Predicting treatment response of patients with extranodal natural killer/T‐cell lymphoma based on levels of PD‐L1 mRNA and soluble PD‐L1

Abstract Appropriate biomarkers may help predict patient response to treatment for extranodal natural killer/T‐cell lymphoma (ENKTL), a subtype of non‐Hodgkin's lymphoma in China. Programmed cell death receptor 1 (PD‐1) and its ligand (PD‐L1) have been investigated in various tumors. However, few studies have addressed expression of PD‐1/PD‐L1 in peripheral blood of ENKTL patients. To identify novel peripheral blood biomarkers for diagnosis and treatment of ENKTL, we retrospectively examined 89 healthy volunteers, 49 patients with ENKTL and 74 patients with diffuse large B‐cell lymphoma treated at West China Hospital from September 2017 to September 2018. Both patient groups showed significantly higher expression of PD‐1 and PD‐L1 on CD4+ T cells, higher levels of PD‐L1 mRNA in peripheral blood mononuclear cells (PBMCs) and higher levels of soluble PD‐L1 in plasma than healthy volunteers (P < .05). In ENKTL patients, levels of PD‐L1 mRNA and soluble PD‐L1 were related to disease stage, level of lactate dehydrogenase, lymphocyte count, and copies of Epstein‐Barr genome in blood. Levels of PD‐L1 mRNA and soluble PD‐L1 were similar between healthy volunteers and ENKTL patients who showed complete remission after treatment, and uni‐ and multivariate analyses identified soluble PD‐L1 as a predictor of treatment response in ENKTL patients. Our results suggest that the levels of PD‐L1 mRNA in PBMCs and soluble PD‐L1 in plasma are useful for ENKTL staging and prediction of treatment response.

activation. 3 This phenomenon has been demonstrated in melanoma, 2 gastric cancer, 4 esophageal cancer, 5 pancreatic cancer, 6 liver cancer, 7 kidney cancer, 8 and ovarian cancer. 9 In fact, the PD-1/PD-L1 signaling pathway helps regulate the peripheral blood immune response in nonsmall cell lung cancer, 10 oral squamous cell carcinoma, 11 gastric cancer, 12 and ovarian cancer. 13 PD-1 level on CD4+ or CD8+T cells in peripheral blood is elevated in patients with Hodgkin's lymphoma 14 and chronic lymphocytic leukemia. 15 Whether PD-1/PD-L1 signaling is involved in ENKTL is unclear. Two studies have reported that elevated levels of soluble PD-L1 (sPD-L1) in serum correlate with poor prognosis in ENKTL. 16,17 Therefore, we examined the expression of PD-1/PD-L1 on circulating lymphocytes as well as levels of PD-L1 mRNA in peripheral blood mononuclear cells (PBMCs) and sPD-L1 in plasma in ENKTL patients.
Our results show that the levels of PD-L1 mRNA in PBMCs and sPD-L1 in plasma are useful for ENKTL staging and prediction of treatment response. The same results were not obtained for patients with another type of lymphoma, diffuse large B-cell lymphoma (DLBCL), suggesting that the two biomarkers may show some specificity for ENKTL.  18 Patients were included if they had not received any treatment prior to initial diagnosis. Peripheral blood samples (4 mL) were collected from each patient at the time of diagnosis and sent to the laboratory within 4 hours after collection. During the patient recruitment period, we also recruited 89 healthy volunteers as controls from the general population and collected their blood samples.

| Study population and treatment
All 49 ENKTL patients received chemotherapy, 26 of whom were administered at West China Hospital. From these 26 patients we again sampled peripheral blood after two courses of chemotherapy.

