Volume 107, Issue 12 p. 2866-2872
Original Article
Free Access

Tumor infiltrating Foxp3+ regulatory T-cells are associated with recurrence in pathologic stage I NSCLC patients

Rebecca P. Petersen MD, MSc

Rebecca P. Petersen MD, MSc

Department of Surgery, Duke University Medical Center, Durham, North Carolina

Search for more papers by this author
Michael J. Campa PhD

Michael J. Campa PhD

Department of Radiology, Duke University Medical Center, Durham, North Carolina

Search for more papers by this author
Justin Sperlazza

Justin Sperlazza

Department of Surgery, Duke University Medical Center, Durham, North Carolina

Search for more papers by this author
Debbi Conlon

Debbi Conlon

Department of Surgery, Duke University Medical Center, Durham, North Carolina

Search for more papers by this author
Mary-Beth Joshi

Mary-Beth Joshi

Department of Surgery, Duke University Medical Center, Durham, North Carolina

Search for more papers by this author
David H. Harpole Jr MD

David H. Harpole Jr MD

Department of Surgery, Duke University Medical Center, Durham, North Carolina

Search for more papers by this author
Edward F. Patz Jr MD

Corresponding Author

Edward F. Patz Jr MD

Department of Radiology, Duke University Medical Center, Durham, North Carolina

Fax: (919) 684-7123

Duke University Medical Center, Department of Radiology, Box 3808, Durham, NC 27710===Search for more papers by this author
First published: 08 December 2006
Citations: 331

Abstract

BACKGROUND.

Early stage lung cancer has a variable prognosis, and there are currently no markers that predict which patients will recur. This study examined the relation between tumor-regulatory T (Treg) cells and total tumor-infiltrating T-cell lymphocytes (TIL) to determine whether they correlated with recurrence.

METHODS.

The authors reviewed all patients in our tissue databank from 1996 to 2001 and identified 64 consecutive pathologic stage I non-small cell lung cancer (NSCLC) patients who had surgical resection and at least a 2.5 years disease-free follow-up or documented recurrence within 2 years. Immunohistochemical analyses were performed on paraffin-embedded lung cancer tissue and the relation between Treg cells, TIL, and disease-specific survival was determined. A risk index was devised deductively for various possible combinations of Treg cells and TIL.

RESULTS.

Treg cells and TIL were detected in 33 of 64 (51%) and 53 of 64 (83%) patients, respectively. When data were analyzed by using a Treg/TIL Combination Risk Index, patients with high-risk and intermediate-risk indices had hazard ratios of 8.2 (P = .007) and 3.3 (P = .109), respectively.

CONCLUSIONS.

Patients with stage I NSCLC who have a higher proportion of tumor Treg cells relative to TIL had a significantly higher risk of recurrence. These data may be useful, particularly if combined with a panel of tumor markers, to suggest at the time of diagnosis which patients with seemingly early-stage NSCLC will relapse. Cancer 2006. © 2006 American Cancer Society.

Lung cancer is the leading cause of cancer death in the world, and more patients die from lung cancer than from cancers of the breast, colon, prostate, and pancreas combined. Approximately 173,000 individuals in the United States were diagnosed with non-small cell lung cancer (NSCLC) in 2004, and the overall 5-year survival rate is a dismal 15%.1 Even when diagnosed at an early stage, patients relapse at a rate as high as 50% after surgical resection.2-4 Despite their identical radiological and histological features, many patients with presumed localized disease have undetectable metastases at the time of diagnosis; and current clinical-pathological staging is inadequate.

Whereas several studies have examined molecular and biochemical pathways of metastasis, there are currently no markers used in clinical practice that predict which patients will have recurrences. Consequently, all patients with early-stage NSCLC are treated the same, despite the existence of several distinct phenotypes. The clinical behavior of any given tumor is unpredictable, but it most likely depends on a complex relation between the tumor genotype and the host response to disease. This is illustrated particularly in patients who simultaneously develop metastases at multiple sites years after primary resection, possibly suggesting a generalized decline in a systemic host response, whose key role is to maintain surveillance of microscopic metastases.5, 6 Disruption in this tumor-host balance may permit metastatic deposits to grow and become clinically apparent.

