Recruitment of CD40 and tumor necrosis factor receptor-associated factors 2 and 3 to membrane microdomains during CD40 signaling.

Signals delivered to antigen-presenting cells through CD40 are critical for the activation of immune responses. Intracellular tumor necrosis factor (TNF) receptor-associated factors (TRAFs) are key elements of the signal transduction pathways of many TNF receptor family members, including CD40. We show for the first time that engagement of CD40 in intact B cells induces the rapid translocation of TRAF2 from the cytoplasm to the plasma membrane. We found that CD40 engagement also results in its recruitment, together with TRAF2 and TRAF3, to membrane microdomains, regions of the plasma membrane enriched in signaling molecules such as the Src family kinases. Using a membrane-permeable chelator of zinc or a mutant TRAF2 molecule, we show that the putative zinc-binding domains of TRAFs contribute to their recruitment to microdomains and to the downstream activation of c-Jun N-terminal kinase. We suggest that the zinc RING and zinc finger domains of TRAFs are required for communication between CD40 and microdomain-associated signaling molecules and may serve a similar role in the signal transduction pathways of other TNF receptor family members.

necrosis factor (TNF) receptor-associated factor (TRAF) family of proteins bind the cytoplasmic domain of CD40 and potentially initiate signal transduction (7)(8)(9)(10). Understanding the roles of TRAF proteins in CD40 signaling requires careful characterization of the molecular interactions between CD40 and TRAF proteins. It has been suggested that TRAF2 is bound to CD40 in unstimulated B cells and is released into the cytoplasm upon CD40 ligation (11). However, using similar techniques, others have concluded that TRAFs are recruited to CD40 during signaling (12). To address the issue of subcellular localization in intact cells, we stably transfected a mouse B cell line with green fluorescent protein (GFP)-labeled TRAF2. Using confocal microscopy, we found that TRAF2 is distributed throughout the cytoplasm of unstimulated cells and is recruited to the plasma membrane only following CD40 engagement with membrane-bound CD154 or anti-CD40 antibody.
Remarkably, we also found that CD154 stimulation of B cells renders CD40, TRAF2, and TRAF3 largely insoluble in nonionic detergents. Recent reports have described the association of important signaling molecules in lymphocytes with low density, detergent-insoluble membrane microdomains (13)(14)(15)(16) or "rafts," enriched in sphingolipids, cholesterol, and glycosylphosphatidylinositol-linked proteins (17). Using density gradient centrifugation, we demonstrate that CD40 engagement results in the rapid, dramatic recruitment of TRAF2 and TRAF3 to these specialized regions of the plasma membrane. In addition, we found that a mutant TRAF2 molecule containing a deletion in its zinc-binding domain displayed reduced raft recruitment, as did wild-type TRAF2 and TRAF3 in cells treated with a membrane-permeable chelator of zinc. Zinc chelation also blocked CD40-mediated JNK activation, but not the activation of JNK by osmotic stress. Taken together, these findings suggest that the zinc-binding domains of TRAFs mediate interactions with raft components critical for downstream signaling events.

EXPERIMENTAL PROCEDURES
DNA Constructs-To construct human TRAF2 with an amino-terminal GFP tag, TRAF2 cDNA was ligated into the pEGFP-C2 vector (CLONTECH, Palo Alto, CA). The insert encoding TRAF2-GFP was subcloned into the inducible expression vector pOPRSVI.mcs1 (18), generating pTRAF2-GFP. LacR-binding sites in the Rous sarcoma virus promoter of this construct allow for the negative regulation of protein production by LacR. pEFLac encoding LacR has been described (19). TRAF2⌬Zn (amino acids 199 -501 of mouse TRAF2) was produced by polymerase chain reaction mutagenesis and placed into pOPRSVI.mcs1.
Transfections-Stable transfections of mouse B cell lines were carried out using electroporation as described previously (21).
