Context.—

An immunohistochemistry (IHC) assay developed to detect lymphocyte-activation gene 3 (LAG-3), a novel immune checkpoint inhibitor target, has demonstrated high analytic precision and interlaboratory reproducibility using a Leica staining platform, but it has not been investigated on other IHC staining platforms.

Objective.—

To evaluate the performance of LAG-3 IHC assays using the 17B4 antibody clone across widely used IHC staining platforms: Agilent/Dako Autostainer Link 48 and VENTANA BenchMark ULTRA compared to Leica BOND-RX (BOND-RX).

Design.—

Eighty formalin-fixed, paraffin-embedded melanoma tissue blocks were cut into consecutive sections and evaluated using staining platform–specific IHC assays with the 17B4 antibody clone. Duplicate testing was performed on the BOND-RX platform to assess intraplatform agreement. LAG-3 expression using a numeric score was evaluated by a pathologist and with a digital scoring algorithm. LAG-3 positivity was determined from manual scores using a 1% or greater cutoff.

Results.—

LAG-3 IHC staining patterns and intensities were visually similar across all 3 staining platforms. Spearman and Pearson correlations were 0.75 or greater for interplatform and BOND-RX intraplatform concordance when LAG-3 expression was evaluated with a numeric score determined by a pathologist. Correlation increased with a numeric score determined with a digital scoring algorithm (Spearman and Pearson correlations ≥0.88 for all comparisons). Overall percentage agreement was 77.5% or greater for interplatform and BOND-RX intraplatform comparisons when LAG-3 positivity was determined using a 1% or greater cutoff.

Conclusions.—

Data presented here demonstrate that LAG-3 expression can be robustly and reproducibly assessed across 3 major commercial IHC staining platforms using the 17B4 antibody clone.

Lymphocyte-activation gene 3 (LAG-3) is a cell-surface immune checkpoint molecule expressed on immune cells (ICs) initiating an inhibitory signal that can impair T-cell activity and attenuate proinflammatory cytokine responses.13  Preclinical studies have shown that dual blockade of LAG-3 and programmed death receptor-1 (PD-1) has synergistic antitumor activity, indicating that LAG-3 is an ideal candidate for novel immune checkpoint inhibitor combinations.4  RELATIVITY-047 (NCT03470922), a phase 2/3, global, randomized, double-blind trial, evaluated combined LAG-3 and PD-1 inhibition with relatlimab, an anti–LAG-3 antibody, and nivolumab, an anti–PD-1 antibody, as a novel fixed-dose combination versus nivolumab in patients with previously untreated metastatic or unresectable melanoma.5  In this study, combined treatment with relatlimab and nivolumab demonstrated superior progression-free survival compared with nivolumab monotherapy regardless of LAG-3 expression.5 

LAG-3 expression may be reflective of tumor inflammation and therefore generally predictive of response to the immuno-oncology therapy class.58  Consequently, assessing LAG-3 expression by immunohistochemistry (IHC) remains of significant interest to the research community despite a lack of clinical utility in informing relatlimab treatment decisions in advanced melanoma.5  A LAG-3 IHC assay using the 17B4 LAG-3 antibody clone on the Leica BOND-III staining platform has been developed in collaboration between Bristol Myers Squibb and LabCorp.9  This assay is currently being used in relatlimab clinical trials, including the recently completed RELATIVITY-047.5  A recent study demonstrated the reproducible intraobserver, interobserver, and interlaboratory performance of the LabCorp LAG-3 IHC assay using the 17B4 antibody clone on the Leica BOND-III platform but did not investigate its performance on other staining platforms.9 

Although assessment of LAG-3 expression is not required in a diagnostic context, a crucial barrier to widespread implementation of an IHC assay is its ability to produce concordant results using different staining platforms.5  Commercial IHC assays that are developed for a specific staining platform can present a significant challenge to testing laboratories, which may not have compatible equipment. Laboratory-developed tests enable the reliable use of an IHC assay on alternative staining platforms by adapting validated assays, but they do not always produce concordant results. Stringent standardization is therefore recommended prior to routine clinical use.1012  This highlights the importance of reliable cross-platform performance for IHC assays to enable broad implementation and maximum utility. The LAG-3 IHC assay described by Johnson et al9  was developed using the 17B4 antibody clone on the Leica BOND-III platform. However, other IHC staining platforms, such as the Agilent/Dako Autostainer Link 48 (ASL-48) and VENTANA BenchMark ULTRA (VBU), are also commonly used. In this study, we assessed the feasibility of developing staining procedures using the 17B4 antibody clone on other staining platforms. The aim of our study was to develop protocols using ASL-48 or VBU that produced results comparable to those for the Leica BOND-RX (BOND-RX) platform. Interplatform concordance was examined using both pathologist evaluation and an investigational digital pathology method.

