Hodgkin lymphoma
Among various hematological malignancies, therapeutic benefits of PD-1 blockade have been best demonstrated in patients with Hodgkin lymphoma (HL). In an early clinical trial of 23 relapsed or refractory HL patients, nivolumab showed an objective response rate (ORR) of 87%, with a complete response (CR) of 17% [
46]. Clinical efficacy and safety profile of nivolumab monotherapy was further demonstrated in the CheckMate 205 trial of 80 HL patients who failed autologous stem-cell transplantation and brentuximab vedotin (an antibody-drug conjugate comprising anti-CD30 mAb conjugated to an anti-microtubule agent) [
47,
48,
61]. Another PD-1 mAb, pembrolizumab, also showed similar efficacy against HL in clinical trials such as KEYNOTE-013 [
49] and KEYNOTE-087 [
50]. These results led to the FDA approval of nivolumab (in 2016) and pembrolizumab (in 2017) for relapsed or refractory HL patients who had failed multiple lines of therapy.
Clinical responsiveness to PD-1 blockade in HL has been explained by multiple unique tumor intrinsic and extrinsic factors. In solid malignancies, high TMB (high TMB; ≥ 20 coding mutations per megabase) has been recognized as an independent predictor for clinical responses to ICB therapies [
62]. However, in HL, it is reported that TMB does not correlate with responsiveness [
63]. Indeed, Liang et al. investigated TMB in 34 HL patients and found that only 15% of the patients had high TMB [
64]. Despite this fact, amplification of 9p24.1 (locus-containing JAK2/PDL1/PDL2), a frequently observed cytogenetic abnormality in classical HL patients, induces aberrant overexpression of PD-L1 on tumor cells [
65], suggesting that tumor PD-L1 might dampen anti-tumor immunity. Indeed, Epstein-Barr virus (EBV) infection can also contribute to PD-L1 upregulation [
66]. While similar frequencies of 9p24.1 amplification in EBV-related or EBV-unrelated cases have been reported in cHL, EBV-positive cHL cases have been shown to be more likely to have higher PD-L1 expression levels [
67]. Additionally, through viral oncoprotein latent membrane protein 1 (LMP1), EBV may also sustain an immune suppressive microenvironment [
68]. Thus, it is possible that the presence of these immune suppressive features in EBV-related cHL may allow patients with this particular sub-type to be more susceptible to ICB.
In addition to the tumor-intrinsic overexpression of PD-L1, the immune microenvironment of HL is critically responsible for responsiveness to PD-1 blockade. The TME of classical HL consists of rare (0.1–1%) malignant cells called Hodgkin Reed-Sternberg (HRS) cells and an abundant immune cell infiltrate which is markedly distinct from the TME observed in non-Hodgkin’s lymphoma (NHL) [
69]. Thus, the unique immunologically “hot” (or inflamed) TME critically contributes to responsiveness to PD-1 blockade [
70]. To understand the immune landscape, Cader et al. recently performed mass cytometry of 7 newly diagnosed classical HL patients [
71]. Notably, they found that CD4
+ T cells especially Th1-polaralized effector cells and Tregs were major T cell subsets in the TME and that MHC-I expression was frequently lost in tumor cells [
71]. This result raises the possibility that CD4
+ T cells, rather than CD8
+ T cells, may be key players in anti-tumor immunity against HL. Indeed, Roemer et al. also reported that β2-microglobulin (β2M) and MHC-I were not expressed on tumor cells in more than 60% of HL patients and that the expression level of MHC-II and PD-L1 predicts therapeutic response to nivolumab in HL patients [
72]. More recently, Patel et al. investigated checkpoint molecules by multiplexed immunofluorescence imaging and observed the expansion of CTLA-4
+ PD-1
− CD4 T cells in close proximity to CD86
+ tumor cells or tumor-associated macrophages (TAMs) [
73]. This suggests that the interaction between CTLA-4 and CD86 might also act as a key negative regulator in HL.
