Inherited Polymorphisms in Hyaluronan Synthase 1 Predict Risk of Systemic B-Cell Malignancies but Not of Breast Cancer

Genetic variations in the hyaluronan synthase 1 gene (HAS1) influence HAS1 aberrant splicing. HAS1 is aberrantly spliced in malignant cells from multiple myeloma (MM) and Waldenstrom macroglobulinemia (WM), but not in their counterparts from healthy donors. The presence of aberrant HAS1 splice variants predicts for poor survival in multiple myeloma (MM). We evaluated the influence of inherited HAS1 single nucleotide polymorphisms (SNP) on the risk of having a systemic B cell malignancy in 1414 individuals compromising 832 patients and 582 healthy controls, including familial analysis of an Icelandic kindred. We sequenced HAS1 gene segments from 181 patients with MM, 98 with monoclonal gammopathy of undetermined significance (MGUS), 72 with Waldenstrom macroglobulinemia (WM), 169 with chronic lymphocytic leukemia (CLL), as well as 34 members of a monoclonal gammopathy-prone Icelandic family, 212 age-matched healthy donors and a case-control cohort of 295 breast cancer patients with 353 healthy controls. Three linked single nucleotide polymorphisms (SNP) in HAS1 intron3 are significantly associated with B-cell malignancies (range p = 0.007 to p = 10−5), but not MGUS or breast cancer, and predict risk in a 34 member Icelandic family (p = 0.005, Odds Ratio = 5.8 (OR)), a relatively homogeneous cohort. In contrast, exon3 SNPs were not significantly different among the study groups. Pooled analyses showed a strong association between the linked HAS1 intron3 SNPs and B-cell malignancies (OR = 1.78), but not for sporadic MGUS or for breast cancer (OR<1.0). The minor allele genotypes of HAS1 SNPs are significantly more frequent in MM, WM, CLL and in affected members of a monoclonal gammopathy-prone family than they are in breast cancer, sporadic MGUS or healthy donors. These inherited changes may increase the risk for systemic B-cell malignancies but not for solid tumors.


Introduction
Hyaluronan synthase 1 (HAS1) produces hyaluronan (HA), a polysaccharide with complex biological effects [1]. Here, we evaluated the influence of inherited HAS1 single nucleotide polymorphisms (SNP) on the risk of having a systemic B cell malignancy in 832 patients and 582 controls. Using targeted sequencing of HAS1 SNPs to unequivocally genotype patient populations and to identify potential low penetrance mutations, we reduced the uncertainties associated with array screening and genome-wide association studies. Herein, we show that three HAS1 intron3 SNPs (rs11084110, rs11084109 and rs11669079) are significantly more frequent in patients with systemic B-cell malignancies or in members of a four-generation Icelandic kindred affected by a monoclonal gammopathy-prone phenotype [2;3] than in healthy donors, unaffected Icelandic family members or patients with solid tumors. Overall, the study of 1414 subjects, including 832 patients and 582 healthy controls, suggests that these inherited changes may predispose to the development of systemic B cell malignancies (multiple myeloma, chronic lymphocytic leukemia or Waldenstrom macroglobulinemia), but do not predispose to sporadic monoclonal gammopathy of undetermined significance (MGUS) or breast cancer.
Hyaluronan synthase 1 products, encoded by the HAS1 gene, appear central to the events giving rise to B cell malignancies and to disease progression [1]. HAS1 is aberrantly spliced in malignant cells from multiple myeloma (MM) and Waldenstrom macroglobulinemia (WM), but not in their counterparts from healthy donors [4]. HA produced by HAS1 splice variants is likely to contribute to mitotic abnormalities and to malignant spread/migration [1;4;5], suggesting that aberrant splicing may be an important contributor to systemic B cell malignancies. The presence of the aberrant HAS1Vb intronic splice variant correlated with poor survival in MM patients [4]. In transfectants, a HAS1 variant is oncogenic in vivo and in vitro [6], possibly resulting from induction of chromosomal instability by aberrant intracellular HA and promotion of malignant spread by extracellular HA [1]. Inherited single nucleotide polymorphisms (SNPs) and mutational hotspots in HAS1 exon3 through intron3 influence HAS1 pre-mRNA splicing [7;8]. A major strength of this study is that we have analyzed the HAS1 polymorphisms in five different disease cohorts and two geographically independent populations (western Canadians and the Icelandic kindred) with their appropriate healthy donor control groups. This is the first report that risk for B-cell malignancies and an inherited monoclonal gammopathy-prone phenotype is directly correlated with intronic HAS1 SNPs, likely by promoting aberrant HAS1 splicing.

