Variations in Chromosomes 9 and 6p21.3 with Risk of Non–Hodgkin Lymphoma

There is growing evidence that common genetic variants play an important role in non–Hodgkin lymphoma (NHL) etiology. Large consortial efforts and genome-wide association studies (GWAS) have implicated a number of immune-related genes with NHL risk. Specifically, polymorphisms in interleukin 10 (IL10) and within several genes located in the 6p21.3 chromosomal region are associated with NHL risk, including the tumor necrosis factor (TNF) and human leukocyte antigen (HLA) genes (1–3). In our multicenter study of NHL in the United States, we have further implicated specific HLA Class I and II alleles with NHL risk, most notably the HLA DRB1*0101 allele with follicular lymphoma, a finding that is consistent with recently published GWAS data (3;4). We have also implicated genes in other pathways including the nuclear factor kappa beta (NF-kB) pathway (5), cell cycle (6), DNA repair (7), and the family of caspase genes (8).

To complement previous a priori and ongoing agnostic approaches to identifying susceptibility genes in NHL etiology, we conducted a detailed literature search to identify additional potential genes of relevance for lymphomagenesis. We included genes implicated from laboratory and clinical research, particularly those based on translational efforts such as gene expression and clinical response data. We present data from a U.S. population-based case–control study of NHL where genotyping for nearly 488 genes (6,679 SNPs) were conducted. Genes and SNPs evaluated are annotated in Supplementary Materials (Supplementary Table S1).

The National Cancer Institute-Surveillance, Epidemiology and End Results (NCI-SEER) NHL case–control study has been previously described (9;10). Briefly, the study included 1,321 incident NHL cases without evidence of HIV infection identified in 4 SEER registries (Iowa; Detroit, MI; Los Angeles, CA; Seattle, WA) aged 20 to 74 years. 1,057 population controls were identified by random digit dialing (<65 years) and from Medicare eligibility files (≥65 years). Overall participation rates were 76% in cases and 52% in controls; overall response rates were 59% and 44%, respectively. Written informed consent was obtained from each participant prior to interview.

All cases were histologically confirmed by the local diagnosing pathologist. NHL subtypes were originally coded according to the International Classification of Diseases for Oncology, 2nd Edition (11) and then updated to the ICD-O-3/World Health Organization classification. In addition to NHL overall, we evaluated 4 B-cell subtypes: diffuse large B-cell lymphoma (DLBCL), follicular lymphoma, marginal zone lymphoma, and chronic lymphocytic leukemia and small lymphocytic lymphoma (CLL/SLL).

Study participants who did not provide a biological specimen, did not have sufficient material for DNA extraction or sufficient DNA for genotyping, or whose genotyped sex was discordant from the questionnaire data were excluded from this analysis. Of the 1,231 cases (820 blood, 411 buccal cell) and 992 controls (692 blood, 300 buccal cell) with biospecimens, 963 cases and 837 controls were genotyped. As previously described, DNA was extracted from blood clots or buffy coats (BBI Biotech) using Puregene Autopure DNA extraction kits (Gentra Systems), and from buccal cell samples by phenol–chloroform extraction methods (12). Genotype frequencies for individuals who provided blood compared with buccal cells were equivalent (13).

Genotyping of tag SNPs from 488 candidate gene regions (Supplementary Table S1) hypothesized to be involved in lymphomagenesis was conducted at the NCI Core Genotyping Facility (Advanced Technology Center; ref. 14) using a custom-designed Infinium assay (Illumina). The Infinium included a total of 7,943 tag SNPs. Tag SNPs were chosen from the designable set of common SNPs [minor allele frequency (MAF) > 5%] genotyped in the Caucasian (CEU) population sample of the HapMap Project (Data Release 20/Phase II, NCBI Build 36.1 assembly, dbSNPb126) using the software Tagzilla, which implements a tagging algorithm based on the pairwise binning method of Carlson et al. (15). For each original target gene, SNPs within the region spanning 20-kb 5′ of the start of transcription (exon 1) to 10-kb 3′ of the end of the last exon were grouped using a binning threshold of r² > 0.8 to define a gene region. When there were multiple transcripts available for genes, only the primary transcript was assessed.

In the current manuscript, we exclude data from 22 genes (5% of total SNPs) for presentation due to prior commitments to ongoing pooling efforts within the InterLymph Consortium. These exclusions do not alter the main results, discussion or conclusion based on our analysis.