Cell Fractionation and Western Blotting-Cells (10 7 /condition/ml) were stimulated for 10 min at 37°C with 10 g/ml anti-CD40 antibody (G28-5 or 1C10) or isotype control mAb (MOPC-21 or EM-95) and then pelleted by centrifugation. Cells were resuspended in 200 l of ice-cold lysis buffer (1% Brij 58, 20 mM Tris (pH 7.5), and 150 mM NaCl with protease and phosphatase inhibitors) and incubated on ice for 30 min. Lysates were centrifuged at 14,000 ϫ g for 25 min at 4°C. After collecting supernatants, the detergent-insoluble pellets were resuspended and briefly sonicated in 200 l of lysis buffer supplemented with 0.5% SDS and 1% ␤-mercaptoethanol. In some experiments, cells were preincubated in serum-free medium (2 ϫ 10 6 /ml) with 10 M TPEN, 10 M BAPTA/AM, or 1:2000 Me 2 SO for 30 min at 37°C. Cells were pelleted, resuspended to 10 7 /ml, and stimulated as described above. Lysate fractions were diluted 1:2 with reducing SDS-PAGE loading buffer, heated for 5 min at 95°C, subjected to SDS-PAGE, and electroblotted onto polyvinylidene difluoride membranes. To allow better detection of hCD40 on Western blots, lysate samples were deglycosylated prior to SDS-PAGE using peptide N-glycosidase F (New England Biolabs Inc., Beverly, MA) according to the manufacturer's protocol. A chemiluminescent substrate (Pierce) was used to detect HRP-labeled antibodies on Western blots.
Raft Isolation-Raft isolation was accomplished using a combination of published protocols (15,24). B cells (2.5 ϫ 10 7 /condition/ml) were stimulated at 37°C for 10 min using 10 g/ml anti-CD40 mAb or 5 ϫ 10 6 CD154-expressing CHO cells. In some experiments, cells were pretreated with TPEN (as described above) or with MCD. Pretreatment of B cells with MCD was essentially as described (14). Briefly, cells were washed in serum-free medium, resuspended in serum-free medium containing 22 mg/ml MCD (ϳ20 mM), incubated for 10 min at room temperature and for 15 min at 37°C, and then stimulated with anti-hCD40 antibody (as described above). Following stimulation, cells (including CHO-hCD154 stimulators, where used) were pelleted and resuspended in 400 l of ice-cold lysis buffer and incubated on ice for 30 min. Cell lysates were diluted 1:2 with 70% Nycodenz in 20 mM Tris (pH 7.5) and 150 mM NaCl and transferred to 3-ml ultracentrifuge tubes. Lysates were overlaid with a Nycodenz step gradient (200 l each of 25, 21.5, 18, 15, and 8% Nycodenz) (24) and centrifuged at 200,000 ϫ g for 4 h at 4°C (Beckman SW 60 rotor). Eleven 200-l fractions were removed from the tubes, starting at the top (lowest density), and then diluted 1:2 in reducing SDS-PAGE loading buffer and heated for 5 min at 95°C. Samples of fractions were subjected to SDS-PAGE and then electroblotted onto polyvinylidene difluoride membranes. For testing ganglioside GM1 content of density gradient fractions, 2.5 ϫ 10 7 unstimulated M12.4.1 cells were incubated for 10 min at room temperature in 1 ml of culture medium containing 3.4 g of HRP-labeled cholera toxin B subunit (or, as controls, untreated cells and cells incubated with an equivalent amount of unconjugated HRP). Cells were then subjected to lysis and density gradient centrifugation as described above. Gradient fractions were assayed for peroxidase activity by mixing 10-l gradient fractions with 100 l of 50 mM sodium phosphate, 25 mM citric acid (pH 5.0), 1 mg/ml o-phenylenediamine dihydrochloride, and 0.012% H 2 O 2 . Samples were incubated for ϳ5 min at room temperature, and the reaction was stopped by adding 150 l of 0.67 M sulfuric acid. Optical density of the samples was read at 405 nm in an enzyme-linked immunosorbent assay plate reader.

