Immunomodulation by durvalumab and pomalidomide in patients with relapsed/refractory multiple myeloma

This study sought to understand how the programmed death ligand 1 (PD-L1) inhibitor durvalumab and the immunomodulatory agent pomalidomide regulate immune cell activation and function in patients with relapsed/refractory (RR) multiple myeloma (MM). Immunologic changes in peripheral blood and bone marrow of patients treated with durvalumab as monotherapy or in combination with pomalidomide with/without dexamethasone were characterized by assessing subsets of immune cells and gene signatures to understand the immunomodulatory effect of the treatment. Soluble PD-L1 levels were elevated at screening in patients with RRMM but did not correlate with response to durvalumab combination therapy. Immune cell subsets were increased in peripheral blood during treatment with durvalumab and pomalidomide, and combination therapy induced significant gene expression changes in the MM tumor microenvironment versus durvalumab alone. Estimation of cell populations based on RNA sequencing data revealed increased monocytes, neutrophils, and natural killer cells with the combination therapy, but not with durvalumab alone. Additionally, multiplex immunofluorescence of bone marrow demonstrated that immune populations were different in responders versus nonresponders to durvalumab plus pomalidomide with dexamethasone therapy. Overall, durvalumab effectively blocked soluble PD-L1; however, durvalumab monotherapy was not associated with immunologic changes, which were observed with combination therapy.

Multiple myeloma (MM) is a chronic cancer of plasma cells in the bone marrow 1 . Although there are effective therapeutic combinations, patients with MM often relapse and become refractory to standard-of-care treatments such as immunomodulatory imide (IMiD) agents, proteasome inhibitors, and anti-CD38 monoclonal antibodies (e.g., daratumumab), resulting in a worsening prognosis 2,3 . Studies have shown that in MM, there is dysregulation of the immune compartment in the bone marrow, including changes in major cell populations such as natural killer (NK) cells, CD4 + and CD8 + T cells, and dendritic cells, and increases in immune suppressive cell populations, including regulatory T cells (T reg ), tumor-associated macrophages (TAMs), and myeloid-derived suppressor cells (MDSCs) [4][5][6][7][8] . It has been demonstrated that programmed cell death protein-1 (PD-1) expression is upregulated on T cells isolated from patients with MM, suggesting that this pathway is of importance in mediating the immunosuppressive state in this patient population 9 . Furthermore, studies have shown that while programmed death ligand 1 (PD-L1) is absent from normal plasma cells, it is expressed in MM cell lines and primary MM tumor cells from patients [9][10][11][12][13] . Together, this suggests a role for immunotherapy in treatment of MM.
Monoclonal antibodies that block inhibitory receptors have shown significant clinical activity across a variety of tumor types [14][15][16][17] . Blockade of immune checkpoints, such as cytotoxic T-lymphocyte-associated antigen 4 (CTLA-4), PD-1, and PD-L1, has shown clinical activity not only in conventionally immune-responsive tumors such as melanoma and renal cell carcinoma, but also in non-small cell lung cancer (NSCLC) and prostate cancer [16][17][18] . These checkpoint inhibitors (CPIs) work to limit and potentially reverse T-cell exhaustion by blocking the inhibitory signals to create an effective immune response.
Durvalumab is a human immunoglobulin G1 kappa monoclonal antibody CPI targeted against PD-L1 that has been approved for the treatment of patients with stage III NSCLC and as first-line therapy with etoposide and either carboplatin or cisplatin for extensive-stage small cell lung cancer [19][20][21][22][23] . PD-L1 expressed on tumor Methods Study design, patients, and treatment. Samples were obtained from patients enrolled in the clinical trial MEDI4736-MM-001 (clinicaltrials.gov identifier: NCT02616640; registered 30/11/15). This was a multicenter, open-label, phase 1b study designed to determine the recommended dose and regimen of durvalumab either as monotherapy or in combination with pomalidomide with or without low-dose dexamethasone in patients with RRMM. Patients must have received at least 2 prior lines of antimyeloma therapy, including lenalidomide and a proteasome inhibitor, and demonstrated disease progression on or within 60 days of completion of the last therapy. This study was conducted in accordance with the International Council for Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use/Good Clinical Practice, the Declaration of Helsinki, and applicable regulatory requirements. The study protocol was approved by local or independent institutional review boards or ethics committees at participating sites. All patients provided written informed consent.
