|Home | About | Journals | Submit | Contact Us | Français|
Mounting evidence suggests that tumor-infiltrating immune cells have prognostic value for patients with solid organ malignancies. Our aim was to investigate the prognostic significance of the immune microenvironment in patients with stage I lung adenocarcinoma (ADC).
Using tissue microarray and immunohistochemistry, we investigated eight types of tumor-infiltrating immune cells in the tumor nest and tumor-associated stroma as well as tumor expression of five cytokines in a uniform cohort of 956 patients with stage I lung ADC (478 each in training and validation cohorts).
Although a high density of stromal forkhead box P3 (FoxP3) –positive cells was associated with shorter recurrence-free probability (RFP; P = .043), the relative proportion of stromal FoxP3 to CD3 was a stronger predictor of recurrence (5-year RFP, 85% for high v 77% for low ratio; P = .004). High expression of tumor interleukin-12 receptor β2 (IL-12Rβ2) was associated with better outcome (5-year RFP, 90% for high v 80% for low expression; P = .026), whereas high expression of tumor IL-7R was associated with worse outcome (5-year RFP, 76% for high v 86% for low expression; P = .001). In multivariate analysis, these immune markers were independently associated with recurrence. Although IL-7R remained significant for poor overall survival, all the markers remained prognostic for recurrence in patients with stages IA and IB disease as well as for patients with tumors ≤ 2 cm.
Our investigation confirms the biologic and prognostic significance of the tumor immune microenvironment for patients with stage I lung ADC and provides support for its use to stratify clinical outcome and immunotherapeutic interventions.
Therapeutic decisions in solid malignancies are dictated by the TNM staging system, which is based on anatomic factors. The need to advance beyond the TNM staging system has been addressed by examining the tumor immune microenvironment, which is influenced by the type, density, and location of tumor-infiltrating immune cells.1–6 For colorectal cancer, an immune score based on the density of cytotoxic CD8+ and memory CD45R0+ lymphocytes in the tumor center and the invasive margin has proven to be a stronger predictor of clinical outcome than the conventional staging system.7 These results have led to discussions about incorporating immunologic parameters into the routine diagnostic and prognostic assessment of tumors.8
More than 160,000 deaths resulting from lung cancer are projected to occur during 2012 in the United States alone, making it the most lethal malignancy.9 In the recent National Lung Screening Trial, computed tomography (CT) screening was shown to decrease mortality related to lung cancer.10 On the basis of this trial, the National Comprehensive Cancer Network recently issued guidelines for lung cancer screening that recommend helical low-dose CT for high-risk patients.11 With the earlier detection afforded by CT screening, it is anticipated that increasing numbers of patients will be diagnosed with early-stage lung cancer. Surgery alone remains the standard of care for patients with stage I lung cancer, yet as many as 27% will experience disease recurrence within 5 years.12
The prognostic utility of tumor-infiltrating immune cells has been investigated for lung cancer13–17; however, these studies investigated heterogeneous stages and histologies and used survival as the main end point, rendering their application to early-stage patients difficult. In our study, we investigated the prognostic utility of tumor-infiltrating immune cells as well as cytokines in a uniform cohort of patients with stage I lung adenocarcinoma (ADC), the most common histologic type of lung cancer. To elucidate the biologic significance of the tumor immune microenvironment, we investigated immune cells that have shown prognostic significance in both the tumor nest and tumor-associated stroma for other solid malignancies—CD3 (pan T cell), CD4 (helper T cell), CD8 (cytotoxic T cell), CD20 (B cell), CD45R0 (memory T cell), forkhead box P3 (FoxP3; regulatory T cell), CD56 (natural-killer cell), and CD68 (macrophage)—as well as tumor expression of cytokines CCR7, CXCL12, CXCR4, interleukin-7 receptor (IL-7R), and IL-12Rβ2.18 Most importantly, we chose recurrence, rather than survival, as the study end point, because recurrence is more clinically relevant to patients with stage I disease and is not confounded by factors that might influence survival.
