Upregulation of CD38 expression on multiple myeloma cells by all-trans retinoic acid improves the efficacy of daratumumab

Daratumumab is an anti-CD38 monoclonal antibody with lytic activity against multiple myeloma (MM) cells, including ADCC (antibody-dependent cellular cytotoxicity) and CDC (complement-dependent cytotoxicity). Owing to a marked heterogeneity of response to daratumumab therapy in MM, we investigated determinants of the sensitivity of MM cells toward daratumumab-mediated ADCC and CDC. In bone marrow samples from 144 MM patients, we observed no difference in daratumumab-mediated lysis between newly diagnosed or relapsed/refractory patients. However, we discovered, next to an expected effect of effector (natural killer cells/monocytes) to target (MM cells) ratio on ADCC, a significant association between CD38 expression and daratumumab-mediated ADCC (127 patients), as well as CDC (56 patients). Similarly, experiments with isogenic MM cell lines expressing different levels of CD38 revealed that the level of CD38 expression is an important determinant of daratumumab-mediated ADCC and CDC. Importantly, all-trans retinoic acid (ATRA) increased CD38 expression levels but also reduced expression of the complement-inhibitory proteins CD55 and CD59 in both cell lines and primary MM samples. This resulted in a significant enhancement of the activity of daratumumab in vitro and in a humanized MM mouse model as well. Our results provide the preclinical rationale for further evaluation of daratumumab combined with ATRA in MM patients.


INTRODUCTION
The introduction of autologous stem cell transplantation as well as novel agents such as bortezomib and the immunomodulatory drugs (IMiDs) thalidomide and lenalidomide has significantly improved long-term outcome of multiple myeloma (MM) patients. 1 However, the increased survival is less evident in patients who present with high-risk disease. In addition, patients with lenalidomide and bortezomib-refractory MM have a very poor outcome with a median overall survival of only 9 months. 2 Altogether, this clearly demonstrates that there is a need for new treatment approaches, especially for these categories of patients. In this respect, several new antimyeloma agents hold promise, including next-generation IMiDs (pomalidomide) and proteasome inhibitors (carfilzomib), but also compounds with different mechanisms of action. 3,4 One of the most promising novel agents is the human IgG1 anti-CD38 monoclonal antibody daratumumab. 5,6 CD38 is highly and uniformly expressed on all MM cells. 7 CD38 is a type II transmembrane glycoprotein with ectoenzymatic activity involved in the catabolism of extracellular nucleotides. 7,8 Other functions ascribed to CD38 include receptor-mediated adhesion by interacting with CD31 or hyaluronic acid, regulation of migration and signaling events. [7][8][9][10] Daratumumab induces killing of MM cells mainly via the activation of potent cytotoxic immune effector functions, including antibody-dependent cellular cytotoxicity (ADCC), antibodydependent cellular phagocytosis and complement-dependent cytotoxicity (CDC). 11 Another mechanism of action is induction of apoptosis upon secondary cross-linking. 12 Anti-myeloma activity has also been demonstrated in mouse xenograft models 11 and, more importantly, in a humanized mouse model. 13,14 Daratumumab is currently being evaluated as a single agent in two clinical studies in heavily pretreated relapsed/refractory MM patients. Preliminary results show that daratumumab monotherapy is well tolerated, and that in the 16 mg/kg cohort at least a partial response can be achieved in 35% of the patients including CR in 10%. 15 Based on preclinical evidence showing potential benefit of combining daratumumab with lenalidomide, [16][17][18] another phase 1/2 study is evaluating the combination of daratumumab plus lenalidomide and dexamethasone in relapsed/refractory MM. Preliminary data show a manageable toxicity profile and high efficacy of this three-drug regimen. 19,20 Several phase 3 trials with daratumumab are underway in both relapse and frontline settings.
Treatment with daratumumab is clinically effective, but there is a marked heterogeneity of response with a fraction of MM patients that does not respond to daratumumab as a single agent. It is currently unknown which mechanisms underlie the differential therapeutic efficacy of daratumumab. As daratumumab is being evaluated as a single agent and in combination with various regimens in the clinical trial setting, 6 it is important to better understand host-and tumor-related factors that predict response. An improved understanding of mechanisms that contribute to innate or acquired resistance may result in the rational design of new daratumumab-based combinations with higher antimyeloma activity.