Study Subjects
Patients with MM, WM or sporadic MGUS were seen at the Cross Cancer Institute and the University of Alberta hospital (Edmonton, AB), and diagnosed according to recommended guidelines [9;10] CLL cells were obtained from the Manitoba CLL Tumor Bank at the Manitoba Institute of Cell Biology. This study was approved by Ethics Review Boards from the University of Alberta, the University of Manitoba, the University of Iceland and Alberta Health Services. Subjects provided written informed consent. Approval for the familial studies was from the Icelandic National Bioethics Committee, license number 8-107-S1, Data Protection Authority: 20096080676. We sequenced DNA from 1,414 individuals (Table 1): this includes a total of 582 healthy control subjects. Controls for B-cell malignancies were 212 anonymous age-matched donors. For breast cancer, 295 anonymous blood samples from patients and 353 matched healthy controls were from the Alberta Research Tumor Bank. For all groups, peripheral blood was taken at the time of diagnosis or at follow-up. We genotyped PB from 34 members from four generations of a monoclonal gammopathy-prone Icelandic family [2;3]. Family members were of both genders; ''controls'' were the unaffected family members. Affected family members were those with MGUS, MM, WM or a hyper-Ig phenotype. Both the Icelandic kindred and the patients from western Canada are predominantly of Caucasian descent.

Genotyping and Analysis
Peripheral blood mononuclear cells were isolated and genomic DNA was purified from 5610 6 PBMCs using QIAamp DNA Blood mini kit (QIAGEN). For two Icelandic samples, DNA was isolated from paraffin-embedded tissue, and amplified with whole genome amplification (WGA). Primers were designed based on the consensus human HAS1 NCBI gene coding sequence (NCBI Reference Sequence: NC_000019.10) (Table S1 in File S1). A HAS1 gene segment from exon3 to intron3 inclusive was amplified in a 50 ml PCR reaction mix (Table S2 in File S1).

Quality Control
To ensure the genotyping results were accurate, we randomly re-sequenced selected DNA samples. For high fidelity Taq DNA polymerase, an enzyme lot was used with an error rate confirmed by us to be 3.0610 24 error/bp. DNA sequencing profiles were analyzed with SeqScape V2.1 for base-calling software/alignment (Applied Biosystems (ABI), CA), confirmed manually by visual inspection of the sequencing profile.

Subcloning and Direct Sequencing
PCR-amplified gene segments were sequenced using two methods: 1) by subcloning in an appropriate vector or 2) by direct sequencing of double or single stranded PCR products. For each patient whose PCR products were subcloned, a minimum of 8 subclones were sequenced in both directions. If a given base was present in all subclones, the two alleles were considered to be homozygous. Heterozygosity was defined as having at least 1/8 subclones with the second allele. 40% of samples were sequenced using plasmid subcloning and 60% were sequenced directly using Big-Dye 3.1 (ABI) on the ABI3130xl DNA analysis system. Obtained sequences were compared to the NCBI HAS1 reference sequence (NC_000019.10).

