Development of a PD-L1 Complementary Diagnostic Immunohistochemistry Assay (SP142) for Atezolizumab.

Cancer immunotherapies, such as atezolizumab, are proving to be a valuable therapeutic strategy across indications, including non-small cell lung cancer (NSCLC) and urothelial cancer (UC). Here, we describe a diagnostic assay that measures programmed-death ligand 1 (PD-L1) expression, via immunohistochemistry, to identify patients who will derive the most benefit from treatment with atezolizumab, a humanized monoclonal anti-PD-L1 antibody. We describe the performance of the VENTANA PD-L1 (SP142) Assay in terms of specificity, sensitivity, and the ability to stain both tumor cells (TC) and tumor-infiltrating immune cells (IC), in NSCLC and UC tissues. The reader precision, repeatability and intermediate precision, interlaboratory reproducibility, and the effectiveness of pathologist training on the assessment of PD-L1 staining on both TC and IC were evaluated. We detail the analytical validation of the VENTANA PD-L1 (SP142) Assay for PD-L1 expression in NSCLC and UC tissues and show that the assay reliably evaluated staining on both TC and IC across multiple expression levels/clinical cut-offs. The reader precision showed high overall agreement when compared with consensus scores. In addition, pathologists met the predefined training criteria (≥85.0% overall percent agreement) for the assessment of PD-L1 expression in NSCLC and UC tissues with an average overall percent agreement ≥95.0%. The assay evaluates PD-L1 staining on both cell types and is robust and precise. In addition, it can help to identify those patients who may benefit the most from treatment with atezolizumab, although treatment benefit has been demonstrated in an all-comer NSCLC and UC patient population.

The programmed-death ligand 1 (PD-L1) and programmed-death 1 (PD-1) pathway, plays a role in immune-mediated destruction of cancer cells,1,2 and is a pivotal immune checkpoint pathway. Tumors can evade antitumor immune activity by exploiting upregulated PD-L1 expression in the tumor microenvironment. The binding of PD-L1 to its receptors PD-1 and B7.1 downregulates T-cell activation and in turn prevents T-cell–induced cytotoxicity.2,3 Preventing this interaction can lead to enhanced T-cell priming and results in immune cells (IC) attacking and killing cancer cells.

Atezolizumab (TECENTRIQ, Genentech Inc., South San Francisco, CA) is an engineered, humanized monoclonal antibody, which inhibits PD-L1 by blocking its interaction with PD-1 and B7.1, and has shown clinical activity in patients with a variety of solid tumors. By targeting PD-L1, the PD-L2/PD-1 interaction is left intact, potentially preserving immune homeostasis in normal tissues.4,5 As a single agent, atezolizumab has shown durable antitumor responses in patients who are chemotherapy-naïve or have been previously treated for advanced or metastatic non–small cell lung cancer (NSCLC),6–9 urothelial cancer (UC),10,11 renal cell carcinoma,12 triple-negative breast cancer,13 melanoma,7,10,14 and other indications. Atezolizumab has received Food and Drug Administration (FDA)15 approval in the United States for the treatment of metastatic UC and previously treated NSCLC, together with the approval of the VENTANA PD-L1 (SP142) Assay (Ventana Medical Systems Inc., Tucson, AZ) as a complementary diagnostic to aid in the benefit/risk assessment of atezolizumab.

PD-L1 is expressed on different cell types, including tumor cells (TC) and tumor-infiltrating IC.7 PD-L1 expression is found in a wide range of different tumor types, including, but not limited to, those originating in the bladder, breast, colon, lung, and kidney.3,16 Higher PD-L1 expression on TC or IC detected in tumor tissue, using the assay shows an association with increased objective response rates, progression-free survival, and overall survival in patients with NSCLC8 and UC11 receiving atezolizumab.17,18 Importantly, PD-L1 expression on IC independently from TC, is associated with clinical benefit from atezolizumab, as demonstrated in both NSCLC8 and UC.11 Given that PD-L1 expression on IC and TC inhibits naïve and memory T-cell responses,19 these data are consistent with the underlying mechanism of reactivation of a preexisting immune response with inhibition of the PD-L1/PD-1 signaling pathway by atezolizumab and underlay the importance of measuring PD-L1 expression on both TC and IC.

