Marrow-Infiltrating Regulatory T Cells Correlate with the Presence of Dysfunctional CD4⁺PD-1⁺ Cells and Inferior Survival in Patients with Newly Diagnosed Multiple Myeloma

Multiple myeloma is a common cancer of plasma cells (PC) which is responsible for 2% of cancer deaths (1). Despite significant progress seen with the inclusion of proteasome inhibitors and immunomodulatory drugs (IMiD) into the mainstay of treatment regimens (2), myeloma remains almost universally incurable. Along with intrinsic drug sensitivities of tumor cells, and genomic drivers of clonal evolution, host factors, including immunological fitness and function, likely also influence clinical outcomes of treatment.

Accumulating evidence points toward a global immune dysregulation in multiple myeloma including impaired antigen presentation (3) and impaired T-cell effector function (4) with accumulation of suppressive cell types (5, 6). These mechanisms appear to converge on disabling T cell–driven antitumor immunity (7) and accordingly alterations in T-cell phenotype and function have been consistently reported in models of multiple myeloma. First, T regulatory cells (Treg) suppress T-cell cytotoxicity and have been reported to be an important driver of disease progression (8). Second, there appears to be a relative reduction in cytotoxic T cells relative to Tregs (8). Third, checkpoint proteins, such as the coinhibitory receptor, PD-1, are reported to be expressed on T cells from patients with multiple myeloma (9, 10) with increased expression of its ligand PD-L1 on tumor cells (11). Despite these reports, the influence of these alterations to T-cell phenotype on patient outcomes remains to be clarified. Data regarding Treg numbers and relationship to clinical outcomes are conflicting (12–14) and reports of increased PD-1 on T cells from patients with multiple myeloma have not been generally corroborated or correlated to outcome (15). Reasons for these discrepancies include different assay systems, examination of peripheral blood (PB) versus marrow or the use of heterogenous patient cohorts. Many studies included relapsed refractory patients, where the host immune system is likely to be affected by prior therapies, repeated infection, and advanced disease.

To resolve some of these issues, we investigated the marrow-infiltrating T-cell populations in untreated patients with multiple myeloma, with focus on Tregs and coinhibitory receptors seeking to understand the influence of these recognized suppressive T-cell populations on the clinical outcomes of first-line treatment.

Bone marrow (BM) aspirates were obtained from patients with newly diagnosed (ND) multiple myeloma with written informed consent (Research ethics committee reference: 07/Q0502/17). Control BM aspirates (n = 15) were collected from healthy volunteers undergoing BM harvesting with Anthony Nolan, and subjects undergoing BM sampling who had no hematologic diagnosis (Supplementary Table S1; REC reference: 15/YH/0311). All BM samples were collected in ethylenediamine-tetraacetic acid and processed within 24 hours. Patients were considered to have adverse risk disease if FISH demonstrated one of these: t(4;14), t(14;16), t(14,20), and del(17p).

BM mononuclear cells (MNC) were isolated by Ficoll Paque (GE Healthcare) centrifugation and cryopreserved in FBS (Gibco) containing 10% DMSO (Sigma-Aldrich). Aliquots were subsequently thawed for antibody staining and flow cytometry, functional studies, or RNA sequencing.

Surface antigen staining was performed using the fluorochrome-conjugated antibodies CD3, CD4, PD-1, ICOS, CD25, CD33, CD11b, CD8, LAG-3, CD4, CD14, CD45RA, CCR7, and fixable viability dye-e780. For intracellular staining, cells were fixed/permeabilized using the FoxP3 Transcription Factor Staining Buffer Set (eBioscience), then stained with FoxP3, CTLA-4, Ki-67, and GzmB. Details of all antibodies are in Supplementary Table S2. Data acquisition was on a BD LSR II Fortessa (BD Biosciences).

Cryopreserved BM MNCs were thawed and cultured at 0.5 × 10¹⁶ cells/mL in RPMI (Lonza), 20%FBS (Gibco), and 1% penicillin/streptomycin (Gibco; complete medium), at 37°C with soluble anti-CD3 (OKT3) and anti-CD28 (0.5 μg/mL, 15E8; Miltenyi Biotec). GolgiPlug (1 μL/mL, BD Biosciences) was added for last 4 hours of incubation. Cells were then stained for surface markers, CD4, CD8, CD69, and fixable viability dye, washed and fixed/permeabilized for staining for intracellular TNFα, IFNγ, IL2, and FoxP3 (Supplementary Table S2).

