Programmed cell death 1 expression is associated with inferior survival in patients with primary central nervous system lymphoma

Programmed cell death 1 (PD-1) and its ligands PD-L1/PD-L2 have been shown to mediate immune evasion in various cancers, but their prognostic implications in patients with primary central nervous system lymphoma (PCNSL) are poorly understood. Therefore, we analyzed 76 PCNSL patients at initial diagnosis who were treated homogenously with high-dose methotrexate-based chemotherapy, and evaluated the prognostic roles of high immunohistochemical PD-1, PD-L1, and PD-L2 expression. The cut-off values for high PD-1 (≥ 70 cells/high power field [HPF]), PD-L1 (≥ 100 cells/HPF), and PD-L2 (≥ 100 cells/HPF) were determined by the area under the receiver operating characteristic curve. Expression of PD-1, PD-L1, and PD-L2 was high in 7.9%, 13.2%, and 42.1% patients, respectively. High PD-1, (P = 0.007) and Memorial Sloan Kettering Cancer Center (MSKCC) prognostic scoring (P = 0.019) were independently associated with inferior overall survival on multivariate analysis. High PD-1 also remained an independent prognostic factor for inferior progression-free survival (P = 0.028), as did MSKCC prognostic scoring (P = 0.041) on multivariate analysis. However, there were no differences in survival according to the expression levels of PD-L1/PD-L2 in PCNSL tumor microenvironment. Our results suggest that PD-1 may be considered a biomarker and potential therapeutic target in PCNSL.


INTRODUCTION
Primary central nervous system (CNS) lymphoma (PCNSL) is an extranodal non-Hodgkin lymphoma (NHL) confined to the CNS, mostly diffuse large B-cell lymphoma (DLBCL) [1]. Currently, the International Extranodal Lymphoma Study Group (IELSG) and Memorial Sloan Kettering Cancer Center (MSKCC) prognostic scoring systems are the best available clinical tools to risk-stratify patients with PCNSL [2,3]. However, these prognostic scores do not take into account underlying tumor biological factors, such as tumor microenvironment of PCNSL. Several biomarkers for PCNSL have been suggested, including multiple www.impactjournals.com/oncotarget/ Oncotarget, 2017, Vol. 8, (No. 50), pp: 87317-87328 Research Paper myeloma oncogene 1/interferon regulatory factor 4 (MUM1/IRF4), B-cell lymphoma-6 (BCL-6) and CD68, but their major limitation is that it is difficult to use those biomarkers as a potential therapeutic target [4][5][6]. Therefore, discovering a novel biomarker that supplements IELSG and MSKCC prognostic scoring system and that can be used as a therapeutic target is desperately needed to improve prognostication and survival of patients with PCNSL.
Tumor microenvironment per se is important for PCNSL development and progression as well as other malignancies [7]. This is also supported by the study that showed characteristic infiltration by tumor infiltrating lymphocytes in perivascular microenvironment of PCNSL, which was associated with survival [8]. Albeit the immunological role of T-cell infiltration in PCNSL is unclear, emerging data are suggesting that tumors have developed evasion mechanisms that exploit immune checkpoints to overcome antitumor immunity [9].
The immune checkpoint molecule programmed cell death 1 (PD-1) and its ligands PD-L1 and PD-L2 have been shown to play key roles in inhibiting T cell activity in the tumor microenvironment, not only in solid cancers but also in hematologic malignancies [10]. PD-1 and its ligands have been highlighted because their blockade showed outstanding clinical responses in advanced hematologic malignancies [11].
Importantly, PCNSL exhibited frequent 9p24.1/ PD-L1/PD-L2 copy number alterations [9] and even PD-1 blockade demonstrated clinical activity in relapsed/ refractory PCNSL [12], although only small number of patients were evaluated. Based on these findings, we sought to evaluate PD-1, PD-L1, and PD-L2 expressions in immunocompetent PCNSL patients at diagnosis, as well as their prognostic implication.

Patients' characteristics
Baseline patient characteristics according to PD-1, PD-L1, and PD-L2 expression levels are summarized in

Survival according to the expression levels of PD-1, PD-L1, and PD-L2
Patients with high expression of PD-1 showed significantly shorter 2-year OS and PFS of 33.3% and 0.0%, in comparison to 81.2% and 50.5% for those with low expression of PD-1 (P = 0.008 for OS; Figure 2a Table 3 and associations of PD-1, PD-L1, and PD-L2 expression levels with survival are summarized in Table 4.