Subcellular Localization of TRAF2 during CD40
Signaling-To examine the localization of TRAF2 in intact B cells, we stably transfected M12.LacR cells with a construct encoding an inducibly expressed GFP-tagged TRAF2 molecule or, as a control, inducible GFP. Confocal microscopy revealed that GFP alone was evenly distributed in cells, including the nucleus (Fig. 1A). Incubation of cells with anti-mCD40 antibody and a rhodamine-labeled secondary antibody (10 min at 25°C) ( Fig.  1, A-C) did not affect the distribution of GFP (unstimulated cells not shown). In contrast, TRAF2-GFP was excluded from the nuclei and localized to the cytoplasm of unstimulated cells (fixed and then surface-stained with anti-mCD40 antibody followed by a rhodamine-labeled secondary antibody) (Fig. 1, D-F). In cells stimulated for 10 min at 25°C with anti-CD40 and rhodamine-labeled secondary antibodies, TRAF2-GFP was localized to small patches in the plasma membrane ( Fig. 1, G-I) and colocalized with CD40 (colocalization in yellow) (Fig. 1I). Recruitment of TRAF2-GFP from the cytoplasm to the plasma membrane began immediately upon the addition of anti-CD40 mAb and was essentially complete after 5-10 min. Interestingly, antibody stimulation of cells at 37°C resulted in significant internalization of CD40 with associated TRAF2 (Fig. 1, J-L). However, internalization appeared minimal when B cells were stimulated at 37°C for 10 min with membrane-bound CD154 (Fig. 1, M-O). We conclude that in intact B cells, TRAF2 associates with CD40 only following CD40 ligation.
Association of TRAF2 and CD40 with Detergent-insoluble Complexes-Although confocal microscopy indicated that CD40 and TRAF2 associate in activated cells, we were unable to consistently demonstrate this association using conventional co-immunoprecipitation techniques. Previously, however, it was reported that the total amount of TRAF2 that could be immunoprecipitated from stimulated cells was considerably less than that obtained from unstimulated cells (12). It was suggested that stimulation resulted in the recruitment of TRAFs to detergent-insoluble complexes or that TRAFs undergo activation-induced degradation. To determine if TRAF2 becomes associated with detergent-insoluble material in activated cells, we examined the effects of CD40 stimulation on a mouse B cell line transfected with hCD40, M12.hCD40 (20). Cells were stimulated with anti-hCD40 mAb and lysed in 1% Brij 58 (a relatively mild, non-ionic detergent), and lysates were centrifuged to pellet detergent-insoluble material. The insoluble material was dissolved in 0.5% SDS and 1% ␤-mercaptoethanol so that it could be subjected to SDS-PAGE/Western blot analysis and compared with samples of the Brij 58soluble material. On Western blots, the majority of TRAF2 from unstimulated B cells appeared in the detergent-soluble fraction ( Fig. 2A). However, in cells stimulated through hCD40 or endogenous mCD40, much of the TRAF2 became associated with the detergent-insoluble fraction ( Fig. 2A). Interestingly, a portion of the TRAF2 in the insoluble fraction appeared to be post-translationally modified, resulting in the appearance of multiple high molecular weight forms. To determine if CD40 itself becomes detergent-insoluble in stimulated cells, we generated a polyclonal antiserum for detecting hCD40 on Western blots. In unstimulated cells, a significant fraction of CD40 was already present in the detergent-insoluble fraction (Fig. 2). Upon stimulation, additional CD40 was recruited to this fraction ( Fig. 2A), although this recruitment was not as dramatic as the activation-induced partitioning of TRAF2 ( Fig. 2A). Association of hCD40 (or TRAF2) with the detergent-insoluble fraction did not occur if the stimulating antibody was instead added post-lysis (data not shown). To determine if the recruit- ment of CD40 to detergent-insoluble material is dependent upon TRAF association, we repeated the experiment using M12.4.1 cells expressing a mutant hCD40 molecule with a 55-amino acid cytoplasmic deletion (hCD40⌬55) (20). hCD40⌬55 lacks the binding sites for TRAF2, TRAF3, and TRAF6 (10,25,26) and has no signaling activity (20). As with full-length CD40, engagement of hCD40⌬55 resulted in its recruitment to detergent-insoluble material, indicating that this activation-dependent event does not require TRAF binding (Fig. 2B). Although engagement of hCD40⌬55 did not mediate recruitment of TRAF2 to the insoluble fraction, engagement of endogenous mCD40 in the same cells induced the transition (Fig. 2B).
To test the possibility that TRAF2 may also undergo degradation in activated cells, we stimulated M12.4.1 cells for various periods of time with anti-mCD40 mAb, and then we prepared cell lysates using either 1% Brij 58 lysis buffer or buffer containing 1% SDS to more completely solubilize cellular proteins. With a stimulation period of as little as 1 min, most of the TRAF2 appeared in the Brij 58-insoluble fraction (Fig. 2C). However, total cellular TRAF2 amounts also decreased with stimulation (Fig. 2D). We conclude that TRAF2 is recruited to Brij 58-insoluble material and undergoes degradation as a result of CD40 stimulation.
