Context.—

With the abundance of therapeutics targeted against programmed death receptor-1 and its ligand (PD-L1) that are currently approved or in clinical development, there is interest in identifying those patients most likely to respond to these drugs. Expression of PD-L1 may be an indicator of an initial and robust inflammatory response to the presence of tumor cells. Therefore, tumors that express PD-L1 may be the most likely to respond to therapies that interrupt the negative feedback mechanism that leads to PD-L1 upregulation.

Objective.—

To develop a prototype immunohistochemistry assay using the anti–PD-L1 antibody clone 22C3.

Design.—

The assay was developed and optimized using commercially available reagents and archival tumor-bank tissue.

Results.—

The optimized immunohistochemistry method had high precision and reproducibility. Using the prototype assay in 142 non–small cell lung cancer and 79 melanoma archival tumor-bank tissue samples, PD-L1 staining was observed at the plasma membrane of nucleated tumor and nontumor cells and, in some cases, as a distinct lichenoid pattern at the tumor-stroma border. Using a preliminary scoring method, 56% (80 of 142) of non–small cell lung cancer and 53% (42 of 79) of melanoma samples were defined as PD-L1+ based on a modified H-score of 1 or more or the presence of a distinctive staining pattern at the tumor-stroma interface.

Conclusions.—

The immunohistochemistry assay using the anti–PD-L1 antibody 22C3 merits further investigation in clinical trials and prevalence assessments to further understand the prognostic and predictive value of PD-L1 expression in cancer.

The programmed death receptor-1 (PD-1) pathway is one of the major immune-control switches engaged by tumor cells to escape from T-cell–mediated immune surveillance.1  The programmed death receptor-1 ligand (PD-L1), which is often expressed by tumor cells,25  binds with PD-1 on T cells and delivers an inhibitor signal to down-regulate T cell proliferation and activation. Its expression has been correlated with poor prognosis and survival in various cancer types, such as renal cell carcinoma,6  pancreatic carcinoma,7  hepatocellular carcinoma,8  ovarian carcinoma,9  melanoma,10  and non–small cell lung cancer (NSCLC).11  Although there are data suggesting a relationship between PD-L1 expression and poor prognosis, there are also reports suggesting no prognostic effect of PD-L1 expression in solid tumors. Several retrospective analyses1214  of NSCLC specimens suggest that PD-L1 expression does not have a positive prognostic effect, and there are also conflicting data regarding the correlation of expression with prognosis and survival in patients with melanoma.10,1517 

Several anti–PD-1 and anti–PD-L1 therapies are currently approved or in development for the treatment of cancer. One of these therapies is pembrolizumab (MK-3475), a potent and highly selective, immunoglobulin (Ig) G4/κ isotype, humanized monoclonal antibody against PD-1. Pembrolizumab is approved in the United States and the European Union for treatment of unresectable or metastatic melanoma and in the United States for metastatic PD-L1+ NSCLC, and it is being investigated in more than 30 tumor types. The ability to reliably and simply measure PD-L1 expression in tumors may identify subsets of patients who might achieve better outcomes with anti–PD-1 or anti–PD-L1 therapeutics. As the initial step toward development of a companion diagnostic, we developed a prototype immunohistochemistry (IHC) assay for measuring PD-L1 expression among patients enrolled in clinical trials to predict those who have a higher likelihood of experiencing a response with pembrolizumab.

Antibody Generation and Specificity Testing

A mouse anti-human PD-L1 IgG1/κ antibody (clone 22C3) was generated at Merck Research Laboratories (Kenilworth, New Jersey) by immunizing mice with a proprietary fusion protein containing the human PD-L1 extracellular domain. Clone 22C3 binding was initially screened using a cell-based, enzyme-linked immunosorbent assay of Chinese hamster ovary cells stably transfected with human PD-L1 as a positive control and nontransfected (parental) Chinese hamster ovary cells as a negative control. A single lot of purified antibody (lot 83AEP) was generated by Aragen Bioscience (Morgan Hill, California) from cryopreserved 22C3 hybridoma cells. As measured by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and size exclusion high-performance liquid chromatography, lot 83AEP had antibody purity of more than 98%. All work described herein was performed with this lot only.