| Flow cytometry analysis
Blood aliquots (500 μL) were transferred to a 1.5-ml Eppendorf tube, 2.3 | Cell separation, RNA extraction, and mRNA reverse transcription Blood aliquots (500 μL) were centrifuged at 3000 rpm to separate plasma from cellular components (lower layer). The upper plasma layer was transferred to a fresh tube and frozen at −80 C for subsequent sPD-L1 assay (see below). PBMCs were separated from the lower cellular components by Ficoll density gradient centrifugation, mixed with 1 mL of Trizol reagent (MRC, Cincinnati, OH, USA) and stored at −80 C until assay. RNA was extracted from thawed samples using a Nanodrop spectrophotometer (Thermo Fisher Scientific, Waltham, USA) and 1 μg of total RNA was reverse-transcribed into cDNA in a 20 μL reaction using of a reverse transcription kit (Takara, Dalian, China).

| ELISA of soluble PD-L1 in plasma
Plasma fractions obtained as described above were thawed and assayed for sPD-L1 using the Human PD-L1 ELISA kits (catalog no. 28-8, Abcam, Cambridge, UK), which has a manufacturer-specified minimal detectable concentration of 2.91 pg/mL. First, samples were added to microplates pre-coated with anti-PD-L1 monoclonal antibodies. After washing out the reagent and unbound antibody, TMB substrate was added to the wells, which were incubated for 10 minutes. Then stop solution was mixed to prevent blue color development, and the color changed to yellow. Absorbance at 450 nm was measured immediately using a microplate reader (R&D, MN, USA). Each test was performed in duplicate.

| Statistical analyses
Data were analyzed using SPSS 24.0 (IBM, Chicago, IL, USA) and Gra-phPad Prism 6.0 software (GraphPad Software Inc., San Diego, CA, USA). Differences between groups in baseline clinical characteristics were assessed for significance using the Mann-Whitney U test. Levels of PD-1 and PD-L1 on CD4+ or CD8+T cells, as well as levels of PD-L1 mRNA and sPD-L1 were expressed as median and interquartile range (IQR). Differences between groups were assessed for significance using the Kruskal-Wallis test. For the subset of patients sampled at diagnosis and after two treatment courses, we assessed differences in levels of PD-L1 mRNA or sPD-L1 using the Wilcoxon matched-pairs signed rank test. Univariate analysis to identify predictors of treatment response was performed using the Mann-Whitney U test for continuous variables or chi-squared test for categorical variables. Variables that were significant in univariate analyses were entered into a binary logistic regression multivariate model. Correlations between variables were analyzed using Spearman rank correlation. Differences associated with a two-sided P < .05 were considered significant.

| Baseline clinical characteristics of patients
Baseline clinical characteristics of patients with ENKTL or DLBCL are summarized in Table 1. In the ENKTL group, 12 (24%) were in stage III or IV, and 16 (33%) had lactate dehydrogenase (LDH) levels>250 U/L.
All patients received at least two cycles of chemotherapy before radiotherapy. Fourteen (29%) patients received anthracyclinecontaining chemotherapy regimens and the others accepted nonanthracycline-containing regimens. After two courses of chemotherapy, 21 patients (43%) achieved CR. We also recruited 89 healthy volunteers (46 men and 43 women) as the control group, whose median age was 41 years (IQR, 33.5-53.5) (data not shown).

| sPD-L1 and relationships with clinical variables
Levels of sPD-L1 (pg/mL) were significantly higher in patients with  Figure 3A]. Levels were significantly higher in ENKTL patients than in DLBCL patients (P < .001, Figure 3A).

| Correlation between PD-1 or PD-L1 biomarkers
Among the ENKTL patients, Spearman rank correlation analysis showed that PD-L1 mRNA levels in PBMCs correlated with PD-L1 levels on CD4 + T cells (P = .0021, Spearman r = 0.4298, Figure 4A) and with sPD-L1 in plasma (P < .001, Spearman r = 0.6671, Figure 4B).   Univariate analysis showed that ENKTL patients with high sPD-L1 levels at diagnosis were more likely to achieve CR (P = .0085,

