TGFβ1 Polymorphisms Predict Distant Metastasis–Free Survival in Patients with Inoperable Non-Small-Cell Lung Cancer after Definitive Radiotherapy

Purpose Transforming growth factor (TGF) -β1 signaling is involved in cancer-cell metastasis. We investigated whether single nucleotide polymorphisms (SNPs) at TGFβ1 were associated with overall survival (OS) and distant metastasis-free survival (DMFS) in patients with non-small cell lung cancer (NSCLC) treated with definitive radiotherapy, with or without chemotherapy. Methods We genotyped TGFβ1 SNPs at rs1800469 (C–509T), rs1800471 (G915C), and rs1982073 (T+29C) by polymerase chain reaction-restriction fragment length polymorphism in blood samples from 205 NSCLC patients who had had definitive radiotherapy at one institution in November 1998–January 2005. We also tested whether the TGF-β1 rs1982073 (T+29C) SNP affected the migration and invasion of A549 and PC9 lung cancer cells. Results Median follow-up time for all patients was 17 months (range, 1–97 months; 39 months for patients alive at the time of analysis). Multivariate analysis showed that the TGFβ1 rs1800469 CT/CC genotype was associated with poor OS (hazard ratio [HR] = 1.463 [95% confidence interval {CI} = 1.012–2.114], P = 0.043) and shorter DMFS (HR = 1.601 [95% CI = 1.042–2.459], P = 0.032) and that the TGFβ1 rs1982073 CT/CC genotype predicted poor DMFS (HR = 1.589 [95% CI = 1.009–2.502], P = 0.046) and poor brain MFS (HR = 2.567 [95% CI = 1.155–5.702], P = 0.021) after adjustment for age, sex, race, performance status, smoking status, tumor histology and volume, stage, receipt of concurrent radiochemotherapy, number of chemotherapy cycles, and radiation dose. Transfection with TGFβ1+29C (vs. +29T) stimulated the migration and invasion of A549 and PC9 cells, suggesting that TGFβ1+29C may be linked with increased metastatic potential. Conclusions TGFβ1 genotypes at rs1800469 and rs1982073 could be useful for predicting DMFS among patients with NSCLC treated with definitive radiation therapy. These findings require validation in larger prospective trials and thorough mechanistic studies.


Introduction
Lung cancer remains the leading cause of cancer death, with an estimated 160,340 deaths in the United States in 2012 [1]. The current standard of care for unresectable non-small-cell lung cancer (NSCLC), by far the most common presentation at diagnosis, is radiotherapy in combination with chemotherapy. However, survival rates after such therapy remain poor because of high rates of recurrence, both local and distant. Commonly used radio(chemo)therapy regimens are also associated with significant normal tissue toxicity, including radiation pneumonitis, which can be lethal. Better treatment strategies designed according to each individual patient's intrinsic sensitivity and risk of toxicity, the tumor's unique expression of molecular target(s), and the risks of locoregional or distant disease recurrence are urgently needed.
One potential molecular biomarker for predicting treatment outcome after radiochemotherapy for NSCLC, in terms of both tumor response and normal tissue damage, is transforming growth factor b1 (TGFb1). Members of the TGFb cytokine superfamily elicit a diverse range of cellular responses, including proliferation, migration, fibrosis, apoptosis, tissue inflammation, wound repair, and angiogenesis [2][3][4]. Changes in TGFb signaling have been reported in several types of cancer in humans, including NSCLC [5,6]. Although TGFb typically acts as a tumor suppressor [7,8], as tumor development progresses, tumor cells acquire resistance to TGFb-induced growth arrest. Of the three known TGFb isoforms, TGFb1 seems to be directly involved in mediating the metastatic activity of cancer cells [9]. Preclinical studies show that inhibition of TGFb1 signaling downregulates the migration of tumor cells and suppresses the development of metastasis [10,11]. Thus, TGFb1 may have a dual role in cancer development, suppressing tumor growth during the initial phases of carcinogenesis but promoting tumor progression and metastasis in more advanced stages [12,13]. Indeed, increasing evidence shows that TGFb signaling has a role in the development of lung cancer [14,15]. TGFb1 has also been shown to induce the epithelial-tomesenchymal transition, which has a crucial role in the invasion and metastasis of some types of tumor cells [16]. TGFb has further been implicated in tumor cell sensitivity to radiation and chemotherapeutic drugs [17][18][19].
