¯uence and washed into osine monoclonal antibody (4G10), anti-PLC-g, anti-PLC-b phosphate-free, serum-free media for 4 h. 32P-o-phosphate antibodies, and GST-PLC-g-N-SH2, GST-PLC-g-C-SH2 and (62.5 mCi/ml) was added in fresh phosphate- and serum-free GST-PLC-g-SH3 were obtained from UBI (Lake Placid, NY, media and cells were incubated for a further 2 h and then USA). Anti-Hsp70/Hsc70 was from Santa Cruz Biotechnol- washed with iced PBS. Cells were lysed with bu€er A as ogy (Santa Cruz, CA, USA). pGEX 2T and 4T3 vectors were described above. Samples were precleared with protein A from Pharmacia Biotechnology (Piscataway, NJ, USA). The beads and then subjected to immunoprecipitation and BCA protein assay, protein A/G beads, secondary antibodies, electrophoresis, then gels were exposed to ®lm. Where and ECL kits were purchased from Pierce (Rockford, IL, indicated, cells were treated with d-ES (10 mg/ml) or CAI USA). EGF was from Collaborative Biomedical Products (10 mM) during the 32P-o-phosphate incubation for the 2 h (Bedford, MA, USA) and phosphate-free media from immediately prior to lysis, or incubated with 100 mg/ml Bio¯uids (Rockville, MD, USA). All other products were phorbol 12-myristate 13-acetate (PMA) for either 30 min or of molecular or analytical grade. 24 h prior to lysis. EGF 100 ng/ml was used for 1 or 5 min prior to lysis. Results are representative of n=3 experiments. Lysate production GST pull down experiments Total cell lysates were prepared from subcon¯uent tumor cells using modi®ed RIPA bu€er A (50 mM Tris-HCl pH 7.6, Commercially available prebound GST fusions containing the 150 mM NaCl, 10 mg/ml aprotinin, 1 mM PMSF, 10 mg/ml N- or C-SH2 domains of PLC-g (25 mg) were subjected to leupeptin, 2 mM Na3VO4,4mM EDTA, 10 mM NaF, 10 mM protein binding overnight using Bu€er A MDA-435 cell Na pyrophosphate, 1% NP-40 and 0.1% sodium deoxycho- lysates (500 mg protein). Bound complexes were washed ®ve late (20)) or a high stringency RIPA bu€er B (bu€er A times with Bu€er A, resuspended in sample bu€er and adjusted to 300 mM NaCl, plus 0.1% Triton X-100 and 0.1% subjected to gel electrophoresis and immunoblotting with SDS) on ice, followed by sonication. Fractionated lysates Ab2. Blots were acid-stripped and then probed with anti- were prepared by hypotonic lysis (Bu€er C: 25 mM Tris-Hcl Hsp70/Hsc70. Identical results were obtained using 25 mgof pH 7.5, 255 mM sucrose, 4 mM EDTA, 400 mM Na3VO4, prebound GST-PLC-g constructs were incubated with 30 mg 100 mM NaF, 10 mM Na pyrophosphate, 1 mM PMSF, lysates in 400 ml binding bu€er (20 mM Tris pH 7.5, 100 mM 10 mg/ml leupeptin, 10 mg/ml aprotinin). These lysates were NaCl, 2 mM EDTA, 0.1% NP-40, 2 mM DTT, 0.05% BSA spun at 500 g for 15 min at 48C to pellet the nuclei. The and 5% glycerol). Complexes were rocked at 48C for 2 h then supernatant was centrifuged at 100 000 g for 30 min at 48C washed ®ve times in 1 ml TENNS bu€er (2.5 mM Tris yielding the