Subgroup analysis of patients without upfront ASCT according to PD-1 expression level
As high PD-1 expression was independently associated with inferior survival in our whole patients, we performed subgroup analysis in patients who did not undergo upfront ASCT according to the level of PD-1   expression to consider further confounding variables. In 60 (78.9%) patients who did not undergo upfront ASCT, high PD-1 expression was observed in 5 (8.3%) patients and they tended to associate with inferior survival ( Figure  3a for OS and Figure 3b for PFS) although not statistically significant.

Survival of patients with high PD-1 expression
We further analyzed survival of patients with high PD-1 expression to better characterize the patients with high PD-1 expression in PCNSL tumor microenvironment. The median survival of patients with high PD-1 expression was only 15.8 months (95% CI: 11.7-19.9) for OS and 10.4 months (95% CI: 3.3-17.5) for PFS. However, the median OS of patients with low PD-1 expression was not reached, and median PFS was 24.7 months (95% CI: 9.8-39.5) in patients with low PD-1. All the PCNSL patients with high PD-1 expression had DLBCL histology and two out of six patients showed positive serum EBV DNA among high PD-1 patients. Among six patients with high PD-1 expression, three patients achieved CR1 and five patients achieved CR after completion of primary chemotherapy.
Among these high PD-1 expressing patients, four patients died of disease progression, although four of whom received WBRT consolidation (median 4.0 [range 30.6-54.0] Gy) and one patient received upfront ASCT consolidation. Among four patients with high PD-1 who died of progression, two patients were intermediate-risk, one patient was low-risk, and one patient had missing data according to the IELSG scoring. Regarding MSKCC scoring, two patients belonged to low-risk, and the remaining two patients belonged to intermediate and high-  Table 5.

Correlation of PD-1 expression with PD-L1 and PD-L2 expression levels
The association of PD-1 expression with PD-L1 and PD-L2 expression in tumor microenvironment of PCNSL is shown in Table 6. The level of PD-1 expression was significantly associated with the level of PD-L1 (P = 0.020) and PD-L2 (P < 0.001), respectively (Table 6).