Tag SNPs that failed manufacturing (ordered but did not convert), failed validation (no amplification or clustering) and assays that had less than 80% completion or 80% concordance with the 90 Hapmap CEU samples used for validation were excluded, resulting in a subset of 6,830 tag SNPs with data for further analysis. Of the 6,830 tag SNPs, SNPs with low completion rate (<90% of samples) were excluded, including 53 SNPs among samples with DNA extracted from blood and 82 SNPs among samples with DNA extracted from buccal cells. Replicate samples (n = 62) from 2 blood donors each and duplicate samples from 93 participants processed in an identical fashion were interspersed for all assays and blinded from the laboratory. For each plate of 368 samples, genotype-specific quality control (QC) samples were also included and comprised 4 each: homozygote wild-type (WT), heterozygote, homozygote variant, and DNA-negative controls. SNPs with concordance less than 95% in QC samples were excluded, including 1 SNP from samples with DNA extracted from blood and 13 SNPs from samples with DNA extracted from buccal cells. Concordance for all remaining QC replicates and duplicates was 99% or more for all assays. We further excluded samples with a low completion rate (<90%; n = 27).

Hardy–Weinberg equilibrium was evaluated among non-Hispanic Caucasian controls. SNPs showing evidence of deviation from Hardy–Weinberg proportions (P < 0.0001) are denoted in Supplementary Materials (Supplementary Table S1).

The final analytic study population included 947 cases and 826 controls with data for 6,679 SNPs.

We obtained a gene region-level summary of association by computing the minP test, which assesses the statistical significance of the smallest P-trend within each gene region (determined by dichotomous logistic regression, comparing NHL or NHL subtypes with controls) by permutation-based resampling methods (10,000 permutations) that automatically adjust for the number of tag SNPs tested within that gene and the underlying linkage disequilibrium pattern (16;17). To account for multiple comparisons within the 488 candidate gene regions, we applied the false discovery rate (FDR) method of Benjamini and Hochberg (18) to the minP test separately for NHL and each subtype.

We calculated odds ratios (OR) and 95% CI estimating the relative risk of NHL and NHL subtypes in relation to each genotype using dichotomous and polytomous unconditional logistic regression models, respectively. The homozygote of the most common allele in the pooled study population was used as the referent group. Tests for trend under the codominant model used a 3-level ordinal variable for each SNP (0 = homozygote common, 1 = heterozygote, 2 = homozygote variant). All models were adjusted for age, race/ethnicity, sex, and study center (categories listed in Supplementary Table S2). We also conducted all analyses restricted to non-Hispanic Caucasians. We assessed heterogeneity among NHL subtypes in the polytomous multivariate unconditional logistic regression models using the Wald chi-square statistic. Analyses were conducted using SAS version 9.1 (SAS Institute).

We conducted haplotype analyses among non-Hispanic Caucasians using 2 methods. First, we evaluated risk of NHL and NHL subtypes associated with haplotypes defined by SNPs within a sliding window of 3 loci across a gene (Haplo Stats, version 1.2.1). A global score statistic was used to summarize the evidence of association of disease with the haplotypes for each window. Second, we visualized haplotype structures using Haploview, version 3.11 (19) based on measures of pairwise linkage disequilibrium between SNPs. For blocks of linkage disequilibrium, we obtained ORs and 95% CIs for the underlying haplotypes under the assumption of an additive model (haplo.glm, minimum haplotype frequency 1%). All haplotype analyses were adjusted for age, sex, and study center.

Descriptive characteristics of the 947 cases and 826 controls are provided in Supplementary Table S2. The distributions of cases and controls were similar with respect to sex, age, and race/ethnicity, with our population largely comprising non-Hispanic Caucasians. The 2 predominant NHL subtypes were DLBCL (30%) and follicular lymphoma (25%).