The sensitivity and specificity of clone 22C3 as an IHC reagent was assessed by several orthogonal methods (messenger RNA [mRNA] analysis, flow cytometry, and IHC) using formalin-fixed, paraffin-embedded (FFPE) blocks of NSCLC cell lines (NCIH23, NCIH226, and HOP92) with differential levels of PD-L1 expression. Cell lines used for testing by orthogonal methods included not only the 3 listed above but also A375, HS578T, LOX IV, PD-L1–transfected Chinese hamster ovary, and Chinese hamster ovary parental. These were selected based on levels of PD-L1 mRNA as recorded in a proprietary Merck-internal gene-expression database to cover a broad spectrum of PD-L1 expression at the RNA level (data not shown). As shown in the overview in Figure 1, cells from each line were expanded to generate aliquots of cells for flow cytometric analysis and cell pellets that were FFPE for PD-L1 IHC and mRNA analysis using the NanoString platform (NanoString Technologies, Seattle, Washington). The prevalence and intensity of PD-L1 staining detected for each cell line using the 22C3 antibody was compared with expression of PD-L1 mRNA detected via the NanoString platform (Figure 2) and measured by anti–PD-L1 clone 29E.2A3 (BioLegend, San Diego, California) using flow cytometry, as shown in Figure 3, for the NCIH23 cell line (Figure 3, A), NCIH226 cell line (Figure 3, B), and HOP92 cell line (Figure 3, C). The appropriateness of the 22C3 PD-L1 IHC signal distribution was corroborated using in situ hybridization (RNAScope, Advanced Cell Diagnostics, Hayward, California) in cell lines, as shown for NCIH23 (Figure 4, A), NCIH226 (Figure 4, B), and HOP92 (Figure 4, C). In Figure 5, PD-L1 IHC expression with the 22C3 antibody is shown for healthy human tonsil crypt epithelium (Figure 5, A) and follicular macrophages (Figure 5, B), with the in situ hybridization results shown in crypt epithelium (Figure 5, C) and follicular macrophages (Figure 5, D). Arrows in Figure 5, D, indicate the presence of low copy number mRNA.

Figure 1.

Overview of process for orthogonal validation of anti–programmed death ligand-1 (anti–PD-L1) clone 22C3 for formalin-fixed, paraffin-embedded (FFPE) non–small cell lung cancer. Abbreviation: IHC, immunohistochemistry.

Figure 2. NanoString analysis of anti–programmed death ligand-1 (anti–PD-L1) clone 22C3 on formalin-fixed, paraffin-embedded non–small cell lung cancer cell pellets. Expression of PD-L1 messenger RNA using the NanoString platform for NCIH23, NCIH226, and HOP92 cell lines.

Figure 1.

Overview of process for orthogonal validation of anti–programmed death ligand-1 (anti–PD-L1) clone 22C3 for formalin-fixed, paraffin-embedded (FFPE) non–small cell lung cancer. Abbreviation: IHC, immunohistochemistry.

Figure 2. NanoString analysis of anti–programmed death ligand-1 (anti–PD-L1) clone 22C3 on formalin-fixed, paraffin-embedded non–small cell lung cancer cell pellets. Expression of PD-L1 messenger RNA using the NanoString platform for NCIH23, NCIH226, and HOP92 cell lines.

Close modal
Figure 3.

Flow cytometry analysis of anti–programmed death ligand-1 (anti–PD-L1) clone 22C3 on formalin-fixed, paraffin-embedded non–small cell lung cancer (NSCLC) cell pellets. Identification of PD-L1 and PD-L1+ NSCLC cell lines by flow cytometry with BioLegend (San Diego, California) clone 29E.2A3 for NCIH23 (A), NCIH226, (B), and HOP92 (C).

Figure 3.

Flow cytometry analysis of anti–programmed death ligand-1 (anti–PD-L1) clone 22C3 on formalin-fixed, paraffin-embedded non–small cell lung cancer (NSCLC) cell pellets. Identification of PD-L1 and PD-L1+ NSCLC cell lines by flow cytometry with BioLegend (San Diego, California) clone 29E.2A3 for NCIH23 (A), NCIH226, (B), and HOP92 (C).

Close modal
Figure 4.

Immunohistochemistry analysis of anti–programmed death ligand-1 (anti–PD-L1) clone 22C3 on formalin-fixed, paraffin-embedded non–small cell lung cancer cell pellets. PD-L1 immunohistochemistry expression analysis with anti–PD-L1 clone 22C3 is shown on NCIH23 (A), NCIH226 (B), and HOP92 (C) cell lines (original magnification ×20 [A through C]).

