|Home | About | Journals | Submit | Contact Us | Français|
Interactions between the malignant plasma cells of multiple myeloma (MM) and stromal cells within the bone marrow (BM) microenvironment are essential for myeloma cell survival, mirroring the same dependence of normal BM-resident long-lived plasma cells on specific marrow niches. These interactions directly transduce pro-survival signals to the myeloma cells and also induce niche production of supportive soluble factors. However, despite their central importance the specific molecular and cellular components involved remain poorly characterized. We now report that the prototypic T cell costimulatory receptor CD28 is overexpressed on myeloma cells during disease progression and in the poor prognosis subgroups, and plays a previously unrecognized role as a two-way molecular bridge to support myeloid stromal cells in the microenvironment. Engagement by CD28 to its ligand CD80/CD86 on stromal dendritic cell (DC) directly transduces a pro-survival signal to myeloma cell, protecting it against chemotherapy and growth factor withdrawal-induced death. Simultaneously, CD28-mediated ligation of CD80/CD86 induces the stromal DC to produce the pro-survival cytokine IL-6 (involving novel crosstalk with the Notch pathway) and the immunosuppressive enzyme indoleamine 2, 3 dioxygenase (IDO). These findings identify CD28 and CD80/CD86 as important molecular components of the interaction between myeloma cells and the bone marrow microenvironment, and point to similar interaction for normal plasma cells as well as suggesting novel therapeutic strategies to target malignant and pathogenic (e.g. in allergy and autoimmunity) plasma cells.
Multiple myeloma (MM) is the second most common hematologic malignancy (1) and remains incurable for most patients. Myeloma cells are the transformed counterpart of the normal, bone marrow-resident long-lived plasma cells (PC) thought to be responsible for long term (in some cases lifelong) maintenance of protective antibody levels. Both normal PC and transformed myeloma cells are critically dependent on interactions with bone marrow stromal cells (BMSC) within specialized niches for their survival, and in the case of myeloma, resistance to therapy (2–3). These interactions, although poorly defined as to their specific cellular and molecular components, are thought to occur at two interwoven levels. The initial interactions occur during cell-cell contact and involve receptor-ligand binding, most clearly studied for the β-integrins (2, 4) and Notch-Jagged (5–6) families of receptors. This direct contact then induces BMSC to secrete soluble growth and supportive factors (e.g.IL-6, APRIL/BAFF, VEGF among others) that further modulate the microenvironment (2, 7–8). IL-6 is the prototypic example of a stromally produced supportive factor induced by myeloma cells (as well as by normal PC (3)), with elevated serum levels clinically correlated with disease severity (9). This cytokine plays a central role in supporting the growth/differentiation/survival of normal B lineage cells (10–11) and primary myeloma cells (2), and also protects MM cells against chemotherapy-induced apoptosis (11). The in vivo importance of IL-6 in PC/MM cell survival is evidenced by the ability of anti-IL-6/IL-6R monoclonal antibodies (mAb) to significantly reduce autoantibody titers and plasma cell numbers in systemic lupus erythematosis patients (12) as well as having anti-myeloma efficacy in both pre-clinical models (13) and clinical trials in combination with chemotherapy (14). However, the specific molecular and cellular mechanisms involved in the induction of stromal-IL-6 by normal or malignant PC remain poorly characterized, although the integrins (2) and Notch-Jagged (15) have been implicated.
It would be predicted that receptor-ligands involved in pro-myeloma cell survival interactions with the microenvironment would be associated with poor prognosis and disease relapse under treatment pressure. One such receptor is CD28, which has been primarily characterized as the prototypic T cell costimulatory receptor. In T cells, CD28 activation upon binding to its ligands CD80 and/or CD86 expressed on professional antigen presenting cells (APC, predominantly myeloid (or conventional) dendritic cells (DC)) in conjunction with T cell receptor activation (signal 1) provides the essential co-stimulatory signal (signal 2) for full T cell activation, proliferation, effector function, metabolic efficiency and augmented survival (16–18). But CD28 is also expressed on both normal PC and myeloma cells (19), and this expression is specifically suppressed by Pax5 (the master regulator of B cell identity) in normal B cells - and is upregulated during B→PC differentiation as Pax5 is downregulated (20). Although this regulated expression suggests specific B-lineage function, CD28’s role in plasma cell biology is only beginning to be characterized. Clinical evidence in myeloma that CD28 expression correlates with disease progression (21) and poor prognosis (22) suggests a pro-survival role, consistent with previous findings by us and others that in vitro activation of CD28 alone (without a signal 1) in myeloma cells triggers downstream NFκB signaling and protects against apoptosis (23) and induces MM cell production of the pro-angiogenic cytokine IL-8 (24).