In addition to its effects on tumor cells, TGFb1 is also involved in the response of normal tissues to radiation or radiochemotherapy. Two groups have shown preliminary evidence that changes in plasma levels of TGFb1 during radiotherapy may predict the risk of developing radiation-induced lung toxicity [20,21]. Strong associations have also been reported between single nucleotide polymorphisms (SNPs) in TGFb1 and radiation-induced fibrosis in patients with breast cancer or gynecologic cancers [22,23] and between SNPs and radiation pneumonitis in patients with lung cancer treated with radiotherapy or radiochemotherapy [24]. Explorations of the prognostic value of TGFb1 SNPs in a variety of cancer types [25,26] suggest that some TGFb1 genotypes predict more aggressive tumor phenotypes and adverse prognosis.
We previously assessed the relationship between three different SNPs of TGFb1 [rs1800469 (C-509T), rs1800471 (G915C), and rs1982073 (T+29C)] and the development of radiation pneumonitis in patients with NSCLC treated with radiochemotherapy and observed that the CT/CC genotypes of TGFb1 rs1982073 (T+29C) were associated with significantly lower risk of radiation pneumonitis [24]. Moreover, rs1982073 (T+29C) and rs1800469 (C-509T) have been linked with increased serum levels of TGFb1 [26] and with the incidence of invasive breast cancer [27]; rs1800469 (C-509T) has also been linked with advanced-stage prostate cancer [26] and colon cancer [28]. On the basis of these published findings, we investigated whether these functional genetic variants of TGFb1 also affect tumor response and outcome in patients with inoperable NSCLC treated with definitive radiation or radiochemotherapy, assessed in terms of overall survival (OS) and distant metastasis-free survival (DMFS).

Methods
Subjects for this retrospective analysis were selected from a large database of 740 patients with NSCLC who had been treated at The University of Texas MD Anderson Cancer Center from November 1998 through 2009. A total of 261 patients were identified as having had whole-blood samples available for analysis; all of these patients had been treated in November 1998 through January 2005. From those 261 patients, 56 were excluded: 25 who had died of other diseases or whose vital status was unknown; 16 who had died of NSCLC without information on distant metastasis or locoregional disease, 2 who had had surgery, 1 who had had small-cell lung cancer, and 12 who had had stage IV disease, leaving a total of 205 patients with complete information for the current analysis.

Genotyping
Genomic DNA was extracted from leukocytes from whole blood samples with a DNA Blood Mini Kit (Qiagen, Valencia, CA) according to the manufacturer's instructions. DNA purity and concentration were determined by spectrophotometric measurement of absorbance at 260 and 280 nm. SNPs in TGFb1 were chosen that met at least two of the following three criteria: (1) reported associations with lung or other types of cancer; (2) a minor allele frequency of .5% in individuals of Caucasian descent; and (3) location in the promotor or coding region of the gene. We genotyped three such SNPs: rs1800469 (C-509T, in the promoter region), rs1800471 (G915C, in exon 1), and rs1982073 (T+29C, also in exon 1) by polymerase chain reaction (PCR)restriction fragment length polymorphism as described elsewhere [24]. In vitro Assessment of rs1982073 TGFb1+29C or TGFb1+29T Stable Transfectants We also tested if TGFb1 rs1982073 (T+29C) C variant genotypes influenced the metastatic potential of lung cancer cells in vitro as follows. For these experiments, A549 and PC9 cells, both originating from human lung adenocarcinoma, were cultured in Ham's F12 (A549) or RPMI 1640 (PC9) containing 10% fetal bovine serum and antibiotics (100 units/mL penicillin and 100 units/mL streptomycin) at 37uC with 5% CO 2 , andmedium was refreshed every 2-3 days. When the cell cultures reached about 50% confluence in fresh serum-free medium, they were transfected with a control virus or with a TGFb1 (+29C or +29T)overexpressing lentivirus (constructed as described below) at a multiplicity of infection of 100, incubated for 48 hours, and cultured further in Ham's F12 (A549) or RPMI 1640 (PC9). The transfection efficiency, measured in terms of cellular expression of green fluorescent protein by fluorescence microscopy (Leica DMI4000B), was found to be .99%. The transfected cells were then tested for their invasiveness and motility as described below.
