Surfactant Protein A Gene Deletion and Prognostics for Patients with Stage I Non–Small Cell Lung Cancer

Although surgical resection is an effective treatment for stage I non–small cell lung cancer (NSCLC), curing ∼60% of patients, ∼40% of the “cured” patients will later die of recurrent disease (1). With the availability of more sophisticated and sensitive radiological imaging studies, more patients with lung cancers will be diagnosed while the disease is still at an early stage I. However, current clinical testing cannot predict whether a patient's cancer can be cured by surgical treatment alone or will require additional and more aggressive treatment to achieve long-term survival. It is therefore desirable that clinically applicable strategies be developed to identify, at the time of surgery, those patients at high risk for cancer recurrence who might benefit from systemic adjuvant chemotherapy (2).

Several immunocytochemical markers have been identified as prognostic factors in early-stage NSCLC, including K-ras/p21, p53, c-erbB-2/p185, bcl-2, Rb, cathepsin B, and Ki-67 protein (3–8). Some molecular markers, including retinoic acid receptor β, cyclooxygenase-2, vascular endothelial growth factor, matrix metalloproteinases-2 and -9, E-cadherin, angiopoietin-2, and CD44, have also been evaluated at the mRNA level (9–11). Additionally, recent methylation analyses of the promoter regions in the DAP, GSTP1, APC, p16, RASSF1A, and MGMT genes suggest that hypermethylation of any of these genes in stage I NSCLC is associated with biologically aggressive cancer (12, 13). However, the role of these molecular markers in clinical diagnosis and therapeutic decision-making remains speculative, and some drawbacks to the techniques used to determine the methylation status limit their clinical application. Thus, more reliable molecular markers for early-stage NSCLC need to be identified.

The pulmonary surfactant system is a phospholipid-protein complex that lowers the surface tension at the air-liquid interface in the alveoli of the lung (14). Surfactant is essential for normal lung function, particularly in infants and children, and a deficiency can lead to various diseases, such as respiratory distress syndrome in the premature infant. Because of the importance of surfactant in mediating local airway conditions and clearing the upper respiratory tract, there is growing interest in the role of the surfactant system in adult pulmonary disease. Surfactant protein A (SP-A) is the most abundant of the four surfactant-specific proteins (A, B, C, and D) known to be part of the pulmonary surfactant system and, therefore, is considered a good candidate as an etiologic factor in lung diseases (15), including, perhaps, lung cancer.

The human SP-A gene locus consists of two highly homologous functional genes, SP-A1 and SP-A2, each ∼4.5 kb long and separated from each other by 59 kb (15). The genes are located on 10q22, which is one of the regions most frequently associated with genomic imbalances in lung cancer (14). Alterations in SP-A in lung cancer have previously been analyzed by reverse transcriptase-PCR, immunoblot analysis, ELISA, and immunohistochemical analysis (16, 17). However, the findings have been inconsistent. Recently, in a high-resolution comparative genomic hybridization analysis of a cDNA microarray, we showed that deletion of the SP-A gene is one of the most common genomic copy changes in primary lung cancer (18). Therefore, we hypothesized that changes in the genomic copies of the SP-A gene might be involved in lung tumorigenesis and believed that it would be worthwhile to use different detection techniques to determine the precise prevalence of SP-A aberrations in human lung carcinomas.