Overall, these results highlight the importance of characterizing the disease-specific TME and T cell phenotypes in order to determine optimal therapeutic targets. Despite the clinical benefits of PD-1 blockade, some HL patients experience recurrence after PD-1 blockade therapy. Currently, the randomized phase 2 study of brentuximab vedotin and nivolumab with or without ipilimumab is ongoing (NCT01896999) to assess the efficacy of dual CTLA-4 and PD-1 blockade. LAG-3 might be a potential target considering the recent single-cell RNA sequence (scRNA-seq) analysis which showed that a subset of HL-associated T cells expressed high levels of this checkpoint molecule [
74]. Alternatively, given that loss of MHC-II confers therapeutic resistance to HL, harnessing innate immunity might be a key approach. The PD-1 and PD-L1 interaction is implicated in negative regulation of NK cells and monocytes/macrophages [
75]. As discussed later, several innate immune checkpoint inhibitors are being developed. The combination of innate checkpoint inhibitors may improve clinical responses against HL with acquired resistance to T cell-mediated anti-tumor immunity.
Non-Hodgkin lymphoma
In contrast to HL patients, PD-1 blockade has not shown remarkable clinical responses in patients with NHL such as diffuse large B cell lymphoma (DLBCL) and follicular lymphoma (FL). In a phase 2 clinical trial of nivolumab in relapsed or refractory DLBCL, ORR of monotherapy was 3% and 10% in transplant-ineligible patients and relapsed patients after autologous SCT, respectively [
51]. The efficacy of nivolumab and ibrutinib (a Bruton’s tyrosine kinase inhibitor) combination has been evaluated in relapsed or refractory NHL (DLBCL and FL) and chronic lymphocytic leukemia patients. However, overall response of the combination was comparable to that of ibrutinib monotherapy [
76], suggesting that PD-1 blockade had limited contribution to disease control. Indeed, PD-1 blockade in combination with rituximab has been shown to be effective in rituximab refractory FL [
52]. However, ICB can also be utilized to prime the immune system prior to such tumor-targeted therapies. In fact, this concept was trialed in a phase 2 study in treatment of naïve FL patients using a combination of nivolumab and rituximab [
53]. Interim analyses indicate an ORR of 84% with 47% achieving CR, suggesting a favorable toxicity profile along with high overall and complete response rates [
53].
Differential responsiveness to PD-1 blockade between DLBCL patients and HL patients can be primarily explained by their different levels of immune infiltration. Indeed, the transcriptional and histological comparison of the immune microenvironment between HL and DLBCL showed an immunologically “cold” (or “non-inflamed”) TME in DLBCL in contrast to the “hot” (or “inflamed”) TME in HL [
77]. Of note, in DLBCL patients treated with rituximab-based standard chemo-immunotherapy (R-CHOP), high infiltration of T cells predicts better prognosis, whereas the presence of PD-1
high T cells predicts poor prognosis [
78,
79] indicating that anti-lymphoma immunity still plays an indispensable role for disease control against immunologically cold DLBCL.
The immunologically “cold” TME is created by tumor-intrinsic factors such as high proliferation rate of lymphoma cells and oncogene-driven immune exclusion [
70]. Various factors are reported to be associated with immunologically “cold” TME such as double rearrangements of
MYC and
BCL2 and/or
BCL6 [
80], deletion or mutation of
PTEN [
81], and Epstein-Barr virus-related subtypes [
82]. Above all, several lines of evidence suggest that
EZH2 mutations critically contribute to immune exclusion phenotypes. It is reported that MHC-I and MHC-II are lost in 40–60% and 20–40% of DLBCL patients, respectively, and
EZH2 mutations are highly enriched in MHC-deficient subsets of patients [
83]. The EZH2-mediated silencing of genes related to MHC-I expression is implicated in immune evasion in a wide range of other malignancies, as Burr et al. also demonstrated this immune evasion mechanism by a whole genome CRISPR/Cas9 screen [
84]. Thus, inhibition of EZH2 activity by a small molecule inhibitor might be a potential approach to sensitize tumor cells to T cell-mediated anti-tumor immunity.
