The Endoplasmic Reticulum Is a Key

Plasma cells (PC) are the effector cells of the humoral Ab response. Unlike other dedicated secretory cells, they exist as two populations with opposite cell fates: short-lived and long-lived PC. Upon transformation they lead to an incurable neoplasia called multiple myeloma. In this study we have explored the molecular mechanism of PC death. Our data show that their apoptotic pathway is unique among other hemopoietic cells inasmuch as neither the death receptors nor the mitochondria play the central role. PC apoptosis is initiated by activation of Bax at the endoplasmic reticulum membrane and subsequent activation of the endoplasmic reticulum-associated caspase-4 before the release of mitochondrial apoptogenic factors. Together, our observations indicate that the cardinal function of PC (i.e., Ig secretion) is also the cause of their death. The Journal of Immunology, 2006, 176: 1340–1347. human Bax (Bax-NT) was purchased from Upstate Biotech- nology. This Ab was used to reveal the activated form of Bax by photonic, confocal, and electron microscopies, as described previously (13). A non- immune rabbit serum from Vector Laboratories was used as a negative staining control. Bax and its control Ab were revealed with 1) a biotinyl- ated goat anti-rabbit Ab (Immunotech), followed by sequential incubation with HRP-coupled extravidin (Sigma-Aldrich) and 3-amino-9-ethylcarba- zole, for immunoenzymatic stainings; 2) FITC-coupled goat anti-rabbit Abs (Jackson ImmunoResearch Laboratories) for immunoﬂuorescence stainings; and 3) goat anti-rabbit IgG conjugated to gold particles (TEBU) for immunoelectron microscopy. The anti-human cytochrome oxidase (Cox) mAb was purchased from BD Pharmingen and was revealed with Cy5-coupled anti-mouse Abs from Jackson ImmunoResearch Laborato- ries. The anti-human protein disulﬁde isomerase (PDI) mAb used for confocal microscopy and the goat anti-mouse Abs conjugated to AlexaFluor 488 used for its revelation were both purchased from Molecular Probes. Ab was used as a loading control. Control extracts were prepared from activated PBL recultured for 3 h with medium (0), staurosporine (STS), or an agonistic anti-CD95 Ab (7C11). The proenzymatic forms of caspase-8 and -3 are visualized as 55- and 35-kDa bands, respectively. Their intermediate cleavage forms migrate as a 41/43-kDa doublet for caspase-8 and a 20/21-kDa doublet for caspase-3. The enzymatically active forms of caspase-8 and -3 are identiﬁed as bands with apparent M of 17 respectively are