TGFb1 cDNAs were amplified by PCR, with a cDNA library from bronchial smooth muscle cells was used as the DNA template.
After transfection, the A549 and PC9 cells were extracted into cold lysis buffer [50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1 mM EDTA, 1 mM MgCl2, 0.5% Triton X-100, and protease inhibitor mix (1 mM PMSF, 2 mg/mL aprotinin, 1 mg/mL leupeptin, and 1 mg/mL pepstatin A)]. The cell lysates were then clarified by centrifugation at 10,000 6 g for 20 min, the supernatants were collected, and the protein concentration of each lysate was determined with a BCA Protein Assay Kit. Equal amount of proteins in lysates were separated on 12% sodium dodecyl sulfate for polyacrylamide gel electrophoresis and transferred to polyvinylidene difluoride membranes (Bio-Rad). TGFb1 was detected by specific antibody #3709 from Cell Signaling Technology.
Cell migration was assessed with a scratch assay and cell invasiveness with a transwell assay as follows. For the scratch assay, A549 and PC9 cells were seeded in 24-well culture plates at 2610 5 cells/well and incubated for 24 hours, after which the monolayer was scratched with a sterile 200-mL pipette tip, the plates were washed twice with phosphate-buffered saline, and the cells incubated with serum-free Ham's F12 (A549) or serum-free RPMI (PC9) for 24 hours, after which images were captured by microscopy at 1006 magnification and the ''healing'' areas were measured with a phase contrast microscope (Leica DMI4000B). For each cell type, three independent transfection experiments were performed, and all experiments were carried out in triplicate.
For the transwell assays, A549 and PC9 cells were suspended in serum-free Ham's F12 or serum-free RPMI 1640 at a density of 3610 5 per mL, and 100-mL aliquots were added to the upper compartments of a 24-well transwell chamber containing polycarbonate filters with 8-mm pores and coated with 60 mL of Matrigel (1:9 dilution; Sigma Aldrich). Five hundred-microliter aliquots of Ham's F12 or RPMI 1640 with 10% FBS were added to the lower chambers, and the chambers were incubated for 24 hours. At that time, cells in the upper compartment were removed with a cotton swab, rinsed with PBS, and fixed in 100% methanol. Cells that had invaded through the Matrigel to the lower surface were stained with 49,6-diamidino-2-phenylindole and quantified by

Statistical Analysis
In the clinical sample correlation studies, OS time was calculated from the first day of radiation treatment until death or the last known follow-up. DMFS times were calculated from the first day of radiation treatment to the first observed day of distant metastasis or the last known follow-up. Associations between the distribution of TGFb1 genotypes and clinical characteristics were assessed by Fisher's exact tests or Pearson's x 2 tests. Cox proportional hazards analysis was used to evaluate the influence of various TGFb1 genotypes on OS and DMFS, calculated as hazard ratios (HRs) with corresponding 95% confidence intervals (CIs). All HRs were adjusted for age, sex, race, Karnofsky performance status, smoking status, tumor histology, gross tumor volume, disease stage, receipt of chemotherapy or concurrent radiochemotherapy, number of cycles of chemotherapy, and radiation dose received. Kaplan-Meier analysis was used to evaluate the effect of TGFb1 genotype on the cumulative probability of DMFS. All reported P values were two-sided, and P,0.05 was considered to indicate statistical significance. Statistical analyses were carried out with SPSS 16.0 (SPSS Inc. Chicago, IL).
The in vitro data were expressed as means 6 SD from three independent experiments (each of which had been performed in triplicate) and compared with Student's t tests. P values of ,0.05 were considered to indicate statistically significant differences.

