Prognostic Immune Markers in Non–Small Cell Lung Cancer Free

Immune responses within the tumor microenvironment are increasingly implicated as a determining factor in tumor progression and aggressiveness. Understanding these responses has allowed investigators to use immune markers to stratify prognosis of colorectal and ovarian cancer patients (1, 2). High levels of intratumoral effector memory cells in colorectal cancer and CD3+ T cells in ovarian cancer are associated with prolonged survival.

Published studies investigating immune markers as prognosticators in non–small cell lung cancer (NSCLC) are heterogeneous in histology (adenocarcinoma, squamous cell, and large cell) and stage (I–IV). In this review, we have identified prognostic immune markers in the tumor microenvironment and peripheral blood, as well as tumor expression of immune genes. The immune genes, from 17 published NSCLC gene-profiling studies, were evaluated for prognostic significance in the NIH Director's Challenge microarray (3), the largest multicenter gene-profiling study to date.

The interaction between tumor and immune cells in the tumor microenvironment is influenced by the type of immune cells [CD8+, CD4+, CD20+, and forkhead box P3 (FoxP3+); see Table 1], their density (as counted microscopically), and location (tumor nest and stroma; ref. 4). Among T lymphocytes, which comprise 80% of tumor-infiltrating lymphocytes (TIL) in NSCLC (5), CD8+ cytotoxic lymphocytes form the effector arm of adaptive immunity and are thought to have protective roles against tumors. However, 5 investigations in NSCLC show no correlation between CD8+ TIL infiltration and survival (6–10). Pertinent to NSCLC patients is the fact that CD8+ T-cell predominance is a characteristic of smoking-related chronic obstructive pulmonary disease (11), and it is thought to drive the progression of emphysema (12). In mouse models, activation of CD8+ T cells is impaired in the presence of cigarette smoke (13, 14), perhaps explaining this finding in NSCLC patients. In fact, studies that report correlation between CD8+ TILs and prognosis contradict each other (15, 16). Ruffini and colleagues showed, in their investigation of 1,290 NSCLC tumors, that TILs, mostly CD8+, were associated with prolonged survival, but only in a subset of squamous cell carcinomas (n = 549; ref. 15). This finding is contrasted by Wakabayashi and colleagues, who correlated tumoral CD8+ and shorter overall survival (OS) in NSCLC, especially in adenocarcinoma (16). Examining the functional significance of CD8+ T cells, Trojan and colleagues analyzed the mRNA ratio of IFN-γ to CD8, in which IFN-γ is an effector cytokine secreted by cytotoxic T cells, specifically in peritumoral areas and within tumor nests. Although the number of peritumoral CD8+ T cells correlated with the IFN-γ/CD8 ratio, this association was not observed intratumorally (8). This study suggests that, whereas the CD8+ T cells are able to infiltrate the tumor, they are not able to mount a robust antitumor response once within the tumor nest.

In contrast to CD8+ lymphocytes, stromal CD4+ (16, 17), CD20+ (17, 18), and colocalization of stromal CD8+ and CD4+ cells (9) have all shown association with improved survival. The significance of immune cell colocalization is highlighted in a study by Dieu-Nosjean and colleagues, who showed the presence of tertiary de novo lymphoid structure in the tumor microenvironment, which they named the tumor-induced bronchus-associated lymphoid tissue (Ti-BALT; ref. 19). In their investigation of 74 stage I to IIA NSCLCs, the presence of mature dendritic cells [DC; lysosome-associated membrane protein (Lamp+)] in Ti-BALT correlates with higher density of T and B lymphocytes and, more importantly, prolonged OS and disease-free survival (DFS; 4-year DFS of 51% versus 88%). The investigators postulated that effector immune cell colocalization within Ti-BALT could play a role in activating an antitumor immune response.