At the genomic level, chromosomal losses have been studied extensively in lung tumors and normal-appearing bronchial epithelium from smokers, and it has been suggested that the genome-wide abnormalities are concentrated on certain chromosome arms (19, 20). For example, we previously showed that changes in chromosome arms 3p, 10q, 5p15, 6, 7p12, and 8q24 were common in lung tumors and that these abnormalities might be detected in bronchial epithelial cells obtained from patients with early-stage NSCLC by using fluorescence in situ hybridization (FISH; ref. 21). Our recent study which used comparative genomic hybridization analysis of cDNA microarrays not only confirmed the previous findings but also defined a narrower region and even identified the individual genes with genetic deletions associated with primary lung cancers, including SP-A (18). To determine whether SP-A aberrations are tumor-specific in early-stage lung cancer and whether such aberrations are associated with the aggressive biological behavior of NSCLC tumors, we tested the touch imprints of lung tumors obtained prospectively from 28 patients with primary NSCLC by performing FISH with SP-A and PTEN probes. We then analyzed archived primary tumor specimens from 130 patients with pathologic stage I NSCLC to determine their SP-A genomic copy number to further define the clinical relevance of SP-A aberrations. We showed that changes in SP-A are common and specific in bronchial epithelial cells adjacent to and within NSCLC tumors. Our findings therefore suggest that SP-A deletions are a new marker of lung cancer recurrence and survival and that determination of the SP-A copy number with FISH might be useful in identifying patients with poor prognoses who might benefit from adjuvant chemotherapy.

Tissue samples. To determine whether the changes in SP-A indicated smoke damage and were cancer-specific, we made touch imprints of resected lung tumors and normal lung tissues located far from the tumor that were obtained prospectively from 28 patients with primary NSCLC who underwent surgical excision, from October 2003 to December 2004, at The University of Texas M.D. Anderson Cancer Center. Table 1 summarizes patient and tumor characteristics.

To determine the prevalence of SP-A aberrations in early-stage lung cancer and whether these aberrations are associated with aggressive biological behavior of the associated tumors, we first examined a total of 505 tissue samples from consecutive patients with stage I NSCLC who underwent definitive surgical resection, defined as a lobectomy or a pneumonectomy, from 1986 to 1996, at our institution to identify those that contained an adequate number of tumor cells and cells from the adjacent normal bronchial epithelium. One hundred and thirty samples which met these criteria were identified. Follow-up information was obtained from chart reviews and reports from the institution's tumor registry service. Table 2 summarizes the patient and tumor characteristics associated with the 130 patients from whom the samples came. Among the 130 patients, 39 had relapses during the study period and known site of disease recurrence. Lung tissue from 63 patients who were treated at the institution during the same period, but who did not have lung cancer, were obtained and used as controls.

No patients received adjuvant chemotherapy or radiation therapy before or after surgery. Tissue sections (4 mm thick) were obtained from each tissue block, stained with H&E, and re-reviewed independently by two pathologists (R.L. Katz and N.P. Caraway) to confirm the diagnosis and the presence of both tumor and adjacent bronchial epithelium. The histologic type of the tumors was graded according to the WHO classification. Areas of interest were marked for future microdissection. All study designs were reviewed and approved by the institution's surveillance committee (Institutional Review Board).

Specific probes. A BAC clone of ∼110 kb containing an SP-A genomic sequence was selected from a BAC-PAC contig (kindly provided by Dr. Peter A. Steck, Department of Neuro-Oncology, M.D. Anderson Cancer Center) in chromosome 10q22. To confirm that genomic sequences of both SP-A1 and SP-A2 were located in the clone and the size of the clone, we first examined it with a PCR-based screening protocol using specific primers for SP-A1 (ACACCAACTGGTACCGACC, TGGAGCCTCAGGATGGAGG) and SP-A2 (CTGCTGATGAACAAATCTGCA, GTGGCACATGGTATGTGCTC); we then sequenced both ends of the clone and placed the sequences in BlastN in the Celera database (http://myscience.appliedbiosystems.com). A BAC clone of ∼120 kb containing a PTEN genomic sequence was identified by using the Celera database, which located it in chromosome 10q23.31. The chromosomal location of the clone was confirmed by control metaphases and interphases prepared from normal human bronchial epithelial cells (Clontech, Palo Alto, CA) in combination with a centromeric probe for chromosome 10 (CEP10; Vysis, Downers Grove, IL).

