Context.—

Identification of gene mutations that are indicative of epithelial-mesenchymal transition and a noninflammatory immune phenotype may be important for predicting response to immune checkpoint inhibitors.

Objective.—

To evaluate the utility of multiplex immunofluorescence for immune profiling and to determine the relationships among tumor immune checkpoint and epithelial-mesenchymal transition genomic profiles and the clinical outcomes of patients with nonmetastatic non–small cell lung cancer.

Design.—

Tissue microarrays containing 164 primary tumor specimens from patients with stages I to IIIA non–small cell lung carcinoma were examined by multiplex immunofluorescence and image analysis to determine the expression of programmed death ligand-1 (PD-L1) on malignant cells, CD68+ macrophages, and cells expressing the immune markers CD3, CD8, CD57, CD45RO, FOXP3, PD-1, and CD20. Immune phenotype data were tested for correlations with clinicopathologic characteristics, somatic and germline genetic variants, and outcome.

Results.—

A high percentage of PD-L1+ malignant cells was associated with clinicopathologic characteristics, and high density of CD3+PD-1+ T cells was associated with metastasis, suggesting that these phenotypes may be clinically useful to identify patients who will likely benefit from immunotherapy. We also found that ZEB2 mutations were a proxy for immunologic ignorance and immune tolerance microenvironments and may predict response to checkpoint inhibitors. A multivariate Cox regression model predicted a lower risk of death for patients with a high density of CD3+CD45RO+ memory T cells, carriers of allele G of CTLA4 variant rs231775, and those whose tumors do not have ZEB2 mutations.

Conclusions.—

Genetic variants in epithelial-mesenchymal transition and immune checkpoint genes are associated with immune cell profiles and may predict patient outcomes and response to immune checkpoint blockade.

Non–small cell lung cancer (NSCLC) accounts for about 85% of all lung cancers. The prognosis of patients with NSCLC is highly dependent on the tumor-node-metastasis (TNM) stage; as the volume of disease increases, survival declines dramatically.1  Surgery is the standard treatment for patients with stage I and stage II NSCLC and for selected patients with stage IIIA disease with the possibility of complete tumor resection. The addition of adjuvant cisplatin-based chemotherapy to surgery improved 5-year survival rates by 5% to 10%, but no significant therapeutic innovation has been established thereafter, and the overall 5-year survival rate remains below 50%.

Over the past few years, immunotherapy with immune checkpoint inhibitors has led to remarkable progress in non–oncogene-addicted metastatic NSCLC.2  Most of these new immunotherapeutic approaches for advanced NSCLC are now being studied in the context of nonmetastatic NSCLC.3  However, not all NSCLC patients receive the same benefit from immunotherapy as a first or second line of treatment. To date, programmed death ligand-1 (PD-L1) is the only validated biomarker for immunotherapy response in NSCLC patients. However, this biomarker is not sufficiently robust and poses many challenges. For example, some patients with more than 50% PD-L1+ tumor cells do not have a response to anti–programmed death receptor-1 (PD-1)/PD-L1 treatment; conversely, some patients whose tumors do not express PD-L1 see good responses.46 

Some studies have suggested that the heterogeneity of PD-1/PD-L1 axis expression and response to PD-1/PD-L1 treatment in NSCLC depends on alterations in other factors, such as immune evasion mechanisms and the tumor immune microenvironment. Expression of genes involved in the epithelial-mesenchymal transition (EMT) and immune checkpoints are also known to control cancer progression.716  However, the associations among these factors and their association with immunotherapy response remain to be determined.

We hypothesized that genetic variants that are indicative of EMT and host immune factors predict patients' outcomes and responses to immune checkpoint blockade. Identification of mutations that can serve as a proxy for a noninflammatory immune phenotype would further enable us to predict response to immune checkpoint blockade. Such information would be particularly important in clinical settings where next-generation sequencing (NGS) is available, but comprehensive studies of tumor-associated immune cells (TAICs) are not. Bearing this in mind, we determined the extent to which immune and genomic profiles predicted clinical outcomes of patients with NSCLC.

PATIENTS AND METHODS

Cases and Specimens

We obtained archival formalin-fixed, paraffin-embedded histologic tumor sections from 164 patients diagnosed with NSCLC who had undergone surgical resection between January 1, 1995, and December 31, 2015, and who had not received neoadjuvant therapy. Patients had been treated at the Hospital das Clinicas of the State University of São Paulo Medical School, the Heart Institute of the University of São Paulo, the Cancer Institute of São Paulo, and the A.C. Camargo Cancer Center, São Paulo, Brazil. The histologic diagnoses were reviewed by 2 experienced lung pathologists, who judged their accuracy based on the World Health Organization 2015 classification system.17  The specimens included 94 adenocarcinomas, 51 squamous cell carcinomas (SCCs), and 19 large cell carcinomas (LCCs). Tissue microarrays were constructed from the primary resected tumors: 3 cores (diameter, 1.5 mm) of representative tumor tissue were taken from a representative formalin-fixed, paraffin-embedded block from each patient case. Clinicopathologic data, including pathologic TNM staging, which was determined according to the guidelines of the International Association for the Study of Lung Cancer, 8th edition,1  were obtained from medical records (Table 1).

Table 1

Clinicopathologic Characteristics of 164 Patients With Non–Small Cell Lung Cancer

Clinicopathologic Characteristics of 164 Patients With Non–Small Cell Lung Cancer
Clinicopathologic Characteristics of 164 Patients With Non–Small Cell Lung Cancer

Our study received approval from the institutional review boards of the institutions involved in the project (process number H-1404-100-572). A waiver of the requirement for informed consent was obtained from the review boards of all participating institutions.

