Inherited Inflammatory Response Genes Are Associated with B-Cell Non-Hodgkin’s Lymphoma Risk and Survival

Background Malignant B-cell clones are affected by both acquired genetic alterations and by inherited genetic variations changing the inflammatory tumour microenvironment. Methods We investigated 50 inflammatory response gene polymorphisms in 355 B-cell non-Hodgkin’s lymphoma (B-NHL) samples encompassing 216 diffuse large B cell lymphoma (DLBCL) and 139 follicular lymphoma (FL) and 307 controls. The effect of single genes and haplotypes were investigated and gene-expression analysis was applied for selected genes. Since interaction between risk genes can have a large impact on phenotype, two-way gene-gene interaction analysis was included. Results We found inherited SNPs in genes critical for inflammatory pathways; TLR9, IL4, TAP2, IL2RA, FCGR2A, TNFA, IL10RB, GALNT12, IL12A and IL1B were significantly associated with disease risk and SELE, IL1RN, TNFA, TAP2, MBL2, IL5, CX3CR1, CHI3L1 and IL12A were, associated with overall survival (OS) in specific diagnostic entities of B-NHL. We discovered noteworthy interactions between DLBCL risk alleles on IL10 and IL4RA and FL risk alleles on IL4RA and IL4. In relation to OS, a highly significant interaction was observed in DLBCL for IL4RA (rs1805010) * IL10 (rs1800890) (HR = 0.11 (0.02–0.50)). Finally, we explored the expression of risk genes from the gene-gene interaction analysis in normal B-cell subtypes showing a different expression of IL4RA, IL10, IL10RB genes supporting a pathogenetic effect of these interactions in the germinal center. Conclusions The present findings support the importance of inflammatory genes in B-cell lymphomas. We found association between polymorphic sites in inflammatory response genes and risk as well as outcome in B-NHL and suggest an effect of gene-gene interactions during the stepwise oncogenesis.


Introduction
Normal B-lymphocyte homeostasis requires survival and proliferation signals provided by cells in the lymph node microenvironment. In the case of B-cell non-Hodgkin's lymphoma (B-NHL) the malignant cells share some similarities with their normal B-cell counterpart linking the immunological inflammatory response to the growth potential of the malignant clone [1][2][3]. Diffuse large B-cell lymphoma (DLBCL) and follicular lymphoma (FL) accounts for the majority of B-NHL cases [4] and genetic analysis of the tumor microenvironment in these diseases has revealed different gene signatures associated to survival [5][6][7]. Functional single nucleotide polymorphisms (SNPs) affect the inflammatory microenvironment nesting malignant tumours [8]; however SNPs may also provide a direct effect on the malignant B-cells in the process of tumorigenesis [9]. Different genetic loci are associated with risk or outcome in B-NHL [10][11][12][13][14], amongst them most notably the IL4RA, IL6, IL10 and TNFA loci [15][16][17][18][19][20][21] and recently, genome wide association studies (GWAS) have suggested an effect of a number of other potential loci encoding inflammatory mediators [22,23]. The focus in most of the previous studies has primarily been on single gene effects, however disease susceptibility and prognosis in complex diseases such as B-NHL may not be caused by single genes but by genes interacting [24] as reported by a small number of investigators [10,[25][26][27][28][29][30]. In the present paper we investigate the effect of single gene, haplotype and gene-gene interactions in different histological subtypes of B-NHL. We further suggest, that the biological effect of such interacting inflammatory response genes with germ line polymorphisms is of pathogenetic impact supported by the investigation of "risk genes" expression in normal B-cell compartments and lymphoma tissue [31,32].

Patients, controls and ethics statement
The current study included 355 B-NHL cases classified according to the World Health Organization (WHO) classification (216 DLBCL and 139 FL). Clinical data was abstracted from the Danish lymphoma database (LYFO) and from medical records. Follow-up information included type of treatment, International Prognostic Index (IPI) score and time to relapse. We included material from 307 healthy blood donors as previously described [33]. Research protocols were approved by the local scientific ethics committee of the North Denmark Region (approval numbers: N-20100059, N-20090018). The use of samples without informed consent was approved by the committee since the result did not result in any interventions and all samples were anonymized and de-identified prior to analysis. For details see S1 Text. acid (EDTA) stabilised whole blood. For gene-expression analysis, normal lymph node tissue and DLBCL lymphoma cells were processed as previously described [34]. For details see S1 Text.

