Polymorphisms of the vascular endothelial growth factor gene and severe radiation pneumonitis in non‐small cell lung cancer patients treated with definitive radiotherapy

Abstract

Vascular endothelial growth factor (VEGF) is a major mediator of angiogenesis and lung cancer progression. We hypothesized that VEGF polymorphisms may modulate the risk of radiation pneumonitis (RP) in non‐small cell lung cancer (NSCLC) patients treated with definitive radiotherapy. We genotyped three potentially functional VEGF single nucleotide polymorphisms (&460 T > C [rs833061], &634 G > C [rs2010963] and +936 C > T [rs3025039]) and estimated the associations of their genotypes and haplotypes with severe radiation pneumonitis (RP ≥grade 3) in 195 NSCLC patients. We found that the VEGF genotypes of rs2010963 and rs3025039 single nucleotide polymorphisms as well as the &460C/&634G/+936C haplotype were predictors of RP development (adjusted hazard ratio [adjHR] = 2.33, 95% confidence interval [CI], 1.01–5.37, P = 0.047 for CC vs GG genotypes; adjHR = 28.13, 95% CI, 5.24–151.02, P < 0.001 for TT vs CC genotypes; and adjHR = 2.51, 95% CI, 1.27–4.98, P = 0.008 for T‐C‐T vs C‐G‐C haplotypes). In addition, there was a trend towards reduced RP risk in patients carrying an increased number of protective VEGF genotypes. Our data suggest that VEGF polymorphisms can modulate the risk of radiation pneumonitis in NSCLC patients treated with definitive radiotherapy. Large and independent studies are needed to confirm our findings. (Cancer Sci 2012; 103: 945–950)

Radiotherapy, alone or combined with concurrent or sequential chemotherapy, is the standard treatment approach for patients with inoperable or unresectable non‐small cell lung cancer (NSCLC) because it improves local control and overall survival. However, radiation pneumonitis (RP), which can occur after radiation therapy as a result of the inflammation of normal lung tissues, has been identified as one of the most common dose‐limiting complications of thoracic radiation, with an incidence rate between 10 and 20% for moderate or severe RP.1 Previous studies have demonstrated an association between RP risk and multiple therapeutic and patient‐related factors, such as Karnofsky performance status (KPS), dosimetric parameters, smoking and plasma inflammatory cytokine levels.2 In addition, it has been hypothesized that polymorphisms of some candidate genes can be used as biomarkers to predict RP susceptibility. It was shown in our previous study that genetic variants of the TGFβ1 gene (e.g. rs1982073:T869C) independently predict RP risk in NSCLC patients.3

Angiogenesis is an essential process for the development, growth and metastasis of malignant tumors, including lung cancer, and is also critical for a number of physiologic processes in normal tissues.4 Radiation therapy can exert an antivascular effect and induces apoptosis of vascular endothelial cells, which can be observed within hours of ionizing radiation in malignant and normal tissues.5, 6, 7 The antivascular and antiangiogenic effects of radiation therapy are associated with increased production of pro‐angiogenic cytokines by tumor and tumor‐associated stromal cells, including vascular endothelial growth factor (VEGF), platelet‐derived growth factor and basic fibroblast growth factor, which can then promote tumor revascularization and thereby counteract the antivascular effects of radiation therapy.8 Hence, the combined use of antiangiogenetic agents with radiotherapy significantly enhances the radiosensitivity of tumors and their surrounding normal tissues, which, can inadvertently lead to increased toxicities in these surrounding normal tissues.9, 10 Although the combination of radiation therapy with the antiangiogenic agent bevacizumab does not appear to increase acute lung toxicity in breast cancer patients,11 increased toxicites have been observed when used with radiation therapy in some lung cancer patients.12 Conversely, high expression levels of angiogenesis‐stimulating cytokines, such as bFGF, have been reported to defend against RP in vitro and in vivo, presumably through the protection of the existing normal vasculature and the re‐growth of endothelial cells and promotion of the repair and revascularization of normal lung tissues.13, 14, 15 Given its central role in angiogenesis, we hypothesized that VEGF single nucleotide polymorphisms (SNP) might also have an impact upon treatment efficacy and tissue toxicity of radiotherapy in the lung.