Here we studied the effect of CD38 expression levels on the efficacy of daratumumab to induce MM cell lysis via ADCC and CDC. We also examined other potential determinants of daratumumab sensitivity including the extent of previous therapy and the frequency of effector cells. Finally, we show that all-trans retinoic acid (ATRA) improves daratumumab-mediated ADCC and CDC against MM cells by upregulation of CD38 expression. Furthermore, ATRA reduces the expression of the complement inhibitors CD55 and CD59 on MM cells, thereby further enhancing daratumumab-mediated CDC.

Antibodies and reagents
Daratumumab was provided by Janssen Pharmaceuticals (Spring House, PA, USA). Human IgG1-b12 (Genmab, Utrecht, The Netherlands), a human mAb against an innocuous antigen (HIV-1 gp120), was used as an isotype control as described previously. 16

Generation of MM cell lines with higher CD38 expression
The luciferase (LUC)-transduced MM cell lines UM9 and L363 were transduced with the human CD38 gene to obtain CD38 expression levels comparable to primary myeloma cells. For this, the amphotropic Phoenix packaging cell line (Phoenix Ampho, a kind gift from Dr G Nolan, Stanford, CA, USA) was transfected using calcium phosphate precipitation, with the pQCXIN vector in which the gene encoding human CD38 was inserted. These cell lines are referred to as UM9-CD38 and L363-CD38.
BLI-based ADCC assay using LUC-transduced MM cell lines LUC-transduced MM cell lines were co-cultured with effector cells (freshly isolated peripheral blood mononuclear cells (PBMCs) from healthy donors) at an effector to target ratio of 25:1 in white, opaque, 96-well flat-bottom plates (Costar, Washington, DC, USA) in the presence of solvent control, IgG1-b12 control antibody or daratumumab for 4 h. The survival of LUC + -MM cells was then determined by bioluminescence imaging (BLI) 10 min after addition of the substrate luciferin (125 μg/ml; Promega, Madison, WI, USA). Lysis of MM cells was determined using the following formula: % lysis = 1 − (mean BLI signal in the presence of effector cells and daratumumab/mean BLI signal in the presence of effector cells and control antibody) × 100%.
BLI-based CDC assays using LUC-transduced MM cell lines Solvent control, IgG1-b12 control antibody or daratumumab were added to MM cell lines in medium supplemented with pooled unheated (native) human serum or pooled heat-inactivated (56°C for 30 min) human serum (10%; Sanquin, Amsterdam, The Netherlands). After a 1-h incubation at 37°C, MM cell survival was determined by BLI 10 min after addition of luciferin (125 μg/ml), and the lysis of cells was calculated using the following formula: % lysis = 1 − (mean BLI signal in the presence of native human serum/mean BLI signal in the presence of heat-inactivated serum) × 100%. In separate experiments, daratumumab was also able to trigger CDC using patients' serum (Supplementary Figure 1). Flow cytometry-based ex-vivo ADCC and CDC assays in BM-MNC Bone marrow mononuclear cells (BM-MNCs) derived from 144 MM patients, containing 2%-57% CD138 + tumor cells, but also autologous effector cells, were used in ADCC and CDC assays. ADCC results were previously reported for 21 of these patients. 17,22 Sample viability at incubation was > 98%, as assessed by using ToPro-3 (Invitrogen/Molecular Probes, Carlsbad, CA, USA). For ADCC assays, BM-MNCs were incubated in RPMI+10% fetal bovine serum with control antibody or daratumumab (0.01-10 μg/ml) in 96-well flat-bottom plates for 48 h. For CDC assays, BM-MNCs were treated with daratumumab (0.3-10 μg/ml) and 10% pooled human serum or autologous patients' serum as a source of complement for 1 h before flow cytometric analysis. The survival of primary CD138 + MM cells in the BM-MNCs was determined by flow cytometry as previously described. 16,21 In both assays, surviving MM cells were enumerated by single-platform flow cytometric analysis of CD138 + cells in the presence of Flow-Count Fluorospheres (Beckman Coulter, Woerden, The Netherlands) and ToPro-3, to determine absolute numbers of viable MM cells. The percentage of daratumumab-mediated ADCC was then calculated using the following formula: % lysis cells = 1 − (absolute number of surviving CD138 + cells in the presence of daratumumab/absolute number of surviving CD138 + cells in the presence of control antibody) × 100%. Complement-dependent lysis was calculated using the following formula: % lysis = 1 − (absolute number of surviving CD138 + cells in the presence of native human serum/absolute number of surviving CD138 + cells in the presence of heat-inactivated serum) × 100%.