RNA was extracted from flow sorted CD3⁺CD4⁺ and CD3⁺CD8⁺ cells from BM MNCs using ReliaPrep RNA Cell Miniprep System (Promega). cDNA libraries were prepared using the SMART-Seq v4 Ultra Low Input RNA Kit (Clontech Laboratories, Inc.). Samples were sequenced on two lanes of the HiSeq 3000 instrument (Illumina) using a 75 bp paired-end run at UCL Institute of Child Health (London, UK).

RNAseq data were processed with a modified version of the nextflow nf-core RNAseq pipeline (https://github.com/nf-core/rnaseq). Reads were trimmed with TrimGalore v0.4.1, aligned against hg19 with STAR v2.5.2a, and duplicated reads were marked with Picard v2.18.9. Read counts per gene were generated with featureCounts v1.6.2 and used for differential gene expression analysis. Gene set enrichment analysis (GSEA) was run using Gene Ontology pathways and previously reported sets of genes differentially expressed by dysfunctional CD4 (16–18) and CD8 T cells (19, 20). Human orthologues of mouse genes were identified using Ensembl and NCBI HomoloGene databases.

Progression-free survival (PFS) was defined as time from start of first-line therapy to first progression or death [as per International Myeloma Working Group criteria (21)]. Flow cytometric data were analyzed with FlowJo version 10 (Tree Star Inc). The percentage of a cell population expressing any given marker is designated as “frequency” (of that marker) within the relevant Treg, CD4 effector, or CD8 populations. Statistical analyses were performed with GraphPad Prism software (Prism 7). P values were calculated using Mann–Whitney U test. PFS was estimated using Kaplan–Meier methods with log-rank test. A multivariate Cox regression model was used to evaluate the independent contribution of variables. All tests of significance were two-sided and P values ≤0.05 considered statistically significant.

Seventy-eight patients with ND multiple myeloma were identified, with median age 59 years (35–86), 64.1% were male (Supplementary Table S3). FISH-defined genetic risk was available in 74 patients, of whom 19.2% were adverse risk. All patients commenced active treatment, most (68, 87.18%) with proteasome inhibitor regimens, and 25 (31.25%) underwent autologous stem cell transplant (ASCT). Overall response rate was 87%, and 53.8% achieved complete response/very good partial response (CR/VGPR). With median follow-up of 22 months (1–43), median PFS was not reached. There was a trend for improved PFS with standard risk genetics (P = 0.075 cf high risk), ASCT (P = 0.06), and in patients with deeper response (CR/VGPR vs. rest, P = 0.09; Supplementary Fig. S1).

We first examined the relative frequencies of T-cell subsets in the BM of patients with multiple myeloma (gating strategy in Fig. 1A). While the frequencies of CD3, CD4, and CD8 cells were comparable with healthy donors (HD; Supplementary Fig. S2A), the frequency of Treg cells (CD4⁺FoxP3⁺) was significantly higher in BM of patients with multiple myeloma (0.51% of live MNCs vs. 0.07% in HD; P < 0.0001, 3.33% of CD4⁺ cells vs. 1.13%; P = 0.0006; Fig. 1B). This was also the case when Tregs were identified as CD4⁺CD25⁺FoxP3⁺ (3.41% of CD4 cells in multiple myeloma BM vs. 1.27% in HD; P = 0.001; Fig. 1B).

The balance between Tregs and effector T cells shapes the antitumor immune response (22). We defined CD4 effectors (CD4^eff) as CD4⁺FoxP3⁻ cells, and observed that the CD4^eff:Treg ratio in patients with multiple myeloma was significantly lower when compared with HD (20.83 vs. 140.2; P = <0.0001), this was also the case for the CD8:Treg ratio (36.34 vs. 170.4; P = <0.0001; Fig. 1C). We found no correlation between Treg cells, CD4^eff:Treg ratio or CD8:Treg ratio with percentage of PCs in BM (Supplementary Fig. S3A). Neither did we find any correlation of CD4:CD8 ratio with PC infiltration.

