Analysis of Programmed Death Ligand 1 (PD-L1) Expression in Diffuse Large B Cell Lymphomas (DLBCL) in HIV-Negative Versus HIV-Positive Patients

Introduction:  Through its interaction with PD-L1, the inhibitory receptor programmed death 1 (PD-1) is a key immune “checkpoint” that regulates adaptive immunity in both infectious and neoplastic diseases.  Co-opted expression of PD-L1 by cancers can thus inhibit  endogenous T cell anti-tumor responses, thereby permitting escape from immune destruction, yet offering a novel means of immunotherapy by antibody blockade of the PD-1/PD-L1 axis. Elevated expression of PD-L1 within melanomas and numerous carcinomas has been associated with adverse prognosis.  We previously described the expression of PD-L1 in a subset of DLBCL cases, particularly in those of non-GCB phenotype (Andorsky et al, Clin. Cancer Res. 2011). Human immunodeficiency virus (HIV)-related DLBCLs are a subgroup of DLBCLs with have a more aggressive clinical course, even in the era of combination antiretroviral therapy. HIV-related DLBCLs are often associated with severe immunosuppression, however, the expression of PD-L1 in HIV-related DLBCL is unknown.  In this study, we examined the relationship between PD-L1 expression, HIV status, and clinical outcomes.

Methods:  A total of 135 cases of incident DLBCL diagnosed between 2000-2007 were included from patients at Kaiser Permanente California, an integrated health care system. Of these, 57 were HIV-pos DLBCL patients, and 78 HIV-neg DLBCL patients were selected age, sex and race matched to the HIV-pos cases.  Patients were characterized according to International Prognostic Index (IPI) risk factors, and cell-of-origin (COO; germinal center B [GCB] versus non-GCB) using the Hans classifier.  PD-L1 staining was performed by immunohistochemistry, with dual staining for the PAX5 B cell marker used to distinguish tumor cell versus non-tumor /stromal macrophage expression of PD-L1.  A 10% cutoff point was used to classify cases as positive for PD-L1 expression by tumor cells. Mortality within 2 years of DLBCL diagnosis and cause of death were ascertained. The association between PD-L1 expression and HIV status was evaluated using the t-test. The association between PD-L1 expression and patient survival was examined using multivariable Cox model, adjusting for IPI, COO and HIV status.

Results:  Tumor cell expression of PD-L1 was noted in 7.7% of HIV-neg cases versus 17.5% of HIV-pos cases (p=0.08).  Expression of PD-L1 within the non-tumor stromal macrophages was more common, seen in 70.5% of HIV-neg and 68.4% of HIV-pos cases (p=0.79) using a 10% cutoff for stromal cell expression.  Using a 30% cutoff for stromal cell expression, fewer cases were PD-L1 positive, but the frequency did not differ significantly between the HIV-neg and HIV-pos cohorts (32.1% and 31.6%, respectively, p=0.95).  Clinical outcome, measured by 2 year lymphoma-specific mortality (LSM), was negatively affected by HIV serologic status, being 11.5% in HIV-neg versus 28.1% in HIV-pos cases (p=0.01).  With regards to tumor cell PD-L1 expression, among HIV-neg cases, both all-cause mortality (ACM) and LSM did not appear to correlate with PD-L1 status (p=0.67 and p= 0.51, respectively).  However, among HIV-pos subjects, tumor cell PD-L1 expression was associated with adverse ACM (p=0.05, HR 3.08 [1.03-9.28]) and LSM (p=0.03, HR 4.01 [1.11-8.48]) when adjusted for IPI and COO.  In an additional overall analysis combining all 135 HIV-neg and HIV-pos cases, tumor cell PD-L1 expression was again correlated with worse outcome in terms of ACM (p=0.04, HR 2.38 [1.02-5.56]), and LSM (p=0.03, HR 3.07 [1.11-8.48]).  In contrast, stromal cell PD-L1 expression did not correlate with outcomes in any of these analyses.  

Conclusions:   Expression of the negative T cell regulatory protein PD-L1 on tumor cells was found on a minority of cases of DLBCL, though at least numerically more often in HIV-pos cases than in HIV-neg in this relatively small series.  Association between tumor cell PD-L1 expression and adverse outcome was seen in HIV-pos but not HIV-neg subjects.   Expression of PD-L1 within tumors in HIV-infected subjects could possibly render these tumors more amenable to immune escape than in HIV-pos subjects, or be a marker for more aggressive intrinsic biology of these tumors.  Lastly, since immunotherapy with PD-1 blockade has recently been shown to have efficacy against some B cell lymphomas (Lesokhin et al, ASH 2014), this form of immunotherapy could possibly improve the less favorable outcomes of patients whose tumor cells express PD-L1.