We chose 3 representative clone of PD-L1 antibody for IHC assay comparison, 405.9A11 (9A11) was first used by Ansell et al[8] in their outstanding study in CHL, 22C3 was considered as a sensitive assay[38] and was the first FDA approved and most widely used Ab recently[18] ,and SP142 were the 1st clinically validate assay for both TC and IC[37, 39]. The RNAscope assay, which was an antibody independent assay on FFPE sections, was recently invented to detect PD-L1 expression at mRNA level and applied recently in breast, lung and gastric tumors [24–26]. Compared to IHC, its specificity, reproducibility and interpretative objectivity has been reported in gastric cancer [25, 26]. Because both HRS cells and reactive IC cells in CHL can express PD-L1 mainly in a membrane pattern, sometimes the PD-L1 positive IC cells are so crowded around the HRS cells, it might be difficult to tell whether the HRS cells expressed PD-L1 or not (as showed in Fig. 3A). In this situation, the PD-L1 mRNA detected by RNAscope as a “dot-like” pattern distributed in the cytoplasm of HRS cells is much easier identified and better evaluated (Fig. 3C). RNAscope assay was successfully applied in our CHL cases and a provisional scoring system was also developed.
After compared with RNAscope assay, for PD-L1 expression on HRS cells, we proved that 9A11 was the most accurate IHC assay, showing nearly perfect correlation with RNAscope assay. 9A11 is a PD-L1 Ab clone binding cytoplasmic domain and thus was more selective for membranous PD-L1, with stronger, more membrane and less cytoplasmic staining [46], made it easier for 9A11 to distinguish the membranous staining of the HRS cells with surrounding ICs for higher intensity and specificity(shown in Fig. 3A).
For PD-L1 expression on HRS cells, nearly half of the cases expressed high level protein or mRNA (48.1% detected by IHC with clone 9A11 and 45.7% detected by RNAscope), and was nearly the same to the results published recently from Veldman et al.[16], although much lower than R/R cases showing 100% positive from Ansell [8]. But for the study from Roemer et al[12], alterations of the 9p24.1 gene encoded PD-L1 in HRS cell of CHL included copy gain (56%) and amplification (36%), patients with amplification was found with significant increased PD-L1 expression and shorter PFS[12]. However, only 4 out of 10 cases were found with gene amplification from the R/R CHL series of Ansell[8]. Discordance in PD-L1 positive rate might be caused by different evaluation criteria among researchers.
While for PD-L1 expression on ICs, in the study from Veldman et al,69% CHL cases was positive, with Ab clone E1L3N [16]. In our study, we found the PD-L1 positive rates on ICs were 85.2% (9A11), 68.5% (SP142) and 74.1% (22C3) respectively, and were most likely to be expressed by TAMs proved by statistical analysis. The majority of CHL cases express high level of PD-L1 on ICs, however, it should take note that different antibody clone would affect the positive rate. 9A11 was most sensitive, SP142 assay was lowest in sensitivity, as reported previously in lung cancer by Tsao et al [38].
In our study, PD-L1 expression, either on tumoral HRS cells or on ICs, showed poor concordance with clinicopathological factors including gender, age morphological subtypes, and even with EBER infection status. Although it was reported by Green et al [1] that EBV-positive CHL would up-regulate PD-L1 expression [1, 40]. But Paydas et al. did not find this association [47] ,which was supported by our study. In another word, high level of PD-L1 expression was so prevalent in CHL, that it was impossible caused solely by EBV.
There were plenty of T-cells and TAMs s subclassified by IHC makers in the microenvironment of our series of CHLs. However, PD-L1 expression status on HRS cells (by 9A11 assay) correlated poorly with the density score of Ths (CD4+), Tregs (FOXP3+), CTLs (CD8+) and TAMs (CD163+). Increase level of PD-L1 expression on ICs correlated best with higher density of CD163 + TAMs, same finding was seen in a study from lung cancer [33], the percentage of PD-L1-positive ICs was significantly higher in TAM-high group than in TAM-low group. In addition, PD-L1 expression on TAM may be derived from HRS cells [27]. There were controversial results about PD-1 + cell densities [9], from small [48] to huge amount [49]. In our study, only about 1/4 cases presented with “high” Level of PD-1 positive cells (cut off value ≥ 10%), in those cases with “high-level(Positive)” of PD-L1 on HRS cells, the PD-1 positivity rate decreased significantly, indicating a reverse correlation to stop over activation of PD-1 pathway and keep the immunosuppressive balance in CHL. Although there were various amount of FOXP3 + Tregs as reported previously [50, 51], different from other tumors[31, 32, 52], they were not associated with PD-L1 expression level (either on HRS cells or ICs) in our series.
Patients with higher density (> 25%) of TMA was adverse predictor of clinical outcome by univariate survival analysis, which was in concordance with previous report [34]. In concordance of the closest relationship between CD163 positive TAM density, patients with higher level of PD-L1 on the ICs was associated with apparently worse OS. In fact, the CD163 positive TAM represent M2 macrophage[34], increased number of tumor-associated macrophages was strongly associated with shortened survival[42] and was also found in various tumors including diffuse large B-cell lymphoma[32, 45, 53].