Multiplex Immunofluorescence Staining

Multiplex immunofluorescence (mIF) staining was performed using methods similar to previously described and validated ones.18  Consecutive 4-μm-thick tissue microarray sections were stained using an automated staining system (BOND-RX, Leica Biosystems, Buffalo Grove, Illinois) using 2 different panels. Panel 1 consisted of pancytokeratin AE1/AE3 (dilution 1:300; epithelial cell positive, Dako, Carpinteria, California), PD-L1 (dilution 1:100; clone E1L3N, Cell Signaling Technology, Danvers, Massachusetts), PD-1 (dilution 1:250; clone EPR4877-2, Abcam, Cambridge, Massachusetts), CD3 (dilution 1:100; T-cell lymphocytes, Dako), CD8 (dilution 1:20; cytotoxic T cells, clone C8/144B, Thermo Fisher Scientific, Waltham, Massachusetts), and CD68 (dilution 1:450; macrophages, clone PG-M1, Dako). Panel 2 consisted of AE1/AE3 (epithelial cell-positive, dilution 1:300; Dako), CD3 (dilution 1:100; T cell lymphocytes, Dako), CD57 (dilution 1:40; natural killer cells, clone HNK-1, BD Biosciences, San Jose, California), CD45RO (ready to use; memory T cells, clone UCHL1, Leica Biosystems), FOXP3 (dilution 1:50; regulatory T cells, clone 206D, BioLegend, San Diego, California), and CD20 (dilution 1:100; B-cell lymphocytes, Dako). All the markers were stained in sequence using their respective fluorophores, which were contained in the Opal 7 kit (Cat #NEL797001KT, Akoya Biosciences/PerkinElmer, Waltham, Massachusetts) after IF validation to obtain a uniform, specific, and correct signal across all channels and ensure a well-balanced staining pattern during the multiplex staining.18  The correct signal from each fluorophore was also defined and optimized between 10 and 30 counts of intensity to maintain a good balance and similar thresholds of intensity across all the antibodies. In parallel, to detect possible variations in staining and optimal separation of the signal, positive and negative (autofluorescence) controls were included during the staining. Autofluorescence controls with an expected spectrum of 488 nm can accurately remove the autofluorescence from all the labeling signals during the analysis process. The stained slides were scanned using a multispectral microscope (Vectra Polaris 3.0 imaging system, Akoya Biosciences/PerkinElmer) under fluorescence conditions (Supplemental Figure 1; see supplemental digital content containing 4 tables and 1 figure at www.archivesofpathology.org in the October 2020 table of contents).

mIF Quantitation

Multispectral images of tumor sections from each core were analyzed with inForm 2.2.1 software (Akoya Biosciences/PerkinElmer). Individual cells, which were defined by nuclear staining and identified by the inForm cell segmentation tool, were used in a phenotype pattern-recognition learning algorithm to characterize the colocalization of the cell populations labeled with 2 panels.19  Panel 1 labeling was as follows: (1) malignant cells (MCs) with the AE1/AE3+ marker, including those with and without PD-L1 expression (AE1/AE3+PD-L1+ and AE1/AE3+PD-L1, respectively); (2) T lymphocytes, including CD3+ and pan–T-cell markers (cytotoxic T cells [CD3+CD8+], antigen-experienced T cells [CD3+PD-1+], and other CD3+ T cells); (3) CD68+ macrophages; and (4) macrophages expressing PD-L1 (CD68+PD-L1+). Panel 2 labeling was as follows: (1) MCs with the AE1/AE3+ marker; (2) T lymphocytes, including CD3+ and pan–T-cell markers (memory T cells [CD3+CD45RO+], natural killer T cells [CD3+CD57+], memory/regulatory T cells [CD3+CD45RO+FOXP3+], and other CD3+ T cells); and (3) B-cell lymphocytes (CD20+).

The individual cell phenotype report produced by the inForm software was processed using Microsoft Excel 2010, and a final summary of the data, which contained the median number of cells/mm2 for each phenotype and the median percentage of macrophages and MCs expressing PD-L1, was created for statistical analysis. If the percentage of MCs or macrophages expressing PD-L1 was greater than the median value, PD-L1 expression was considered positive. If the percentage of macrophages or MCs expressing PD-L1 in a sample was lower than or equal to the median, PD-L1 expression was considered negative. Likewise, if the density of TAICs in a cell population was greater than the median, it was considered high density; if the density was less than or equal to the median, it was considered low.

In addition, as proposed by Teng and colleagues,20  we identified 4 different types of tumor microenvironments on the basis of the density of CD3+ tumor-infiltrating lymphocytes (TILs) (a density greater than the median was considered high) and the PD-L1 expression level of MCs. The 4 microenvironment types were type I, adaptive immune resistance (PD-L1+/TIL+); type II, immunologic ignorance (PD-L1/TIL); type III, intrinsic induction (PD-L1+/TIL); and type IV, tolerance (PD-L1/TIL+).

Genomic Analysis

Following previously published procedures,21  we extracted DNA from fresh-frozen lung cancer tissue obtained from 70 of the 164 patients in our study; 33 of the fresh-frozen samples were from patients with adenocarcinoma, 24 were from patients with SCC, and 13 were from patients with LCC. To study the tumor genomic landscape from different angles, we used targeted NGS to evaluate single-nucleotide variants and short deletions/insertions (indels) in NSCLC samples, including 13 genes, 121 targets, and 372 amplicons associated with these genes. We verified the results with DNA sequencing analysis with an Illumina TruSeq Custom Amplicon Panel (Supplemental Table 1). The genes examined were grouped according to their functions as immune checkpoints (CD274 [PD-L1], CD276, CTLA4, PDCD1LG2 [PD-L2], LAG3, and VTCN1 [B7H4]) or EMT genes (CD44, CDH1 [E-cadherin], TGFβ1, VIM, MMP2, ZEB1, and ZEB2). Libraries were sequenced with an Illumina MiSeq platform using MiSeq v2 (300 cycles). Primary data analysis was performed using the TruSeq Amplicon pipeline (alignment to reference genome hg19) via the Basespace Sequence Hub. Illumina VariantStudio v2.2 software was used for filtering and variant annotation. Variants with frequency lower than 1% in the study population were considered single-nucleotide variants, and variants with frequency higher than 1% were defined as somatic variants. Filtered retained variants had to have a total coverage depth of 100 reads or more and a variant allele frequency of at least 5% to be considered of adequate quality to be interpreted. The potential pathogenicity of the allele variants was evaluated with 3 publicly available algorithms (Polyphen2, Sift, ClinVar). The distribution of mutations in EMT and immune checkpoint genes in our study cohort was compared with data from The Cancer Genome Atlas. All mutations were tested for correlations with patients' clinicopathologic characteristics, TAIC levels, and outcomes.