Samples and Staining Procedures

Eighty formalin-fixed, paraffin-embedded (FFPE) melanoma tissue blocks were obtained from commercial vendors (BioIVT, Detroit, Michigan; Discovery Life Sciences, Huntsville, Alabama). Each tissue block was cut at 4-μm thickness into 4 consecutive sections for LAG-3 IHC staining on each platform, with additional sections cut for hematoxylin-eosin staining and isotype controls. Consecutive sections were evaluated using BOND-RX (section 1), ASL-48 (section 2), VBU (section 3), and BOND-RX (section 4), with 2 sections assessed using BOND-RX to measure intraplatform agreement for evaluation (hereafter referred to as BOND-RX 1 [section 1] and BOND-RX 2 [section 4]; Figure 1). Deparaffinization and melanin removal were performed as described by Johnson et al,9  with the exception that a PT200 PreTreatment module (Agilent/Dako, Santa Clara, California) was used to incubate slides rather than a Decloaking Chamber NxGen (BioCare Medical, Pacheo, California). All experiments across all platforms were performed using a monoclonal LAG-3 antibody, clone 17B4 (catalog No. LS-C18692, LSBio, Seattle, Washington). Staining procedures were initially developed on BOND-RX (Supplemental Figure 1, see the supplemental digital content containing 2 figures, a methods section, and 1 table at https://meridian.allenpress.com/aplm in the November 2023 table of contents), after which protocols were iteratively optimized on ASL-48 and VBU to produce staining visually similar to that of BOND-RX, before performing comparisons (staining procedure optimization is described in the Supplemental Methods section). All staining for the interplatform comparison experiments was performed in a single laboratory to avoid interlaboratory bias. Staining procedures and reagents were specific to each platform and are summarized in the Table. The study was performed in accordance with the Bristol Myers Squibb Bioethics policy (https://www.bms.com/about-us/responsibility/position-on-key-issues/bioethics-policy-statement.html) and adhered to the World Medical Association Declaration of Helsinki for Human Research.

Figure 1

Diagram showing the consecutive sectioning design. Formalin-fixed, paraffin-embedded (FFPE) melanoma tissue blocks were cut into consecutive sections and stained using the lymphocyte-activation gene 3 (LAG-3) 17B4 antibody clone as shown. Abbreviations: ASL-48, Agilent/Dako Autostainer Link 48; BOND-RX 1, Leica BOND-RX [section 1]; BOND-RX 2, Leica BOND-RX [section 4]; VBU, VENTANA BenchMark ULTRA.

Figure 1

Diagram showing the consecutive sectioning design. Formalin-fixed, paraffin-embedded (FFPE) melanoma tissue blocks were cut into consecutive sections and stained using the lymphocyte-activation gene 3 (LAG-3) 17B4 antibody clone as shown. Abbreviations: ASL-48, Agilent/Dako Autostainer Link 48; BOND-RX 1, Leica BOND-RX [section 1]; BOND-RX 2, Leica BOND-RX [section 4]; VBU, VENTANA BenchMark ULTRA.

