Checkpoint inhibitors in hematological malignancies

For proper T cell activation, two separate signals are required (Fig. 1) [8]. The first signal is mediated by antigen-dependent T cell receptor (TCR) binding to the major histocompatibility complex (MHC) molecule of an antigen-presenting cell (APC). The second signal is antigen-independent, co-stimulatory, or co-inhibitory signal delivered by the APCs. The second signal modulates TCR signaling and determines the T cell’s fate. Several co-stimulatory or co-inhibitory molecules on T cells with their respective ligands are collectively known as B7-CD28 family. The prototypical co-stimulatory molecule is CD28 on resting naïve T cells, which induces cell-cycle progression, interleukin-2 (IL-2) production, and clonal expansion is constitutively expressed in resting naïve T cells [9]. Without co-stimulatory second signals, T cells fall into anergy. On the other hand, cytotoxic T-lymphocyte antigen-4 (CTLA-4) is a co-inhibitory receptor on T cells that induces T cell tolerance [10]. Additional second signal molecules include programmed death-1 (PD-1), lymphocyte activation gene-3 (LAG-3, CD223), T cell immunoglobulin and mucin domain-containing protein-3 (TIM-3), T cell immunoreceptor with immunoglobulin and ITIM domains (TIGIT), or B- and T-lymphocyte attenuator (BTLA). In this review, we discuss biology of CTLA-4, PD-1, LAG-3, and TIM-3, dysregulation of these molecules in hematologic malignancies, clinical trials, and biomarkers.

CTLA-4 (also known as CD152) was first discovered by Brunet et al. (Fig. 2) [10]. It is a protein encoded by the 4-exon CTLA4 gene on chromosome 2q33.2. It belongs to the immunoglobulin superfamily, with a single immunoglobulin V-like domain containing ligand binding sites [10, 11]. It consists of 223 amino acids, and with a calculated molecular weight of 24.6 kDa. CTLA-4 mainly resides in the cytoplasm in naïve resting T cells, but its expression on the surface of T cells can be detected within 1 or 2 days after activation [12]. On the other hand, rapid induction of CTLA-4 expression is seen in memory T cells upon activation, and its expression lasts longer compared with naïve resting T cells [13]. In regulatory T cells, CTLA-4 is constitutively expressed [14].

Although their functions are opposite, CLTA-4 and CD28 share the same ligand, B7-1 and B7-2. They share the MYPPPY motif for ligand binding [15]. Of note, CTLA-4 expression is 30- to 50-fold less than that of CD28 even in its maximum state upon activation. However, the affinity and avidity for CTLA-4 and its ligands are much greater than CD28 because the former homodimerizes and can bind to B7 molecules bivalently [16]. Upon activation by ligand binding, CTLA-4 molecules migrate from the cytoplasm to the cell surface, and this migration is dependent on the strength of T cell receptor signaling and phosphorylation of the Y¹⁶⁵VKM motif in the cytoplasmic domain of CTLA-4 [17‐20]. Furthermore, redistribution of CTLA-4 to the immunological synapse was shown to be highly dependent on B7-1, but only slightly dependent on B7-2 [21].

T cell inactivation by CTLA-4 can be explained by two mechanisms. Once redistribution of CTLA-4 to the proximity of immunological synapse occurs, it can sequester B7-1/B7-2 owing to its higher avidity and affinity so that the CD28-mediated co-stimulatory signal would be reduced (competitive antagonism) [22]. The second mechanism is for CTLA-4 to deliver an inhibitory signal via the cytoplasmic tail. Although the precise mechanism is not unequivocally determined, CTLA-4 signal inhibits nuclear accumulation of activator protein 1 (AP-1), NF-κB, and nuclear factor of activated T cells (NFAT) in activated T cells [23, 24]. Furthermore, CTLA-4 halts cell cycle progression by direct inhibition of cyclin-dependent kinase 4 (CDK4), CDK6, and cyclin D3 [25]. CTLA-4 also selectively inactivates microtubule-associated protein kinase (MAPK), extracellular signal-regulated kinase-1 (ERK), and c-Jun NH2-terminal kinase (JNK), which are required for stimulation of IL-2 production [26].

