PD1 blockade with pembrolizumab is highly effective in relapsed or refractory NK/T-cell lymphoma failing l-asparaginase

Natural killer (NK)/T-cell lymphomas are aggressive malignancies, with anthracycline regimens resulting in dismal outcome.¹  L-asparaginse regimens,¹  particularly SMILE (dexamethasone, methotrexate, ifosfamide, l-asparaginase, and etoposide), have significantly improved the outlook^2,3  and are now the most commonly employed initial therapy.^1,4  However, SMILE or SMILE-like regimens still fail in 20% to 40% of cases.^1-3  There is no known effective salvage, and outcome is virtually always fatal.

NK/T-cell lymphomas are invariably infected by Epstein-Barr virus (EBV), which exists in a latency II state, expressing immunogenic antigens EBNA1, LMP1, and LMP2.⁵  Furthermore, EBV-infected lymphoma cells upregulate programmed death ligand 1 (PDL1), ligand of the inhibitory receptor programmed death 1 (PD1) on T cells.⁶  Ligation of PDL1 on lymphoma cells with PD1 on effector T cells suppresses T-cell cytotoxicity. The PDL1/PD1 axis is therefore a potential mechanism for NK/T-cell lymphomas to avert effector T-cell targeting.

Here we report retrospectively results of PD1 blockade in relapsed/refractory NK/T-cell lymphomas failing l-asparaginase regimens, which provide proof-of-concept data that the PDL1/PD1 axis⁶  is highly relevant in this malignancy.

Patients with relapsed/refractory NK/T-cell lymphomas failing l-asparaginase regimens from Hong Kong, Singapore, and Seoul, treated with the anti-PD1 antibody pembrolizumab and reported annually to the Asia Lymphoma Study Group since 2015, were analyzed. On a named-patient off-label basis, pembrolizumab at 2 mg/kg every 3 weeks (based on efficacy in Hodgkin lymphoma)⁷  was used in all but 1 patient, who received pembrolizumab every 2 weeks. Patients provided informed consent for therapy.

[¹⁸F]Fluorodeoxyglucose (FDG) positron emission tomography computed tomography (PET/CT) was used to assess response according to standard criteria (5-point Deauville score).⁸  For lesions not observed in pretreatment or other preceding scans, accessible ones were biopsied, whereas inaccessible ones were evaluated by standard criteria⁸  and immune-related response criteria.⁹  Because this study was retrospective, schedule of PET/CT varied, but early assessment after 3 to 4 cycles¹⁰  was adopted in most cases. Circulating EBV DNA was quantified with quantitative polymerase chain reaction.¹¹ 

Seven men (median age, 49 years; range 31-68, years) were treated (Tables 1 and 2). Most had advanced-stage disease with widespread organ involvement (initial diagnosis: I_E, n = 2; IV, n = 5; before pembrolizumab: I_E, n = 1; IV, n = 6; Table 1). Systemic HPS was present in 5 patients before pembrolizumab treatment. SMILE or SMILE-like regimens had failed for all patients. Two patients had relapsed after allo-HSCT. The median number of prior regimens including allo-HSCT was 2 (range, 1-5; Table 2). Details of previous treatment are shown in the supplemental Data available on the Blood Web site.

A median of 7 (range, 2-13) cycles of pembrolizumab were administered. Objective response was observed in every patient. After a median follow-up of 6 (range, 2-10) months, 5 patients remained in CR.

Patient 6 had biopsy-proven esophageal involvement and radiologic pulmonary infiltration. Pembrolizumab treatment led to immediate improvement, and PET/CT scan after 6 cycles showed metabolic CR. He has since received 6 more cycles. EBV DNA monitoring was not performed. Patient 7 relapsed after allo-HSCT, presenting with marrow involvement, cytopenia, and HPS. Symptoms and laboratory abnormalities resolved after the first cycle. EBV DNA fell from 3.9 × 10³ copies/mL to undetectable after 2 cycles (Figure 1A). Marrow examination and PET/CT after 3 cycles confirmed CR. He has since received 5 cycles of treatment.