Clinical Sample Correlation Studies
The clinical characteristics of the 205 analyzed patients are summarized in Table 1. Patient-, disease-, and treatment-related characteristics according to TGFb1 polymorphisms were well balanced in terms of age, race, sex, Karnofsky performance status, tobacco use, tumor histology, gross tumor volume, disease stage, number of cycles of chemotherapy, receipt of chemotherapy vs. radiation only, and radiation dose received (not shown).  Table 3. TGF-b1 genotype and brain metastasis-free survival in 102 NSCLC patients without distant metastasis and 39 NSCLC patients with brain metastasis. The median follow-up times were 17 months for the entire group and 39 months for patients who were alive at the time of analysis (range, 1-97 months). The median OS time was 21 months. A total of 103 patients (50%) developed DM at 134 sites; 27 patients had DM at more than 1 site (23 at 2 sites and 4 at 3 sites). The sites of metastases included brain (n = 39), bone (n = 26), lung (n = 26), adrenals (n = 13), liver (n = 14), and other unspecified sites (n = 16). Figure 1 illustrates the Kaplan-Meier analysis of the effect of TGFb1 genotype and cumulative probability of DMFS. The TGFb1 rs1800469 (C-509T) CT+TT genotypes were associated with significantly shorter median DMFS (11 months vs. 31 months for the CC genotype, P = 0.010; Fig. 1A). The TGFb1 rs1982073 (T+29C) CT+CC genotypes were also associated with significantly shorter median DMFS (compared with the TT genotype) (P = 0.013; Fig. 1C). However, no associations were found between any genotype (GG, CG, CC, or CG+CC) of the TGFb1 rs1800471 (G915C) SNP and DMFS (Fig. 1B).
We further analyzed the association of TGFb1 rs1800469 (C-509T), rs1800471 (G915C), and rs1982073 (T+29C) genotypes with the organ sites in which metastases developed by using Cox proportional hazard models. The CT or TT genotypes of rs1800469 (vs. CC) and the CT or CC genotypes pf rs1982073 (vs. TT), were both associated with a twofold increase in HRs for brain, bone, and lung metastases in univariate analyses. However, in multivariate analyses, the only significant association was between rs1982073 CT or CC (vs. TT) and brain metastases (HR = 2.567 [CI = 1.155-5.702], P = 0.021) ( Table 3).

In vitro Findings
Transfection efficiency for both A549 and PC9 lung cancer cell lines was approximately 99% (Figs. 2A,B). Transfection of A549 and PC9 cells with TGF-b1+29C or +29T induced the expected increases in the expression of both precursor and active forms of TGFb1 (Figs. 2C,D). Both cell lines, upon transfection with the TGFb1+29C construct, demonstrated increased motility (Figs. 3A,B,C,E) and increased invasion (Figs. 3D,F) relative to the +29T transfectants. These results suggest that the TGFb1+29C genotype in both of these lung cancer cell lines induced a phenotype with increased potential for metastasis.

Discussion
Metastasis remains a major cause of treatment failure for patients with NSCLC treated with radiotherapy, with or without chemotherapy [29,30]. Factors that predict the biologic behavior and metastatic potential of NSCLC have been actively sought for the past 3 decades. The most significant finding from the current study is the association between variant TGFb1 genotypes at rs1800469 and rs1982073 and worse DMFS among patients with NSCLC given radiotherapy; this finding suggests that SNPs in TGFb1 genes could be used as biomarkers to prescribe personalized radiochemotherapy according to the expected pattern of recurrence before the initiation of treatment. We further validated the association of the TGF-b1 SNP rs1982073 T+29C with Further analyses of our results showed a statistically higher risk of brain metastasis in patients with the TGFb1 rs1982073 (T+29C) C variant genotypes. Brain metastasis is a common complication of locally advanced NSCLC, with 21% to 54% of patients developing brain metastasis during the course of the disease [31,32]. The OS time of patients with brain metastasis generally ranges from about 3 to 6 months [33][34][35], and 30%-50% of patients with brain metastases die from neurologic causes [36,37]. Prophylactic cranial irradiation (PCI) has been shown in two randomized studies to reduce the incidence or delay the onset of brain metastasis in patients with locally advanced NSCLC after primary therapy, and use of PCI can prolong OS in such patients [38,39]. The Radiation Therapy Oncology Group (RTOG), in an attempt to extend these results, opened a large trial (RTOG 0214) of PCI versus observation for patients with locally advanced NSCLC. In patients with stage III disease without progression of disease after therapy, PCI decreased the rate of brain metastasis but did not improve OS or disease-free survival [40]. Ideally, one would prescribe PCI only for patients at high risk of developing brain metastases. However, so far, no reliable way has been found to identify which patients will develop brain metastasis and which will not-hence the intense search for biomarkers. Therefore, our finding that TGFb1 rs1982073 (T+29C) C variant genotypes were associated with brain metastasis is of particular importance in the management of NSCLC, because this SNP might serve as a biomarker to identify patients who would benefit from PCI.