The cytoplasmic tail of CTLA-4 does not contain an immune receptor tyrosine-based inhibitory motif (ITIM) and does not have intrinsic enzymatic activity. Instead, CTLA-4 inhibitory effects (phosphatase activity) are thought to be mediated with other molecules including serine/threonine phosphatase PP2A and/or Src homology 2 domain-containing phosphatases (SHPs). PP2A is bound to newly synthesized CTLA-4 molecules and makes CTLA-4 inactive [27]. Upon ligand binding in the vicinity of TCR, the scaffolding subunit of PP2A is phosphorylated and PP2A is dissociated from CTLA-4. The dissociated PP2A inhibits the phosphatidylinositol 3-kinase (PI3K)/Akt pathway via directly inactivating protein kinase B/Akt [28]. In addition, Guntermann and Alexander demonstrated that the majority of phosphatase activity of CTLA-4 was attributed to SHP-1 [29]. Because CTLA-4 lacks ITIM, which is a direct binding site of SHP-1, it is thought that adapter proteins might be needed for interaction between CTLA-4 cytoplasmic domains and SHP-1.

PD-1 (also known as CD279) was first discovered by Ishida et al. from Tasuku Honjo’s group in 1992 in search of a gene inducing apoptosis [30]. PD-1 is a transmembrane protein with 288 amino acids and is encoded by PDCD1 gene on chromosome 2q37.3. PD-1 contains a single immunoglobulin V-like domain, a transmembrane domain, and an intracellular domain. The intracellular domain has an ITIM (S/I/V/LxYxxI/V/L) and an immunoreceptor tyrosine-based switch motif (ITSM; TxYxxV/I) [31, 32]. Expression of PD-1 is present in effector T cells, regulatory T cells (Treg), naïve and activated B cells, natural killer cells, myeloid dendritic cells, and monocytes with low intensity. In resting T cells, PD-1 expression is not present, but it can be induced within 24 h of T cell activation [33].

Programmed death-ligand 1(PD-L1) and programmed death-ligand 2 (PD-L2) are ligands for PD-1. Search of ligand for PD-1 was actively sought by Dr. Honjo’s group, but PD-L1 (also known as CD274 or B7-H1) was independently discovered by Dong et al. in Lieping Chen’s group in 1999 [34]. Dr. Chen’s group identified a molecule named B7-H1, but was not aware it was actually a ligand to PD-1. 1 year later, Gordon Freeman’s group, in collaboration with Honjo’s group, discovered a ligand for PD-1 (PD-L1) and demonstrated that PD-L1 is identical to B7-H1 [35]. PD-L1 is encoded by CD274 gene on chromosome 9p24.1. In non-pathologic lymphoid tissue, PD-L1 expression is observed in follicular T cells, macrophages, and a subset of dendritic cells. PD-L1 is also seen in placental syncytiotrophoblasts and dendritic cells/monocytes in the lung and liver [33, 34, 36]. By collaborative study in Honjo, Freeman, and Arlene Sharpe’s group, PD-L2 (also known as CD273 or B7-DC) was identified in 2001 [37]. In the same year, Tseng et al. in Drew Pardoll’s group independently discovered PD-L2 [38]. PD-L2 is encoded by PDCD1LG2 gene on chromosome 9p24.1. Of note, CD274 gene and PDCD1LG2 gene are 42 kB apart from each other. Compared with PD-L1, PD-L2 expression is more restricted. It is only seen in activated CD4⁺ or CD8⁺ T cell subsets, myeloid dendritic cells, monocytes, endothelial cells, and placental syncytiotrophoblasts [39]. Expression of PD-L1 and PD-L2 can be induced by interferon gamma (IFN-γ), granulocyte macrophage colony-stimulating factor (GM-CSF), and IL-4 [37, 40‐42].