Recruitment of CD40, TRAF2, and TRAF3 to Membrane Rafts-T cell receptor ligation induces recruitment of Zap-70, T cell receptor ␣, CD3⑀, and CD3 to membrane rafts, which may facilitate their interactions with raft-associated Src family kinases (14,15). Potentially, the stimulation-induced insolubility of CD40 and TRAF2 results from a similar association with low density, detergent-insoluble lipid rafts. To test this hypothesis, we stimulated mouse B cells for 10 min at 37°C with hCD154bearing CHO cells, lysed the cells in 1% Brij 58, and fractionated the lysates by density gradient ultracentrifugation. Insoluble membrane rafts are found in the low density portion of the gradient, whereas soluble membrane and cytoplasmic proteins are found in higher density fractions (15). TRAF2 from unstimulated M12.hCD40 cells was found in the high density fractions (Fig. 3A). CD40 engagement resulted in a marked shift of TRAF2 into lower density fractions, suggesting an association with membrane rafts. Similar results were obtained using anti-mCD40 or anti-hCD40 antibody or CHO-mCD154 cells as stimuli and in a second mouse B cell line, CH12.LX (data not shown). Preliminary experiments indicate that this recruitment also occurs in normal mouse splenic B cells. Blots were also examined for two additional CD40-binding proteins, TRAF3 and TRAF6, both of which appeared in the high density fractions from unstimulated cells (Fig. 3, B and C). In CD154-stimulated cells, the behavior of TRAF3 was similar to that of TRAF2, whereas there was no detectable association of TRAF6 with membrane rafts. Stimulation of M12.hCD40 cells with hCD154 also induced recruitment of hCD40 to low density fractions, although a small amount of CD40 was found in the low density fractions of unstimulated cells as well (Fig.  3D). In addition, we confirmed that TRAF2-GFP was recruited to low density membrane rafts in CD40-stimulated cells (data not shown) and that the raft recruitment of TRAF2 also occurred at 25°C (Fig. 3, E and F). To examine the localization of a known constituent of membrane rafts, Western blots were reprobed for Lyn (Fig. 3G). As expected based upon previous reports (27), Lyn localized to low density fractions. Lyn distribution was similar in unstimulated cells (data not shown). The ganglioside GM1 has also been reported to be enriched in membrane microdomains (28). To test density fractions for GM1 content, cells were incubated with cholera toxin (which binds GM1) conjugated to HRP, washed, and then subjected to lysis and density gradient centrifugation. As shown in Fig. 3H, peroxidase activity was detected in all fractions, but was considerably enriched in the low density material. Taken together, these results support the hypothesis that CD40, TRAF2, and TRAF3 are recruited to membrane rafts in activated B lymphocytes.
Effects of Methyl-␤-cyclodextrin on CD40 Signaling-It has been reported that cholesterol-binding agents such as MCD can disrupt membrane microdomains (29). Treatment of live cells with MCD disrupts T cell receptor-mediated Ca 2ϩ fluxes, and it has been argued that this type of experiment demonstrates the importance of membrane microdomains in T cell receptor signaling (14,30). However, we found that MCD treatment, while rendering Ͼ10% of B cells permeable to trypan blue (data not shown), only partially disrupted membrane rafts. As shown in Fig. 4A, MCD treatment of M12.hCD40 cells dissociated only a portion of Lyn from low density, Brij 58-insoluble complexes. A previous report similarly demonstrated that MCD was not able to entirely dissociate Fyn (another Src family kinase) from detergent-insoluble complexes, and it was suggested that MCD may only inefficiently extract cholesterol from the inner leaflet of the plasma membrane (29). We also found that the raft association of TRAF2 in CD40-stimulated cells (Fig. 4B) was only partially disrupted by MCD. However, the combination of MCD treatment and lysis in a stronger detergent solution (1% Triton X-100) did lead to more complete dissociation of membrane microdomains (Fig. 4C). We next examined the effects of MCD on the activation of JNK, an enzyme rapidly activated by CD40 stimulation. In addition to its inability to fully disrupt membrane microdomains (unless combined with a sufficiently strong detergent), MCD treatment, by itself, activated JNK (Fig. 4D). MCD treatment appeared not to specifically inhibit JNK activation mediated by CD40 (Fig. 4D), but it is difficult to know if there was a partial inhibition masked by the MCDinduced activation of JNK or if MCD was unable to sufficiently disrupt raft complexes involved in CD40 signaling. Taken together, our results indicate it would be inappropriate to draw firm conclusions from CD40 signaling experiments performed using MCD-treated cells.