Patients were randomized to 1 of 3 treatment arms. In the durvalumab monotherapy arm (Arm A), patients received 1500 mg intravenous (IV) durvalumab on day 1 of each 28-day cycle. Patients randomized to Arm B received durvalumab as in Arm A plus oral pomalidomide 4 mg/day on days 1 to 21. Patients in Arm C received oral dexamethasone 40 mg/day (≤ 75 years of age) or 20 mg/day (> 75 years of age) on days 1, 8, 15, and 22 of the 28-day cycle in addition to durvalumab and pomalidomide. At the investigator's discretion, patients randomized to Arms A or B could switch to Arm C upon progression or lack of efficacy.
Responders were defined as patients with a partial response or better using International Myeloma Working Group Uniform Response Criteria 25 . The focus of this paper is on exploratory endpoints, including pharmacodynamic, mechanistic, and predictive biomarkers of durvalumab, both as a single agent and in combination with pomalidomide with or without dexamethasone.
Soluble PD-L1 (sPD-L1) analysis. Blood was collected at screening in serum separator tubes, inverted, and allowed to stand at room temperature (RT) for 60 to 120 min. Tubes were centrifuged at 1500 g for 10 min at RT, and serum was transferred to a vial and stored at − 70 °C. Serum samples for electrochemiluminescence assays were shipped frozen to the central laboratory (Laboratoires Cerba, Saint Ouen l' Aumône, France), and to the analytical laboratory (Intertek, London, UK) where the validated quantitative sPD-L1 electrochemiluminescence assay was performed. The upper and lower limits of detection were 1000 pg/mL and 15.6 pg/mL, respectively. Analyses provided screening sPD-L1 values for each evaluable patient sample and were compared with samples from healthy volunteers. The healthy serum samples were acquired through Intertek (San Diego, CA); all samples from these volunteers were consented.
Flow cytometric analyses of peripheral blood. Peripheral blood was obtained at screening, cycle 1 day 1 (C1D1; pretreatment), cycle 1 day 8 (C1D8), cycle 1 day 15 (C1D15), cycle 2 day 1 (C2D1), and cycle 2 day 15 (C2D15). Whole blood samples for immunophenotyping were collected in sodium-heparin tubes and shipped same-day at ambient temperature directly to the analytical laboratory (Quintiles: Marietta, GA or Livingston, Scotland). Cells were resuspended and stained with 6 separate panels using antibody-fluorophore conjugates (BD Biosciences, San Jose, CA) on a BD FACSCanto II (8 color, 3 laser) instrument (see Supplementary Table for a full listing of 6 panels with fluorophores and antibody clones used). Flow cytometry was performed at Quintiles Laboratories (Marietta, GA, or Livingston, Scotland), and data were analyzed using FACSDiva software (BD Biosciences). Analysis included evaluation of immune cell subsets including T cells, B cells, NK cells, MDSCs, and T reg s. Ki67 staining was used to monitor proliferation of T cells and NK cells. Additional panels provided the evaluation of suppressor cells, checkpoint molecule expression, T-cell activation, and T-cell exhaustion. Patient samples were evaluated longitudinally through C2D15, and reported data included percentage of population, absolute cell number, and mean fluorescence intensity as appropriate.
RNA sequencing of bone marrow. RNA from bone marrow samples was obtained from patients at screening and after 6 weeks of treatment (at the C2D15 visit) using tubes with either dimethylsulfoxide (DMSO) or TRIzol to stabilize the material. RNA sequencing was performed at EA Genomics (Q 2 Solutions, Morrisville, NC) using the Qiagen Micro RNeasy kit (Hilden, Germany). Quality control checks included spectrophotometric measurements and agarose gel analysis. RNA sequencing libraries were prepared using polyA enrichment www.nature.com/scientificreports/ and strand-specific library construction using barcodes. Samples were sequenced on an Illumina HiSeq 2500 with 2 × 50 bp read lengths using TruSeq SBS v4 chemistry (Illumina, San Diego, CA). RNA-sequencing count data were normalized using trimmed mean of M values (TMM) and then transformed to log 2 counts per million along with observation level weights using voomWithQualityWeights from the limma R package 26 . Genes with a TMM count of at least 1 in 10% of libraries, resulting in approximately 19,000 genes, were retained for further analysis. No libraries were removed because of poor quality control. To find differentially expressed genes between screening and C2D15, a linear model with a random effect for individual was run within each cohort. To study cell type proportion changes, cell markers were used from the xCell signature gene sets (University of San Francisco Institute of Computational Health Sciences), which are based on pure cell types from multiple sources 27 . With the xCell gene sets, competitive gene set analysis was performed between screening and C2D15 using the camera function from the limma R package. Finally, visualization of the results was done with gene set variation analysis (GSVA) to calculate a summary metric for each gene set. The P value and adjusted P value threshold were calculated using limma (R package). Linear models included the visit and the bone marrow mononuclear cell storage condition (TRIzol vs. DMSO), plus a blocking term (with duplicate correlation calculated) for patients who had samples matched between screening and C2D15.
