A functional variant at the miRNA binding site in HMGB1 gene is associated with risk of oral squamous cell carcinoma

Oral squamous cell carcinoma (OSCC) is a common malignancy that has been causally associated with both hereditary and acquired factors. The high mobility group box 1 (HMGB1) gene plays an important role as a DNA chaperone to help maintain nuclear homeostasis. Altered expression of HMGB1 has been implicated in a wide range of pathological processes, including inflammation and cancer. The present study explores the impact of HMGB1 gene polymorphisms, combined with environmental risks regarding susceptibility to oral tumorigenesis. Four single-nucleotide polymorphisms (SNPs) of the HMGB1 gene, rs1412125, rs2249825, rs1045411, and rs1360485, were evaluated in 1,200 normal controls and 772 patients with OSCC. We found an association between the wild-type allele of rs1045411 and genotypes CT and CT/TT (AOR=0.754, 95% CI=0.582-0.978 and AOR=0.778, 95% CI=0.609-0.995, respectively). Additionally, bioinformatics analysis was used to characterize the functional relevance of these variants for the miRNA-505-5p binding site and transcriptional regulation by the HMGB1 3’-UTR and promoter regions. Moreover, in considering behavioral exposure to environmental carcinogens, the presence of the four HMGB1 SNPs, combined with/without betel quid chewing and smoking showed, profoundly synergistic effects on the risk of OSCC. In conclusion, we present a potential clinical relevance for HMGB1 variants in OSCC, as well as associations between HMGB1 polymorphisms, haplotypes and environmental risk factors. The finding may help in development of optimal therapeutic approaches for OSCC patients.


INTRODUCTION
Oral squamous cell carcinoma (OSCC) is the sixth most common malignancy worldwide, with 599,000 new cases and 325,000 cancer deaths in 2012 [1,2]. OSCC, which accounts for more than 90% of all mouth malignancies and approximately 40% of head and neck tumours has, a high potential for local invasion and lymph node metastasis. Indeed, the estimated 5-year overall survival rate of less than 50% has not improved significantly over the last three decades [3]. OSCC often causes postsurgical dysfunctions in chewing, swallowing, and speech, as well as aesthetics loss [4]. Moreover, although early-stage OSCCs has a high treatment success rate approximately 70% of patients with progressive disease cannot be successfully treated due to relatively high local and regional recurrence rates [5]. Therefore, early detection and clarification of the detailed molecular mechanism of OSCC are urgently needed [6,7].
The HMGB1 gene belongs to the high-mobility group protein family, and the protein product contains two 80-amino acid DNA-binding domains (A-box and B-box) and an negatively charged C-terminus [8]. The gene sequence is located on the long arm of chromosome 13; the transcriptional region spans approximately 6,000bp, with a promoter region of at least 1,700-bp, and 5 exons of approximately 2,600-bp [8]. HMGB1 functions in the nucleus as a chromatin structural protein and extracellularly as a pro-inflammatory cytokine [9,10]. Nuclear HMGB1 is a non-histone DNA-binding protein, which suggests that it facilitates the assembly of sitespecific DNA targets by acting as a DNA chaperone [11]. In contrast, extracellular HMGB1 functions as a damageassociated molecular pattern that propagates infection-or injury-elicited inflammatory responses [12]. Moreover, evidences confirm that HMGB1 over-expression is closely related to tumour development through functions in proliferation, invasion and migration of cancer cells [13][14][15][16]. Therefore, HMGB1 may serve a biomarker of inflammation and/or a prognostic marker for OSCC progression [9,17,18].
Although the development of OSCC may take several decades years, early detection of this cancer is seldom achieved due to the lack of reliable markers [19]. Therefore, the disease runs a largely asymptomatic course until it becomes too advanced for successful treatment [20,21]. Aberrations in some genes may be responsible for certain clinical features of OSCC [22]. For instance, differences in the level of HMGB1 expression have been demonstrated between precancerous and malignant lesions [23][24][25] yet little is known regarding the joint effects of HMGB1 gene variants and behavioural exposure to cancercausing substances on predisposition to OSCC. Moreover, previous studies have reported the effect of HMGB1 gene polymorphisms on human cancer susceptibility, and the polymorphisms may efficiently predict the risk of cancers [26][27][28][29][30]. Thus, we hypothesized that four polymorphisms (rs1412125, rs2249825, rs1045411, and rs1360485; Table 1) in HMGB1 are associated with susceptibility to OSCC. The aim of this study was to examine the potential clinical relevance of four SNPs in the HMGB1 gene in patients with OSCC, as well as associations between HMGB1 polymorphisms, haplotypes and environmental risk factors.