The role of the immune system in destroying cancer cells was first proposed over 50 years ago,7, 8 and there is increasing evidence to implicate an immune response in maintaining disease stability, particularly in early-stage disease where the tumor burden is considered minimal. More recently, several investigators have studied an opposing aspect of the immune system, and the role of T regulatory (Treg) cells.5, 6 In principle, Treg cells may down-regulate the immune response by attenuating the host's antitumor T cells, potentially permitting unrestricted growth and subsequent metastasis. Thus, increased Treg cells within the primary tumor may predict a poor prognosis. These seemingly incongruous activities of the immune system underscore the complexity of the host immune response, and the current study specifically focused on the relation between Treg cells, total tumor CD3+ tumor-infiltrating T-cell lymphocytes (TIL) and overall survival in patients with resectable stage I NSCLC.

MATERIALS AND METHODS

Study Population

This study was approved by our institutional review board, and informed consent was obtained from all patients. From our prospectively maintained tissue databank, we selected 64 consecutive patients with pathologic stage I NSCLC (T1N0M0 or T2N0M0) in accordance with the International System of Staging for Lung Cancer.7 All patients underwent a lobectomy at our institution between January 1996 and December 2001, and all patients had at least 2.5 years of follow-up or documented recurrence within 2 years after initial diagnosis. We identified 34 stage I NSCLC patients who remained disease free at least 2.5 years after resection and 30 stage I patients from the same time period who developed recurrence within 2 years after surgery. Baseline demographics, histopathologic data, overall survival, cancer-specific survival, and pathologic specimens preserved in paraffin were available for all patients. Patients who received neoadjuvant or any type of postoperative therapy, or had another malignancy were excluded.

Immunohistochemistry

Immunohistochemical analyses were performed on resected, paraffin-embedded lung cancer tissue. After microtome sectioning (4 to 6 μm), slide labeling, and deparaffinization with xylene and ethanol, antigen retrieval was accomplished by microwaving and washing with phosphate-buffered saline. The sections were incubated with primary monoclonal antibodies to the Foxp3 protein (ab22510, Abcam Inc, Cambridge, MA), which is a transcription factor protein and a marker of Treg activation,10 and the T-cell surface glycoprotein CD3 (AZ452, Dako Carpinteria, CA), a comprehensive total T-cell marker. Incubation with a horseradish peroxidase conjugated secondary antibody was subsequently performed, followed by development with diaminobenzidine and counterstaining with hematoxylin. An experienced pathology technician reviewed and scored 10 separate fields for each slide, and an average immunohistochemical score was recorded for each patient. The technician was blinded to patient outcome. Immunohistochemical staining was graded on a scale of 0 to 3 for both Foxp3 and CD3. The histochemistry scale is defined in Table 1. In addition, a risk index was devised deductively for the various possible combinations of Foxp3 and CD3 for each patient and is shown in Table 2. The risk index was stratified into 3 categories based on the theoretical consideration that patients with no T cells, or a high proportion of Tregs relative to total T cells, were at highest risk of recurrence, whereas patients with abundant T cells and no Tregs were at the lowest risk of recurrence. All other combinations were considered to indicate intermediate risk of relapse.

