Durvalumab Combined with Immunomodulatory Drugs (IMiD) Overcomes Suppression of Antitumor Responses due to IMiD-induced PD-L1 Upregulation on Myeloma Cells

The programmed cell death 1 (PD-1; also known as CD279)–programmed death-ligand 1 (PD-L1; also known as CD274 and B7-H1) pathway (PD-1/PD-L1 signaling pathway) plays a key role in immune suppression in the tumor microenvironment (1–4). PD-L1 expressed on various tumor cells interacts with PD-1 on tumor-infiltrating T cells, inhibits functional T-cell activation through the inhibitory signal of T-cell receptor signaling, and reduces adaptive immune responses (5–7). Its blockade promotes endogenous adaptive anti-immunity, and PD-1/PD-L1–targeted therapy has shown clinical success in several types of cancers, including melanoma, non–small cell lung cancer, renal cell carcinoma, Hodgkin lymphoma, and microsatellite instability–high cancers (8, 9).

Multiple myeloma (MM) is an incurable hematologic malignancy characterized by expansion of abnormal plasma cells (PCs) caused by multistep genetic and bone marrow (BM) microenvironmental changes (10, 11). PD-L1 expression levels on PCs from patients with MM are markedly upregulated compared with cells from monoclonal gammopathy of undetermined significance (MGUS) patients and healthy volunteers (12, 13). On the other hand, high levels of PD-1 are expressed on CD4⁺ and CD8⁺ T-cell subsets from the peripheral blood (PB) and BM of patients with MM (14, 15). The PD-1/PD-L1 interaction in MM attenuates the production of MM-specific cytotoxic T lymphocytes (12). Furthermore, our previous study showed that PD-L1 on MM cells is associated with aggressive myeloma behaviors including cell proliferation capacity and drug resistance, and the interaction of PD-L1 with PD-1 molecules induces drug resistance in MM cells by antiapoptotic responses through the Akt signaling pathway (12, 16). Thus, PD-L1 on MM cells is associated with not only immune dysfunction in the BM microenvironment but also disease progression in patients with MM.

Recently, the availability of immunomodulatory drugs (IMiDs), such as thalidomide (THAL), lenalidomide (LEN), and pomalidomide (POM), has dramatically prolonged the overall survival rates of MM patients. IMiDs directly bind to cereblon (CRBN), which is part of an E3–ubiquitin ligase complex, and lead to the ubiquitination and subsequent degradation of Ikaros (encoded by Ikaros family zinc finger protein 1; IKZF1) and Aiolos (encoded by IKZF3; refs. 17–20). Ikaros and Aiolos are the key regulators of both transcriptional factors and repressors in hematopoietic cells, and their degradation mediated by IMiDs leads to anti-MM effects through downregulation of interferon regulatory factor 4 (IRF4) and c-Myc (19, 21). However, many patients with MM develop IMiD resistance over time, necessitating long-term survival to overcome it. Hence, it is crucial to elucidate the mechanism of this resistance.

In this study, we investigated the regulation of PD-L1 expression by IMiDs and antimyeloma effects of the anti–PD-L1 antibody durvalumab in combination with IMiDs.

Human MM cell lines were cultured in RPMI-1640 medium (Wako Chemical Industries) containing 10% fetal bovine serum, 100 U/mL of penicillin, and 100 mg/mL of streptomycin (Thermo Fisher Scientific) at 37°C under 5% CO₂. KMS-18, KMS-27, and KMS-28PE cell lines were kindly provided by Dr Takemi Otsuki in 2004 (Kawasaki Medical School, Okayama, Japan). U266, RPIM 8226 (obtained in 2005), and MM.1S cells (obtained in 2010) were obtained from the American Type Culture Collection, and OPM2 cells were purchased from DSMZ in 2010. The human MM cell line MOSTI-1 was established using BM samples from LEN/bortezomib double-refractory MM patients in our laboratory (obtained in 2012; ref. 16). Cell characteristics and phenotypes of these MM cell lines are shown in Supplementary Table S1. PD-L1–transfected KMS28-PE (PD-L1.KMS-28PE) cells were established, as described previously (12). Mycoplasma was tested and was negative in all cells using the e-Myco Mycoplasma PCR Detection Kit (iNtRON Biotechnology). These cell lines were not examined by short tandem repeat analysis. The cells were cultured to passages 3 to 5, frozen in the cell-freezing medium TC-protector (DS Pharma Biomedical), and used within 20 passages in all experiments.