PD-1 negatively regulates IL-2 production and T cell proliferation [43, 44]. Upon ligand binding, ITIM and ITSM in the cytoplasmic domain of PD-1 are phosphorylated by the Src-family tyrosine kinases and SHPs are further recruited to the phosphorylated tyrosine residue. SHPs dephosphorylate downstream signal pathways including PI3K/Akt or RAS/MEK/ERK pathway, blocking cell cycle progression [28, 45, 46]. SHPs also inactivate zeta-chain-associated protein kinase 70 (ZAP70) and protein kinase C-θ (PKC-θ), essential for T cell activation and IL-2 production, respectively [47, 48]. However, PD-1-mediated inhibitory signals can be overcome by strong T cell stimulation with CD28 or exogenous IL-2 [49].

Inhibitory function is not the only role of PD-1 pathway. Francisco and colleagues demonstrated that PD-L1 converts naïve CD4⁺/forkhead box P3 (FOXP3)- T cells to CD4⁺/FOXP3⁺ regulatory T cells (Tregs) in vitro, even without transforming growth factor beta (TGF-β) which is a well-established stimulator of Treg induction [50]. Induction and maintenance of Tregs by PD-L1 was also shown by the same group in vivo. Although this is not an inhibitory function in cellular level, PD-1 pathway enhances immune suppression by inducing immunosuppressive Tregs in the level of organism.

While CTLA-4 and PD-1 both deliver the co-inhibitory second signal, they execute their roles at different time points in the life cycle of immune response [51, 52]. CTLA-4 functions early in the life cycle of immune response during T cell priming in lymphoid organs (central checkpoint) and affects the global impact on the immune system. CTLA-4:B7-1/B7-2 interaction diminishes CD4⁺ T effector cells, increases CD4⁺ T-helper cells and enhances immunosuppressive activity of regulatory T cells, resulting in peripheral T-cell tolerance or anergy [53]. CTLA-4-deficient mice developed fatal lymphoproliferation and multiorgan autoimmunity [54, 55]. On the other hand, PD-1 plays a role in T-cell activation in peripheral tissue containing target cells (peripheral checkpoint). PD-1:PD-L1/PD-L2 interaction attenuates TCR signaling in T cells, inducing T cell exhaustion. PD-1-deficient mice developed lupus-like autoimmune disease inflammation [56, 57].

These checkpoints in immune response are often exploited in many cancers including hematologic malignancies [58, 59]. The concept that CTLA-4 blockade can be used to enhance anti-tumor activity was first shown by Leach and colleagues [60]. The role of PD-1 pathway in tumor immunity was independently shown by Dong and Iwai, promoting PD-1 blockade in cancer therapy [40, 61].

Following clinical success of targeting CTLA-4 and PD-1, other co-inhibitory molecules receive more attention; LAG-3, and TIM-3. The lymphocyte activation gene-3 (LAG-3, CD223) was discovered by Triebel and colleagues in 1990 [62]. It is encoded by 8-exon LAG3 gene, located at 12p13.31. LAG-3 has 498 amino acids and has structural similarity to CD4, containing one immunoglobulin-like V-type domain and three immunoglobulin-like C2-type domains. The intracellular domain of LAG-3 contains a unique KIEELE motif, which is essential for T cell modulation by LAG-3 [63]. Expression of LAG-3 is present in activated T cells, NK cells, activated B cells, and plasmacytoid dendritic cells [62, 64, 65]. The major ligands of LAG-3 are class II MHC molecule on APCs and liver and lymph node sinusoidal endothelial cell C-type lectin (LSECtin) on tumor cells or hepatocytes [66]. LAG-3 is a negative regulator in CD4 and CD8 T cell expansion in vitro as well as in vivo [67]. However, precise mechanisms remain to be elucidated. Co-expression of LAG-3 and PD-1 has been seen in tumor infiltrating lymphocytes (TILs) in tumor mouse models as well as human tissue, suggesting its role similar to PD-1 [68‐70]. Inhibition of both PD-1 and LAG-3 showed augmented anti-tumor activity of CD8⁺ T cells compared to targeting either of them [68, 70].