We sought to determine whether the presence of Treg cells in the BM of ND patients had any influence on clinical outcomes. We used PFS, a common primary endpoint for studies in patients with multiple myeloma (23). Identifying Tregs as CD4⁺FoxP3⁺ cells, we observed that patients with multiple myeloma with a high frequency of Tregs (>median, Treg^hi) had significantly shorter PFS when compared with patients with multiple myeloma with low frequency of Tregs [≤median, Treg^lo; HR, 2.91; 95% confidence interval (CI), 1.21–7.04; P = 0.021; Fig. 2A]. Similar findings were also seen when Tregs were identified as CD4⁺FoxP3⁺CD25⁺ cells (P = 0.022; Supplementary Fig. S2B). We used surv_cutpoint function from the “survminer” R package (https://github.com/kassambara/survminer) to determine the optimal cut-off value for Treg frequency, and ascertained this to be 3.31%, which is the median value.

Having noted that the ratios of effector cells to Tregs in patients with multiple myeloma are low compared with HD, we next examined the association with PFS. We observed that patients with low CD4^eff:Treg ratios (≤median) had significantly shorter PFS compared with high CD4^eff:Treg ratios (>median; HR, 4.22; 95% CI, 1.79–10.15; P = 0.005; Fig. 2B). There was a weaker association of CD8:Treg ratio with PFS (P = 0.067; Fig. 2B). Triple color IHC was performed on BM trephine biopsies to confirm presence of Tregs in representative Treg^hi and Treg^lo patients (Fig. 2C). There were no associations between CD4 effectors, CD8 cells, or CD4:8 ratio with PFS (Supplementary Fig. S2B).

We next examined the phenotype of marrow-infiltrating Tregs, to better understand their influence on clinical outcomes. We observed higher expression of CD25 on Tregs from patients with multiple myeloma compared with HD suggesting higher level of activation of multiple myeloma Tregs (Fig. 3A), as CD25 expression is associated with Treg activity and suppressive function (24). In this cohort of patients with multiple myeloma, both the abundance of CD25^hi cells and expression intensity of CD25 [mean fluorescence intensity (MFI)] was greater among Tregs compared with CD4 effectors and CD8 T cells (Fig. 3B). While there were no significant differences in frequencies of PD-1, LAG-3, or CTLA-4 on Tregs from patients with multiple myeloma compared with HD (Fig. 3A), there was a greater frequency of PD-1 and LAG-3 on Tregs from Treg^hi patients compared with Treg^lo (Fig. 3C). These differences in checkpoint protein expression suggest that functional as well as quantitative features of marrow-infiltrating Tregs in patients with multiple myeloma may be important (25, 26). We further explored the differentiation status of BM Tregs in a separate cohort of patients with ND multiple myeloma, observing that the majority are CD45RA⁻ indicating that marrow Tregs in these patients have an activated phenotype (Fig. 3D).

Next we asked whether altered Treg frequency and activation state were reflected in effector T-cell function in multiple myeloma BM. Examining coinhibitory and coactivation receptors on CD4^eff and CD8 T cells, we observed that frequencies of LAG-3 and Ki-67 were higher on both CD4^eff and CD8 T cells from patients with multiple myeloma compared with HD (P = 0.001, P = 0.009, P = 0.0001, P = 0.0001, respectively; Fig. 4A and B) with no significant differences in ICOS or CTLA-4 (Fig. 4A and B). In addition, a higher percentage of CD8 T cells from patients with multiple myeloma expressed PD-1 (P = 0.045) and the cytotoxic granule GzmB compared with HD (P = 0.01; Fig. 4B). There was no correlation between the frequency of any coinhibitory or coactivation receptor on CD4^eff or on CD8 T cells with disease burden in the BM, except for frequency of LAG-3 on CD8 T cells (r = 0.27, P = 0.028; Supplementary Fig. S3B).