Figure 4.

Immunohistochemistry analysis of anti–programmed death ligand-1 (anti–PD-L1) clone 22C3 on formalin-fixed, paraffin-embedded non–small cell lung cancer cell pellets. PD-L1 immunohistochemistry expression analysis with anti–PD-L1 clone 22C3 is shown on NCIH23 (A), NCIH226 (B), and HOP92 (C) cell lines (original magnification ×20 [A through C]).

Close modal
Figure 5.

Programmed death ligand-1 (PD-L1) expression in healthy human tonsil crypt epithelium (A and C) and follicular macrophages (B and D) assessed by immunohistochemistry with the 22C3 antibody (A and B) and in situ hybridization (C and D), with the arrows indicating the presence of low copy number messenger RNA signal in follicular macrophage (D) (original magnification ×60 [A through D]).

Figure 5.

Programmed death ligand-1 (PD-L1) expression in healthy human tonsil crypt epithelium (A and C) and follicular macrophages (B and D) assessed by immunohistochemistry with the 22C3 antibody (A and B) and in situ hybridization (C and D), with the arrows indicating the presence of low copy number messenger RNA signal in follicular macrophage (D) (original magnification ×60 [A through D]).

Close modal

PD-L1 Immunohistochemistry Assay Optimization

Multitissue “sausage” blocks,18  also known as multitumor blocks, were used for assay optimization. Multitumor blocks include larger pieces of tissue than those found in typical tissue microarrays but still permit the inclusion of several FFPE samples on a single slide. In addition to human tonsil FFPE samples, used as controls, multitumor blocks used for assay optimization included 16 NSCLC samples (8 adenocarcinoma and 8 squamous cell carcinoma samples) and 8 melanoma samples. The 22C3 antibody concentration, antibody incubation time, antigen-retrieval reagents and method, and the antibody detection system were all tested as part of the optimization process. The 22C3 concentrations tested ranged from 0.5 to 10 μg/mL (to convert to micromoles per liter, multiply by 0.00667), and the antibody incubation times ranged from 1 hour to overnight. Multiple antigen-retrieval methods were tested, including heating in citrate-based (acid pH or neutral pH) or Tris and/or chelator-based (basic pH) buffers, either alone or in combination with digestion by a weak protease (proteinase K; Dako, Carpinteria, California). The different methods of antigen-retrieval tested included steam, microwave, and an antigen-retrieval module. Both biotin-dependent and biotin-independent antibody-detection systems were tested. Progressive, iterative steps were employed, based on the results of prior staining runs, to identify the conditions that demonstrated accurate cellular localization of PD-L1, a broad dynamic range of PD-L1 expression, an appropriate signal-to-noise ratio, and acceptable performance in positive and negative tissue controls.

Other than the 22C3 antibody, all reagents tested were commercially available. Mouse IgG1/κ (Dako) was used at the same concentration as 22C3 to determine any nonspecific (ie, antibody constant region) staining inherent in the detection reagents or tissues or arising in tissues; mouse IgG1/κ does not control for the unique PD-L1 antigen-binding region of clone 22C3. Temperature and maintenance logs were maintained for all refrigerators and freezers used in storing assay components. Lot numbers and expiration dates were recorded for all assay components.

Optimized PD-L1 22C3 IHC Protocol

The FFPE tumor tissue sections of 4- to 5-μm thickness were cut onto positively charged slides (Fisher ProbeOn Plus, Thermo Fisher Scientific, Pittsburgh, Pennsylvania), baked at 60°C (dry heat) for 45 minutes less than 24 hours before use, deparaffinized in 4 changes of 100% xylene, and rehydrated with a graded ethanol series (100%, 70%, and 30%) to distilled water.

Prepared slides were incubated for 20 minutes at more than 90°C in EnVision FLEX+ low-pH target-retrieval solution (Dako), using a commercial steamer as the heat source. After cooling for 5 minutes, automated staining was performed using a TechMate 500 or 1000 automated IHC staining platform (Roche Tissue Diagnostics, Tucson, Arizona) and WorkMate software, version 3.96 (Aveva, Cambridge, England). This automated platform uses a capillary gap process19  for all reagent changes, including antibody incubation, detection steps up to and including counterstaining, and intervening washes. All procedures were carried out at room temperature (25°C). On the TechMate instrument, antigen retrieval was further advanced by incubating cells for 10 minutes in proteinase K (1/160 dilution in EnVision FLEX+ wash buffer). Following a 10-minute incubation with EnVision FLEX+ peroxidase block, slides were incubated with the anti–PD-L1 antibody clone 22C3 at a concentration of 2 μg/mL in a commercially available primary-antibody diluent for a mean (SD) of 16 (1) hours in a dark, humidified chamber.