T cell immunoglobulin and mucin domain-containing protein-3 (TIM-3) was discovered by Monney and colleagues in 2002 [71]. TIM-3 is also called hepatitis A virus cellular receptor 2 (HAVCR2) and is encoded by HAVCR2 gene. HAVCR2 is located at 5q33.3 and consists of seven exons. TIM-3 is a transmembrane protein, containing signal peptide sequence, immunoglobulin-like V-type domain, mucin domain, and cytoplasmic tail [71]. TIM-3 expression is present in cytotoxic T cells, T helper 1 cells, regulatory T cells, NK cells, monocytes, and dendritic cells. Ligands of TIM-3 are many, including galectin-9, high mobility group protein B1 (HMGB1), and phosphatidyl serine [72, 73]. In the absence of ligands, BAT3 (HLA-B associated transcript 3) is bound to tyrosine residues in the cytoplasmic domain, forming a complex with TIM-3. Upon bindings to ligands, BAT3 is dissociated from the tyrosine residues and FYN, which can induce T cell anergy, could replace them [74, 75]. Similar to LAG-3, co-expression of TIM-3 and PD-1 was observed in CD8⁺ TILs [76, 77]. Interestingly, TILs with PD-1-/TIM-3- and showed the most severe dysfunction, compared to TILs with PD-1⁺/TIM-3- (weak dysfunction) or TILs with PD-1⁺/TIM-3⁺ (good function) [76, 78]. Although tyrosine residues in the cytoplasmic domain of TIM-3 are thought to cooperate with downstream signaling pathways, precise mechanisms are yet to be determined. Targeting TIM-3 showed significant anti-tumor activity in tumor mouse models [79]. Inhibition of both PD-1 and TIM-3 also demonstrated enhanced anti-tumor activity of CD8⁺ TILs [76].

CTLA4 gene overexpression was observed in bone marrow sample from patients with PCM, suggesting additional immune-evasive signals are mediated with CTLA-4 in T cells [110]. A recent study showed that low expression of PD-1, CTLA-4, LAG-3, and TIM-3 is present on T cell clones in bone marrow and peripheral blood samples of myeloma patients, suggesting the T cells are not exhausted [111].

PD-L1 expression in myeloma cells and myeloma-propagating pre-plasma cells detected by flow cytometry has been reported in several studies [112‐115]. Similar to CHL, increases in copy number of PD-L1 correlates with PD-L1 protein expression in myeloma cells [112]. However, one study demonstrated that there was no difference regarding PD-L1 expression between normal plasma cells from healthy donors and malignant plasma cells from patients with newly diagnosed monoclonal gammopathy of undetermined significance (MGUS) or PCM [116]. Having said that, available data supports that PD-1 pathway is implicated in development of plasma cell myeloma. Bone marrow myeloma burden and serum lactate dehydrogenase level was higher in patients with PD-L1 expression in myeloma cells compared to patients without PD-L1 expression [113]. PD-L1-expressing myeloma cells are resistant to melphalan [117]. High serum soluble PD-L1 was associated with worse progression-free survival (PFS) [118]. PD-L1 expression is higher in patients with relapsed refractory plasma cell myeloma [113]. An in vitro study demonstrated that myeloma cells with expression of PD-L1 could produce exhausted T cells (CD8^+/PD1^+/TIM-3^+), instead of functional cytotoxic T cells [119]. Furthermore, co-culture of primary myeloma cells with CD4⁺/CD25−/FOXP3− T cells induced increased amount of inducible Tregs (CD4⁺/CD25^+/FOXP3⁺) [120]. The tumor microenvironment (TME) of plasma cell myeloma is conducive to immune evasion. PD-1 overexpression was observed in T cells in patients with newly diagnosed PCM and relapsed refractory PCM [112, 116, 121]. PD-L1 expression can be induced in myeloma cells when cultivated with autologous stromal cells or human stromal cell line (HS-5) [113]. Interestingly, PD-1 expression in T cells was normalized after stem cell transplant. Additionally, PD-L1 expression is present in plasmacytoid dendritic cells or myeloid-derived suppressor cells in the TME of patients with PCM [122, 123]. PD-1 blockade showed improved survival in a myeloma murine model [119]. Unlike PD-L1, PD-L2 expression is not present in myeloma cells [112].

