Genetic variants of SP‐D confer susceptibility to radiation pneumonitis in lung cancer patients undergoing thoracic radiation therapy

Abstract Background Surfactant protein D (SP‐D) is an innate immunity molecule in the alveoli. However, the associations between genetic variants of SP‐D and radiation pneumonitis (RP) have never been investigated. Methods The Linkage disequilibrium of SP‐D and tagSNPs were analyzed by using Haploview 4.1. Eight tagSNPs were genotyped among 396 lung cancer patients who received thoracic radiation therapy with follow–up time (median [P25, P75]: 11[6, 18]) using improved multiplex ligation detection reaction (iMLDR). The associations between clinical characteristics, tagSNP alleles, genotypes, haplotypes and onset time of grade ≥2 or ≥3 RP were evaluated by using univariate and multivariate Cox proportional hazard regression model. Results Three tagSNPs of SP‐D (rs1998374, rs911887 and rs2255326) were significantly associated with grade ≥2 RP in multivariate analysis with multiple testing (Q test). The rs199874 had a protective effect for grade ≥2 RP in the dominant model (Hazard ratio (HR), 0.575; 95% confidence interval (CI), 0.378‐0.875). The homozygous mutant genotype for rs911887 had risk effect for grade ≥2 RP (HR, 2.209; 95% CI, 1.251‐3.902). The A mutant allele of rs2255326 also showed an elevated risk for grade ≥2 RP (HR, 1.777; 95% CI, 1.283‐2.461) and this risk effect was still significant in the recessive genetic model (HR, 3.320; 95% CI, 1.659‐6.644) and dominant genetic model (HR, 1.773; 95% CI, 1.166‐2.696). Compared to the lung cancer patients bearing the most common haplotype C‐G‐T, the patients bearing the haplotype T‐A‐C (rs1998374‐rs2255326‐rs911887) showed a significant risk of both grade ≥2 RP (HR, 1.885; 95% CI, 1.284‐2.765) and grade ≥3 RP (HR, 2.256; 95% CI, 1.248‐4.080). Conclusions Genetic variants of SP‐D were associated with risk of RP development in lung cancer patients receiving thoracic radiotherapy.


| INTRODUCTION
With the increasing lung cancer patients diagnosed, more and more lung cancer patients receive thoracic radiotherapy as a part of standard treatment regimen. 1 However, approximately 16%-30% of lung cancer patients experience moderate to severe radiation-induced pneumonitis (RP) within the first 3 months of thoracic radiation therapy, even for those who were treated with intensity-modulated radiation therapy (IMRT). 2 RP limits the escalation of radiation dose necessary to achieve curative effects while reducing the quality of life for patients. [3][4][5] Therefore, establishing reliable predictors of RP is a critical step towards maximizing efficacy of treatment for lung cancer patients while minimizing adverse effects associated with thoracic radiation.
Multiple therapeutic and patient-related RP risk factors include performance status (PS), smoking status, chemotherapy, and dosimetric parameters. 6 However, they are not sufficient to fully explain why under similar radiotherapy doses fraction some patients develop RP while the rest never do, which suggests a genetic basis for RP. RP genetic association studies had identified some RP susceptibility genes, further supporting a genetic basis in RP development. These RP susceptibility genes identified thus far were involved in the DNA repair pathways, [7][8][9] oxidative stress pathways, 10 cellular signaling pathways, 11 and inflammatory response to ionizing radiation. [12][13][14] Surfactant protein D (SP-D), a component of lung surfactants, which reduces surface tension at the pulmonary air-liquid interface [15][16][17] and enhances defense against pathogens as the first line of innate pulmonary immunity, [18][19][20][21] has never been investigated with regard to the association between its genetic variants and the risk of RP. Previous study showed that serum levels of SP-D in RP patients were elevated. [22][23][24] More specifically, the genetic polymorphisms in SP-D were associated with some lung injuries such as COPD, 25 interstitial pneumonia, 26 asthma, 27 lung cancer, 26 and infectious lung diseases. 28,29 Therefore, we suspected that single-nucleotide polymorphisms (SNPs) of SP-D would also be associated with RP.
In this study, two endpoints of grade ≥2 or grade ≥3 RP were observed and the potential associations of SP-D were explored in lung cancer patients treated with thoracic radiotherapy. To comprehensively investigate this association, the alleles, genotypes and haplotypes of tagSNPs in SP-D were analyzed after clinical-risk RP factors were evaluated.