The EnVision FLEX+ polymer kit, which is biotin independent and reduces the potential for background or nonspecific staining from endogenous biotin, was used for primary antibody detection. The steps included were a 15-minute incubation with EnVision FLEX+ Mouse Linker, a 25-minute incubation with EnVision FLEX+ horseradish peroxidase polymer, a 10-minute incubation with EnVision FLEX+ diaminobenzidine (DAB) chromogen, and a 10-minute incubation with a Ni2+Cl2-DAB enhancer. Between all incubation steps, slides were extensively washed with EnVision FLEX+ wash buffer. The slides were counterstained with hematoxylin for 1 minute, rinsed in distilled water, dehydrated off platform in an ethanol series (95% and 100%) and 4 changes of 100% xylene, and permanently coverslipped (Cytoseal XYL mounting media, Thermo Fisher).

Assessment of PD-L1 Expression in Tumor Samples

The optimized PD-L1 IHC prototype assay was used to evaluate PD-L1 expression patterns in archival FFPE tissue samples of NSCLC (n = 142) and melanoma (n = 79) obtained from a tumor bank, along with hematoxylin-eosin–stained and negative reagent-control slides as an aid to the PD-L1 scoring. A board-certified pathologist scored results in the total area of viable tissue section available; areas of necrosis or obviously poorly fixed areas of tissue were not evaluated. Based on the pathology observations (described in the Results), the ultimate scoring method included a quantitative assessment of PD-L1 positivity within tumor nodules, as well as the presence or absence of other recognizable staining patterns at the tumor-stroma interface or within the tumor nodules.

A modified H-score (MHS) was used to semiquantitatively assess tumor PD-L1 expression, with the modification to the typical H-score being that mononuclear inflammatory cells (MICs) within tumor that express PD-L1 were scored in conjunction with the tumor cells rather than just PD-L1 tumor cells alone. Full or partial PD-L1 plasma membrane staining was scored, whereas cytoplasmic staining was not scored. The MHS was calculated using the following formula:

formula

in which the intensity of the PD-L1 expression was reported as 0 for null, negative, or nonspecific membrane staining; as 1+ for low or weak membrane staining; as 2+ for medium or moderate membrane staining; and as 3+ for high or strong membrane staining. The range for MHS values was 0 to 300.

For this assay, PD-L1 positivity was defined as an MHS of 1 or more or the presence of distinctive PD-L1 interface expression (ie, PD-L1 staining of MICs present at the leading edge or margin of the tumor mass or nodules/nests or closely associated with tumor cells in the stromal environment). The MICs sometimes appeared intermixed with tumor cells at the interface. If tumor cell staining was clearly present, that staining was captured in the MHS value. The PD-L1 interface pattern was not assessed for intensity but, rather, as either present or absent and did not contribute to the MHS.

Concordance proficiency test slides for 3 trainees (all board-certified pathologists) were collated from slides run during validation to confirm use of this scoring method for assessing PD-L1 expression. Concordance, defined as agreement among pathologists on PD-L1 positivity and negativity, was based on review of 30 glass-slide samples.

PD-L1 IHC Assay Precision

Precision of the PD-L1 IHC assay was assessed in a panel of 15 NSCLC and 9 melanoma samples that covered the expected range of PD-L1 expression in clinical samples. Within-run and among-run precision were determined from multiple staining runs performed on different days. At least 2 operators were involved, and different automated-staining platforms were used. The tissues used in each run were replicate serial sections, with 3 sections per sample for PD-L1 expression and 1 section per sample as a negative control. Assay precision was documented following a review of all stained slides by a board-certified pathologist using the tumor MHS.

