Immunotherapy of Lymphoma and Myeloma: Facts and Hopes

Blocking inhibitory surface receptor–ligand pairs, which function to limit T-cell activation and autoimmunity, has revealed a critical role for immune checkpoints in aiding cancer's evasion of host immunity (1–3). Blockade of immune checkpoints cytotoxic T-lymphocyte antigen-4 (CTLA-4) and programmed cell death protein 1 (PD-1) is revolutionizing treatment in many types of solid tumors by stimulating endogenous antitumor immune responses (4). Immune checkpoint blockade therapy (CBT) is also under development in several subtypes of hematologic malignancies, with impressive responses seen in relapsed/refractory (R/R) classical Hodgkin lymphoma (cHL) and recent promising results in multiple myeloma by combining CBT with immunomodulatory drugs (IMiD). Herein, we will review the development of CBT for the treatment of lymphoid cancers to date and discuss opportunities for future progress.

Hodgkin lymphoma can be cured in the majority of cases; however, despite optimal therapy, salvage autologous hematopoietic stem cell transplant (auto-HSCT), and brentuximab vedotin (BV), additional treatment options are needed for a subset of patients who relapse. cHL is characterized by the presence of an inflammatory immune infiltrate surrounding the malignant Hodgkin Reed Sternberg (HRS) cell and near universal genetic amplification of the 9p24.1 locus that encodes the PD-1 ligands as well as JAK2, which in a dose-dependent fashion can further upregulate PD-L1 expression via JAK2–STAT signaling (5). These observations formed the rationale for exploring CBT in this patient population. Patients with cHL treated with anti–PD-1 experienced objective response rates that were higher than expected, suggesting a potential intrinsic sensitivity to PD-1 blockade directly correlated with the degree of 9p24.1 amplification (5–11).

Patients with R/R cHL after auto-HSCT and BV receiving nivolumab on the phase I CheckMate 039 study (7) had an 87% overall response rate (ORR), with 17% reaching a complete response (CR) and 70% achieving partial response (PR). The phase II CheckMate 205 study (9) demonstrated an ORR of 66%, with seven patients reaching CR and 26 patients reaching PR. The phase I study of pembrolizumab (KEYNOTE-013) showed an ORR of 58% and a CRR of 19%, and 12% of patients reached PR (12, 13). In the phase II study exploring pembrolizumab (KEYNOTE-087) among three cohorts defined by history of auto-HSCT and exposure to BV, there was an ORR of 65.4% to 68.3% and a CRR of 21.7% to 20%, and 93.7% of patients had a reduction in their tumor burden by radiographic assessment (14). In patients with R/R cHL after auto-HSCT and BV, the landmark clinical trials of immune checkpoint blockade led to accelerated approval of nivolumab and pembrolizumab by the FDA for this indication.

Beyond its use in R/R patients, PD-1 pathway blockade in combination therapies is being rapidly explored in other cHL populations, including newly diagnosed patients, autologous transplant in the salvage setting, transplant-ineligible patients, brentuximab-naïve patients, and patients with localized early-stage disease with unfavorable characteristics (Table 1). Early data are encouraging. Interim results from the phase I/II study of nivolumab combined with BV as first salvage therapy after frontline chemotherapy before auto-HSCT showed a complete response rate (CRR) of 63% among the 59 evaluable patients—a rate significantly higher than expected with use of either agent alone (15). In R/R cHL, early data from the phase I ECOG-ACRIN E4412 study recently presented showed a CRR of 61% in 18 evaluable patients among 19 treated with the combination of nivolumab plus one of two dose levels of BV (n = 10 with 1.2 mg/kg and n = 9 with 1.8 mg/kg; ref. 16). In the brentuximab plus ipilimumab arms, patients treated with BV 1.8 mg/kg plus one of two dose levels of ipilimumab (1 or 3 mg/kg) responded at a rate of 67%, with five of 12 (42%) achieving CR, with responses seen at both dose levels (17). BV plus nivolumab will be further evaluated in a pending phase III clinical trial in auto-HSCT–ineligible or R/R patients (CheckMate 812; NCT03138499).