Our training cohort comprised patients diagnosed with pathologic stage I lung ADC at Memorial Sloan-Kettering Cancer Center between 1995 and 2005 and has been well characterized19; patients in the validation cohort received the same diagnosis between 2002 and 2009. Institutional review board approval was obtained. Data were obtained from the prospectively maintained thoracic surgery database. No patients received neoadjuvant chemotherapy or radiation therapy.
All hematoxylin and eosin–stained slides were re-reviewed by a pathologist (K.K.), and problem cases were reviewed by two pathologists (K.K., W.D.T.). All slides were evaluated for lymphatic, vascular, and visceral pleural invasion19 as well as predominant histologic subtype according to the new IASLC/ATS/ERS (International Association for the Study of Lung Cancer/American Thoracic Society/European Respiratory Society) classification of lung ADC.20 Staging was based on the seventh edition of the TNM Cancer Staging Manual.21 All patients were observed until death or last follow-up (assessed April 2011). Slides for the validation cohort were reviewed by the same pathologists.
Formalin-fixed, paraffin-embedded tumor specimens were used for tissue microarray construction. For each tumor, the slide with the most severe inflammatory reaction was chosen. From each tumor in the training cohort, four representative cores with the most abundant inflammatory reaction, 0.6 mm in size, were marked—two each from the tumor nest and tumor-associated stroma. For the validation cohort, the number of cores was increased—six from the tumor nest and three from the tumor-associated stroma. Lung tissues from eight normal samples were included as controls. The standard avidin-biotin-peroxidase complex technique was used for immunohistochemical staining for the antihuman antibodies (Appendix Table A1, online only).
Representative images and scoring of immunohistochemistry are shown in the Data Supplement. Under a high-power field (magnification, ×400), each core was scored for degree of immune-cell infiltration in the tumor nest and tumor-associated stroma semiquantitatively. Normal controls showed sparse staining for all stains. The scores for each core were averaged to give one score per patient. For each patient, immune-cell infiltration was defined by a score of 1 (average, 1 to 1.67), 2 (average, 1.67 to 2.33), or 3 (average, > 2.33). For statistical analysis, a score of 1 was considered to be low, and 2 and 3 were considered to be high.
For cytokines, we scored the tumor stain on the basis of intensity and distribution, as previously described.22 For CCR7, CXCL12, CXCR4, and IL-7R, a score of < 1 was regarded as negative and ≥ 1 as positive. For IL12-Rβ2, the score was based on intensity only, because the distribution was diffuse. IL12-Rβ2 intensity of < 1 was regarded as negative and ≥ 1 as positive. Normal controls demonstrated staining for IL12-Rβ2, as previously described,23 but not for the other four cytokines.
Associations between clinicopathologic variables and each marker were analyzed using Pearson's χ2 test. Recurrence-free probability (RFP) was estimated using the Kaplan-Meier method, with follow-up starting at the time of surgery. Patients whose disease did not recur during study follow-up were censored at the time of last contact or death without documented recurrence. Differences in RFP between subgroups of patients were compared using the nonparametric log-rank test. Multivariate analyses were performed using the Cox proportional hazards regression model to estimate the effect of the immune markers of interest on RFP, with adjustment for clinicopathologic factors. All significance tests were two sided, and all used a 5% level of significance. Statistical analyses were conducted using SAS statistical software (version 8.02; SAS Institute, Cary, NC).
A total of 956 patients were included in the study, 478 each in training and validation cohorts. Clinicopathologic variables are listed in Table 1. The validation cohort had a higher percentage of patients with stage IA disease, likely as an effect of this cohort being more recent, resulting in a higher percentage of wedge resections, tumors without lymphatic and pleural invasions, and lepidic-predominant morphology.
Of the available clinicopathologic variables, sex (P = .002), stage (IA v IB; P < .001), lymphatic invasion (P = .013), vascular invasion (P = .01), and tumor morphology (P < .001) were significantly associated with recurrence (Table 2), in concordance with published results. The presence of visceral pleural invasion reflected a tendency for higher rates of recurrence (P = .075).