BM samples were obtained from patients with newly diagnosed MM and MGUS for diagnostic purposes at Nippon Medical School after informed written consent was obtained according to the Institutional Review Board–approved protocol (Supplementary Table S2). The diagnoses were made according to International Myeloma Working Group criteria (22). PB samples were obtained from healthy adult volunteers. Mononuclear cells (MNCs) were separated from BM and PB samples with Histopaque (Sigma-Aldrich) density centrifugation.

Cells were suspended in 50 μL of fluorescence-activated cell sorter (FACS) buffer (PBS with 5% bovine serum albumin and 10 mmol/L sodium azide) and blocked with Human TruStain FcX (BioLegend). Then, the cells were stained with flow cytometry (FCM) antibodies for 30 minutes. Data acquisition was performed in an LSRFortess X-20 flow cytometer with FACSDiva software version 8.0.1 and analyzed using FlowJo software (BD Biosciences). PCs were identified as CD138-positive and CD38-strong positive cells. Monocytes and lymphocytes were gated using the CD45-gating method, and the gated cells were confirmed by CD14 and CD3 positivity, respectively.

Total RNA was extracted from CD138⁺ PCs isolated from BM MNCs of MM patients and MM cell lines. The cDNA from total RNA was synthesized, and then quantification of mRNA using real-time PCR was performed (12, 16). Primer sequences are shown in Supplementary Table S3.

Cells were lysed using RIPA lysis buffer (Merck Millipore) plus protease inhibitor and phosphatase inhibitor (Roche Diagnostics). Twenty micrograms of protein was electrophoresed on an SDS–12.5% polyacrylamide gel and then transferred to a PVDF membrane. The target protein was detected using the ECL Prime Western Blotting Detection System (GE Healthcare) combined with LuminoGraph I (ATTO Corporation). The chemiluminescence of targeted protein was quantified using ImageJ version 1.50e software (NIH).

MM cells were exposed to LEN and POM for 3 days and then the cells were stained with annexin-FITC (BioLegend) and propidium iodide (PI; Wako Chemical Industries) for the analysis using FCM (23). For assessing drug sensitivity to LEN and POM, 2 × 10⁴ MM cells in 100 μL of complete medium were seeded in 96-well plates and then treated with LEN or POM (0–500 μmol/L) for 3 days. Following exposure to IMiDs, cell viability was determined in the 2-(2-methoxy-4-nitrophenyl)-3-(4-nitrophenyl)-5-(2,4-disulfophenyl)-2H-tetrazolium monosodium salt (WST-8) assay using the Cell Counting Kit-8 (CCK-8; Dojindo; ref. 16). The IC₅₀ is the concentration of drug that inhibits cell proliferation by 50% of that of the untreated controls.

We used the lentiviral short-hairpin RNA (shRNA) expression vector pRSI9-U6-(sh)-UbiC-TagRFP-2A-Puro vector (Addgene) for knockdown experiments and the lentiviral expression vector pLV-EF1a-IRES-Hygro (Addgene) for overexpression experiments. The targeted sequences of the shRNAs are shown in Supplementary Table S3. Ikaros cDNA fragments (isoform 1) were amplified from human MM cell line–derived cDNA using the primer shown in Supplementary Table S3. shRNA DNA fragments and Ikaros cDNA were inserted into each vector. These plasmids and ViraPower Lentiviral packaging mix (Thermo Fisher Scientific) were cotransfected into 293FT cells. The supernatant containing lentivirus 2 days posttransfection was harvested, and subsequently the MM cells were transduced with lentiviral supernatant. To establish stable cell lines, the cells were cultured with complete medium containing 1 μg/mL of puromycin (Sigma-Aldrich) or 200 μg/mL of hygromycin (Wako Chemical Industries) 48 hours after transduction. The empty vectors were used as the controls.