Among the non-Hodgkin lymphomas (NHL), PD-L1 overexpression is observed in many entities, including primary mediastinal large B-cell lymphoma (PMBL), primary central nervous system (CNS) lymphoma, primary testicular lymphoma, plasmablastic lymphoma, HHV-8–associated primary effusion lymphoma, T-cell/histiocyte-rich B-cell lymphoma, both Epstein–Barr Virus (EBV)-positive and EBV-negative posttransplant lymphoproliferative disorders, and EBV-associated diffuse large B-cell lymphoma and extranodal natural killer (NK)/T-cell lymphoma (ENKL; refs. 18, 19). Some NHL subtypes, such as PMBL, derive PD-L1 overexpression from 9p24.1 mutations or copy-number alterations (5, 19). In other entities, EBV drives PD-L1 overexpression through a mechanism independent of 9p24.1 amplification through effects of the EBV-encoded latent membrane protein-1 (LMP1), which promotes AP1 and JAK–STAT signaling and increases PD-L1 expression via an AP-1–dependent enhancer (Fig. 1; refs. 19–21). Recent studies have focused on entities with PD-L1 expression, and promising activity was observed in the phase Ib study with PMBL (an ORR of 41% among 17 patients) as well as a phase II study in mycosis fungoides/Sézary syndrome (an ORR of 38% among 24 patients; refs. 10, 22, 23). In addition, impressive activity was reported in small retrospective series of patients with ENKL and CNS lymphoma (20, 21). Building upon these data, a prospective study in CNS lymphoma is underway (NCT02857426), and further analysis in ENKL is certainly warranted. Apart from PD-1, markers of immune exhaustion LAG-3 and TIM-3 are coexpressed in T-cell infiltrates in NHL and represent potential additional targets for checkpoint blockade with in vitro data supporting this approach (24, 25).

Despite remarkable activity of anti–PD-1 in cHL and several variants of NHL, a subset of patients experience progressive disease after an initial response or are primary refractory to PD-1 blockade, underscoring the importance of elucidating mechanisms of response and resistance beyond 9p24.1 amplification. Studies from solid tumors highlight a need for tumor cell recognition by T cells for efficacy of CBT, a process that requires relevant antigens and antigen presentation machinery (26). A retrospective series found decreased or absent expression of β2M and/or MHC-I in 80% of patients with cHL and decreased or absent MHC class II in 70% of patients with cHL; β2M is the most frequently mutated gene in cHL (27). A retrospective analysis of 108 newly diagnosed patients with cHL treated with conventional chemotherapy plus modified involved field radiotherapy found that those with reduced or absent β2M or MHC class I expression on HRS cells had poor outcomes independent of 9p24.1 status (28). Loss of MHC-II expression on HRS cells is also found more commonly in patients with relapsed cHL compared with newly diagnosed patients (29). Although the relationship between β2M mutations and response to CBT has not yet been described in cHL, β2M mutations and loss of MHC-I in melanoma have been described in patients with progressive disease and resistance to PD-1 blockade (30). Identification of tumor antigens in cHL is complicated by the relative rarity of HRS cells in the tumor microenvironment and requires enrichment techniques such as laser-capture microdissection or cell sorting using flow cytometry (27). As such, associations between antigen-specific immune response against either shared or mutation-derived neoantigens and efficacy of PD-1 blockade are not known. Additional research is needed to better define mechanisms of resistance to PD-1 blockade in cHL to inform design of rational clinical trials aimed toward achieving durable remissions in a larger proportion of patients.