T o fulfill their function of effectors of the humoral Ab response, plasma cells (PC) 3 are equipped for the production of massive quantities of Igs. Accordingly, they share the morphological and functional attributes of dedicated secretory cells, including expansion of their endoplasmic reticulum (ER) and Golgi networks. However, PC are unique among other professional secretory cells inasmuch as the ER is also instrumental in their differentiation. As recently demonstrated by the groups of Glimcher and Brewer (1,2), production of the spliced active form of the XBP-1 mRNA, a transcription factor essential for PC commitment, requires a multimolecular system called the unfolded protein response (UPR) that operates in the ER. The UPR is activated when correct folding of proteins is compromised. It elicits several biological responses that concur in restoring the equilibrium between the rate of protein synthesis and the folding capacity of the ER (3). It was long believed that PC, unlike most other professional secretory cells, are short-lived. However, this dogma has been challenged by two seminal studies demonstrating that some PC escape deletion and form a compartment of long-lived PC in the bone marrow (BM) that constitutes one of the cellular components of B cell memory (4,5). The possibility of manipulating PC survival is thus an important issue for therapeutic purposes. The present study was undertaken to dissect the apoptotic pathway used by PC.
There are two main routes to cellular apoptosis, designated the intrinsic and the extrinsic pathway (see Ref. 6 for a review). The intrinsic pathway is initiated by internal or external stress signals that lead to the translocation and/or activation of proapoptotic members of the Bcl-2 family (Bax and Bak) to the mitochondria. Their insertion in the outer mitochondrial membrane promotes its permeabilization and ultimately induces the release of mitochondrial apoptogenic factors such as cytochrome c (cyt c) or apoptosis-inducing factor (AIF), in the cytosol. Cyt c together with the adaptor protein Apaf-1 and the initiator caspase-9 form a multimolecular signaling platform called the apoptosome, which initiates caspase activation downstream of the mitochondria (7). Work over the past 5 years has revealed that, in fact, many cellular organelles can sense stress signals and trigger apoptotic mediators (8). The ER, for example, can activate its own initiator caspase (caspase-4 in humans, caspase-12 in the mice) (9,10). Cell death induced by the intrinsic pathway is prevented by overexpression of the antiapoptotic members of the Bcl-2 family. The extrinsic pathway involves the death receptor (DR) family, whose prototypic member is the CD95 (Fas, APO-1) molecule. Upon oligomerization, these receptors bind an adaptor molecule (Fas-associated death domain protein) that, in turn, recruits the proenzymatic form of an initiator caspase (caspase-8 or -10). The oligomerized receptors, Fas-associated death domain protein, and procaspase-8/10 form a multimolecular signaling complex called death-inducing signaling complex that promotes autoproteolytical cleavage of caspase-8/10 and initiation of the apoptotic cascade (11). Alterations of the mitochondrial functions also occur during the extrinsic pathway, but they play an ancillary role in execution of the death program.
In this study we show that spontaneous PC apoptosis is initiated by signals that emanate from the ER. This conclusion is based upon three lines of evidence. First, Bax is primarily activated at the ER, not at the mitochondria membrane, during PC apoptosis. Second, PC death relies neither on caspase-8 nor caspase-9, but involves the ER-associated initiator caspase-4. Third, activation of

Reagents
The soluble recombinant human FLAG-CD95-ligand (CD178; Alexis) was used at 50 ng/ml together with the anti-FLAG mAb M2 (Sigma-Aldrich) at 1 g/ml. The soluble trimerized form of TL1A (R&D Systems) was used at 0.5 g/ml. Soluble chimeric forms of CD95, DR3, and TNFRI in which the extracellular domain of each receptor has been fused with the Fc portion of human IgG1 were used to prevent binding of the corresponding death ligands to their cellular receptors. CD95-Fc and DR3-Fc were purchased from Alexis and were used at 10 and 5 g/ml, respectively. TN-FRI-Fc was provided by Dr. H. Waldmann (University of Oxford, Oxford, U.K.) and was used at 20 g/ml. The caspase inhibitors were purchased from the following manufacturers: Bachem for z-VAD-fluoromethylketone (fmk); Calbiochem for z-IETD-fmk, z-LEHD-fmk, and z-DEVD-fmk; and MBL for z-LEVD-fmk. They were used at the concentrations indicated in the text. To estimate the efficiency and specificity of the caspase-8 and caspase-9 inhibitors, both antagonistic peptides were tested at optimal concentration for their capacity to inhibit DNA fragmentation in PHA-activated PBL recultured with staurosporine (inducer of the intrinsic mitochondrial pathway) or the anti-CD95 mAb 7C11 (inducer of the extrinsic pathway). z-LEHD-fmk was found to be effective on staurosporine-induced apoptosis, but not on anti-CD95-induced apoptosis, whereas z-IETD-fmk displayed the opposite pattern of inhibitory effect (data not shown). IL-2 was purchased from PeproTech, and IL-10 was provided by Dr. F. Brière (Schering-Plough, Dardilly, France). They were used at 10 U/ml and 50 ng/ml, respectively, throughout the study. Staurosporine (1 M), used as an inducer of the mitochondrial intrinsic apoptotic pathway, was purchased from Sigma-Aldrich.