In-vivo efficacy of the combination of ATRA and daratumumab against MM tumors growing in a humanized microenvironment.
Hybrid scaffolds consisting of three 2-to 3 mm biphasic calcium phosphate particles were coated in vitro with human mesenchymal stromal cells (2 × 10 5 cells/scaffold). After a week of in-vitro culture in an osteogenic medium, humanized scaffolds were implanted subcutaneously into RAG2 − / − γ c − / − mice, as described previously. 13,23 Eight weeks after implantation, mice received a sublethal irradiation dose (3 Gy, 200 kV, 4 mA) and XG1 cells were injected directly into the scaffold (1 × 10 6 cells/ scaffold). Three weeks after inoculation, when there was visible tumor growth in the scaffolds by BLI, different groups of mice were treated with (1) vehicle, (2) ATRA plus T-cell-depleted PBMC as effector cells (PBMC-T), (3) daratumumab plus PBMC-T and (4) daratumumab plus ATRA plus PBMC-T. Daratumumab (8 mg/kg) was given intraperitoneally on days 23, 30 and 37; PBMC-T (8 × 10 6 cells/mouse) were given intravenously on days 24, 31 and 38; and ATRA (10 mg/kg) was given via intraperitoneal injection on days 21-24, 28-31 and 35-38. PBMC-T were prepared by Ficoll-Hypaque density-gradient centrifugation of buffy coats and subsequent depletion of T cells by CD3 beads using the EasySep technology (Stemcell Technologies, Vancouver, BC, Canada). Tumor growth was monitored by weekly BLI measurements as described previously. 13 The investigator was not blinded to the group allocation when mice were analyzed. All animal experiments were conducted after acquiring permission from the local ethical committee for animal experimentation and were in compliance with the Dutch Animal Experimentation Act.

Statistics
Comparisons between variables were performed using two-tailed (paired) Student's t-test. The statistical differences between the different treatment groups in the mice experiments were calculated using a Mann-Whitney test. A total of four mice per treatment group was needed to demonstrate a difference in tumor growth of 47% (two-sided significance level α = 0.05 and power 1 − β = 0.80). P-values below 0.05 were considered significant.

RESULTS
Daratumumab-mediated lysis is not affected by extent of prior treatment During their treatment course, MM cells become increasingly resistant toward anti-myeloma agents. To evaluate whether clinical resistance to bortezomib and/or lenalidomide also results in resistance to daratumumab, we analyzed daratumumabmediated killing of tumor cells from different groups of MM patients. The characteristics of these 144 patients are shown in Table 1. ADCC and CDC assays with 10 μg/ml daratumumab were performed in 127 and 56 patients, respectively. Sufficient BM-MNCs for both assays were obtained from 39 patients. In these experiments, BM-MNCs, containing tumor cells as well as autologous effector cells, were treated with daratumumab, after which survival of MM cells was determined by enumeration of viable CD138 + cells by flow cytometric analysis.
Daratumumab-mediated ADCC against the MM cells was variable and ranged from − 19.9% (negative values indicate MM cell growth) to 80.6% (median 30.8%). In addition, complementdependent lysis of primary MM cells was very heterogeneous and ranged from − 18.5% to 93.6% (median 27.4%). Importantly, there were no significant differences in ADCC or CDC between patients with newly diagnosed disease or relapsed/refractory MM ( Figure 1). Moreover, in the subgroup of heavily pretreated lenalidomide-and bortezomib-refractory MM patients, the efficacy of daratumumab was similar to that observed in the newly diagnosed patients or in relapsed patients with less previous therapies. These data suggest that resistance to classic antimyeloma agents such as steroids, anthracyclines and alkylators, as well as novel agents (IMiDs and proteasome inhibitors), is not associated with decreased sensitivity to daratumumab-mediated ADCC and CDC in vitro.