Seven-week-old NOD/Shi-scid, IL2RγKOJic (NOG) female mice (CLEA Japan Inc.) were used for the in vivo assays. NOG mice were inoculated subcutaneously in the right flank with 2 × 10⁶ MM.1S cells in 100 μL of PBS. Tumor sizes were measured with calipers every 3 to 4 days, and the volumes were calculated using the formula α² × β × 0.5, where α is the shortest diameter, and β is the diameter perpendicular to α. After the tumor volume reached 1,000 mm³, mice were administered 25 mg/kg of LEN orally daily for 12 days. The animal study protocols were approved by the Institutional Animal Ethical Committee of Nippon Medical School.

The Student t and Mann–Whitney U tests were used to analyze the data. P values of less than 0.05 were considered to represent statistically significant differences. Statistical analyses were performed using the GraphPad Prism 8 (GraphPad Software) and SPSS version 23 software (SPSS. Inc.).

Reagents and other conventional techniques are described in the Supplementary Information.

We first assessed the impact of IMiDs on cell-surface PD-L1 expression in MM cells after their coculture. Of the IMiDs, LEN and POM, but not THAL, enhanced cell-surface expression of PD-L1 in MOSTI-1 cells in a concentration-dependent manner (Fig. 1A). Its induction by LEN and POM in MOSTI-1 cells peaked at 3–4 days (Fig. 1B). In plasma from patients who received 25 mg LEN or 4 mg POM, the maximum concentration (C_max) was approximately 2 and 0.2 μmol/L, respectively, and the C_max of POM was reported to be 10-fold higher than that of LEN (24). Thus, we set the concentration of LEN and POM at 10 and 1 μmol/L, respectively, and the treatment period at 3 days for the in vitro assay. In addition to MOSTI-1 cells, the mRNA expression of PD-L1 in U266, MM1.S, KMS-18, and KMS-27 cells was also upregulated by POM (Fig. 1C). The cytotoxicity of these MM cell lines was hardly affected by POM, except for RPMI 8226 cells (Fig. 1D). POM strongly promoted the apoptosis of RPMI 8226 cells, although their PD-L1 expression was not increased by POM (Fig. 1C and D). MOSTI-1, U266, MM1.S, KMS-18, and KMS-27 cell lines showed lower sensitivity to LEN and POM than RPMI 8226 cells, and the IC₅₀ values for POM were 322, 252, 250, 217, 224, and 39 μmol/L, respectively (Supplementary Fig. S1). On the other hand, the 50% inhibitory activity of LEN was not reached in MOSTI-1, U266, MM.1S, KMS-18, and KMS-27 cell lines (Supplementary Fig. S1). The PD-L1 induction levels in these MM cell lines were significantly correlated with the IC₅₀ values for POM (r = 0.813, P = 0.0491; Fig. 1E). Furthermore, CRBN expression in MOSTI-1, U266, KMS-18, and KMS-27 cells was lower than in MM.1S and RPMI 8226 cells, and IC₅₀ values for POM tended to be negatively correlated with CRBN expression levels (r = −0.887, P = 0.0147; Supplementary Fig. S2). Although PD-L1 induction levels by POM in MOSTI-1 cells were the same as in the other MM cell lines, the level of PD-L1 expression in MOSTI-1 was markedly higher than in other MM cell lines because PD-L1 gene amplification caused cell-surface overexpression in MOSTI-1 cells (Fig. 1F; Supplementary Fig. S3). Moreover, the expression of PD-L2, which is located in close proximity to the PD-L1 gene locus, was scarcely induced by LEN and POM (Supplementary Fig. S4).

