Concordance of Programmed Death-Ligand 1 Expression between SP142 and 22C3/SP263 Assays in Triple-Negative Breast Cancer

Triple-negative breast cancer (TNBC), defined by the lack of estrogen receptor and progesterone receptor expression and the absence of HER2/neu (ERBB2) gene amplification, accounts for approximately 10%–20% of invasive breast cancers [1, 2, 3]. TNBC represents a clinically and molecularly heterogeneous group. Currently, TNBC lacks clinically approved targeted therapies and is treated with traditional chemotherapy. TNBC represents a major clinical challenge due to its aggressive and metastatic behavior. Therefore, therapeutic strategies are needed to improve TNBC patient management.

Immunotherapy has changed the treatment landscape for melanoma and non-small cell lung cancer (NSCLC). Currently, nivolumab is used for NSCLC in the second-line treatment, pembrolizumab is used for NSCLC with high programmed death-ligand 1 (PD-L1) expression (≥ 50%) in the first-line treatment or in the second-line treatment for tumors with ≥ 1% PD-L1 expression, and atezolizumab is used for all subtypes of NSCLC in the second-line treatment [4, 5].

Several studies have reported that 20%–58% of TNBCs express PD-L1. Higher expression of PD-L1 has been observed in TNBC than in non-TNBC [6, 7, 8, 9, 10, 11, 12]. Recently, the Food and Drug Administration (FDA) approved the use of atezolizumab and nab-paclitaxel chemotherapy for the first-line treatment of patients with locally advanced and metastatic TNBC [13]. Accelerated approval was granted based on Impassion130 (NCT02425891), a multicenter, international, double-blinded, placebo-controlled, randomized trial that included 902 patients with unresectable, locally advanced or metastatic TNBC [14]. Furthermore, several clinical trials are being conducted with PD-1 inhibitors, such as nivolumab and pembrolizumab, and PD-L1 inhibitors, such as atezolizumab and durvalumab, for TNBC treatment [15, 16].

Therefore, the identification of patients who may benefit from immune checkpoint inhibition in TNBC is a critical issue. PD-L1 expression is an important biomarker for the prediction of the response to anti-PD1 and anti-PD-L1 immunotherapy. Currently, there are multiple PD-L1 immunohistochemistry (IHC) assays to determine the expression of PD-L1 in tumor cells (TCs). Each assay is linked to a specific therapeutic agent. In NSCLC, the Dako PD-L1 IHC 22C3 assay was approved for use with pembrolizumab [17, 18], and the Ventana PD-L1 IHC SP142 assay was approved for use with atezolizumab [19]. The Ventana PD-L1 IHC SP263 assay is in development for use with durvalumab [20].

In TNBC, the FDA also recently approved the Ventana PD-L1 IHC (SP142) assay as a companion diagnostic device for selecting TNBC patients for atezolizumab treatment [14]. However, each PD-L1 IHC assay was found to have many unresolved issues such as the use of different antibody clones, different IHC protocols and platforms, different cut-off values for PD-L1 positivity, and different types of cells in which PD-L1 is assessed (TCs vs. immune cells [ICs]). It is not clear whether the cut-off points for PD-L1 positivity are valuable in predicting the response to immunotherapy in TNBC. Furthermore, there is a question of whether PD-L1 FDA-approved assays have comparable performance. In this study, the analytical performance of the PD-L1 22C3/SP263 assay was evaluated and compared with the SP142 assay for ≥ 1% ICs.

Immune checkpoint inhibitors are emerging as therapeutic options for TNBC [14, 15]. Recently, the FDA approved the combination of atezolizumab and nab-paclitaxel chemotherapy for the first-line treatment of patients with locally advanced or metastatic TNBC with PD-L1 positivity [14]. The Impassion130 trial showed that atezolizumab plus nab-paclitaxel chemotherapy prolonged progression-free survival among patients with metastatic TNBC in the PD-L1-positive group. In PD-L1-positive patients, the response rates were 58.9% with atezolizumab–nab-paclitaxel and 42.6% with placebo–nab-paclitaxel. In this trial, PD-L1 positivity was defined as ≥ 1% of infiltrating ICs in the PD-L1 SP142 assay. Of the 902 patients, 369 (41%) were classified as the PD-L1-positive group, 243 (27%) had low levels at 1% to < 5%, and 126 (14%) had levels in their ICs at ≥ 5%. In contrast, only 81 (9%) of the patients in the study exhibited PD-L1 expression on TCs [14]. In the present study, 52.6% of patients showed PD-L1 expression at a 1% cut-off value, 28.4% at a 5% cut-off value, and 15.8% at a 10% cut-off value using the SP142 assay for infiltrating ICs. On the contrary, only 16.8% of the patients showed PD-L1 expression at a 1% cut-off value in TCs. These findings are in agreement with the expression rate of PD-L1 in the Impassion130 trial. The Ventana SP142 assay consistently showed fewer positive rates in TCs compared with the other assays in NSCLC and urothelial carcinoma [4, 21, 22, 23, 24]. Furthermore, the concordance rate between the TCs and ICs was lower in the Ventana SP142 assay compared with the other assays. Unlike NSCLC and urothelial carcinoma, the high PD-L1 expression rate of ICs found with the SP142 assay may be related to high levels of tumor-infiltrating lymphocytes (TIL) in TNBC. PD-L1 expression in ICs may reflect an association with a TIL-mediated antitumor inflammatory response [25].