Cells
Tonsillar germinal center (GC) B cells were isolated as described previously (14). PC were generated in vitro from purified GC B cells using a modification of the two-step culture system initially described by Arpin et al. (15). In the first culture, GC B cells are stimulated for 3 days with mouse Ltk Ϫ cells stably transfected with CD154 in the presence of IL-2 and IL-10. Viable B cell blasts are next recultured for 3 additional days in the presence of IL-2, IL-10, and the anti-CD154 mAb LL2 (provided by Dr. P. Garonne, Schering-Plough, Dardilly, France) to block residual CD154 activity. At the end of the secondary culture, PC were purified by immunomagnetic negative selection using an Ab mixture including anti-CD2, CD20, DQ, CD16, CD56, CD3, and CD8 mAbs. To generate polyclonal mouse PC, C57BL/6 mice were immunized i.p. with SRBC (5 ϫ 10 8 SRBC/mouse). Splenocytes recovered 6 days after immunization were double stained with PE-conjugated anti-CD138 and allophycocyanin-conjugated anti-B220 rat mAbs. Splenic PC were sorted by gating the CD138 high B220 low cells. The purity of the sorted murine PC was 99% on the average. Protein extracts prepared from human myeloma cell lines IM9 and U266 (ATCC) and PBL activated for 4 days with PHA and IL-2 were used as controls for Western blotting.

Measurement of apoptosis
The potential sensitive fluorescent dye 3,3Ј-dihexyloxacarbocyanine iodide (DiOC 6 ; Molecular Probes) was used to reveal disruption of the mitochondrial transmembrane potential (⌬m). Phosphatidylserine exposure was quantified by surface binding of annexin V according to the instructions of the manufacturer (Bender MedSystems). DNA fragmentation was assessed by TUNEL assay, using the In Situ Cell Death Detection Kit (Roche). The CaspACE assay system (Promega) was used to determine the global status of cellular caspases activation by flow cytometry.

Microscopy
For the immunoenzymatic and immunofluorescence stainings of Bax, cyt c, Cox, or PDI, PC were fixed in paraformaldehyde 4% for 15 min at room temperature, permeabilized with 0.2% Triton X-100, then processed for immunostaining. Immunofluorescence stainings were observed under an LSM 510 laser scanning confocal microscope (Zeiss), and images were processed with Adobe Photoshop software 6.0. Preparation of the cell samples and their analysis by transmission and immunoelectron microscopies were conducted as described previously (16).

RNA and protein analysis
Expression of the DR family transcripts was measured by multiprobe RNase protection assay using the RiboQuant kit (BD Pharmingen) following the instructions of the manufacturer. The quantity of protected RNAs was determined using a PhosphorImager and ImageQuant software (both from Molecular Dynamics). For Western blot analysis, total protein extracts were prepared as described previously (14). Cytosolic fractions of PC were generated using a digitonin-based subcellular fractionation technique as described previously (17).

Spontaneous death of PC can be assimilated to apoptosis
To circumvent the difficulty caused by the poor representation of PC in lymphoid tissues, we generated PC in vitro from human tonsillar B lymphocytes using a previously described two-step culture system (15). In vitro-generated PC display the major phenotypical features of ex vivo PC, as revealed by their strong expression of CD38, CD27, and CD138 (syndecan-1) and downregulation of CD20, surface Igs, and the CD79a/CD79b heterodimer ( Fig. 1A). They also exhibit the expected morphology and cardinal function of PC, i.e., the capacity to spontaneously secrete large amounts of Igs as determined by dual ELISA/ ELISPOT assays (data not shown).
As shown in Fig. 1B, spontaneous PC death is accompanied by membrane, mitochondrial, and nuclear damage as well as caspase activation. By 48 h of culture, Ͼ90% of PC are dead according to any of these four criteria. By electron microscopy ( Fig. 1C), in vitro-generated PC display the classical morphological features of bona fide PC: abundant cytoplasm and enlarged reticulum cisternae. Some of them even present the morphological characteristics of Mott cells, with multiple vesicles (Russel bodies) filled with stored Igs. PC undergoing spontaneous cell death present some of the typical features of apoptosis, such as condensation of chromatin to the periphery of the nucleus, piknotic nuclear fragmentation, and cell shrinkage, but preservation of plasma membrane integrity. They also exhibit a dramatic microvesiculation and reorganization of the ER/Golgi network. Together, these findings suggest that spontaneous PC death can be regarded as apoptosis.