| Patients
Lung cancer patients receiving thoracic radiotherapy were recruited at the Department of Radiation Oncology at Daping Hospital (Chongqing, China) between February 2006 and March 2011, and were continually recruited at the Cancer Center of Chengdu Military General Hospital (Chengdu, China) between April 2011 and April 2017. The inclusion criteria were (a) histological and cytological confirmation of lung cancer; (b) genomic DNA samples and clinical data availability; (c) patients who did not develop RP during the follow-up period must have received a radiation dose ≥40 Gy; (d) patient records with at least a 6month follow-up period. The exclusion criteria included (a) patients who had previous thoracic irradiation; (b) patients who had severe cardiopulmonary diseases; (c) patients who had blood relatives already enrolled in the study. Written informed consent was obtained from every patient before radiotherapy treatment. Key patient characteristics, including PS, smoking history, lung cancer stage, and chemotherapy, were all recorded. A 1-ml peripheral blood sample was collected from each patient before radiotherapy, and genomic DNA was extracted from the blood samples using Wizard Genomic DNA Purification Kit (Promega, USA). The isolated DNA samples' purity and concentration were determined by spectrophotometric measurement of absorbances at 260 and 280 nm.

| Treatment and follow-up
All patients received radiotherapy with 6-MV X-rays from a linear accelerator (Varian, USA). The dosimetric parameters genetic model (HR, 1.773; 95% CI, 1.166-2.696). Compared to the lung cancer patients bearing the most common haplotype C-G-T, the patients bearing the haplotype T-A-C (rs1998374-rs2255326-rs911887) showed a significant risk of both grade were obtained from the treatment planning system. The median total radiation dose was 55.1 Gy (range: 22 to 72 Gy), with a mean of 2.2 Gy (range: 1.5 to 3 Gy) administered per radiation treatment. 96.2% of the patients (n = 381) received IMRT (intensity-modulated radiation therapy) and others received three-dimensional conformal radiotherapy with total radiation dose ranging from 40 Gy to 70 Gy. The dosimetric parameters for the risk of RP such as the percentage of lung volume receiving greater than 5, 10, 20, 30 Gy (V5, V10, V20, V30) were obtained from the dose-volume histograms (DVH).
During radiation therapy, patients were monitored weekly. They were checked every month from 1 to 3 months after radiotherapy and then every 3 months afterwards during follow-up observations. Follow-up evaluations included interval history, physical examination, chest CT or PET/CT, pulmonary functional tests, and routine blood tests. RP was assessed firstly at each follow-up visit and diagnosed by clinical presentation and radiographic abnormalities, including ground-glass opacity, attenuation, or consolidation changes within the radiation field. Then, RP was graded by two radiation oncologists independently who were blinded from the genotyping results according to the Common Terminology Criteria for Adverse Events version 3.0. Patients with clinical symptoms such as shortness of breath, dry cough, low-grade fever, chest tightness and/or pain were graded >2 RP. The RP patients with oxygen therapy in medical history were graded >3 RP. The time to end-point development was calculated from the beginning of the radiation therapy for each patient to the final follow-up date October, 2017. NSCLC, nonsmall cell carcinoma; PS, performance status of ECOG score standard; SCLC, small cell carcinoma; V10 (%), percentage of the lung volume that received more than 10 Gy; V20 (%), percentage of the lung volume that received more than 20 Gy; V30 (%), percentage of the lung volume that received more than 30 Gy; V5 (%), percentage of the lung volume that received more than 5 Gy. percentage of the lung volume that received more than 10 Gy; V20 (%), percentage of the lung volume that received more than 20 Gy; V30 (%), percentage of the lung volume that received more than 30 Gy; V5 (%), percentage of the lung volume that received more than 5 Gy.