Preclinical data support a role for the PD-1/PD-L1 pathway in myeloma via expression of the PD-1 receptor on T and NK cells in patients with multiple myeloma and expression of PD-1 ligands on malignant plasma cells (PC; ref. 31). T cells have been shown to recognize abnormal PCs, as supported by detection of marrow-infiltrating T cells in monoclonal gammopathy of undetermined significance (MGUS) capable of mounting anti-PC immune responses, and presence of immunity against shared antigens is associated with prolonged progression to over-symptomatic multiple myeloma. However, once symptomatic multiple myeloma develops, marrow T-cell responses have not been observed without ex vivo expansion steps (32–34). The reasons for the loss of antigen-specific T-cell activity in vivo in multiple myeloma compared with precursor disease is not well understood but could be due to progressive immunosuppression by the tumor microenvironment during disease progression from MGUS to multiple myeloma in contrast to the proinflammatory milieu present in the cHL tumor microenvironment (Fig. 1). Perhaps the relative paucity of antigen-specific T cells is one reason that anti–PD-1 monotherapy using nivolumab had limited clinical activity (10). Interestingly, lenalidomide administration appeared to have transient efficacy immediately following nivolumab during a period where prolonged receptor occupancy of the PD-1 receptor was expected (35).

IMiDs (thalidomide, lenalidomide, and pomalidomide) enhance T-cell responsiveness to antigen-presenting cells (APC) and polarize T cells toward a Th1 phenotype, inhibit myeloid-derived suppressor cells (MDSC) and regulatory T cells (Treg), and downregulate PD-L1 on tumor cells (36–39). These observations suggested the hypothesis that IMiD and PD-1 blockade combinations could result in clinically relevant antimyeloma immune responses in R/R multiple myeloma (Table 2). The KEYNOTE-023 study evaluating pembrolizumab, lenalidomide, and dexamethasone demonstrated an ORR of 44% (n = 50), with a strict CR (sCR) of 4%, a very good partial response (VGPR) of 12%, and a PR of 28%. Lenalidomide-refractory patients responded to pembrolizumab plus lenalidomide and dexamethasone at a rate of 35%, with 5.4% achieving sCR, 8.1% reaching VGPR, and 21.6% achieving PR (40). A phase II study of pembrolizumab, pomalidomide, and dexamethasone demonstrated an ORR of 60% (29/48), with four patients (8%) reaching sCR/CR, nine (19%) reaching VGPR, and 16 (33%) reaching PR. Although limited by a small sample size, correlative analyses of pretreatment tissue biopsies demonstrated that presence of CD3⁺/PD-1⁺ marrow-infiltrating lymphocytes was associated with shorter progression-free survival (41). Patients expressing PD-L1 in the bone marrow before treatment had a trend toward a higher rate of responses of VGPR or better (41). An alternative hypothesis for the failure of PD-1 monotherapy in multiple myeloma proposes that clonal bone marrow T cells expressing PD-1 in multiple myeloma exhibit a telomere-independent senescence phenotype and are unable to respond to reinvigoration with immune checkpoint blockade (42). Additional biomarker studies are needed to better understand the association between response and PD-L1 expression in multiple myeloma marrow, and if PD-1⁺ T cells in multiple myeloma are senescent T cells or can be reinvigorated (43).

Encouraging clinical activity observed with IMiDs and anti–PD-1 combinations has spurred evaluation of agents capable of shifting the tumor microenvironment toward immune activation while inducing myeloma cell killing. In this regard, CD38 has emerged as an interesting target in multiple myeloma due to high levels of expression on PCs, a contribution to T-cell anergy through ectoenzyme function that leads to adenosine production and expression on inhibitory cell populations such as MDSCs and Tregs (44, 45). Targeting CD38 with daratumumab kills malignant PCs through traditional antibody-dependent cellular cytotoxic mechanisms. In responding patients, treatment with daratumumab also appears not only to deplete subpopulations of Tregs and MDSCs in the myeloma microenvironment but also to result in T-cell expansion and increased T-cell clonality suggestive of an immune mechanism of myeloma disease control (46). These observations have provided rationale for investigation of daratumumab in combination with PD-1/PD-L1 blockade with or without IMiDs (NCT01592370, NCT03000452, and NCT02431208).