DRs are not involved in spontaneous PC apoptosis
To explore the contribution of the extrinsic apoptosis pathway to PC death, we first compared tonsillar B cells with in vitro-generated PC for the expression of several transcripts encoding receptors, ligands, or signaling molecules associated with the DR pathway. As illustrated in Fig. 2A, three transcripts are up-regulated in PC: caspase-8, CD178 (CD95 ligand), and DR3 mRNAs. Expression of the CD178 transcript was also found in sorted ex vivo tonsillar PC by RT-PCR (data not shown). Expression of TRAIL-R1, TRAIL-R2, TNFRI, and CD95 proteins was next analyzed by flow cytometry on ex vivo tonsillar PC and in vitro-generated PC. Both PC populations display a similar staining pattern, characterized by a high density of CD95 expression and virtually no expression of the other three DR (Fig. 2B). Due to the unavailability of Abs suitable for immunofluorescence staining, immunoblotting was used to compare the expression of DR3 and DR6 in tonsillar B lymphocytes and in vitro-generated PC (Fig. 2C). DR6 is expressed in neither of these populations, but DR3 protein is clearly up-regulated (ϳ9-fold according to densitometry scanning of the bands) in PC compared with tonsillar B lymphocytes. Together, these findings indicate that PC express two of the six known DR: CD95 and DR3.
To address the possible contribution of endogenously produced CD95 or DR3 ligands to PC death, we tested the impact of soluble DR3 and CD95-Fc fusion proteins on spontaneous apoptosis of PC in vitro. The TNFRI-Fc fusion protein was used as a negative control in these experiments. As shown in Fig. 2D, none of the three DR-Fc fusion proteins tested was able to prevent spontaneous disruption of the mitochondrial transmembrane potential in cultured PC. The combination of DR3-Fc and CD95-Fc equally failed to protect PC from spontaneous apoptosis (data not shown).
To check the functionality of the soluble chimeric CD95 and DR3-Fc proteins, we also analyzed their capacity to inhibit PC apoptosis induced by their respective soluble trimeric ligands. Fig.  2D shows that both CD178 and TL1A, the ligand of DR3 (18), raised the proportion of apoptotic PC above the level of spontaneous apoptosis. PC death promoted by CD178 and TL1A was selectively blocked upon addition of their corresponding DR-Fc fusion protein, whereas the TNFRI-Fc repressed neither spontaneous nor death ligand-induced PC apoptosis. This demonstrates that 1) both DR3-Fc and CD95-Fc efficiently prevent binding of CD178 and TL1A to their cellular receptor; and 2) the use of chimeric receptor-Fc fusion proteins does not introduce artifactual modulation of apoptosis in our experimental system. These results testify that PC death is not driven by endogenous production of the DR3 or CD95 ligands.