Specificity of 22C3 as an IHC Reagent and Assay Optimization

The specificity of the anti–PD-L1 antibody clone 22C3 as an IHC reagent was confirmed using several orthogonal methods (Figure 1). As assessed by the NanoString platform (Figure 2) and flow cytometry (Figure 3) in NSCLC cell lines, PD-L1 mRNA expression and protein, respectively, was absent in NCIH23 cells, moderately expressed in NCIH226 cells, and highly expressed in HOP92 cells. Using IHC with FFPE cell pellets and the 22C3 antibody, a similar expression pattern was observed, such that no staining was observed in NCIH23 cells, and almost all HOP92 cells showed staining for PD-L1 at the cell membrane (Figure 4, A through C). A comparison of PD-L1 protein expression as detected by IHC and PD-L1 mRNA as detected by in situ hybridization also confirmed the specificity of clone 22C3 as an IHC reagent by corroborating the distribution of the IHC signal within intact tissues (Figure 5). Based on a robust serial process, the optimal protocol for using 22C3 as an IHC reagent for detecting PD-L1 expression was determined (described in the Materials and Methods).

PD-L1 Immunohistochemistry Staining Patterns

As assessed by the optimized IHC assay in a series of archival NSCLC (n = 142) and melanoma (n = 79) samples, full and partial staining of the plasma membrane of nucleated tumor and nontumor cells was observed (Figure 6, A through H). Cytoplasmic staining was occasionally observed, although its presence or absence was not considered part of the semiquantitative scoring scheme.

Figure 6.

Representative programmed death ligand-1 (PD-L1) tumor staining with the 22C3 antibody. A, Non–small cell lung cancer (NSCLC); modified H-score (MHS), 0. B, NSCLC; MHS, 40. C, NSCLC; MHS, 150. D, NSCLC; MHS, 250. E Melanoma; MHS, 0. F, Melanoma; MHS, 90. G, Melanoma; MHS, 130. H, Melanoma; MHS, 300 (original magnification ×20 [A through H]).

Figure 6.

Representative programmed death ligand-1 (PD-L1) tumor staining with the 22C3 antibody. A, Non–small cell lung cancer (NSCLC); modified H-score (MHS), 0. B, NSCLC; MHS, 40. C, NSCLC; MHS, 150. D, NSCLC; MHS, 250. E Melanoma; MHS, 0. F, Melanoma; MHS, 90. G, Melanoma; MHS, 130. H, Melanoma; MHS, 300 (original magnification ×20 [A through H]).

Close modal

For many of the NSCLC and melanoma samples, a distinctive PD-L1 expression pattern was noted, being present or absent either at the leading edge or margin of the tumor mass or appositionally associated with tumor cells in the stroma microenvironment (Figure 7, A through C). Such PD-L1 staining in, at, or close to the tumor-stroma border was observed even at magnifications as low as ×4. This pattern, termed the interface or stroma pattern, appeared under magnification as a lichenoid pattern of membrane-stained PD-L1+ cells staining at (demarcation or closely associated) or within (intercalating) the stroma bordering the leading edge or margin of tumor nodules. When PD-L1+ MICs infiltrated the stroma, the infiltrate tended to be more intense adjacent to the tumor nests, such that, at low magnification, the staining pattern tended to outline the nests. At higher magnification, the inflammatory cells appeared to be predominantly macrophages. However, juxtaposed tumor cells at the interface can also be PD-L1+ and, on occasion, could be difficult to differentiate from macrophages by PD-L1 alone or by hematoxylin-eosin staining.

Figure 7.

Programmed death ligand-1 (PD-L1) immunohistochemistry (22C3 antibody) staining patterns. Interface pattern in non–small cell lung cancer (A and B) and melanoma (C). D, Dendritic pattern, melanoma. E and F, Melanin interference in melanoma, where PD-L1 staining (E) is not distinguishable from the isotype-negative control (F) (original magnification ×20).

Figure 7.

Programmed death ligand-1 (PD-L1) immunohistochemistry (22C3 antibody) staining patterns. Interface pattern in non–small cell lung cancer (A and B) and melanoma (C). D, Dendritic pattern, melanoma. E and F, Melanin interference in melanoma, where PD-L1 staining (E) is not distinguishable from the isotype-negative control (F) (original magnification ×20).