Regulatory T cells (Treg) suppress the host immune response and are thought to promote tumor growth. The protumor association of Treg tumor infiltration was examined by Shimizu and colleagues in a study of stage I to III 100 NSCLC tumors, showing that tumor-infiltrating FoxP3+ Tregs correlate with cyclooxygenase-2 (COX-2) expression and increased tumor recurrence (20). This finding is supported by preclinical findings from Sharma and colleagues, who showed that tumor-derived COX-2 and its product prostaglandin E₂ (PGE₂) induced in vitro lymphocyte expression of FoxP3. In vivo, inhibition of COX-2 reduced Treg activity, attenuated FoxP3 expression in TILs, and decreased tumor burden (21). Furthermore, urinary PGE-M, the major metabolite of PGE₂, was proposed as a biomarker to predict response to COX-2 inhibitors in NSCLC patients (22). Petersen and colleagues showed that the Treg/TIL combination risk index (the ratio of FoxP3+ to CD3+) correlates with disease-specific survival (DSS) in patients with stage I NSCLC (23). Patients with high-risk tumors (30% of tumors) experienced worse DSS (median 53 months) when compared with patients with intermediate (63 months) and low-risk tumors (>72 months), who had low FoxP3 and high CD3.

Tumor-associated macrophages (TAM) show both pro- and antitumor effects in the tumor microenvironment (24). Antitumor TAMs in NSCLC are of the M1 phenotype and accumulate intratumorally, whereas protumor TAMs of the M2 phenotype accumulate in the stroma and express interleukin-8 (IL-8), IL-10, and triggering receptor expressed on myeloid cells (TREM-1). IL-8 is an angiogenic factor, and the angiogenic role of TAM in NSCLC has been shown by correlating macrophage density with intratumor microvessel counts and poor patient outcomes (25). By doing reverse transcriptase (RT)-PCR for IL-8 in resected specimens, the study concludes that TAMs lead to poor patient outcome by their angiogenic role through IL-8 production. IL-10 is an immunosuppressive cytokine, and its expression by TAM has been observed more commonly in stages II, III, and IV, thus correlating with decreased OS (26). TAM expression of TREM-1, which can initiate and amplify an inflammatory response, is increased in malignant pleural effusions of NSCLC patients (27). Furthermore, in 68 stage I to III NSCLC patients, an increased level of TREM-1 high TAMs in resected specimens was an independent predictor of shorter OS.

In contrast to these findings, TAMs have also been associated with prolonged survival in a study of 144 stage I to IV NSCLCs (28). Recent studies have shed light on these contradictory findings, in which TAMs have been classified according to their location and phenotype. In an investigation of 175 stage I to III NSCLC tumors, Welsh and colleagues found that CD68+ macrophages in the stroma are associated with poor prognosis (5-year OS of 27% versus 35%), whereas intratumoral macrophages are associated with increased survival (5-year OS of 53% versus 8%; ref. 29).

Phenotypically, TAMs are comprised of 2 distinct types: M1, proinflammatory with antitumor activity, and M2, which are immunosuppressive, angiogenic, and protumor (30, 31). In a nested case–control study, 20 long-survival patients (median 93 months) had high tumor infiltration of the M1 phenotype compared with 20 poor-survival patients (median 8 months). M1 was characterized by the expression of human leukocyte antigen (HLA)-DR, inducible nitric oxide synthase (iNOS), myeloid-related protein (MRP) 8/14, and tumor necrosis factor (TNF)-α (32). Contrasting the antitumor association of M1 TAMs, high numbers of CD204+ M2 TAMs in stroma have been shown to correlate with decreased OS (33). CD204+ TAMs are also associated with increased tumor expression of IL-10 and monocyte chemoattractant protein (MCP)-1, 2 molecules that are known to attract macrophages.

Similar to TAMs, tumor-associated neutrophils (TAN) also have polarized functions. Although no published studies have investigated the role of TAN in human NSCLC, recent preclinical work by Fridlender and colleagues showed that in mouse models of lung cancer, TGF-β induces a population of TAN with protumor function (N2), whereas TGF-β blockade results in antitumor neutrophils (N1; ref. 34). Depletion of these N1 neutrophils resulted in increased tumor growth.