Caspase requirements for spontaneous PC apoptosis
The results presented above rule out the participation of endogenously produced death ligands in PC suicide. They do not exclude the possibility that PC death could be driven by ligand-independent self-aggregation of DR, as previously described (19). To address this question, we determined the kinetics of activation of caspase-8 and caspase-3 in cultured PC by immunoblotting. If caspase-8 activation occurs within a DR signaling complex, its active cleavage product should be released early on and should precede appearance of the cleaved form of caspase-3. As illustrated in Fig. 3, the kinetics of activation of caspase-8 in apoptotic PC do not fit with this scenario. Its p18 active cleavage product appeared only after 3 h of culture, whereas the p17 active cleavage product of caspase-3 was already detectable by 30 min of culture and peaked by 3 and 6 h of culture. Caspase-8 is thus unlikely to initiate PC apoptosis because its activation is subsequent to that of caspase-3.
Finally, to definitely rule out contribution of the extrinsic pathway to spontaneous PC death, we investigated whether caspases intervene upstream or downstream of the mitochondria during this process. Mitochondrial perturbations are expected to be caspase dependent in the extrinsic pathway and caspase independent in the intrinsic pathway. In PC generated in vitro, the pan-caspase inhibitor z-VAD prevents DNA degradation in a dose-dependent fashion, but fails to block phosphatidylserine exposure and the ⌬m drop (Fig. 4A). The same holds true for in vivo-generated mouse polyclonal PC (Fig. 4B). These findings demonstrate that caspase activation during spontaneous PC apoptosis contributes to the nuclear damage and exclusively takes place downstream of the mitochondria.
To confirm the identities of the caspases involved, in vitro-generated human PC were cultured in the presence or the absence of 1) the broad-range caspase inhibitor z-VAD-fmk, 2) the caspase-8-specific inhibitor z-IETD-fmk, 3) the caspase-9-specific inhibitor z-LEHD-fmk, and 4) the caspase-3-specific inhibitor z-DEVDfmk. As shown in Fig. 4C, only the pan-caspase inhibitor z-VAD and the caspase-3-specific inhibitor significantly prevented DNA fragmentation in PC. The inability of both the caspase-8 and caspase-9 antagonistic peptides to block nuclear apoptosis in PC is compatible with the hypothesis that neither of these initiator caspases is a key player in the PC apoptotic process. Together, these findings argue against implication of DR in spontaneous apoptosis of PC.

Mitochondria contribute to, but do not initiate, PC apoptosis
As shown above, the caspase-9 inhibitory peptide is unable to prevent nuclear PC death. Moreover, although intermediate cleavage products of caspase-9 were found in apoptotic PC, its active p17/ p18 cleavage product remained undetectable whatever the time point considered (data not shown). Nevertheless, these findings do not formally exclude a role for caspase-9 in initiating the caspase cascade in apoptotic PC, because this initiator caspase does not necessarily require proteolytic processing to be enzymatically active (20). To decipher the role of the mitochondria in the PC death pathway, we next examined the activation of Bax with an Ab recognizing an epitope that is exposed in the activated (i.e., membrane-inserted) form of the molecule. As illustrated by Fig. 5, A and B, apoptotic PC express the activated form of Bax. The pancaspase inhibitor z-VAD does not reduce the proportion of PC with active Bax in unstimulated cultures, whereas it prevents the rise of Bax ϩ cells induced by the agonistic anti-CD95 Ab (Fig.  5C). This finding is coherent with our previous observation that the ⌬m drop in apoptotic PC is caspase independent. We next explored the subcellular location of Bax during PC apoptosis. In vitro-generated PC were double stained with the Ab recognizing the activated form of Bax (green fluorescence) and with an Ab directed against the mitochondrial marker Cox (red fluorescence). PC were stained before (Fig. 5D) and after a 6-h culture period conducted in the absence (Fig. 5E) or the presence (Fig. 5F) of an agonistic anti-CD95 Ab. As expected, freshly isolated PC displayed punctate staining with the Cox Ab, but did not express the active form of Bax. After 6 h of culture in the absence of exogenous stimulus, apoptotic PC expressed activated Bax, but, unexpectedly, no colocalization was found between the active form of Bax and Cox. This finding is not imputable to a technical flaw, because the Cox and activated Bax stainings colocalized (yellow staining) when PC were stimulated with an agonistic anti-CD95 Ab known to promote translocation of Bax to the mitochondria.
Together, these results suggest that Bax activation takes place in an organelle other than the mitochondrion during PC apoptosis.