Effect of CD38 expression on ADCC and CDC in MM cell lines As previous therapy does not explain the variability in daratumumab-mediated ADCC and CDC, we examined other factors that may affect the susceptibility of tumor cells toward daratumumab. We hypothesized that cell surface expression of CD38 on MM cells is associated with the extent of daratumumabmediated ADCC and CDC. We first tested this hypothesis in a controlled experimental setting in which CD38 expression level was the only variable. To this end, we generated clones of the MM cell lines UM9 (UM9-CD38) and L363 (L363-CD38) with higher levels of CD38 expression ( Figure 2a). The expression of CD138 and the complement-regulatory proteins (CD46, CD55 and CD59) was similar between the non-transduced parental cell lines and the clones (Figure 2b). Importantly, in ADCC as well as CDC assays, daratumumab-mediated lysis of CD38 transgenic clones was significantly better compared with the non-transduced parental cell lines (Figures 2c and d).
Effect of CD38 expression on ADCC and CDC in primary MM cells The impact of CD38 expression levels on daratumumab-mediated killing was further examined by using patients' samples. As expected, all MM cells expressed CD38 antigen in these patients' samples (n = 144), but there was a marked heterogeneity in intensity of CD38 expression with median fluorescence intensity ranging from 19.99 to 2642 (median 401.3). No significant difference in CD38 expression was observed between newly diagnosed and relapsed/refractory patients. Next, we divided the patients into tertiles according to CD38 expression on their MM cells. As illustrated in Figure 3a, ADCC against primary MM cells (n = 127 patients' samples) mediated by 10 μg/ml daratumumab was only 14.2% in the lowest tertile of CD38 expression, whereas it was significantly higher in the mid-tertile (33.5%) and the highest tertile (45.6%). We also evaluated the association between CDC induced by 10 μg/ml daratumumab and CD38 expression in 56 patients. In the lowest tertile of CD38 expression, daratumumab-mediated CDC was significantly worse (13.9%), when compared with the mid-tertile (35.5%) and highest-tertile (44.7%; Figure 3b).
In 25 patients, we collected enough BM-MNCs to evaluate different concentrations of daratumumab in ADCC and CDC assays. The dose-response curve of patients in the lowest tertile  P-values between the indicated groups were calculated using a Student's t-test. *P o0.05, **P o0.01, ***P o0.001, ****P o0.0001; ns, not significant.
the effector:target ratio (E:T ratio) in the BM-MNCs. The frequency of total CD3 − CD56 + NK cells in the BM-MNCs from 25 M patients ranged from 1.1% to 8.9% (median 3.8%). There was also a great variation in the frequency of MM cells in these samples (range 2%-57%, median 17.7%). Daratumumabmediated ADCC against MM cells was significantly inferior (12.0%) in the lowest tertile according to total NK cell:MM cell ratio, whereas it was 37.9% and 51.8% in the mid-tertile and highest tertile, respectively (Figure 3e). Similar results were obtained when we considered the activated fraction of NK cells (defined as CD3 − CD56 + CD16 + ) as effector cells (data not shown). In addition, a high monocyte:MM cell ratio was associated with improved ADCC (Figure 3f).

ATRA increases CD38 expression
Our experiments demonstrate that CD38 expression on MM cells is an important determinant of susceptibility to both daratumumab-mediated ADCC and CDC. Therefore, we hypothesized that an increase in CD38 expression may enhance the efficacy of daratumumab. As interaction of ATRA with nuclear retinoic acid receptors results in altered expression of target genes including induction of CD38 expression, 8   with 10 nM ATRA or solvent control for 48 h, followed by incubation with or without daratumumab in the presence of human serum as complement source for CDC (Figure 5a), or in the presence of PBMCs as effector cells in ADCC assays (Figure 5b). Pretreatment of MM cell lines with ATRA alone induced no or only a minimal increase