Radiotherapy may also be an effective combination partner with PD-1 blockade by taking advantage of in situ vaccination caused by immunogenic cell death. Radiotherapy has been shown to result in epitope spreading and augmented antigen presentation by local APC. These effects have been associated with abscopal (distant) clinical effects in a variety of diseases (47–49). Temporal upregulation of PD-L1 in the irradiated tumor suggests intrinsic mechanisms that inhibit immune responses after radiotherapy, and provides rationale for blockade of PD-L1 in combination with radiotherapy (50) to overcome this mechanism. Several reported cases of systemic responses in patients with multiple myeloma and plasmacytomas irradiated while receiving anti–PD-1 suggest potential induction of abscopal effects (10, 41), which previously have been reported to occur spontaneously in very rare instances (51–53). We have recently begun enrollment of a combination trial using radiotherapy plus PD-1 pathway blockade in patients with solitary bone plasmacytoma and limited clonal bone marrow plasmacytosis (NCT03196401) with the aim to elicit systemic immunity and the abscopal effect.

Both auto-HSCT and allogeneic HSCT (allo-HSCT) are commonly used for treatment of patients with hematologic malignancies. In addition to antitumor responses produced by immunologic graft-versus-tumor (GVT) effects after allo-HSCT, immune responses by the donor immune system against nontumor host tissue can result in acute graft-versus-host disease (aGVHD) and chronic GVHD (cGVHD), leading to morbidity and treatment-related mortality. The normal function of immune checkpoints limits T-cell–mediated immune responses against host tissues. Relapse after allo-HSCT represents a significant clinical dilemma, and CBT is also being explored in this patient population. Preclinical studies examining PD-1 axis blockade after allo-HSCT demonstrated not only potentiation of GVT effects (54, 55) but also evidence supporting exacerbation of GVHD (56).

The feasibility of immune checkpoint inhibition for treatment of hematologic cancers relapsing after allo-HSCT was first explored using CTLA-4 blockade with ipilimumab in two studies with responses observed in both lymphoid and myeloid malignancies without high rates of treatment-emergent GVHD (57, 58). Several series further elaborate on efficacy and toxicity of PD-1 inhibitor use before or after allo-HSCT (Table 3). Based on early reports suggesting a toxicity signal of hyperacute, severe acute, and chronic GVHD, and four treatment-related deaths observed among 39 patients who received PD-1 blockade before allo-HSCT (59), a warning was added to the FDA package insert for nivolumab (60). The FDA recommends that patients receiving allo-HSCT after PD-1 blockade be closely monitored for early evidence of transplant-related complications, such as hyperacute GVHD, severe acute GVHD, steroid-requiring febrile syndrome (as a potential harbinger of severe acute GVHD), hepatic veno-occlusive disease, and other immune-mediated reactions.

For patients with relapsed cHL after allo-HSCT, limited treatment options have led to increased off-label usage of PD-1 inhibitors. These data suggest that patients can achieve objective responses to PD-1 blockade after allo-HSCT (ORR, 77%–95%), but this is complicated by a significant risk of developing severe treatment-emergent GVHD in up to 30% to 55% of patients. Among 53 total patients with publicly reported outcomes following receipt of PD-1 inhibitors after allo-HSCT, the observed rate of treatment-emergent GVHD was 47.2%, with 30.2% of treated patients developing grade 3 to 4 acute or severe chronic GVHD (61–63).

Given the potential risk involved with the use of PD-1 inhibitors before or after allogeneic stem cell transplant, this approach should be pursued only in the context of a clinical trial. PD-1 blockade is being formally explored in prospective studies as maintenance therapy after allo-HSCT (NCT02985554). Perhaps these studies will provide greater insight into predictors of GVH risk versus GVT benefit of this approach and define appropriate patient populations in which clinicians can safely harness the potential of PD-1 blockade to maintain a meaningful GVT response while minimizing the risk of developing treatment-emergent GVHD after CBT.

