Checkpoint inhibitors in AML: are we there yet?

Summary

Immunotherapy is distinct from traditional chemotherapy in that it acts on immune cells rather than cancer cells themselves. Monoclonal antibodies targeting immune checkpoints on T cells – CTLA-4 and PD-1 – and PD-L1 on the cells of immune microenvironment are now approved for clinical use in several solid tumors and hematological malignancies. This article provides a general overview of the use of checkpoint inhibitors in hematologic malignancies with a special focus in acute myeloid leukemia.

Immunotherapy is distinct from traditional chemotherapy in that it acts on immune cells rather than cancer cells themselves (Batlevi et al., 2016; Khalil et al., 2016; Weiss et al., 2019). Monoclonal antibodies targeting immune checkpoints on T cells – CTLA-4 and PD-1 - and PD-L1 on the cells of immune microenvironment are now approved by the Food and Drug Administration (FDA) and European Medical Agency (EMA) for the treatment of malignancies including melanoma (Hodi et al., 2010), non-small cell lung cancer (Brahmer et al., 2015), renal cancer (Motzer et al., 2018) and Hodgkin lymphoma (Ansell et al., 2015). Inhibition of checkpoints activates T cells and recognition of tumor-associated antigens leads to lysis of tumors and regression of cancer. Thus immune checkpoint inhibition enhances recognition of tumors as non-self and an expanding body of evidence in solid tumors identified several factor for successful applications of immune checkpoint blockers, including recognition of tumor-related antigens presented in the context of human leukocyte antigen (HLA) by T cells, presence of the checkpoints on T cells and the ligands of checkpoints on the antigen presenting cells or the tumor cells (Rizvi et al., 2015). These therapies have demonstrated antitumor responses in several malignancies either alone or in combination with other anti-cancer therapies (Ansell et al., 2015; Younes et al., 2016; Younes et al., 2019). This article provides a general overview of the use of checkpoint inhibitors in hematologic malignancies with a special focus in acute myeloid leukemia. We also discuss their treatment-related toxicity as well as the use of these agents in specific populations such as allogeneic hematopoietic stem cell transplantation (alloHCT) recipients.

Immune escape of AML after allogeneic hematopoietic stem cell transplantation

AlloHCT is a potentially curative option for patients with high-risk AML (Gooley et al., 2010; Jenq & van den Brink, 2010). Despite increases in overall survival due to improved patient selection, HLA-matching and supportive care, relapse remains a major cause of death in patients with AML who undergo alloHCT (Wayne et al., 2013). In addition to the effectiveness of high-intensity chemotherapy with or without radiation, alloHCT recipients benefit from the ability of donor-derived T cells to recognize host AML cells, which bear non-self HLA, as foreign and thus mount a graft versus leukemia (GVL) effect that eliminates any residual disease after alloHCT and maintains lasting immunosurveillance to prevent relapse (Copelan, 2006; van den Brink et al., 2010). However, host tissue recognition as foreign by donor T cells can also lead to a broader immune reaction causing potentially lethal graft-versus-host disease (GVHD) (Ferrara & Reddy, 2006).

In preventing relapse after AML, the relative contributions of the immunological effects of alloHCT, namely GVL effects and the cytotoxic effects of high-intensity conditioning have not been completely elucidated. A detailed analysis of leukemic cells that relapse after alloHCT may however provide useful insights into mechanisms of relapse. An early study of patients with myeloid leukemia that relapsed after alloHCT demonstrated decreased ability to stimulate donor T cell proliferation in 4 of 5 patients (Dermime et al., 1997). The relapsed cells also had decreased MHC-expression, and decreased susceptibility to T and NK cell lysis.

With the advent of high-dimensional analsysis, a comprehensive analysis of samples from 15 patients with AML who relapsed after alloHCT was performed. The study included whole exome sequencing, whole transcriptome analysis and multicolor flow cytometry (Christopher et al., 2018). Curiously, whole exome sequencing revealed relapse after alloHCT was not associated with the acquisition of new AML-specific mutations or structural variations in immune-related genes. Whole transcriptome studies by RNAseq and flow revealed involvement of immune-related pathways and most notably HLA class II regulation. In a second study, serial whole transcriptome and genome analysis was performed on leukemic blasts at different time points (Toffalori et al., 2014). A signature consistent with immune dysregulation was found that affected both antigen presentation and T cells. The AML blasts at relapse after alloHCT expressed checkpoints including PD-L1, and loss of HLA class II genes concurrently with down-regulation of its regulator CIITA.