Close modal

In melanoma samples only, a dendritic-type pattern was occasionally observed (Figure 7, D). This pattern appeared to be driven by dendritic macrophages and tumor cells along cell processes that also expressed PD-L1. In some melanoma samples, quantities of melanin were sufficient to obscure PD-L1 expression even in the context of the negative control (Figure 7, E and F). Although bleaching was tested as a means of reducing the masking effects of melanin, bleaching also reduced the PD-L1 signal and caused tissue loss, damage, or both. Alternative detection/chromogen methods were also tried without success, particularly because melanin content varied among samples. Because the non-DAB chromogens tested nonspecifically bound melanin and staining was present in the negative control, DAB was chosen as the chromogen for this assay for its crisp localization pattern and lack of nonspecific melanin binding. Overall, samples that were heavily melanocytic were not evaluable by PD-L1 expression with the optimized IHC protocol and the 22C3 antibody because of melanin interference.

Scoring of PD-L1 Expression and Analytical Validation

Based on the PD-L1–staining patterns observed in the NSCLC and melanoma tumor samples, the scoring scheme developed to assess PD-L1 positivity captured membranous PD-L1 expression and patterns of interface or, in the case of melanoma, dendritic PD-L1 staining. The MHS, which captures membranous PD-L1 expression only, ranged from 0 to 300, with a score of 1 or more indicative of PD-L1+ expression. The presence of the interface pattern was captured as a Yes/No variable, and samples were considered PD-L1+ if the MHS was 0, but they showed stromal staining. Applying this scoring scheme across 30 samples, there was 100% concordance among the 3 board-certified pathologists in identifying PD-L1+ and PD-L1 samples, based on MHS 1 or more and whether the interface pattern was present. Of note, the intensity of the PD-L1 membrane reactivity could not be considered or inferred with this scoring system; similarly, determination of whether the stromal expression was focal, regional, or global was not possible.

In addition to the 100% interpathologist concordance, there was 100% concordance of PD-L1 status in all replicate samples tested in both within-run and among-run precision-assessment experiments. There was no impact of the instrument used to perform the staining or the operator who performed the staining runs. Positive and negative controls performed as expected in these studies.

Tumor Survey

A focused assessment of PD-L1 staining patterns in the 142 FFPE NSCLC (53 squamous cell carcinoma [37%], 78 adenocarcinoma [55%], and 11 large cell carcinomas [8%]) and 79 FFPE melanoma samples revealed robust performance of the assay for both tumor types. A dynamic range of MHS from 0 to 300 was seen for both melanoma and NSCLC samples. For NSCLC samples, tumor nests comprised virtually all malignant neoplastic cells, although an occasional MIC may have been present and included in the MHS.

As shown in the Table, 80 of 142 NSCLC samples (56%) were PD-L1+, including 56 (39%) that were positive by MHS and 75 (53%) that showed stromal staining. By histologic type, PD-L1 positivity was observed for 33 of the 53 squamous cell carcinoma samples (62%), 39 of the 78 adenocarcinoma samples (50%), and 8 of the 11 large cell carcinoma samples (73%). Four of 83 melanoma samples (5%) were heavily melanocytic and not evaluable for PD-L1 expression. Of the 79 evaluable melanoma samples, 42 (53%) were considered to be PD-L1+, including 23 (29%) that were positive by MHS and 30 (38%) that were positive by stroma (Table). The range of MHS positivity and the presence/absence of stroma positivity are shown for NSCLC (Figure 8, A) and melanoma (Figure 8, B).

Programmed Death Ligand-1 (PD-L1) Expression in Non–Small Cell Lung Cancer (NSCLC) and Melanoma Tumor Samples

Programmed Death Ligand-1 (PD-L1) Expression in Non–Small Cell Lung Cancer (NSCLC) and Melanoma Tumor Samples
Programmed Death Ligand-1 (PD-L1) Expression in Non–Small Cell Lung Cancer (NSCLC) and Melanoma Tumor Samples
Figure 8.

Relationship among staining patterns and modified H-score (MHS) in non–small cell lung cancer (A) and melanoma (B) tissue samples.

Figure 8.

Relationship among staining patterns and modified H-score (MHS) in non–small cell lung cancer (A) and melanoma (B) tissue samples.