We next investigated whether PD-L1 expression on primary PCs from five patients newly diagnosed with MM who had not received chemotherapy, was also induced by LEN and POM. The percentage of CD138⁺CD38⁺ PCs in BM MNCs from MM patients was 24.34% ± 14.8% [mean ± standard deviation (SD); range, 10.8%–49.2%] in this study (Supplementary Table S2). CD138⁺CD38⁺ PCs and monocytes expressed high levels of PD-L1, but lymphocytes did not (Fig. 1G). Its expression on PCs, but not on monocytes and lymphocytes, obtained from patients with MM was significantly enhanced by LEN and POM (Fig. 1G and H). On the other hand, PD-L1 expression on PCs from patient with MGUS was low and not upregulated by LEN or POM (Fig. 1G).

Next, to clarify the mechanism of PD-L1 induction by LEN and POM, we performed gene and microRNA (miRNA) expression profiling analysis in the MM cell lines MOSTI-1 and U266, with and without POM coculture. In gene-expression microarray analysis, a total of 404 genes with differential expression showing more than a twofold difference after coculture with POM in both MOSTI-1 and U266 cells (366 upregulated and 38 downregulated genes) were identified, as described in Supplementary Table S4. The 366 upregulated genes were defined in seven categories by gene ontology (GO) molecular function enrichment analysis (Fig. 2A). Because both U266 and MOSTI-1 cells were HLA-DR–positive cell lines, MHC class II–antigen-presentation pathway-related genes were upregulated by POM treatment (Fig. 2A). Furthermore, the category of receptor ligand activity contained 17 genes, among which we focused on APRIL [also known as tumor necrosis factor ligand superfamily member 13 (TNFSF13)], which is related to MM tumor growth (Fig. 2B). On the other hand, we could not identify the miRNA associated with PD-L1 upregulation by LEN and POM by human miRNA array.

APRIL is a protein of the tumor necrosis factor (TNF) superfamily recognized by the cell-surface receptor transmembrane activator and CAML interactor (TACI) and B-cell maturation antigen (BCMA; ref. 25). The gene expression of APRIL showed 2.3-fold and 5.2-fold changes with POM treatment in U266 and MOSTI-1 cells, respectively (Supplementary Table S4). We confirmed that the expression of mRNA was increased in MOSTI-1, U266, and MM1.S cells cocultured with POM, but not increased in KMS-18, KMS-27, and RPMI 8226 cells (Fig. 2C). The cell-surface expression of BCMA and TACI was not changed after POM coculture (Supplementary Fig. S5A and S5B). When MM cell lines and primary PCs from newly diagnosed MM patients were cultured with recombinant human APRIL for 3 days, PD-L1 was induced by APRIL as well as by LEN and POM (Fig. 2D–F). Its induction on MOSTI-1 cells was suppressed by both TACI-Ig and BCMA-Ig (Fig. 2G), by the antibody against BCMA but not that against TACI (Fig. 2H) and, furthermore, by the MEK/ERK inhibitor U0126 (Fig. 2I). Gene set enrichment analysis (GSEA) showed that the MAPK/ERK signal cascade was significantly upregulated in MM cells treated with POM (Supplementary Fig. S6).

To confirm the in vitro assay data, we treated MM.1S xenograft‐bearing NOG mice with 25 mg/kg LEN orally daily for 12 days (Fig. 3A). The cell-surface PD-L1 in MM.1S cells was increased in a concentration-dependent manner by POM in vitro (Fig. 1C). The mRNA and cell-surface PD-L1 expression levels in MM.1S cells engrafted in NOG mice were significantly upregulated by LEN treatment in comparison with controls (Fig. 3B and C). Similarly, the APRIL expression level in LEN-treated mice was markedly higher than that in controls (Fig. 3D and E). Next, while receiving LEN, MM.1S xenograft-bearing NOG mice were administered 100 μg anti-BCMA antibody intraperitoneally every 4 days (Fig. 3F). The upregulation of PD-L1 in MM.1S cells in NOG mice treated with LEN was suppressed by anti-BCMA antibody (Fig. 3G). However, the tumor size did not differ among the three treatment groups (Supplementary Fig. S7). The results indicated that the PD-L1 induction in MM cells by LEN treatment might be associated with the APRIL–BCMA pathway.