An investigation into the donor T cell compartment in the bone marrow infiltrating CD4+ and CD8+ T cells from patients with AML after alloHCT demonstrated an exhaustion signature (Noviello et al., 2019). Bone marrow infiltrating T memory stem cells and central memory T cells, expressing the irreversibly exhausted PD-1+Eomes+T-bet—predicted relapse. Accumulation of regulatory T cells was also noted in patients with relapse. In HLA-matched alloHCT, CD8+ T cells infiltrating the bone marrow of patients expressing PD-1, CTLA-4, and/or TIM-3 was higher in patients who relapsed than in patients without disease.

Given the primacy of immune dysregulation associated with AML relapse after alloHCT, T cell-based therapies are a reasonable approach to treat or prevent relapse. Adoptive transfer of donor T cells in the form of donor lymphocyte infusion (DLI) is one such approach to overcome the exhausted T cells. Early success with DLI in chronic myelogenous leukemia (Kolb et al., 1995; Collins et al., 1997) led exploration of its use in AML and MDS with more limited success (Depil et al., 2004; Schmid et al., 2007). One key limitation of the use of DLI to treat a full-blown relapse is the high burden of disease. Donor T cells, driven by expansion of alloreactive T cells, are slower than chemotherapy and also dependent on the expression of HLA. Hence hypomethylating agents like azacytadine have also been used in conjunction with DLI (Schroeder et al., 2013; Tessoulin et al., 2014; Steinmann et al., 2015; Motabi et al., 2016). In one of the larger retrospective analyses of outcomes of intensive chemotherapy with that of hypomethylating agents in relapsed AML and MDS after alloHCT, 56% of patients in the chemotherapy group and 33% of patients in the hypomethylating agents group received at least one DLI after treatment. The use of chemotherapy followed by DLI offered better outcomes (ORR, 68%; CR, 59%, 1-year OS, 44%; and median OS, 9·8 months) than hypomethylating agents (Motabi et al., 2016).

Checkpoint inhibitors in lymphoid malignancies

After some early clinical trials of PD-1 blockade using the mAb pidilizumab (Berger et al., 2008; Westin et al., 2014) and the anti-CTLA-4 mAb ipilimumab (Ansell et al., 2009) suggested activity in lymphoid malignancies, recent studies have focused on two different PD-1 mAbs, nivolumab and pembrolizumab (Perales et al., 2016). The most promising results have been observed in Hodgkin lymphoma, which has frequent amplification of genetic material at 9p24, which results in the over-expression of the PD-1 ligands, PD-L1 and PD-L2 (Green et al., 2010). Due to this and additional mechanisms, the very high frequency of PD-L1/PD-L2 expression on the surface of the HL tumor cell surface suggested that HL is a uniquely vulnerable target for PD-1 blockade (Chen et al., 2013). The clinical results in Hodgkin lymphoma have lead to approval of both drugs in this disease. Initial studies in Hodgkin lymphoma reported ORR and CR rates of 65% and 16% with pembroluzimab (Armand et al., 2016), and 87% and 17% with nivolumab (Ansell et al., 2015), respectively (Table I). These results lead to two large studies testing each drug in several cohorts. The Checkmate 205 study included patients who had not received brentuximab vedotin (BV), had received it after an autologous stem cell transplant (ASCT), and had received BV both pre and post ASCT (Younes et al., 2016; Armand et al., 2018). The objective response rate was 69% overall and 65– 73% in each cohort (Armand et al., 2018). The median duration of response was 16·6 months (95% CI, 13·2– 20·3 months), and median progression-free survival was 14·7 months (95% CI, 11·3–18·5 months). The KEYNOTE-087 study examined 3 cohorts of patients treated with pembroluzimab, patients who progressed after ASCT and BV, patients who failed salvage chemotherapy and BV, were ineligible for ASCT, and those who failed ASCT but did not receive BV (Chen et al., 2017; Chen et al., 2019). The ORR was 69·0% (95% CI, 62·3–75·2%), and the CR rate was 22·4% (95% CI, 16·9–28·6%). ORRs were 73·9% for cohort 1, 64·2% for cohort 2, and 70·0% for cohort 3. More recently, nivolumab has been tested in the upfront setting in combination with AVD (Ramchandren et al., 2019). The ORR was 84% (71–93%), with 67% (52–79%) achieving CR. Additional studies are ongoing and nivolumab has also been combined with BV (Brockelmann & Engert, 2017; Herrera et al., 2018). The exact sequence of treatments in patients with Hodgkin lymphoma, including the role of ASCT and alloHCT remains an evolving picture as new data becomes available (Perales et al., 2015). BV has now moved upfront in selected patients, and is also being incorporated in salvage regimens pre ASCT. Checkpoint inhibitors are typically reserved for patients who have failed an ASCT, and are being used in BV naïve as well as BV refractory patients. While the prior standard of care in ASCT failures was to proceed with alloHCT in patients responding to salvage therapy, this paradigm has shifted with BV and checkpoint inhibitors, where many clinicians are deferring alloHCT until progression or in patients with incomplete responses.