A pro-survival role for CD28 points to CD80/CD86+ BMSC as the cellular partners in the myeloma niche. Cells expressing CD80 and CD86 are predominantly B cells and professional APC such as monocyte/macrophages and dendritic cells. Conventional myeloid DC are best characterized as the primary regulators of T cell activation (18), but are also centrally involved in normal plasma cell differentiation (25) through cell contact-mediated interactions as well as DC production of the pro-survival cytokines IL-6 and APRIL/BAFF (8). Consistent with this, we and others have found that both myeloid DC and plasmacytoid DC (pDC), as well as monocyte/macrophages, are selectively increased in myelomatous regions of patient bone marrow and support the survival of primary myeloma cells in a cell-contact dependent manner (23, 26–28). Earlier studies also found that the myeloid DC in the bone marrow of myeloma patients were being induced to make IL-6 (26), although how this occurs remains unknown. Identification of CD28 as a potential component in the pro-myeloma survival interaction with DC (23) raises a possible molecular mechanism for induction of the stromally produced “soluble” microenvironment, as previous work in the completely different context of DC-mediated T cell activation demonstrated that CD28-mediated ligation of CD80 and/or CD86 induced myeloid DC to make IL-6 necessary to fully activate the T cell (29). Furthermore, this same T cell literature found that CD80/CD86 ligation by the CD28 family member CTLA4 induces DC to produce the immunosuppressive enzyme indoleamine 2, 3 dioxygenase (IDO) (30). IDO catabolizes the essential amino acid tryptophan in the microenvironment into the toxic metabolite kynurenine, resulting in T cell anergy via GCN2-kinase mediated sensing of depleted intracellular tryptophan pools (30–33). Whether CD28 on myeloma cells (or normal PC) can also induce DC production of IL-6 and/or IDO through CD80/CD86 ligation has not been previously examined; however, lower levels of tryptophan/higher levels of kynurenine have been observed in the majority of patients with multiple myeloma (34).
Given these preceding observations with T cells, we have examined whether CD28 expressed on myeloma cells interacts with CD80/CD86 on stromal myeloid DC to function as previously unrecognized molecular bridge by which MM cells modulate the microenvironment to generate a pro-survival and immunosuppressive milieu.
All primary human cells were obtained under protocols approved by the Institutional Review Board of the Roswell Park Cancer Institute. Total bone marrow mononuclear cells from de-identified cryopreserved samples from normal allogeneic bone marrow transplant donors or myeloma patients were analyzed for expression for CD11b, CD138 and CD28 in the viable cell gate (FSC/SSC) as previously described (23). Primary CD138+ plasma cells/myeloma cells were isolated as previously described (23). The myeloma cell lines RPMI 8226 and U266 were obtained from American Type Culture Collection (ATCC, Manassas, VA) and the MM1.S cell line was a gift from Dr. Stephen Rosen (Robert H. Lurie Cancer Center, Chicago, IL). Antibodies used: Anti human-CD4, CD25, CD14, CD11b, CD1a, CD19, CD80, CD83 and CD86 (all from Beckman Coulter, (Fullerton, CA), anti-FoxP3 (BD Biosciences, San Jose, CA), murine mAb anti-Notch-1 (Thermo Fisher, Fremont, CA), rabbit polyclonal anti-Jagged 1 or Jagged 2 (Santa Cruz, CA), anti-mouse Ig-PE (VMRD, Pullman, WA), anti-rabbit Ig-APC or PE (Jackson Immunoresearch, PA), anti-human CD28 mAb (mAb 9.3, kind gift of Dr. Bruce Levine, University of Pennsylvania, PA,), CTLA4-Ig and CD28-Ig (R&D Systems, Minneapolis, MN), anti-human IDO antibody (Chemicon, Ternecula, CA), polyclonal anti-IL6 antibodies (Abcam, Cambridge, MA). hCD28 mAb 9.3 (Fab)2 fragments were made with Pierce (Fab)2 kit (ThermoScientific, Rockford, IL), and the purity of the Fab preparations by SDS-PAGE were greater than 95%.
Monocytes were purified from normal human blood using EasySep™ (Stem Cell Technologies, Vancouver, Canada) or the MACS kit (Miltenyi Biotec Inc. Auburn, CA) as per the manufacturer’s instructions and differentiated to DC in RPMI 1640 media (10% FBS, 1000 U/ml penicillin/streptomycin, 4 mM L-glutamine, 0.1 mM non-essential amino acids, 1 mM sodium pyruvate) with GM-CSF (10 ng/ml) and IL-4 (1000 U/ml) for 7–11 days with media changes every 2–3 days. DC were analyzed by flow cytometry as previously described (23) for CD14, CD11b, CD1a, CD19, CD80, CD83 and CD86 Expression of Notch-1, Jagged 1 or Jagged 2 was analyzed by flow cytometry after staining with the appropriate antibodies. Bone marrow aspirates from myeloma patients were co-stained with anti-CD11b-APC and with PE-conjugated antibodies against DC surface markers.