The VEGF gene is located on chromosome 6p21.3 and consists of eight exons.16 At least 30 SNPs have been described in this gene, among which −460 T > C (rs833061), −634 G > C (rs2010963) and +936 C > T (rs3025039) have been found to be involved in the regulation of VEGF expression.17, 18, 19, 20 However, few published studies have investigated the influence of VEGF variants upon radiation toxicities in irradiated tissues. Only one paper reports an association between the VEGF G1154A polymorphism and severe hematological toxicities in irradiated esophageal cancer patients.21

Previous studies have primarily focused on the three common functional SNPs of the VEGF gene (i.e. −460T > C [rs833061], −634G > C [rs2010963] and +936 C > T [rs3025039]; minor allele frequency [MAF] = 0.422, 0.431 and 0.222 in Caucasians, respectively, according to the Hapmap database). The −460 T > C SNP is located in the promoter region and may influence promoter activity;18 the −634 G > C SNP lies within the 5′‐untranslated region and may affect transcriptional factor binding affinity;22 and the +936 C > T SNP is located in the 3′‐untranslated region and has been associated with lower VEGF plasma levels.17 In a recent study, we found an association between VEGF −460 C genotypes and survival for locally advanced NSCLC after radiotherapy.23 We hypothesized that one of the possible mechanisms for such an association is that VEGF variants might modulate cellular response to radiotherapy. In the present study, we evaluated the association of the above three potentially functional VEGF SNPs with radiation pneumonitis, a well‐known dose‐limiting factor in the treatment of NSCLC patients who receive definitive radiotherapy.

Materials and Methods

Study populations

Clinical data were derived from a larger dataset of 261 patients with histopathologically confirmed NSCLC, who were treated with definitive radiation at M.D. Anderson Cancer Center between 1999 and 2005. Among these 261 patients, 195 patients had documented information on RP with complete follow‐up information and radiation dosimetric data. The median total radiation dose was 63 Gy (range, 50.4–84.0 Gy) given between 1.2 and 2.0 Gy/fraction. Of the 195 patients, 20 received radiation only and 175 received radiation in combination with chemotherapy. The details of the radiation treatment planning, follow‐up schedule and tests, guidelines for RP scoring and dosimetric data analysis have been described previously.24, 25 The current investigation was approved by The University of Texas M. D. Anderson Cancer Center Institutional Review Board.

Genotyping

Genomic DNA was extracted from the buffy coat fraction of each blood sample using a Blood Mini kit (Qiagen, Valencia, CA, USA) according to the manufacturer's instructions. DNA purity and concentrations were determined by spectrophotometric measurement of absorbance at 260 and 280 nm using a UV spectrophotometer. The selected three VEGF SNPs (−460 T > C [rs833061], −634 G > C [rs2010963] and +936 C > T [rs3025039]), which are relatively common in Caucasian populations (MAF > 0.05), were genotyped using the PCR‐restriction fragment length polymorphism (RFLP) method. The primer sequences, restriction enzymes and PCR conditions used for the experiments are described elsewhere.23

Statistical analysis

Clinical end points (i.e. the development of RP grade 3 or higher) were assessed and scored using Common Terminology Criteria for Adverse Events version 3.0 (National Cancer Institute, Bethesda, MD, USA). The times to development of end points were calculated from radiotherapy starting time, and patients not experiencing the end point were censored at the final follow up. Patients were grouped according to their genotypes. Statistical analysis was performed using SAS 9.1 statistical software (SAS, Chicago, IL, USA). Cox proportional hazards regression analysis was performed to calculate the hazard ratio (HR) and the 95% confidence interval (CI) to evaluate the influence of genotypes on RP risk. In addition, multivariate Cox regression was performed to adjust for other covariates. Kaplan–Meier analysis was performed to estimate the cumulative RP probability. Haplotype frequencies and individual haplotypes were generated using SAS PROC HAPLOTYPE. Likelihood ratio tests were performed for each multivariate Cox regression to assess the goodness‐of‐fit. A P‐value of 0.05 or less was considered statistically significant.