Close modal

Pembrolizumab is a potent, highly selective, humanized monoclonal antibody that was designed to block the interaction between PD-1 and its ligands, PD-L1 and PD-L2. This work describes steps taken to develop a prototype assay suitable for analyzing PD-L1 expression in Merck-sponsored clinical trials and epidemiologic studies to assess the utility of PD-L1 as a potential prognostic biomarker, predictive biomarker, or both. The described assay has a wide dynamic range, shows consistent results among runs, and allows for interpathologist concordance, making it a valuable tool in addressing questions related to the role of PD-L1 as a biomarker in advanced malignancies. This assay is now being used to determine eligibility in select clinical studies of pembrolizumab with the intent of enriching the population of patients with NSCLC who may be more responsive to pembrolizumab20,21  and for defining the prognostic role of PD-L1 in melanoma,22  NSCLC,20,21  head and neck cancer,23  gastric cancer,24  and Hodgkin lymphoma.25 

This assay does have some limitations for its potential broad use, including a long procedure time (2-day protocol because of overnight incubation) and the use of a staining platform that is no longer available commercially. Furthermore, the scoring methods may be simplified in the future, based on the role of PD-L1 patterns, the intensity of PD-L1 staining in the cell membrane, and the percentage of membranous PD-L1 positivity, depending on the predictive or prognostic value of those scoring components individually or together. Nonetheless, this prototype assay is important because it was used in several epidemiologic studies and early clinical trials of pembrolizumab and because it served as the basis for the development of a companion diagnostic for pembrolizumab.