Nanda and colleagues [15] reported the results of the first single-agent anti-PD-1 therapy (pembrolizumab) in the metastatic TNBC cohort within KEYNOTE-012. This trial screened 111 metastatic TNBC patients for PD-L1 positivity. PD-L1 positivity was defined as positive staining in ≥ 1% of TCs, which was immunohistochemically assessed using the PD-L1 22C3 assay. Of the 111 patients, 58.6% had PD-L1 positive tumors in this trial [15]. In the present study, 50.5% of patients showed PD-L1 expression at a 1% cut-off value, 17.9% at a 5% cut-off value, and 6.3% at a 10% cut-off value in the 22C3 assay. The PD-L1 expression rate sharply decreased as the cut-off value increased. The PD-L1 22C3 assay had the lowest total positive expression rate at all cut-off values in the present study.

For early-stage TNBC, several studies are testing durvalumab in combination with other agents (NCT02826434; NCT02489448; NCT02685059). PD-L1 expression was assessed on TCs using the PD-L1 SP263 assay. In the present study, 52.6% of patients showed PD-L1 SP263 expression at a 1% cut-off value, 45.3% at a 5% cut-off value, and 31.6% at a 10% cut-off value in the SP263 assay. The PD-L1 SP263 assay had the highest total positive expression rate at all cut-off values in present study.

Identifying which patients will respond to immune checkpoint inhibitors is a significant challenge. This identification has been mainly based on IHC evaluation of PD-L1 expression. However, variations in PD-L1 expression rates have been attributed to different antibody clones, staining protocols and platforms, cut-off values, and scoring algorithms [26, 27, 28]. Different PD-L1 IHC assays have been developed and approved in parallel with different therapeutic agents, with different cut-offs determined according to clinical response. The PD-L1 expression rates of TCs and ICs were diverse in previous studies on TNBC as well [6, 14, 25]. The PD-L1 expression of TCs was 19% at a 5% cut-off with clone 5H1 [29] and 64%–80% at a 1% cut-off with clone E1L3N [6]. Sun et al. [30] reported that PD-L1 expression was significantly higher in ICs with all 3 PD-L1 assays including 28-8, E1L3N, and SP142 and varied at different cut-off values. In the present study, the PD-L1 expression rate ranged from 49.0%–51.0% with a 1% cut-off value, 17.3%–43.9% with a 5% cut-off value, and 6.1%–30.6% with a 10% cut-off value.

The analytical performance of inter-PD-L1 assays should be comparable to allow for the appropriate interpretation of the use of PD-L1 diagnostic assays in selecting immune checkpoint inhibitors. Therefore, the SP263 and 22C3 assays were evaluated for performance in comparison with the PD-L1 SP142 assay, the currently FDA-approved assay. The concordance rate between assays was highest at a 1% cut-off value and decreased when the cut-off value increased. This may be due to the pathologist's tendency to interpret a result as positive at a 1% cut-off, even with low-level intensity and slight expression (1+), and their tendency to make more subjective interpretations as cut-off values increase. This finding was similar to those previously reported [30]. The concordance rate between the SP142 and SP263 assays at a 1% cut-off was high, while the concordance rate between the SP142 and 22C3 assays at a 1% cut-off was relatively lower. Relatively high degrees of agreement (OPA: 85.3% at a 5% cut-off value, 84.2% at a 10% cut-off value) were observed between the 22C3 and SP142 assays at cut-off values of 5% and 10%, likely due to the high NPAs (NPA: 97.1% at a 5% cut-off value, 96.3% at a 10% cut-off value). The concordance and discordance between the SP142 assay (≥ 1% of ICs) and the 22C3/SP263 assays were analyzed in detail. At a 1% cut-off value for the SP263 and 22C3 assays, few cases were exclusively positive by a specific assay, but relevant amounts of cases were exclusively classified as positive by the SP142 assay at a 5% cut-off value and the SP263 and 22C3 assays at a 10% cut-off value. Considered together, these data indicate that the 22C3 assay at a 1% cut-off value exhibits comparability with the SP142 assay for the clinically relevant cut-off, but it could not be used interchangeably, while the SP263 assay at a 1% cut-off value is interchangeable with the SP142 assay at the clinically relevant cut-off.

A limitation of this study was the potential difference in the PD-L1 expression between the TMAs and the whole cancer tissue sections, although the TMAs were constructed by collecting 2 cores for each case to limit the impact of the heterogenous expression of PD-L1. Furthermore, an analysis comparing the therapeutic responses to the immune checkpoint inhibitors could not be performed. Therefore, further studies on TNBC are needed to define the clinically relevant cut-off value based on therapeutic response.

This is the first study to compare the analytical performance of PD-L1 22C3/SP263 diagnostic assays in comparison with the recently FDA-approved SP142 assay for ≥ 1% ICs in TNBC.

In conclusion, this study demonstrates that although the analytical performance of the 22C3 assay at a 1% cut-off value is comparable to but not interchangeable with the PD-L1 SP142 assay at the clinically relevant cut-off, the analytical performance of the SP263 assay at a 1% cut-off value shows interchangeability with the PD-L1 SP142 assay at the clinically relevant cut-off.