Finally, we used classification and regression tree (CART) analysis to identify higher‐order interactions between clinical factors and genetic variants with the rpart package in S‐PLUS Version 8.0.4 (TIBCO, Palo Alto, CA, USA). CART creates a decision tree to assess how well each genotype or clinic‐pathologic variable predicts the class (e.g. RP). Specifically, the recursive procedure starts at the root node and uses a log‐rank statistic to determine the first optimal split and subsequent split of the dataset. This process continues until the terminal nodes have no subsequent statistically significant splits or the terminal nodes reach a pre‐specified minimum size (n > 10).

Results

Population characteristics and risk of radiation pneumonitis

Clinical and pathological characteristics of the patients included in the current study are shown in Table 1. Among the 195 patients, there were 112 men and 83 women, with a median age of 63 years (range, 35–88 years), of whom 71.3% were Caucasian, 85.1% had stage III/IV diseases, 89.7% were treated with a combination of chemotherapy and radiation therapy, and 96.8% received radiation doses between 60 and 70 Gy. The median follow‐up time for assessment for RP of any grade after radiation therapy, with or without chemotherapy, was 21 months, and the median time to RP (grade ≥3) development was 3.5 months (95% CI, 3.3–4.5 months). We estimated the associations between RP and clinic‐pathologic characteristics to identify any confounding factors, including age, sex, race, KPS, disease stage, tumor histology, smoking history, use of chemotherapy, radiation dose and mean lung dose (MLD). In the univariate analysis, we found that sex and MLD were significantly associated with severe RP (grade ≥3) (HR for men vs women 2.28, 95% CI 1.07–4.86, P = 0.034; and HR for MLD ≥ 19.7 vs MLD < 19.7 2.16, 95% CI 1.04–4.50, P = 0.04). No other clinic‐pathologic characteristics were found to be associated with severe RP risk in this study population.

Table 1.

Patient demographics and association with radiation pneumonitis (grade ≥ 3) (N = 195)

Parameter	Number (%)	HR	95% CI	P a
Age (years)
< 63	96 (49.2)	1.00
≥63	99 (50.8)	1.23	0.63–2.40	0.565
Sex
Female	83 (42.6)	1.00
Male	112 (57.4)	2.28	1.07–4.86	0.034
Race
White	139 (71.3)	1.00
Black	41 (21.4)	0.68	0.26–1.77	0.430
Other	15 (7.7)	1.45	0.44–4.83	0.531
KPS
< 80	36 (18.5)	1.00
≥80	159 (81.5)	0.66	0.30–1.45	0.301
Stage groups
III, IV	166 (85.1)	1.00
I, II	29 (14.9)	0.97	0.40–2.34	0.942
Histology
Adenocarcinoma	66 (33.7)	1.00
NSCLC, NOS	67 (34.2)	1.17	0.54–2.53	0.692
Squamous cell	62 (32.1)	0.77	0.33–1.84	0.561
Smoking status
Ever	176 (90.7)	1.00
Never	18 (9.3)	1.36	0.77–2.40	0.292
Chemotherapy
No	20 (10.3)	1.00
Yes	175 (89.7)	1.35	0.41–4.43	0.617
Mean lung dose (Gy)
< 19.7	88 (45.1)	1.00
≥19.7	107 (54.9)	2.16	1.04–4.50	0.040
Radiation dose (Gy)
< 63	36 (18.5)	1.00
≥63	159 (81.5)	0.71	0.31–1.63	0.423

^a P‐values were calculated using the Cox proportional hazard model and univariate analysis. CI, confidence interval; HR, hazard ratio; KPS, Karnofsky performance status; NOS, not otherwise specified NSCLC, non‐small cell lung cancer.

Radiation pneumonitis and vascular endothelial growth factor genotypes

Table 2 shows univariate and multivariate analyses of the association between VEGF genotypes and grade ≥3 RP using the Cox proportional hazards regression model. In the univariate analysis, only the TT genotype of the +936 C > T (rs3025039) polymorphism was associated with significantly increased risk of grade ≥3 RP (crude HR = 42.70, 95% CI, 8.36–218.16, P < 0.001). In the multivariate analyses performed with adjustment for sex, KPS, TNM stage and MLD, which were potential confounding factors of RP, as suggested by the current and previous studies,2, 3 both the CC genotype of the −634 G > C (rs2010963) polymorphism and the TT genotype of the +936 C > T (rs3025039) polymorphism were associated with significantly increased RP risk (CC vs GG: adjusted HR = 2.33, 95% CI, 1.01–5.37, P = 0.047; and TT vs CC: adjusted HR = 28.13, 95% CI, 5.24–151.02, P < 0.001, respectively). We also found a trend of additive effects of the risk alleles of −634 G > C (rs2010963) and +936 C > T (rs3025039) SNPs.