Table I.

Select trials of anti-PD-1 antibody therapy in Hodgkin lymphoma.

Checkpoint inhibitor	Additional agents	Patients	n	Main outcomes	Citation
Nivolumab		Relapsed/refractory HL	23	ORR: 87%; CR: 17%	Ansell et al., 2015
Pembrolizumab		cHL after Brentuximab vedotin failure (KEYNOTE-013)	31	ORR: 65%; CR: 16%	Armand et al., 2016
Nivolumab	Treatment history: Cohort A: Brentuximab vedotin (BV)-naïve Cohort B: BV received after auto-HSCT Cohort C: BV received before and/or after auto-HSCT	Relapsed/refractory cHL after auto-HCT treatment failure (CheckMate 205)	243 Cohort A: 63 Cohort B: 80 Cohort C: 100	ORR: 69%; CR: 16% Cohort A: ORR: 65%; CR: 18% Cohort B: ORR: 68%; CR: 10% Cohort C: ORR: 73%; CR: 12%	Armand et al., 2018
Pembrolizumab		Relapsed or refractory classic Hodgkin lymphoma (KEYNOTE-087)	210	ORR: 719%; CR: 276%	Chen et al., 2019
Nivolumab	Doxorubicin, vinblastine, dacarbazine (N-AVD)	Newly diagnosed cHL (CheckMate 205) Cohort D		ORR: 84%; CR: 67%	Ramchandren et al., 2019

In contrast to the excellent results in Hodgkin lymphoma, the activity of nivolumab in other hematologic malignancies has been quite variable, with response rates of 36–40% noted in follicular lymphoma, DLBCL, and T-cell lymphoma (Lesokhin et al., 2014; Armand et al., 2015).

Checkpoint inhibitors in AML

Studies have shown that polyclonally expanded T cells might recognize tumor cells that have a higher mutation burden (i.e., the total number of mutations per coding area of the tumor genome), which have higher odds of presenting antigens that evoke recognition as non-self (Yarchoan et al., 2017). In several solid tumors, a significant correlation exists between tumor mutation burden and objective response rates, and tumor mutation burden is therefore an important predictor of responses to immunotherapy (Snyder et al., 2014; Rizvi et al., 2015; McGranahan et al., 2016; Samstein et al., 2019).

AML and MDS express somatic mutations that could be recognized as non-self, but the tumor mutation burden is relatively lower than most other tumors (Lawrence et al., 2013). Immune checkpoint expression is generally low at baseline. Additionally, AML and MDS are associated with a dysfunctional and suppressive immune compartment (Kordasti et al., 2007; Wei et al., 2013; Arandi et al., 2018; Han et al., 2018). Simlar to the approach discussed above in patients receiving DLI, investigators have proposed combining agents that can enhance antigen presentation and/or increase immune checkpoint expression with immune checkpoint inhibitors (Table II).

Table II.

Immune checkpoint therapy for AML.