Gene expression data from normal plasma cells (NPC, n=22), monoclonal gammopathy of undetermined significance (MGUS, n=44), smoldering myeloma (SM, n=12) and newly diagnosed multiple myeloma (MM, n=538) were generated from gene expression profiles performed at the University of Arkansas (NCBI Gene Expression Omnibus GSE5900 and GSE4204) (35). Based on dataset annotation, the 538 MM samples were sorted into 8 subsets as previously described (35). Data for CD28 probe set were collected and analyzed using Prism software (GraphPad Software Inc). The unpaired t-test was used to assess significance between the different groups and subtypes.
MM cells (5 × 104 cells/well) were co-cultured with DC (5 × 104, 2.5 × 104 or 1.25 × 104) in RPMI 1640 media with or without arsenic trioxide (5 µM), melphalan (10–20 µM), dexamethasone (10 µM) or in serum free media for 48–72 hr. 2-chamber Transwell (HTS Multiwell Insert System, BD Falcon, Bedford, MA) co-cultures were done using a 1 µm pore size filter membrane to separate the DC from the MM cells. Blockade of the CD28–CD80/CD86 interaction was done using anti-hCD28 mAb 9.3 (Fab)2 fragments (50 µg/ml) or CTLA4-Ig + CD28-Ig (200 µg/ml) added prior to co-culture. At 48 and 72 hrs, co-cultured cells were stained with CD11b-PE + annexin V-FITC + 7AAD, and the % viable CD11b-negative myeloma population was determined by flow cytometry. DC-MM clustering was visualized by using CFSE-labeled (5 µM) U266 cells.
DC alone (5 × 105 cells/well) were cultured with CTLA4-Ig (50 µg/ml, as a positive control) or co-cultured with mitomycin-treated (100 µg/ml × 15 min) MM1.S myeloma cells (5 × 105 cells/well) ± αCD28 (Fab)2 (50 µg/ml). Primary purified myeloma cells were used without mitomycin C treatment. In other experiments, DC were preincubated × 2 hr with the gamma secretase inhibitor DAPT (GSI, 25 µM, Calbiochem, San Diego, CA) prior to co-culture. Cell fixation was done with 2% paraformaldehyde in PBS for 15 min. After 48 hrs, cell free supernatants were assayed by sandwich ELISA in Nunc Immuno™ Maxisorp™ plates (VWR Scientific, Rochester, NY) using capture and detection antibodies against human IL-6 (R&D Systems), per the manufacturer’s protocol or by using Luminex assays (Luminex® 100, Luminex Crop, Austin, TX). Total RNA from the co-cultures were prepared using TRIZOL™ (Invitrogen, Carlsbad, CA) according to manufacturer’s protocol. RT-PCR was done as described earlier (36) using specific primer sets for human IL-6 (5′ – AGAAAACAACCTGAACC TTCC and 3′ – CTACATTTGCCGAAGAGCC, product size 382 bp) and for DC specific marker Fascin-1 (5′ – ACC AAAAAGTGTGCCTTCC and 3′ – CACCGTCCAGTATTTGCC, product size 402 bp). Densitometric analysis was performed with ImageJ software (NIH, Bethesda, MD).
DC alone (1 × 105 cells/well) were cultured with CTLA4-Ig (50 µg/ml as a positive control) or with myeloma cells (U266 or CD138+CD28+ primary myeloma cells) ± IFNγ (1000 U/ml) ± anti-CD28 (Fab)2 (40 µg/ml) in media containing 100 µM tryptophan. After 48 hr, supernatants were assayed for L-kynurenine content as a measure of IDO activity as previously described (37). IDO expression was determined by Western blot (36) using a polyclonal anti-human IDO antibody, and HEK293-IDO cells stably expressing human IDO were used as positive control (kind gift of Dr. Kunle Odunsi, Roswell Park Cancer Institute). Intracellular staining for IDO was performed as previously described (33), using CD11b-FITC and anti-human IDO antibody (5 µg/ml) and the Cytoperm/Cytofix kit (BD Biosciences, Franklin Lakes, NJ). To demonstrate the immunosuppressive activity of IDO induced in DC-MM co-cultures on T cell proliferation, supernatants were collected from co-cultures of paraformaldehyde fixed U266 + DC in 500 µl media ± the IDO inhibitor INCB014943 (1 µM, provided under MTA by Incyte Corporation, Wilmington, DE).