cytosolic fraction. The high spin pellet was pH 7.5, 2.5 mM EDTA, 250 mM NaCl, 1% NP-40 and 2.5% incubated in bu€er C with added 1% Triton X-100 and 1% sucrose) (Takayama et al., 1995) and resuspended in sample sodium deoxycholate for 30 min on ice to solubilize the bu€er, subjected to electrophoresis then immunoblotted. membrane components. The supernate of a 10 min centrifu- The CAIR-1/BAG-3 region containing the C terminal 3 or gation at 800 g (48C) constituted the membrane fraction. All all 4 PXXP putative SH3 binding domains (Figure 1) were lysates were aliquoted and subjected to no greater than 1 ampli®ed by PCR (5' primer for PXXP4: CTG GAT CCC freeze/thaw cycle. CTG AAA ACA AAC CAG AA; 5' primer for PXXP3: GGA TCC ATC CGC AA GAG GTG GAT; 3' primer for both: CTA GAA TTC TCA CAC TTT CAG CAC TCC Preparation of anti-peptide antibodies, immunoprecipitation, TGG; PCR conditions: 948C61 min; 548C645 s; and immunoblot 728C645 s, 30 cycles) and subcloned into PGEX-2T Polyclonal rabbit antisera were raised against keyhole limpet (PXXP3) or PGEX-4T3 (PXXP4). Clone sequences were hemocyanin (KLH)-conjugated CAIR-1 peptides 2 (SSPKSV- veri®ed, then bacterial fusion protein from PXXP or control ATEERAAPS) and 8 (DKGKKNAGNAEDPHT). Anti- null vector-transduced clones were produced according to peptide antibodies were puri®ed over peptide-conjugated manufacturer's instructions. Fusion proteins were immobi- Agel-10 columns and titered for activity by immunoblot lized on glutathione-Sepharose 4B beads preblocked in 0.5% analysis. Speci®city of Ab2 and Ab8 in immunoblot and nonfat milk and 0.05% BSA (Takayama et al., 1995). immunoprecipitation was demonstrated by peptide competi- Preblocked beads were washed three times with PBS- tion assays using cognate peptide and, as a control, the containing protease inhibitors (PBS+PI; 0.001% aprotinin, alternate peptide. For immunoprecipitation, equal quantities 0.001% leupeptin, 10 mM Na pyrophosphate, 4 mM EDTA of lysate protein (range: 200 ± 1000 mg) were precleared with and 1 mM PMSF). Beads (100 ml) were bound with GST or protein A or A/G beads and control IgG, then immunopre- GST-PXXP fusion proteins in PBS+PI at 48C for 2 h. cipitated with control IgG or speci®c antibody for 2 h at 48C. Bound beads (25 ml) were incubated with 30 mg MDA 435 Immunocomplexes were captured with protein A or A/G cell lysates in 400 ml binding bu€er at 48C for 2 h, then beads for 2 h at 48C and then washed ®ve times with lysis washed ®ve times in TENNS bu€er. The beads were bu€er (A or B) as indicated. Protein lysate or immunopre- resuspended in sample bu€er and analysed by electrophoresis cipitates were subjected to reducing gel electrophoresis and immunoblotting for PLC-g. The blot was acid-stripped followed by immunoblotting. Bands were detected with and reprobed for Hsp70/Hsc70. Results are representative of ECL following standard protocols. Protein partner binding n53 replicate experiments.