Each immune cell in the tumor nest and tumor-associated stroma was independently assessed for its ability to predict recurrence (Table 2). A high density of FoxP3-positive cells in the stroma was significantly associated with recurrence (P = .043; Fig 1A). Because FoxP3-positive regulatory T cells are a subset of the entire T-cell population, we next investigated the relative proportion of FoxP3-positive to CD3+ cells. Interestingly, we observed that among patients with high stromal FoxP3, those with high stromal CD3 had better RFP compared with those with low stromal CD3. In fact, the cohort with high stromal FoxP3 and concurrent high-level stromal CD3 infiltration demonstrated a recurrence rate similar to that of the group with low stromal FoxP3 (data not shown). On the basis of this observation, we devised a stromal FoxP3 risk index in which tumors containing high stromal FoxP3 and low stromal CD3 are considered high risk and tumors with low stromal FoxP3 are considered low risk, as are tumors with high stromal FoxP3 and concurrent high-level stromal CD3 infiltration. We found stromal FoxP3 risk index to be a strong predictor of recurrence; low-risk patients had a 5-year RFP of 85%, compared with 77% for high-risk patients (P = .004; Fig 1B). These results were replicated in the validation cohort (Figs 1E, E,1F).1F). None of the other immune cells had significant prognostic value.
Although the five investigated cytokines are known to be expressed on both tumor and immune cells, our immunohistochemical analysis was optimized for staining on tumor cells. Of the five cytokines, IL-12Rβ2 and IL-7R were found to be prognostic (Table 2). Higher-level expression of IL-12Rβ2 was associated with reduced risk of recurrence (5-year RFP, 90% for high v 80% for low level; P = .026; Fig 1C), whereas higher-level expression of IL-7R was associated with increased risk of recurrence (5-year RFP, 76% for high v 86% for low level; P = .001; Fig 1D). These results were replicated in our validation cohort (Figs 1G, G,1H).1H). The associations between these two cytokine expressions and the densities of immune cells are shown in the Data Supplement. IL-12Rβ2 expression was associated with low tumor density of CD68 (P = .006), whereas IL-7R expression was associated with high density of stromal CD3 (P = .02), tumor CD68 (P < .001), and stromal CD68 (P = .002; CD68 is the marker for tumor-associated macrophages). We did not find significant associations between expression of the other three cytokines (CCR7, CXCL12, and CXCR4) and recurrence.
After identifying three prognostic immune markers—stromal FoxP3 risk index, tumor IL-12Rβ2, and tumor IL-7R—we next performed a multivariate analysis, with adjustment for other currently known prognostic clinicopathologic factors, including sex, disease stage (IA v IB), and lymphatic invasion. The multivariate analysis confirmed that all three immune markers remained independently associated with recurrence (Table 3).
To gain additional biologic insights, we next investigated the association of the three immune markers with clinicopathologic factors (Fig 2). Combining results from the two cohorts, we found a significant association between stromal FoxP3 risk index score and lymphatic invasion (P = .038), vascular invasion (P < .001), and high-grade morphology (P < .001). IL-12Rβ2 expression had a significant association with low-grade morphology (P = .019) and presence of EGFR mutation and lack of KRAS mutation (P = .0075). IL-7R expression had a significant association with higher stage (P < .001), larger tumor size (P < .0013), lymphatic invasion (P < .001), vascular invasion (P < .001), high-grade morphology (P < .001), and presence of KRAS mutation and lack of EGFR mutation (P < .001).
The current TNM staging system relies solely on anatomic factors and is limited in its ability to discriminate a subset of patients with stage I disease with poor clinical outcome. In fact, for stage I lung ADC, tumor size is the only standard prognosticator available. In our study, we have demonstrated the prognostic power of immunologic parameters for stage I lung ADC, identifying three immune markers that are predictive of recurrence. This immunologic observation has both prognostic and therapeutic implications.