Statistical Analysis

The χ2 test or Fisher exact test was used to examine differences in categorical variables, and the Wilcoxon rank sum test and Kruskal-Wallis test were used to detect differences in continuous variables among groups of patients. In addition, a general linear regression model was used to test the relationships between continuous variables and several other variables, and the residuals were examined to ensure that they were approximately normally distributed. The overall survival (OS) distributions for the patients were estimated using the Kaplan-Meier method. Overall survival was defined as the interval from surgery to death and was estimated using the Kaplan-Meier method, and the log-rank test was used to compare survival among the groups of interest. Regression analysis of the OS data was performed using the Cox proportional hazards model. Variables shown by univariate analysis to be significantly associated with survival were entered into a Cox proportional hazards regression model for multivariate analysis. The statistical software programs SPSS (version 22, IBM, Armonk, New York) and S-Plus (version 8.04, TIBCO, Palo Alto, California) were used to perform the analyses. P values less than or equal to .05 were deemed statistically significant.

RESULTS

Clinicopathologic Characteristics

The clinical characteristics of the 164 patients in our cohort are summarized in Table 1 by histologic tumor type. Similar distributions of age and sex were found among the histologic types. Of the 106 patients with a history of tobacco smoking, 51 (48%) had adenocarcinoma, 48 (45%) had SCC, and 7 (7%) had LCC. The median follow-up duration was 43 months (range, 6–180 months). During the follow-up period, disease progression (local recurrence and distant metastasis to liver, bones, lungs, or brain) occurred in 30 of the 164 patients: 16 (53%) with adenocarcinoma, 12 (40%) with SCC, and 2 (7%) with LCC. All patients with distant metastases also had lymph node metastases. Of the 164 patients, 34 (21%) received adjuvant therapy, and 56 (34%) had died of their disease by last follow-up.

PD-L1 Expression in MCs and Macrophages

The median cell densities of MCs and macrophages by histologic tumor type are shown in Table 2. Image analysis and mIF revealed no significant differences in the median densities of PD-L1+ MCs among the histologic types: the median density was 0.39 cells/mm2 for adenocarcinoma, 0.48 cells/mm2 for SCC, and 1.59 cells/mm2 for LCC (P = .60) (Table 2). Using the median percentage of MCs expressing PD-L1 across all the samples as a cutoff, we identified 45 of 94 (47%) PD-L1+ adenocarcinoma samples, 25 of 51 (49%) PD-L1+ SCC samples, and 6 of 19 (31%) PD-L1+ LCC samples (Figure, A through F). For all 3 histologic types, the density of CD68+PD-L1+ TAMs was higher than that of PD-L1+ MCs.

Table 2

Median Densities of Various Immune Marker–Expressing Cells According to Histologic Tumor Type (N = 164)

Median Densities of Various Immune Marker–Expressing Cells According to Histologic Tumor Type (N = 164)
Median Densities of Various Immune Marker–Expressing Cells According to Histologic Tumor Type (N = 164)

Multiplex immunofluorescence images of representative non–small cell lung cancer core tumor sections analyzed for panel 1 (pancytokeratin, cyan; PD-L1, orange; CD3, red; CD8, pink; PD-1, green; CD68, yellow; DAPI, blue) and panel 2 (pancytokeratin, cyan; CD3, red; CD8, pink; CD45RO, yellow; GranzymeB, orange; FOXP3, green; DAPI, blue) markers. The images reflect the variations in cell phenotypes observed in adenocarcinoma (A and D, panels 1 and 2, respectively), squamous cell carcinoma (B and E, panels 1 and 2, respectively), and large cell carcinoma (C and F, panels 1 and 2, respectively), especially in regard to PD-L1 expression by malignant cells (original magnification ×200).

Multiplex immunofluorescence images of representative non–small cell lung cancer core tumor sections analyzed for panel 1 (pancytokeratin, cyan; PD-L1, orange; CD3, red; CD8, pink; PD-1, green; CD68, yellow; DAPI, blue) and panel 2 (pancytokeratin, cyan; CD3, red; CD8, pink; CD45RO, yellow; GranzymeB, orange; FOXP3, green; DAPI, blue) markers. The images reflect the variations in cell phenotypes observed in adenocarcinoma (A and D, panels 1 and 2, respectively), squamous cell carcinoma (B and E, panels 1 and 2, respectively), and large cell carcinoma (C and F, panels 1 and 2, respectively), especially in regard to PD-L1 expression by malignant cells (original magnification ×200).

Clinicopathologic Correlations With PD-L1 and TAICs

Certain immune cell phenotypes were significantly associated with patients' clinicopathologic characteristics, as shown in Supplemental Table 2. A general linear regression model demonstrated that low PD-L1 expression on MCs was significantly associated with tobacco use (P = .04) and adenocarcinoma (P = .04). Low density of natural killer T cells (CD3+CD57+) was also associated with tobacco use (P = .04). Low density of cytotoxic T cells (CD3+CD8+) was associated with LCC (P = .04). In addition, low density of memory T cells (CD3+CD45RO+) was associated with female sex (P = .01) and age 65 years or older (P = .04). Importantly, high density of antigen-experienced T cells (CD3+PD-1+) was associated with brain metastasis (P = .04), and, curiously, high density of B cells (CD20+) was significantly more common among patients who did not receive adjuvant therapy than among those who did (P = .04). All these associations remained statistically significant after allowing for the contribution of other clinicopathologic characteristics, as determined by multivariate analysis.