Notably, we observed a positive correlation between Treg frequency and the fraction of PD-1⁺ CD4^eff and CD8 cells (Supplementary Fig. S4), but no correlation with the frequency of any other coinhibitory or coactivation receptors. Accordingly, PD-1 expression on CD4 effectors also correlated with PD-1 on CD8 cells (Supplementary Fig. S4), and a positive correlation was also noted between PD-1 expression on Tregs and on CD4 effectors (Supplementary Fig. S4D). To understand the relationship between PD-1 expression and differentiation status of marrow-infiltrating effector cells, we further studied a similar cohort of patients with ND multiple myeloma. Interestingly, while terminally differentiated effector memory cells reexpressing CD45RA comprise a large proportion of CD8 cells, this subset comprises only a minority of CD4 effectors, with the effector memory (EM) subset being dominant in most patients (Supplementary Fig. S5A). PD-1⁺CD4 effectors were enriched for central memory (CM, CCR7⁺CD45RA⁻), and EM (CCR7⁻CD45RA⁻) cells when compared with PD-1-CD4 effectors (Supplementary Fig. S5B).

Finally, the frequency of monocytic myeloid-derived suppressor cells (M-MDSC) in the BM of patients with multiple myeloma was higher when compared with HD (P = 0.006; Supplementary Fig. S6A). The frequency of M-MDSCs showed only a weak correlation with CD4^effPD-1⁺ levels (Supplementary Fig. S6).

Next we examined the association of coinhibitory receptor expression on CD4 and CD8 effectors with clinical outcomes. When we divided patients into two groups based on the frequency of PD-1 on CD4^eff, we observed that patients with multiple myeloma with more CD4^effPD-1⁺ cells (>median, termed CD4^effPD-1^hi) had significantly shorter PFS compared with those with less CD4^effPD-1⁺ cells (≤median, CD4^effPD-1^lo; HR, 3.98; 95% CI, 1.66–9.55; P = 0.007; Fig. 4C). In contrast, there was no correlation between frequency of PD-1 on CD8 T cells and PFS (Fig. 4C). There was no correlation between frequency of LAG-3, ICOS, or CTLA4 on either CD4^eff or CD8 T cells and PFS (Supplementary Fig. S7A–S7C). Similarly, no correlation was found between GzmB or Ki-67 or on either CD4^eff or CD8 T cells and PFS (Supplementary Fig. S7D and S7E).

Given the association with clinical outcomes, we examined the CD4^effPD-1⁺ cell fraction in multiple myeloma in more detail. This subset coexpressed the exhaustion markers LAG-3/CTLA-4 and the terminal differentiation marker GzmB more frequently in CD4^effPD-1^hi compared with CD4^effPD-1^lo patients (P = 0.0035, P = 0.046, P = 0.034, respectively; Fig. 4D), suggesting this subset is characterized by a dysfunctional state that is more pronounced among CD4^effPD-1^hi patients.

Having identified immune features with prognostic value, we examined both CD4^eff:Treg ratio and CD^effPD-1⁺ cell frequency for associations with known clinical prognostic parameters. We found no association between ISS, genetic risk, ASCT, or response depth with either CD4^eff:Treg ratio or CD4^effPD-1⁺ cells (Supplementary Fig. S8). A multivariate Cox regression model was built including genetic risk, ASCT, ISS and depth of response, and the immune features identified above. In this model, CD4^eff:Treg ratio retained independent prognostic value, along with CD4^effPD-1⁺ cells, genetic risk, ASCT, and depth of response (Fig. 5A). A risk model was bulit including CD4^eff:Treg ratio, CD4^effPD-1⁺ cells, and genetic risk, stratifying patients into three risk groups based on diagnostic features. Patients with two or more risk factors had significantly shorter PFS (Fig. 5B).