| TagSNPs selection and genotyping
Linkage disequilibrium of SP-D and tagSNPs in SP-D were analyzed by using HaploView 4.0 software (Supplementary Figure S1). A total of eight tagSNPs (rs721917, rs2243639, rs726288, rs1923536, rs1998374, rs911887, rs2255326 and rs75074551), which captured 100% of 36 alleles with MAF >0.1 with a mean r 2 of 0.959, were genotyped by experts who were blinded to the clinical information using improved multiplex ligation detection reaction (iMLDR) method (Genesky Biotechnologies Inc, Shanghai, China). A 5% blind, random sample of study subjects was genotyped twice and the genotype concordance rate was 100% and the call rate was 100%.

| Statistical methods
The Stepwise multivariate Cox proportional hazard regression model was used to select independent risk factors and the probabilities of entry and removal were 0.05 and 0.10. 95% CI, 95% confidence interval; HR, hazard ratio; RP, Radiation Pneumonitis; V10 (%), percentage of the lung volume that received more than 10 Gy; V30 (%), percentage of the lung volume that received more than 30 Gy.

| Univariate and multivariate analysis of tagSNPs association with grade ≥2 RP or grade ≥3 RP
The results of Pearson χ² goodnessof-fit test showed that allele and genotype frequencies of all tagSNPs were in Hardy-Weinbergequilibrium (P > 0.05) in both the patient and the control populations, indicating our samples were a random mating population without selection. With respect to grade ≥2 RP, the associations were statistically significant for rs1998374, rs911887 and rs2255326 in multivariate analysis after adjusting confounding factors and with correction with Q value 30 (

| SNPs and cumulative probability of grade ≥2 or ≥ 3 RP
The overall probabilities of grade ≥2 or grade ≥3 RP were assessed with the Kaplan-Meier curve and log-rank test (Figure1). Patients bearing mutant allele C of rs1998374 showed lower risk for grade ≥2 RP and grade ≥3 RP compared to wild-type allele T. This protect effect was also seen in dominant model ( Figure 1A,D). Patients with mutant allele of rs911887 ( Figure 1B,E) or of rs2255326 ( Figure  1C,F) had significantly higher risk for both grade ≥2 RP and grade ≥3 RP compared to wild-type allele. This risk effect was also seen in homozygous mutant genotype compared to wild-type genotype for grade ≥2 RP ( Figure 1H,I) and in dominant model for grade ≥3 RP ( Figure 1K,L).

| Haplotype association analysis
To further investigate the combined effect of the 3 tag-SNPs (rs1998374, rs911887, and rs2255326) on the risk of RP, we estimated the frequencies of each haplotype by PHASE software and performed haplotype analysis using Cox proportional hazard regression model. There were only 4 haplotypes with over 10% frequency among eight haplotypes in the patients, other rare haplotypes were combined together as "others" in our statistical analysis (