The possibility of using TGFb1 SNPs as predictive biomarkers in other kinds of cancer has been studied by others. In one such study, the TGFb1 rs1800469 C-509T homozygous TT genotype was found to be associated with an increased risk of having an aggressive form of prostate cancer, but the rs1982073 T+29C (T+29C) C variant genotypes were not [26]. In contrast, another group showed that patients with breast cancer who carried the T+29C C allele had lower 5-year disease-free survival rates than did those with the TT genotype [25]. Other evidence is emerging that SNPs in some other genes may predict outcome in NSCLC. For example, genetic aberrations in genes for matrix metalloproteinase (MMP) -3 and MMP-9 or the x-ray repair protein XRCC1 have been reported to influence the clinical behavior and risk of metastasis of lung cancer [41][42][43]. These biomarkers have yet to be widely translated into clinical practice. Our findings are in line with these reports: DM was a major pattern of failure in our study (DMFS rate, 50.2%), and variant genotypes at TGFb1 rs1800469 (C-509T) and TGFb1 rs1982073 (T+29C) seemed to be associated with shorter DMFS times. If validated in larger studies, our findings could serve as the basis for using these genotypes as a prognostic marker of clinical outcome, particularly predisposition to brain metastasis, in patients with NSCLC.
The mechanistic basis of the findings from the current study needs further investigation. Some investigators have suggested that TGb1 protein levels might independently predict survival in patients with adenocarcinoma of the lung [18,[44][45][46]. In those studies, TGFb1 expression in primary lung cancer tissues was higher among patients with pulmonary metastases than that among patients without such metastases. Earlier reports suggested that polymorphisms in the signal sequence at +915 [47], C-509T [48], and T+29C (encoding Leu10Pro) in the TGFb1 gene were associated with increased TGFb1 production. Those studies may shed light on our finding that variant genotypes at TGFb1 rs1800469 (C-509T) and TGFb1 rs1982073 (T+29C) were associated with shorter DMFS times in patients with NSCLC treated with definitive radiation therapy.
In our study, having the TGFb1 rs1800469 (C-509T) CT or TT genotypes seemed in univariate analysis to be associated with shorter OS, and no associations were found between OS and TGFb1 rs1982073 (T+29C) genotypes. This finding is inconsistent with a report by Mu et al. [49] showing that patients with earlystage breast cancer and the T/T genotype or high intratumoral TGFb1 levels had shorter OS times than did those without T/T or with low intratumorial TGFb1 levels. However, this association was different for patients with late-stage disease: for those patients, having the T/T genotype was associated with lower risk of disease recurrence. These findings indicate that TGFB1 has complex roles in breast cancer progression, which supports the findings of other studies that TGFb1 has conflicting effects on tumor growth and metastasis.
Our study had several limitations; its single-institution and retrospective nature require that its findings be validated in larger, preferably multi-institutional prospective trials. Another limitation was that we did not measure TGFb1 protein levels in plasma before, during, or after radiation therapy, and so we could not determine if having the TGFb1 rs1982073 (T+29C) CT genotype influenced the expression of TGFb1 protein levels in vivo, nor could we determine whether the risk of DM correlates with changes in this variable before or after treatment. Thus clarification is still needed regarding the relationships among polymorphisms in TGFb1 and TGFb1 protein levels in plasma and prognosis in patients with NSCLC. Finally, our research was limited to patients who received definitive radiation therapy. Whether the TGFb1 rs1982073 variant genotypes also predict outcome among patients who undergo surgery or other nonradiation-based therapies remains unclear.
Our previous work [24] showed that having the CT or CC variants in TGFb1 rs1982073 was associated with lower risk of severe (grade $3) radiation pneumonitis, an observation suggesting that radiation dose intensification to increase locoregional tumor control in patients with this genotype would be feasible. However, the results we report here showed that patients with those genotypes were at increased risk of DM, specifically brain metastases, an observation suggesting that systemic treatment and PCI are indicated. The results from our study suggest, for the first time, that the TGFb1 rs1982073 variant genotypes may be useful as a genetic biomarker to help physicians tailor the treatment according to the patient's risk of radiation pneumonitis and DM. As such, additional studies are needed to validate this finding and to investigate the mechanisms by which TGFb1 polymorphisms influence the course of lung cancer development and progression. These kinds of studies will help to provide a better understanding of the cancer microenvironment and metastasis and may be instrumental in tailoring treatment for NSCLC.