To gain mechanistic insight into the potential dysfunction of effector T cells from CD4^effPD-1^hi patients, we sorted CD4 and CD8 cells from CD4^effPD-1^hi and CD4^effPD-1^lo patients for RNA sequencing. GSEA carried out using gene sets from previous studies of impaired CD4 function (16–18) revealed that CD4 cells from CD4^effPD-1^hi patients have transcriptional features of CD4 dysfunction. Among three gene sets tested, all were enriched among genes differentially expressed by CD4 cells from CD4^effPD-1^hi patients, although only the Tilstra and colleagues signature reached statistical significance (P <0.001; Fig. 6A). Similarly, CD8 T cells from CD4^effPD-1^hi patients also displayed transcriptional features of dysfunction (Fig. 6A). We then performed GSEA to identify pathways enriched in T cells from CD4^effPD-1^hi versus CD4^effPD-1^lo patients. Pathways related to activation downstream of T-cell receptor (TCR) signalling, proliferation, and regulation of apoptosis were enriched in CD4 cells from CD4^effPD-1^hi patients (Supplementary Fig. S9A). Similar pathways of activation and proliferation were also upregulated in CD8 T cells from CD4^effPD-1^hi patients (Supplementary Fig. S9A), as described previously for dysfunctional CD8 T cells (27, 28).

To further explore the notion that T cells from CD4^effPD-1^hi patients are functionally impaired, we next assessed cytokine secretion by stimulating whole BM MNCs with anti-CD3 and anti-CD28 antibodies. We found that after 6-hour stimulation, there was a trend toward higher TNFα, IFNγ, and IL2 production in activated CD4 effectors (CD4⁺FoxP3⁻CD69⁺) from CD4^effPD-1^lo patients compared with CD4 effectors from CD4^effPD-1^hi patients; however, only the frequency and intensity (MFI) of TNFα reached statistical significance (P = 0.0043; Fig. 6B; P = 0.0411; Supplementary Fig. S9B). A similar pattern was observed with activated CD8 T cells (CD8⁺CD69⁺) from patients with CD4^effPD-1^lo; these effectors produced more TNFα compared with those from CD4^effPD-1^hi patients (P = 0.026; Fig. 6B), with a trend toward higher IFNγ and IL2 production.

Collectively, these data suggest that CD4 effectors and CD8 T cells from CD4effPD-1^hi patients display transcriptional and functional features of dysfunction that may contribute to poorer outcomes.

We present data correlating the phenotype and function of BM CD4 T-cell subsets at diagnosis to clinical outcomes of first-line treatment in a large cohort of patients with multiple myeloma. Specifically, we report for the first time that patients with a high frequency of marrow-infiltrating Tregs at diagnosis have poorer clinical outcomes. Beyond numerical differences, high frequency of Tregs is accompanied by phenotypic changes (increased PD-1 and LAG-3) suggestive of increased suppressive capacity. Tregs contribute to cancer progression by directly suppressing the effector T-cell activity and here we also report that CD4^eff:Treg ratio may be independently prognostic in multiple myeloma. We are also the first to present data in multiple myeloma correlating PD-1 expression on CD4 T cells to patient outcomes, and to impaired cytokine production as well as transcriptional signatures of dysfunctional CD4 and CD8 cells. This supports a growing body of evidence underpinning the role of CD4 T cells in the antitumor immune response (29), and suggests the independent importance of immune dysregulation on prognosis.

Myeloma cells have been shown to promote Treg expansion in vivo (8) and in vitro (30). In addition, Treg depletion improves survival in a syngeneic murine model of multiple myeloma (8), indicating that this is a key immunosuppressive population that facilitates disease progression. Previous studies report higher levels of Tregs in PB in patients with multiple myeloma compared with age-matched controls (6, 12) and in BM compared with patients with MGUS (14). One study reported that higher levels of Tregs in PB correlated with shorter time to progression (14), but no study has systematically examined Treg numbers and phenotype in the BM of ND patients. Ours is the first study to examine BM-infiltrating Tregs at diagnosis and significantly extends these earlier reports because we show for the first time that the CD4^eff:Treg ratio in the tumor environment independently associates with clinical outcomes. We also observed that increased Treg numbers associated with greater frequencies of the checkpoint proteins, PD-1 and LAG-3 (on Tregs), consistent with murine models of multiple myeloma (8). Previous work has confirmed the suppressive function of Tregs from BM of patients with multiple myeloma (31, 32), while expression levels of these checkpoint proteins is reported to associate with Treg suppressive function in other cancers (25, 26, 33). Further functional and molecular studies on PD-1–expressing Tregs from BM of patients with multiple myeloma are planned, to provide mechanistic insights.