Zinc-dependent JNK Activation and Recruitment of TRAFs to Rafts-In addition to its carboxyl-terminal receptor (CD40)binding domain, TRAF2 contains amino-terminal zinc RING and zinc finger motifs (31). Although the functions of these putative zinc-binding domains are unclear, similar domains in other proteins can mediate protein-protein interactions (32). Mice transgenic for a mutant TRAF2 molecule lacking the zinc RING domain (33) display defective activation of JNK by TNF receptor family members, including CD40. These results and the mutational analysis of TRAF2 (34) strongly suggest that the zinc-binding features of TRAFs are important for interactions with downstream signaling molecules. In support of this hypothesis, we found that TPEN, a membrane-permeable chelator of zinc (35), abrogated the CD40-mediated activation of JNK (Fig. 5A), but did not inhibit the activation of JNK by osmotic stress (sorbitol). A cell-permeable chelator of Ca 2ϩ , BAPTA/AM, did not specifically inhibit the activation of JNK by CD40, indicating that the action of TPEN is not due to its low affinity for Ca 2ϩ . Although TPEN may affect other steps in the CD40 signal transduction pathway, its primary effects appear to be on TRAF function. Zinc chelation abolished the activation-induced post-translational modification of TRAF2 and significantly inhibited the recruitment of TRAF2 (Fig. 5B), but not CD40 (data not shown), to raft complexes. Similarly, a mutant TRAF2 molecule in which we deleted most of the zincbinding domain (TRAF2⌬Zn) displayed reduced recruitment to membrane rafts (Fig. 5C). Importantly, TPEN treatment did not further reduce recruitment of TRAF2⌬Zn to membrane rafts, indicating that zinc chelation does not interfere with the binding of the TRAF domain of TRAF2 to the cytoplasmic domain of CD40. Together, these results suggest the zincbinding features of TRAF2 interact with membrane raft-associated molecules and that these interactions are required for downstream signaling events. Further support for this hypothesis is presented in Fig. 6. Cells were stimulated with anti-CD40 antibody and then solubilized in 1% Brij 58 or lysis buffer in which Brij 58 was replaced with 0.5% Nonidet P-40. Although Nonidet P-40 was able to entirely solubilize CD40, a significant proportion of the TRAF2 and TRAF3 from stimulated cells remained in the detergent-insoluble material, indicating that TRAFs may bind a Nonidet P-40-insoluble microdomain component. Pretreatment of cells with TPEN abolished CD40-stimulated recruitment of TRAFs to the Nonidet P-40insoluble fraction, demonstrating that the zinc-binding domains of TRAFs are critical for their association with the insoluble fraction. DISCUSSION We have shown that engagement of CD40 results in the rapid recruitment of TRAF2 and TRAF3 from the cytosol to the with anti-hCD40 mAb, and then lysates were prepared using 1% Brij 58. Lysates were subjected to density gradient fractionation, SDS-PAGE, and Western blotting for Lyn (A). Western blot membranes from A were reprobed for TRAF2 (B). In C (showing a Western blot for Lyn), cells were treated as described for A, except that 1% Triton X-100 was substituted for Brij 58 in the lysis buffer. To examine the effects of MCD on the activation of JNK, M12.hCD40 cells were incubated with or without MCD and stimulated for 5 min at 37°C with anti-hCD40 mAb (ϩ), an isotype control mAb (Ϫ), or 0.6 M sorbitol (S). Cell lysates were then tested for activated JNK using GST-c-Jun-(1-79) as a substrate. Phosphorylated GST-Jun appears as a band on the autoradiogram shown in D.
plasma membrane, where they associate with membrane rafts. CD40 stimulation also induces the association of CD40 itself with membrane rafts, but its recruitment does not appear to be TRAF-dependent. TRAF6 has also been characterized as a CD40-binding protein, yet in contrast to TRAF2 and TRAF3, it showed little if any raft recruitment during stimulation of B cells. However, it remains possible that TRAF6 is recruited only after prolonged stimulation or at low but functionally significant levels.
