Expression of PD-1, PD-L1 and PD-L2 in Lymphomas in Patients with Pre-Existing Rheumatic Diseases—A Possible Association with High Rheumatoid Arthritis Disease Activity

Simple Summary Immunotherapy blocking programmed cell death protein 1 (PD-1) and its ligands (PD-L1, PD-L2) is less effective in non-Hodgkin lymphoma (NHL) than classical Hodgkin lymphoma. However, NHL is a heterogeneous group and current research seeks to identify subgroups of NHL patients responsive to PD-1 blocking agents. Whether patients with pre-existing rheumatic diseases might constitute such a subgroup is unknown. We investigated the expression of PD-1 and its ligands in lymphoma patients with pre-existing rheumatic diseases. Our key findings include that in patients with pre-existing rheumatoid arthritis (RA) and subsequent diffuse large B-cell lymphoma, an association between RA disease severity and increased expression of PD-L1 in tumor cells was seen. This warrants further studies of the PD-1 pathway in lymphoma in other chronic inflammatory conditions. Abstract Current research seeks to identify subgroups of non-Hodgkin lymphoma (NHL) patients responsive to PD-1 blocking agents. Whether patients with pre-existing rheumatic diseases might constitute such a subgroup is unknown. We determined intratumoral expression of PD-1 and its ligands in lymphoma patients with pre-existing rheumatic diseases. We included 215 patients with rheumatoid arthritis (RA), systemic lupus erythematosus (SLE) or Sjögren’s syndrome with subsequent lymphoma and 74 diffuse large B-cell lymphoma (DLBCL) controls without rheumatic disease. PD-1 and PD-ligand immunohistochemical markers were applied on tumor tissue microarrays. The number of PD-1+ tumor infiltrating leukocytes (TILs) and proportions of PD-L1+ and PD-L2+ tumor cells and TILs were calculated and correlated with clinical data. Expression of PD-L1 in tumor cells and TILs was highest in classical Hodgkin lymphoma and DLBCL. In DLBCLs, expression of PD-1 in TILs and PD-L1 in tumor cells was similar in RA, SLE and controls. In RA-DLBCL, high expression of PD-L1 in tumor cells was significantly more common in patients with the most severe RA disease and was associated with inferior overall survival in multivariable analysis.


Introduction
The inhibitory immunoreceptor programmed cell death protein 1 (PD-1) and its ligands PD-L1 and PD-L2 play crucial roles in maintaining immunological homeostasis, self-tolerance and preventing autoimmunity, mainly by suppressing T-cell activity. Many malignancies, including lymphomas, can exploit the PD-1 pathway by expressing PD-L1, leading to evasion of host anti-tumor immune response. In classical Hodgkin lymphoma (cHL), PD-L1 is commonly overexpressed by Hodgkin and Reed-Sternberg (HRS) cells [1], Figure 1. Flowchart of patients with pre-existing rheumatoid arthritis, systemic lupus erythematosus or Sjögren's syndrome and subsequent lymphoma development included in the study with complete clinical information and with sufficient tumor tissue for tissue microarray construction. * Of 85 originally identified DLBCL patients without any rheumatic disease, specimens suitable for TMA construction remained in 74 patients used as controls for comparison with the rheumatoid arthritis-and systemic lupus erythematosus-DLBCL cases.

Clinical Information
Detailed clinical information of both the rheumatic disease and lymphoma was collected from medical records in previous studies [26][27][28]. For RA patients, disease activity had been individually scored repeatedly over time based on swollen and tender joint counts, erythrocyte sedimentation rate (ESR), and the treating physician's global assessment of disease activity from onset of RA until lymphoma diagnosis, as described in detail previously [26]. Values of cumulative disease activity for each patient were calculated as the area under the curve (AUC). In order to investigate a possible impact of RA disease severity on the expression of PD-1 and its ligands, RA patients with DLBCL were divided into two groups depending on their AUC values, using the upper quartile as the cut-off for the highest disease activity group. Active treatment for the rheumatic disease was defined as the treatment at the time of lymphoma diagnosis, provided it had been used for four or more consecutive weeks.

Clinical Information
Detailed clinical information of both the rheumatic disease and lymphoma was collected from medical records in previous studies [26][27][28]. For RA patients, disease activity had been individually scored repeatedly over time based on swollen and tender joint counts, erythrocyte sedimentation rate (ESR), and the treating physician's global assessment of disease activity from onset of RA until lymphoma diagnosis, as described in detail previously [26]. Values of cumulative disease activity for each patient were calculated as the area under the curve (AUC). In order to investigate a possible impact of RA disease severity on the expression of PD-1 and its ligands, RA patients with DLBCL were divided into two groups depending on their AUC values, using the upper quartile as the cut-off for the highest disease activity group. Active treatment for the rheumatic disease was defined as the treatment at the time of lymphoma diagnosis, provided it had been used for four or more consecutive weeks.