Tregs actively suppress cytolytic T-cell activity (8), and the ratio of Tregs to effector cells has been reported to correlate with survival outcomes (6). In this series of patients, the high frequency of Tregs in the BM resulted in lower effector T cell: Treg ratios; however, only the CD4^eff:Treg ratio significantly correlated with PFS. In comparison, there was only a trend of CD8:Treg ratio to outcome (P = 0.067) which challenges the prevailing view that CD8⁺ T cells are the dominant contributors to antitumor immunity (34). Indeed, the antitumor functions of the CD4 tumor compartment are increasingly recognized (29) which encompasses their helper function for cytotoxic CD8⁺ T cells (35) as well as the ability to directly eliminate tumor (36). In multiple myeloma, CD4-mediated cytolysis of autologous tumor cells has been demonstrated in vitro (37) and in a syngeneic murine myeloma model, direct CD4-mediated cytotoxicity was demonstrated even in the absence of tumor MHC II expression (38). Moreover, in a recent in vivo autograft model, significant reduction in tumor control was observed on depletion of either CD4 or CD8 T cells (39).

PD-1 is an early marker of the T-cell dysfunction observed in chronic infections and cancer characterized by a hierarchical loss of effector function and proliferation. Classically, analysis of this dysfunctional immune state has focused on CD8 T cells (40). Studies in small patient cohorts report increased PD-1 levels on CD8 cells in the PB and BM of patients with multiple myeloma (10, 41), but we are the first to show that PD-1 on CD4 cells is prognostic of clinical outcomes. Despite a correlation between PD-1 on CD4 effectors and CD8 cells, we did not find any association of CD8 parameters with clinical outcomes. On the other hand, the CD8 compartment from CD4^effPD-1^hi patients also (as well as CD4 effectors) manifested reduced cytokine secretion and transcriptional features of dysfunction, suggesting that the presence of increased PD-1⁺CD4 effectors is indicative of a broader, pan-T-cell dysfunctional phenotype.

Examining transcriptomic profiles of T cells from CD4^effPD-1^hi patients, we observed enrichment of pathways that are characteristic of T-cell dysfunction. Among both CD4 and CD8 T cells, we found enrichment of both TCR and nonclassical NF-κB pathways indicative of ongoing antigen stimulation and activity of costimulatory pathways (42), respectively. In keeping with upregulated TCR signalling, we found enrichment of pathways related to transcription and cell cycle, suggestive of cell activation. While initial reports of T-cell dysfunction in murine models of chronic infection indicated a near total loss of T-cell effector function (43), it is increasingly clear from studies of solid malignancy that the effector potential of dysfunctional T cells is reduced but not absent and active cell proliferation is a key feature of this state (28, 44). Consistent with previous reports of T-cell dysfunction (45, 46), we additionally observed enrichment of metabolic pathways including oxidative phosphorylation among both subsets and an expression profile indicative of heightened sensitivity to apoptosis among CD4 but not CD8 T cells.

Impaired cytokine production by dysfunctional T cells has been reported previously (28) and we extend this finding to BM-infiltrating T cells in multiple myeloma. Here we tested T-cell cytokine production and found this to be reduced in both CD4 and CD8 effectors from CD4^effPD-1^hi patients that reached statistical significance only for TNFα. Larger studies that take into account several variables such as stimulus, duration of stimulation, and cell population are required to confirm these observations. We observed increased numbers of MDSCs in multiple myeloma but further work is needed to explore the contribution of the myeloid compartment to the immune dysfunction in untreated multiple myeloma marrow.

In this work, we used patient BM as opposed to PB as we wished to examine the multiple myeloma–driving, immune changes within the tumor microenvironment. Recent in vivo multiple myeloma models report differences in the immune phenotype of circulating and BM-infiltrating T cells (8) in disease, and indicate earlier changes within the BM immune microenvironment. Similarly, a study in patient samples also reported functional differences between BM and PB effector T cells (47). In addition, we found the age of patients did not correlate with CD4^eff:Treg ratio or CD4^effPD-1 cells. However, as our cohort of healthy donors were younger, comparisons with patients with myeloma need to be interpreted with caution. Another point to note is that a minority of patients (10%) had >80% BM PC infiltration, which may have amplified differences in marker expression, thus our findings await confirmation in further patient cohorts.