Close modal

Pathologist Scoring

Slides were scored for LAG-3–positive IC content within the tumor region by a single expert pathologist trained on a LAG-3 scoring methodology (reference slides are provided in Supplemental Figure 2), which has been previously described.9  Scoring was performed by a single pathologist to avoid interobserver bias. The tumor region included 100 or more tumor cells (TCs [confirmed using a hematoxylin-eosin–stained slide]), intratumoral stroma, and peritumoral stroma (the band of stromal elements directly contiguous with the outer tumor margin) and did not include normal and/or adjacent uninvolved tissues. LAG-3 expression was recorded using a numeric score defined as the percentage of LAG-3–positive ICs that morphologically resembled lymphocytes relative to all nucleated cells (ICs [lymphocytes and macrophages], stromal cells, and TCs). The scoring scale was (in %) 0, 1, 2, 3, 4, 5, and 10, and further increments of 10 up to 100. Samples with LAG-3–positive IC percentage scores of 1% or greater were reported as LAG-3–positive. Slides were randomized prior to evaluation. For each staining platform, all 80 slides were scored over 2 days (40 slides per day) followed by a washout period of 5 days. All 320 slides were scored by the same pathologist.

Digital Scoring

Images of all consecutive sections from the same tissue block evaluated using BOND-RX 1, ASL-48, VBU, or BOND-RX 2 were analyzed in Visiopharm (Visiopharm, Hoersholm, Denmark). The analysis was performed using a single in-house scoring algorithm modified for use on the staining platforms used in this study, with a simple thresholding change and minor clean-up steps to eliminate some pigmentation. To ensure that a consistent region of interest (the tumor region containing the tumor and tumor-associated stroma) was captured for serial sections, the region of interest was annotated on 1 image and copied onto the other 3 images, with minor adjustments to account for interslide variability. Areas that were unevaluable on 1 image (eg, due to tissue detachment or excess melanin pigmentation) were not taken into consideration on the other images. If a section stained on 1 platform was completely unevaluable it was excluded, the sections stained on the other 3 platforms were analyzed, and a comment was made. If all sections from a block were unevaluable, then none were analyzed, and a comment was made. The digital scoring algorithm recorded LAG-3 expression as the sum of the positive 3,3′-diaminobenzidine tetrahydrochloride hydrate signal as a fraction of the area of analysis.

Statistical Analysis

LAG-3 IHC scores were analyzed on the ratio scale, and no additional normalization and/or standardization was applied. All analyses were performed using R software (version 4.0.5).

Analysis of LAG-3 Using a Numeric Score

For the bar plot of LAG-3 score by case, cases were ordered from lowest to highest based on the values from the BOND-RX 2 samples. The prevalence as a function of cutoff was calculated by counting the percentage of cases exceeding (greater than or equal to) the given cutoff. For the scatterplot analyses of the manual and digital scoring data, the square-root transformation was applied to both x and y axes for better visualization in the low-score range. For each scatterplot comparison, the strength of correlation was assessed using the coefficient of determination (R2), Pearson correlation, and Spearman correlation.

Analysis of LAG-3 Positivity Using a 1% or Greater Cutoff

Agreement between platforms with a 1% or greater cutoff to determine LAG-3 positivity was assessed using the pairwise percentage agreement as well as Cohen κ per US Food and Drug Administration guidelines and as previously described.13,14  For this analysis, (1) in “A versus B percentage agreement,” A was used as the reference; (2) the discordance metric was calculated as (100 − overall percentage agreement [OPA])%; (3) the 95% CIs for the percentage agreements were calculated using the Clopper-Pearson method15 ; and (4) the 95% CI for the Cohen κ was calculated using the variance estimates by the Fleiss-Cohen-Everitt method.16  The Venn diagram was generated using the R package VennDiagram (version 1.6.0).

After adapting the BOND-RX staining procedure for use on ASL-48 and VBU, we examined the similarity of LAG-3 IHC staining patterns and intensities across the different staining platforms using consecutive sections cut from 80 melanoma FFPE tissue blocks, as shown in Figure 1. LAG-3 staining was assessed by visual inspection by a pathologist and revealed similar LAG-3 IHC staining patterns and intensities across the different staining platforms (Figure 2).

Figure 2

Visual comparison of lymphocyte-activation gene 3 (LAG-3) staining in melanoma samples across 3 staining platforms. Representative images of LAG-3 immune cell immunohistochemistry staining in melanoma samples at ×8 (A) BOND-RX 1, (B) BOND-RX 2, (C) ASL-48, (D) VBU; and ×24 (E) BOND-RX 1, (F) BOND-RX 2, (G) ASL-48, (H) VBU. Abbreviations: ASL-48, Agilent/Dako Autostainer Link 48; BOND-RX 1, Leica BOND-RX (section 1); BOND-RX 2, Leica BOND-RX (section 4); VBU, VENTANA BenchMark ULTRA.