Clinical evidence for PD-L1 as a predictive marker has resulted in a number of PD-L1 immunohistochemistry (IHC) assays being used clinically, with a variety of formats and scoring approaches.6,20,21 IHC is widely used and allows pathologists to assess the expression of PD-L1 in the context of tissue architecture and the tumor microenvironment. Understanding these assays and the interpretation of the results has become acute, given the data from the front-line NSCLC trials for pembrolizumab and nivolumab, in patients with PD-L1 expression. The KEYNOTE-024 (NCT02142738) study evaluating pembrolizumab in a first-line setting for patients with advanced NSCLC and PD-L1 expression on at least 50% of TC (Dako 22C3 assay), demonstrated improved progression-free survival [hazard ratio=0.50; 95% confidence interval (CI), 0.37 to 0.68; P<0.001; median, 10.3 vs. 6 mo] and overall survival (hazard ratio=0.60; 95% CI, 0.41 to 0.89; P=0.005) compared with the standard of care (chemotherapy).17 Conversely, the CheckMate 026 study (NCT02041533) for nivolumab, failed to show significant clinical benefit in patients with advanced NSCLC and PD-L1 expression on at least 1% of TC, using the Dako 28-8 complementary diagnostic assay.18

Here we describe the characteristics of the FDA-approved complementary VENTANA PD-L1 (SP142) Assay that was codeveloped with atezolizumab. The analytic characteristics of the assay including repeatability and reproducibility were verified, as well as the effectiveness of pathologist training on the assessment of PD-L1 staining on both TC and IC using the assay. In addition, we performed dual chromogenic IHC staining to evaluate the characteristics of PD-L1 staining IC in the tumor microenvironment.

MATERIALS AND METHODS

Tissue Specimens and Tissue Processing

Tissue samples were derived from formalin-fixed, paraffin-embedded (FFPE) NSCLC and UC tissues. The tissues were commercially procured tissues for analytic verification as well as tissues from patients enrolled in Genentech-sponsored clinical trials. Decalcified bone metastases and cytology samples were not tested in clinical trials as validation studies in these tissue types have not been performed. Lymph node metastases samples and other metastases were tested in clinical trials. The FFPE specimens (biopsies or resections) from primary and metastatic samples were sectioned at ∼4 μm and mounted on positively charged glass slides. The cut sections were stored at room temperature and stained with the VENTANA PD-L1 (SP142) Assay within 2 months, which was within the established maximum period of antigen stability to retain characteristic assay performance (data on file). A pathologist assessed the patient samples for evidence of viable tumor tissue before evaluating PD-L1 (SP142) staining.

Assay Development and Optimization

The VENTANA PD-L1 (SP142) rabbit monoclonal primary antibody (Ventana Medical Systems Inc.) was optimized for use as a fully automated IHC assay on the BenchMark ULTRA (Ventana Medical Systems Inc., Tucson, AZ) staining platform using the OptiView DAB IHC Detection Kit and OptiView Amplification Kit (Ventana Medical Systems Inc., Tucson, AZ). The assay was optimized for detection of PD-L1 expression in NSCLC, where TC and IC are predictive8 and in UC, where IC is predictive.11 Parameters evaluated during optimization included antibody concentration, antibody diluent, multiple antigen retrieval methods, antibody incubation conditions, and counterstain conditions. In addition, signal amplification using the OptiView Amplification Kit (Ventana Medical Systems Inc.) was tested to evaluate its efficacy in visualizing IC staining. The optimal conditions for staining both TC and IC in NSCLC and UC tissues on the BenchMark ULTRA instrument are outlined in Table 1. Briefly, antigen retrieval was undertaken for 48 minutes, the primary antibody was applied for 16 minutes at 36°C, amplification was done for 8 minutes amplifier/8 minutes multimer, and samples were counterstained for 4 minutes with hematoxylin II and postcounterstained for 4 minutes.