Our study suggests that immune parameters in BM of untreated patients with multiple myeloma may inform risk of relapse, and that combining such immune features with genetic risk in a new risk model identifies patients likely to have very poor outcomes. In this patient cohort, we used the median frequency of Tregs (3.31%) as a cut-off value (confirmed using “survminer”). Pending confirmation in a larger validation cohort, this measure could be used to identify patients with inferior treatment outcomes who may benefit from adjunctive immune-directed therapies, for example, Treg depletion strategies. Promising agents include IFNα/IFNβ receptor antagonists and the use of CD25 antibodies optimized for depletion (22). Blockade of the PD-L1/PD-1 axis has already been explored in multiple myeloma (48), but in the relapsed refractory setting, and it remains to be established if checkpoint blockade could overcome immune dysfunction in ND patients, for example, with high CD4 effector levels of PD-1 either as a monotherapy or in combination with Treg-depleting agents. The disappointing results of single-agent checkpoint blockade in multiple myeloma has been suggested to relate to T-cell senescence rather than exhaustion (49). These authors, however, only examined CD8⁺ T cells, thus the question of the effect of PD-1 blockade on CD4 effector function remains unanswered. Interestingly, only 3 of 78 patients in our cohort received the IMiD lenalidomide, which acts to enhance cytokine release, augmenting T-cell costimulation signals (50). Thus, the prognostic impact of PD-1 expression on CD4 cells remains to be confirmed in the context of lenalidomide therapy.

In conclusion, our work demonstrates that increased Tregs in association with dysfunctional CD4 effectors identified by high PD-1 expression correlate with significantly shorter PFS in patients with ND multiple myeloma. These data support the importance of CD4 T cells as mediators of antitumor immunity in myeloma and prompt further mechanistic studies to gain better understanding of the biology of CD4 dysfunction and Treg function, and open up therapeutic opportunities for these patients.

M. Pule is an employee/paid consultant for and holds ownership interest (including patents) in Autolus Ltd. K.L. Yong reports receiving research grants from Takeda, Amgen, Sanofi, and Celgene, and speakers bureau honoraria from Sanofi, Janssen, and Amgen, and is an advisory board member/unpaid consultant for Janssen. No potential conflicts of interest were disclosed by the other authors.

Conception and design: N. Alrasheed, M. Rodriguez-Justo, M. Pule, S.A. Quezada, K.L. Yong

Development of methodology: N. Alrasheed, L. Lee, D. Patel, M. Pule, K.L. Yong

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N. Alrasheed, M. Chin, D. Galas-Filipowicz, A.J.S. Furness, S.J. Chavda, H. Richards, O.C. Cohen, D. Patel, A. Brooks

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N. Alrasheed, E. Ghorani, J.Y. Henry, L. Conde, A.J.S. Furness, S.J. Chavda, H. Richards, A. Brooks, M. Rodriguez-Justo, M. Pule, J. Herrero, S.A. Quezada, K.L. Yong

Writing, review, and/or revision of the manuscript: N. Alrasheed, L. Lee, E. Ghorani, A.J.S. Furness, O.C. Cohen, D. Patel, M. Rodriguez-Justo, S.A. Quezada, K.L. Yong

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Chin, D. Galas-Filipowicz, D. De-Silva, D. Patel

Study supervision: S.A. Quezada, K.L. Yong

The authors thank Prof. David Linch (Formerly Head of the Department of Haematology at University College London) for valuable comments on this manuscript, and Dr. Nicholas Counsell at Cancer Research UK and UCL Cancer Trials Centre for statistical assistance. This study was supported by grants from the Medical Research Council (MR/S001883/1), and by King Faisal Hospital and Research Centre, Saudi Arabia. This work was undertaken with support from the Cancer Research UK (CRUK)-UCL Centre (C416/A18088), and a Cancer Immunotherapy Accelerator Award (CITA-CRUK; C33499/A20265), at University College London/University College London Hospitals, which is a National Institute for Health Research Biomedcial Research Centre, and a Bloodwise Research Centre of Excellence. S.A. Quezada is a CRUK Senior Cancer Research Fellowship (C36463/A22246) and is funded by a CRUK Biotherapeutic Program Grant (C36463/A20764).

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.