Lymphoma Subtyping and Analysis of Epstein-Barr Virus Status
DLBCLs were classified as either germinal center B-cell-like (GCB) or non-GCB according to the Hans algorithm [32]. The presence of Epstein-Barr virus (EBV) in lymphoma tissue was analysed using EBV-encoded RNA (EBER) in situ hybridization [33].

Tissue Samples
Formalin-fixed, paraffin-embedded (FFPE) diagnostic biopsies were gathered and TMAs containing two cores of 1 mm in diameter per case were constructed using standard techniques [26,31].

Immunohistochemical Stainings
To identify PD-1, mouse monoclonal antibody (mAb) NAT105/ab52587 (Abcam, Cambridge, UK) was used, citrate buffer was used to retrieve the antigen and positive signals were envisioned with MACH3 mouse HRP reagents (brown) (Biocare Medical, Walnut Creek, CA, USA). Rabbit mAb E1L3N/13684 (Cell Signaling Technology, Danvers, MA, USA) was used to identify PD-L1 and rabbit mAb D7U8C/82723 (Cell Signaling Technology) was used to identify PD-L2. As described in previous studies, E1L3N/13684 [34] and D7U8C/82723 [35] detect membranous expression of tumor cells and immune cells, respectively. In order to accurately distinguish between PD-ligands expressed on tumor cells or TILs, double immunostaining for PD-ligands and the B-cell biomarker PAX-5 was performed. Mouse mAb M7307/DAK-Pax5 (Dako, Santa Clara, CA, USA) was used to identify PAX-5. Antigens for PD-L1, PD-L2 and PAX-5 were retrieved in Tris-EDTA buffer (pH 9.0) in a pressure cooker. Positive signals for PD-L1 and PD-L2 were envisioned with the Betazoid DAB detection kit (brown) (Biocare Medical), and PAX-5 was detected using Warp Red chromogen (red) (Biocare Medical). Lastly, the sections were counterstained with Intellipath FLX hematoxylin. All antibodies were diluted 1:50.

Evaluation of Immunohistochemical Stainings
The TMA slides were scanned in high-power magnification (×400). Each TMA core was 1 mm in diameter and consisted of an area of 0.785 mm 2 (radius * radius * π=area) (0.5 * 0.5 * π=0.785) and the total area for 2 cores was 1.57 mm 2 . The scanned slides were viewed and analyzed with the Visiomorph software (Visiopharm ® , Hørsholm, Denmark). Different image analysis application algorithms were constructed for PD-1, PD-L1 and PD-L2. First, representative positive and negative areas were marked manually to train each application. The applications were trained to designate cells with a brown membranous staining of all intensities (PD-1, PD-L1 or PD-L2) as positive, while cells without brown membranous staining were designated as negative. In addition, for PD-L1 and PD-L2, the software was manually trained to designate cells with a red nuclear staining with PAX-5 as tumor cells, and PAX-5 negative cells as leukocytes. Next, using a TMA workflow application with each image analysis application applied, the software calculated the number of positive and negative cells for each application and calculated numbers and proportions of positive cells. The numbers and proportions for each marker were calculated as an average count of the two cores from that case. Areas with fibrosis, artifacts or reactive germinal centers were manually excluded from the analyses. Proportions of PD-L1 or PD-L2 positive leukocytes was calculated as the number of positive leukocytes divided by the total number of leukocytes (positive and negative). Proportions of PD-L1 or PD-L2 positive tumor cells was calculated as the number of positive tumor cells divided by the total number of tumor cells (positive and negative). In order to present results comparable with manual evaluations with a microscope, we present the number of PD-1 positive cells per 400× high power field (HPF) (0.0625 mm 2 ). The ratio between the area of the 2 cores and the area of 1 HPF was calculated (1.57/0.0625=25.13). Thus, to calculate the number of PD-1 positive cells per HPF, the number of positive cells in the 2 cores were divided by 25.13 for each case. In order to compare our findings with previous studies on cHL [3], PD-1 in TILs was also calculated as a proportion of positive cells in patients with cHL. On morphological examination, PD-1 was mainly expressed by lymphocytes ( Figure 2G). cHL [3], PD-1 in TILs was also calculated as a proportion of positive cells in patients with cHL. On morphological examination, PD-1 was mainly expressed by lymphocytes ( Figure  2G).