| DISCUSSION
In the present study, three tagSNPs (rs1998374, rs911887, rs2255326) were significantly associated with grade ≥2 RP and haplotype (T-A-C) in SP-D was significantly associated with both grade ≥2 RP and grade ≥3 RP in lung cancer patients treated with radiotherapy. To our knowledge, this is the first report on an association between SP-D and RP. It is worth noting that the eight tagSNPs tested captured 100% of 36 alleles in SP-D with mean r 2 of 0.959, which means 36 SNPs in SP-D were investigated in this study. RP, a complication after radiotherapy, is involved clinical characteristics, therapy-related factors and follow-up tasks. Therefore, the design and measures of RP genetic association study would be more complex than common association studies of disease. 31,32 One critical issue is the confounders. In our study, nongenetic factors including clinical information and radiotherapy-related factors were comprehensively considered before investigation of novel genetic risk factors. In addition to nongenetic factors usually analyzed in other RP genetic association studies, 7,9,11,33,34 tumor location and more precise dosimetric parameters such as V5, V10 and V20 mentioned in some clinic studies as prediction factors for RP 5,[35][36][37] were also analyzed in this study. We found that PS, smoking status, age and dosimetric parameters were significantly associated with RP. Whether chemotherapy was performed or not was regarded as an influencing factor of RP in some studies, 38 but in our study only 2.5% patients received radiotherapy alone and the results did not reach statistical significance (P grade ≥2 = 0.841 and P grade ≥3 = 0.097). Further analysis about different chemotherapy drugs used among patients showed no association with RP risk. Another critical issue is the stringency of RP diagnosis criteria. First, to ensure case group are strictly defined, RP diagnosis and grading should be defined properly and accurately recorded. In our study, 2 radiation oncologists diagnosed and graded RP independently and were blinded to the genetic information. Second, our follow-up schedule was considered to record RP occurrence accurately. Lastly, to reduce the possibility of false non-RP phenotypes in the control group, the follow-up period in our study was extended to a median of 11.4 months, longer than normal occurrence time of RP (<6 months). 6 The allele and genotype analysis in this study consistently revealed that the genetic variants of SP-D were associated F I G U R E 1 Kaplan-Meier curve and log-rank test comparing the cumulative RP-free probability of time to an adverse outcome between patients with alleles and genotypes of rs1998374, rs911887 and rs2255326. K-M allele analysis of rs1998374, rs911887 and rs2255326 for grade ≥2 RP were (A-C) and for grade ≥3 RP were (D-F) respectively. K-M genotype analysis of rs1998374, rs911887 and rs2255326 for grade ≥2 RP were (G-I) and for grade ≥3 RP were (J-L), respectively with RP. The mutant alleles of rs911887 and rs2255326 were risk factors for RP while the mutant allele of rs1998374 was a protect factor for RP. Their different effect for RP roused our interest to further analyze their combined effect for RP using haplotype analysis. Lung cancer patients treated with radiotherapy bearing C-G-T haplotype had about twofold higher risk of RP than those bearing T-A-C haplotype, strongly suggesting that the genetic variants of SP-D could be a genetic biomarker in predicting RP development among lung cancer patients. In our cohort of 396 samples, there was only one patient with a homozygous mutant genotype TT of rs1923536. Although this patient developed high-grade RP, the association between rs1923536 and RP must be reinvestigated in a large sample. Therefore, this tagSNP was excluded from haplotype analysis.
Since RP susceptibility loci could be predictors before radiotherapy, screening RP susceptibility genes has attracted researchers' interest lately. However, maybe for the reason that complex RP follow-up tasks limited the sample size, to explore RP susceptibility loci usually used not genome-wide association study (GWAS) but only candidate gene approach. Until now, RP susceptibility loci were identified in DNA repair-related, inflammation-related, angiogenesis-related and stress response-related pathways with different underlying mechanisms. 32 SP-D, a surfactant proteins known for its contribution to the host's lung immunity, 39 has been mentioned to be biomarker of severe RP after radiotherapy. 40 This study expanded the range of RP candidate genes to a new gene encoded pulmonary surfactant protein D, suggesting another mechanism may underlying the pathogenesis of RP. To further determine the effects of SP-D SNPs and its role in RP, the functional identification of SNPs linked to variable expression levels of SP-D and more biological studies would be performed later.
Although our patients were recruited from two hospitals, studies with larger sample size or more patients from multicenter studies are needed to validate our findings.
Despite its limitations, this study is the first to identify novel RP susceptibility gene SP-D. Three tagSNPs (rs1998374, rs911887, rs2255326) were identified as significantly associated with RP risk in the lung cancer patients treated with thoracic radiotherapy. The findings of our study may be useful in the development of genetic testing for the prediction of RP.

CONFLICT OF INTEREST
None declared. Multivariate analysis adopted Cox regression model with adjustment for V30 (%). c Multivariate analysis adopted Cox regression model with adjustment for age group and V10 (%).