in MM cell death, but significantly enhanced daratumumab-mediated CDC in XG1 cells and ADCC in XG1 and UM9 cells, compared with solvent control. In RPMI8226 cells, there was no significant improvement in daratumumab-mediated ADCC and CDC, which may be explained, in part, by the significantly smaller relative increase in CD38 expression levels in this cell line when compared with the other cell lines (Figure 4a). We subsequently used MM cells derived from patients, including those with lenalidomide-and bortezomib-refractory disease, to further explore the combination of ATRA and daratumumab. ATRA alone for 48 or 96 h did not significantly affect MM cell viability when compared with solvent control (non-viable cells at 48 h: 13% vs 15% and at 96 h: 12% vs 15% for ATRA and solvent control, respectively). However, also in primary MM cells, pretreatment with ATRA for 48 h resulted in a significant increase in their susceptibility to daratumumab-mediated CDC in 13 out of 16 patients and ADCC in 8 out of 11 patients. Pooled results from these patients show that ATRA improved CDC mediated by 10 μg/ml daratumumab from 16.1% to 43.9% (P o 0.0001) and ADCC from 25.1% to 39.5% (P = 0.0315) (Figures  5c and d, respectively). ADCC and CDC data, as well as CD38 levels with or without ATRA treatment from individual patients' samples, are shown in Supplementary Figures 2 and 3a, respectively. Importantly, ATRA augmented daratumumab-mediated CDC and ADCC against MM cells with low, intermediate or high levels of CD38 expression, as well as against MM cells derived from patients with double-refractory disease. Furthermore, ATRA improved the efficacy of daratumumab in MM cells, which were resistant to daratumumab as a single agent in CDC and/or ADCC assays (Supplementary Figure 2). Similar improvement in CDC was seen when autologous patients' serum was used (see Supplementary Figure 4). Altogether, this suggests that ATRA is an attractive strategy to enhance CD38 expression and to improve daratumumab activity in MM.
ATRA reduces CD55 and CD59 expression on MM cells As ATRA improved CDC to a higher extent than ADCC, we evaluated the effect of ATRA on the complement-inhibitory proteins, CD46, CD55 and CD59, which confer protection against several therapeutic antibodies. 25 Interestingly, ATRA also reduced expression levels of CD55, CD59 and CD46 in a dose-dependent way in MM cell lines (Figures 6a and b). Similarly, in primary MM cells, ATRA (10 nM, 48 h) reduced CD55 (mean reduction: 21.3%; P = 0.019) and CD59 expression levels (mean reduction: 37.5%; P = 0.0047), without significantly affecting CD46 levels ( Figure 6c). There was a significant correlation between ATRA-induced increase in CD38 expression and downregulation of CD59 (r = − 0.503; P = 0.0047), but not with reduction of CD55 expression. The CD46, CD55 and CD59 expression levels with or without ATRA treatment of the 16 patients' samples tested in the CDC assays are depicted in Supplementary Figure 3b. We also analyzed the effect of ATRA on effector cells. Preincubation of PBMCs with ATRA before ADCC assays did not result in enhancement of daratumumab-mediated ADCC against MM cell lines (Supplementary Figure 5). whereas ATRA as single agent had no effect. Furthermore, also in this humanized mouse model ATRA significantly enhanced the anti-MM effect of daratumumab (Figure 7c).

DISCUSSION
Immunotherapy with daratumumab is clinically effective, but there is a significant variability in quality of response among patients. 15 Mechanisms that influence daratumumab efficacy will most likely be multifactorial and include both host-and tumorrelated factors. In this study, we evaluated daratumumabmediated CDC and ADCC against primary MM cells, in relation to CD38 expression on the tumor cells, frequency of effector cells in the bone marrow and extent of previous treatment of the patient. To this end, we used mononuclear cells isolated from BM derived from MM patients, which contain not only MM cells but also autologous effector cells and stromal cells.