Figure 2

Visual comparison of lymphocyte-activation gene 3 (LAG-3) staining in melanoma samples across 3 staining platforms. Representative images of LAG-3 immune cell immunohistochemistry staining in melanoma samples at ×8 (A) BOND-RX 1, (B) BOND-RX 2, (C) ASL-48, (D) VBU; and ×24 (E) BOND-RX 1, (F) BOND-RX 2, (G) ASL-48, (H) VBU. Abbreviations: ASL-48, Agilent/Dako Autostainer Link 48; BOND-RX 1, Leica BOND-RX (section 1); BOND-RX 2, Leica BOND-RX (section 4); VBU, VENTANA BenchMark ULTRA.

Close modal

Next, we sought to compare pathologist scoring of LAG-3 expressed on ICs using a numeric score across the 3 staining platforms. Consecutive sections were cut from 80 melanoma FFPE tissue blocks and stained for LAG-3. Overall, for all cases, LAG-3 expression on ICs was comparable between consecutive sections cut from the same tissue block when stained using the different IHC platforms (Figure 3, A). Similarly, consecutive sections cut from the same tissue block displayed comparable LAG-3 expression on ICs when both sections were stained using the BOND-RX IHC platform. The prevalence of LAG-3 expression on ICs at different cutoffs was similar across platforms and between runs on the same platform for the 2 BOND-RX runs (Figure 3, B). For example, the percentage of samples determined as having LAG-3 IC expression 1% or greater was 42% (34 of 80) for BOND-RX 1, 48% (38 of 80) for ASL-48, 56% (45 of 80) for VBU, and 51% (41 of 80) for BOND-RX 2. LAG-3 IC expression was strongly correlated between the 2 BOND-RX runs and between BOND-RX and both ASL-48 and VBU (Figure 4, A through C; raw data are provided in Supplemental Table 1). Notably, intraplatform concordance (BOND-RX 2 versus BOND-RX 1: Pearson correlation = 0.91 [P < .001], Spearman correlation = 0.82 [P < .001], slope = 1.2) was comparable to interplatform concordance (ASL-48 versus BOND-RX 1: Pearson correlation = 0.90 [P < .001], Spearman correlation = 0.86 [P < .001], slope = 1.2; VBU versus BOND-RX 2: Pearson correlation = 0.88 [P < .001], Spearman correlation = 0.75 [P < .001], slope = 0.9).

Figure 3

Comparison of lymphocyte-activation gene 3 (LAG-3) immune cell (IC) immunohistochemistry staining results across 3 staining platforms. A, Bar charts show LAG-3 IC expression assessed in consecutive sections on different platforms. Cases were ordered from lowest to highest based on the values from the BOND-RX 2 samples. B, Bar charts show the percentage of samples exceeding (greater than or equal to) the indicated LAG-3 expression cutoffs. Two slides were run on the BOND-RX platform to assess intraplatform variability. Abbreviations: ASL-48, Agilent/Dako Autostainer Link 48; BOND-RX 1, Leica BOND-RX (section 1); BOND-RX 2, Leica BOND-RX (section 4); VBU, VENTANA BenchMark ULTRA.

Figure 3

Comparison of lymphocyte-activation gene 3 (LAG-3) immune cell (IC) immunohistochemistry staining results across 3 staining platforms. A, Bar charts show LAG-3 IC expression assessed in consecutive sections on different platforms. Cases were ordered from lowest to highest based on the values from the BOND-RX 2 samples. B, Bar charts show the percentage of samples exceeding (greater than or equal to) the indicated LAG-3 expression cutoffs. Two slides were run on the BOND-RX platform to assess intraplatform variability. Abbreviations: ASL-48, Agilent/Dako Autostainer Link 48; BOND-RX 1, Leica BOND-RX (section 1); BOND-RX 2, Leica BOND-RX (section 4); VBU, VENTANA BenchMark ULTRA.