In some MM cell lines, APRIL was induced by POM, while in others it was not (Fig. 2C). Furthermore, PD-L1 induction by POM was not completely suppressed by the APRIL inhibitors and antibody against the APRIL receptors (Fig. 2G and H). Thus, we attempted to confirm whether PD-L1 induction by LEN/POM was directly associated with the degradation of Ikaros/Aiolos meditated via cereblon (CRBN; Fig. 4A). We constructed two pGL3-promoter vectors carrying the PD-L1 promoter regions, −353 to +1 and −1,013 to +1, and investigated their activity using a luciferase reporter assay in APRIL- and BCMA-negative HeLa cells. The luciferase activity of the −353-bp promoter induced was 2.2 ± 0.6- and 3.5 ± 0.7-fold higher following LEN and POM treatment, respectively, compared with that in untreated controls (Fig. 4B). Similarly, the highly inducible activity of the −1,013-bp promoter was conserved in response to LEN (2.6 ± 0.4-fold change) and POM (3.3 ± 0.5-fold change; Fig. 4B). Next, to clarify the participation of CRBN, Ikaros, and Aiolos, we performed the knockdown of these proteins in MM cells. Upregulation of PD-L1 by LEN and POM was not seen in stable CRBN knockdown MOSTI-1 cells (Fig. 4C). On the other hand, the knockdown of Ikaros increased expression levels of PD-L1 protein and mRNA in MOSTI-1 cells, and PD-L1 upregulation was markedly enhanced by LEN and POM (Fig. 4D). PD-L1 expression was conversely downregulated in Ikaros-overexpressing MOSTI-1 cells (Fig. 4E). However, the expression of PD-L1 protein in Aiolos-knockdown cells after culture with LEN or POM was the same as in controls, although the induction levels of mRNA expression in knockdown cells were lower than in controls (Fig. 4F). In other MM cell lines, CRBN-knockout MM cells decreased PD-L1 induction, whereas Ikaros-knockdown MM cells increased PD-L1 expression (Supplementary Fig. S8). Analysis of Ikaros ChIPseq data from the human GM12878 lymphoblastoid cell line demonstrated one Ikaros-binding peak in the PD-L1 promoter region, suggesting that Ikaros might directly bind to the upstream promoter region (−1 to −353; Supplementary Fig. S9A). To confirm this, we constructed a pGL3-promoter vector containing a nonbinding mutation at the Ikaros-binding site of the −353-bp promoter region for the luciferase reporter assay. The promoter activity of the mutant type was upregulated in comparison with the wild-type in the absence of LEN and POM (Fig. 4G). Therefore, Ikaros might function as a transcriptional repressor of the PD-L1 gene, and PD-L1 might be directly induced by LEN and POM via CRBN-mediated Ikaros degradation as well as via the APRIL–BCMA pathway.

When untreated or POM-pretreated U266 sells were cocultured with T cells, we ascertained the impact of each cell type on T-cell activity. POM-pretreated U266 cells significantly suppressed T-cell proliferation and IFNγ production from T cells as compared with untreated U266 cells at the effector:target (E:T) ratios of 1:0.25 and 1:0.5 (Fig. 5A and B). Therefore, PD-L1 upregulation on MM cells robustly induced T-cell suppression. However, untreated U266 cells also showed strongly suppressed T-cell activation identical to POM-treated U266 cells at the E:T ratio of 1:1, and therefore the difference in immune suppression between the two cell types was difficult to determine. In this study, it appeared that T-cell suppression may be associated with cell–cell contact signaling such as the PD-1–PD-L1 pathway but not with other mechanisms because immunosuppressive soluble factors were not generated from irradiated MM cells.