Table 2.

Univariate and multivariate analyses of different genotypes and radiation pneumonitis (grade ≥3) for patients with non‐small cell lung cancer

			Crude			Adjusted
VEGF genotypes	Patient number	Event	HR	95% CI	P a	HR	95% CI	P b	P trend
−460 T > C (rs833061)
TT	58	10	1.00			1.00
CT	99	21	1.25	0.59–2.66	0.557	1.07	0.49–2.34	0.861
CC	38	5	0.62	0.19–1.97	0.415	0.65	0.20–2.11	0.471	0.577
CC vs CT + TT	157	31	0.53	0.19–1.51	0.236	0.62	0.21–1.83	0.387
−634 G > C (rs2010963)
GG	96	13	1.00			1.00
CG	58	12	1.63	0.73–3.62	0.234	1.43	0.63–3.26	0.390
CC	41	11	2.21	0.97–5.00	0.058	2.33	1.01–5.37	0.047	0.050
CC + CG vs GG	99	23	1.86	0.93–3.74	0.082	1.77	0.87–3.59	0.113
+936 C > T (rs3025039)
CC	145	23	1.00			1.00
CT	48	11	1.45	0.70–2.99	0.316	1.40	0.67–2.91	0.376
TT	2	2	42.70	8.36–218.16	<0.001	28.13	5.24–151.02	<0.001	0.024
TT + CT vs CC	50	13	1.70	0.86–3.38	0.129	1.62	0.80–3.25	0.178
−460CC/−634GG/+936CC‡
0	29	11	1.00			1.00
1	83	13	0.38	0.17–0.84	0.017	0.45	0.20–1.03	0.059
2	53	8	0.37	0.15–0.92	0.033	0.40	0.16–1.01	0.053
3	30	4	0.26	0.07–0.92	0.037	0.34	0.09–1.27	0.109	0.025

^a P‐values were calculated using the Cox proportional model and univariate analysis.

^b P‐values were calculated with adjustment for sex, Karnofsky performance status, tumor stage and mean lung dose. 0 (1, 2, 3): indicates the number of combined protective genotypes of −460CC, −634GG and +936CC in individual patients. CI, confidence interval; HR, hazard ratio; KPS, Karnofsky performance status; VEGF, vascular endothelial growth factor.

To understand the joint effect of these three SNPs upon individual risk of RP, we combined protective genotypes of the −460 CC (rs833061), −634 GG (rs2010963) and +936 CC (rs3025039) genotypes. We found that patients with an increasing number of protective genotypes showed a trend towards reduced frequencies of severe RP, compared with those without protective genotypes (P _trend  = 0.025) (Table 2). Figure 1 plots the incidence of RP grade ≥3 as a function of time since radiation therapy started according to the selected SNP.

Figure 1.

Risk of radiation pneumonitis (grade ≥3) by selected polymorphisms of the VEGF gene. (a) −460 T > C (rs833061); (b) −634 G > C (rs2010963); (c) +936 C > T (rs3025039); and (d) combined protective genotypes of −460 CC (rs833061), −634 GG (rs2010963) and +936 CC (rs3025039). RP, radiation pneumonitis.

Radiation pneumonitis and vascular endothelial growth factor haplotypes

We explored VEGF haplotypes to determine which haplotype was most likely to be associated with RP risk (grade ≥3). Among all cases, there were four haplotypes with frequencies >5%. Other less common haplotypes (frequencies <5%) were combined into one haplotype group. The four most common haplotypes in the patients were −460C/−634G/+936C (C‐G‐C), T‐C‐C, T‐G‐C and T‐C‐T, with respective frequencies of 40.0, 24.1, 19.7 and 7.9%, which were similar to those reported in other Caucasian populations.26 Compared with the most common C‐G‐C haplotype, the T‐C‐T haplotype was associated with significantly increased RP risk in both univariate and multivariate models (crude HR = 2.95; 95% CI = 1.50–5.80, P = 0.002; adjusted HR = 2.51; 95% CI = 1.27–4.98, P = 0.008) (Table 3).