Checkpoint inhibitor	Additional agents	Patients	N	Main outcomes	Citation
Pre alloHCT
Nivolumab	Cytarabine and idarubicin	Newly diagnosed AML	44	Median EFS not reached (95% CI 7·93-NR). Median OS and RFS in responders 18·5 months	Ravandi et al., 2019
Nivolumab	Azacitidine	Relapsed/refractory AML	70	ORR: 33%; CR/iCR: 22%. Median OS 6·3 months	Daver et al., 2019
Post alloHCT
Ipilimumab		Relapsed heme malignancies after conventional alloHCT	29 (including 2 AML)		Bashey et al., 2009
Ipilimumab		Relapsed heme malignancies after conventional alloHCT	28 (including 12 AML)	CRs in 4 patients with extramedullary AML and 1 patient with MDS transformed to AML.	Davids et al., 2016

One such approach has been to utilize chemotherapy to release antigens that might help prime T cells. In a phase 2 single-arm study reported from MD Anderson Cancer Center, nivolumab was used in combination with idarubicin and cytarabine as a frontline therapy for AML or high risk MDS (Ravandi et al., 2019). Induction included cytarabine 1·5 g/ m² by continuous infusion daily on days 1–4 (3 days in patients >60 years) and idarubicin 12 mg/m² daily on days 1–3. Nivolumab 3 mg/kg was started on about day 24 (±2 days) and continued every 2 weeks for up to a year in responders. Of the 44 patients enrolled in this study, the primary endpoint of median event free survival was not reached at a median follow up of 17·25 months. The median relapse free survival of the responders and overall survival were 18·54 months. Six patients had seven grade 3–4 immune-related adverse events which included rash, colitis, transaminitis, pancreatitis, and cholecystitis. Nineteen (43%) of 44 patients achieved a response and proceeded to allogeneic stem cell transplantation, with grade 3/4 graft-versus-host disease observed in five (26%).

Another promising strategy has been to use hypomethylating agents such as azacitadine, which are a standard antineoplastic therapy in AML/MDS. In a single arm, phase II study, a combination of nivolumab and azacytidine was tested in 70 patients with relapsed/refractory AML (Daver et al., 2019). Azacytidine 75 mg/m² was administered on days 1–7 and nivolumab 3 mg/day on days 1 and 14 every 4–6 weeks. The results were encouraging with an ORR rate of 33%, including 4 (6%) complete remissions (CR), 11 (16%) CR with incomplete blood count recovery (CRi), 1 partial remission (PR) and 7 showing hematologic improvement. Six patients (9%) had stable disease. The combination was well tolerated with grade 2, and grade 3–4 immune-related adverse effects observed in eight (11%) patients respectively, which were mostly responsive to steroids or infliximab. Majority of the immune-related adverse effects happened in the first 8 weeks after initial treatment of nivolumab, and two patients died due to steroid and infliximab-resistant toxicities.

While the results of some of these early studies are encouraging, at the present time, checkpoint inhibitors are not recommended as part of standard of care in patients with AML. Ongoing studies should help to clarify the role of checkpoint inhibitors in myeloid malignancies.

Toxicities of checkpoint inhibitors

The significant benefits in clinical outcomes obtained with the use of checkpoint inhibitors in several indications have been hampered by their treatment-related toxicities. Most of these adverse events occur as a result of the dysregulation in the immune system balance produced by these drugs (Pardoll, 2012). Because these immune-related adverse events (irAERs) differ from classical chemotherapy-related side effects both in the clinical presentation and in their treatment, a multidisplinary approach to manage them has been proposed. In many centers, specific working groups where gastroenterologists, hepatologists, infectious disease specialists and neurologists (among others) work together with oncologists and hematologists, have been develop in order to improve patient care. Patient education is also important for early diagnosis and treatment of these complications, since it has been shown that early intervention of irAEs reduces both their severity and duration (Hodi et al., 2010).

The spectrum of irAEs is diverse and can involve almost any tissue or organ (Bair et al., 2019). These events include but are not limited to complications in skin (rash, vitiligo, toxic epidermal necrolysis) (Abdel-Rahman et al., 2015), the lungs (pneumonitis, pleural effusion) (Naidoo et al., 2017), thyroid gland (hypothyroidism) (Ryder et al., 2014), the gastrointestinal tract (colitis, pancreatitis, celiac disease) (Beck et al., 2006; Peyrony et al., 2019), the nervous system (encephalopathy, Guillain-Barre syndrome, myelitis) (Liao et al., 2014), the liver (hepatitis) (Lleo et al., 2019), the heart (myocarditis) (Quagliariello et al., 2019) and hematological cells (immune cytopenias) (Friedman et al., 2016). Most of these events can occur simultaneously or at different intervals and may involve different organs or tissues. Time to irAEs is also quite variable and can range from a few weeks to up to 2 years after the first treatment administration (Wolchok et al., 2010). Moreover, these events can occur even after discontinuation of the drug. Whereas most of these side effects are manageable, they can be potentially fatal in some cases (Johnson et al., 2016; Brahmer et al., 2018).