ER stress is primarily responsible for PC apoptosis
We next examined the subcellular localization of Bax during PC apoptosis by immunoelectron microscopy. As shown in Fig. 6, A and B, gold particles bound to the Ab recognizing the activated form of Bax were predominantly associated with the ER and were rarely found in the mitochondria or nucleus of apoptotic PC. These data were confirmed by confocal microscopy using the same anti-Bax Ab (red staining) and the ER marker PDI/protein disulfide isomerase (green staining). The activated form of Bax was not detectable in freshly produced PC (Fig. 6C), but was expressed in most PC after 6 h of culture (Fig. 6F). PDI showed a perinuclear distribution typical of the ER-Golgi network in both viable (Fig.  6D) and apoptotic (Fig. 6G) PC. In most cases, the fluorescent signal from the anti-Bax Ab overlapped with that from PDI in cultured PC (Fig. 6H). On the average, 60 and 30% of the Baxexpressing cells after 6 h of culture displayed complete or partial overlap of PDI and Bax stainings, respectively (data not shown).  This indicates that Bax either relocates to the ER or is activated at the ER during PC apoptosis.
To further document the participation of the ER in PC apoptosis, we monitored the activation of caspase-4 during apoptosis of human PC. Human caspase-4 has recently been demonstrated to be localized to the ER membrane and is likely to fulfill the same function as murine caspase-12 inasmuch as it mediates apoptosis induced by ER stress and amyloid ␤ treatment in human cells (9). As shown in Fig. 7A, the cleaved form of caspase-4 appears by 1 h of culture in apoptotic PC. We also analyzed the kinetics of the cytosolic release of cyt c and AIF in cultured PC. As shown in Fig.  7B, cyt c was undetectable in the cytosol after 1 h of culture, but appeared at later time points (4 h). By contrast, AIF, although present in the total cellular extracts, was not released by the mi-tochondria into the cytosol during the first 6 h of culture. Taken together, these data indicate that the first sign of ER stress (i.e., caspase-4 activation) precedes the release of apoptogenic factors by the mitochondria.
Moreover, the processing of the effector caspase-3 in apoptotic PC was strictly under the control of caspase-4, because addition of the caspase-4 inhibitory tetrapeptide z-LEVD in PC cultures completely prevented proteolytic cleavage of caspase-3 (Fig. 7C). In agreement with this observation, z-LEVD, as the pan caspase antagonist z-VAD and the caspase-3 antagonist z-DEVD, inhibited DNA fragmentation in dying PC (Fig. 7D). Together, these results indicate that the ER-associated caspase-4 connects ER to effector caspases during PC apoptosis.