Table 3.

Univariate and multivariate analyses of different haplotypes and radiation pneumonitis (grade ≥3)

Haplotype	Number	Event	HR	Crude HR	P a	HR	Adjusted HR	P b
				95% CI			95% CI
C‐G‐C	152	26	1.00			1.00
T‐C‐C	94	17	1.10	0.59–2.05	0.756	1.14	0.61–2.13	0.687
T‐C‐T	31	13	2.95	1.50–5.80	0.002	2.51	1.27–4.98	0.008
T‐G‐C	77	10	0.81	0.39–1.69	0.575	0.77	0.37–1.61	0.486
Others	36	6	1.00	0.41–2.45	0.999	1.07	0.44–2.63	0.886

^a P‐values were calculated using the Cox proportional hazards model and univariate analysis.

^b P‐values were obtained using the Cox hazards model with adjustment for sex, Karnofsky performance status, tumor stage and mean lung dose. CI, confidence interval; HR, hazard ratio.

Association of multiple factor interaction with radiation pneumonitis risk (classification and regression tree analysis)

To assess the impact of higher‐order interactions on the risk of developing RP, we built a decision tree using clinico‐pathologic variables (i.e. sex, disease stage, KPS group and MLD) and all SNPs included in the current study. In the CART analysis, the first split in the decision tree was MLD, indicating that MLD was the strongest predictor for RP among the factors considered. Further inspection of the tree structure showed distinct patterns between low MLD (<19.7 Gy) and high MLD (≥19.7 Gy) groups (Fig. 2), which were split by KPS and sex, respectively. Genetic factors of −634 G > C (rs2010963) and +936 C > T (rs3025039) polymorphisms were identified as the third splitter. Compared with the lowest RP risk group (+936 CC [rs3025039] genotype with MLD < 19.7 Gy and KPS ≥ 80), male patients of −634 CG/CC (rs2010963) genotypes with MLD ≥ 19.7Gy had the highest risk of developing severe RP (HR = 11.12, 95% CI 2.49–49.72) (Fig. 2).

Figure 2.

Classification and regression tree analysis for predictors of grade ≥3 radiation pneumonitis in 195 patients with non‐small cell lung cancer. HR, hazard ratio; KPS, Karnofsky performance status; MLD, mean lung dose.

Discussion

To our knowledge, this is the first study evaluating associations between potentially functional VEGF SNPs and severe RP in lung cancer patients treated with definitive radiotherapy. Our data indicate that some VEGF variants (e.g. −634 G > C [rs2010963] and +936 C > T [rs3025039]) may independently modulate RP risk, whereas a joint effect of multiple genetic variants (−460 T > C [rs833061], −634 G > C [rs2010963] and +936 C > T [rs3025039]) may have a substantial effect on RP risk in NSCLC patients treated with definitive radiotherapy. The −460T/−634C/+936T haplotype was associated with a significantly higher risk of severe RP, and might be useful biomarkers to predict RP susceptibility.

Thoracic radiation is one of the major approaches taken to control lung malignancies and to reduce symptoms and improve survival. However, it might also cause severe adverse effects by producing tissue‐damaging free radicals, and these adverse effects can limit thoracic radiation doses and volumes and, therefore, restrict use of thoracic radiation treatment. RP is a dysregulated wound‐healing inflammatory reaction to normal lung tissue injury. Previous investigations have identified MLD and being male to be independent risk factors associated with severe RP development,3, 27 and these findings are confirmed in the current study. In addition, there is evidence that substantial vascular damage occurred during acute radiation‐induced pneumonitis in a mouse lung model.28 It is reasonable to speculate that endothelial cells of normal human lung tissues are also significantly damaged in severe RP patients. Damaged vascular endothelial cells initiate early inflammatory responses, increase vascular permeability and facilitate leukocyte infiltration into the inflammatory site, where active interactions between cellular and humoral factors take place, involving expression of inflammatory cytokines, such as tumor‐necrosis factor, interleukin and prostaglandins. Genetic factors that could cause prolonged and intensified inflammatory reactions in response to radiotherapy are potential candidate genes to predict RP sensitivity.