Auto-HSCT avoids the challenges of GVHD, but absence of GVT is thought to be a limitation to durability of responses. Nevertheless, the dynamics of immune reconstitution early after autologous stem cell transplant alters the immune-regulatory network to favor autologous GVT response that may be further augmented by immune checkpoint inhibition (64). For example, Treg populations decline as CD8⁺ T cells expand during early lymphocyte recovery after autologous stem cell transplant. Seeking to harness this potentially favorable immune phenotype, a trial testing autologous lymphocyte infusions and combined CTLA-4 and PD-1 pathway blockade in concert with auto-HSCT for multiple myeloma is ongoing (NCT02716805). Recent studies showing that T cells produced by the autograft are able to respond to APC and develop into antigen-specific cytotoxic T lymphocytes (CTL) as early as 12 days after auto-HSCT support vaccine strategies in this setting as well (64). Several ongoing studies aiming to improve durability of disease control following auto-HSCT via induction of multiple myeloma–directed immune responses include a dendritic cell (DC)–tumor cell fusion vaccine (NCT02728102), a WT1-directed vaccine (NCT01827137), and an RNA-electroporated DC vaccine (NCT01995708). Future combination trials incorporating vaccines with CBT in the postautologous transplant space are a logical extension of these studies.

Immune checkpoint blockade is well tolerated in many patients, but immune-mediated toxicities do develop. Three phase I studies in hematologic malignancy trials reported a drug-related grade 3 adverse event (AE) rate ranging from 18% to 20%, a small number of grade 4 AEs, and a single case of fatal pneumonitis (7, 10, 12).

The phase II studies of pembrolizumab and nivolumab in R/R cHL demonstrated acceptable safety profiles consistent with prior PD-1 inhibitor phase I studies. In the phase II study of nivolumab, 13 of 210 (5.4%) of patients had a treatment-related grade 3 AE, and there were no treatment-related grade 4 or 5 AEs reported. In the phase II study of pembrolizumab, 22 of 80 patients had grade 3 AEs by investigator assessment, two patients had grade 4 increased lipase, one patient developed grade 4 neutropenia, and there were no reported treatment-related deaths (9, 14).

In multiple myeloma, pembrolizumab plus pomalidomide and dexamethasone did not appear to result in additive toxicity greater than that seen in solid tumors (65). Six patients developed immune-mediated pneumonitis, the majority of which were grade 1 to 2 in severity, and only one patient developed grade 3 pneumonitis (41) despite pomalidomide's association with pneumonitis (66). Of note, a hold on accrual of subjects to the phase III KEYNOTE-183 and KEYNOTE-185 studies evaluating the additive benefit of pembrolizumab to lenalidomide and dexamethasone or pomalidomide and dexamethasone was instituted by Merck in June 2017 due to excess deaths in the pembrolizumab treatment arm. Further evaluation of this safety signal is pending. In our experience, early detection and treatment of immune-related AEs is critical, as the severity of these events seems to be inversely proportional to the time from onset of symptoms to treatment. As clinicians become accustomed to the patterns of toxicities seen with CBT, it is expected that the severity of toxicities should diminish.

Clinical successes with blockade of the PD-1 pathway in cHL have led to regulatory approvals and significant excitement among clinicians in evaluating the utility of these treatments earlier in disease natural history. Genetically driven increases in the 9p24.1 locus in HRS cells appear to have a clear association with depth of response, underscoring an intrinsic sensitivity to PD-1 blockade in cHL. However, the absence of antigen presentation machinery in most HRS cells highlights that additional study is needed to understand precise mechanisms of activity of PD-1 blockade in this disease. A broad range of combination trials currently ongoing will undoubtedly define how to best use PD-1 blockade within the landscape of cHL therapy over the coming years. It is hoped that further study of mechanisms of activity in cHL will enable tailoring of better patient selection for specific combination approaches and perhaps address emergent resistance. Beyond cHL, PD-1 blockade is active in several virally driven NHL subtypes and entities with 9p24.1 abnormalities; prospective clinical studies of immune checkpoint inhibitors are ongoing to follow up these observations. In subtypes of lymphoma with limited response to checkpoint blockade, development of reliable biomarkers to predict which subsets of patients might respond to these agents is needed.