Specimen preparation and FISH. From each patient with primary lung cancer, six touch imprints from resected lung tumor and the paired normal lung tissue were made. Four were fixed for FISH in methanol and acetic acid at a ratio of 3:1, and one was stained with Papanicolaou stain and one was stained with Diff-Quik (Baxter Scientific, Deerfield, IL), respectively. Before its analysis by FISH, each touch imprint was evaluated for normal and malignant cells by comparing its staining by Papanicolaou stain and Diff-Quik with the histologic diagnosis made on the basis of H&E-stained sections obtained from the same tissue mass. Tricolor FISH was done using the SP-A probe, which was directly fluorescence-labeled in green; the PTEN probe, which was directly fluorescence-labeled in yellow; and the CEP10, which was directly fluorescence-labeled in red and used as an internal control probe. One hundred nanograms of each probe was mixed with a 30-fold excess of human Cot-1 DNA (Life Technologies, Rockville, MD) in 10 μL of LSI hybridization buffer (Vysis) and mounted on a slide. Hybridization was done by incubating the slides at 37°C overnight. Nuclei were counterstained with 6-diamidino-2-phenylindole (Vysis).

To isolate whole nuclei of tumor cells and adjacent bronchial tissue cells from the formalin-fixed, paraffin-embedded tissue blocks, representative tumor and adjacent bronchial tissues were first located microscopically and then marked on H&E-stained sections. Corresponding areas were then identified on the paraffin blocks, and a 22-gauge needle (Biopunch, Buffalo, NY) was used to obtain core tissue samples, 1.0 mm in diameter, from each block. To avoid possible contamination of tumor tissues by nonneoplastic cells and of normal tissues by tumor cells, we took deep core tissue samples at a range of 10 to 50 μm deep, depending on the thickness of the block, because such deep tissue samples, dissected by a 1- to 1.5-mm needle core, allow sampling from the exact areas of interest and yield an isolate with sufficient numbers of intact nuclei that is free of contaminating cells (22–28). The paraffin-embedded tissue samples were deparaffinized by immersion in xylene for 1 hour, followed by immersion in various dilutions of ethanol (100%, 90%, and 70%) for 10 minutes each. The specimens were then incubated with 0.1% protease (Sigma P8038; Sigma Chemical Co., St. Louis, MO) at 37°C for 10 minutes. Cytospin preparations were generated on glass slides using a cytocentrifuge (Cytospin 2; Shandon, Pittsburgh, PA) and then fixed with methanol and acetic acid at a ratio of 3:1. Papanicolaou and Diff-Quik preparations were also made from each cytospin preparation and then evaluated for normal and malignant cells and correlated with the histologic diagnosis of the original marked H&E-stained sections. For the slides prepared from adjacent normal-appearing bronchial tissue, only those uncontaminated by tumor cells were processed for further FISH analysis. Dual-color FISH was done using a digoxigenin-labeled SP-A probe as described by our previous publication (29), and a CEP10 was used as an internal reference.

Scoring FISH signals. Slides were examined under microscopes equipped with appropriate filter sets (Leica Microsystems, Inc., Buffalo, NY). Two hundred cells were counted on each slide. To ensure random analysis of the cells, at least two areas were examined on each slide. Overlapping cells and cells that had indistinct or blurry signals were not scored, and care was taken not to interpret a split signal as two signals. More or fewer signals from the SP-A probe or the PTEN probe than from the CEP10 indicated SP-A gene or PTEN gene gain or loss, respectively. FISH-signaling patterns of SP-A and PTEN were defined as abnormal if there were more or fewer copies of the SP-A or PTEN signals than of the CEP10 signals, beyond a certain cutoff (baseline value). The cutoff value was calculated from normal tissue samples and was defined as the mean number of cells having an abnormal SP-A or PTEN signal pattern on FISH analysis, ± 3 SD.