Study population
A total of 1,972 participants, including 772 OSCC cases and 1,200 controls were recruited to explore the effects of the HMGB1 gene on OSCC risk. Demographic characteristics including mean age, betel quid chewing, cigarette smoking and alcohol drinking are shown in Table  2. We discovered significant differences when grouping by betel quid chewing (P<0.001), cigarette smoking (P<0.001) and alcohol drinking (P<0.001) in the healthy controls and OSCC patients ( Table 2). Table 3 summarizes the basic characteristics of the HMGB1 SNPs in the study population. In both the OSCC patients and healthy control subjects, genotypes T/T, C/C, C/C, and T/T exhibited the highest frequencies for rs1412125, rs2249825, rs1045411 and rs1360485, respectively. In these controls, the genotypic frequency of HMGB1 SNP rs1412125 met the Hardy-Weinberg equilibrium (P=0.282, χ2 value: 1.159). The frequencies of HMGB1 SNPs rs2249825, rs1045411 and rs1360485 were also in the Hardy-Weinberg equilibrium (P=0.678, χ2 value: 0.172; P=0.451, χ2 value: 0.569; and P=0.537, χ2 value: 0.382, respectively). According to the adjusted odds ratios (AORs) and 95% confidence intervals (CIs) in a multiple logistic regression model for HMGB1 polymorphism and OSCC, compared with their corresponding wild-type homozygotes (C/C), only rs1045411 C/T or C/T+T/T presented a significant (P<0.05) protective role after adjusting confounding factors: 0.754-fold (95% CI, 0.582-0.978) and 0.778-fold (95% CI, 0.609-0.995), respectively (Table 3).

Synergistic effects of genetic variants and betel quid chewing behavior with smoking
The habit of betel quid chewing widespread in Taiwan, and chewers are often smokers. Thus, we analyzed the synergistic effects of cigarette smoking and betel chewing combined with the four SNPs in the OSCC patients. Among the 1,322 smokers in our study, the estimated AORs of the four selected HMGB1 polymorphisms were elevated in individuals who were both betel quid chewers and smokers (10.071-10.659fold) compared to abstainers (Table 4). However, the risk of OSCC tended to decline among those who had quit betel quid chewing, suggesting that the four selected HMGB1 variants are associated with betel quid chewing and OSCC susceptibility in cigarette smokers (Table 4).

Haplotypes analysis for the HMGB1 gene
HMGB1 polymorphisms were further characterized using linkage disequilibrium (LD) and haplotype analyses. LD was determined pairwise among all 4 SNPs and the haplotype structure of the HMGB1 gene was analyzed (D' and r 2 ) according to 1000 Genomes Project data [37] for the East Asian population (CHB+JPT, Figure 2). Haplotype blocks divided by the D' confidence interval method with a D' value of 95% CI 0.70-0.98 in adjacent SNPs was classified as the same haplotype block. One LD block was detected by Solid Spine [38], a haplotype phasing technique. Block1 (3 kb) consisted of two closely selective SNPs showing strong linkage, rs1045411 and rs1360485, in the 3'-UTR of HMGB1 ( Figure 2). Furthermore, a haplotype-based association study was performed to show association between HMGB1 haplotypes and OSCC risk ( Table 5). Block1 of the 3'-UTR SNPs constituted virtually four haplotypes of approximately equal frequencies in the control subjects (75.0%, 22.5%, 2.3% and 0.1%). In contrast, one haplotype in the OSCC cases, "C-C", was associated with increased susceptibility to OSCC (AOR=1.808; 95% CI, 1.257-2.601, P=0.001) compared with the most common "C-T" haplotype ( Table 5).