Close modal
Figure 4

Scatterplots of platform comparison using pathologist scoring and a digital scoring algorithm. A through C, Scatterplots comparing pathologist scoring of lymphocyte-activation gene 3 (LAG-3) expression across platforms (see the Pathologist Scoring section in Materials and Methods). D through F, Scatterplots comparing LAG-3 scores using a digital scoring algorithm (see the Digital Scoring section in Materials and Methods). On graphs A through F, the line of identity is in blue and the linear regression line is in black. The coefficient of determination (R2), Pearson correlation, and Spearman correlation are displayed on the scatterplots. Two slides were run on the BOND-RX platform to assess intraplatform variability. Abbreviations: ASL-48, Agilent/Dako Autostainer Link 48; BOND-RX 1, Leica BOND-RX [section 1]; BOND-RX 2, Leica BOND-RX [section 4]; DAB, 3,3′-diaminobenzidine tetrahydrochloride hydrate; VBU, VENTANA BenchMark ULTRA.

Figure 4

Scatterplots of platform comparison using pathologist scoring and a digital scoring algorithm. A through C, Scatterplots comparing pathologist scoring of lymphocyte-activation gene 3 (LAG-3) expression across platforms (see the Pathologist Scoring section in Materials and Methods). D through F, Scatterplots comparing LAG-3 scores using a digital scoring algorithm (see the Digital Scoring section in Materials and Methods). On graphs A through F, the line of identity is in blue and the linear regression line is in black. The coefficient of determination (R2), Pearson correlation, and Spearman correlation are displayed on the scatterplots. Two slides were run on the BOND-RX platform to assess intraplatform variability. Abbreviations: ASL-48, Agilent/Dako Autostainer Link 48; BOND-RX 1, Leica BOND-RX [section 1]; BOND-RX 2, Leica BOND-RX [section 4]; DAB, 3,3′-diaminobenzidine tetrahydrochloride hydrate; VBU, VENTANA BenchMark ULTRA.

Close modal

Because observer variation is a documented limitation of quantitative IHC analysis by pathologists’ visual IC scoring, we also assessed LAG-3 expression by digital image analysis.1719  This approach eliminates intraobserver variation and enables a more objective assessment of staining variability between platforms. Of the 80 samples, LAG-3 expression was analyzed using the digital scoring algorithm in 77 samples on BOND-RX and ASL-48 and 75 samples on VBU. Although melanin was removed from all samples using the same method (Table), excess melanin pigmentation made 3 samples unevaluable by digital image analysis on all platforms, and a further 2 samples were unevaluable by digital image analysis on VBU. Notably, each of these samples had sufficient melanin removal for manual evaluation by a pathologist. Intraplatform and interplatform correlations increased when using the digital scoring algorithm compared with pathologist scoring (BOND-RX 2 versus BOND-RX 1: Pearson correlation = 0.98 [P < .001], Spearman correlation = 0.93 [P < .001], slope = 0.97; ASL-48 versus BOND-RX 1: Pearson correlation = 0.96 [P < .001], Spearman correlation = 0.88 [P < .001], slope = 0.91; VBU versus BOND-RX 2: Pearson correlation = 0.93 [P < .001], Spearman correlation = 0.88 [P < .001], slope = 0.85; Figure 4, D through F; raw data are provided in Supplemental Table 1).

Finally, because clinical trials often focus on stratifying patients using a defined cutoff, we investigated interplatform and intraplatform agreement when LAG-3 IC positivity was determined from pathologist scores using a 1% or greater cutoff. The level of agreement was high across all platforms, with point estimates for comparisons across all platforms greater than 75% for positive percentage agreement, negative percentage agreement, and OPA, and a low level of discordance (<25% across all platforms; Figure 5, A). Across all platforms, 27 of 80 samples were determined as LAG-3–negative, and 26 of 80 samples were determined as LAG-3–positive (Figure 5, B).

Figure 5

Agreement between 3 immunohistochemistry (IHC) staining platforms using a 1% or greater cutoff to determine lymphocyte-activation gene 3 (LAG-3) positivity. A, Intraplatform and interplatform agreement between 3 IHC staining platforms. B, A Venn diagram showing positive calls using a 1% or greater cutoff to determine LAG-3 immune cell positivity across 3 IHC staining platforms. Two slides were run on the BOND-RX platform to assess intraplatform variability. Abbreviations: ASL-48, Agilent/Dako Autostainer Link 48; BOND-RX 1, Leica BOND-RX (section 1); BOND-RX 2, Leica BOND-RX (section 4); NPA, negative percentage agreement; OPA, overall percentage agreement; PPA, positive percentage agreement; VBU, VENTANA BenchMark ULTRA.