Tumor-infiltrating immune cells have shown prognostic value for several solid malignancies, including colorectal,1,2,7 ovarian,3 and breast cancers4 (Appendix Table A2, online only). Galon et al1 have advocated, through their work in colorectal cancer, the use of three important parameters of tumor-infiltrating lymphocytes (TILs)—type, density, and location—to predict clinical outcome. In our study of a uniform cohort of patients with stage I lung ADC, we investigated eight markers of TILs and found the FoxP3/CD3 ratio in tumor-associated stroma to be significantly associated with recurrence. FoxP3 is a marker of regulatory T cells, a subset of lymphocytes known to suppress the host immune response. In patients with lung cancer, regulatory T cells are thought to play protumor roles,24 and their association with worse prognosis has been shown for all histologic types, including ADC.17,25 Interestingly, FoxP3 in the stroma only—and not in the tumor nest—was associated with recurrence, emphasizing the importance of assessing the location of TILs within the tumor microenvironment. In fact, the significance of immune cells in the tumor stroma has been shown in non–small-cell lung cancer. In patients with stages I to IIA disease, Dieu-Nosjean et al26 demonstrated the presence of tertiary de novo lymphoid structure in the tumor microenvironment, a structure they termed the tumor-induced bronchus-associated lymphoid tissue (Ti-BALT). The presence of mature dendritic cells in Ti-BALT correlated with prolonged overall and disease-free survival.
Furthermore, we demonstrate that in addition to type, density, and location, a fourth characteristic of TILs—the relative proportion of pro- and antitumor immune cells—is an important parameter. Our close examination of patients with high densities of stromal FoxP3 revealed that among these patients, a concurrent high density of stromal CD3 predicted better outcome. This suggests that even in the presence of high stromal FoxP3, a high density of CD3 may overcome the protumor effects of FoxP3-positive regulatory T cells. In addition to revealing prognostic value, this finding has significant implications for devising potential immunomodulatory therapy for patients with lung ADC; an intervention that decreases FoxP3 and increases CD3 would likely be beneficial. Of interest, cyclophosphamide has been shown to modulate the tumor immune microenvironment by depleting regulatory T cells.27 Also, in a murine melanoma model, activation of a T-cell costimulatory receptor, 4-1BB, has been shown to result in decreased tumor infiltration of regulatory T cells.28 Because FoxP3-positive regulatory T cells are thought to be a subset of CD4+ T cells, we also investigated stromal FoxP3 density and its relation to CD4 density. Although stromal densities of CD4 and FoxP3 showed significant correlation, combining CD4 and FoxP3 densities did not result in significant prognostic findings (Data Supplement).
In our analysis of chemokine expression on tumor cells, we found two to be of prognostic significance, one with antitumor associations (IL-12Rβ2) and one with protumor associations (IL-7R). IL-12Rβ2 is one of the two subunits that form the receptor for IL-12. Tumors expressing IL-12Rβ2 were associated with low-grade morphology, EGFR mutation, and less recurrence. Given that IL-12Rβ2 expression is observed on normal lung epithelium,23 it is plausible that lung ADC progression is accompanied by loss of IL-12Rβ2 expression. In fact, in preclinical models, mice deficient in IL-12Rβ2 have been shown to spontaneously develop lung ADC.29
After observing that IL-12Rβ2 expression is associated with less-aggressive tumors, we investigated how tumor expression of IL-12Rβ2 is associated with the protumor stromal environment. In a group of patients identified as high risk by the stromal FoxP3 risk index, we observed that patients with high-level expression of IL-12Rβ2 (n = 91) experienced less-frequent recurrence than those with low-level expression (n = 25; 5-year RFP, 88% for high-level v 73% for low-level expression; P = .086). This suggests that even in the presence of unfavorable stromal immune-cell infiltrates, high-level expression of IL-12Rβ2 on tumor cells may play a protective role. Thus, therapies targeted to maintain IL-12Rβ2 expression on tumor cells are of interest. In patients with stages I to IV lung ADC, methylation of the IL-12Rβ2 gene was shown to be associated with less mRNA expression in vivo and shorter survival.30 The recent publication of a phase I/II study of DNA methyltransferase inhibitor and a histone deacetylase inhibitor in patients with recurrent metastatic lung cancer is promising.31 Furthermore, in preclinical models, T cells genetically modified to secrete IL-12 have shown intrinsic resistance to regulatory T cells.32 IL-12–secreting T cells are of special interest, because this resistance could potentially overcome the protumor associations of regulatory T cells. In addition, the association of IL-12 and IL-12Rβ2 expression on tumor cells warrants further investigation.