Table 1. Immunohistochemisty Scale for Foxp3 and CD3
Immunohistochemistry scoring system scale
0 No staining
1 Positive staining in <20% of cells
2 Positive staining in 21–50% of cells
3 Positive staining in >50% of cells
Table 2. Foxp3+/CD3+ Combination Risk Index According to IHC Score
Risk index IHC score
CD3, Foxp3
Low risk 3,0; 2,0
Intermediate risk 1,0; 2,1; 3,1; 3,2
High risk 0,0; 1,1; 2,2; 3,3

Statistical Analysis

All marker analyses were blinded to patient outcome. Univariate and multiple regression analyses using a Cox proportional hazards model were carried out to examine the relation between potential prognostic factors and cancer-specific survival, including the individual immunohistochemistry scores for Foxp3 and CD3. More importantly, we report a Foxp3/CD3 Combination Risk Index as the relation between Treg cells and total T cells that is more informative in understanding the effect of the host immune response than either cell marker independently (Table 2). As above, tumors with low Treg cells and low TIL are predicted to do worse than those with low Treg cells and abundant TIL. This index was evaluated in relation to cancer-specific survival as described in Table 2.

Univariate predictors with a P-value of ≤.1 were examined in a multiple Cox proportional hazards model. Cancer-specific survival was defined as the time between surgery and the last follow-up date or date of recurrence and/or cancer death. If a patient died without cancer recurrence, the patient's survival time was censored at the time of death. In addition, Kaplan-Meier survival analysis was performed, by stratifying significant predictor variables identified in the Cox proportional hazards model.

RESULTS

Study Population

Baseline demographics, including age, sex, tobacco use, histopathologic data, and pathologic stage are reported in Table 3.

Table 3. Baseline Demographics, N = 64
Characteristic n (%)
Age
 Mean ± SD 67 ± 9
Sex
 Female 30 (47)
Tobacco use
 Pack-y 51± 33
Tumor size, cm
 Mean ± SD 3.3 ± 1.8
Pathologic stage
 IA 32 (50)
 IB 32 (50)
Cell type
 Adenocarcinoma 30 (47)
 Squamous cell carcinoma 22 (34)
 Other 12 (19)
  • SD indicates standard deviation.

Foxp3+ Treg Cells

Intratumoral infiltration of Foxp3+ Treg cells with an immunohistochemistry score ≥1 were detected in 33 of 64 (51%) patients. Univariate analysis, dichotomizing the Foxp3 immunohistochemistry score at the median (≥1), did not reveal intratumoral infiltration of Foxp3+ Treg cells to be associated with cancer recurrence. However, when analyzed in the context of overall TIL, the Foxp3/CD3 Combination Risk Index was found to be associated with cancer-specific survival (as below) (Fig. 1).

CD3+ T Cells

Intratumoral infiltration of CD3+ T-cells with an immunohistochemistry score ≥1 was detected in 53 of 64 (83%) patients. Univariate analysis, dichotomizing the CD3 immunohistochemistry score at the median (≥2), revealed intratumoral infiltration of CD3+ T cells to be protective against cancer recurrence, with a hazard ratio (HR) of 0.39, P = .005, and it remained significant after statistically controlling for tumor size (CD3+ T cells: HR 0.21, P = .005; tumor size in centimeters: HR 1.3, P = .026) in the multivariate Cox proportion hazards model. Kaplan-Meier cancer-specific survival analysis stratifying on tumors with a CD3+ score of <2 revealed patients to have a median survival of 49 months compared with >72 months (not met) for patients with a CD3+ score of ≥2, log-rank P = .004 (Fig. 2).

Details are in the caption following the image

Kaplan-Meier survival-free of cancer recurrence analysis by CD3 and Foxp3 Combination Risk Index, log-rank P = .0148.

Details are in the caption following the image

Kaplan-Meier survival-free of cancer recurrence analysis by CD3 Immunohistochemistry score ≥2, log-rank P = .0004.