The myeloma cell line RPMI 8226 was cultured in fresh media mixed (1:1) with media conditioned by DC:MM co-culture ± dexamethasone (2.5 µM for proliferation assays and 1 µM for viability assays). For proliferation assays, [H3] thymidine was added after 48 hrs as described earlier (23). In some conditions, azide-free neutralizing polyclonal anti-IL6 antibodies (25 µg/ml) were added. Viable myeloma cells were quantified by flow cytometric analysis with AnnexinV-FITC/7AAD staining.
Purified normal human T cells (1 × 105/well) were cultured in media conditioned by DC-MM co-cultures diluted 1:1 with fresh, tryptophan-free media (Hyclone Trp− RPMI 1640 without FBS), activated with PMA (5 ng/ml) + ionomycin (200 ng/ml) and proliferation assayed as previously described (38). DC that were previously cocultured with myeloma cells in the presence of IFNγ (IDO activity assays), were isolated by negative selection of the myeloma cells with CD138 coated magnetic beads (Stem Cell Technologies, Vancouver, Canada). These DC were co cultured with naïve allogeneic T cells for 7 days, and then analyzed by flow cytometry for Treg induction by T cell surface expression of CD4, CD25 and intracellular FoxP3 (FoxP3 Staining Kit, BD Biosciences, San Jose, CA). DC cultured in media with or without IFNγ was used as controls.
Previous flow cytometric studies have shown that CD28 expression on myeloma cells correlates with tumoral expansion (21), is associated with the poor-prognosis t(14:16) and t(14:20) translocations subgroups that have activated c-MAF and MAFB proto-oncogenes (35, 39), and predicts worse outcome with high dose therapy (40). Similarly, we have found in analyzing bone marrow mononuclear cells from newly diagnosed MM patients and normal allogeneic bone marrow transplant donors using CD138 as a plasma cell marker that a substantial and similar percentage of both normal PC and myeloma cells were positive for CD28 expression (Table I, representative FACS plot in Fig. 1A). We extended our observations by examining CD28 expression by genome-wide expression profiling of purified plasma cells from normal donors (NPC), newly diagnosed patients with monoclonal gammopathy of undetermined significance (MGUS), smoldering myeloma (SM) and active multiple myeloma (MM). As seen in Fig. 1B and Table II, there was significantly increased CD28 expression in SM and MM vs. NPC, with a trend towards increased expression in MGUS vs. NPC. Further analysis of the MM group (n=538) subsetted by the 8 different genetic signatures defined previously (activation of cyclin D family members (CD-1, CD-2), MMSET (MS), c-MAF and MAF-B (MF), hyperdiploidy (HY), proliferation (PR), low bone disease (LB), and myeloid contamination (MY) (35)) demonstrated that CD28 expression was not equally expressed among the subgroups, but was markedly higher in the poor-prognosis MF subgroup (consistent with the previous flow cytometric based studies) compared to all the other subgroups (Fig. 1C, Table II). These findings further support a clinically relevant pro-survival role for CD28 in myeloma cells.
A role for CD28 in supporting myeloma cell survival suggests the presence of stromal cells expressing the CD28 ligands CD80 and CD86 in the microenvironmental niche. This is consistent with previous histological studies of patient BM biopsies that CD80/CD86+ myeloid APC (DC, monocytes, macrophages) were more abundant in the myelomatous regions compared to uninvolved regions (23, 27). We have now found that compared to normal donor controls, there was a significantly higher percentage of bone marrow mononuclear cells expressing the myeloid marker CD11b in MM patients (Fig. 2A and Table I) despite the fact that the normal marrow elements are being displaced by the myeloma itself. This data suggests that myeloid APC are being actively recruited into the myeloma pro-survival niche. Flow cytometric analysis of this CD11b+ myeloid population in the bone marrow aspirates from four myeloma patients demonstrated a phenotype consistent with an unactivated tissue-resident DC expressing CD11c and CD40 - with low-intermediate levels of CD80, CD83 and CD86, CD40 and MHC class II (Supplemental Fig. 1B). The degree of expression of CD83, CD80 and CD86 was somewhat variable between patients (although all patients had CD80+ and CD86+ DC subpopulations), possibly due to the different stages of myeloma in these patients. To further characterize the phenotype of the DC infiltrating the myelomatous areas in patient bone marrow biopsies, we examined 5 different patient bone marrow biopsies by immunohistochemistry for fascin+ and CD83+ cells (which are both markers characteristic of more mature DC)(Table III). Consistent with the flow cytometric analysis and our previous findings (23), we found that the DC in the myeloma infiltrates were of a more mature fascin+ /CD83+ phenotype.