Although the functions of the TRAF proteins remain enigmatic, they appear to serve as adapter molecules linking CD40 to important components of its signaling pathway. Previous work indicates that the zinc RING and zinc fingers of TRAF2 are not required for its binding to CD40 (7,26). However, we found that disrupting the function of the zinc-binding domains (with TPEN or by mutation) partially inhibited the raft recruitment of TRAF2 and TRAF3, suggesting that raft-associated molecules (in addition to CD40) contribute to the avidity of TRAF-microdomain interactions. This hypothesis is supported by the fact that the CD40 from activated cells was completely solubilized in 0.5% Nonidet P-40, whereas the same detergent only partially solubilized TRAF2 and TRAF3.
Although our results indicate that TRAF2 and TRAF3 may associate with membrane raft constituents, these molecules remain to be identified. Recently, putative TRAF-interacting proteins have been identified, including apoptosis signal-regulating kinase 1 (36), nuclear factor B-inducing kinase (37), germinal center kinase (38), and MEKK1, which, unlike the other kinases, may interact with the zinc-binding features of TRAF2 (39). However, the interactions of these kinases with TRAFs have been almost exclusively demonstrated in epithelial cell lines overexpressing transiently transfected TRAFs and candidate molecules. Using commercial antisera, we have been unable to detect the endogenous forms of these putative TRAF-interacting molecules in detergent-insoluble TRAF-containing fractions from CD40-stimulated B cells. Further work is necessary to determine if there are very low levels of these kinases present in the CD40 signal transduction complex or if other proteins (perhaps related to the candidates previously identified) mediate CD40 signaling in B cells. Interestingly, one raft-associated kinase, Lyn, has been postulated to participate in CD40 signaling (40,41), although there is currently no evidence that it directly interacts with any of the TRAF molecules.
One key to identifying TRAF-interacting molecules in membrane microdomains may be the post-translational modification displayed by raft-associated TRAFs (Figs. 2-6). The modification appears to require the zinc-binding features of TRAFs in that it is largely ablated by TPEN treatment and is not evident in TRAF2⌬Zn. If this modification is important for the binding of raft-associated molecules, such molecules may have been overlooked in previous attempts to identify TRAF-binding partners. Furthermore, the modification may require the activity of raft-associated proteins, and determining the type of modification should facilitate their identification. Interestingly, several zinc RING-containing proteins such as Cbl participate in the ubiquitination of themselves and other proteins, targeting them for degradation (42). In light of the activationinduced modification and degradation of TRAF2 we observed, it seems reasonable to suggest that one or more of the zinc RINGcontaining TRAFs similarly promote ubiquitination and perhaps help to limit the duration of CD40 signaling. We are currently testing this hypothesis.
The discovery of raft-associated CD40, TRAF2, and TRAF3 represents a major advance in the understanding of CD40 signal transduction, indicating where and how to look for other molecules that mediate the proximal steps in CD40 signaling. Although our work focuses on the CD40 signal transduction pathway, the results presented here have broad implications for signaling through a variety of related receptors. TRAF To determine if TPEN also inhibits the CD40-mediated raft recruitment of TRAF2, M12.hCD40 cells were pretreated with TPEN or Me 2 SO and then stimulated with isotype control mAb or anti-hCD40 mAb. Cell lysates (1% Brij) were subjected to density gradient centrifugation and Western blot analysis for TRAF2 (arrowheads) (B; density of fractions increases from left to right). To confirm that the zinc-binding features of TRAF2 contribute to raft recruitment, M12.4.1 cells transfected with inducible TRAF2⌬Zn were cultured for 48 h with 100 M isopropyl-␤-D-thiogalactopyranoside to up-regulate production of the TRAF2 mutant, incubated for 30 min with TPEN or Me 2 SO, and then stimulated for 10 min at 37°C with isotype control mAb (Ϫ) or anti-mCD40 antibody (ϩ). Detergent-soluble (Sol.) and -insoluble (Ins.) fractions were examined for TRAF2 content on Western blots (C). WT, wild-type.