T-cell receptor analysis. T-cell receptor (TCR) beta chain sequencing was performed using the immu-noSEQ platform (Adaptive Biotechnologies, Seattle, WA). Bone marrow mononuclear cell (BMMC) samples with 400,000 to 5 million cells were frozen (n = 74; 32 matched patients between screening and cycle 2). Adaptive Biotechnologies performed the DNA extraction, TCR beta amplification, and sequencing. The immunoSEQ Analyzer 3.0 software was used to calculate the clonality for each sample and the pairwise comparisons between screening and cycle 2. Clonality is the inverse of the normalized Shannon entropy such that values near 1 represent samples with many copies of only a few clones, while values near zero are highly polyclonal.

Multiplex immunofluorescence analysis. Bone marrow biopsies for multiplex immunofluorescence
(mIF) were isolated according to standard institutional procedures. Tissue was fixed in 10% formalin, transferred to 70% ethanol, and shipped at ambient temperature to the central laboratory. Two panels were used to characterize the cellular content of the bone marrow biopsy samples using mIF with a PerkinElmer Vectra instrument (PerkinElmer, Waltham, MA). Visiopharm was used to calculate the cell marker densities based on the hematoxylin and eosin-stained nuclei and marker intensities. The innate immune panel had antibodies for CD138, CD38, PDL1, CD163, CMAF, and CD11c. The lymphocyte panel had antibodies for CD3, CD8, CD20, PD1, FOXP3, and CD138. Cell densities for pairs of markers were used to look at more specific populations. On the tumor/myeloid panel, these pairs were used: CD38/CD138, CD138/PDL1, CD38/PDL1, CD163/CMAF, CD163/PDL1, CD11c/PDL1, and CMAF/PDL1. On the lymphocyte panel, these pairwise densities were calculated: CD3/CD8, CD3/FOXP3, CD3/PD1, and CD8/PD1. Each panel was applied on a different slide; cell densities from each slide were analyzed independently. The comparison between responders and nonresponders was done using a Wilcoxon rank sum test using GraphPad Prism (GraphPad Software, San Diego, CA).
Ethical approval. The

Results
Soluble PD-L1 (sPD-L1) levels. At the C1D1 timepoint, sPD-L1 was measured in 18 patients in the durvalumab monotherapy arm (Arm A), 7 patients in the durvalumab + pomalidomide arm (Arm B), and 73 in the durvalumab + pomalidomide + dexamethasone arm (Arm C). The results demonstrated that sPD-L1 is significantly elevated in patients with RRMM compared with 28 healthy volunteers of a similar age (P = 0.00005; Fig. 1a). However, there was no association of sPD-L1 between screening (C1D1) and response to therapy (Fig. 1b). There was also no association of sPD-L1 levels when comparing patients with high-risk cytogenetics versus other patients. Following durvalumab treatment, sPD-L1 levels were below the limit of quantitation in all patients, indicating complete target coverage, even in the presence of elevated screening levels of sPD-L1 (Fig. 1c).
