Precision in PD-L1 Assessment in Diffuse Large B Cell Lymphoma: Greater Biological Insight Using in Situ Hybridization Approach

PD-L1 expression in diffuse large B-cell lymphoma (DLBCL) has been associated with the activated B-cell (ABC) type, gains/ amplifications at the PD-L1/ 9p24 locus, and worse clinical outcome in the setting of conventional (R-CHOP) chemotherapy. Most of the prior studies have evaluated PD-L1 expression by immunohistochemistry (IHC), utilizing a variety of antibody clones and with various cut-offs for determining positivity. IHC methods for evaluating PD-L1 expression are inherently suboptimal due to the differing sensitivities/ specificities of the various clones and staining platforms and the lack of well-established interpretive criteria. Moreover, quantifying levels of PD-L1 expression by percent of positive cells and/or staining intensity are difficult and highly subjective. Alternative, highly sensitive and quantitative approaches for studying PD-L1 expression are available (e.g. gene expression profiling, flow cytometry), but these methods lack the ability to distinguish PD-L1 expression in lymphoma cells from those of the surrounding immune microenvironment.

RNA-based, in situ hybridization on formalin-fixed tissues is a sensitive, specific and more quantitative approach for determining levels of expression on morphologically preserved tissue. We analyzed 69 of DLBCL cases by RNAScope in situ hybridization (Advanced Cell Diagnostics) on an automated Leica platform. PD-L1 expression was quantitated in lymphoma cells using established scoring guidelines from 0 to 4 , where 0 corresponds to no RNA expression detected or only one hybridization signal and 4 corresponds to 10 or more signals per cell. From 69 cases of DLBCLs studied 45% had the score of 0, 17% score of 1, 23% score of 2, 6% score of 3 and 9% score of 4. Results were correlated with IHC (using a variety of commercial antibodies), cell of origin classification and gains/ losses at the 9p24 locus (PD-L1 locus).

We observed that PD-L1 expression quantified by RNAScope highly correlated with IHC (p<0.0001, Chi-square test) method. In the lower range of PD-L1 expression, RNAscope had higher sensitivity and dynamic range with 23 cases demonstrating no PD-L1 expression using IHC (score of 0) revealing low levels of amplification using RNAscope: with 15 cases at score 0, 7 cases at score 1 and 1 case at score 2. We also observed that high RNAScope scores categorized as high (3,4) were strongly correlated with non-GCB type and also with samples showing gain of 9p24 (p=0.0061 Chi-square test, p=0.0002, Chi-square test respectively). Interestingly, PD-L1 expression based protein detection using IHC correlated with EBV infection (p=0.008 Chi-square test), while RNAscope score demonstrated no significant correlation (p=0.6 Chi-square test). Comparison of IHC scores and RNAscope scores in EBV positive cases demonstrated that PD-L1 overexpression was not linked to 9p24 amplification, thus suggesting a different biological mechanism of PD-L1 overexpression in EBV-associated cases.

These findings shed additional light into potential mechanism of PD-L1 overexpression in DLBCLs and offer a complimentary and highly sensitive practical approach of assessing PD-L1 overexpressing tumors in FFPE tissue.