The universal availability of peripheral blood lymphocyte count has led to its investigation as a prognostic marker in NSCLC. In resected NSCLC patients (n = 177; 72% stages I and II), increasing total lymphocyte counts were associated with lower hazard ratios (HR) for death [HR 0.62, confidence interval (CI), 0.43–0.9; P = 0.012; ref. 35]. The same study further identified neutrophil to lymphocyte ratio (NLR) as a superior predictor of survival compared with pathologic stage (P = 0.001; ref. 35). Specifically, an NLR of >3.81 was found to be a significant predictor of survival in patients with stage I NSCLC. As for the tumor-promoting Tregs, they have been observed at an increased level in the peripheral blood of NSCLC patients compared with normal healthy volunteers (36, 37). These reports correlate with elevated serum and plasma levels of TGF-β and IL-10, both known promoters of Treg development, in NSCLC patients compared with healthy controls (38, 39). Peripheral blood levels of Tregs, TGF-β, and IL-10, however, have not been investigated in predicting clinical outcome of NSCLC patients.

Myeloid-derived suppressor cells (MDSC) represent a heterogeneous population of myeloid cells comprising immature macrophages, granulocytes, and DCs at early stages of differentiation that have protumor effects. MDSCs are mobilized from bone marrow into the peripheral blood by tumor-derived factors and accumulate in the tumor microenvironment, in which they exert their protumor effect by inhibiting T-cell proliferation and activation (40). Similar to Tregs, increased levels of MDSCs, marked by CD11b+/CD14−/CD15+/CD33+, were observed in the peripheral blood of advanced stage NSCLC patients (n = 87) compared with healthy controls (41). In addition, high levels of MDSCs were associated with decreased levels of CD8+ T cells, further supporting the protumor effect of MDSCs. No study to date, however, has investigated the prognostic significance of MDSCs in the tumor microenvironment of patient samples.

Stromal CD4+ lymphocytes, especially colocalized with CD8+ lymphocytes, stromal CD20+ B lymphocytes, Ti-BALT, and intratumoral macrophages (M1), are associated with prolonged survival in NSCLC. In contrast, FoxP3+ Tregs, stromal macrophages (M2), and their associated cytokines, IL-8, IL-10, and TREM-1, are associated with poor prognosis. Tumor expression and secretion of COX-2 and IL-10, respectively, lead to shorter survival. These prognostically significant findings are summarized in Fig. 1.

In the large, multisite, blinded validation study examining lung adenocarcinoma (NIH Director's Challenge), 1 of the four prognostic gene clusters contained genes related to immunologic function (3). Further pathway analysis showed immune function to be important in stratifying lung adenocarcinomas into prognostic subgroups (42). In addition to the NIH Director's Challenge study, we reviewed all published genomic studies done on lung cancer between 2000 and 2010 that used mRNA microarray or RT-PCR. From the 17 studies reviewed (total of 1,615 patients; Supplementary Table S1; refs. 3, 43–58), we identified the prognostic genes related to immune and/or inflammatory response (search and/or selection criteria shown in Supplementary Fig. S1), resulting in a list of 84 genes (Supplementary Table S2).

As shown in Fig. 2, 7 overlapping immune genes are identified from the 17 reviewed studies, and 2 genes are from a well-investigated chemokine axis in NSCLC, C-C motif chemokine 19 (CCL19)/C-C chemokine receptor type 7 (CCR7). CCR7 is a chemokine receptor expressed on naïve T cells, DCs, natural killer cells, and B cells, and its interaction with its ligands, CCL19 and C-C motif chemokine 21 (CCL21), plays a central role in lymphocyte trafficking and homing to lymph nodes (59, 60). In lung cancer, CCR7 expression on tumor cells by mRNA has been shown as an independent predictor of lymph node metastasis in an investigation of 71 NSCLC patients (61). It has been hypothesized that, when expressed on tumor cells, the homing mechanism of CCR7 contributes to the tumor's increased potential for lymphatic metastasis.