In contrast to IL-12Rβ2 expression, IL-7R expression was associated with aggressive tumor features: high-grade morphology, lymphovascular invasion, larger tumor size, KRAS mutation, and more-frequent recurrence. IL-7 and IL-7R are implicated in lung cancer lymphangiogenesis via c-Fos/C-Jun–dependent vascular endothelial growth factor D (VEGF-D) upregulation.33 In breast cancer, IL-7R has been shown to induce tumor growth and lymphangiogenesis through upregulation of VEGF-D.34,35 Its ligand, IL-7, is produced by stromal and epithelial cells and plays a central role in T-cell development, in addition to providing a potent lymphocyte-survival factor through the JAK-STAT pathway.36 The role of the IL-7/IL-7R axis in the tumor immune microenvironment in lung ADC warrants more investigation.
The prognostic significance of the immune markers was further strengthened by their ability to stratify within currently known prognosticators—stage (IA and IB; Figs 3A to to3F),3F), tumors ≤ 2 cm (Figs 3G to to3I),3I), and morphologic grade (Data Supplement). Patients with stage IB disease with a high stromal FoxP3 risk index or high-level expression of IL-7R experienced outcomes similar to those of patients with stage II disease, for whom adjuvant chemotherapy is recommended. The ability to prognosticate within tumors ≤ 2 cm is especially important, because this represents a population in which incidence is expected to increase with the adoption of more-widespread CT screening, but in which a standardized prognostic factor is lacking. When we assessed overall survival as an end point, IL-7R remained a significant prognosticator in both the training (P = .007) and validation cohorts (P = .02), whereas findings were not significant for FoxP3/CD3 risk index (P = .21) or IL-12Rβ2 (P = .51). The ability of IL-7R to prognosticate both recurrence-free and overall survival merits further investigation of its biologic role (Data Supplement).
One limitation of our study is the semiquantitative nature of immunohistochemical scoring. Although a digital analysis was attempted using the Aperio ScanScope XT (Aperio, Vista, CA), the results were confounded by anthracotic pigments picked up as positive stains, a unique problem encountered in lung specimens. Accurate analysis and discrimination of the tumor from the tumor-associated stroma were best achieved by direct visualization under the microscope.
Our findings shed light on the complex tumor immune microenvironment in stage I lung ADC. First, we demonstrated that in the tumor-associated stroma, immune infiltrates rich in FoxP3-positive regulatory T cells create a protumor environment, and this environment may be overcome when there is a concurrently high density of CD3+ lymphocytes. Second, IL-12Rβ2 is expressed on normal lung epithelium and on less-aggressive tumors but to a lesser extent on more-aggressive tumors. Furthermore, tumors expressing IL-12Rβ2 tend to do well despite an unfavorable immune environment. Third, IL-7R expression on tumors is associated with aggressive features.
In conclusion, we have identified prognostic immune factors in this first, to our knowledge, large-scale study of the tumor immune microenvironment in patients with stage I lung ADC, a population anticipated to increase with widespread CT screening. For a population in which tumor size is currently the main prognostic factor, our results provide important prognostic tools and demonstrate the feasibility of using a multidisciplinary approach to advance beyond the limitations of the current staging system. More importantly, these findings provide a crucial foundation for future investigations into immunomodulatory therapies for lung ADC.