The effects of rs1412125 and rs2249825 on HMGB1 transcriptional regulation
As a preliminary assessment of the putative functional role of these SNPs, we investigated whether rs1412125 and rs2249825 are associated with differential expression of HMGB1, According to the Genome-Tissue Expression (GTEx) database [39], statistically significant down-regulation of HMGB1 mRNA expression with rs2249825 variant genotype (C/G+G/G) compared with the wild-type homozygous genotype (C/C, P=0.026, Figure 3A) is observed in whole blood. Moreover, OSCC-risk-associated The odds ratio (OR) with their 95% confidence intervals were estimated bylogistic regression models. ‡ The adjusted odds ratio (AOR) with their 95% confidence intervals were estimated by multiple logistic regression models after controlling for betel quid chewing, alcohol and tobacco consumption. environmental factors were further explored by examining functional annotation in the Encyclopedia of DNA elements (ENCODE) data [40] for sites in the 3,500-bp promoter and intron 1 sequences ( Figure 3B and 3C). We determined that rs1412125 and rs2249825 are situated at a site for transcription factor binding, with histone modification patterns and DNA hyposensitivity characterized in the promoter or enhancers in several cell type ( Figure 3C). The risk allele [T] of rs1412125 in located in the core of a CCAAT/Enhancer box [41] (Figure 3D), which is one of the most common elements in eukaryotic promoter. The effect of rs2249825 may be attributed to the suboptimal avian myeloblastosis viral oncogene homolog-like 2 (v-Myb) binding site [42] (Figure 3E), with the consensus motif of (C/T)AA(G/T)G surrounding intron 1 of the predicted transcriptional start site of human HMGB1 gene ( Figure 3B), which enables modulating of initiation rates in response to transcriptional status. Accordingly, these promoter polymorphisms in the CCAAT/Enhancer box and v-Myb binding site might collaborate in the regulation of HMGB1 expression.  10.659 (7.073-16.065) † The odds ratio (OR) with their 95% confidence intervals were estimated bylogistic regression models. ‡ The adjusted odds ratio (AOR) with their 95% confidence intervals were estimated by multiple logistic regression models after controlling for betel quid chewing, alcohol and tobacco consumption. a. Individuals are homozygous for ancestral allele but without betel quid chewing. b. Individuals with either at least one polymorphic allele or betel quid chewing. c. Individuals with both at least one polymorphic allele and betel quid chewing. www.impactjournals.com/oncotarget DISCUSSION Cancer is the result of a series of genetic or epigenetic regulations, and environmental risk factors may contribute to the development and progression of tumors [21,43]. Thus, in-depth understanding of the mechanisms underlying carcinogenesis is needed to characterize genetic alterations linked to OSCC development [44]. Such information may provide relevant strategies for prevention, for the surveillance of patients at risk, and for early detection to reduce morbidity and mortality from OSCC [45]. Several studies have suggested that genetic alternations harbored on chromosome 13q12 are involved in multiple tumor types [46][47][48][49][50] and this locus is highly susceptibility to genomic instability in OSCC [51]. Recent studies have reported that HMGB1 polymorphism is associated with susceptibility to several carcinomas including breast cancer [52], melanoma [53], gastric cancer [54] and colorectal cancer [55]. Loss of HMGB1 increase DNA damage due to decreases in DNA chaperone efficiency in response chemotherapy, cell stress, and environmental risk factors [56]. Our data revealed an increased risk of OSCC among patients with the HMGB1 polymorphism rs1045411 C/C compared with those with the C/T or C/T+T/T genotype (Table 3). We present additional evidence for a role for HMGB1 in OSCC by showing a significant interaction between betel quid chewing and at least one polymorphic allele of four HMGB1 SNPs in individuals, higher incidence of OSCC with smoking behavior (Table 4).   The present study confirms the association of the clinically examined rs1045411 and rs1360485 haplotype in the HMGB1 3'-UTR region with OSCC risk most likely because a putative miRNA-505 binding site (Figure 1). We confirmed that these two HMGB1 3'-UTR variants are synergistically involved in the etiopathogenesis of OSCC in a haplotype-specific fashion rather than as single genetic variants (Table 5). Therefore, an SNP haplotype corresponding to the RNA bulge region may effects the binding strength of specific miRNA/target duplexes, resulting in low minimum free energy and modulating mRNA stability (Figure 1). Although directly testing this hypothesis is beyond the scope of the current study, evidence suggests that mRNA stability is associated with susceptibility to OSCC, with HMGB1 downregulation being associated with an increased susceptibility to OSCC. Previous studies demonstrate that miRNA-505 acts as tumor suppressor in endometrial carcinoma [57]. Matamala et al. also reported that circulating miRNA-505 was significantly overexpressed in breast cancer patients [58]. Moreover, miRNA-505 suppresses proliferation and invasion in hepatoma cells by directly targeting HMGB1 has been previously reported [59]. Hence, the present analyses increased our understanding of naturally occurring HMGB1 variants, a lesion category that, while not infrequent, has been relatively neglected in terms of elucidation of the underlying pathogenic mechanisms. HMGB1 expression is considered as a novel and independent predictor of decreased disease-free survival of patients with prostate cancer [25,60], and it may also be used as a novel and independent predictor of prognosis for OSCC patients.
Converging lines of evidence support that OSCC development is a complicated process regulated by both environmental and genetic factors. In the present study, we estimated the magnitude of the statistical interaction between two measurable exposures with respect to a given outcome with HMGB1 genetic variants and found that the use of betel quid with smoking conferred a higher susceptibility to OSCC (Table 4). Strikingly, polymorphisms altering DNA chaperone activity, which maintains nuclear homeostasis, may lead to synergistic effects in areca nut carcinogen-induced OSCC risk [61]. However, the effects of environmental risks on predisposition to OSCC may be underestimated because of the inability to quantitatively access the degree of betel quid chewing and DNA chaperone gene variations with smoking. Nevertheless, this study provides comprehensive evidence for the association of several genetic variation in the DNA-chaperon HMGB1 gene and multiple environmental exposures in OSCC development [62]. However, there are several limitations in our investigation. One of the limitations is that information on exposures of carcinogens is dichotomized into ''ever-user'' versus ''never-user.'' More detailed analysis based on amount on exposures of carcinogens should be performed. Furthermore, the molecular functional role in the HMGB1 3'UTR mRNA with a miRNA-505-binding site of OSCC requires further investigation.
To our knowledge, there are no previous studies of the effects of gene-environment interactions on DNA chaperon gene polymorphisms and betel-chewing habits in oral tumorigenesis. Nonetheless, the coeffects of genetic and environmental factors interact to facilitate OSCC development. Overall, we present a possible role for HMGB1 variants and provide deeper insights into naturally occurring haplotype-based variants. Furthermore, characterizing the molecular basis of polymorphisms in cancer provides insight into tumorigenesis. Accurate biomarkers for such types of variants are required for developing optimal therapeutic approaches that will eventually ameliorate the clinical phenotype in patients harboring the corresponding lesions.