Oncogene An EGF-regulated complex of CAIR-1/BAG-3, PLC-g and Hsc70 H Doong et al 4395 Acknowledgments Statistical analysis The authors thank Ms V Kulpa for technical support and Where indicated Student's t-test analysis was done using Dr H Krutsch for peptide production. This work was Microsoft Excel (Seattle, WA, USA). All reported P values supported in part by the G Harold and Leila Y Mathers are 2-tailed. Charitable Foundation and the Florence and Max Kohn Fund.

Abbreviations CAI, carboxyamido-triazole; EGF, epidermal growth fac- tor; PMA, phorbol 12-myristate 13-acetate; d-ES, d- erythrosphingosine.

References

Bar-Sagi D, Rotin D, Batzer A, Mandiyan V and Schles- Lee J-H, Takahashi T, Yasuhara N, Inazawa J, Kamada S singer J. (1993). Cell, 74, 83 ± 91. and Tsujimoto Y. (1999). Oncogene, 18, 6183 ± 6190. Berridge MJ, Bootman MD and Lipp P. (1998). Nature, 395, Morimoto RI. (1998). Genes Dev., 12, 3788 ± 3796. 645 ± 648. Nishizuka Y. (1988). Nature, 344, 661 ± 665. Bukau B and Horwich AL. (1998). Cell, 92, 351 ± 366. Pawson T and Scott JD. (1997). Science, 278, 2075 ± 2080. Chang J-S, Noh DY, Park IA, Kim MJ, Song H, Ryu SH and Pilon M and Schekman R. (1999). Cell, 97, 679 ± 682. Suh P-G. (1997). Cancer Res., 57, 5465 ± 5468. Schlessinger J. (1994). Curr. Opin. Genet. Dev., 4, 25 ± 30. Chen D, Waters SG, Holt KH and Pessin JE. (1997). J. Biol. Smith MR, Court DW, Kim HK, Park JB, Rhee SG, Rhim Chem., 271, 6328 ± 6332. JS and Kung HF. (1998). Carcinogenesis, 19, 177 ± 185. Felder CC, Ma AL, Liotta LA and Kohn EC. (1991). J. Stuart JK, Myszka DG, Joss L, Mitchell RS, McDonald SM, Pharmacol. Exp. Therap., 257, 967 ± 971. Xie Z, Takayama S, Reed JC and Ely KR. (1998). J. Biol. Gusovsky F, Lueders JE, Kohn EC and Felder CC. (1993). J. Chem., 273, 22506 ± 22514. Biol. Chem., 268, 7768 ± 7772. Takayama S, Bimston DN, Matsuzawa S, Freeman BC, Hampson IN, Pope L, Cowling GJ and Dexter TM. (1992). Aime-Sempe C, Xie Z, Morimoto RI and Reed JC. (1997). Nucleic Acids Res., 20, 2899. EMBO J., 16, 4887 ± 4896. Hata A, Akita Y, Suzuki K and Ohno S. (1993). J. Biol. Takayama S, Sato T, Krajewksi S, Kochel K, Irie S, Millan Chem., 268, 9122 ± 9129. JA and Reed JC. (1995). Cell, 80, 279 ± 284. Hu Y and Bowtell DDL. (1996). Oncogene, 12, 1865 ± 1872. Takayama S, Xie Z and Reed JC. (1999). J. Biol. Chem., 274, Khoshyomn S, Penar PL, Rossi J, Wells A, Abramson DL 781 ± 786. and Bhushan A. (1999). Neurosurgery, 44, 568 ± 577. Turner T, Chen P, Goodly LJ and Wells A. (1996). Clin. Exp. Koch CA, Anderson D, Moran MF, Ellis C and Pawson T. Metastasis, 14, 409 ± 418. (1991). Science, 252, 668 ± 674. Turner T, Epps-Fung MV, Kassis J and Wells A. (1997). Kohn EC, Alessandro R, Spoonster J, Wersto R and Liotta Clin. Cancer Res., 3, 2275 ± 2282. LA. (1995). Proc. Natl. Acad. Sci. USA, 92, 1307 ± 1311. Wahl MI, Daniel TO and Carpenter G. (1988). Science, 241, Kohn EC, Felder CC, Jacobs W, Holmes KA, Day AF, Freer 968 ± 970. R and Liotta LA. (1994a). Cancer Res., 54, 935 ± 942. Yang H, Shen F, Herenyiova M and Weber G. (1998). Kohn EC, Jacobs W, Kim YS, Alessandro R, Stetler- Anticancer Res., 18, 1399 ± 1404. Stevenson WG and Liotta LA. (1994b). J. Biol. Chem., Yang LJ, Rhee SG and Williamson JR. (1994). J. Biol. 269, 21505 ± 21511. Chem., 269, 7156 ± 7162. Kohn EC and Liotta LA. (1990). J. Natl. Cancer Inst., 82, Zwick E, Hackel PO, Prenzel N and A U. (1999). Trends 54 ± 60. Pharmacol. Sci., 20, 408 ± 412. Kohn EC, Sandeen MA and Liotta LA. (1992). Cancer Res., 52, 3208 ± 3212.

Oncogene