TABLE 1

Assay Parameters Evaluated for Optimization and Optimal Conditions Selected for the Assay

Evaluation of Tumor Samples

Board-certified pathologists conducted all tumor sample evaluations. The workflow for the evaluation of tissue specimens is described in Supplemental Figure 1 (Supplemental Digital Content 1, http://links.lww.com/AIMM/A186). All tissues were stained with hematoxylin and eosin (H&E), rabbit monoclonal negative immunoglobulin as a negative reagent control, and with the VENTANA PD-L1 (SP142) Assay. The H&E staining was performed to determine the adequacy of tumor. A tissue sample was adequate for the VENTANA PD-L1 (SP142) Assay interpretation if it contained at least 50 viable TC. Tumor-associated stroma was not required for TC scoring, but was essential for scoring IC. Prequalified human benign tonsil tissue was used as positive and negative tissue control for each staining run. Tonsil tissue stained with PD-L1 (SP142) was assessed for staining in lymphocytes and macrophages in germinal centers, and scattered PD-L1 staining cells among PD-L1-negative cells in paracortical regions. Tonsil tissue was also assessed for the presence of diffuse staining observed in the reticulated crypt epithelial cells with the absence of staining of superficial squamous epithelial cells. Patient-matched tissue stained for negative reagent control was evaluated for the presence and acceptability of nonspecific background staining. Once the H&E and the control slides were deemed acceptable, the PD-L1 stained slide was assessed.

PD-L1 (SP142) Scoring Method

PD-L1 expression in TC was assessed as the proportion of TC showing membrane staining of any intensity; expression in IC was assessed as the proportion of tumor area occupied by PD-L1-positive IC of any intensity. Only IC staining in the tumor microenvironment was evaluated, including different patterns of staining (aggregates and single cells dispersed among TC) and staining in different immune cell types (lymphocytes, macrophages, dendritic cells, and granulocytes).

Tumor area was defined as the area containing viable TC, their associated intratumoral stroma and contiguous peritumoral stroma (Supplemental Fig. 2A, Supplemental Digital Content 2, http://links.lww.com/AIMM/A187 showing NSCLC example). The staining protocol for IC is shown in Supplemental Figure 2B (Supplemental Digital Content 3, http://links.lww.com/AIMM/A188) (NSCLC only). An IHC-scoring convention was developed to categorize PD-L1 expression in TC and IC (Table 2).

TABLE 2

PD-L1 Scoring for Both TC and IC

Dual Chromogenic IHC Staining

FFPE human NSCLC tissues were sectioned at 4 μm and mounted on positively charged glass slides. Using the BenchMark ULTRA platform (Ventana Medical Systems Inc.), tissue sections were deparaffinized and exposed to antigen retrieval. Sequential staining was performed with the PD-L1 (SP142) primary antibody, OptiView DAB IHC Detection Kit, and OptiView Amplification Kit, followed by CD8 or CD163 primary antibody (Ventana) with UltraView Red detection. Counterstaining was performed with hematoxylin and bluing.

Reader Precision

Interreader and intrareader reproducibility were assessed using 60 UC and 80 NSCLC samples, spanning the range of PD-L1 expression, including samples around the diagnostic cut-offs for UC [IC1 (1% cut-off) and IC2 (5% cut-off)] and NSCLC (TC1/IC1, TC2/IC2, and TC3/IC3). Samples stained with the VENTANA PD-L1 (SP142) Assay were scored twice with a minimum 2-week washout period. For each read, slides were blinded, randomized, and evaluated by 3 pathologists.

Before inclusion in the reader precision study, a consensus score for PD-L1 status was established for each case by a separate group of 3 pathologists. The consensus score served as a reference standard to calculate the positive percent agreement (PPA), negative percent agreement (NPA), and overall percent agreement (OPA). The results from each of the 3 pathologists participating in the precision study were compared with the consensus score for each case. The overall agreement for intrareader average was calculated by averaging the agreement rate between the first and second read scores, across 3 readers, and stratified by the positive/negative group consensus to calculate PPA/NPA.