Characterization of Tumor Immune Microenvironments

We observed significantly higher densities of CD3+ T cells (P = .02), CD3+CD8+ cytotoxic T cells (P = .04), and CD3+PD-1+ antigen–experienced T cells (P = .01) in adenocarcinoma and LCC than in SCC (Table 2). No other statistically significant differences were found among the histologic types.

By examining the PD-L1 expression of MCs in conjunction with the density of CD3+ TILs, as proposed by Teng and colleagues,20  among the different histologic types, we identified all 4 types of tumor microenvironments in our NSCLC samples (Table 3). Type I, adaptive immune resistance (PD-L1+/TIL+), was observed in 11 of 94 (12%) adenocarcinoma samples, 7 of 51 (14%) SCC samples, and 4 of 19 (21%) LCC samples. Type II, immunologic ignorance (PD-L1/TIL), was found in 39 of 94 (41%) adenocarcinoma samples, 28 of 51 (55%) SCC samples, and 7 0f 19 (37%) LCC samples. Type III, intrinsic induction (PD-L1+/TIL), was found in 3 of 94 (3%) adenocarcinoma samples, 4 of 51 (88%) SCC samples, and 3 of 19 (16%) LCC samples. Type IV, tolerance (PD-L1/TIL+), was found in 41 of 94 (42%) adenocarcinoma samples, 12 of 51 (24%) SCC samples, and 5 of 19 (26%) LCC samples.

Table 3

Distribution of Types of Immune Microenvironments in 164 Non–Small Cell Lung Cancer Specimensa

Distribution of Types of Immune Microenvironments in 164 Non–Small Cell Lung Cancer Specimensa
Distribution of Types of Immune Microenvironments in 164 Non–Small Cell Lung Cancer Specimensa

Correlations Between Genomic Profiles, Clinicopathologic Characteristics, and Immune v

Using NGS, we examined the occurrence of variants in several cancer-related genes in fresh-frozen lung cancer tissue specimens obtained from 70 of the 164 NSCLC patients included in our study, and we correlated the mutation status of these genes with patients' clinicopathologic characteristics and immune cell populations. The distribution of mutations in EMT and immune checkpoint genes in our study cohort was compared with data from The Cancer Genome Atlas (Supplemental Table 3). We found similar frequencies of mutations in immune checkpoint and EMT genes in our NSCLC cohort and the Cancer Genome Atlas cohort. The most frequently mutated genes in our cohort were ZEB2 (n = 10; 14%), MMP2 (n = 4; 6%), ZEB1 (n = 2; 3%), CDH1 (n = 2; 3%), and CD44 (n = 2; 3%), all of them involved in EMT. On multivariate analysis, CD44 mutations were significantly associated with tobacco use (P ≤ .05). The small number of mutations in TGFβ1 and VIM did not allow for multivariate analysis to test for associations with clinicopathologic characteristics.

The genotype and allele frequencies of 5 variants, rs7854303 (PDCD1LG2 [PD-L2]), rs2297136 (CD274 [PD-L1]), rs870849 (LAG3), rs10754339 (VTCN1 [B7H4]), and rs231775 (CTLA4), in our cohort are summarized in Table 4. Only genotype CC and the C allele of PD-L2 variant rs7854303 were identified in all patients. On multivariate analysis, the only significant difference in clinicopathologic characteristics by genotype was that the AA genotype of the CTLA4 rs231775 variant was more common in stage II and III than in stage I tumors (P = .02) (Supplemental Table 2). No association was found between rs2297136 (PD-L1), rs870849 (LAG3), or rs10754339 (VTCN1) and any clinicopathologic characteristics.

Table 4

Distribution of Genetic Variants in 70 Patients With Non–Small Cell Lung Cancer and Association of These Variants With Overall Survival

Distribution of Genetic Variants in 70 Patients With Non–Small Cell Lung Cancer and Association of These Variants With Overall Survival
Distribution of Genetic Variants in 70 Patients With Non–Small Cell Lung Cancer and Association of These Variants With Overall Survival

Our investigation of the association between genomic profiles and immune cell profiles revealed significant correlations between high density of CD3+PD-1+ antigen–experienced T cells and the G allele (GG+GA) of CTLA4 rs231775 (R = 0.33; P = .01), high density of CD3+ TILs and ZEB1 mutation (R = 0.27; P = .04), and high density of CD3+CD8+ cytotoxic T cells and ZEB1 mutation (R = 0.34; P = .01) (Supplemental Table 4).

Survival Analysis

Preliminary examination of Kaplan-Meier survival curves (data not shown) demonstrated that in this study, patients with pathologic disease stages II and III had approximately the same hazard for survival, with a median survival time of 40 months for both groups. Thus, we coded overall pathologic stage as a single dummy variable with a value of 0 for stage I and a value of 1 for stages II and III. The results of the Cox proportional hazards regression models appear in Tables 4 and 5.

Table 5

Univariate and Multivariate Cox Proportional Hazards Models of Associations of Overall Survival With Clinicopathologic Characteristics and Immune Cell Profiles

Univariate and Multivariate Cox Proportional Hazards Models of Associations of Overall Survival With Clinicopathologic Characteristics and Immune Cell Profiles
Univariate and Multivariate Cox Proportional Hazards Models of Associations of Overall Survival With Clinicopathologic Characteristics and Immune Cell Profiles

The following clinicopathologic variables were significantly associated with poor OS: female sex, tobacco use, tumor size greater than 4.5 cm, N1 status, tumor stage higher than II, SCC histologic type, and absence of adjuvant treatment (Table 5). We also found that lower-than-median densities of CD3+CD8+, CD3+CD45RO+, and CD3+CD45RO+FOXP3+ T cells in the tumor microenvironment predicted shorter OS, as did ZEB2-mutant tumors. In addition, CTLA4 variant rs231775 with AA genotype was associated with shorter OS, whereas allele G was significantly associated with longer OS (adjusted hazard ratio = 2.091, 95% CI = 0.233–2.219, P = .05) (Table 4). No other variables included in the multivariate analysis were significantly associated with survival.