Among the various myeloid populations, we and others have found that dendritic cells support the survival of myeloma cells in a contact dependent fashion (23, 41). To determine if CD28 and CD80/CD86 are molecular components of the MM-DC pro-survival interaction, we first co-cultured CD28+ MM cells with human monocyte-derived dendritic cells in vitro. The DC generated were phenotypically more characteristic of tissue resident unactivated/immature DC, similar to myeloid DC in myeloma bone marrow (Supplemental Fig. 1A and 1B), and expressed CD11b, lower levels of CD80 and CD86 and the immature DC marker CD1a (Supplemental Fig. 1A). Co-cultures of CFSE-labeled myeloma cells and unstained DC reveal distinct clustering of MM cells with the DC (Fig. 2B). The death of MM1.S myeloma cells caused by two chemotherapeutic agents with differing mechanisms of action (arsenic trioxide or melphalan) was significantly reduced by co-culture with DC (Fig. 2C). Co-culture with DC also protected MM1.S cells and primary CD138+ CD28+ myeloma cells (Pt 1–3) against growth factor withdrawal-induced death in serum-free conditions (Fig 2D), which induces apoptosis through different mechanisms than the chemotherapeutic agents. This DC-mediated protection required cell-cell contact, as it was completely lost for both MM1.S and primary myeloma cells when the DC and MM cells were separated by a Transwell membrane (Fig. 2E).
Involvement of CD28 in the complex DC-MM cell-cell interaction was examined by blocking CD28 with anti-CD28 (Fab)2 fragments (which did not alter expression of CD80/CD86 on DC (Supplemental Fig. 1C)), which had no effect on survival of MM1.S cells alone but significantly diminished the ability of DC to protect MM cells in co-cultures under serum starvation conditions (Fig. 3A). Consistent with this, blocking CD80 and CD86 on the DC with the soluble chimeric fusion receptors CTLA4-Ig and CD28-Ig (recombinant proteins of the extracellular domain of CTLA4 or CD28 genetically fused to the Fc portion of human IgG) also significantly reduced DC-mediated protection of myeloma cells against serum starvation. Interestingly however, this reduction was less compared to blocking CD28, and the possibility that ligation of CD80 and/or CD86 by CTLA4-Ig/CD28-Ig might be eliciting DC responses that offset the loss of survival signals through CD28 activation was explored further below. When cell death was induced by melphalan or dexamethasone instead of serum starvation, blockade of CD28 with αCD28 (Fab)2 or CD80/CD86 with CTLA4-Ig had the same effect of abrogating DC-mediated protection of co-cultured MM1.S (Fig. 3B).
Previous evidence that CD28-Ig-mediated crosslinking of CD80/CD86 induced DC production of IL-6 (29) and our observation that CD28-Ig + CTLA4-Ig was less effective in blocking the pro-survival effects of DC co-culture led us to examine whether myeloma cells could induce DC IL-6 production in a CD28–CD80/CD86− dependent fashion. While DC and MM1.S made little to no IL-6 when cultured alone, MM-DC co-culture significantly increased IL-6 to levels comparable to the positive control of ligating CD80/86 with CTLA4-Ig, and this was completely inhibited by blocking CD28 with anti-CD28 (Fab)2 (Fig. 4A). Upregulation of IL-6 protein was mirrored by a similar increase in IL-6 mRNA level (Fig. 4B, normalized to fascin as a DC-specific housekeeping gene). Chemical fixation of DC but not of the myeloma cells completely abrogated IL-6 production (Fig. 4C), indicating that the DC are the cells producing the cytokine. Primary CD28+ myeloma cells also induced DC IL-6 production (Fig. 4D, Pt 5 and 6), whereas CD28-negative myeloma cells from Pt 7 induced considerably less IL-6. Similar to the cell lines, upregulation of IL-6 by primary MM cells appeared to be at the mRNA level (Fig. 4E).
Our findings parallel previous observations that Jagged 2 expressed on myeloma cells ligates Notch receptors on BM stroma and induces BMSC IL-6 production (15, 42). Given that anti-CD28 (Fab)2 nearly abolishes MM-induced DC IL6 production, if IL-6 is also being induced via Notch signaling in DC this would indicate a previously unrecognized interaction between these two pathways. DC used in our study expressed Notch-1, Jagged 2 but not Jagged 1, while MM1.S cells expressed Notch-1, Jagged 1 and 2 (Fig. 4F top), consistent with previously published reports (15, 43). Induction of IL-6 production in the MM-DC co-cultures was significantly inhibited by blocking Notch signaling with the gamma secretase inhibitor (GSI) DAPT (Fig. 4F, bottom), and similar results were seen using blocking Notch-1 or Jagged 2 antibodies (not shown). This demonstrates that myeloma cells can also elicit IL-6 production from myeloid DC through Jagged-Notch interactions. Given that the marked inhibition of IL-6 production by blocking either the CD28–CD80/CD86 or Jagged -Notch-1 pathways individually cannot readily be accounted for if the pathways were independent, these findings suggest there is essential crosstalk between two pathways.