The current study shows that there is a significant positive association between CD38 expression levels on MM cells from patients and the efficacy of daratumumab to induce cell death by ADCC, as well as CDC. The importance of CD38 expression was further strengthened by the observation that an enforced increase in CD38 expression levels on UM9 and L363 cells resulted in an increase in daratumumab-mediated CDC and ADCC. This indicates that CD38 may be useful as a biomarker in daratumumab-based therapies. Several other studies have also demonstrated that efficacy of monoclonal antibodies is partly dependent on expression levels of their target. For example, preclinical studies show that rituximab-mediated killing of chronic lymphocytic leukemia and lymphoma cells is dependent on CD20 expression, [26][27][28][29] and also clinical studies show an inferior outcome in patients with weak CD20 expression, treated with rituximab-based immunochemotherapy. 30,31 Furthermore, ofatumumab-mediated CDC effects on lymphoma and CLL cells are largely dependent on CD20 expression levels. 28 The association between the levels of the target antigen and clinical outcome has also been demonstrated for alemtuzumab 32 and trastuzumab. 33,34 The variability in daratumumab-mediated ADCC and CDC was not solely explained by the differential expression of CD38. For this reason, we also evaluated CD38-independent factors influencing ADCC and/or CDC, including effector cell frequencies and extent of previous therapy. Importantly, the efficacy of daratumumab in terms of ADCC and CDC did not differ between newly diagnosed and heavily pretreated lenalidomide and bortezomib double-refractory patients. Similarly, we have previously shown in a humanized mouse model that daratumumab is highly effective in killing MM cells derived from double-refractory patients. 17,18 This suggests that mechanisms of resistance toward prior therapies, such as IMiDs and proteasome inhibitors, do not affect the susceptibility of MM cells to daratumumab-mediated ADCC and CDC. These findings are in agreement with the high efficacy of daratumumab as single agent in heavily pretreated relapsed/refractory MM patients. 15 In addition, we found that the ratios between NK cells, activated NK cells or monocytes, to MM cells are positively associated with susceptibility to ADCC. It is possible that the activity of these effector cells may be affected by preceding or concomitant therapy such as steroids and IMiDs. 16 In addition, inhibitory signals transmitted to NK cells by MM cells may result in NK cell dysfunction. 35 Modulation of determinants of daratumumab sensitivity with novel therapeutic approaches may lead to more effective daratumumab-based regimens with increased quality of response and improvement in survival. As CD38 levels may influence the efficacy of both daratumumab-mediated CDC and ADCC against MM cells, we hypothesized that upregulation of CD38 expression levels could increase MM cell kill and thereby enhance the response rate of the antibody. Indeed, we demonstrated that ATRA increased expression levels of CD38 on MM cells. This resulted in enhanced ADCC and CDC in both cell lines and patient samples, including those with low CD38 expression or complete resistance to daratumumab-mediated CDC and/or ADCC. In addition, the anti-MM activity of daratumumab was also significantly enhanced by ATRA in our recently developed humanized mouse model. Interestingly, the improvement in CDC was more pronounced than the enhancement of ADCC, suggesting that ATRA also modulates CD38-independent determinants of CDC but not of ADCC. Indeed, we did not observe enhancement of ADCC by preincubating effector cells with ATRA, but we did find a significant downregulation of the complementinhibitory proteins CD55 and CD59 with ATRA. It is possible that ATRA has additional CD38-independent activities that result in improvement of daratumumab-mediated lysis of MM cells. Importantly, the enhancing effect of ATRA was evident at a dose of 10 nM, which is a clinically achievable and safe concentration of ATRA.  (4) daratumumab plus ATRA plus PBMC-T. Shown are BLI images before (day 21; top panels), 3 weeks after the initiation of treatment (day 42; middle panels) and at the end of the experiment (day 56; lower panels). (c) Analysis of tumor load per treatment group with four mice per group and each mouse with four scaffolds. The statistical differences between mice treated with daratumumab and mice treated with daratumumab plus ATRA were calculated using the Mann-Whitney U-test. **Po 0.01, ***P o0.001; ns, not significant.
Retinoic acids can influence gene expression and protein production in different ways. 36 Previous studies have demonstrated that the retinoic acid receptor has an important role in the induction of CD38 by ATRA. 8,24,37 Indeed, the CD38 gene contains a retinoic acid-responsive element in the first intron. 38 However, also nonclassical retinoic acid signaling, independent of the conventional retinoic acid receptor pathway, has been demonstrated in CD38 upregulation. 37 This includes ATRAinduced CD38 induction via response elements in the 5′-flanking region, which is mediated by protein kinase Cδ. 37 In addition, phosphatidylinositol 3-kinase is involved in ATRA-induced upregulation of CD38 on human hematopoietic cells. 39 Similar retinoic acid receptor -dependent and -independent mechanisms may be involved in ATRA-mediated suppression of CD55 and CD59.
In conclusion, this study has identified multiple factors that influence the extent of MM cell lysis mediated by daratumumab. These factors may serve as biomarkers to predict response in daratumumab-based regimens. A better understanding of mechanisms underlying variability in sensitivity to daratumumabmediated killing may also lead to novel strategies to enhance the effectiveness of daratumumab therapy. Here we provide the rationale for clinical evaluation of ATRA and daratumumab in MM patients.