There were 5 SYK SNPs statistically significant with a P-trend less than 0.001 for all NHL, including the most significant SNP observed in our overall analysis (rs2991216, P-trend = 0.00005; Table 1). The associations were consistently in the same direction across the 4 subtypes evaluated, though most significantly associated with DLBCL. On the basis of the haplotype structure, statistically significant SNPs for SYK fall within 2 larger blocks. The first block includes SYK SNPs rs2035072 and rs290213 (D′ = 0.93) and the second block includes rs2991216, rs965892, and rs290203 (D' = 0.88–0.99). Linkage between the SNPs in the 2 blocks was low (D′ = 0.24–0.36), suggesting that 2 independent regions within SYK may be associated with NHL etiology. Haplotype analyses by hapwalk of 3 SNPs yielded the same 2 regions of high significance for NHL risk. Haplotype analyses of the blocks as defined by Gabriel et al. did not reveal any further regions of significance as all haplotypes associated with NHL included the variant allele for rs2035072 (data not shown).

Corresponding SNPs for the 3 gene regions in chromosome 6p21.3 were statistically significantly associated with NHL at P < 0.001, including SNPs in RING1/RXRB (rs2855429, rs213213), AIF1 (rs2857597), and BAT4 (rs3115667; Table 1). By subtype, all 4 SNP associations were significant in both DLBCL and follicular lymphoma (Table 1). The AIF1 and BAT4 SNPs were also significantly associated with marginal zone lymphoma, but the 2 SNPs in RING1/RXRB were not. All associations remained statistically significant in models adjusted for the implicated TNF -308A and HLA DRB1*01010 alleles (RING1/RXRB (rs2855429 TNF/HLA-adjusted P-trend = 0.0003, rs213213 TNF/HLA-adjusted P-trend = 0.001), AIF1 (rs2857597 TNF/HLA-adjusted P-trend = 0.02), and BAT4 (rs3115667 TNF/HLA-adjusted P-trend = 0.01)). We note, however, that power was limited in these adjusted models because they are restricted to the two-third of samples with data for the HLA allele. In general, risk estimates for each SNP were modest (e.g., less than a 2-fold increase or decrease in risk; Supplementary Table S3).

We plotted the P-trends for the 288 SNPs from the 27 gene regions that were evaluated in the 6p21.3 region for this analysis with NHL (Fig. 1) and NHL subtypes (Supplementary Figure S1). For NHL and DLBCL, we show that the statistical significance of all 3 SNPs is greater than that for the confirmed TNF G-308A polymorphism (rs1800629) in our data. Hapwalk analysis of 3-SNPs yielded the same regions of significance in AIF1, BAT4, and RING1/RXRB for NHL (Fig. 1). Importantly, the haplotype structure reveals that the implicated regions are in distinct haplotypic blocks from the TNF/LTA region. The haplotype structure further shows the RING1/RXRB region to be distinct from the AIF1 and BAT4 SNPs, which is consistent with our observation that SNPs in AIF1 and BAT4 are potentially distinct from RING1/RXRB based on their different subtype-specific patterns.

Haplotype analyses of blocks as defined by Gabriel et al. did not yield any further information; no haplotypes were more significant than the individual SNP associations (Fig. 1).

Our SNP-based results are supported by our evaluation of gene region-based analyses. Of the 488 gene regions evaluated, 14 were significant with a P < 0.01 for all NHL. The 3 top gene regions of significance (RING/RXRB; AIF1; and BAT4) are located on chromosome 6p21.3 (Table 2) and had FDR-adjusted P values 0.25. By subtype, the RING1/RXRB region was significantly associated with follicular lymphoma, the BAT4 region with DLBCL, and the AIF1 region with both DLBCL and follicular lymphoma. The gene region encompassing SYK was also significantly associated with NHL (P = 0.004, FDR P = 0.27), and also with both DLBCL and follicular lymphoma (P = 0.01). Results for all gene regions evaluated are shown in Supplementary Table S4.

Other regions of potential importance include genes on chromosome 2 (BCL11A, UGT1A8), chromosome 12 (KRAS), and chromosome 21 (ITGB2) where SNPs and gene regions were both statistically significant at P-trend < 0.001 and permutation P < 0.01, respectively, for NHL overall (Table 1 and 2). In subtype-specific results, the association between PGR and DLBCL was also noteworthy (P-trend < 0.0001; permutation P < 0.001).