Description of the participants
This study recruited 772 male OSCC patients between 2008 and 2015 at Chung Shan Medical University Hospital, Taiwan. Clinically information including TNM staging, primary tumor size, lymph node involvement, and histologic grade was examined according to American Joint Committee on Cancer (AJCC) classification [63,64]. Subjects with neither self-reported history of cancer of any site nor oral precancerous disease such as oral submucous fibrosis, leukoplakia, erythroplakia, or verrucous hyperplasia were selected to the control group from Taiwan Biobank. Before the study began, approval was obtained from the Institutional Review Board of Chung Shan Medical University Hospital, and informed written consent was obtained from each individual.

SNPs selection and genotyping
Genomic DNA was isolated from peripheral blood using the QIAamp DNA blood mini kit as described in detail previously [65,66]. The final preparation was stored at -20°C, quantified by optical density measurement at 260 nm and used as the template for polymerase chain reaction (PCR). Genotyping of four HMGB1 SNPs (rs1412125, rs2249825, rs1045411, and rs1360485; Table 1) with minor allele frequencies >5% in the HapMap Chinese Han Beijing (CHB) population was performed using the TaqMan SNP genotyping assay (Applied Biosystems, Foster City, CA, USA) [67]. Allele frequencies were determined by the ABI SDS software. www.impactjournals.com/oncotarget

Bioinformatics analysis
Several semi-automated bioinformatics tools were applied to assess whether SNPs or their linked genetic variants are associated with a putative function that might affect patient outcomes. HaploReg [68] v4 and the GTEx database [39] from the ENCODE project [69] were used to identify the regulatory potential of candidate functional variants. Particular sites of interest were examined including transcription factor (TF)-ChIP signals, DNase peaks, DNase footprints and predicted DNA sequence motifs for TFs. The GTEx data were used to identify correlations between SNPs and whole blood-specific gene expression levels. The publically available cBioPortal for Cancer Genomics [70] and UCSC Cancer Genomics Browser [71] for OSCC were utilized to analyse HMGB1 gene expression, enhancer modulation, molecular features, and clinical outcomes.

Characteristics of miRNA candidates
We predicted targets using the web-based tools RNAhybrid on BiBiServ2 (http://bibiserv.techfak.unibielefeld.de/rnahybrid). RNAhybrid [72] determines the most energetically favorable hybridization patterns using the MFE of two RNA fragments with different lengths, i.e. long (3'-UTR of HMGB1) and short (mature miRNA sequences). The parameters used in the analysis were as follows: number of hits per target, 3 nucleotides, maximum mismatch size, 1 nucleotide, overhangs 2 nucleotides. The MFE considered for each miRNA/target duplex was higher than -15 kcal/mole as assessed for a perfect match between the mature miRNA and its target.

Statistical analysis
The Mann-Whitney U test and Fisher's exact test were used to compare differences in the distribution of age and demographic characteristics between the controls and OSCC patients. ORs with 95% CIs were estimated using logistic regression models. AORs with 95% CIs were used to assess association between genotype frequencies with OSCC risk and clinical factors. P values less than 0.05 were considered significant. The data were analysed with SPSS 12.0 statistical software (SPSS Inc., Chicago, IL, USA).

Author contributions
YFL and CWL conceived and designed the experiments; SFY, CMY and YEC performed the experiments; YFL and SFY analyzed the data; SFY and CYC contributed samples; YFL and CWL wrote the paper.