As seen in HL, EBV has also been linked to a number of malignant NHLs [
85]. In a recent study, Kim et al. reported that more patients with EBV-positive NHL responded to pembrolizumab than EBV-negative subtypes [
86]. Moreover, high PD-L1 expression was reported in EBV-positive NHL as compared to EBV-negative NHL [
86]. In agreement with this, Kataoka et al. recently reported a high frequency of PD-L1/PD-L2 containing somatic aberrations in various EBV-positive lymphoma subtypes [
87]. In addition, a distinct pattern of somatic alterations in EBV-positive DLBCL was reported where the genetic profile was found to be distinct from EBV-negative DLBCL [
87]. While ICB in treating virus-associated cancers appears promising, more clinical trials are needed to verify if viral infections such as EBV can effectively predict ICB efficacy.
Another important tumor-intrinsic factor is gene alterations (amplifications or translocations) of
PD-L1 that are observed in ~ 25% of DLBCL patients (especially the non-germinal center type) [
88]. This DLBCL subset with aberrantly overexpressed PD-L1 is characterized by high infiltration of clonal T cells and low expression of tumor MHC-I. While patients with this type of DLBCL show inferior progression-free survival following front-line chemo-immunotherapy, they show, strikingly, good responsiveness to PD-1 blockade [
88]. Thus,
PD-L1 alterations may be a useful biomarker in predicting responsiveness to PD-1 blockade in DLBCL patients. Still, it remains unknown how we can overcome therapeutic resistance in patients with the immunologically “cold” DLBCL (low T cell infiltration and low MHC expression on tumor cells). Blockade of innate checkpoint molecules in combination with rituximab might be a possible approach to augment NK cell-mediated antibody-dependent cellular cytotoxicity (ADCC) and macrophage-mediated antibody-dependent cellular phagocytosis (ADCP).
NK/T cell lymphomas have a distinct immunophenotype, and PD-1 blockade has shown some efficacy in relapsed or refractory NK/T cell lymphomas [
89]. In a recent retrospective study of seven patients with relapsed or refractory NK/T cell lymphomas treated with pembrolizumab at diagnosis, the ORR was reported to be 57.1% with a complete response occurring in 2 patients [
90]. Large scale trials are warranted to further assess long-term clinical responses. Notably, modest responses to single agent PD-1 blockade have been observed in peripheral T cell lymphomas (PTCL) [
91]. Indeed, mutations in epigenetic modifier genes are often observed in PTCL, and such mutations may promote immune escape. To this end, a recent phase I/II trial combined pembrolizumab and a histone deacetylase inhibitor in relapsed or refractory PTCL patients [
54]. From the 15 evaluable patients, 3 were complete responders that remained in remission for at least 10 months [
54]. These early findings are encouraging, and therefore, further effort in this direction should be pursued. Furthermore, some responses to ICB have also been reported in patients with common subtypes of cutaneous T cell lymphomas. PD-1 blockade led to an ORR of 38% in patients with mycosis fungoides or Sézary syndrome in a phase 2 trial demonstrating a favorable safety profile and modest anti-tumor activity [
55]. Notably, some flare reactions such as worsening of erythema were reported in patients with Sézary syndrome, and such skin flare reactions were found to be associated with high PD-1 expression on circulating Sezary cells before therapy [
55]. Although high PD-1 expression may be an effective predictor of this reaction, further validation is warranted. On the contrary, PD-1 blockade led to rapid progression in patients with adult T cell leukemia/lymphoma (ATLL) in a recent study [
92]. Rapid clonal expansion of malignant T cells was observed in patients treated with PD-1 blockade, suggesting a tumor-intrinsic regulatory role of PD-1 in ATLL [
93]. This is in line with a previous report from a T cell lymphoma mouse model where PD-1 activity enhanced PTEN levels, and PD-1 deletion after an oncogenic insult resulted in aggressive lymphomas in vivo [
94]
. These data highlight the growing need to mechanistically understand the underlying mechanisms of ICB as we attempt to extend their promise to an ever-increasing list of malignancies.