Although CCR7 may increase the metastatic potential of tumor cells, investigators have attempted to exploit the effect of this chemokine axis on the immune cells to promote tumor regression. CCL19, or Epstein-Barr virus–induced molecule 1 ligand chemokine (ELC), which is expressed in the T-cell zone of lymph nodes and DCs, is known to attract immune cells that express CCR7. Although no clinical study has investigated the role of CCL19 in NSCLC, its ability to promote tumor infiltration of T cells and DCs has been shown in preclinical models to decrease tumor burden. Intratumoral injection of CCL19 led to increased influx of CD4+ and CD8+ T cells and DCs, increase in antitumor factors such as IFN-γ and IL-12, and a decrease in the immunosuppressive molecule TGF-β (62, 63).

CCL21, the other ligand of CCR7, has been investigated as a potential adjunct to DC-based therapy, owing to its ability in promoting colocalized tumor infiltration of DCs and lymphocyte effector cells (64). Intratumoral injection of DCs transduced to express CCL21 reduced tumor burden in a mouse model of spontaneous bronchoalveolar carcinoma. On the basis of these findings, a phase I clinical trial is currently investigating the intratumoral injection of CCL-21–expressing DCs in stage IIIb to IV NSCLC patients refractory to standard therapy (65).

A Cox proportional hazards model was used to evaluate the association between DFS and the expression profiles of the 84 identified prognostic immune genes in the Director's Challenge data set (3). This analysis revealed 17 genes that were significantly associated with recurrence (Table 2). We employed the Ingenuity Pathway analysis to examine the networks shared by the 17 genes (Fig. 3). STAT3 was a common signal transduction pathway highlighted in this analysis, which has been implicated in cancer inflammation and immunity (66).

Although prognostic immune genes are identified in NSCLC, the cells of origin (tumor cells, tumor-infiltrating immune cells, or both) of these genes remain to be explored. One of the 17 prognostic genes, IL-7R, is expressed by both the tumor and immune cells. When expressed on the tumor, IL-7R expression leads to upregulation of VEGF-D in vitro and also shorter OS in NSCLC patients (67). On the other hand, it has been shown in vitro that the expression of IL-7R on CD8+ T cells signifies a long-lived memory phenotype (68). How the expression of these genes modulates the tumor immune microenvironment is a future area of investigation. CXCR4 is the receptor for stromal-derived factor-1alpha (CXCL12). In lung adenocarcinoma, tumor expression of CXCL12 has been shown to correlate with accumulation of CXCR4-expressing immune cells, 30% of which were protumor regulatory T cells (69).

The immune microenvironment surrounding the tumor is influenced by interactions among tumor, immune cells (TILs, TAMs, TANs, and MSDCs), and cytokines, which shift the environment to pro- or antitumor. In NSCLC, stromal CD4+ T lymphocytes, especially colocalized with CD8+ T cells, stromal CD20+ B lymphocytes, Ti-BALT, and intratumoral M1 macrophages predicted prolonged patient survival. FoxP3+ Tregs and stromal M2 macrophages are associated with poor clinical outcome. We examined gene expression profiling studies for insights into the NSCLC tumor microenvironment. In reviewing 17 gene profile studies, we identified 17 prognostic immune genes by comparing their expression to tumor recurrence in lung adenocarcinoma patients. Further studies using RT-PCR and immunohistochemical analysis are necessary to determine the source cell (tumor versus immune) and how they influence the tumor immune microenvironment. Future studies investigating a large patient series uniform in histology and stage and assessing tumor-immune interaction in both tumor nest and tumor-associated stroma may yield significant information.

No potential conflicts of interest were disclosed.

We thank Eduardo Morales (figure) and Erlin Daley (editorial) for their assistance in preparation of the manuscript.

Supported in part by American Association of Cancer Research (AACR) Lung Cancer Translational Research Award; American Association for Thoracic Surgery (AATS),Third Edward D. Churchill Research Scholarship; International Association for the Study of Lung Cancer (IASLC) Young Investigator Award; National Lung Cancer Partnership/LUNGevity Foundation Research Grant; William H. Goodwin and Alice Goodwin and the Commonwealth Foundation for Cancer Research and the Experimental Therapeutics Center; New York State Empire Clinical Research Investigator Program (ECRIP); and Stony Wold-Herbert Fund.