We thank Joe Dycoco, BA, of the Memorial Sloan-Kettering Cancer Center (MSKCC) Thoracic Surgery Division for his help with the thoracic service lung cancer database; David Sewell, MA, MFA, of the MSKCC Thoracic Surgery Division for editing the manuscript; and Irina Linkov of the MSKCC Department of Pathology for performing the immunohistochemical staining.
|CD3||Mouse||Monoclonal||Dako (Glostrup, Denmark)||1:1,600|
|CD4||Goat||Polyclonal||R&D Systems (Minneapolis, MN)||1:100|
|FoxP3||Mouse||Monoclonal||Abcam (Cambridge, United Kingdom)||1:2,000|
|CD56||Mouse||Monoclonal||Lab Vision (Fremont, CA)||1:50|
|CCR7||Rabbit||Monoclonal||Epitomics (Burlingame, CA)||1:100|
|IL-7R||Mouse||Polyclonal||Santa Cruz Biotechnology (Santa Cruz, CA)||1:2,000|
|IL12-Rβ2||Goat||Polyclonal||Santa Cruz Biotechnology||1:100|
Abbreviations: FoxP3, forkhead box P3; IL-7R, interleukin-7 receptor; IL-12Rβ2, interleukin-12 receptor β2.
|Cancer Type||Study||No. of Patients||Disease Stage||End Point||Validation Cohort||Location of Immune-Cell Infiltration||No. of Immune Markers Studied||Prognostic Immune Marker|
|Colorectal||Pagès et al2||377||I to IV||Survival||No||Tumor center, invasive margin||2||CD45RO|
|Galon et al1||603||I to IV||Survival||Yes||Tumor center, invasive margin||4||CD3, CD8, CD45RO ratio|
|Pagès et al6||602||I, II||Recurrence, survival||Yes||Tumor center, invasive margin||3||CD8, CD45RO|
|Mlecnik et al7||599||I to IV||Recurrence, survival||Yes||Tumor center, invasive margin||5||CD8, CD45RO|
|Breast||Mahmoud et al4||1,334||Not given||Survival||Yes||Intratumoral, adjacent and distant stroma||1||CD8|
|Lung||Al-Shibli et al16||335||I to IIIA (NSCLC)||Survival||No||Tumor, stroma||5||Stromal CD4, CD8|
|Current study||956||I (ADC)||Recurrence||Yes||Tumor, stroma||13||Stromal FoxP3/CD3, tumor IL-12Rβ2, IL-7R|
Abbreviations: ADC, adenocarcinoma; FoxP3, forkhead box P3; IL-7R, interleukin-7 receptor; IL-12Rβ2, interleukin-12 receptor β2; NSCLC, non–small-cell lung cancer.
Supported in part by Grants No. 1R21CA164568-01A1, 1R21CA164585-01A1, and U54CA137788/U54CA132378 from the National Cancer Institute; Grants No. PR101053 and W81XWH-11-LCRP-PCRA from the Department of Defense; the New York State Empire Clinical Research Investigator Program; the American Association for Thoracic Surgery Third Edward D. Churchill Research Scholarship; an International Association for the Study of Lung Cancer Young Investigator Award; a research grant from the National Lung Cancer Partnership/ LUNGevity Foundation; William H. and Alice Goodwin and the Commonwealth Foundation for Cancer Research and Experimental Therapeutics Center; the Stony Wold-Herbert Fund; and a grant from the Mesothelioma Applied Research Foundation in Memory of Lance S. Ruble.
Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.
The author(s) indicated no potential conflicts of interest.
Conception and design: Kei Suzuki, William D. Travis, Michel Sadelain, Prasad S. Adusumilli
Financial support: William D. Travis, Prasad S. Adusumilli
Administrative support: William D. Travis, Prasad S. Adusumilli
Provision of study materials or patients: Valerie W. Rusch, William D. Travis, Prasad S. Adusumilli
Collection and assembly of data: Kei Suzuki, Kyuichi Kadota, Jun-ichi Nitadori, Prasad S. Adusumilli
Data analysis and interpretation: Kei Suzuki, Camelia S. Sima, Jun-ichi Nitadori, Valerie W. Rusch, William D. Travis, Michel Sadelain, Prasad S. Adusumilli
Manuscript writing: All authors
Final approval of manuscript: All authors