Multivariate analysis controlling for CD3+CD8+ T cell density, CD3+CD45RO+ T cell density, CTLA4 variants, and ZEB2 mutations revealed that tumor size, N stage, histologic type, and adjuvant treatment significantly predicted survival time. However, when CD3+CD45RO+ T cells, CTLA4 rs231775 G allele carrier genotype, and ZEB2 wild type were included as covariates, their relationship to survival was much stronger. Whereas the overall likelihood ratio of the Cox model using tumor size, N staging, histologic type, and treatment was just 21.83, the likelihood ratio with stage, CD3+CD45RO+ T cells, CTLA4 variant rs231775 G allele carrier genotype, and ZEB2 wild-type tumors was 31.18.

DISCUSSION

In the present study, we explored the quantitative relationships among TAIC density and EMT and immune checkpoint genomic profiles in 164 patients with nonmetastatic NSCLC. Our results indicate that incorporating genomic profile data into quantitative immune profiling produced several important findings. First, our results validate mIF for immune profiling in NSCLC. Second, we discovered associations between clinicopathologic characteristics of NSCLC and PD-L1 expression by MCs and immune cells. Third, our analyses revealed that higher densities of antigen-experienced T cells were associated with metastasis. Fourth, we found that microenvironment types II (immunologic ignorance) and IV (tolerance) were the most frequent microenvironments in the tested NSCLC tissues. Fifth, we found that patients with early-stage disease, carriers of the G allele of CTLA4 variant rs231775, patients with wild-type ZEB2, and patients with a high density of CD3+CD45RO+ T cells had longer OS than other patients. Finally, we found that ZEB2 mutations were indicative of EMT and served as a proxy for noninflammatory immune phenotypes (immunologic ignorance and immune tolerance). This information may be important for predicting response to immune checkpoint blockade, particularly in clinical settings where NGS is available but immune profiling is not.

To better understand the role of immune and genomic profiling in early-stage NSCLC, our study was carried out in a cohort of patients with nonmetastatic disease. Our investigation included adenocarcinoma, SCC, and LCC, and, to our knowledge, it is the first study to compare genetic and immune cell profiles among the 3 major histologic types of NSCLC and to determine their impact on patient outcomes.

Using mIF and image analysis, we found that PD-L1 expression by MCs and other cells was associated with several clinicopathologic characteristics and with disease outcomes. We found no differences in PD-L1 expression on MCs among the histologic tumor types, but our results showing that 31.51% of LCC samples had positive PD-L1 expression on MCs agreed with those of several studies of large cell neuroendocrine carcinoma of the lung.22,23  This similarity suggests that patients with the large cell subtype of NSCLC may benefit from checkpoint inhibitor therapies. In agreement with findings reported by Calles and colleagues,24  we found that high expression of PD-L1 on MCs was associated with tobacco use and the adenocarcinoma histologic type. Interestingly, we found that low densities of natural killer T cells were also associated with tobacco use; this finding was in concordance with Hogan and colleagues'25  study, which showed a significant reduction of natural killer cells in tobacco users. That reduction was accompanied by significant defects in cytokine production, suggesting that depletion of natural killer T cells can interfere with antitumor immune responses.

Similarly, our study showed that low densities of cytotoxic T cells were associated with LCC and that low densities of memory T cells were associated with female sex and older age, suggesting that alterations in the regulation of the immune system can increase the likelihood of tumor invasion and progression. As suggested by Prado-Garcia and colleagues,26  alterations in the immune system induced by tumor cells can lead to T-cell dysfunction and reduce T cells' ability to attack tumors. It is also known that increasing age can produce downregulation of immune cells.27 

Interestingly, shorter survival time was associated with high densities of CD3+PD-1+ antigen–experienced T cells, suggesting that PD-1 plays an important role in facilitating NSCLC metastasis. A similar finding was made by Wang et al,28  who found that, in cervical cancer, the interaction between PD-1 and PD-L1 can trigger immune responses and facilitate tumor growth and metastasis. Most, if not all, malignancies trigger different immune responses through intricate interactions between tumor cells and the host's immune cells,29  suggesting that the progression of cancer is influenced by complex tumor–3immune cell interactions driven by characteristics of the tumor, the behavior of inflammatory cells, and genetic mutations.

Our study also identified differences in the immune profiles of the histologic types of NSCLC that may affect the choice of therapy. Densities of different subpopulations of T cells, including cytotoxic T cells, were higher in adenocarcinoma and LCC than in SCC; these results agreed with those of our previous study.30  In addition, we also observed higher densities of CD3+PD-1+ antigen–experienced T cells in adenocarcinoma and LCC than in SCC, suggesting that adenocarcinoma and LCC provoke more immune exhaustion and, thus, that immunotherapeutic intervention may be warranted for these 2 histologic types. Previous studies have shown that activation of the PD-1 pathway mediates inhibition of T cells and that PD-1 expression is an independent predictive factor for prognosis.31  Using the tumor microenvironment criteria proposed by Teng and colleagues,20  we determined that the most common microenvironments in our cohort were immunologic ignorance (PD-L1/TIL) and immunologic tolerance (PD-L1/TIL+), suggesting that complementary strategies, as proposed by Teng et al,20  will be necessary to induce a response to immunotherapy in patients with these microenvironments.