Statistical Analysis
Optimal cut-off values for high and low expression of PD-1 in TILs and PD-L1 and PD-L2 in tumor cells and TILs were determined by receiver operating characteristic (ROC) curves with Yoden s index calculated for each marker with dead at 12 months as the outcome in RA-DLBCL patients. With this approach, the following cut-off values for high proportion were used: ≥21 PD-1 positive cells/HPF for TILs, ≥17% PD-L1+ and ≥8% PD-L2+ for tumor cells, and ≥10% PD-L1+ and ≥2% PD-L2+ for TILs in patients with DLBCL ( Figure 2). Overall survival (OS) was defined as time from lymphoma diagnosis to the time of last follow up or death of any cause. All patients were followed up for survival until death or 28 August 2020. Survival curves and univariate survival analyses were performed using the Kaplan-Meier method, log-rank test and Cox proportional hazards regression to compare differences between groups. Variables of statistical significance (p < 0.05) in the univariate survival analyses were included in the multivariable Cox regression model. The proportional hazards assumption was tested and was not validated. Tabulated values were compared using the Chi square or Fisher's exact test. Wilcoxon signed rank-test was used to compare non-tabulated values between groups. Continuous variables were compared using the Spearman Rank Order Correlation test. A p-value of <0.05 was considered significant. Statistical analysis was performed using Rstudio 1.3.1093 and R with the Rcmdr package (R version 3.6.2).

Patient Characteristics
Of the 215 included cases with a diagnosis of rheumatic disease before lymphoma development, 188 (87%) had RA, 18 (8%) SLE and 9 (4%) pSS (Table 1). Of these, 62% were women, the mean age at lymphoma diagnosis was 68 years (range 31-88) and the mean duration from onset of the rheumatic disease until lymphoma diagnosis was 20 years (range 1-59). The most common lymphoma subtype was DLBCL (54%), of which the non-GCB subtype dominated (71%). At the time of lymphoma diagnosis, 39% of the RA-DLBCL patients received active RA treatment defined as a disease modifying anti-rheumatic drug (DMARD) and/or corticosteroids. Twenty-three percent of the RA-DLBCL patients received a DMARD. None were treated with rituximab or any other biological DMARD during the course of RA. Detailed information about the different DMARDs used is presented in Supplementary Table S1. Most RA patients without active RA treatment at lymphoma diagnosis received non-steroidal anti-inflammatory drugs (NSAIDs).

PD-1, PD-L1, PD-L2 and Lymphoma Subtypes
PD-1 was rarely expressed in tumor cells in patients with a pre-existing rheumatic disease (four cases) as well as in DLBCL controls (two cases). PD-1+ TILs were most frequent in FL (median 236 cells/HPF). In cHL, the median proportional expression of PD-1 by TILs was 9%. The highest median proportions of PD-L1+ tumor cells and TILs were observed in cHL (16% and 20%, respectively) and DLBCL (2% and 10%, respectively). PD-L2 expression was rare in tumor cells (nine cases) and generally low in TILs in most lymphoma subtypes (Figure 3A-C and Supplementary Tables S2-S4).  disease (four cases) as well as in DLBCL controls (two cases). PD-1+ TILs were most frequent in FL (median 236 cells/HPF). In cHL, the median proportional expression of PD-1 by TILs was 9%. The highest median proportions of PD-L1+ tumor cells and TILs were observed in cHL (16% and 20%, respectively) and DLBCL (2% and 10%, respectively). PD-L2 expression was rare in tumor cells (nine cases) and generally low in TILs in most lymphoma subtypes ( Figure 3A-C and Supplementary Tables S2-S4).

Clinical and Lymphoma Characteristics in RA-DLBCL
Clinical and lymphoma-related characteristics of the 103 RA-DLBCL cases and the 74 DLBCL controls without rheumatic disease are shown in Table 2. In RA-DLBCL patients, the highest cumulative RA-related disease activity was significantly more common in pa-

Clinical and Lymphoma Characteristics in RA-DLBCL
Clinical and lymphoma-related characteristics of the 103 RA-DLBCL cases and the 74 DLBCL controls without rheumatic disease are shown in Table 2. In RA-DLBCL patients, the highest cumulative RA-related disease activity was significantly more common in patients with high expression of PD-L1 in tumor cells compared to patients with low expression (45% vs. 20%, p = 0.04) ( Table 3). There were significantly more women in the group with the highest RA-disease activity but no difference was seen regarding the proportion of patients with high tumor PD-L1 expression between men and women ( Table 3). There were no significant differences with respect to other clinicopathological variables between the RA patients with the highest disease activity and the group with less severe RA, including EBV status (Table 4). High PD-L1 expression in tumor cells was significantly associated with EBV-positivity and high numbers of PD-1+ TILs were associated with active RA treatment at lymphoma diagnosis, while no other significant associations between expression of PD-1 or PD-L1 and clinicopathological variables were observed (Table 3).