Statistical analysis. All statistical analyses were done using SAS software (version 6.12; SAS Institute, Inc., Cary, NC). Overall, disease-specific and disease-free survival rates were calculated using the Kaplan-Meier method. All survival times were calculated from the date of surgery. Patients were stratified into two groups based on the time to relapse from the date of surgery. An early relapse was defined as a relapse that occurred within 24 months of the date of surgery, and a late relapse was defined as a relapse that occurred >24 months after the date of surgery. Disease-specific survival time was calculated from the date of surgery to the date of death from a lung cancer-related cause. The χ² test was used to evaluate associations between categorical variables, and the Mann-Whitney test was used to assess differences in continuous variables by a dichotomous relapse status. The univariate Cox proportional hazards model was used to evaluate the unadjusted association between survival times and various clinically and histopathologically interesting risk factors [i.e., age, sex, race, smoking history, alcohol consumption, histologic subtype, and tumor-node-metastasis (TNM) stage]. Multivariate Cox models were used to model the increased risk associated with increases in SP-A deletions on survival time, after adjusting for the same clinical and histopathologic variables mentioned previously. Tukey's honestly significant difference test was also used to analyze the relationship between relapses and risk factors. All P values were determined by two-sided tests. P values <0.05 were considered statistically significant.

Two green signals (corresponding to the SP-A probe), two yellow signals (corresponding to the PTEN probe), and two red signals (corresponding to the CEP10) were detected in the control metaphases and interphases from the normal cells (Fig. 1A).

Tricolor FISH with the SP-A probe, the PTEN probe, and the CEP10 was used to determine whether the changes in SP-A and PTEN were specific and indicated smoke damage in the touch preparation from resected lung tumors and normal lung tissues. In normal tissues, the percentage of cells that showed a loss of SP-A signals was between 0% and 3%, and the percentage of cells that showed a loss of PTEN signals was between 0% and 2%. Therefore, a tumor specimen was considered to have a SP-A deletion when >6% of cells showed a deletion of SP-A and to have a PTEN deletion when >4% of cells showed a deletion of PTEN. Of the 28 tumors analyzed, SP-A deletions occurred in 89% of the tumor samples. Deletions of SP-A occurred more frequently in the tumors of patients who smoked than in tumors of patients who never smoked (P < 0.001). PTEN was deleted from 16% in the cancer tissues and the deletion was not related to the smoking status of patients (P > 0.05) (Fig. 1B).

Dual-color FISH with the SP-A probe and the CEP10 was also done on all the archived, formalin-fixed, paraffin-embedded tissue blocks. In normal control cells, the percentage of nuclei found to have a loss of SP-A signals was between 0% and 2%. Therefore, a tissue-block specimen was considered to have a SP-A deletion when >4% of cells had a deletion of SP-A. In each sample of 200 nuclei, the mean rates of deletion were 12% for adjacent normal-appearing bronchial tissues and 38% for tumors. In addition, 32% of adjacent normal-appearing bronchial tissues and 87% of tumors analyzed, had >4% of cells with SP-A deletions (Fig. 2). There was a statistically significant correlation between deletions of SP-A in tumor cells and deletions in adjacent normal-appearing bronchial epithelial cells (P < 0.001).

We tested for correlations between the SP-A copy numbers in tumors or adjacent bronchial tissue and various clinicopathologic features. There were no statistically significant correlations between the frequency of SP-A deletions and patient age, patient sex, tumor histologic type, or tumor TNM stage (e.g., T₁N₀M₀ versus T₂N₀M₀). However, deletion of SP-A occurred more frequently in the tumors and normal-appearing bronchial epithelium of patients who smoked than in tumors of patients who never smoked (P < 0.001 for both).

When we used the univariate Cox proportional hazards model to assess the effects of these variables on time to relapse, we found that only SP-A deletions in the tumors and adjacent normal-appearing bronchial tissues significantly increased the risk of relapse (P = 0.035 and P < 0.0001, respectively; Table 3). This result was confirmed by the honestly significant difference mean-rank test, which showed that SP-A deletions in both tumors and their adjacent bronchial tissues were strongly correlated with relapse (P < 0.001 for both).