Repeatability and Intermediate Precision

Precision studies for the VENTANA PD-L1 (SP142) Assay in NSCLC and UC samples were completed to demonstrate intraday repeatability, interday precision, interinstrument precision, and interlot precision. For intraday repeatability, consecutively sectioned replicate slides from a minimum of 8 samples were stained with VENTANA PD-L1 (SP142) antibody on a single BenchMark ULTRA instrument in a single day and evaluated for PD-L1 expression. Interday precision was assessed with consecutively sectioned replicate slides from a minimum of 8 samples, evaluated on a single BenchMark ULTRA instrument across 5 nonconsecutive days over a minimum 20-day period. In addition, interinstrument and interlot precision were evaluated with consecutively sectioned replicate slides from a minimum of 18 samples for both NSCLC and UC. The slides were stained using 3 lots of VENTANA PD-L1 (SP142) antibody and 3 paired lots of OptiView DAB IHC Detection Kit and OptiView Amplification Kit, on 3 BenchMark ULTRA instruments. All slides were blinded and randomized and then evaluated for PD-L1 expression in TC and/or IC.

Interlaboratory Reproducibility

Interlaboratory reproducibility studies for the VENTANA PD-L1 (SP142) Assay were conducted to demonstrate the reproducibility of the assay in determining the PD-L1 status in UC and NSCLC tissue samples. In each study, 28 unique samples (equal number of cases above and below the cut-off) were stained at 3 external College of American Pathologists accredited laboratories on 5 nonconsecutive days over a period of at least 20 days. Staining was conducted on a single BenchMark ULTRA instrument at each site, using 1 lot of PD-L1 (SP142) antibody, OptiView DAB IHC Detection Kit, and OptiView Amplification Kit. Before staining, slides were blinded and randomized. At each site, the stained slides were independently evaluated by 2 pathologists. The sample set comprised a total of 420 case slides (140 slides per site, 3 sites) and 840 reads (420 case slides, 2 pathologists) generated from 28 unique samples.

Pathologist Training

Pathologist training was performed at multiple sites for the purposes of clinical trial testing as well as for commercialization. Training consisted of either live or webinar sessions that encompassed review of the comprehensive interpretation guide, didactic training on PD-L1 staining characteristics, and multihead scope session review of example cases or review of respective images. The training was followed by proficiency testing with a set of glass slides from at least 40 cases and a review of discrepancies. Although training of pathologists was conducted for multiple cut-offs, the methods, and results sections of this manuscript focus on IC2 cut-off for UC and TC3/IC3 cut-off for NSCLC given the commercialization in the United States.

Statistical Analyses for Precision Studies

The interreader and intrareader precision was evaluated by PPA, NPA, and OPA. Acceptance criteria were defined as PPA and NPA of at least 85% point estimate for interreader and intrareader precision. Two-sided 95% CI were calculated using generalized estimating equation method to adjust for case correlation.

For interinstrument and interlot precision studies for TC2, IC1, IC2, and IC3 in NSCLC, agreements were assessed by average positive agreement (APA), average negative agreement (ANA), and OPA using pairwise comparison. Two-sided 95% CI were calculated using the percentile bootstrap method. For IC1, IC2 in UC and TC1, TC3 in NSCLC interinstrument and interlot precision studies, and for all intraday repeatability/interday precision studies, agreements were assessed by PPA, NPA, and OPA. The case-level most frequent result served as a reference standard to calculate PPA/NPA/OPA. Two-sided 95% CI were calculated using the Wilson score method. Acceptance criteria were defined as OPA of at least 90% for intraday repeatability and interday precision and OPA of at least 85% for interinstrument and interlot precision.

The overall agreement for interlaboratory reproducibility was evaluated by PPA/NPA/OPA using the Ventana consensus score as a reference standard. Two-sided 95% CI were calculated using the Wilson score method. Acceptance criteria were defined as PPA and NPA of at least 85% compared with the consensus score. In addition, interreader and intersite agreements were assessed by APA, ANA, and OPA using pairwise comparison. Two-sided 95% CI were calculated using the percentile bootstrap method for APA and ANA, and were calculated using the Wilson score method for OPA.

Effectiveness of training was assessed by achieving the predefined training criterion of ≥85% OPA compared with consensus scores for assessment of PD-L1 expression on IC for UC, and TC and IC for NSCLC.