Genotyping
Genotyping was performed using a TaqMan OpenArray genotyping system from Applied Biosystems (ABI, Foster City, CA, USA). Seven SNPs were genotyped using custom-designed assays and 43 SNPs were genotyped using predesigned TaqMan SNP assays (see S1 Table for detailed assay information). The arrays were read using the OpenArray NT Imager and the allele calls and scatter plots were generated with the Biotrove OpenArray SNP Genotyping Analysis Software package version 1.0.3. The default threshold for the Quality value was set to 0.95. For details see S1 Text.
Analysis of gene expression profiles of normal and malignant lymphnodes DLBCL lymphoma samples and and flow-sorted B-cell populations from Ficoll-purified mononuclear cells from human tonsils were hybridized to Affymetrix GeneChip Human Genome U133 Plus 2.0 Arrays (Affymetrix, Santa Clara, CA) as previously described [3]. The CEL files are deposited in the National Center for Biotechnology Information Gene Expression Omnibus repository (GSE56315) [3]. Gene expression profiles were analyzed using the statistical software system R, version 2.15.3. (http://www.r-project.org) and Bioconductor R-packages [35]. The CEL files were normalized by the just.rma function from the Bioconductor package affy, using customized cdf-files from brainarray [36], which summarized the probesets to ENGS symbols. The ENSG symbols were mapped to the HGNC symbols of interest by the function getBM from the Bioconductor package biomaRt.

Statistical analysis
Statistical analyses were carried out by the statistical software Stata Version 12.1 (Stata Corporation, College Station, TX, USA) and the statistical software system R, version 2.15.3. (http:// www.r-project.org). PHASE 2.1 software (University of Chicago, USA) was used to construct haplotypes. Logistic regression analysis was used to estimate the association between odds of lymphoma and the single genotypes, alleles, haplotypes, and gene-gene interactions, respectively. Survival analysis for 10 year overall survival (OS) was used to analyse the effect of single SNPs, haplotypes, and gene-gene interactions between SNPs corresponding to different genes. Cox proportional hazards models adjusted for sex, age, IPI, and treatment was used for survival analysis. P-values below 0.05 were generally considered statistically significant. The purpose of performing gene-gene interaction analyses was to identify pairs of genes for which the interaction between them was significantly associated with the risk of getting lymphoma (performed by logistic regression) or significantly associated with the survival among lymphoma patients (performed by Cox regression). We performed the gene-gene interaction analyses by comparing the models with and without the interaction term using a likehood ratio test. Highly significant interactions were selected using the p-value from the likelihood ratio test corrected for the number of individual loci (31 loci) investigated why only interactions with a p-value of < 0.002 were considered significant. As possibly 9 combinations of genotypes exist for each pair of genes we have for the survival part reported hazard ratio estimates for each combination of genotypes with the combination of the most frequent homocygotes as reference. For the geneexpression data where a Bonferroni corrected p-value < 0.05 was considered significant.

Study population
The patient population consisted of 139 FL and 216 DLBCL patients with a median age of 63 years including 53% males. The median OS was 75.5 months (58.7 months for DLBCL and 101.7 months for FL). In this population, age (p < 0.0001) and IPI (p < 0.0001), but not sex (p = 0.39) were prognostic markers for OS. The control group consisted of 307 healthy blood donors. Of the controls 46% were males and the median age was 41 years.

Gene expression analysis in normal B-cell subsets and DLBCL lymphoma cells
We selected the most significant genes from the gene-gene interaction analysis and analyzed the expression of these genes in the normal lymph node B-cell hierarchy and DLBCL lymphoma cells. We explored genes related to the IL-4 and IL-10 cytokines; Figs 7-10 illustrates differences in expression between normal germinal centre (GC) B-cells (defined as centrocytes (CC) and centroblasts (CB)) and non-GC subtypes (naïve (N), plasmablasts (PB), memory (M) B-cells) and DLBCL cells using Bonferroni corrected p-values. The IL10 gene (Fig 7) and the IL10RB (Fig 8) gene were up regulated in DLBCL cells when compared to normal GC subtypes. The increased expression of IL10 was only seen in DLBCL cells whereas increased IL10RB expression could also be shown in pre-and post germinal center subpopulations. The IL4 gene was equally expressed in all normal B-cell subpopulations (Fig 9) whereas the expression in DLBCL cells was slightly down regulated, a similar expression profile was seen for the IL4R gene (Fig 10) however this gene was also found to be down regulated in post GC subpopulations.