We next questioned whether PD-L1 upregulation on MM cells by LEN and POM could inhibit T-cell immune responses and whether durvalumab (MED14736), a human immunoglobulin G1 (IgG1) kappa monoclonal antibody that binds to PD-L1 (26), could induce T-cell activation in combination with LEN or POM. To resolve these questions, we examined T-cell activation after cocultivating POM-pretreated MM cells in the presence of durvalumab and POM (Fig. 5C). When T cells were cocultured with POM-pretreated U266 cells, the cell proliferation of CD8⁺ T cells was inhibited, whereas this T-cell suppression was not prevented by durvalumab alone (Fig. 5D). However, its suppression was recovered with the combination of durvalumab and POM during cocultivation of POM-treated U266 cells with T cells (Fig. 5D). Furthermore, the IFNγ production from T cells and population of both CD25⁺CD45RO⁺ double-positive CD8⁺ T cells (activated T cells) cocultured with POM-treated U266 cells were significantly increased after combined POM and durvalumab treatment (Fig. 5E and F). Similarly, the inhibition of CD8⁺ T-cell proliferation by cocultivation with LEN-treated U266 cells was disarmed by the combination of durvalumab with LEN (Supplementary Fig. S10).

Previously, we demonstrated that the binding of PD-1 with PD-L1 can induce drug resistance through Akt activation in MM cells (16). Thus, we investigated whether the ligation of PD-1-Fc with PD-L1 upregulated by IMiDs could increase phosphorylation of Akt in MM cells. The amount of PD-1-Fc bound was increased by POM, although the binding was completely blocked by durvalumab (Fig. 6A). The binding of PD-1-Fc to PD-L1 on MM cells decreased the percentage of apoptotic cells induced by bortezomib, similar to our previous findings, although durvalumab overcame the drug resistance induced by PD-1−PD-L1 signaling (Fig. 6B). Expression levels of Akt phosphorylation did not increase with PD-L1 upregulation (Fig. 6C). Consistent with those results, the number of bortezomib-induced apoptotic cells did not decrease when POM-treated MM cells, which had increased PD-L1 expression, were cultured with PD-1-Fc. However, the POM/bortezomib combination increased the number of apoptotic cells in the presence of durvalumab through the blockade of PD-1−PD-L1 signaling (Fig. 6D).

PD-L1 on MM cells inhibits the effector function of immune cells such as T and NK cells in patients with MM (1, 3). In the present study, we showed that PD-L1 expression on IMiD-insensitive MM cells and PCs from newly diagnosed MM patients is upregulated by LEN and POM via ERK activation induced by the APRIL–BCMA pathway and Ikaros degradation through CRBN (Fig. 6E). Furthermore, T-cell suppression in MM cells after cultivation with LEN/POM and drug resistance via PD-L1–PD-1 interaction is disarmed by durvalumab. Our findings suggest that PD-L1 upregulation by IMiDs might induce aggressive myeloma behaviors and tumoral immune escape in MM patients.

Several researchers reported that PD-L1 in the MM cell line RPMI 8226 and in CD138⁺ PCs in the BM from relapsed/refractory MM patients was downregulated after cultivation with LEN (27, 28). However, our results showed that LEN and POM upregulated PD-L1 on MOSTI-1, U266, MM1.S, KMS-18, and KMS-27 cells but not on RPMI 8226 cells. Iwasa and colleagues also reported that PD-L1 on the MM cell lines MM.1S, KMS-11, and OPM-2 was increased by LEN and POM (29). There was no difference in cell characteristics and phenotypes among these MM cell lines (Supplementary Table S1). In particular, the MOSTI-1 cell line was established from BM PCs with LEN/bortezomib double-refractory MM patients. Its IC₅₀ value of IMiDs and the rate of PD-L1 induction by IMiDs were higher than those in other MM cell lines. Furthermore, the IC₅₀ values of POM were negatively correlated with CRBN expression. Lopez-Girona and colleagues demonstrated that the proliferation of RPMI 8226 cells was strongly inhibited by LEN in comparison with other MM cell lines including U266 cells (30). They further reported that CRBN knockdown in U266 cells induced resistance to the antiproliferative effects of LEN (31). CRBN reduction decreased the antiproliferative potency of IMiDs mediated via induction of the cell-cycle progression inhibitor p21 and the suppression of IRF4, a gene critical for myeloma cell growth (31). Therefore, because CRBN expression was higher in RPMI 8226 cells than in other MM cell lines, RPMI 8226 cells might be IMiD sensitive. Thus, it is likely that the phenomenon of PD-L1 upregulation by IMiDs may be observed in IMiD-insensitive MM cells because it is difficult to induce apoptosis via IMiDs in these MM cell lines except for RPMI 8226. On the other hand, PD-L1 expression on PCs from newly diagnosed MM patients was increased by LEN and POM. Two of five patients with MM received LEN-based therapy and were refractory to Rd (LEN and dexamethasone; Supplementary Table S2). In the future, we need to analyze the relationship of PD-L1 upregulation by IMiDs with disease progression in patients with refractory MM using a large sample size.