Comparison of PD-1, PD-L1 and PD-L2 in DLBCLs
When comparing the RA-and SLE-DLBCLs with DLBCL controls, we found no significant differences regarding PD-1 in TILs or PD-L1 and PD-L2 in tumor cells (Table 6). High expression of PD-L1 and PD-L2 in TILs was less common in RA-DLBCL patients compared to controls (51% vs. 77%, p < 0.001 and 23% vs. 74%, p < 0.001, respectively). A similar trend for PD-L1 in TILs was seen in SLE. A high proportion of PD-L2+ TILs was significantly more common in SLE-DLBCL compared to RA-DLBCL (92% vs. 24%, p < 0.001) and close to significantly more common compared to DLBCL controls (92% vs. 74% p = 0.057).  Table 5. Relative risk of death from any cause estimated as HRs with 95% CIs and p-values in RA-DL-BCL patients exposed to any lymphoma treatment and DLBCL controls by putative prognostic factors.

Discussion
This population-based study was the first to investigate the tumor expression of PD-1, PD-L1 and PD-L2 in a cohort of patients with pre-existing rheumatic diseases and subsequent lymphoma development. High expression of PD-L1 in tumor cells in RA-DLBCL patients was more common in the RA group with the highest cumulative disease activity, indicating an association between tumor PD-L1 expression and the severity of the rheumatic disease.

RA Disease Severity and PD-L1 Expression in RA-DLBCL
Besides RA disease activity, no other significant difference was seen between the RA-DLBCL patients with the most severe RA and the lower disease activity group regarding analyzed factors possibly affecting PD-L1 expression. Lymphoma stage, EBV status and RA treatment were similar between the groups ( Table 4). The proportion of patients treated with DMARDs and/or corticosteroids did not differ. Only 23 (22%) of all RA-DLBCL patients were treated with a DMARD, several different types were used, and the number of patients treated with each single drug was small. This makes a more significant impact of any individual DMARD than others on the PD-L1 expression in the study group unlikely. The use of proresid was numerically more common in the lower disease activity group, whereas antimalarials were significantly more common in the highest disease activity group, but the numbers of patients were too few to draw any conclusions from this finding (Supplementary Table S1). As previously described, assessment of RA disease activity was based on repeated measures of joint activity, ESR and the physicians judgement of global disease activity, reflecting the cumulative inflammatory burden of the rheumatic disease from onset until lymphoma diagnosis. In a previous study including 51 RA patients, the level of PD-L1 expression on synovial cells in joints of RA patients was associated with factors of more severe disease, including rheumatoid factor positivity, higher C-reactive protein and higher synovitis score (p < 0.001) [36], which support a correlation between RA disease activity and PD-L1 expression. Possibly, RA-associated proinflammatory cytokines such as IFNγ [23], TNF-α [21] and IL-6 [22], which are also involved in the regulation of PD-L1 [24], might contribute to higher expression of PD-L1 in the group with the most severe RA disease. Whether DLBCL patients with severe RA might be a subgroup of patients benefitting from PD-1 inhibitors needs to be addressed by future studies.

Expression of PD-1 and PD-L1 in Relation to Clinicopathological Parameters in RA-DLBCL
High PD-L1 expression in RA-DLBCL tumor cells was associated with EBV-positivity, consistent with findings in previous studies of DLBCL [37,38], and other EBV-associated malignancies including cHL and posttransplant lymphoproliferative disorders [39]. Active RA treatment at lymphoma diagnosis was associated with higher numbers of PD-1+ TILs in the RA-DLBCL group, but no difference was seen regarding PD-1+ TILs between the groups treated with either DMARDs or corticosteroids in monotherapy or both in combination. A previous study in RA patients (n = 44) found an association between increased accumulation of PD-1+ T-cells in synovial fluid and higher inflammatory joint activity (p = 0.043) [40]. Supposedly, patients received RA treatment due to a more severe RA, which could possibly affect PD-1 expression in TILs, although the proportion of patients belonging to the highest disease activity group did not differ between the treated and untreated cases.