Immune cell subsets in peripheral blood. Results of flow cytometry in the peripheral blood conducted at screening, C1D1, C1D8, C1D15, C2D1, and C2D15 are summarized by arm in Fig. 2, with up to 19, 11, and 84 samples analyzed at each visit in arms A, B, and C, respectively. There was a 50% increase in the proportion of proliferating (Ki67 +) CD4 T cells following durvalumab monotherapy (Fig. 2a). In patients treated with pomalidomide in Arm B or Arm C, larger increases (150-500%) in the proportion of proliferating CD4 T cells, CD8 T cells, and NK cells were detected as early as C1D8 (Fig. 2a-c), and for Arm C this was followed by a decline at the start of cycle 2 (prior to dosing) and then an increase again by C2D15. Both cohorts that received pomalidomide in combination with durvalumab had additional increases in T reg s (CD4 + FoxP3 + CD127 +) (not shown) and CD4 + ICOS + helper T-cell subsets (Fig. 2d) as a proportion of CD4 cells. In the CD8 + compartment of Arm C patients, there were increases in CD8 + ICOS + (Fig. 2e) subsets and effector memory (Fig. 2f,g), and at C1D8 or C1D15, while the central memory cells had a slight decrease initially (C1D8) but then rebounded to higher than initial levels at the start of cycle 2. Some T-cell subsets had a decrease in Arm C patients: there was a 50% decrease in the absolute counts of naive CD8 + T cells and both central memory and naive CD4 + T cells were also decreased by approximately 50% on treatment (Fig. 2f,g). These decreases were not seen in Arm A patients, and Arm B patients trended similarly, but have wider confidence intervals due to the lower number of patients compared with Arm C.
Estimation of cell populations based on RNA sequencing data. Whole bone marrow aspirate samples from patients in all arms were analyzed by RNA sequencing at screening and C2D15. We estimated the relative abundance of different cell populations in each of the 3 arms by applying gene set enrichment analysis of immune cell types based on gene sets from xCell (Fig. 3) 27 and comparing C2D15 to screening samples. Following durvalumab monotherapy, there were inferred increases in CD8 + T-cell, NK-cell, and T reg genes (Fig. 3a). In the combination arms with pomalidomide (Arms B and C; Fig. 3b,c), an increase in genes from dendritic cells, monocytes, and macrophages, in addition to CD8 + T cells and T reg s, was detected. In addition, in Arms B and C, many of the B-cell-related gene sets were enriched in the screening relative to C2D15 samples, consistent with a decrease in tumor plasma cells on treatment.
Gene expression changes in the MM tumor microenvironment. We used the RNA sequencing data to look for individual transcripts that changed within the first 2 months of treatment. Figure 4a shows a volcano plot of the top differentially expressed genes in the durvalumab + pomalidomide + dexamethasone arm (Arm C) between screening and C2D15 such that genes which have higher expression at C2D15 compared with screening have positive fold changes along the x-axis. Figure 4b,c show the screening versus C2D15 volcano plots for Arms A and B, with the top 200 genes from Arm C highlighted in red and blue. Durvalumab alone did little to change the transcriptional profile of bone marrow, as there are no individual genes that have adjusted P values lower than 0.05, and the differential genes from Arm C are not consistently changed in the direction of their fold change (Fig. 4b). The addition of pomalidomide increased the number of differential genes on treatment in Arms B and C (Fig. 4a,c), and the specific gene expression changes on treatment in Arm B were highly consistent with those seen in Arm C (Fig. 3c), despite only 2 samples at C2D15 in Arm B. In Arms B and C, the chemoattractants IL1B, CXCL1, and CXCL9 were upregulated on C2D15, as well as other inflammatory factors such as TNF, PD-L1 (CD274), and PD-1 (PDCD1). Downregulated genes included those from B cells, CD19, CD79B, and BLK. Furthermore, CXCR3, Tim-3 (HAVCR2), and IL7 were also increased, suggesting a role for T-cell function. Separating the Arm C patients by responders and nonresponders did not reveal any significant gene differences at screening or on treatment (not shown).

IFN-γ signature.
Previous studies have used an IFN-γ-based 4-gene signature as a predictive biomarker of efficacy in NSCLC 28 . To determine if this gene signature was induced in any of the patients in this study upon treatment, we used the RNA sequencing data to compare screening and C2D15 bone marrow samples that were also separated into nonresponder and responder groups. There were insufficient samples from Arms A and B to draw inferences about the effect of treatment on the IFN-γ-based gene signature (Fig. 5a,b). However, in Arm C, the increased numbers of samples allowed us to divide the patients by response, with the signature score significantly upregulated in both responder and nonresponder populations (Fig. 5c). Although the IFN-γ-based gene signature increased on treatment, we found that the level at screening was not associated with treatment response.