Details are in the caption following the image

This figure illustrates a sixty-six-year-old man with pathologic stage IB non-small cell lung cancer who is a long-term survivor without evidence of recurrence. Coronal (left panel) and axial (right panel) positron emission tomography images with flurodeoxyglucose demonstrate a hypermetablolic left hilar mass (arrow) consistent with malignancy (Panel A). Note is also made of normal left ventricular myocardial activity (green curved arrow). Below are immunohistochemistry slides illustrating a low-risk CD3/Foxp3 Combination Index (Panel B). CD3+ immunohistochemistry score = FOXP3+ immunohistochemistry score = 0. Figures 3 and 4 illustrate two patients with early-stage disease and identical radiological findings (see PET images) and histology. Although their PET images were similar, the CD3/Foxp3 Combination Index suggested the patient in Figure 3 would have a better prognosis.

Details are in the caption following the image

This figure depicts a sixty-four year old man with pathologic stage IA non-small cell lung cancer who died of recurrent disease. Coronal (left panel) and axial (right panel) positron emission tomography images demonstrate a hypermetablolic left lower lobe mass (arrow) consistent with malignancy (Panel A). Below are immunohistochemistry slides illustrating a high-risk CD3/Foxp3 Combination Index (Panel B). Figures 3 and 4 illustrate two patients with early-stage disease and identical radiological findings (see PET images) and histology. Although their PET images were similar, the CD3/Foxp3 Combination Index suggested the patient in Figure 4 would have a worse prognosis.

Foxp3/CD3 Combination Risk Index

Eleven (17%) patients were classified as low risk, 34 (53%) as intermediate risk, and 19 (30%) as high risk on a Foxp3/CD3 Combination Risk Index. Univariate analysis revealed the Foxp3/CD3 Combination Risk Index to be associated with cancer-specific survival with hazard ratios (HRs) for the high- and intermediate-risk indices of 6.3 (P = .017) and 3.0 (P = .15), respectively. The Foxp3/CD3 Combination Risk Index remained significant after statistically controlling for tumor size (intermediate-risk index: HR 3.3, P = .109; high-risk index: HR 8.2, P = .007; tumor size in centimeters: HR 1.2, P = .05) in the multivariate Cox proportion hazards model (Table 4). Furthermore, Kaplan-Meier cancer-specific survival analysis stratified by the 3 risk-index groups revealed significantly different curves (log-rank, P = .0148), with median survival times of 53 months, 63 months, and >72 months (not met) for the high-, intermediate-, and low-risk index patients, respectively (Figs. 3, 4).

Table 4. Cox Proportional Hazards Model, N = 64
Predictor HR (95% CI) P n
Risk Index
 Intermediate 3.3 (0.8–14.1) 0.109 34
 High 8.2 (1.8–38.4) 0.007 19
 Tumor size, cm 1.2 (1.01–1.5) 0.05 64
  • HR indicates hazard ratio; CI, confidence interval; cm, centimeter.

DISCUSSION

Lung cancer continues to be a major healthcare problem worldwide, and most patients present with advanced-stage disease. Screening for lung cancer with chest radiographs or sputum cytology has failed to demonstrate a reduction in disease-specific mortality, and although CT screening studies find smaller tumors, these smaller tumors do not necessarily indicate early-stage disease. This is clearly evidenced by the high rate of recurrence in resected stage I patients. Many of these patients will have tumors with identical radiographic and pathologic features, but these findings are inadequate in predicting outcomes.

To optimize treatment for patients with early-stage disease, numerous studies have attempted to use a combination of clinical, pathological, and molecular properties of tumors, but there is currently no way to determine which patients will recur and when recurrence will become clinically apparent. In fact, it appears that patients with stage I NSCLC have 3 diverse phenotypes. First, there are patients who never recur after resection. They have indolent tumors, possibly supported by a host response that maintains tight surveillance of any microscopic metastatic deposits. There are also those individuals with this indolent phenotype who die from other causes, and have undiagnosed, clinically irrelevant lung cancer discovered at autopsy.11-13

On the opposite end of the spectrum, there are patients with clinical-pathological stage I NSCLC, who undergo curative surgery, but die with diffuse metastatic disease shortly after resection. This phenotype is typified by the Lewis lung cancer model, where the metastatic deposits grow only after the primary tumor is removed.14 Finally, there are patients with early-stage tumors who present with metastases years after resection. It is in these individuals that a host response suppressing metastatic tumor growth is most suggestive; multiple sites are often simultaneously detected, and tumor growth rates are much more rapid at the time of recurrence than would be predicted if metastases had been constantly growing from the time of resection.