Statistical Analysis
Hardy-Weinberg equilibrium (HWE) tests were performed for each SNP. We assessed for deviations from HWE using the chisquare test, with 5 degrees of freedom. The SNPs were only considered for further analysis if the allele frequency conformed to HWE expectation. This ensures that any loss of heterozygosity that might be present in malignant cells is excluded from analysis [11]. The difference in genotype and allele frequencies between patients and healthy controls were analyzed by the Fischer exact test: probability values (P values) of ,0.05 were considered statistically significant. The HAS1 alleles designated as ''minor'' by NCBI are the most frequent in the populations studied, with the exception of the Icelandic kindred, but we followed the accepted NCBI dbSNP database convention for naming.
Unconditional logistic regression analysis was performed to obtain odds ratios (OR) and 95% confidence interval (CI) to independently estimate relative risks for each SNP, using Statistical Analysis Software (SAS institute). The independent cohorts were

HAS1 Polymorphisms
HAS1 gene segments from 1414 individuals were sequenced ( Table 1). In the dbSNP database, ten HAS1 SNPs occur in the region sequenced ( Figure 1, Table-S3 in File S1); of these only five met the Hardy-Weinberg Equilibrium criterion of p.0.05; these were two SNPs in exon 3 and three SNPs in intron 3.

HAS1 Genotype and Minor Allele Frequencies in an Icelandic Kindred
We analyzed 34 members of an Icelandic family predisposed to monoclonal gammopathies, including IgM gammopathy, CLL, MM and WM [2;3] ( Table 2). Twelve family members showed increased production of immunoglobulin in vitro, referred to as hyper-responders [3]. The 12 hyper-responders plus two family members diagnosed MGUS, two with WM and one with MM (a total of 17 family members) were termed ''affected'' family members. Controls were 17 unaffected family members who lacked any evidence of monoclonal gammopathies and did not have a hyper-responder phenotype. In the Icelandic kindred, the frequencies of the linked intron3 SNPs (rs11084110, rs11084109 and rs11669079) for affected members were significantly different from those of unaffected members (genotype frequencies p = 0.007 for the intron3 SNPs). In contrast, the exon3 SNP rs11084111 is not significantly different between affected and unaffected family members, providing an internal control. The minor allele frequency for the linked intron3 SNPs was 0.74 for affected and 0.32 for unaffected family members (Table 2), a highly significant difference (p = 0.0005).  Minor Allele Frequencies (MAF) for Study Cohorts The allele frequencies for patients and their respective control groups were determined (Table 3). For HAS1 exon3 SNPs the allele distribution between the cancer groups and their controls are not significantly different. In contrast however, the linked HAS1 intron3 SNPs, the MAF are significantly higher in CLL and WM than they are in the age-matched control group. For all of the intron3 SNPs, the MAF is 0.80 for CLL and 0.81 for WM compared to 0.75 for MM and 0.63 for sporadic MGUS. For MM, sporadic MGUS and the breast cancer group (0.71), the MAFs for the linked HAS1 intron3 SNPs are not significantly different from their respective control groups. For HAS1 exon3 SNPs the allele distribution between the cancer groups and their controls are not significantly different.

Association between HAS1 SNP Genotypes and B-cell Malignancies
Two-tailed unconditional logistic regression was used to determine the effect of HAS1 SNP genotypes on the presumptive risk of B-cell malignancy (Table 4). No assumptions were made about the effect of the SNPs prior to analysis. Odds ratios (OR) measure the impact of a given HAS1 SNP on the risk of having a B-cell malignancy (Table 4). Each cancer group was compared to its control group.
As compared to the healthy controls, for the linked HAS1 intron3 SNPs, the genotypes having the ''minor'' allele were each associated with an increased risk of B-cell malignancy but not of MGUS, perhaps because only 1-2% of sporadic MGUS transform to MM. For the exon 3 SNPs, there was no association between HAS1 SNP genotype and B-cell malignancies or MGUS. The intronic HAS1 SNP rs11084110 was associated with 1.5-fold increase in the risk for MM (OR = 1.53; p = 0.008). The OR was 2.0 for CLL and 2.1 for WM, both of which have a higher familial incidence than does MM. Similar results were observed for the other two intronic SNPs (Table 4). To determine whether or not the linkage between intronic HAS1 SNPs and risk was restricted to B cell malignancies, we also evaluated a cohort of breast cancer patients. There was no difference between breast cancer patients and their matched healthy control group, indicating a lack of association with HAS1 SNPs ( Table 5). Comparison of the B-cell cancer group (MM, CLL and WM) with the breast cancer subjects shows that for all the three linked SNPs, the systemic B cell malignances have a significantly greater association than does breast cancer (p$0.01). Thus, the solid tumor group and its controls provide a negative control for the patient groups harboring systemic B-cell malignancies. Our findings showed that the linked HAS1 intron3 SNPs, had a significant association with all B-cell malignancies studied (Table 4), an association that appears restricted to B-cell malignancies.