Figure 5

Agreement between 3 immunohistochemistry (IHC) staining platforms using a 1% or greater cutoff to determine lymphocyte-activation gene 3 (LAG-3) positivity. A, Intraplatform and interplatform agreement between 3 IHC staining platforms. B, A Venn diagram showing positive calls using a 1% or greater cutoff to determine LAG-3 immune cell positivity across 3 IHC staining platforms. Two slides were run on the BOND-RX platform to assess intraplatform variability. Abbreviations: ASL-48, Agilent/Dako Autostainer Link 48; BOND-RX 1, Leica BOND-RX (section 1); BOND-RX 2, Leica BOND-RX (section 4); NPA, negative percentage agreement; OPA, overall percentage agreement; PPA, positive percentage agreement; VBU, VENTANA BenchMark ULTRA.

Close modal

This study investigated the comparability of IHC assays using the 17B4 antibody clone to assess LAG-3 expression across different staining platforms, which is an important consideration for the broad implementation of an IHC assay. Although the data from RELATIVITY-047 demonstrated that combined treatment with relatlimab and nivolumab prolongs progression-free survival compared with nivolumab monotherapy regardless of LAG-3 status in patients with previously untreated metastatic or unresectable melanoma, there remains significant interest in assessing LAG-3 expression by IHC among the research community.5 

Here, we show that LAG-3 IHC assays using the 17B4 antibody clone produce visually similar staining patterns and intensities across the BOND-RX, ASL-48, and VBU staining platforms. Across all 3 platforms, we observed statistically significant agreement when LAG-3 expression on ICs was measured by a pathologist using a numeric score or when a 1% or greater cutoff was used to determine LAG-3 positivity, ranging from 77.5% to 87.5% OPA. These findings are consistent with previous publications and in line with what would be expected in clinical settings.9  Intraobserver variability of manual scoring is considered to be the main factor affecting the degree of agreement observed within this study and may indicate practical limitations of manual scoring, in line with previously published reports.18,19  This task is more challenging than TC scoring because of IC size and variability, and is made more difficult by the distribution of LAG-3 expression relative to the 1% clinical cutoff. Importantly, intraplatform correlation and interplatform correlation were similar, highlighting the concordance of the LAG-3 IHC assays using the 17B4 antibody between staining platforms when performed manually by a pathologist.

In recent years, digital pathology techniques have been developed to address some of the inherent challenges within traditional manual pathology methods. In particular, digital pathology techniques are less time consuming and reduce interobserver variability compared with manual pathology methods.17,20  In this study, the use of a digital scoring algorithm to analyze LAG-3 IC expression increased both intraplatform and interplatform correlation compared with pathologist scoring. This indicates that there is little true intraplatform and interplatform variance in LAG-3 IC staining, because the limited variance observed using the digital scoring algorithm contains contributions from both technical variance and biologic variance between consecutive sections of the same tissue block. A challenge with digital scoring is that simple algorithms focused on color detection are susceptible to artifacts, due to a low signal-to-noise ratio caused by melanin presence.17  Improvements to digital scoring could be facilitated by further improvements in the removal of melanin, better color separation between the IHC chromogen and brown melanin pigment, or by developing artificial intelligence–based image analysis algorithms that contain more sophisticated models of positive ICs.

Previous IHC comparison studies have shown that specific antibodies can be used on alternative platforms in order to overcome barriers to implementation of IHC assays.2124  For example, Hendry et al21  found that the 22C3 assay produced consistent programmed death ligand-1 (PD-L1) scoring on the ASL-48 platform and the VBU platform. Data presented here describe the development of LAG-3 IHC assays using the 17B4 antibody clone that produce comparable LAG-3 staining on the BOND-RX, VBU, and ASL-48 staining platforms.