Discussion
DLBCL and FL share a common dependence of the inflammatory tumor microenvironment for their proliferation and growth [7] however we generally observed a specific association with disease risk supporting DLBCL and FL being two different diseases with respect to genetic risk factors. Interleukins (IL) and toll-like receptors (TLR) play a major role in B-cell differentiation and proliferation. The TNFA gene have been intensively studied and in accordance with others, including the large pooled Interlymph study and a recent GWAS study including largely the same patients [10,37] we found the TNFA (rs1800629) A allele associated with increased risk of DLBCL (OR = 1.46 (1.06-2.00)). The functional IL1B (rs1143627) C allele and IL1B (rs16944) A allele were associated with a decreased risk of FL but not DLBCL, a trend for an association with the IL1B gene was observed in the Interlymph study [10], however substantial difference was observed between the included cohorts. [10] The IL10RB (rs1058867) SNP was earlier investigated in a pooled study [12] reporting an association with FL; results we failed to Among other notable findings, the (rs241447) SNP in the TAP2 gene, coding for a HLA-II associated transport protein [40], was significantly associated with decreased risk of DLBCL, OR = 0.61 (0.44-0.84). Vijai et al. reported an adjacent SNP in the TAP2 gene to be associated with risk of DLBCL and FL, a finding confirmed by Cerhan et al. [23,40]. The cumulative findings emphasize the possible importance of genetic variation in the HLA-II region and further  evaluation of the TAP2 gene in larger cohorts is awaited. The GALNT12 gene is linked to pathological aberrant glycosylation, and was associated with outcome in FL in a recent GWAS [41] study including 244 FL cases. We observed a decreased risk of FL for carries of the GALNT12 (rs10987898) G allele whereas there was no apparent effect on DLBCL risk, supporting a specific role for this gene in the pathogenesis of FL and emphasising the need to explore this gene in relation to the glycosylation patterns in FL B-cells. Another intensively studied locus [13,15,42] is the FCGR2A gene. The FCGR2A gene encodes a low affinity Immunoglobulin-G (IgG) receptor (FcγRIIa) and genetic variation affects inflammatory cytokine production [43]. The association between the FCGR2A (rs1801274) SNP and DLBCL was also confirmed in our population (OR = 1.37 (1.07-1.77)) whereas the association with FL was not statistically significant (OR = 0.76 (0.57-1.01)).  When we analyzed 10 year OS, we observed an association between the CHI3L1 (rs4950928) C allele and a worse outcome in FL (CG genotype compared to the CC genotype) -the rare GG genotype group was unfortunately small and did not add sufficient information as presented the Kaplan-Meier plot (Fig 1). The CHI3L1 gene encodes the pro-inflammatory  Genetic Polymorphisms in B-NHL YKL-40 protein [44] and we have previously reported circulating levels of YKL-40 to be associated with dismal outcome in NHL [45] and the existence of the functional (rs4950928) SNP in the CHI3L1 gene affecting serum-YKL-40 [33], however to the best of our knowledge the CHI3L1 gene have not previously been investigated in this setting. The SELE gene encodes the selectin-E molecule expressed on endothelial cells known to promote metastasis [46]; the SELE (rs5361) SNP was found to be associated with OS in FL and DLBCL in a recent study [47] and we found an equal effect on the combined DLBCL and FL cohort, however, the effect was not obvious when analyzed at specific diagnostic entity levels. IL12A (rs485497) (Fig 2) was associated with decreased risk in FL, as well as outcome in DLBCL and we also confirmed the earlier suggested association between B-NHL outcome and the CX3CR1 gene [47] (Fig 3) and we did observe a novel association between the IL5 (rs2069812), encoding B-cell growth factor IL-5 [48], and outcome of DLBCL ((HR = 1.94 (1.12-3.39)) for carriers of the TT genotype, Fig 4). Surprisingly, we observed a DLBCL specific association between the TAP2 (rs241447) gene in GG genotype and decreased OS however the G allele is significantly associated with decreased risk of DLBCL (OR = 0.61 (0.44-0.84)) we do not have an explanation for this findings, but it could suggest that this protein could play a different role in disease initiation and response to treatment (Fig 5). The suggested effect are however in line with the general believe of genes in the HLA-II region of chromosome 6p as one of the most interesting areas in the genome in relation to B-cell lymphomas and our findings of a possible relation to OS in DLBCL must encourage to further studies of this region. The MBL2 gene has been studied in a number of malignant and infectious diseases with conflicting results. The MBL2 (rs7096206) is associated with risk of hepatocellular carcinoma [49] and febrile neutropenia [50] however the genotype have not been studied in relation to NHL [51]. Whether the effect of this genotype on DLBCL survival (Fig 6) is related to disease progression or increased risk of infectious complication needs to be investigated in future studies.