Pooled Association Analysis
For increased power to demonstrate a relationship between HAS1 SNPs and the risk of a B-cell malignancy, three case cohorts (MM, CLL and WM) were pooled for analysis ( Figure 2). When MM, CLL and WM were evaluated as a single group in comparison with the healthy control group, we found a significant association between the linked HAS1 intron3 SNPs, and the risk of systemic B-cell malignancy. For the three HAS1 intronic SNPs (OR = 1.78-1.80; p,0.001), the overall risk was 1.78 ( Figure 2, left column of panels) for the group of B cell malignancies. For the Icelandic family members the overall risk was 5.8; no such risk was seen for breast cancer or for the cohort of patients with sporadic MGUS (OR = 0.85 or 0.88, respectively) ( Figure 2, right column of panels).

Discussion
Extensive inherited and acquired mutations characterize HAS1 in patients with B-cell malignancies [8]. HAS1 overexpression has also been reported for some solid tumors, including prostate, ovarian and bladder cancers [1;12-14]. To avoid the uncertainties inherent in genome-wide association studies, we sought susceptibility polymorphisms in a candidate gene, HAS1, based on the biology of the gene in cancer patients. By sequencing in 832 patients and 582 healthy controls (a total of 1,414 individuals) a segment of the HAS1 gene known to influence aberrant HAS1 splicing [7;8], we show that the presence of HAS1 SNPs is strongly associated with the diagnosis of a systemic B-cell malignancy (range of p = 0.005 to p = 10 25 , OR = 1.5 ranging to OR = 2.1, overall risk = 1.78), but not with a solid tumor. For the Icelandic family members the overall risk was 5.8; no such risk was seen for breast cancer or for the cohort of patients with sporadic MGUS that have no known familial predisposition. The observation that HAS1 intronic SNPs do not confer risk to patients with sporadic MGUS likely reflects the fact that only a very small proportion are at risk of transformation to frank malignant disease. It seems likely from the Icelandic familial analysis that if biomarkers were available that preferentially select for only those MGUS that will transform to frank malignancy, a risk association would become detectable. In the Icelandic family kindred, HAS1 SNPs were more frequent in affected members (those with monoclonal gammopathies or hype-responder phenotype) than in unaffected family members (p = .0005), confirming the association with risk in a four generation family having a shared genetic heritage.
The increased risk associated with minor allele genotypes in HAS1 intron 3, as first shown here, is supported by our previous work [1;8;15]. In some blood cancer patients, HAS1 is a hypermutated gene [8] that undergoes aberrant splicing to generate splice variants that correlate with poor survival [4]. In vitro mutagenesis of HAS1 intron3 alters HAS1 splicing patterns, causing transfectants to acquire the splicing pattern seen in MM patients [7;8]. In silico and in vitro mutagenesis analyses predict that HAS1 intron3 plays a central role in clinically relevant splice site selection and consequent aberrant intronic splicing of HAS1 [7;8]. Intronic SNPs can have a functional impact by affecting gene regulation and splicing mechanisms [16][17][18], confirming that intronic SNPs have a potentially strong influence on oncogenic events. In MM and WM, HAS1 genetic variations are distributed in or near splicing elements [8;15] and can direct aberrant splicing of HAS1 [7;8]. In transfected cells, the properties of aberrant HAS1 variants are dominant over those of the normally expressed HAS1 full length form [6].
The independent Icelandic kindred is a small but relatively homogeneous population with a common ancestry. Affected family members exhibit hyper-Ig synthesis in vitro and some have MGUS, WM or MM. The linked HAS1 intron3 SNP alleles and genotypes of affected members, but not an exon3 SNP, were significantly different from those of unaffected members (genotype frequencies p = 0.007, allele frequencies p = 0.0005).