Multiple myeloma
Multiple myeloma (MM) remains an incurable malignancy, and new immunotherapeutic approaches are actively being tested. In relapsed and refractory MM patients, an early clinical trial supported good efficacy (ORR 44%) with an acceptable safety profile of pembrolizumab, lenalidomide (an immunomodulatory imide drug: IMiD), and low dose dexamethasone [
56]. However, the combination of pembrolizumab and IMiD showed unfavorable benefit-risk profile in following phase III trials in relapsed and refractory MM (KEYNOTE-183) and newly diagnosed MM (KEYNOTE-185) [
57,
58].
It remains poorly understood why PD-1 blockade failed to show clinical benefits in MM patients. As MM predominantly grows in the bone marrow (BM), the unique TME in the MM BM might be implicated in therapeutic resistance [
95]. One of the key factors is an altered T cell phenotype in the MM BM. In general, T cell subsets with a senescent phenotype represent dysfunctional T cells in the TME and often result in hypo-responsiveness to ICB therapy [
96]. A recent scRNA-seq analysis of the MM immune microenvironment showed that a senescent T cell subset emerges in a premalignant state (i.e., smoldering MM) [
97], indicating that symptomatic MM patients already have dysfunctional T cells at diagnosis. Moreover, dysfunctional T cells are increased in patients who experience relapse after autologous SCT [
98] indicating that there is difficulty in reinvigorating T cells in relapsed and refractory MM patients. Another key barrier for anti-tumor immunity is immunosuppressive subsets in the TME. As myeloid cells and their progenitors abundantly exist in the BM, MM progression triggers conversion of these cells into myeloid-derived suppressor cells (MDSCs). Indeed, the transcriptional landscape in MM patients revealed an inverse correlation between MDSC-related genes and cytotoxic lymphocyte-related genes suggesting that MDSCs might contribute to T cell exclusion in the MM BM [
99]. MM-associated Tregs are important targetable immunosuppressive cells since CD38
+ Tregs can be depleted by daratumumab (anti-CD38 mAb) [
100]. However, a phase 1 clinical trial of daratumumab in combination with anti-PD-1 mAb (JNJ-63723283) in relapsed and refractory MM (MMY2036, NCT03357952) was terminated due to limited clinical responses with increased adverse events in the combination group compared to daratumumab monotherapy.
While PD-1 blockade has not demonstrated clinical benefits in MM patients, it is noteworthy that some patients achieved long-term remissions after stopping pembrolizumab in clinical trials [
101]. Thus, further investigations are necessary to predict responder patients. Alternatively, T cell immunoreceptor with Ig and ITIM domains (TIGIT) has emerged as an important immune checkpoint on MM T cells [
102]. The inhibitory mechanisms of this new immune checkpoint will be discussed later in this review.
Myeloid malignancies
Early phase 1 trials showed lack of efficacy of anti-PD-1 mAb as monotherapy in patients with acute myeloid leukemia (AML) [
103] or high-risk myelodysplastic syndrome (MDS) [
104]. Similarly, CTLA-4 blockade using ipilimumab failed to show clinical benefits in high-risk MDS patients [
105].
Although these early trial results are disappointing, ipilimumab treatment after allogeneic-hematopoietic stem cell transplantation (HSCT) showed good responses in 22 patients with various hematological malignancies including 12 AML patients (CR 23% in 4 patients with extramedullary AML and 1 with MDS developing into AML) [
59], suggesting that ipilimumab can augment graft-versus-leukemia (GvL) effects. Of course, the key concern remains to be immune-related adverse events (irAEs), especially graft-versus-host disease (GVHD). However, among 22 patients, 3 patients in total developed GVHD (1 patient with grade II acute GVHD of the gut and 2 patients with chronic GVHD of the liver) which was able to be controlled by glucocorticoids [
59]. PD-1 blockade after allogeneic HSCT is also being tested, but so far, it has shown modest clinical benefits [
106].