Our findings also highlighted the role of EMT in genomic profiling to predict outcomes in NSCLC. The EMT is activated in cancer cells and is induced by transcription factors such as Snail, Twist, and ZEB1/ZEB2 to modify the transcriptional machinery, alter translation and protein stability, and promote invasion and metastasis.32  In fact, we demonstrated that ZEB1 mutations were inversely correlated with levels of CD3+ and CD3+CD8+ T cells, suggesting a molecular link between EMT and immunosuppression, 2 key drivers of cancer progression. We also found that mutations in the EMT-associated gene CD44 were associated with adenocarcinoma and SCC and with tobacco use, in agreement with other studies.3234  Interestingly, tobacco use induces metaplastic bronchial squamous epithelium that exhibits increased hyaluronan and CD44 expression in the proliferating basal layers. In premalignant bronchial dysplasia, the entire epithelial thickness shows aberrant hyaluronan-CD44 expression, indicating that squamous malignant transformation is closely associated with CD44 expression and with tobacco use.35  We also found that CD44-mutant tumors were associated with high densities of CD20+ B cells, which suggests that CD44 also may participate in several other immune-related processes, as previously demonstrated by Zittermann and colleagues.36 

Over the last decade, significant progress has been made in systemic therapies to improve the length of good-quality survival in patients with metastatic NSCLC. However, many questions remain about the behavior of nonmetastatic NSCLC. One such issue relates to the expected 5-year survival rate, which ranges from 36% to 92% for patients with early-stage (ie, stage I, II, or IIIA) disease.3  Another question relates to current studies of immunotherapy in the context of nonmetastatic NSCLC.3  Not all NSCLC patients have a response to immunotherapy because the heterogeneity of immune cell profiles in NSCLC tumors probably depends on other factors, for instance EMT and immune checkpoint genes' control of cancer progression.7,37  In this context, in our cohort, multivariate analysis reliably predicted longer OS for patients with a high density of CD3+CD45RO+ memory T cells and those with the G allele of CTLA4 variant rs231775 and shorter OS when nonmetastatic NSCLC tumors have ZEB2 mutations. In fact, the presence of a high density of CD3+CD45RO+ immune cells in the tumor region is correlated with favorable clinical outcomes in epithelial lung cancers.38  In addition, several studies have assessed the contribution of CTLA4 polymorphisms in the context of immune checkpoint inhibition.3944  Our multivariate Cox analysis revealed that the CTLA4 rs231775 AA genotype is an adverse prognostic indicator for NSCLC patients. Although associations between CTLA4 and outcomes have been reported, these lacked independent validation. As shown by Deng and colleagues,44 CTLA4 expression was not significantly prognostic in their cohort of 1432 patients on univariate analysis but was significant in their cohort of patients with available data for multivariate analysis.

Our analysis has a number of limitations that could not be addressed in this first quantitative study. The first of these limitations is the small cohort, which included 164 retrospectively collected cases. However, the maximum follow-up time was 168 months, the robustness of clinical variables was ascertained, and survival information was available. A second limitation is that we included only 70 tissue samples in our genomic analysis. However, a preliminary examination of Kaplan-Meier survival curves (data not shown) demonstrated that patients with pathologic stages II and III disease had approximately the same hazard for survival, with a median survival time of 40 months for both groups. A third limitation of the study concerns the noninclusion of normal tissue or blood samples for the genomic analysis, which may have affected the allele frequency determination. A fourth limitation was the tissue microarray format we used, which could have induced underrepresentation or overrepresentation of PD-L1 due to tumor heterogeneity and the small area of analysis. Fifth, we lacked available data regarding responses to anti–PD-1/anti–PD-L1 monoclonal antibodies in our patient population. Sixth, even with accumulating evidence supporting the contribution of germline genetics to host immunity, knowledge about host genetic factors as predictive biomarkers of clinical outcomes is limited. Moreover, at present, no systematic study of genetic variants as surrogates for immunotherapy outcomes at the genome-wide level is available. Larger study cohorts will be required for the discovery of low-penetrance germline loci associated with immunotherapy outcomes. This clearly indicates the need for a large international collaboration pooling patient resources. The implementation of genome-wide association studies coupled with large-scale immune phenotyping will help to identify common genetic variations associated with a large population–based scale.42  Finally, mIF can be used to simultaneously identify specific proteins on a single slide. This technique allows study of the immune contexture using paraffin-embedded tumor tissues. In addition, multiplexed image analysis methods are highly advantageous for investigating immune evasion mechanisms and for discovering potential biomarkers to assess mechanisms of action and to predict response to a given treatment.45,46  The combination of NGS data with mIF may facilitate disease management and the appropriate application of immunotherapeutic agents.

In conclusion, incorporating a genomic profile into quantitative mIF revealed a variety of factors associated with the behavior of nonmetastatic NSCLC, including specific clinicopathologic characteristics and immunomodulatory control of local disease progression. Thus, these tools might be useful for predicting whether patients with early-stage, nonmetastatic NSCLC will benefit from combinations of targeted therapy, chemotherapy, and immunotherapy.

Editorial support was provided by Amy Ninetto, PhD, ELS, of Scientific Publications, Research Medical Library, The University of Texas MD Anderson Cancer Center.