To determine the potential biological relevance specifically of IL-6 in the context of all the known and unknown soluble factors induced by MM-DC co-culture, we examined the ability of the IL-6 in the supernatants from these co-cultures to overcome the anti-proliferative and apoptotic effects of dexamethasone, which is widely used in the treatment of multiple myeloma. As seen in Fig. 5A, dexamethasone-mediated suppression of RPMI 8226 proliferation was significantly ameliorated when these cells were cultured in conditioned supernatants from the DC-MM cocultures, and this effect was lost when neutralizing anti-IL-6 antibodies were added. The viability of RPMI 8226 cells in the presence of DC-MM co-culture supernatants was also significantly increased in the face of dexamethasone treatment, which was also neutralized by anti-IL6 antibodies (Fig. 5B). These findings suggest that myeloma-induced DC IL-6 production may be contributing to chemotherapy resistance by supporting both MM cell survival and proliferation, although our initial data (not shown) in both myeloma cells and normal plasma cells indicates that IL-6 is not essential for PC survival if they are in direct contact with stromal DC.
The same body of work that identified CD80/CD86-mediated induction of DC IL-6 production also found that ligation of CD80/CD86 on DC by CTLA4 expressed on regulatory T cells (Treg) or by the soluble CTLA4-Ig molecule alone, in the presence of interferon (IFN), induces myeloid DC production of the immunosuppressive enzyme indoleamine 2, 3 dioxygenase (IDO) (30). IDO catabolizes the essential amino acid tryptophan in the microenvironment into the toxic metabolite L-kynurenine, and the depleted intracellular tryptophan pools are sensed by GCN2 kinase and result in T cell inactivation (32). Given that myeloma patients characteristically have suppressed T cell responses, decreased serum tryptophan and elevated serum kynurenine levels (30), as well as increased BM infiltration of pDC (28, 34) (which have been implicated as the source of IFN in Treg-induced IDO production(44–46)), we examined whether myeloma CD28 ligation of CD80/CD86 on DC also induced IDO. When DC were co-cultured with U266 myeloma cell line + IFNγ, IDO activity as measured by kynurenine levels in the media was up-regulated 3 to 9 fold over that of DC + IFNγ alone, comparable to IDO activity induced by the positive control CTLA4-Ig (Fig. 6A). Other myeloma cell lines (RPMI 8226, MM1.S and KMS11) also induced significantly greater IDO activity when co-cultured with DC compared to cells cultured alone (supplemental fig. 2). Similar IDO induction was also observed when primary CD138+CD28+ myeloma cells (Pt 9, Pt 10) were co-cultured with DC + IFNγ (Fig. 6B). This induction was largely CD28 dependent, as blockade with αCD28 (Fab)2 fragments resulted in a ~7-fold reduction in IDO activity (Fig. 6C).
The level of IDO activity is both a reflection of its expression level and enzymatic activity, as IDO exists in both inactive and active states (47) and requires IFNγ for activity (33). We found that the increase in IDO activity in the DC + U266 + IFNγ co-cultures correlated with an increase in IDO protein expression vs. DC + IFNγ (Fig. 6D). While IDO protein is induced in DC + U266 co-culture without IFNγ (Fig. 6D), there was only low level IDO activity as determined by kynurenine production (Supplemental Fig. 2B). Intracellular staining of co-cultures indicates the CD11b+ DC population was the sole producer of IDO (Fig. 6E), and that the entire population of DC are producing the enzyme. This finding also indicates that the same DC induced to make IL-6 via CD80/CD86 engagement are also making IDO, which is consistent with evidence for the simultaneous induction of DC IL-6 and IDO production by myeloma cell coculture (Supplemental Fig. 2B). While other groups have shown that the IDO2 isoform of IDO (which is produced by a small subset of tissue-resident DC, but has not been shown to be produced by human monocyte-derived DC used in our co-cultures) can block IL-6 gene expression in murine pDC (48), our data suggests this may be different for human myeloid DC.