DISCUSSION
In this study, we found that high immunohistochemical PD-1 expression in biopsy specimen of patients with PCNSL at diagnosis was significantly associated with inferior OS and PFS.
Much information has been accumulated regarding the pathogenesis and tumor microenvironment of PCNSL [13]. However, in addition to the rarity of the disease, obtaining adequate numbers of specimens is difficult and thus prognostic roles of studied biomarkers have been unclear. Therefore, IELSG or MSKCC scoring is thus far considered the best available clinical tool for risk-stratifying PCNSL patients [2,3], although biological prognostic markers that also consider the tumor microenvironment and that can be used as a therapeutic target are needed to improve prognostication and survival of patients with PCNSL. Although our group recently reported prognostic importance of CD68 expression in PCNSL microenvironment [6], the major limitation was regarding therapeutic implication as it is difficult to modulate immune response via CD68 positive macrophages in PCNSL tumor microenvironment.
In this regard, PD-1 is an attractive emerging therapeutic target because it has been shown to be expressed in various cancers as well as in hematologic malignancies [14,15]. Regarding PD-1 expression in PCNSL, Berghoff et al. [16] were among the first to demonstrate PD-1 and PD-L1 expression in PCNSL. Of 20 PCNSL patient specimens, 2 (10.0%) showed moderate intensity and 1 (5.0%) had high intensity. Besides, Four et al. [17] reported expression of high PD-1 (2+) in 6.2% of patients with PCNSL. The proportions of high PD-1 expressing patients are similar to our study, as 7.9% of our patients expressed high PD-1. Compared to those studies, our study recruited more number of patients who were initially homogenously treated with HD-MTX-based chemotherapy. We also performed subgroup analysis according to the upfront ASCT, as the role of upfront ASCT in high-risk patients with PCNSL was emphasized [18]. One of our patient with high PD-1 expression died of progression even after receiving upfront ASCT. Interestingly, the poor prognosis of patients with high PD-1 expression was initially indistinguishable using the IELSG or MSKCC scoring systems; because among four high PD-1 expressing patients who died of disease progression, none of them belonged to the high-risk group according to IELSG scoring and only one of them belonged to the high-risk group regarding MSKCC scoring. However, if we knew that high PD-1 expression predicted poor survival in PCNSL, we would have been able to get more help regarding selection of treatment for these high-risk patients.
There were high expressions of PD-L1 and PD-L2 in 13.2% and 42.1% of our cohort, respectively. Regarding PD-L1 expression, our data also showed quite similar rate to that of Four et al. [17], as they reported 18.7% patients were PD-L1 high. However, these rates are relatively low compared to a recent study of PD-L1 expression in    glioblastoma, which is a completely different disease but also the malignant case in the CNS [19]. PD-L1 expression might have been induced by genetic aberrations within tumor cells such as Hodgkin lymphoma and primary mediastinal large B-cell lymphoma, which harbor amplification of 9p24.1, a genomic region that encodes PD-L1 and PD-L2. Interestingly, gain of 9p24.1 was also observed in PCNSL [7]. Indeed, PCNSL is associated with immunodeficiency [1], and PCNSL might evade the immune system utilizing PD-1 pathway [9], resulting in poorer outcome compared to non-CNS disease. Our observation suggests a possible link between high PD-1 expression and poor survival, which could explain the more aggressive behavior associated with high PD-1 expression. Furthermore, we demonstrated significant correlation of PD-1 expression with PD-L1 and PD-L2, which is in line with previous studies [17,20]. Therefore, our study might be used as a basis for future clinical trial targeting PD-1 pathway in PCNSL.
We acknowledge our limitation as the expression of PD-1 and PD-L1/PD-L2 have not been defined in specific cell type due to the lack of double staining with CD20, CD3 or other markers for macrophages. However, expression of PD-1 in tumor infiltrating lymphocytes and PD-L1 in tumor cells were distinguished with microscopic visualization by two independent experienced hematopathologists. Therefore, our claim that high PD-1 expression in tumor microenvironment of PCNSL is associated with poor survival is assuming that high PD-1 expression on tumor infiltrating lymphocytes in PCNSL tumor microenvironment is associated with poor survival of patients with PCNSL at diagnosis.
We also estimated PD-1, PD-L1, and PD-L2 expression visually followed by calculation of the mean  curve for analytic purposes. However, we acknowledge the fundamental limitation to standardize the way of determining the optimal cut-off values for PD-1 and its ligands expression from tumor biopsy samples is difficult. Therefore, the predictive relevance of our method and determining optimal cut-off values remain to be validated in future studies.
In conclusion, we showed that high PD-1 expression in PCNSL tumor microenvironment is significantly associated with inferior survival. Better knowledge of the PD-1 pathway in PCNSL, along with future trials that include PD-1 based biological risk-stratification and therapeutic targeting, are necessary for this challenging disease.

Patients
Seventy-six biopsy-proven PCNSL patients between December 2004 and March 2016 at Severance Hospital, South Korea for whom archived formalin-fixed and paraffin-embedded (FFPE) tissue blocks at initial diagnosis were available were retrospectively analyzed. The patient cohort was previously described [6]. All the patients included in this study were immunocompetent and did not have a history of immunosuppressive drug use nor malignancies other than PCNSL. We excluded human immunodeficiency virus-related PCNSL. The diagnosis was made histologically by surgical resection or stereotactic brain biopsy. All the patients received HD-MTX-based chemotherapy as an initial treatment homogenously without receiving steroid treatment before pathologic confirm.
PCNSL was stipulated as histologically confirmed NHL confined to the CNS [1]. Baseline clinical data were retrospectively collected including age, gender, IELSG prognostic score, MSKCC prognostic score, serum EBV positivity by quantitative polymerase chain reaction, serum human immunodeficiency virus positivity by enzyme immunoassay, type of treatment including consolidative upfront ASCT and WBRT, initial response to treatment, and survival.
The ECOG performance status was determined at the time of diagnosis. Pre-treatment evaluation included contrast enhanced magnetic resonance imaging of the brain, positron emission tomography-computed tomography to exclude systemic NHL, bone marrow aspiration and biopsy with histologic, cytologic, and immunocytologic examination, ocular examination including a slit lamp examination to distinguish ocular involvement, and lumbar puncture for cerebrospinal fluid (CSF) analysis, unless contraindicated. Elevated CSF protein concentration was defined as a level more than 45 mg/dL in patients younger than 60 years of age, and a level more than 60 mg/dL in patients older than 60 years of age [2]. Involvement of deep brain structures was defined as involvement of the