In contrast, single-agent PD-1 pathway blockade in multiple myeloma was underwhelming. Fortunately, rationally designed combination trials with IMiDs in multiple myeloma have had encouraging results and opened the door to pivotal phase III trials whose results are eagerly awaited. Additional immunotherapeutic interventions in multiple myeloma, including monoclonal antibody therapy with daratumumab or elotuzumab, vaccine strategies, and highly encouraging early data from chimeric antigen receptor–modified T-cell therapies form unique opportunities to rapidly evaluate rational combination strategies.

Numerous additional questions remain on the use of immune CBT in these two distinct diseases. Can stem cell transplant, radiotherapy, and other chemotherapies routinely used in cHL and multiple myeloma combinations result in a favorable efficacy/safety profile? Will evaluating PD-L1 expression, T-cell clonality, or other biomarkers derived from studies in solid tumor malignancies have applicability in cHL, multiple myeloma, and other lymphoid malignancies? What is the role of antigen-specific immunity in these diseases in the context of checkpoint blockade, and will shared antigens or neoantigens emerge as potential predictors of activity? Are there additional immune checkpoints or agonists whose modulation will also be therapeutically effective for these diseases? The emerging paradigm has been to evaluate combinations on a PD-1 blockade backbone, but perhaps this approach will mask unique biology or augment toxicity of other immune modulatory pathways.

Partnership of immune checkpoint antibodies with other immune-based approaches, such as adoptive cellular immunotherapy such as chimeric antigen receptor–modified T cells or antibody engineering products such as bispecific T-cell engagers, might exhibit synergistic activity. Vaccine-based approaches aimed at stimulating antigen-specific immunity to shared tumor antigens or neoantigens potentially through DC-based platforms could also be rationally combined with immune checkpoint blockade to amplify antitumor immune responses.

Tumor immunotherapy originated more than 120 years ago by William Coley and his induction of inflammation by direct tumor inoculation of bacterial products at the turn of the 20th century. Years of basic science investigations since that time have delineated pathways of immune activation and regulation, and ultimately, have yielded the realization that negative regulators of immune activation are dominant pathways of cancer immune evasion. As such, checkpoint blockade has in effect reinvigorated the entire field of tumor immunotherapy.

A.J. Moskowitz reports receiving commercial research grants from Bristol-Myers Squibb and Seattle Genetics and is a consultant/advisory board member for Bristol-Myers Squibb, Merck, Seattle Genetics, and Takeda. A.M. Lesokhin reports receiving commercial research grants from Bristol-Myers Squibb and Serametrix Inc., speakers bureau honoraria from Aduro, Bristol-Myers Squibb, Janssen, and Juno, holds ownership interest (including patents) in Serametrix Inc., and is a consultant/advisory board member for Bristol-Myers Squibb. No potential conflicts of interest were disclosed by the other author.

M.J. Pianko is supported in part by the Memorial Sloan Kettering Cancer Center (MSKCC) Mortimer J. Lacher Fellowship supported by the Lymphoma Foundation and the MSK Sawiris Foundation, and by a grant from the NIH/National Center for Advancing Translational Sciences (UL1TR00457), administered by the Clinical and Translational Science Center at Weill Cornell Medical Center and MSKCC. A.M. Lesokhin is a member of the Parker Institute for Cancer Immunotherapy, which supported the MSKCC Cancer Immunotherapy Program. He also received support from the MSK Sawiris Foundation and Cycle For Survival. This work was also supported in part by the Memorial Sloan Kettering MSKCC NCI core grant (P30 CA008748; to A.M. Lesokhin, principal investigator: Craig B. Thompson).