Myeloma-induced DC IDO production could suppress anti-myeloma T cell responses through two potential mechanisms – generation of regulatory T cells (Treg, which are increased in MM patients (49–50)) and/or direct suppression of effector T cell activation. To begin to examine the first possibility, we assessed the ability of DC first cocultured with myeloma cells, subsequently purified (via negative immunomagnetic selection of the CD138+ myeloma cells) and then re-cocultured with naïve resting T cells to generate CD4+CD25+FoxP3 Treg in vitro (Fig. 6F). We observed the generation of a CD4+CD25+FoxP3+ population when T cells were cocultured with DC previously cocultured with myeloma cells + IFNγ. However, Treg generation was no greater than in cocultures with DC treated with IFNγ alone, which also are induced to express IDO (albeit at a level significantly less than DC from the MM + IFNγ conditions). However, even though Treg generation was equivalent, only the DC + MM + IFNγ conditions were able to suppress human T cell activation and proliferation induced by PMA + ionomycin (Fig. 6G), where IDO activity in DC + U266 + IFNγ supernatants was sufficient to cause a 6.5-fold reduction in T cell proliferation over that DC + IFNγ alone. IDO activity could be completely inhibited by the addition of a highly selective IDO inhibitor INCB014943 (provided by Incyte Corporation, Wilmington, DE) to the MM-DC co-cultures (Fig. 6G, left panel), which was reflected in the significant de-repression of T cell proliferation (Fig. 6G, right panel). These findings suggest that maximal Treg generation may occur at low levels of DC IDO production (such as produced by DC treated with IFNγalone), but direct suppression of effector T cell activation requires the induction of higher DC IDO production induced by myeloma cell coculture.
Survival of myeloma cells is critically dependent on interactions with specific bone marrow niche elements that recapitulate those that support the survival of normal bone marrow-resident long lived plasma cells (LLPC). Although the specific molecular and cellular components involved remain largely undefined, a unique characteristic of these interactions is the simultaneous direct transduction of survival signals to the myeloma/LLPC coupled with the induction from the stromal niche cells of supportive factors that further (and perhaps farther) influence the surrounding microenvironment. On the stromal side, the archetypal response is the induction of stromal IL-6 production by cell-cell contact with myeloma cells or LLPC. Although normal and malignant PC are not solely dependent on IL-6 for their survival (3), it is clear that this cytokine is a central component of the network of soluble mediators (including APRIL/BAFF, interferons, etc (51)) that modulate PC differentiation (which may be relevant even in myeloma in the context of the putative MM stem cell (52)) and survival.
Thus the key pro-survival interactions are likely to involve molecules that form the signaling linkages between the myeloma/LLPC and the stromal niche. We have previously demonstrated that activation of CD28 transduces a pro-survival signal directly to myeloma cells, and now have found that myeloma cells interaction with stromal myeloid DC through CD28–CD80/CD86 engagement also induces DC production of both the supportive cytokine IL-6 and the immunosuppressive enzyme IDO. While DC production of IL-6 and IDO induced by CD80/CD86 ligation by CD28 (and CTLA4, which is not expressed on myeloma cells) has been well characterized in the context of T cell activation/inactivation, this is the first report to our knowledge that the same interaction is involved in supporting cells of the B lineage. The specific importance of this interaction is suggested by findings by us and others that CD28 overexpression on myeloma cells correlates with disease progression and most highly with the poor-prognosis MF patient subgroup. There is currently no evidence that implicates c-MAF in the direct regulation of CD28 expression (or vice versa), although c-MAF over-expression in myeloma cells enhances their integrin-mediated adhesion to BMSC (4) and suggests a mechanism for selection of cell contact-mediated pro-survival interactions. The importance of the CD28–CD80/CD86 interaction in supporting BM-resident myeloma cells survival is further supported by our in vivo findings in mice that B lineage knockout of CD28 (as well as global knockout of CD80 or CD86) causes a significant and selective loss of normal bone marrow LLPC without having any effect on the splenic (short lived) PC population (CHR, manuscript submitted).
Identification of CD28’s involvement in PC/MM survival points to cells bearing the CD80/CD86 ligands as components of the microenvironment, in particular macrophage/monocytes and dendritic cells which have been previously identified as being part of myeloma niche (18). Although myeloma cells also express CD86 (but not CD80(53)), our findings that blocking CD28–CD86 between myeloma cells alone had little effect on survival (e.g. Fig. 3A) suggests any pro-survival contribution of cis interaction between myeloma cells is relatively modest. We focused on myeloid DC because they are the predominant cell type that presents CD80/CD86 to T cells, selectively infiltrate myelomatous areas in patient BM (23, 26) and can support PC and MM survival in vitro (41). Consistent with this, we found that the myeloid compartment is increased in the myeloma bone marrow compared to normal controls, and that direct MM-DC contact protects myeloma cells against apoptosis that was significantly dependent on CD28 and CD80/CD86. Moreover, flow cytometric analysis of this CD11b+ myeloid compartment suggests that they are similar to unactivated tissue resident DC, although somewhat more mature based on their expression of CD80, CD86 and CD83. In this context it is interesting to note that the anti-myeloma agent lenalidomide, which is thought to target (currently unknown) components of the BM microenvironment, has been shown to disrupts myeloid differentiation by downregulating PU.1 in early myeloid precursors (54).