CLTA-4 plays a role in immune escape of AML. Using a murine myelogenous leukemia cell line (C1498) with expression of either CD80 or CD86, LaBelle et al. found that progressive tumor growth of C1498/CD80, but complete regression of C1498/CD86 after in vivo injection in naïve mice. They demonstrated that immune escape of C1498/CD80 is dependent on CTLA-4 [124]. A mouse model of relapsed AML study demonstrated that CTLA-4 blockade enhanced CTL-mediated killing of residual leukemic cells [125]. A CTLA-4 polymorphism CT60 AA genotype, located in the 3′-UTR of CTLA4 gene, was shown to be associated with relapse in AML patients [126].

Preclinical studies demonstrated that PD-1 pathway was dysregulated in acute myeloid leukemia (AML). Murine leukemic cell C1498 shows low level PD-L1 expression when grown in vitro, but demonstrates upregulation of PD-L1 expression when grown in vivo, suggesting the microenvironment is conducive to expression of PD-L1 in leukemic cells [127]. Tregs and CD8⁺ T cells with PD-1 expression significantly increased in the liver where C1498 leukemic cells disseminate following C1498 inoculation [128]. Similar finding is also observed in the bone marrow of AML patients [129]. Tregs have suppressive effect on CD8⁺ T cell proliferation and secretion of IFN-γ from CD8⁺ T cells. However, in PD-1 knock-out (KO) mice or in wild-type mice injected with anti-PD-L1 antibody, the suppressive effect of Tregs was abrogated [128]. When C1498 leukemia cells were inoculated to PD-1 KO mice, enhanced anti-tumor response was observed with longer survival compared with C1498 inoculation to wild-type mice [127, 128]. Similar anti-tumor activity was seen with in vivo administration of anti-PD-L1 antibody to C1498-challenged wild-type mice [127, 128]. In human, mRNA expression of PD-L1 and PD-L2 is observed in many AML cell lines. However, PD-1 and CTLA-4 mRNAs were only detected in KG-1 cells [130].

Clinical data also supports dysregulated PD-1 pathway in AML. Compared to healthy individuals, PD-1 expression on T cells was significantly higher in patients with AML [131]. By quantitative polymerase chain reaction (Q-PCR), upregulation (≥twofold) of PD-L1 and PD-L2 mRNA in CD34-positive cells was observed in 36 and 12% of patients with myelodysplastic syndrome (MDS) [130]. Similarly, upregulation of abovementioned mRNAs in CD34-positive cells were seen in 25 and 33% of patients with AML, respectively. By immunohistochemistry, PD-L1 protein expression in leukemic blasts was seen in 20% of patients with MDS, chronic myelomonocytic leukemia, or AML. Of interest, upregulation of PD-L1, PD-L2, PD-1, and CLTA-4 was observed in 66% of patients with myeloid neoplasms who underwent epigenetic therapy.

In addition to PD-1 pathway and CTLA-4, another immune inhibitory molecule, TIM-3, is explored. Both human and mouse AML cells express galectin-9, a ligand of TIM-3. In a mouse model, exhausted T cells co-expressing PD-1 and TIM-3 were found, and they have reduced production of INF-γ, TNF-α, and IL-2 in reaction to their ligands-expressing AML cells. Blocking PD-1 or TIM-3 alone was not sufficient to reduce tumor burden, but combined blockade showed increased tumor rejection and improved survival [132]. The role of immune escape function of TIM-3 is also seen in AML patients as well. TIM-3 in bone marrow T cells is more frequently present in relapsed AML patients compared to those in remission or healthy donors [133].