An important limitation is that LAG-3 positivity determined using a 1% or greater cutoff is not predictive of response to relatlimab, and therefore assessment of LAG-3 expression is for exploratory use only.5  Another limitation is that preanalytic factors that may have an impact on LAG-3 staining, such as fixation time, the fixative used, and sample ischemia time, could not be evaluated with the commercial samples used in this manuscript.25  Additionally, in this study all samples were stained in a single laboratory with 1 pathologist performing the scoring, and we did not assess interobserver variability within and across IHC staining platforms, or interantibody clone variability, as has been done in harmonization studies of PD-L1 IHC assays.21,23  Although this ensured accurate assessment of interplatform reproducibility by preventing interlaboratory, interobserver, and interantibody clone bias, it does not reflect the variability that may be observed in the real-world use of the assays. Robust interobserver and interlaboratory reproducibility of the LAG-3 IHC assay using the 17B4 antibody clone have been recently demonstrated on the Leica BOND-III staining platform.9  Nevertheless, future studies examining interlaboratory, interobserver, and interantibody clone variability within and across additional platforms would provide further valuable evidence that LAG-3 IHC assays are reliable and suitable for use across a range of different staining platforms.

Here, we demonstrate that LAG-3 IHC assays based on the 17B4 clone perform reproducibly across 3 different and widely used IHC staining platforms. Together with previously published data demonstrating the analytic precision and interlaboratory reproducibility of the LAG-3 17B4-based IHC assay on the Leica BOND-III,9  results presented here show that this assay is robust and could help overcome barriers to LAG-3 testing implementation in future studies.

The authors acknowledge Iryna Shnitsar, PhD, of Bristol Myers Squibb for helpful discussions and contributions to the preparation of the manuscript. Medical writing and editorial support were provided by Peter Harrison, PhD, and Matthew Weddig of Spark Medica Inc, funded by Bristol Myers Squibb.