T-cell receptor sequencing. TCR     www.nature.com/scientificreports/ sequences were compared and each clone was determined to be increasing, decreasing, unchanged, or too low to determine. Patients who responded to therapy had a higher number of clones that expanded on treatment compared with nonresponders (P = 0.044; Fig. 6a). Additionally, the clonality was increased overall on treatment (P = 0.032; Fig. 6b).

Proliferating CD8+ T cells Proliferating CD4+ T cells Proliferating NK cells CD4+ ICOS+ T cells CD8+ ICOS+ T cells Central memory, effector memory, and naive T cells: comparing visits in Arm C
Immunofluorescence of bone marrow. We examined immunofluorescence staining of bone marrow biopsies using 2 panels: one that focused on the innate immune infiltrates and another that was used to interrogate the lymphocyte compartment, including T-cell checkpoint expression. For this analysis, only the durvalumab + pomalidomide + dexamethasone group (Arm C) was evaluable. Each response group was assessed with both panels at screening and on C2D15 (Fig. 7a,b). Potentially due to the low sample numbers, there were few significant differences between responders and nonresponders at screening or C2D15. Responders appeared to have a slight increase in CD163 + PDL1 + cells at both screening and C2D15, although this was not signifi- There were no other differences in the measured cell densities between responders and nonresponders, including no difference for PD-L1 + tumor cells. There were significant changes in the T-cell infiltrates of tumors: responders had an increase in CD8 + T cells at screening and C2D15 (P = 0.0158), including those that also expressed PD-1 + (P = 0.020) (Fig. 7c). There was also an increase in CD3 + PD1 + T cells as a whole (P = 0.015) at C2D15 in the responders. Although not significant, there also appeared to be a slight increase in Foxp3 + T cells in responders at screening and C2D15 (P = 0.100). Taken together, these results suggest that treatment with durvalumab, pomalidomide, and dexamethasone induced a more robust T-cell response in patients that responded to treatment as opposed to those who did not.

Discussion
CPIs have led to positive outcomes in several solid tumor types [16][17][18][21][22][23] . However, recent trials with the anti-PD-1 pembrolizumab in combination with an IMiD agent and dexamethasone in MM were stopped due to decreased survival and increased adverse events 29,30 ; subsequently, all studies with CPIs and IMiD agents were temporarily stopped based on these findings. The future use of CPIs in hematologic malignancies is dependent on a complete understanding of the immunologic impact of these agents in this setting. The current study was designed to investigate the immunomodulatory effects of checkpoint blockade with durvalumab alone and in combination with pomalidomide and dexamethasone. Serum sPD-L1 shed from the surface of several immune lineages, including plasmacytoid dendritic cells, macrophages, and MDSCs, has been shown to be a predictive and prognostic biomarker in various human cancers including MM, diffuse large B-cell lymphoma, NSCLC, and hepatocellular carcinoma 31,32 . In MM, the overall response rate to older therapies, including proteasome inhibitors, has been shown to be higher in patients with low sPD-L1 expression than in those with high sPD-L1 expression 33 . Here, we confirmed previous reports of elevated sPD-L1 in patients with RRMM compared with age-matched healthy volunteers. Durvalumab binds to circulating sPD-L1, which may limit exposure of tumor cells to the drug. As a result, sPD-L1 has also been used as a biomarker of target coverage in solid tumor trials of durvalumab, where treatment has resulted in dose-dependent suppression of free sPD-L1 34,35 . In contrast with these previous trials, in this study sPD-L1 levels were below the limit of quantitation following durvalumab treatment, indicating complete target coverage, even in the presence of elevated screening levels of sPD-L1 in patients with MM.
Similar to the known pharmacodynamic effects of durvalumab monotherapy in solid tumors, we saw modest effects of durvalumab on T-and NK-cell proliferation in peripheral blood, which was further increased with addition of pomalidomide. Of interest, the addition of dexamethasone to the durvalumab + pomalidomide combination did not inhibit the immune response; instead, the onset of immune cell activation was delayed by about 7 days and was not sustained.