A phase 1 study with nivolumab single therapy included 27 patients with relapsed/refractory PCM (NCT01592370). With the median follow-up of 65.6 weeks, stable disease was the best response in 17 (63%) patients, which lasted a median of 11.4 weeks (range 3.1–46.1 weeks) [145]. In a different arm of the same study (NCT01592370, CheckMate 039), nivolumab and ipilimumab combination therapy was tried in seven patients with relapsed/refractory PCM [149]. Only one patient (14%) showed stable disease and four patients died due to disease progression. A phase 3 study with pomalidomide and dexamethasone with or without nivolumab for patients with relapsed/refractory plasma cell myeloma is ongoing (NCT02726581, CheckMate 602). A notable phase 2 study with nivolumab, lenalidomide, and low-dose dexamethasone is underway in patients with high-risk smoldering plasma cell myeloma (NCT02903381).

A phase 1 study of pembrolizumab given in combination with lenalidomide and low-dose dexamethasone to patients with relapsed/refractory plasma cell myeloma showed responses in 20 of 40 patients (50%), including 38% of patients who were refractory to lenalidomide (KEYNOTE-023, NCT02036502) [150, 151]. Similarly, in a phase 2 study with pembrolizumab, pomalidomide, and dexamethasone given to 48 patients with relapsed/refractory plasma cell myeloma, the ORR was 56% (27 patients) including 4, 3, 6, and 14 patients with stringent CR, near CR, very good PR, and PR, respectively (NCT02289222) [152]. A similar, smaller scale study with the same regimen given to patients with relapsed/refractory plasma cell myeloma also showed clinical activity with acceptable toxicity [153]. A phase 3 study with pomalidomide and low-dose dexamethasone with or without pembrolizumab for patients with relapsed/refractory plasma cell myeloma is currently recruiting patients (NCT02576977 KEYNOTE-183) [154]. Another phase 3 study designed for patients with newly diagnosed, treatment naïve plasma cell myeloma, who are ineligible for autologous stem cell transplantation (NCT02579863, KEYNOTE-185) [155]. Similar to nivolumab, pembrolizumab is also tried to patients with intermediate- or high-risk smoldering plasma cell myeloma, but as a single therapy (NCT02603887).

There are several clinical trials with anti-PD-L1 antibodies as a single therapy or combined with others in patients with plasma cell myeloma (NCT01375842, NCT02431208, NCT02616640, NCT02685826, NCT02716805, NCT02784483, and NCT02807454), but results have not been reported yet.

Available data is limited regarding CTLA-4 blockade in patients with plasma cell myeloma. Twenty-nine patients including 6 with myeloma were enrolled in a study of ipilimumab to treat relapse after allogeneic stem cell transplant. No objective response was seen in patients with myeloma [156]. The previously described phase 1/1b trial with ipilimumab in patients with relapsed hematologic malignancies after allogeneic stem cell transplant included one patient with lung plasmacytoma, who showed a partial response without progression for more than 21 months (NCT01822509) [148]. A phase 1/2 study of combined checkpoint inhibition with nivolumab and ipilimumab in patients with plasma cell myeloma or lymphoma who are status post autologous stem cell transplant at high risk for post-transplant recurrence is underway (NCT02681302, CPIT001). A phase 1 study of tremelimumab with durvalumab is ongoing in patients with autologous stem cell transplant for plasma cell myeloma (NCT02716805).

Ipilimumab appears to be efficacious in relapsed AML patients after allogeneic stem cell transplant. The phase I/Ib study with ipilimumab (10 mg/kg) in patients with relapsed hematologic malignancies after allogeneic stem cell transplant (NCT01822509) included 16, 2, and 1 patients with AML, MDS, and myeloproliferative neoplasm, respectively. Among 22 patients treated with 10 mg of ipilimumab per kilogram, 5 patients (23%) who showed a complete including 3 with leukemia cutis, 1 with myeloid sarcoma, and 1 with AML showed a complete response. Additional four patients with AML did not achieve an objective response, but showed a reduction in the tumor burden [148].