To determine whether SP-A deletions could predict the survival times of patients with pathologic stage I NSCLC, we first did a univariate analysis using the Cox model. We found that only SP-A deletions in adjacent normal-appearing bronchial tissues, compared with the absence of such deletions, were predictive of a shorter survival time (P = 0.026; Table 4). Interestingly, patient age may be another predictive factor for survival because older patients had a marginally shorter overall survival time than did younger patients (P = 0.037). However, when stepwise selection was used to choose the appropriate multivariate model, a one unit (1%) increase in SP-A deletions in adjacent bronchial epithelial cells increased the risk of dying by 19%, whereas a one unit (1 year) increase in age increased the risk of dying by only 4%. There was also some evidence that having a high SP-A deletion rate in adjacent bronchial epithelial cells resulted in a shorter survival time, as shown by the Kaplan-Meier curve analysis of the probability of lung cancer–specific survival for patients with SP-A deletions versus patients without SP-A deletions (Fig. 3).

We also examined the SP-A deletion level from the standpoint of the site of disease recurrence in a subset of patients (n = 39) for whom this information was known. The average SP-A deletion level was highest in specimens from patients who subsequently developed contralateral or distant metastases, including six brain metastases. In contrast, the SP-A deletion level was lowest in specimens from patients whose metastases were restricted to the ipsilateral lung. The differences in deletion levels of SP-A were significant (P = 0.023). Collectively, the data suggested that the level of SP-A deletions in cells from the bronchial epithelial tissue adjacent to the tumor is the most significant prognostic indicator for disease-specific survival.

Our current study clearly showed that SP-A deletion is a statistically significant predictor of poor outcome in patients with early-stage NSCLC. In addition, we found that SP-A deletions occur with equal frequency among all histological types of NSCLC, suggesting that changes in SP-A are not histologic type–specific, and that detection of the abnormalities may be used to predict prognosis in all histological types of NSCLC.

Furthermore, one of the more intriguing outcomes of our study is the demonstration of differential SP-A deletions in the ipsilateral lung versus the contralateral lung or distant metastases: a higher level of SP-A deletions occurred in patients whose disease progressed to contralateral or distant metastases, and a lower level in patients whose metastases were restricted to the ipsilateral lung. These findings have an interesting implication, namely, that deletions of SP-A in tumor specimens of primary lung cancer can help identify patients likely to develop metastases outside the ipsilateral lung and also predict the recurrence of disease. Therefore, our findings imply that the SP-A gene copy number could be an ideal marker for determining the prognosis of patients with early-stage NSCLC and evaluating the effects of chemotherapeutic agents.

SP-A alterations have previously been studied in human lung cancers. O'Reilly et al. (16) first observed synthesis of SP-A in a cell line derived from a lung adenocarcinoma by using immunoblot analysis and ELISA. Linnoila et al. (30) found a high level of SP-A expression in adenocarcinomas by immunohistochemical analysis, particularly in those found to have a papillolepidic growth pattern. These results were confirmed by others using reverse transcriptase-PCR and RNA in situ hybridization techniques (31, 32). Furthermore, Shijubo et al. (33, 34) found that 27 of 67 patients with lung adenocarcinomas had high levels of SP-A in their pleural effusions, whereas patients with adenocarcinomas originating from different primary sites, other histologic types of lung cancers, and tuberculosis had low levels, suggesting that detection of SP-A in malignant effusions might help distinguish primary lung adenocarcinoma from other adenocarcinomas of miscellaneous origins. However, Fujita et al. (35), using both immunohistochemical analysis and reverse transcriptase-PCR, detected SP-A expression in only 1 of 16 lung cancer cell lines, 6 of which were adenocarcinomas. Furthermore, Zamecnik and Kodet (36) observed significantly less SP-A immunoreactivity in poorly differentiated than in well-differentiated lung adenocarcinomas. Moreover, Tsutsumida et al. (17) recently reported that a high mucin 1 to SP-A ratio could predict the recurrence rate and disease outcome in patients with pulmonary adenocarcinomas.

We believe that these inconsistent findings might result from the different techniques used to detect SP-A aberrations or from the small sample sizes and mixed stages of the tumors studied. In addition, none of these studies evaluated the prognostic value of SP-A in stage I NSCLC. By using FISH to measure the SP-A genomic copy numbers in a large population of patients with stage I NSCLC for whom complete follow-up information was available, we were able to address these issues.