One mechanism of its induction by IMiDs was associated with ERK signaling transmitted through the BCMA–APRIL pathway. A recent report has demonstrated that BCMA-overexpressing RPMI8226 cells implanted into severe combined immunodeficiency mice enhance not only MM tumor growth but also PD-L1 expression, and PD-L1 upregulation by recombinant APRIL on MM cell lines was suppressed by antagonistic anti-APRIL monoclonal antibody in vitro (32). Furthermore, APRIL, secreted by osteoclasts and myeloid cells, increased PD-L1 expression on MM cell lines via the BCMA receptor in the MEK/ERK pathway (33). However, our data showed that PD-L1 expression on MM cells was induced by autocrine or paracrine secretion of APRIL enhanced by LEN/POM from MM cells. Although PD-L1 enhancement by IMiDs in the in vitro and in vivo assays was inhibited by anti-BCMA antibody, APRIL was not detected in cell culture supernatants of IMiD-treated MM cells and serum of MM.1S xenograft‐bearing NOG mice treated with IMiDs. The reason for this may be that the APRIL concentration in the samples was too low or because upon release it immediately binds to BCMA on MM cells in the tumor microenvironment. The promoter activity of the APRIL gene is regulated by specificity protein 1 (Sp1) and nuclear factor kappa B (NF-κB; ref. 34). In GSEA, the NF-κB signaling pathway-related gene sets were not enriched in POM-treated MM cells. Furthermore, APRIL expression was increased in CRBN-knockdown cells, but was not upregulated by IMiDs in CRBN-, Ikaros-, and Aiolos-knockdown cells (Supplementary Fig. S11). These results suggest that APRIL induction by IMiDs might not be associated with CRBN, Ikaros, and Aiolos degradation. In this study, the APRIL gene was only upregulated by IMiDs in some MM cell lines, and therefore the APRIL–BCMA pathway may be involved in some subsets of MM cells (Fig. 2C). Thus, the mechanism of APRIL induction by IMiDs remains unclear. The BCMA–APRIL signaling pathway has been shown to promote cell growth, survival, and immunosuppression (25). Similarly, PD-L1–positive MM cells have a greater proliferative capacity and drug resistance in comparison with negative cells (12, 16). Therefore, BCMA/APRIL blockade might be effective in suppressing not only PD-L1 expression but also antimyeloma activity.

In addition to the APRIL–BCMA pathway, PD-L1 was upregulated directly by the CRBN-dependent degradation of Ikaros but not of Aiolos. PD-L1 expression was strongly induced in Ikaros-knockdown cells by IMiDs, whereas it was downregulated by Ikaros overexpression. These results suggest that Ikaros acts as a suppressor of PD-L1 gene expression. The upstream promoter region of PD-L1 (−1 to −353 bp) contains one Ikaros-binding element (AGGAA), and its site is included in the sequence of the NF-κB binding site (Supplementary Fig. S9B). PD-L1 expression is upregulated by the MyD88 and TRAF6 adaptor proteins stimulated by Toll-like receptor ligands, and its signaling activates transcranial factor NF-κB (13, 35). Ikaros may inhibit the binding of NF-κB to the promoter of PD-L1, and therefore the degradation of Ikaros mediated by IMiDs disarms its PD-L1 suppression. Our data suggest that PD-L1 upregulation by IMiDs on IMiD-insensitive MM cells might promote immune escape in the BM microenvironment even if IMiDs augment both the adaptive and innate immune systems.