References

References
1.
Goldstraw
P,
Chansky
K,
Crowley
J,
et al.
The IASLC Lung Cancer Staging Project: proposals for revision of the TNM stage groupings in the forthcoming (eighth) edition of the TNM Classification for Lung Cancer
.
J Thorac Oncol
.
2016
;
11
(1)
:
39
51
.
2.
Ghysen
K,
Vansteenkiste
J.
Immunotherapy in patients with early stage resectable nonsmall cell lung cancer
.
Curr Opin Oncol
.
2019
;
31
(1)
:
13
17
.
3.
Deslypere
G,
Gullentops
D,
Wauters
E,
Vansteenkiste
J.
Immunotherapy in non-metastatic non-small cell lung cancer: can the benefits of stage IV therapy be translated into earlier stages?
Ther Adv Med Oncol
.
2018
;
10
:
1758835918772810
.
4.
Yu
H,
Boyle
TA,
Zhou
C,
et al.
PD-L1 expression in lung cancer
.
J Thorac Oncol
.
2016
;
11
:
964
975
.
5.
Sun
JM,
Zhou
W,
Choi
YL,
et al.
Prognostic significance of PD-L1 in patients with non-small cell lung cancer: a large cohort study of surgically resected cases
.
J Thorac Oncol
.
2016
;
11
(7)
:
1003
1011
.
6.
Reck
M,
Rodríguez-Abreu
D,
Robinson
AG,
et al.
Pembrolizumab versus chemotherapy for PD-L1–patients with non-small cell lung cancer: a large cohort study of surgically resected cases
.
N Engl J Med
.
2016
;
375
:
1823
1833
.
7.
Kim
S,
Koh
J,
Kim
MY,
et al.
PD-L1 expression is associated with epithelial-to-mesenchymal transition in adenocarcinoma of the lung
.
Hum Pathol
.
2016
;
58
:
7
14
.
8.
Marwitz
S,
Scheufele
S,
Perner
S,
Reck
M,
Ammerpohl
O,
Goldmann
T.
Epigenetic modifications of the immune-checkpoint genes CTLA4 and PDCD1 in non-small cell lung cancer results in increased expression
.
Clin Epigenetics
.
2017
;
9
:
51
.
9.
Papanicolau-Sengos
A,
Yang
Y,
Pabla
S,
et al.
Identification of targets for prostate cancer immunotherapy
.
Prostate
.
2019
;
79
(5)
:
498
505
.
10.
O'Leary
K,
Shia
A,
Schmid
P.
Epigenetic regulation of EMT in non-small cell lung cancer
.
Curr Cancer Drug Targets
.
2018
;
18
(1)
:
89
96
.
11.
Chae
YK,
Chang
S,
Ko
T,
et al.
Epithelial-mesenchymal transition (EMT) signature is inversely associated with T-cell infiltration in non–small cell lung cancer (NSCLC)
.
Sci Rep
.
2018
;
8
(1)
:
2918
.
12.
Chen
L,
Gibbons
DL,
Goswami
S,
et al.
Metastasis is regulated via microRNA-200/ZEB1 axis control of tumour cell PD-L1 expression and intratumoral immunosuppression
.
Nat Commun
.
2014
;
5
:
5241
.
13.
Kurimoto
R,
Ebata
T,
Iwasawa
S,
et al.
Pirfenidone may revert the epithelial-to-mesenchymal transition in human lung adenocarcinoma
.
Oncol Lett
.
2017
;
14
(1)
:
944
950
.
14.
Alsuliman
A,
Colak
D,
Al-Harazi
O,
et al.
Bidirectional crosstalk between PD-L1 expression and epithelial to mesenchymal transition: significance in claudin-low breast cancer cells
.
Mol Cancer
.
2015
;
14
:
149
.
15.
Ma
Y,
Adjemian
S,
Mattarollo
SR,
et al.
Anticancer chemotherapy-induced intratumoral recruitment and differentiation of antigen-presenting cells
.
Immunity
.
2013
;
38
(4)
:
729
741
.
16.
Mazzaschi
G,
Madeddu
D,
Falco
A,
et al.
Low PD-1 expression in cytotoxic CD8(+) tumor-infiltrating lymphocytes confers an immune-privileged tissue microenvironment in NSCLC with a prognostic and predictive value
.
Clin Cancer Res
.
2018
;
24
(2)
:
407
419
.
17.
Travis
WD,
Brambilla
E,
Nicholson
AG,
et al.
The 2015 World Health Organization Classification of Lung Tumors: impact of genetic, clinical and radiologic advances since the 2004 classification
.
J Thorac Oncol
.
2015
;
10
(9)
:
1243
1260
.
18.
Parra
ER,
Uraoka
N,
Jiang
M,
et al.
Validation of multiplex immunofluorescence panels using multispectral microscopy for immune-profiling of formalin-fixed and paraffin-embedded human tumor tissues
.
Sci Rep
.
2017
;
7
(1)
:
13380
.
19.
Gorris
MAJ,
Halilovic
A,
Rabold
K,
et al.
Eight-color multiplex immunohistochemistry for simultaneous detection of multiple immune checkpoint molecules within the tumor microenvironment
.
J Immunol
.
2018
;
200
(1)
:
347
354
.
20.
Teng
MW,
Ngiow
SF,
Ribas
A,
Smyth
MJ.
Classifying cancers based on T-cell infiltration and PD-L1
.
Cancer Res
.
2015
;
75
(11)
:
2139
2145
.
21.
Machado-Rugolo
J,
Fabro
AT,
Ascheri
D,
et al.
Usefulness of complementary next-generation sequencing and quantitative immunohistochemistry panels for predicting brain metastases and selecting a treatment outcomes of non-small cell lung cancer
.
Hum Pathol
.
2019
;
83
:
177
191
.
22.
Eichhorn
F,
Harms
A,
Warth
A,
Muley
T,
Winter
H,
Eichhorn
ME.
PD-L1 expression in large cell neuroendocrine carcinoma of the lung
.
Lung Cancer
.
2018
;
118
:
76
82
.
23.
Tsuruoka
K,
Horinouchi
H,
Goto
Y,
et al.
PD-L1 expression in neuroendocrine tumors of the lung
.
Lung Cancer
.
2017
;
108
:
115
120
.
24.
Calles
A,
Liao
X,
Sholl
LM,
et al.
Expression of PD-1 and its ligands, PD-L1 and PD-L2, in smokers and never smokers with KRAS-mutant lung cancer
.
J Thorac Oncol
.
2015
;
10
(12)
:
1726
1735
.
25.
Hogan
AE,
Corrigan
MA,
O'Reilly
V,
et al.