The ability to predict which pathologic stage I patients are at highest risk of recurrence could facilitate patient selection for adjuvant therapy while limiting the exposure of low-risk patients to unnecessary toxicity. In this study, we focused on 1 immune response mechanism to malignant disease and the relation between Treg cells and TIL in predicting tumor recurrence in early-stage disease. We were particularly interested in patients with stage I NSCLC, as it has been suggested that the immune system is probably most effective at controlling malignancies when there is minimal disease.

Previous reports, and data from this study, have shown total TIL to be associated with improved prognosis in several types of cancers;15-23 however, it is clear not all TILs have an antitumor effect. As suggested by the current study, additional classification of TIL, by identifying the Treg cell component, may improve phenotypic stratification. Taams and colleagues have shown that a small subset of CD4+ cells that uniquely express the transmembrane protein CD25+, exist in various proportions of the total number in NSCLC patients.5 Although some of these may be Treg cells, activated CD4+ T cells can have a similar phenotype (ie, CD4+CD25+). These 2 different types of T cells theoretically have opposing effects on the tumor, therefore this study evaluated expression of Foxp3+, a Treg activation marker, as Treg cells normally function to protect the host against development of autoimmunity.24 However, in patients with a malignancy, Treg cells have been associated with tumor progression and suppression of an antitumor immune response possibly through secretion of TGF-β or direct contact inhibition,5, 6 although the precise complex mechanism remains to be defined.25, 26

The stimulus for an immune reaction, particularly for Treg cells, to any given tumor remains to be characterized. In a recent report, it was shown that COX-2/PGE2 expression by NSCLC induces the Treg cell-specific transcription factor, Foxp3, a member of the forkhead family, which subsequently increases Treg activity.5, 25-28 The activation of the Treg cells results in production of interleukin 10 and TGF-β, which may in turn inhibit reactive T cells. COX-2 is often overexpressed in NSCLC, and a correlative clinical study demonstrated an association between COX-2 expression and a poor outcome in patients with early-stage NSCLC.29, 30 Whereas regulation of COX-2 in tumors is unclear, several different mechanisms, including hypoxia and HIF-1, the WNT pathway, and iNOS have been proposed.31-33

Although this is a retrospective study with potential limitations in selection bias, the current data demonstrate that Treg cells are present as a varying proportion of total TIL in early-stage NSCLC tumors and that a higher ratio of Treg cells to TIL is associated with development of metastasis and a reduction in cancer-specific survival. In addition, we confirm that pathologic stage I NSCLC patients who have undergone resection and who have no or few TILs are at higher risk of recurrence. Although several other studies in NSCLC have shown TIL to be associated with improved outcomes, this is the first study to investigate the correlation between Foxp3+ Treg cells, TIL, and lung cancer-specific survival. It is not surprising that when the presence of tumor infiltrating Treg cells is analyzed independently of total TIL that there is no association with prognosis. Because Treg cells exert their effects through inhibition of antitumor T cells, the presence of tumor Treg cells correlates with survival only when considered in the context of total TIL. Thus, we describe a Combined Risk Index that provides an integrated approach to these 2 essential components.

The immune response is but 1 element of complex mechanisms that dictate tumor behavior. A comprehensive integration of other components, including the tumor genotype and protein expression, will undoubtedly lead to more accurate characterization of NSCLC and improve stratification of patients for optimal outcomes.

Acknowledgements

We thank Liz Gottlin, PhD for her editorial contributions