RESULTS

PD-L1 (SP142) IHC Analysis

NSCLC and UC tissues stained with the PD-L1 (SP142) antibody showed staining on TC and/or IC. TC often showed moderate to strong partial or circumferential membrane staining, with PD-L1 (SP142) occasionally associated with cytoplasmic staining. IC staining highlighted a heterogenous population of IC, most of which were morphologically consistent with lymphocytes, macrophages, dendritic cells, and granulocytes. IC staining is characterized by dark brown punctate or linear membrane staining and was often observed as aggregates in the intra or peritumoral stroma. Occasionally, IC staining was also observed in the form of focal or diffuse scattered single cells or small aggregates dispersed among TC. This pattern was seen in association with aggregates in tumor stroma. A range of expression from TC0 to TC3 and IC0 to IC3 in NSCLC tumor tissue is shown in Figure 1.

FIGURE 1

A range of PD-L1 expression on both TC and IC in NSCLC. IC indicates tumor-infiltrating immune cells; NSCLC, non–small cell lung cancer; PD-L1, programmed-death ligand 1; TC, tumor cells.

The use of the OptiView Amplification Kit enhanced the visual contrast of both TC and IC (Fig. 2). This additional signal amplification was particularly desirable to assist with IC scoring, which has not previously been included in any predictive diagnostic algorithm.

FIGURE 2

PD-L1 staining of TC and IC in NSCLC and UC: without amplification in NSCLC (A); with amplification in NSCLC (B); without amplification in UC (C); with amplification in UC (D). IC indicates tumor-infiltrating immune cells; NSCLC, non–small cell lung cancer; PD-L1, programmed-death ligand 1; TC, tumor cells; UC, urothelial cancer.

Assessment of the chromogenic dual staining demonstrated coexpression of PD-L1 and CD8 on T cells as well as PD-L1 and CD163 on macrophages. PD-L1 staining is visualized by a brown color, and CD8 and CD163 staining visualized by red color. Coexpression of PD-L1 with CD8 and PD-L1 with CD163 is shown in Figure 3.

FIGURE 3

Coexpression of PD-L1 SP142 (brown) and CD8 (red) (A) and CD163 (red) (B) in macrophages in the tumor microenvironment of human NSCLC tissue. Images taken at ×20, insets at ×60. Arrows represent regions of cell that express PD-L1 on cell specific marker (CD8=T cell, CD163=antigen presenting cell). NSCLC indicates non–small cell lung cancer; PD-L1, programmed-death ligand 1.

Interday precision in UC samples showed an OPA of 97.0% and 96.0% and intraday repeatability of 98.0% and 100.0% in IC1 and IC2, respectively. Interinstrument and interlot precision studies showed an OPA of 98.2% in IC1 and 99.6% in IC2 (Fig. 4A).

FIGURE 4

Precision of VENTANA PD-L1 staining of UC (A) and NSCLC (B and C) samples. Interreader and intrareader precision of VENTANA PD-L1 (SP142) Assay in UC (D) and NSCLC (E). Interlaboratory reproducibility of VENTANA PD-L1 (SP142) Assay in UC (F) and NSCLC (G). CI indicates confidence interval; IC, tumor-infiltrating immune cells; NSCLC, non–small cell lung cancer; PD-L1, programmed-death ligand 1; TC, tumor cells; UC, urothelial cancer.

Similar to UC samples, high interday precision and intraday repeatability were observed in NSCLC samples across TC1 to TC3 and IC1 to IC3 (Fig. 4B). Additional evaluations of interantibody, interdetection, and interinstrument precision using NSCLC samples also demonstrated high OPAs ranging from 93.4% to 99.6% across TC1 to TC3 and IC1 to IC3 (Fig. 4C).

In UC samples, the interreader precision showed OPAs of 88.8% and 95.8% in IC1 and IC2, respectively. The intrareader precision showed an OPA of 90.9% and 93.6% in IC1 and IC2, respectively (Fig. 4D). In NSCLC samples, interreader precision demonstrated OPAs of 92.7%, 93.8%, and 93.5%, in TC1/IC1, TC2/IC2, and TC3/IC3, respectively. The intrareader precision showed OPA of 94.6%, 93.3%, and 93.6% in TC1/IC1, TC2/IC2, and TC3/IC3, respectively (Fig. 4E).