PD-1/PD-L1 blockade alone dramatically improves the status of some patients with non–small cell lung, prostate and renal cell cancer, and melanoma (36, 37). Anti–PD-1 antibodies such as nivolumab and pembrolizumab bind to PD-1–expressing T and natural killer cells and reverse their dysfunction (38). On the other hand, anti–PD-L1 antibodies including durvalumab, avelumab, and atezolizumab bind to several types of tumor cells, dendritic cells, and myeloid-derived suppressor cells expressing PD-L1 and reverse the inhibition of T-cell proliferation (38). De Sousa Linhares and colleagues reported that the functional and binding assay results after therapeutic PD-L1 antibody administration are superior to those with PD-1 antibodies (39). In clinical trials of anti–PD-1 inhibitors in MM, although almost no MM patients showed a response to treatment with nivolumab alone (NCT01592370; ref. 40), the combination of pembrolizumab with IMiDs and dexamethasone yielded a more than 50% objective response rate including complete remission, very good partial response, and partial response in patients with relapsed/refractory MM (NCT02036502 and NCT02289222; refs. 41, 42). Clinical trials of durvalumab in combination with POM and dexamethasone in relapsed/refractory MM are in progress (NCT02616640 and NCT02807454; ref. 43). In our study, T-cell suppression induced by PD-L1 upregulation was recovered after coculture with durvalumab combined with IMiDs but not after that with durvalumab alone. LEN enhances the PD-1/PD-L1 blockade-induced anti-MM immune response in comparison with blockade alone (27). These data suggest that IMiDs may be necessary for the functional recovery of tumoral T-cell activation because their dysfunction is induced by not only the PD-1/PD-L1 pathway but also by other mechanisms.

Our study demonstrated that IMiDs enhance PD-L1 expression on MM cells as well as APRIL production, and one mechanism may involve the APRIL–BCMA interaction (Fig. 6E). Another mechanism may be directly involved in PD-L1 upregulation by Ikaros degradation through CRBN, resulting in AKT signaling via reverse PD-1−PD-L1 signaling (Fig. 6E). Interestingly, these results suggested that the induction of PD-L1 expression by IMiDs might be associated with disease progression in IMiD-insensitive MM patients with the acquisition of aggressive myeloma behaviors and tumor immune evasion. However, because the number of patients with MM was small in this study, in the future, we must investigate PD-L1 expression on PCs from patients with MM before and after IMiD-based treatment, or PD-L1 induction in responders and nonresponders to IMiDs. Furthermore, T-cell inhibition induced by PD-L1–upregulated MM cells was recovered after combination treatment with durvalumab and IMiDs, suggesting that this combination therapy may be a reasonable treatment for patients with relapsed/refractory MM. Our findings provide new insight into the mechanism of disease progression in patients with IMiD-insensitive MM.

J. Yamamoto reports affiliation with an endowed course from Celgene Corporation and personal fees (honoraria) from Celgene Corporation outside the submitted work. H. Handa reports receipt of an international joint research grant from Celgene/BMS, USA. H. Tamura reports grants and personal fees from Celgene during the conduct of the study; grants and personal fees from Ono Pharmaceutical Co., Ltd.; personal fees from Janssen Pharmaceutical K.K.; Takeda Pharmaceutical Co., Ltd.; and Sanofi outside the submitted work. No disclosures were reported by the other authors.

M. Ishibashi: Conceptualization, data curation, software, formal analysis, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing, analysis and interpretation of data. J. Yamamoto: Resources, methodology, analysis and interpretation of data. T. Ito: Resources, methodology, analysis and interpretation of data. H. Handa: Resources, methodology, analysis and interpretation of data. M. Sunakawa: Resources. K. Inokuchi: Resources. R. Morita: Resources, methodology, writing–review and editing. H. Tamura: Conceptualization, resources, supervision, funding acquisition, project administration, writing–review and editing, analysis and interpretation of data.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.