1.
Long
L,
Zhang
X,
Chen
F,
et al.
The promising immune checkpoint LAG-3: from tumor microenvironment to cancer immunotherapy
.
Genes Cancer
.
2018
;
9
(5–6)
:
176
189
.
2.
Wang
J,
Sanmamed
MF,
Datar
I,
et al.
Fibrinogen-like protein 1 is a major immune inhibitory ligand of LAG-3
.
Cell
.
2019
;
176
(1–2)
:
334
347.e12
.
3.
Workman
CJ,
Cauley
LS,
Kim
IJ,
Blackman
MA,
Woodland
DL,
Vignali
DAA.
Lymphocyte activation gene-3 (CD223) regulates the size of the expanding T cell population following antigen activation in vivo
.
J Immunol
.
2004
;
172
(9)
:
5450
5455
.
4.
Woo
S-R,
Turnis
ME,
Goldberg
MV,
et al.
Immune inhibitory molecules LAG-3 and PD-1 synergistically regulate T-cell function to promote tumoral immune escape
.
Cancer Res
.
2012
;
72
(4)
:
917
927
.
5.
Tawbi
HA,
Schadendorf
D,
Lipson
EJ,
et al.
Relatlimab and nivolumab versus nivolumab in untreated advanced melanoma
.
N Engl J Med
.
2022
;
386
(1)
:
24
34
.
6.
Wu
S,
Shi
X,
Wang
J,
et al.
Triple-negative breast cancer: intact mismatch repair and partial co-expression of PD-L1 and LAG-3
.
Front Immunol
.
2021
;
12
:
561793
.
7.
Yarchoan
M,
Xing
D,
Luan
L,
et al.
Characterization of the immune microenvironment in hepatocellular carcinoma
.
Clin Cancer Res
.
2017
;
23
(23)
:
7333
7339
.
8.
Saleh
RR,
Peinado
P,
Fuentes-Antrás
J,
et al.
Prognostic value of lymphocyte-activation gene 3 (LAG3) in cancer: a meta-analysis
.
Front Oncol
.
2019
;
9
:
1040
.
9.
Johnson
L,
McCune
B,
Locke
D,,
et al.
Development of a LAG-3 immunohistochemistry assay for melanoma [published online ahead of print May 9, 2022]
.
J Clin Pathol
.
2022
.
10.
Vyberg
M,
Nielsen
S.
Proficiency testing in immunohistochemistry–experiences from Nordic Immunohistochemical Quality Control (NordiQC)
.
Virchows Arch
.
2016
;
468
(1)
:
19
29
.
11.
Buttner
R,
Gosney
JR,
Skov
BG,
et al.
Programmed death-ligand 1 immunohistochemistry testing: a review of analytical assays and clinical implementation in non-small-cell lung cancer
.
J Clin Oncol
.
2017
;
35
(34)
:
3867
3876
.
12.
Torlakovic
E,
Lim
HJ,
Adam
J,
et al.
“Interchangeability” of PD-L1 immunohistochemistry assays: a meta-analysis of diagnostic accuracy
.
Mod Pathol
.
2020
;
33
(1)
:
4
17
.
13.
Batenchuk
C,
Albitar
M,
Zerba
K,
et al.
A real-world, comparative study of FDA-approved diagnostic assays PD-L1 IHC 28-8 and 22C3 in lung cancer and other malignancies
.
J Clin Pathol
.
2018
;
71
(12)
:
1078
1083
.
14.
US Food and Drug Administration.
Guidance for industry and FDA staff. Statistical guidance on reporting results from studies evaluating diagnostic tests
. .
15.
Thulin
A.
The cost of using exact confidence intervals for a binomial proportion
.
Electron J Statist
.
2014
;
8
(1)
:
817
840
.
16.
Fleiss
JL,
Levin
B,
Cho Paik
M.
Statistical Methods for Rates and Proportions
. 3rd ed.
Hoboken, NJ
:
John Wiley & Sons Inc
;
2003
:
598
626
.
17.
Bera
K,
Schalper
KA,
Rimm
DL,
Velcheti
V,
Madabhushi
A.
Artificial intelligence in digital pathology - new tools for diagnosis and precision oncology
.
Nat Rev Clin Oncol
.
2019
;
16
(11)
:
703
715
.
18.
Adam
J,
Le Stang
N,
Rouquette
I,
et al.
Multicenter harmonization study for PD-L1 IHC testing in non-small-cell lung cancer
.
Ann Oncol
.
2018
;
29
(4)
:
953
958
.
19.
Rimm
DL,
Han
G,
Taube
JM,
et al.
A prospective, multi-institutional, pathologist-based assessment of 4 immunohistochemistry assays for PD-L1 expression in non-small cell lung cancer
.
JAMA Oncol
.
2017
;
3
(8)
:
1051
1058
.
20.
Aeffner
F,
Zarella
MD,
Buchbinder
N,
et al.
Introduction to digital image analysis in whole-slide imaging: a white paper from the digital pathology association
.
J Pathol Inform
.
2019
;
10
(1)
:
9
.
21.
Hendry
S,
Byrne
DJ,
Wright
GM,
et al.
Comparison of four PD-L1 immunohistochemical assays in lung cancer
.
J Thorac Oncol
.
2018
;
13
(3)
:
367
376
.
22.
Torlakovic
E,
Albadine
R,
Bigras
G,
et al.
Canadian multicenter project on standardization of programmed death-ligand 1 immunohistochemistry 22C3 laboratory-developed tests for pembrolizumab therapy in NSCLC
.
J Thorac Oncol
.
2020
;
15
(8)
:
1328
1337
.
23.
Ilie
M,
Khambata-Ford
S,
Copie-Bergman
C,
et al.
Use of the 22C3 anti-PD-L1 antibody to determine PD-L1 expression in multiple automated immunohistochemistry platforms
.
PLoS One
.
2017
;
12
(8)
:
e0183023
.
24.
Neuman
T,
London
M,
Kania-Almog
J,
et al.
A harmonization study for the use of 22C3 PD-L1 immunohistochemical staining on Ventana’s platform
.
J Thorac Oncol
.
2016
;
11
(11)
:
1863
1868
.
25.
Ramos-Vara
JA,
Miller
MA.
When tissue antigens and antibodies get along: revisiting the technical aspects of immunohistochemistry–the red, brown, and blue technique
.
Vet Pathol
.
2014
;
51
(1)
:
42
87
.

Author notes

Supplemental digital content is available for this article at https://meridian.allenpress.com/aplm in the November 2023 table of contents.

Avraam, Giacomazzi, Vandebroek, and Rypens are employees of CellCarta. Wojcik, Desai, Dillon, and Benci are employees of and own stock in Bristol Myers Squibb.

Supplementary data