Cigarette smoke alters the invariant natural killer T cell function and may inhibit anti-tumor responses
.
Clin Immunol
.
2011
;
140
(3)
:
229
235
.
26.
Prado-Garcia
H,
Romero-Garcia
S,
Aguilar-Cazares
D,
Meneses-Flores
M,
Lopez-Gonzalez
JS.
Tumor-induced CD8+ T-cell dysfunction in lung cancer patients
.
Clin Dev Immunol
.
2012
;
2012
:
741741
.
27.
Fulop
T,
Larbi
A,
Kotb
R,
de Angelis
F,
Pawelec
G.
Aging, immunity, and cancer
.
Discov Med
.
2011
;
11
(61)
:
537
550
.
28.
Wang
S,
Li
J,
Xie
J,
et al.
Programmed death ligand 1 promotes lymph node metastasis and glucose metabolism in cervical cancer by activating integrin beta4/SNAI1/SIRT3 signaling pathway
.
Oncogene
.
2018
;
37
(30)
:
4164
4180
.
29.
Mantovani
A,
Allavena
P,
Sica
A,
Balkwill
F.
Cancer-related inflammation
.
Nature
.
2008
;
454
(7203)
:
436
444
.
30.
Parra
ER,
Behrens
C,
Rodriguez-Canales
J,
et al.
Image analysis-based assessment of PD-L1 and tumor-associated immune cells density supports distinct intratumoral microenvironment groups in non-small cell lung carcinoma patients
.
Clin Cancer Res
.
2016
;
22
(24)
:
6278
6289
.
31.
Paulsen
EE,
Kilvaer
TK,
Khanehkenari
MR,
et al.
Assessing PDL-1 and PD-1 in non-small cell lung cancer: a novel immunoscore approach
.
Clin Lung Cancer
.
2017
;
18
(2)
:
220
233
e8.
32.
Gonzalez
DM,
Medici
D.
Signaling mechanisms of the epithelial-mesenchymal transition
.
Sci Signal
.
2014
;
7
(344)
:
re8.
33.
Leung
EL,
Fiscus
RR,
Tung
JW,
et al.
Non-small cell lung cancer cells expressing CD44 are enriched for stem cell-like properties
.
PLoS One
.
2010
;
5
(11)
:
e14062
.
34.
Chou
YE,
Hsieh
MJ,
Chiou
HL,
Lee
HL,
Yang
SF,
Chen
TY.
CD44 gene polymorphisms on hepatocellular carcinoma susceptibility and clinicopathologic features
.
Biomed Res Int
.
2014
;
2014
:
231474
.
35.
Penno
MB,
August
JT,
Baylin
SB,
et al.
Expression of CD44 in human lung tumors
.
Cancer Res
.
1994
;
54
(5)
:
1381
1387
.
36.
Zittermann
SI,
Achino
BI,
Agriello
EE,
Halperin
N,
Ramhorst
RE,
Fainboim
L.
Modulation of CD44 in acute lymphoblastic leukemia identifies functional and phenotypic differences of human B cell precursors
.
Eur J Haematol
.
2001
;
66
(6)
:
377
382
.
37.
Koh
J,
Go
H,
Keam
B,
et al.
Clinicopathologic analysis of programmed cell death-1 and programmed cell death-ligand 1 and 2 expressions in pulmonary adenocarcinoma: comparison with histology and driver oncogenic alteration status
.
Mod Pathol
.
2015
;
28
(9)
:
1154
1166
.
38.
Sheng
SY,
Gu
Y,
Lu
CG,
Zou
JY,
Hong
H,
Wang
R.
The distribution and function of human memory T cell subsets in lung cancer
.
Immunol Res
.
2017
;
65
(3)
:
639
650
.
39.
Breunis
WB,
Tarazona-Santos
E,
Chen
R,
Kiley
M,
Rosenberg
SA,
Chanock
SJ.
Influence of cytotoxic T lymphocyte-associated antigen 4 (CTLA4) common polymorphisms on outcome in treatment of melanoma patients with CTLA-4 blockade
.
J Immunother
.
2008
;
31
(6)
:
586
590
.
40.
Queirolo
P,
Morabito
A,
Laurent
S,
et al.
Association of CTLA-4 polymorphisms with improved overall survival in melanoma patients treated with CTLA-4 blockade: a pilot study
.
Cancer Invest
.
2013
;
31
(5)
:
336
345
.
41.
Hamid
O,
Schmidt
H,
Nissan
A,
et al.
A prospective phase II trial exploring the association between tumor microenvironment biomarkers and clinical activity of ipilimumab in advanced melanoma
.
J Transl Med
.
2011
;
9
:
204
220
.
42.
Kirchhoff
T,
Ferguson
R.
Germline genetics in immuno-oncology: from genome-wide to targeted biomarker strategies
.
Methods Mol Biol
.
2020
;
2055
:
93
117
.
43.
Lim
YW,
Chen-Harris
H,
Mayba
O,
et al.
Germline genetic polymorphisms influence tumor gene expression and immune cell infiltration
.
Proc Natl Acad Sci U S A
.
2018
;
115
:
E11701
E11710
.
44.
Deng
L,
Gyorffy
B,
Na
F,
et al.
Association of PDCD1 and CTLA-4 gene expression with clinicopathological factors and survival in non-small-cell lung cancer: results from a large and pooled microarray database
.
J Thorac Oncol
.
2015
;
10
(7)
:
1020
1026
.
45.
Hofman
P,
Badoual
C,
Henderson
F,
et al.
Multiplexed immunohistochemistry for molecular and immune profiling in lung cancer—just about ready for prime-time?
Cancers (Basel)
.
2019
;
11
(3)
:
283
.
46.
Parra
ER,
Francisco-Cruz
A,
Wistuba II. State-of-the-art of profiling immune contexture in the era of multiplexed staining and digital analysis to study paraffin tumor tissues
.
Cancers (Basel)
.
2019
;
11
(2)
:
247
.

Author notes

Supplemental digital content is available for this article at www.archivesofpathology.org in the October 2020 table of contents.

Supported by The University of Texas Lung Cancer Specialized Programs of Research Excellence (grant P50CA70907) and the Foundation for the Support of Research of the State of São Paulo (FAPESP 2013/14277-4, FAPESP 2018-20403-6).

The authors have no relevant financial interest in the products or companies described in this article.

Supplementary data