Patient 1 had biopsy-proven cutaneous relapse. Skin lesions responded after the first cycle, and EBV DNA became undetectable (Figure 1A). On PET/CT after 4 cycles, metabolic CR was achieved (Figure 1B). However, biopsy of the index lesion (scar-like by that time) showed scanty CD56⁺, EBER⁺ cells among CD3⁺CD8⁺ T cells (Figure 1C). He remained in clinical and molecular remission after 2 more cycles, with EBV DNA remaining undetectable. Patient 3 on relapse had complete destruction of hard palate and all midline nasal structures (Figure 1B), associated with marrow involvement and HPS. After the first cycle, HPS resolved, with gradual normalization of cell counts and liver biochemistry. EBV DNA fell from 2 × 10² copies/mL to undetectable (Figure 1A). On PET/CT scan after 3 cycles, preexisting nasal/nasopharyngeal lesions disappeared, but new lesions unexpectedly appeared in other areas (Figure 1B). A new lesion in the abdominal wall (Figure 1B) was biopsied, showing a predominance of CD3⁺CD4⁺ T cells with very scanty CD56⁺EBER⁺ cells (Figure 1C). Biopsy of an apparently new nasal lesion showed similar findings (Figure 1C). These lesions disappeared after 6 cycles, but newer nasal lesions appeared. The same phenomenon was seen after 9 cycles. After 12 cycles, FDG-avid lesions were no longer detected, and clinical, molecular, and metabolic CR was finally achieved. Patient 4 presented with involvement of liver and marrow and concomitant HPS. After the first cycle, symptoms and laboratory abnormalities resolved. EBV DNA fell from 6.9 × 10⁴ copies/mL to undetectable after 4 cycles (Figure 1A). After 6 cycles, marrow examination and PET/CT confirmed CR. Afterward, EBV DNA slowly rose to 10² to 10³ copies/mL. After 13 cycles of pembrolizumab, he was still in clinical and radiologic remission.

Patient 2 presented with marrow infiltration and HPS 7 years after allo-HSCT. Clinical and biochemical improvement were immediate after the first cycle. After 3 cycles, marrow was in morphologic remission. However, EBV DNA remained persistently detectable (Figure 1A). PET/CT showed 3 residual hypermetabolic splenic foci (Figure 1B), which on splenectomy showed EBER⁺ cells confined by CD4⁺ and CD8⁺ T cells (Figure 1C). After 3 more cycles, marrow showed reappearance of EBER⁺ cells. He then succumbed to a chest infection. Patient 5 presented with HPS. PET/CT showed diffuse hepatic involvement. After the first cycle, HPS resolved, with normalization of laboratory findings, and EBV DNA fell from 3 × 10³ copies/mL to undetectable (Figure 1A). However, after a second cycle, there was an increase in EBV DNA. Its significance could not be assessed, because he died as a result of sepsis consequent on gastrointestinal bleeding caused by gastric ulcers.

Patient 7 (post–allo-HSCT) developed grade 2 rash, which on biopsy was consistent with acute graft-versus-host disease. There was a quick response to corticosteroid, and further pembrolizumab therapy was unhindered. Otherwise, no treatment-related adverse events were seen in other patients.

In 4 patients (cases 1-4), there was uniformly strong PDL1 expression (supplemental data). In case 5, weaker staining was found in ∼20% of cells. Data were not available for 2 patients (cases 6 and 7).

Responses after pembrolizumab treatment require clinical, radiologic, pathologic, and molecular assessments. Histologic evaluation might show residual EBER⁺ cells among infiltrating CD4⁺ and CD8⁺ T cells, appearing as FDG-avid lesions on PET/CT. With EBV DNA remaining undetectable, these lesions resolved after further treatment, consistent with pseudoprogression^12-14  observed in solid tumors treated with immune checkpoint inhibitors. Similarly, in the presence of continued clinical remission, detectable EBV DNA was not necessarily associated with disease progression. This is different from chemotherapy treatment, where persistently positive EBV DNA portends a poor prognosis.^15-18  Strong PDL1 expression correlated with excellent responses.¹³  The low-dose pembrolizumab regimen (2 mg/kg)⁷  was not associated with significant adverse events, even after allo-HSCT, consistent with safety of immune-checkpoint blockade after allo-HSCT.^19-21 

The authors thank Annie Pang for performing quantification of EBV DNA and Siok-Bian Ng for histopathologic evaluation of a patient.

The Singapore National Medical Research Council Translational and Clinical Research Flagship Programme provided funding for PDL1 staining.

The authors thank members of the Asia Lymphoma Study Group for constructive comments.

Contribution: Y.-L.K. designed the study, treated the patients, and wrote and approved the manuscript; T.S.Y.C., D.T., S.J.K., L.-M.P., and B.M. treated the patients and approved the manuscript; P.-L.K. performed the PET/CT scans and approved the manuscript; F.L, R.A.-Y., and J.I. performed the histopathological examination and approved the manuscript; C.P. treated the patients and approved the manuscript; and E.T. designed the study, treated the patients, and approved the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Yok-Lam Kwong, Department of Medicine, Professorial Block, Queen Mary Hospital, Pokfulam Rd, Hong Kong, China; e-mail: ylkwong@hkucc.hku.hk.