Another potential approach to improve ICB efficacy is using a hypomethylating agent (HMA) such as azacytidine. Azacytidine is clinically approved for MDS and AML and has shown immunomodulatory effects including upregulation of MHC-I. In a recent trial, the combination of nivolumab and azacytidine showed an ORR of 33% and CR rate of 22% in 70 patients with relapse/refractory AML [
60]. Intriguingly, while high infiltration of effector T cells before therapy predicted good responsiveness to the combination therapy, an increase in CTLA-4
+ effector lymphocytes after treatment predicted non-responder patients [
60]. The combination of HMAs and PD-1 blockade are currently being tested in several trials [
106], but co-blockade of PD-1 and CTLA-4 may potentially overcome therapeutic resistance.
Immune dysregulation in AML remains incompletely understood as it might be differentially organized and depends on multiple factors such as tumor subtypes, presence or absence of the MDS state, patients’ age, and treatment history. Kong et al. originally showed high TIGIT expression on CD8
+ T cells in AML patients, and that siRNA mediated TIGIT knockdown can reinvigorate cytokine production suggesting that TIGIT is expressed on dysfunctional T cells [
107]. Another independent group also showed that PD-1
+ TIGIT
+ CD8
+ T cells with low expression levels of DNAM-1 (DNAX accessory molecule 1 or CD226) represent dysfunctional T cells [
108], further providing evidence that TIGIT represents a key checkpoint molecule in AML. In terms of mechanisms of immune escape after HSCT, Christopher et al. recently performed exome sequencing on paired samples obtained at diagnosis and at post-HSCT relapse demonstrating that downregulation of MHC II-related genes is a key feature of relapsed AML [
109]. This result suggests that CD4
+ T cells might play a critical role for controlling relapse. Thus, it might be possible to augment CD4-mediated GvL effects by blockade of LAG-3-MHC-II interaction. Another immune checkpoint TIM-3 is also expressed on AML T cells [
110], but tumor TIM-3 has a unique tumor-intrinsic role, as discussed later. Thus, TIGIT, LAG-3, and TIM-3 are differentially implicated in AML immune-regulation and are rational therapeutic targets in combination with PD-1 blockade and/or CTLA-4 blockade.
Compared to MDS and AML, therapeutic benefits of ICB may be better achieved in myeloproliferative neoplasms (MPNs), a heterogeneous group of diseases characterized by clonal expansion of myeloid cells. Significant advances have been made to control disease progression by targeting JAK2, one of the key driver mutations in MPNs. However, it is still challenging to treat high-risk patients who are refractory to JAK1/JAK2 inhibitors [
111]. Prestipino et al. showed that JAK2 mutations can trigger the overexpression of PD-L1 on myeloid cells, leading to metabolic and functional impairment of T cells [
112]. Importantly, PD-1 blockade improved survival in human MPN xenograft and primary murine MPN models [
112] suggesting potential roles for immunotherapy in MPNs. Another notable aspect is immunogenicity of MPNs. The JAK2V617F mutation is seen in > 50% of patients with MPNs, and this mutation-associated neoantigen can elicit tumor-specific CD8
+ T cell responses [
113]. Somatic frameshift mutation of the calreticulin gene (
CALR) is another MPN-restricted key driver mutation found in 67% and 88% of patients with essential thrombocythemia and primary myelofibrosis respectively [
114]. These
CALR mutations induce a shared MPN-specific neoantigen leading to the generation of mutant calreticulin–specific T cell responses [
115,
116]. Thus, due to the unique immunogenicity and induction of neoantigen-specific T cell responses, ICB might play an important role for disease control in MPNs. However, further clinical investigations are warranted.