Patients receiving checkpoint inhibitors seem to be at a higher risk of developing infections, especially if they require corticosteroids or other immunosuppressant agents to treat immune-related adverse events (Redelman-Sidi et al., 2018). In a recent study, 7.3% of patients receiving checkpoint blockade for the treatment of melanoma developed severe infections (Del Castillo et al., 2016). In hematological patients, this risk might be higher due to previous immunosuppressive therapies administered to the patient and the nature of the malignancy (Ansell et al., 2015).

Risk factors for the development of irAEs events include the type of checkpoint inhibitor, combination therapy and previous autoimmune disease. In terms of the type of antibody, PD1/PDL1 axis inhibitors seem to have lower incidence of adverse events than CTLA-4 inhibitors whereas the combination of 2 inhibitors has a higher rate of immune-related events than either monotherapy (Friedman et al., 2016). Most of the trials evaluating the use of these drugs in cancer patients excluded patients with previous history of autoimmune diseases. There is limited data showing that checkpoint inhibitors can be safely administered to patients with autoimmune disease and the decision for the indication should be taken with caution (Kyi et al., 2014).

Role of checkpoint inhibitors in the context of alloHCT

While checkpoint inhibitors have a theoretical potential during peri-alloHCT to enhance donor-derived T cells, concerns for the development of uncontrolled GVHD and other fulminant immune adverse events have hampered wider use of these agents.

Given these concerns, the first trial to study Ipilimumab was attempted with very low doses to treat relapsed disease (Bashey et al., 2009). In this phase I trial, 29 patients with hematological malignancies including 2 with AML who relapsed after alloHCT were treated with 0·3–1 mg/kg in a single infusion. Responses were noted in only three of the patients with lymphoid malignancies. Importantly, dose-limiting toxicity was not encountered and no evidence of GVHD or graft rejection was found. This led to a phase I/II trial of 28 patients with post-alloHCT relapsed hematological malignancies that enrolled to receive ipilimumab at 3 or 10 mg/kg every 3 weeks for up to 4 doses (Davids et al., 2016). Additional doses were given every 12 weeks for up to 60 weeks in patients who showed clinical benefit. No responses or significant immune adverse effects were seen in the 3 mg/kg arm. In the 10 mg/kg arm, CRs occurred in 4 patients with extramedullary AML and 1 patient with the MDS transformed into AML. Four patients had a durable response that lasted more than 1 year. Responses were associated with infiltration of CD8+ T cells, decreased regulatory T cells, and expansion of effector T cells subsets.

In contrast to ipilimumab, the use of anti-PD1 antibodies in the context of alloHCT has been hampered with reports of significant GVHD and other irAEs. The effects of PD-1 when used to treat lymphoma prior to alloHCT resulted in grade II–IV and grade III–IV acute GVHD in 44% and 23% of the patients, respectively (Merryman et al., 2017). While the rates of II-IV aGVHD were expected in this patient population, the rates of grade III-IV were higher than predicted. Furthermore, a number of patients required steroids for treatment of a persistent febrile syndrome. Additionally Hodgkin lymphoma patients who relapsed after nivolumab and underwent alloHCT also had high rates of acute GVHD (Armand et al., 2018). In spite of the setbacks with anti-PD1 therapies, there are ongoing trials with nivolumab and pembrolizumab after alloHCT with encouraging early results that continue to accrue.

Future directions

The success and enthusiasm for immune checkpoint inhibitors have hitherto been less in AML compared to other cancers like solid tumors and Hodgkin lymphoma. However greater understanding of the immunological basis for relapse and differences between the mechanisms behind commonly used immune checkpoint inhibitors will likely see greater use of these and newer agents. Since the greatest concern in the use of immune checkpoints is development of uncontrollable GVHD, alloHCT platforms with low GVHD potential could be used. In vivo and in vitro T cell depleted alloHCT are associated with lower incidence of acute and chronic GVHD (Papadopoulos et al., 1998; Jakubowski et al., 2007; Jakubowski et al., 2011; Barba et al., 2017; Malard et al., 2018; Cho & Perales, 2019) (Papadopoulos et al., 1998; Jakubowski et al., 2011; Jakubowski et al., 2017). Further studies will be required to identify the optimal treatment combinations with checkpoint inhibitors in AML as well as their use in the context of alloHCT.