PD-1, PD-L1 and Overall Survival in RA-DLBCL and Comparison with Previous Studies
In RA-DLBCL patients that received active lymphoma treatment, high expression of PD-L1 in tumor cells was associated with inferior OS in multivariable analysis, whereas high expression of PD-1 in TILs was associated with superior OS in univariate analysis. Age ≥ 60 years and Ann Arbor stage III-IV were associated with inferior OS in the multivariable analysis, but no significant correlation was seen between these variables and high expression of PD-1 or PD-L1 in RA-DLBCL. High tumor cell expression of PD-L1 might be associated with inferior survival in DLBCL due to immune evasion of the tumor cells, and high numbers of PD-1+ TILs in DLBCL might reflect an immunologically active TME, which is prone to respond to chemotherapy. However, the findings from the survival analyses should be interpreted cautiously, mainly since the patients did not receive modern treatment regimens and survival was markedly poor. In the hitherto largest study of PD-1 and PD-L1 in DLBCL by Kiyasu et al., including 1253 DLBCL patients, high tumor cell expression of PD-L1 was associated with inferior OS (p = 0.0009), while PD-1+ TILs were not associated with OS [37].

PD-1 and PD-L1 in RA-DLBCL, SLE-DLBCL and DLBCL Controls
The expression of PD-1 in TILs and PD-L1 in tumor cells was similar in patients with pre-existing RA, SLE and in DLBCL controls, whereas the expression of PD-L1 in TILs was significantly higher in the DLBCL controls compared to the RA-DLBCL group and close to significantly higher than the SLE-DLBCL group. These findings are somewhat unclear and might be explained by an altered inflammatory tumor milieu in DLBCL patients with RA or SLE. In addition, we cannot rule out that these findings may be related to the fact that the RA-DLBCL material was older than the DLBCL controls, since the age of FFPE material may influence the results of the immunohistochemical markers [41]. However, when older (diagnosed 1965-1986) versus newer (diagnosed 1987-2005) RA-DLBCL cases were compared, no differences were found in expression of PD-1 or PD-L1.

Comparison of PD-1, PD-L1 and PD-L2 between Various Lymphoma Subtypes
Among the various included lymphoma subtypes, the expression of PD-L1 in tumor cells and TILs was highest in cHL. PD-L1 is commonly overexpressed by HRS cells due to frequent genetic alterations in the chromosome 9p24.1 region containing the PD-L1 locus [42] and both PD-1+ and PD-L1+ TILs frequently occur in the TME. Median expression of PD-L1 in TILs, PD-L1 in HRS cells and PD-1 in TILs were higher in the present study compared to a previous study from our research group where 387 cases with cHL were included irrespective of pre-existing rheumatic diseases (20%, 16% and 9% vs. 12%, 0% and 2%, respectively) [3]. Our findings suggest a generally higher expression of PD-1 and PD-L1 in cHL patients with pre-existing rheumatic diseases, possibly induced by a higher inflammatory activity in patients with rheumatic diseases. However, due to the low number of cases in the present study, these findings should be interpreted with great caution. Among the NHL subtypes, DLBCL had the highest expression of PD-L1 in both tumor cells and TILs. Expression of PD-1 in TILs was highest in FL followed by cHL and DLBCL. This is consistent with results from previous studies showing PD-1 expression at high levels on germinal center follicular cells in FL [43].

Strengths and Weaknesses
Strengths of our study include the population-based setting with histopathological review of the cases, a large study group in the context of lymphoma developed in patients with a pre-existing rheumatic disease and the detailed clinical information of the rheumatic diseases including a cumulative disease activity score for RA-DLBCL cases. Furthermore, the high proportion of patients untreated for their rheumatic disease reduces the risk of anti-rheumatic drugs affecting the results. Our digital image analysis approach and the use of PD-ligand/PAX-5 double staining technique allowed a reliable and consistent identification of tumor cells (PAX-5 positive) and leukocytes (PAX-5 negative). Weaknesses include the fact that most of the cases with pre-existing rheumatic diseases were diagnosed with lymphoma before the 1990s, making comparison with the more recent control group difficult and in particular affecting the reliability of the survival analyses.

Conclusions
In conclusion, our study is the first to investigate the expression of PD-1, PD-L1 and PD-L2 in lymphoma patients with a pre-existing rheumatic disease. Our results indicate an association between a more severe RA disease and increased expression of PD-L1 in DLBCL tumor cells. This is of mechanistic pathophysiological interest and might also have implications for the prognosis of the lymphoma. Whether DLBCL patients with severe RA might be a subgroup of patients benefitting from PD-1 inhibitors needs to be addressed by future studies. The findings warrant further studies also in other chronic inflammatory conditions in order to elucidate whether this association is restricted to RA-associated DLBCL or if a more general mechanism related to chronic inflammation could be suspected.