In a phase I study, patients with high risk MDS (n = 11) who failed with hypomethylating agents were treated with ipilimumab monotherapy. Although objective response was not reported in any patients, disease stabilization was seen in five patients (45%) [157]. Many other clinical trials with anti-CTLA-4 antibodies are explored in patients with MDS or AML as single therapy or in combination with others (NCT01757639, NCT02117219, NCT02846376, and NCT02890329).

A single-center, phase 1b/2 study of nivolumab combined with azacitidine in patients (n = 51) with relapsed AML demonstrated superior survival compared to historical survival data derived from patients with relapsed AML treated with azacitidine-based salvage protocols. Among 35 patients who were evaluable for response, 6 patients (18%) with complete remission (CR) or complete remission with insufficient recovery of counts (Cri), 5 (15%) with hematologic improvement (HI), 9 (26%) had 50% bone marrow blast reduction, and 3 (9%) had stable disease. Of note, 12 patients (34%) had disease progression [158]. A preliminary result of a phase 2 study with various combinations of nivolumab, ipilimumab, and azacitidine in MDS patients (NCT02530463) was reported [159]. In the cohort of treatment-naïve MDS patients who were treated with azacitidine plus nivolumab, the ORR was 69% (9/13) with 2 CR, 5 morphologic CR and hematologic improvement (HI), and 2 HI. In the cohort of MDS patients with hypomethylating agent failure, ipilimumab single therapy showed some response (ORR 22%). However, in the same cohort, nivolumab single therapy demonstrated no response and enrollment was stopped. Preliminary result of a similar study with pembrolizumab (KEYNOTE-013, NCT01953692) in patients with MDS who failed with hypomethylating agents was also reported. The ORR was 4% (1/27) with no CR and 1 PR [160]. Other clinical trials of anti-PD-1 antibody in combination with hypomethylating agent(s) in patients with MDS or AML patients are ongoing (NCT02845297 and NCT02599649).

Anti-PD-1 antibody can be explored with chemotherapeutic agents in patients with AML. NCT02464657 and NCT02768792 are such studies. In another angle, PD-1 blockade can be tried in AML patients who are in remission (NCT02275533, NCT02532231, and NCT02708641). A phase 2 study of pembrolizumab in patients with non-favorable risk AML who underwent lymphodepletion with fludarabine and melphalan followed by autologous transplantation will be interesting (NCT02771197).

Among anti-PD-L1 antibodies, durvalumab is actively studied in patients with MDS or AML. A phase 2 study with oral azacitidine with durvalumab in patients with MDS who failed with hypomethylating agents is underway (NCT02281084). A similar phase 2 study, but with subcutaneous azacitidine in combination with durvalumab in treatment-naïve MDS or elderly (≥65 years) AML patients is also underway (NCT02775903). A phase 1 study with durvalumab single therapy or in combination with tremelimumab with or without azacitidine to patients with MDS is ongoing (NCT02117219).

Checkpoint inhibitors, like any other drugs, do not provide benefits to patients without risks. Immune-related adverse events (irAEs) are a spectrum of side effects including gastrointestinal, dermatologic, hepatic, or endocrine events. It is usually recommended for patients with grade 2 irAEs to withhold checkpoint inhibitor transiently. For patients with grade 3 or higher irAEs, checkpoint inhibitor should be stopped and treated with systemic corticosteroids (1 to 2 mg/kg or equivalent) daily. Other immune modulatory agents such as Infliximab can be considered for patients without improvement with steroids [161].