| DISCUSSION
In this study, we found that levels of PD-L1 mRNA and sPD-L1 decreased to healthy controls after CR in ENKTL patients, suggesting that dynamic monitoring of PD-L1mRNA and sPD-L1 can be used to evaluate treatment response. In addition, patients in our sample who had higher sPD-L1 levels before treatment were less likely to achieve CR, suggesting that sPD-L1 levels can predict treatment response in patients with ENKTL.
Our results are consistent with a study showing that ENKTL patients had higher sPD-L1 levels than healthy controls, and that higher levels correlated with worse prognosis. 16 Our results are also consistent with work showing that ENKTL patients in stage I or II of the disease who had a high concentration of serum sPD-L1 showed lower rate of CR to treatment and worse survival than those with a low sPD-L1 concentration. 17 The present work substantially extends those two studies because we measured biomarker levels before and after treatment in many of our patients, and we found higher sPD-L1 Our study found that PD-L1 was up-regulated on CD4+ T cells in ENKTL patients. PD-L1 can also be up-regulated d on CD4+ regulatory T cells in Hodgkin lymphoma, which then inhibit the function of T cells expressing PD-1. 21 Therefore, we speculate that CD4+ T cells that strongly express PD-L1 may bind to PD-1 on other T cells to down-regulate their response. Levels of PD-1 on CD4+ or CD8+ T cells are increased in the peripheral blood of patients with gastric cancer 12 or chronic lymphocytic leukemia. 15 We found that PD-1 and PD-L1 expression on CD4+ T cells was elevated in ENKTL, which positively correlated with EBV DNA load, similar to a previous study. 22 We further found a significant correlation between the expression of PD-1 and PD-L1 on CD4+T cells in ENKTL patients, and higher PD-1 levels on CD4+ T cells correlated with more advanced high disease, higher ECOG score, higher copy number of the EBV genome, and higher Ki-67 index. These results indicate that in ENKTL, CD4+ T cells expressing more PD-1 are more important drivers of PD and tumor immune escape than T cells expressing less PD-1. We found that PD-L1 was barely expressed on circulating CD8+ T cells in patients with ENKTL or DLBCL and healthy volunteers. A previous study of patients with non-small cell lung cancer also reported low rated of PD-L1 expression ratio on CD8+ T cells (0.08%-8.78%). 23 We found higher PD-L1 mRNA levels in ENKTL patients than in healthy controls, and the same has been reported in chronic lymphocytic leukemia 24 and breast cancer. 25 We further found that these higher PD-L1 mRNA levels varied with disease stage, hemoglobin level, lymphocyte count, lactate dehydrogenase, and number of EBV genome copies. These results suggest that ENKTL patients expressing more PD-L1 on circulating lymphocytes may have weaker systemic immunity and may therefore be more prone to PD.
Our study and several others suggest that sPD-L1 level has prognostic potential. A multicenter clinical study demonstrated that patients with patients with higher plasma PD-L1 level had worse prognosis than those with a lower level. 26 Another study linked high sPD-L1 level with more advanced gastric cancer staging and greater likelihood of lymph node metastasis. 27 We also observed higher sPD-L1 levels in patients with B symptoms, lymphopenia, more advanced disease, higher scores on the international prognostic index, higher Ki-67 index, higher serum levels of lactate dehydrogenase, and lower hemoglobin concentration.
The EBV protein EBNA2 has been shown to up-regulate expression of PD-L1 in B-cell lymphomas. 28 Our study also found that the level of PD-L1 mRNA in PBMCs and levels of sPD-L1 in plasma correlated with EBV genome copy number of in peripheral blood. These findings further support the idea that the virus up-regulates PD-L1.
The source of up-regulated sPD-L1 in ENKTL is unclear. Myeloid suppressor cells have been proposed to be the main source of sPD-L1 in this disease, 29 but we found that sPD-L1 level correlated positively with levels of PD-L1 mRNA in PBMCs. This suggests that sPD-L1 may be produced not only by tumor cells but also by certain non-tumor immune cells.
Although small, our study proposes biomarkers that may facilitate prediction of treatment response in ENKTL. It also justifies further studies to elucidate how the PD-1 signaling pathway is up-regulated in ENKTL.