Growing evidence suggest, that inflammatory mediators act in networks why co-occurrence of genotypes -gene-gene interaction -analysis needs to be considered in complex diseases as B-NHL [24,52]. Of most notable interest, was a highly significant interaction between two risk alleles; IL10 (rs1800872) and IL4RA (rs1801275). IL-4 and IL-10 play a major role in the GC reaction, B-cell proliferation is enhanced by IL-10 in the presence of IL-4 and both cytokines are involved in the class switch recombination process [53][54][55]. We further explored gene-gene interactions in relation to FL and DLBCL specific OS. Most notably, we discovered a highly significant (p = 0.0002) interaction between IL10 (rs1800890) and IL4RA (rs1805010) in DLBCL. Carriers of IL10 (rs1800890) AA in combination with the IL4RA (rs1805010) AA genotype had a significantly improved outcome (HR = 0.11 (0.20-0.50)). Several SNPs in the IL4, IL4R and IL10 genes has been proposed associated with outcome in earlier studies [20]. Analysis of the effect of these individual genes have been somewhat contradictory however the results of attempts to include these loci in multi-gene models [47], including ours, argues for a combined effect of these loci not readily identified in a single gene model emphasizing the interaction between these cytokines to be of importance and in need of exploration in future studies. For FL the most notable effect of gene-gene interactions was seen for interaction between the IL1RN (rs2637988) and IL4 (rs2243248) (p = 0.0004). As it was the case for DLBCL, a SNP related to the IL-4 cytokine was involved in gene-gene interactions in the FL group. We did explore survival estimates for gene-gene combinations (S8, S9 and S10 Tables), however risk estimates are imprecise since the number of events in each group was small.
In order to study the functional aspect of these genes in the B-cell hierarchy, we combined the SNP study with analysis of gene expression in normal B-cell subpopulations and DLBCL lymphoma cells. We restricted the analysis to the most notable genes from the gene-gene expression analysis which suggested a prominent effect of IL-4 and IL-10 related genes. Increased expression of IL-10 related genes has earlier been shown in chronic lymphocytic leukaemia (CLL) and in Waldenström's macroglobulinemia [56]. We found a up regulation of the IL10 and the IL10RB gene in DLBCL cells when compared to normal GC B-cell subpopulations supporting the hypothesis that this cytokine/receptor could play a role for the survival of germinal centre derived lymphomas, as supported by the earlier findings in CLL [56] and further supporting the protective effect of the low producer IL10 (rs1800872) AA genotype as well as the protective effect of the IL10RB (rs1058867) G allele [10,12]. The IL4 gene was slightly down regulated in DLBCL cells when compared to all normal B-cell subpopulations whereas the IL4R gene was down regulated in post GC B-cell subpopulations suggesting that these genes could be important at different stages of the stepwise oncogenetic process [57] and further suggesting that such genes with oncogenetic potential may be important in the multi step process of malignant transformation even if they are not found to be expressed in malignant tissue [9]. These findings needs to explored further by investigating the expression of these genes in the different B-cell subpopulation thought to be the cell of origin of the investigated B-cell malignancy [31,58]. The functional properties of these pathways obviously needs further Genetic Polymorphisms in B-NHL investigation; as an example, the functional properties of IL4 gene and the distal IL10 promotor variations still remains largely unknown and so does the combined effect of these interactions on protein level [59]. Our study holds some limitations providing a possible bias. We do acknowledge that the sample size in our study was too small for in depth investigation of genegene interactions making our risk estimates less precise. We however do encourage that the findings from our exploratory investigation of gene-gene interactions is investigated in a larger cohort, preferably a multicenter study. The age distribution in the control group and patient group was not identical and the sample material differed between groups. We found no association between age and genotypes within the groups and we included a cohort with no reported BM involvement, however these differences between groups still could have introduced bias to our results.

Conclusion
We analysed 50 SNPs in inflammatory response genes with respect to risk and outcome in B-NHL. We reported a significant association between thirteen SNPs and risk and reported seven SNPs associated with outcome. Moreover we here suggest a gene-gene interaction effect in relation to risk and survival and when we combined these findings with gene expression analysis in normal and malignant B-cells we suggest that these inherited variations could play a role at different stages of the multi step process of B-cell oncogenesis. These findings further strengthened the discovered association between interacting key inflammatory response genes and B-lymphoma and we propose further studies exploring the functional aspects of these interactions. As for earlier studies on genetics in lymphoma, our results were not always consistent with findings in other B-NHL cohorts. This can be a result of sample size as well as ethnical differences between study populations [51] although we have a well defined, ethnically homogenous population with a long follow up period. Despite limitations of our study, we believe that our findings confirms some of the earlier reported findings and adds new knowledge to the immunogenetics as an important factor in relation to risk and outcome in B-NHL.   Table. Gene-gene interactions in relation to overall survival in DLBCL (DOCX) S10 Table. Gene-gene interactions in relation to overall survival in FL (DOCX) S1 Text. Supplementary methods.