A number of studies have reported correlation between various molecular markers, including expression level of PD-L1 in tumor and immune cells and IFN-γ gene expression signatures, and the clinical activity of CPIs, including durvalumab 17,36 . Experience with durvalumab in solid tumors showed that greater responses were observed in patients with PD-L1-positive tumors, with much lower rates of response in patients with PD-L1-negative tumors 37 . In addition, tumors that had an elevated IFN-γ 4-gene signature (IFN-γ, CD274, LAG3, and CXCL9) responded better to treatment 28 . In the study reported here, there was an increase of this signature on treatment in Arm C; there were insufficient data to draw conclusions in the other arms. Notably, the magnitude of the IFN-γ-based 4-gene signature at screening was not associated with treatment response to durvalumab combination therapy in RRMM.
Pomalidomide has shown efficacy in combination with dexamethasone for patients with RRMM who experienced relapse on prior lenalidomide and bortezomib 38,39 . Here, in the triplet regimen arm, the IFN-γ 4-gene signature was significantly upregulated in both responder and nonresponder populations following treatment. The changes in gene signatures observed during treatment with pomalidomide resembled those that have been observed with IMiD agent treatment alone in other hematologic malignancies, and are indicative of the strong immunomodulatory effects of IMiD agents 40 . As a result, this study provided us with additional insight into the mechanism of action of IMiD agents in the MM tumor microenvironment. Pomalidomide also had a clear immunostimulatory effect, increasing cycling T-cell and NK-cell populations (including T reg , helper T-cell, effector memory cell, and central memory cell subsets), dendritic cells, monocytes, and macrophages, as well as an IFN-γ signature, similar to previous studies with other IMiD agents [41][42][43][44][45] . The addition of pomalidomide likely increased the expression of these genes, owing to increased T-cell activation in these samples. This is consistent Figure 4. Difference in transcriptional profile of whole bone marrow samples at screening and cycle 2 day 15 (C2D15) from patients in all three cohorts using RNA sequencing. In panel A, the volcano plot shows each gene as a point and its log 2 fold change (x-axis) between screening and C2D15 for patients in Arm C (durvalumab, pomalidomide, dexamethasone), where positive values mean higher expression of the gene in C2D15 compared to screening, and negative values are the inverse; and, the y-axis shows the negative log 10 of the raw P value with a horizontal line drawn where the multiple-testing adjusted P value (False Discovery Rate -FDR) is at 0.05. In Arm C there were 50 samples at C2D15 and 53 samples at screening. In B and C, the volcano plots are based on the differences between C2D15 and screening for Arm A and Arm B, respectively. In those panels, the top 200 differentially expressed genes from Arm C are colored red if they were upregulated on treatment in Arm C and blue if they were downregulated on treatment in Arm C so that it could be determined if they were changing in the same direction in those smaller cohorts. In Arm A (durvalumab), there were 15 patients at screening and 7 patients at C2D15. In Arm B (durvalumab + pomalidomide), there were 6 patients at screening and 2 patients at C2D15. C2D15, cycle 2 day 15; FDR, false discovery rate.  www.nature.com/scientificreports/ with our earlier work that observed increased activated CD4 + and CD8 + T cells in bone marrow in patients treated with pomalidomide compared with treatment-naive patients 46 . Because all arms in this study included durvalumab, it was not possible to determine whether its addition to pomalidomide had any differential effect. TCR beta chain sequencing of bone marrow samples showed that there was an overall expansion in T-cell clonality in patients receiving treatment. Furthermore, responders had more individual clonal expansion than nonresponders. While this is consistent with observations in patients with solid tumors treated with checkpoint inhibitors 47 , it is not clear whether this was due to durvalumab since all patients in the study were treated with the drug. The combination of durvalumab and pomalidomide induced significant inflammatory changes in the blood and bone marrow, whereas there was little evidence for immunostimulatory activity with durvalumab monotherapy. The apparent lack of immunomodulation with durvalumab when given in combination with pomalidomide in MM may further explain the limited clinical efficacy of these agents in this setting.
A more complete understanding of the clinical benefit of immuno-oncology agents in this setting due to enhanced immune cell competency will inform optimal therapeutic sequencing and use of rational combination regimens. An understanding of the interplay of the distinct mechanisms of cereblon E3 ligase modulators, chimeric antigen receptor T cells, and T-cell engagers will be key to development of the next generation of combination therapies for RRMM.