To determine whether the changes in SP-A are specific indicators of smoke damage and thus of lung cancer, we did tricolor FISH directly comparing genetic changes of SP-A with those of PTEN. SP-A gene deletion occurred solely in the tumors and not in the normal tissues located far from the tumors and was seen more frequently in the tumor specimens of patients who smoked than in patients who never smoked. PTEN deletion, however, were found far less frequently and did not show association with smoking history. PTEN is a tumor-suppressor gene and is inactivated in several types of human tumors. Studies from others suggest that loss of PTEN expression is not uncommon in NSCLC (37, 38), however, the prognostic value of its low protein expression in NSCLC have been inconsistent (39, 40). The results of PTEN aberration from the current study is consistent with previous reports from others, which suggested that genetic alterations of the PTEN are rare in NSCLC detected by direct DNA sequencing and Southern blot analysis, and lack of PTEN expression may be partially explained by promoter methylation (41, 42). Therefore, although PTEN locates on 10q23, one of the sites with the most concentrated genomic losses in lung tumors, PTEN is not likely a target of the genomic deletion. Taken together, our findings imply that the changes in SP-A are specific to lung tumors and might be used as an indicator of smoke damage in the airways of smokers.

The main value of prognostic markers is to identify patients at high risk for cancer-related events, such as recurrence or survival, and to determine the need for and effects of adjuvant therapy. Although many genomic aberrations have been described in smoking-related lung cancers (19), none have been shown to predict the prognosis of patients with primary lung cancer. For example, we suggested in our previous study of the same set of archived, formalin-fixed, paraffin-embedded tissue blocks that deletions of 3p21.3 were also an early event in smoking-related lung tumorigenesis but were not associated with an increased risk of disease relapse or shorter survival (43).

Most known prognostic biomarkers are present within the lung tumor itself, but the methods used to detect them are invasive. Bronchial brushings and sputum specimens can provide bronchial epithelial cells from distinct areas of the airway, which can be examined for various morphological features and biomarkers. In addition, collection of these cells is less expensive and invasive than surgical removal of tissues and, therefore, could be routinely done. If genetic aberrations in the bronchial epithelium reflect lung cancer development and can predict patient prognosis, then testing for these genetic changes in bronchial brushings or even sputum specimens containing exfoliated bronchial epithelial cells can be used to monitor the clinical outcome of patients.

Our current study showed that SP-A deletions in the normal-appearing bronchial tissue adjacent to tumor were closely related to SP-A deletions in the tumor cells and, most importantly, were strongly correlated with patient survival and time to disease relapse. These findings might have clinical importance because the SP-A probe could be useful as a surrogate biomarker of recurrent disease and because FISH analysis might be a reliable and easily done method of testing for SP-A deletions in cells obtained from easily accessible tissues (e.g., those obtained from bronchial brushings or sputum specimens), with less harm caused to patients. For example, patients with persistently high levels of SP-A deletion in cells from their bronchial brushings or sputum specimens might benefit from adjuvant chemotherapy.

In conclusion, a higher number of SP-A deletions are associated with malignant tumor progression and smoking, thereby making the number of SP-A deletions a potentially valuable prognostic marker. For early-stage NSCLC, the SP-A copy number shown by FISH might be useful in distinguishing patients with poor prognoses from those with better prognoses and in deciding on the appropriateness and effectiveness of adjuvant chemotherapy for patients.

In an ongoing study, we are collecting data from sputum, oral brushing, and lung tumor tissue specimens, obtained prospectively from patients with newly diagnosed lung cancer, using FISH with an SP-A probe and comparing these data with clinical covariates to determine the sensitivity and specificity of the assay and to establish diagnostic criteria for the oral brushing and sputum specimens. We are also planning to perform an independent, double-blinded, prospective, randomized clinical trial to validate our current findings.

We thank Ms. Kathryn B. Carnes and Dr. Ellen M. McDonald of the Department of Scientific Publications for editorial review of this manuscript.