In the pooled analysis of SNP genotypes for all patient and control groups, the odds ratio for B cell malignancy (OR trend ) was 1.78 and for affected Icelandic family members, OR was 5.8. These odds ratios, especially when considered in the context of the familial analysis, suggest minimal or no influence of cryptic population stratification. Risk predictions may be directly correlated with the identified SNPs, or may result from unknown  Table 4. HASI SNP genotype associates with risk of B-cell malignancy*.    Family members of patients with MGUS or MM have a twothree-fold higher risk of developing MGUS and multiple myeloma [19][20][21][22]. CLL and WM include both sporadic and familial forms of the disease [23][24][25][26]. Inherited polymorphisms are associated with both CLL and WM [15;25;27-31], and a strongly recurrent somatic mutation with WM [32]. The odds ratios found here for CLL and WM were considerably more pronounced than those found for MM which appears to have less familial influence. However, this work demonstrates for the first time that an increased frequency of HAS1 intron3 SNPs is common among all three systemic B-cell malignancies, possibly indicating shared genetic predispositions that culminate in malignant transformation of B-cells.
We employed sequencing for this study because it provides for unequivocal allele calling in each patient, as well as identification The left column of three panels show the odds ratios of B cell malignancy for the three linked intronic HAS1 SNPs. The OR denotes the odds of having the minor allele as assessed by genotype. The X-axis corresponds to odds ratio. The horizontal line represents 95% confidence interval. Each box represents the OR point estimate. The diamond represents the odds ratios obtained from pooled analysis with 95% confidence interval. Each group is compared to the appropriate control group. The right column of three panels show odds ratios for the Icelandic kindred, MGUS and breast cancer; each individual group is compared to its own control group as indicated in results. The cumulative risks were not estimated for these last three groups as they are unrelated to each other. These plots show that there were no associations of risk with the three linked SNPs for sporadic MGUS or breast cancer. There was a very strong association of the minor allele with risk for the B cell malignancies (pooled OR = 1.78) and the Icelandic Kindred (OR = 5.8). Note difference in scale values between panels in the two columns. doi:10.1371/journal.pone.0100691.g002 of any closely linked rare mutations that might influence risk. An advantage in SNP detection by direct DNA sequencing is the complete information it yields, including haplotype relationships. Recent literature supports our approach [33] and emphasizes the use of sequencing to detect direct associations between disease and casual SNPs [34]. Our analysis was restricted to polymorphisms in HAS1 exon3 and intron3, a mutation-rich region that is involved in aberrant splicing of HAS1 exons and introns [7;8]. Our work suggests that the three linked HAS1 intronic SNPs may predispose to aberrant splicing, or are in linkage disequilibrium with a causative functional variant located outside exon3 and intron3 of HAS1. This study provides strong evidence to support the hypothesis that HAS1 contributes to a genetic risk of B-cell malignancy.
In conclusion, our findings suggest that genetic variants in HAS1 are associated with risk of B-cell malignancy in MM, CLL and WM. Because the HAS1 intron3 SNPs are present at significantly higher frequencies in individuals who have developed a systemic B-cell cancer, these results suggest that intronic HAS1 SNPs predispose to systemic B-cell cancers and as inherited characteristics may thus act at an early stage of oncogenesis. Although multiple genes are certainly involved, this work supports the idea that HAS1 intron3 SNPs have a strong impact on molecular events, particularly on aberrant HAS1 pre-mRNA splicing events that contribute to the malignant phenotype in B-cell cancers. Molecular therapy to target this region of HAS1 warrants evaluation.

Supporting Information
File S1 Suppporting Tables. Table 1, Primer sequences used for PCR or sequencing reactions.