Spontaneous PC apoptosis does not involve the extrinsic DR pathway
In the present work we have gathered several pieces of evidence indicating that the extrinsic apoptosis pathway does not contribute to PC death. First, capase-8 activation in apoptotic PC is a late event, posterior to caspase-3 cleavage. Second, neither the caspase-8 inhibitory peptide z-IETD nor the soluble forms of DR3 and CD95 DR were able to prevent PC death. Third, both the mitochondrial ⌬m loss and the conformational change in Bax occur independently of caspase activation during PC apoptosis. Our findings are in contradiction with those of Ursini-Siegel et al. Bax is not activated at the mitochondrion membrane during PC apoptosis. A-C, Immunoenzymatic staining of apoptotic PC with an anti-Bax Ab. In vitro-generated human PC were cultured for 3 h before being stained with a nonimmune rabbit control serum (A) or a rabbit Ab revealing the conformational change associated with the membrane insertion of Bax (B). C, In vitro-generated human PC were cultured for 3 h in the presence or the absence of z-VAD-fmk (150 M) with or without the agonistic anti-CD95 Ab 7C11 before immunoenzymatic staining with the anti-activated Bax Ab. Bax ϩ cells were scored by photon microscopy. Results are expressed as the percentages of cells with activated Bax, calculated as the mean Ϯ SD of duplicate determinations. D-F, Subcellular localization of activated Bax in apoptotic PC. In vitro-generated PC were stained with the rabbit Ab recognizing the active conformation of Bax (in green) and with an mAb directed against the mitochondrial marker Cox (in red). Stainings were analyzed by ␣ confocal microscopy. PC were stained immediately after isolation from secondary cultures (D) or after a 6-h culture in complete medium with (F) or without (E) the agonistic anti-CD95 Ab 7C11. Note that there is no overlapping (yellow staining) in PC cultured for 6 h in complete medium, whereas there is a significant proportion of yellow patches in PC treated with the agonistic anti-CD95 when the two fluorescence stainings are merged. The data shown are representative of three separate experiments. (21), who documented that PC death implies an autocrine or paracrine loop involving the death ligand TRAIL. It is doubtful that this discrepancy could merely be imputable to our in vitro model of PC production. We have confirmed in in vivo-generated mouse PC (Fig. 4B) that the mitochondrial alterations accompanying spontaneous PC death are caspase independent, which argues against the contribution of DRs. Beyond the differences pertaining to the experimental models used in our study and that by Ursini-Siegel et al. (21), the possibility that PC attrition operates through different mechanisms depending on their maturational stage is worthwhile considering. It is noteworthy that a minor population of TRAIL-R1 ϩ cells is found within ex vivo tonsillar PC, whereas in vitro-generated PC completely lack expression of this receptor. Similarly, some ex vivo tonsillar PC have retained expression of CD79b, whereas this marker is lost in in vitro-generated PC. This suggests that PC that develop in vivo exhibit a higher degree of heterogeneity than the PC population we generated in our in vitro culture system. TRAIL-sensitive PC could thus be more represented within the postimmunization splenic CD138 ϩ population studied by Ursini-Siegel et al. (21) than within our in vitro-generated human PC population. This raises the question of the in vivo counterpart of the PC population produced in our culture system. Three considerations lead us to propose that our in vitro-generated PC population is essentially constituted of fully mature PC. First, a large proportion of these cells express high levels of CD138. In humans, CD138 is one of most reliable markers that allows discrimination between plasmablasts and mature PC (22,23). Second, our culture system generates occasional cells with prominent Russel bodies, which represent the ultimate PC differentiation stage (Fig. 1C). Third, our use of anti-CD20 and DQ Abs for PC enrichment at the end of the secondary culture favors elimination of early PC that still express these markers at low density. Autocrine or paracrine DR-mediated apoptosis could thus be associated with plasmablasts rather than mature PC. This point deserves investigation. Nonetheless, in vitro-generated human PC are prepared for autonomous triggering of the extrinsic apoptosis pathway. They express two functional DRs as well as the transcript for the CD95 death ligand (CD178), as previously observed by Strater et al. (24). Nevertheless, the function of CD178 in PC physiology remains elusive. If the death pathway responsible for the demise of PC in the periphery is repressed in the BM, an alternate apoptosis mechanism could be required to maintain homeostasis of the BM memory PC compartment confronted with the constant feeding with newly produced PC. The CD95/CD178 couple could then be instrumental in promoting contraction of the resident BM PC pool to accommodate establishment of the incomers.