A high level of VEGF expression stimulates vascular endothelial cell growth, proliferation and survival and, therefore, might help maintain endothelial integrity and protect against RP development. However, VEGF can also activate transcriptional factors, such as NF‐kappaB in damaged endothelial cells,29 which might induce the synthesis of pro‐inflammatory cytokines and chemokines.30 As demonstrated previously, VEGF is a potent inducer of capillary permeability and an important contributor to inflammatory response.31 Hence, the effect of VEGF in RP development seems to depend upon a balance between maintenance of vascular endothelial cell integrity and stimulation of an inflammatory response.

According to the Common Terminology Criteria for Adverse Events version 3.0, there are five grades of radiation‐treatment related pneumonitis. Previous studies have used different types of criteria in RP investigation, including RP of any grade (grade ≥1),32, 33 mild or moderate RP (grade ≥2)34 and severe RP (grade ≥3).3 In the present study, we selected severe RP as our study end point, because mild and moderate RP (grade 1 and 2) are not likely to have significant clinical consequences. In our study, there was an increased severe RP risk in patients with the CC genotype of the −634 G > C (rs2010963) polymorphism and the TT genotype of the +936 C > T (rs3025039) polymorphism. These two genotypes were associated with higher levels of VEGF expression in previous investigations.20, 22, 35 Because VEGF might have dual roles in RP development, patients with these genotypes might have intensified inflammatory responses after radiotherapy, as observed in the current study. However, we could not rule out a chance finding of an association with the +936 C > T (rs3025039) polymorphism due to the small number of subjects with the TT genotype in our study population. Notably, the CC genotype of the −460 T > C (rs833061) polymorphism, which was associated with better survival for locally‐advanced NSCLC in our previous study,23 was also associated with a non‐significant decrease of severe RP risk, probably due to the reduced statistical power. Therefore, larger studies are required to validate our results.

In this study, we were interested to find that there appeared to be an additive/synergistic effect against RP development in those carrying the protective genotypes of three selected SNPs. Patients with two or three favorable genotypes showed a dramatically reduced risk of developing RP compared with those not carrying these favorable genotypes. This is biologically plausible because the effect of a single SNP would be mild or modest, whereas a joint effect of multiple SNPs in a gene would likely have a more substantial effect. A trend test further supported this point, although statistical significance was not reached in the multivariate analyses likely due to the small sample size. Furthermore, patients with the T‐C‐T haplotype had a significantly increased RP risk, compared with those carrying the most common −460T/−634C/+936T haplotype.

The CART analysis was an explorative, non‐parametric approach, which required no assumption of a genetic model. Through this decision tree‐based data mining, we identified evidence to support MLD as the most important risk factor for severe RP development. Through a combination of genetic, clinical and dosimetric factors, we divided the patients into several groups with different levels of RP susceptibility. Compared with the lowest RP risk group (+936 CC [rs3025039] genotype with MLD < 19.7 Gy and KPS ≥ 80), patients with high RP risk showed over seven‐fold increased RP risk, suggesting the need for a close surveillance of such a high‐risk group during and after thoracic radiotherapy.

Several limitations of the current study need to be addressed. First, the mechanisms by which the VEGF polymorphisms influence the RP risk were not fully clarified, as they were beyond the scope of the current study. Second, we used a candidate polymorphism or gene approach that allowed us to focus on potentially functional SNPs in the gene, but this approach did not comprehensively cover the entire gene. Some important but rare SNPs may have been missed or the observed associations may have been due to genetic linkages with other untyped SNPs. Therefore, additional investigations of the tagging SNPs are warranted and will be the subject of future studies. Finally, we did not adjust the P‐values for multiple tests, as the current study was primarily exploratory with a limited study power. We plan to confirm our current findings in our ongoing prospective expansion studies with more stringent conditions and larger sample sizes.

In summary, we found that selected VEGF SNPs (i.e. −460 T > C [rs833061], −634 G > C [rs2010963] and +936 C > T [rs3025039]) may have independent or joint effects on RP development in NSCLC patients treated with definitive radiotherapy. These findings, once validated in future prospective studies with large sample sizes and improved study designs, will guide tailored therapeutics for individual lung cancer patients.

Disclosure Statement

The authors have no conflict of interest.