In general, IrAEs with anti-PD-1 antibodies are less common than anti-CTLA-4 antibody. In 298 melanoma patients treated with ipilimumab (3 mg/kg), irAEs of any grade were reported in 85% of patients [162]. Grade 3 or higher irAEs are seen in 112 patients (38%), diarrhea being the most common irAE followed by hepatotoxicity, dermatitis, hypophysitis, and uveitis. Approximately 1/3 of patients were treated with systemic corticosteroids, but it did not affect OS or time-to-treatment failure indicating generous use of corticosteroid for irAEs. In a pooled analysis of 576 melanoma patients treated with nivolumab (3 mg/kg), 71% of patients suffered irAEs of any grade [163]. Grade 3 or higher irAEs were seen in 57 (10%) of patients including neurologic AEs, autoimmune neuropathy, central demyelination, Guillain-Barré syndrome, and involuntary muscle contractions. Similar to ipilimumab, management of irAEs with systemic corticosteroids did not affect treatment response of nivolumab. Grade 3 or higher irAEs were more common in melanoma patients who were treated with combined nivolumab and ipilimumab compared with those treated with either ipilimumab or nivolumab single therapy (55, 25, and 16%, respectively) [164, 165].

Although the clinical efficacy of anti-PD-1 therapy has been proven, not all cancer types respond to anti-PD-1 therapy. In solid tumors, immunologically responsive tumors vs. immunologically ignorant tumors are recognized based on immune cell infiltration in the TME. The former tend to be seen with numerous T cells in the TME (inflamed tumors) and to have a high mutational load in tandem with neoantigens with higher quantity. Immunologically responsive tumors are more likely responsive to anti-PD-1 therapy [166]. However, anti-PD-l therapy is not effective in all patients with responsive tumors and even in those with response, delayed, or mixed tumor regression can be seen [167]. Furthermore, manipulation of immune checkpoints with anti-PD-1 agents not uncommonly causes irAEs. Therefore, biomarkers to selectively identify best candidates are much needed.

Several methods are currently available (Table 2). PD-L1 expression in tumor cells assessed by immunohistochemistry has been associated with better response to anti-PD-1 therapy in solid tumors as well as in CHL [167, 168]. However, there are different types of clones for PD-L1 immunohistochemical antibodies and standardization has not been achieved. Similarly, the serum level of soluble PD-L1 measured by enzyme-linked immunosorbent assay (ELISA) can be a potential predictive biomarker in patients with DLBCL or PCM [118, 169]. However, these patients were treated with conventional chemotherapy, so investigation targeting the PD-1 pathway must be conducted. TILs, particularly with PD-L1 expression, were associated with higher response to PD-1-targeting therapy in patients with solid tumors [5, 170]. However, data are not available regarding TILs with PD-L1 expression in patients with lymphoma. Assessment of dynamics in immune cell profiles in the TME of biopsy samples using immunohistochemistry at different time points during the treatment schedule sheds light on the prediction of response. Chen et al. has demonstrated that immune cell profiles early in treatment, not before treatment, are predictive of treatment response. They also showed that gene expression profiling using a 795-gene NanoString panel recapitulates the result [171]. Immune cell profiles can also be evaluated with peripheral blood using flow cytometry [172, 173].

The higher the mutational load in cancer cells, the more neoantigens are produced in them. Neoantigens generally have high antigenicity, which attracts immune cells (inflamed tumors). High mutational load is associated with a better response to anti-PD-1 therapy [174]. However, the number of mutations in cancer cells does not directly correlate with the production of high-quality neoantigens. A computational genomic tool has been developed to predict immunogenicity of mutagen-derived neoantigens or cancer germline antigens and their binding affinity to immune cells. It can further provide prediction of response when treated with anti-PD-1 or anti-CTLA-4 agents [175, 176]. The computational genomic tool was shown to be feasible with solid tumor, yet applicability of mutational load as a biomarker in lymphomas is questionable due to lack of data [174, 177].

Not all of abovementioned methods can be applicable to hematologic malignancies because most data was driven from patients with solid tumors or solid tumor models. However, efforts to detect intrinsic overexpression of PD-L1 or PD-L2 are recommended to identify possible good responders to anti-PD-1/PD-L1 therapy. That is, immunohistochemistry for PD-L1/PD-L2 protein expression, chromosome analysis, or fluorescence in situ hybridization to detect aberrations in 9p24.1/PD-L1/PD-L2 locus or RT-PCR to detect gene rearrangements involving PD-L1 or PD-L2 could be a minimum step.