The ER initiates PC apoptosis
Although both mitochondria and the ER are obviously involved in PC apoptosis, our experimental results support the idea that mitochondria are minor contributors. First, immunoelectron and confocal microscopies show that the active form of Bax is almost exclusively found at the ER and not at the mitochondria membrane during PC apoptosis. Second, the ER-associated caspase-4 is proteolytically cleaved during PC apoptosis before apoptogenic factors are released from the mitochondria. Third, the caspase-4, but not the caspase-9, antagonist completely blocks the cleavage of the effector caspase-3, which strongly suggests that initiation of the caspase cascade is subordinated to the ER rather than to the apoptosome. Nevertheless, more than one apoptotic pathway can be elicited in response to ER stress. This is suggested by the observation that caspase-12 Ϫ/Ϫ mouse embryonic fibroblasts are only partially resistant to ER stress-induced apoptosis (10). At face value, our finding that caspase inhibitors do not fully prevent DNA fragmentation in PC points to the possibility that both a caspasedependent and a caspase-independent mechanism cooperate to promote PC apoptosis. The mitochondrial apoptogenic factor AIF could be a key mediator of the caspase-independent pathway, because it possesses an intrinsic DNA fragmentation activity that does not require caspase activation (25). Although we found no evidence for the cytosolic release of AIF in apoptotic PC during Total protein extracts prepared from in vitro-generated PC after 30 min, 1 h, and 3 h of culture were immunoblotted with an anti-caspase-4 Ab. The myeloma cell lines IM9 and U266, which exclusively exhibit the proform and the p30 cleaved form of caspase-4, respectively, were used as positive controls (first left two lanes). B, Kinetics of production of mitochondrial apoptogenic factors in apoptotic PC. Cytosolic extracts prepared from in vitro-generated PC after 1, 4, and 6 h of culture were immunoblotted with Abs recognizing cyt c, AIF, Cox, and ␤-actin. Total cell extracts prepared from freshly isolated PC (first lane on left) were used as controls. The absence of Cox in the cytosolic fractions guarantees that there has been no accidental release of inner mitochondrial material during preparation of the cytosolic extract. C, Influence of a caspase-4 antagonist on the activation of caspase-3. In vitro-generated PC were cultured for 6 h in the presence or the absence of the caspase-4 antagonist z-LEVD (150 M) and stained with an mAb recognizing specifically the cleaved active form of caspase-3 (plain lines) or with an isotype-matched unrelated mAb (dotted lines). D, Impact of the caspase-4 antagonistic peptide on different apoptosis parameters of PC. In vitro-generated PC were cultured for 12 h with or without z-LEVD before being assessed by flow cytometry for membrane, mitochondrial, and nuclear alterations and for expression of the active form of caspase-3. the first 6 h of culture, this does not exclude that this factor could intervene at later stages of the process. The nature of the signaling mediator that connects ER to the mitochondria in PC is unknown, but Ca 2ϩ stands as a plausible candidate. The respective contributions of the ER and mitochondria to ER stress-induced apoptosis may vary depending on the volume these organelles occupy within the cell. It is conceivable that the massive surface area of the ER in PC allows it to elicit a strong caspase-12/4 signal in response to ER stress, thus rendering the contribution of the mitochondria dispensable. This situation could be compared with that described for CD95-induced apoptosis, in which the mitochondria only contribute to execution of the death program when the initiator caspase signal provided by the CD95 death-inducing signaling complex is too weak (26).

Is PC longevity subordinated to efficiency of the UPR?
By analogy with the neuronal cell loss driven by the accumulation of misfolded proteins in Alzheimer's (10) and Parkinson's (27) diseases, it can be envisaged that an ER stress signal is generated in PC when their Ig secretion rate exceeds the folding capacity of the ER. The UPR, which is responsible for homeostasis of the ER function, attenuates general protein translation, specifically up-regulates the expression of ER chaperones, and eventually contributes to promote cell death (28). The existence of long-lived PC demonstrates that apoptosis is not necessarily an inescapable issue for fully differentiated PC. Thus, under certain conditions, PC can behave as pancreatic ␤ islet cells, which can survive the ER stress associated with their secretory function. It follows that the UPR could be more efficient in long-lived PC than in short-lived PC. Because BM stromal cells (29) or the factors they produce (30) are required to support the survival of isolated BM PC, it is unlikely that their exceptional longevity is determined by an intrinsic genetic differentiation program. This points to the possibility that the factors produced in the BM environment could modulate the UPR in PC.
In conclusion, our present results suggest that the ER is a crucial regulator of PC survival. Additional experiments will determine whether targeting certain components of the UPR system or ERassociated antiapoptotic proteins, such as Bax inhibitor-1 (31), can sensitize PC for apoptosis. These studies might later be exploited for therapeutic purposes to promote cell death in diseases such as multiple myeloma or Ab-mediated autoimmune diseases that involve aberrant survival of PC.