In our evaluation of a priori candidate genetic polymorphisms in NHL etiology, the SYK SNPs comprised the most statistically significant SNP of all variants evaluated and were associated with NHL with P-trends < 0.0001. Our results require replication but if confirmed, are of interest for several reasons. SYK is important in B-cell development and recent evidence supports its potential as a therapeutic target (20–22). SYK is a protein-tyrosine kinase gene encoded on chromosome 9 that is expressed in hematopoietic cells and predominantly in the lymphoid tissues of the spleen and thymus. Syk is activated by the B-cell receptor and mediates a variety of critical cellular responses. Syk expression is necessary for the survival of human NHL cell lines, and its disruption leads to cellular apoptosis in NHL mouse models (21). Friedberg and colleagues demonstrated that an oral Syk inhibitor provided to patients with B-cell NHL improved progression-free survival, particularly for CLL/SLL (20). If our results are replicated, determining the functional consequences of SNPs in SYK is warranted.

Our results also provide further evidence that variation among additional genes in the 6p21.3 chromosomal region is important for NHL etiology across multiple NHL subtypes. Notably, the SNPs within the implicated gene regions — RING1/RXRB, AIF1, and BAT4 — are neither correlated nor in linkage disequilibrium with the confirmed TNF G-308A polymorphism (1;2) and reside in different haplotype blocks as shown in Fig. 1C. These genes are also not linked to the well-established ancestral haplotype 8.1 that comprises TNF-308A-HLA A*01-B-07*DRB1*03 that is associated with NHL and remain statistically significantly associated with NHL even in logistic regression models that include both the implicated TNF-308A and HLA DRB1*0101 alleles. Still, our alleles may represent surrogates for other causal genes associations with NHL and NHL subtypes. These collective data along with emerging evidence for HLA (3;4) and non-HLA genes (3) located in the same region in NHL etiology strongly suggest that a detailed evaluation of genetic variation in the 6p21.3 chromosomal region is warranted.

Of the genes in 6p21.3 we found to be significantly associated with NHL etiology, BAT4 has previously been linked to NHL susceptibility, specifically for CLL (23). In our data, we found a different BAT4 SNP (rs3115667) to be significantly associated with all NHL, DLBCL, follicular and marginal zone lymphomas but not CLL/SLL. The remaining gene regions implicated in our data have not previously been implicated in NHL etiology, though they have been implicated in other aspects of lymphomagenesis. Lower expression of RXRB and other MHC Class II genes were reported to be correlated with poor patient survival in mediastinal DLBCL (24) and the AIF1 is one of a group of genes (including BCL6, LMO2, GCET1) found to define the germinal center B-cell signature (25), also with implications for DLBCL survival. A brief description of all top-ranked genes associated with NHL is shown in Table 3.

Our study strengths include the population-based design of this study and the tagging algorithm intended to allow for gene-based and SNP-based analyses. We note that our results are consistent when restricted to non-Hispanic Caucasians and thus are unlikely to be biased by population stratification. Still, we cannot fully exclude the potential for false positive associations in our results and they therefore require replication in other independent and larger populations such as in pooled efforts. Other study limitations include loss of eligible subjects to death, illness, and refusal to participate. We also had limited power to detect modest associations for rare alleles and less common NHL subtypes.

In summary, our data suggest genetic variations in the SYK gene on chromosome 9 and 3 additional genes in the 6p21.3 chromosomal region to be associated with NHL etiology. Our results further support the need for in-depth analyses of genetic variants in the 6p21.3 chromosomal region. In particular, combining HLA allele information with data emerging from genome-wide association studies should contribute significantly to identifying specific regions of functional importance to NHL etiology. Additional studies will be needed to delineate gene–environment interactions, particularly as a number of autoimmune conditions, some of which are potential risk factors for NHL (26), are also associated with susceptibility alleles in HLA and chromosome 6p21.3 immune response genes. Our results for SYK require replication, but are of great interest due to its importance in lymphomagenesis and its growing potential as a therapeutic target.

No potential conflicts of interest were disclosed.

We thank Peter Hui, Michael Stagner, and Mary McAdams of the Information Management Services, Inc. for their programming support. We also gratefully acknowledge the contributions of the staff and scientists at the SEER centers of Iowa, Los Angeles, Detroit, and Seattle for the conduct of the study's field effort.

The NCI-SEER study was supported by the Intramural Research Program of the NIH (NCI), and by Public Health Service (PHS) contracts N01-PC-65064, N01-PC-67008, N01-PC-67009, N01-PC-67010, and N02-PC-71105. DNA extraction, genotyping and statistical analysis for this project were supported by the Intramural Research Program of the NIH [National Cancer Institute (NCI)]. This project has been funded in part with federal funds from the NCI, NIH, under contract no. HHSN261200800001E. The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.