1
Disis
ML.
Immune regulation of cancer
.
J Clin Oncol
.
2010
;
28
(
29
):
4531
4538
.
2
Brown
JA,
Dorfman
DM,
Ma
FR,
et al.
Blockade of programmed death-1 ligands on dendritic cells enhances T cell activation and cytokine production
.
J Immunol
.
2003
;
170
(
3
):
1257
1266
.
3
Dong
H,
Strome
SE,
Salomao
DR,
et al.
Tumor-associated B7-H1 promotes T-cell apoptosis: a potential mechanism of immune evasion
.
Nat Med
.
2002
;
8
(
8
):
793
800
.
4
Francisco
LM,
Sage
PT,
Sharpe
AH.
The PD-1 pathway in tolerance and autoimmunity
.
Immunol Rev
.
2010
;
236
:
219
242
.
5
Sharpe
AH,
Freeman
GJ.
The B7-CD28 superfamily
.
Nat Rev Immunol
.
2002
;
2
(
2
):
116
126
.
6
Thompson
RH,
Webster
WS,
Cheville
JC,
et al.
B7-H1 glycoprotein blockade: a novel strategy to enhance immunotherapy in patients with renal cell carcinoma
.
Urology
.
2005
;
66(5)(suppl):10–14.
7
Nomi
T,
Sho
M,
Akahori
T,
et al.
Clinical significance and therapeutic potential of the programmed death-1 ligand/programmed death-1 pathway in human pancreatic cancer
.
Clin Cancer Res
.
2007
;
13
(
7
):
2151
2157
.
8
Gao
Q,
Wang
XY,
Qiu
SJ,
et al.
Overexpression of PD-L1 significantly associates with tumor aggressiveness and postoperative recurrence in human hepatocellular carcinoma
.
Clin Cancer Res
.
2009
;
15
(
3
):
971
979
.
9
Hamanishi
J,
Mandai
M,
Iwasaki
M,
et al.
Programmed cell death 1 ligand 1 and tumor-infiltrating CD8+ T lymphocytes are prognostic factors of human ovarian cancer
.
Proc Natl Acad Sci U S A
.
2007
;
104
(
9
):
3360
3365
.
10
Hino
R,
Kabashima
K,
Kato
Y,
et al.
Tumor cell expression of programmed cell death-1 ligand 1 is a prognostic factor for malignant melanoma
.
Cancer
.
2010
;
116
(
7
):
1757
1766
.
11
Mu
CY,
Huang
JA,
Chen
Y,
Chen
C,
Zhang
XG.
High expression of PD-L1 in lung cancer may contribute to poor prognosis and tumor cells immune escape through suppressing tumor infiltrating dendritic cells maturation
.
Med Oncol
.
2011
;
28
(
3
):
682
688
.
12
Sun
JM,
Zhou
W,
Choi
SJ,
et al.
PD-L1 expression and survival in patients with non–small cell lung cancer (NSCLC) in Korea
[abstract].
J Clin Oncol
.
2014
;
32(15)(suppl 5):8066.
13
Sorensen
S,
Zhou
W,
Dolled-Filhart
M,
et al.
Antitumor activity of pembrolizumab (pembro; MK-3475) and correlation with programmed death ligand 1 (PD-L1) expression in a pooled analysis of patients (pts) with advanced non–small cell lung carcinoma (NSCLC1328P) [abstract]
.
Ann Oncol
.
2014
;
25
(
suppl 4
):
1328
.
14
Wang
A,
Wang
HY,
Liu
Y,
et al.
The prognostic value of PD-L1 expression for non–small cell lung cancer patients: a meta-analysis
.
Eur J Surg Oncol
.
2015
;
41
(
4
):
450
456
.
15
Gadiot
J,
Hooijkaas
AI,
Kaiser
AD,
van Tinteren
H,
van Boven
H,
Blank
C.
Overall survival and PD-L1 expression in metastasized malignant melanoma
.
Cancer
.
2011
;
117
(
10
):
2192
2201
.
16
Taube
JM,
Anders
RA,
Young
GD,
et al.
Colocalization of inflammatory response with B7-H1 expression in human melanocytic lesions supports an adaptive resistance mechanism of immune escape
.
Sci Transl Med
.
2012
;
4(127):127ra37.
17
Daud
A,
Hamid
O,
Robert
C,
et al.
Relationship between programmed death ligand 1 (PD-L1) expression and clinical outcome in patients with melanoma treated with pembrolizumab (MK-3475) [abstract]
.
Eur J Cancer
.
50(suppl 6):48–49
.
18
Battifora
H.
The multitumor (sausage) tissue block: novel method for immunohistochemical antibody testing
.
Lab Invest
.
1986
;
55
(
2
):
244
248
.
19
Reed
JA,
Manahan
LJ,
Park
CS,
Brigati
DJ.
Complete one-hour immunocytochemistry based on capillary action
.
Biotechniques
.
1992
;
13
(
3
):
434
443
.
20
Gandhi
L,
Balmanoukian
A,
Hui
R,
et al.
MK-3475 (anti-PD-1 monoclonal antibody) for non–small cell lung cancer: antitumor activity and association with tumor PD-L1 expression
[abstract].
Cancer Res
.
2014
;
74(19)(suppl 1):CT105.
21
Garon
E,
Gandhi
L,
Rizvi
N,
et al.
Antitumor activity of pembrolizumab (pembro; MK-3475) and correlation with programmed death ligand 1 (PD-L1) expression in a pooled analysis of patients (pts) with advanced non–small cell lung carcinoma (NSCLC)
[abstract].
Ann Oncol
.
2014
;
25(suppl 4):LBA43.
22
Daud
AI,
Hamid
O,
Antoni-Ribas
F,
et al.
Antitumor activity of the anti-PD-1 monoclonal antibody MK-3475 in melanoma (MEL): correlation of tumor PD-L1 expression with outcome
[abstract].
Cancer Res
.
2014
;
74(19)(suppl 1):CT104.
23
Chow
LQ,
Burtness
B,
Weiss
J,
et al.
Phase Ib study of pembrolizumab (pembro; MK-3475) in patients (pts) with human papillomavirus virus (HPV)-positive and negative head and neck cancer (HNC)
.
Ann Oncol
.
2014
;
25(suppl 4):LBA31.
24
Muro
K,
Bang
Y,
Shankaran
V,
et al.
A phase 1b study of pembrolizumab (pembro; MK-3475) in patients with advanced gastric cancer
[abstract].
Ann Oncol
.
2014
;
25(suppl 4):LBA15.
25
Moskowitz
CH,
Ribrag
V,
Michot
JM,
et al.
PD-1 blockade with the monoclonal antibody pembrolizumab (MK-3475) in patients with classical Hodgkin lymphoma after brentuximab vedotin failure: preliminary results from a phase 1b study [keynote 013; abstract]
.
Blood
.
2014
;
124
(
21
):
290
.

Author notes

Editorial support for this manuscript was provided by Melanie Leiby, PhD, at the APO Group (Yardley, Pennsylvania), and was funded by Merck Sharp & Dohme Corp, a subsidiary of Merck & Co, Inc.

Drs Dolled-Filhart, Wu, Yearley, Pierce, Weiner, and Emancipator are current or former employees of Merck & Co, Inc; Drs Locke and Murphy are current or former employees of QualTek Molecular Laboratories; Dr Frisman is a consultant for QualTek Molecular Laboratories; Dr Lynch is an employee and has equity ownership in QualTek Molecular Laboratories; and Drs Dolled-Filhart, Yearley, and Emancipator own stock in Merck & Co, Inc and have stock options. The authors have no other relevant financial interest in the products or companies described in this article.