However, the observation that blocking CD80/CD86 did not have the same pro-death effect in these MM-DC co-cultures as blocking CD28, at least under serum starvation conditions, suggested that CD80/CD86 were not simply acting as ligands for CD28, and that their continuing ligation by CTLA4-Ig/CD28-Ig was inducing a separate biological response. The most likely response is the induction of DC IL-6 production, which had been originally identified in the context of DC-mediated T cell activation (55) and has not been examined in the B cell lineage. We found that MM cells can also elicit DC IL-6 production in a CD28–CD80/CD86 dependent fashion, which we believe uncovers a major mechanism by which this cytokine is elicited from the microenvironment in this disease. Unexpectedly, this induction appears to involve previously unrecognized cross-talk with the Notch1 signaling pathway. Prior studies have demonstrated that Notch receptor signaling on BMSC (which were not defined) elicits their production of IL-6, is important for the survival of MM cells that express the Notch ligand Jagged 2 (15), and blockade of Notch signaling inhibits the ability of BMSC to protect MM cells against chemotherapy-induced apoptosis (43). We find that when the BMSC are DC, inhibition of either the CD80/CD86 or Notch pathways abrogated DC IL-6 production to a greater extent than can be accounted for if these pathways were independent. The same crosstalk is seen when ligating CD80/CD86 with soluble CD28-Ig alone instead of with myeloma cells (CK, data not shown), indicating that these two pathways may in fact be a single integrated one.
The finding that CD28 on myeloma cells ligates CD80/CD86 to induce DC IL-6 production raised the possibility that the same ligation also induces DC production of IDO, which was also initially defined in context of DC-mediated T cell anergy and generation of Treg (30, 33, 47, 56) and has also not been examined in the B lineage. Consistent with this, CD28+ myeloma cell lines or primary myeloma isolates induced DC IDO production in a CD28 and IFNγ dependent manner. The in vivo source of IFNγ is unclear, but it is interesting to note that pDC are a component of the pro-survival myeloma microenvironment (28), and separately that pDC are thought to be the primary source of IFN in the context of IDO-mediated T cell tolerance (57). Alternatively, activated T cells that infiltrate myelomatous regions of the bone marrow (58) may also be a source of IFNγ. Our in vitro findings demonstrate that IDO production from DC can generate both Treg (Fig. 6F) as well as directly inhibit T cell activation (although the latter only occurs at the higher levels of IDO induced when DC are cocultured with myeloma cells) (Fig. 6G), which point to specific molecular mechanisms underlying the characteristic immunologically suppressed state seen in myeloma patients. Supportive of this, earlier clinical studies demonstrated increased kynurenine metabolites and lower tryptophan levels correlated with increased IFN activity (34), as well as dysfunctional T cell responses with increased frequencies of regulatory T cells in myeloma patients (59) - all pointing to an immunosuppressive role for myeloma-induced IDO that could be relieved by small molecule IDO inhibitors.
In summary, we propose that CD28 expressed by myeloma cells serves as a central molecular bridge within a complex and integrated cellular and soluble factor microenvironment necessary for MM cell survival. CD28 directly delivers a pro-survival signal to the myeloma cell, and by ligating CD80/CD86 on myeloid DC backsignals to these stromal cells to elicit both the supportive cytokine IL-6 (in conjunction with Notch1 signaling) as well as the immunosuppressive enzyme IDO. Other integral cellular components include pDC, which are induced by myeloma cells to produce APRIL/BAFF (which is complementary to IL-6 in supporting PC survival and function) and activated T cells, either of which could be the primary source of interferon required for the induction of myeloid DC IDO production. Although undoubtedly incomplete, this model begins to point to novel therapeutic targets for the treatment of multiple myeloma. In addition, we have found the interactions described here for myeloma cells very closely mirror the interactions that support the survival of normal long-lived plasma cells in the bone marrow. Thus, therapeutic inhibitors of CD28 originally developed to block T cell activation (e.g. Abatacept (CTLA4-Ig), which is FDA approved for the treatment of rheumatoid arthritis) may be effective in multiple myeloma as well as against pathogenic PC in allergic and autoimmune diseases.
Thanks are due to Dr. Kunle Odunsi, RPCI for providing us the HEK293 cell lines. LHB is a Georgia Cancer Coalition Distinguished Scholar. .
1This study was supported in part by NIH grants CA121044, CA121044, CA085183 and the Multiple Myeloma Research Foundation.
Disclosures: JPS is the founder and owns stock in Signal Genetics, LLC. The other authors declare no conflict of interest.