In the study testing UC samples, the overall agreement (IC1: 92.4%; IC2: 92.8%), intersite agreement (IC1: 91.4%; IC2: 89.6%), and interreader agreement (IC1: 96.4%; IC2: 88.1%) between laboratories showed a high concordance (Fig. 4F). Similarly, NSCLC samples showed high concordance between laboratories in overall agreement (TC1/IC1: 93.5%; TC2/IC2: 91.2%; TC3/IC3; 93.2%), intersite agreement (TC1/IC1: 87.6%; TC2/IC2: 87.6%; TC3/IC3; 91.0%), and interreader agreement (TC1/IC1: 88.4%; TC2/IC2: 90.5%; TC3/IC3; 94.7%) (Fig. 4G).

A total of 52 pathologists at 17 sites were trained on how to interpret UC tissues for the IC2 scoring algorithm, and 31 pathologists were trained on interpreting NSCLC tissues for the TC3 to IC3 scoring algorithm. The pathologists met the predefined training criterion of ≥85% OPA compared with consensus scores for assessment of PD-L1 expression. Pathologists achieved an average OPA of 96.0% for UC samples and 94.0% for NSCLC samples.

DISCUSSION

We developed and optimized the VENTANA PD-L1 (SP142) Assay to assess PD-L1 staining in both TC and IC. This is the first PD-L1 assay to validate and incorporate scoring of IC as part of a scoring algorithm for a predictive assay. Assessment of staining on IC, in addition to TC, is an important consideration, which takes into account the mechanism of action of anti-PD-L1 and anti-PD-1 drugs in the tumor microenvironment. In conventional targeted therapies, such as epidermal growth factor receptor or anaplastic lymphoma kinase inhibitors, the drug molecule directly binds to the TC expressing these markers and kills the cells. In contrast, anti-PD-1 and PD-L1 therapies enable the unleashing of the body’s immune response to allow effector cytotoxic T cells to attack and eliminate TC.

The optimization and validation of the VENTANA PD-L1 (SP142) Assay that detects PD-L1 on both TC and IC, enables a comprehensive capture of PD-L1 status in the patient’s tumor sample. Intratumoral preexisting immunity, reflected by IC infiltration and PD-L1 expression in the tumor microenvironment, influences the efficacy of atezolizumab in NSCLC8 and UC.11 The VENTANA PD-L1 (SP142) Assay thus is able to identify patients for atezolizumab with the strongest clinical efficacy.11

Staining on IC with the assay was further characterized by using dual chromogenic IHC. Coexpression of CD8 and CD163 was observed in IC showing staining for PD-L1. These data were consistent with previous observations.7 Further characterization studies are needed to understand the type, extent, and distribution of IC staining in relation to clinical characteristics including outcome.

The VENTANA PD-L1 (SP142) Assay is highly reproducible in measuring PD-L1 expression on both TC and IC.9,11 A number of precision and repeatability measures were evaluated, the results of which demonstrated high agreement rates. In addition, pathologists at various sites showed high concordance rates achieving 96% for IC in UC and 94% for TC/IC in NSCLC.

There is no specific guidance from the health authorities on the requirements for complementary diagnostic development. Hence, the assay was developed using the guidance for companion diagnostic tests to detect PD-L1 expression on both TC and IC with high specificity and sensitivity. A number of different assays are used in the clinical setting to assess PD-L1 expression. These include the Dako 28-8 pharmDx (complementary Dx) and Dako 22C3 (companion Dx) assays and both are approved by the FDA, for use with nivolumab and pembrolizumab, respectively. However, all of these assays comprise different antibodies, scoring criteria, and clinical cut-offs. Therefore, ongoing assessments to evaluate the similarities and differences between various PD-L1 IHC assays in NSCLC cases are being undertaken.22 Large and comprehensive studies are warranted to better understand the similarities and differences between the assays, especially in terms of clinical outcomes.

CONCLUSIONS

The VENTANA PD-L1 (SP142) Assay is the first PD-L1 assay to clinically validate both TC and IC, which provides a comprehensive assessment of the tumor microenvironment. Multiple precision and repeatability studies demonstrate that the assay is reproducible and precise in detecting PD-L1 expression in both TC and IC in NSCLC and UC tissue samples at multiple cut-offs. In addition, pathologists at various sites achieved high concordance for both cell types indicating the trainability of the assay. Further studies to understand the differences between multiple PD-L1 assays that incorporate clinical outcome are needed.

Supplemental Digital Content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s website, www.appliedimmunohist.com.