Unraveling the interactions between environmental factors and genetic polymorphisms in non-Hodgkin lymphoma risk

Abstract

The etiologic causes for non-Hodgkin lymphoma (NHL) remain largely unexplained. Challenges include the heterogeneous nature of NHL, the difficulty in accurately assessing exposures and the few genetic variations with confirmed associations to date, with none yet identified as causal. These challenges are compounded when evaluating gene–environment interactions. However, there are some well-characterized NHL risk factors where pursuing gene–environment interactions could shed important information on the mechanisms relevant for lymphomagenesis, such as whether common mechanisms of action exist across different exposures, specific gene families or pathways are of importance for certain exposures, or susceptible subgroups can be identified. As new associations from genome-wide association studies are confirmed, broader agnostic and data-mining approaches may yield additional clues. These results could provide a platform for identifying biological mechanisms of interest and identifying new and/or clarifying known exposures important for lymphomagenesis. This will require large consortial efforts to attain adequate sample size and power to detect meaningful biological and statistical interactions.

Keywords::

• environment
• gene
• interaction
• joint effect
• lymphoma
• non-Hodgkin lymphoma

In epidemiology, ‘interaction’ is defined as “differences in the effects of one or more factors according to the level of the remaining factors” [1]. For evaluating interactions between genes and the environment, researchers search for differences in the risk effects of an environmental exposure given different genotypes (e.g., wildtype, heterozygous or homozygous variant) or vice versa where there is differential genotypic risk among varying levels of an environmental exposure. There are only a handful of well-characterized gene–environment interactions in the epidemiologic literature. One of the best known examples is the interaction between the metabolizing gene, N-acetyl transferase 2 (NAT2), smoking behavior and bladder cancer risk [2]. In this example, there is a modest association (main effect) between the NAT2 slow acetylators with increased risk of bladder cancer. When stratified by smoking status, however, the association between NAT2 slow acetylation and bladder cancer is stronger among current smokers, compared with former and never smokers. There are several things to note about this example. First, the disease (bladder cancer) is clearly defined. Second, the environmental exposure (smoking) is accurately assessed. Third, the NAT2 genotype has a well-characterized phenotype – slow acetylation. To unravel similar interactions between genes and the environment in non-Hodgkin lymphoma (NHL) and any other disease will similarly require clearly defined disease, accurate exposure measurement and genetic variations with a well-characterized phenotype. Here, we discuss the present knowledge of NHL in terms of these characteristics and how research to uncover gene–environment interactions in NHL may be pursued to further our understanding of NHL etiology (Box 1). For in-depth statistical discussion on conducting gene–environment interactions, we refer the reader to recent manuscripts published on this topic [3–5].

Non-Hodgkin lymphoma is the fifth most common cancer diagnosis in the USA [6] and predominantly affects white, non-Hispanic men with rates a third lower among women and African–American and Hispanic populations. Incidence rates differ considerably across geographic regions with higher rates (11–16 per 100,000 people) in western Europe, Australia and North America, and rates below four out of 100,000 in Asia and Africa. The distribution of NHL subtypes further varies between geographic regions with higher proportions of B-cell NHL (diffuse large B-cell lymphoma [DLBCL], follicular lymphoma and chronic lymphocytic leukemia) in North America and Europe, and higher proportions of T-cell NHL in Asia and Burkitt’s lymphoma in Africa [7]. NHL incidence rose in the latter half of the 20th Century (and is still rising among men and women of older ages) [6], but the etiologic causes that account for this rising incidence remain unexplained. One hypothesis regarding why exposures have not yet been identified is that they are not being evaluated in a susceptible population, such as that defined genetically. Similarly, studies to identify genetic variants associated with NHL risk may prove more fruitful if evaluated in the context of the culpable exposure. In other words, the “failure to incorporate both genetic and environmental factors in joint analysis will weaken the observed associations between a true risk factor and disease occurrence” [8]. As a result, identifying risk factors may be particularly challenging in a combined population of susceptible and nonsusceptible individuals because risk estimates would be biased toward the null.

Challenges

Defining the outcome

Non-Hodgkin lymphoma is a heterogeneous disease entity comprising of many subtypes, each recognized to be clinically and histologically well-defined, with distinct incidence rates and patterns. Although there are risk factors that appear to be relevant for NHL etiology in general, there is an increasing recognition that NHL subtypes may have different etiologies [9] as exemplified by some well-known examples, including infection with Epstein–Barr virus (EBV) and Burkitt lymphoma, human T-lymphotropic virus (HTLV1), peripheral T-cell lymphoma, infection with Helicobacter pylori and mucosa-associated lymphoid tissue lymphoma (MALT) [10–12]. Most epidemiologic studies on NHL to date have been designed with sufficient power and sample size to evaluate all NHLs as a single disease entity (e.g., target of ∼1000 cases and 1000 controls). However, as it has become apparent that individual subtypes may have distinct etiologies, individual studies have found that they are not sufficiently powered to evaluate risk factors by subtypes, particularly for risk factors conferring a modest (≤twofold) risk. To do so would require each individual subtype to amass a thousand cases. However, for some tumor sites, in situ disease or intermediate disease markers along the causal pathway (e.g., monoclonal gammopathy of undetermined significance for multiple myeloma) can be used as an intermediate outcome to increase sample size and to further enable analyses that dissect their roles between development of an intermediate outcome from specific involvement in disease progression. For chronic lymphocytic leukemia, for example, monoclonal B-cell lymphocytosis may be considered an emerging precursor [13,14]. Pooled and meta-analyses will play important roles in evaluating risk factors by NHL subtypes with sufficient power. Moreover, standardization of NHL subtype definitions for epidemiologic studies, such as that published by the International Lymphoma Epidemiology (InterLymph) Consortium will be essential for such large-scale pooling efforts [15].

In recent years, researchers have used gene-expression profiling to identify molecular subtypes for DLBCL, follicular lymphoma and Burkitt lymphoma, potentially adding another layer of complexity to defining NHL. Most notably, distinct gene-expression signatures for DLBCL have been identified, which divides DLBCL into at least two subgroups, the germinal-center (GC) B-cell signature and activated B-cell-like subgroup [16,17]. These molecular subtypes have differential prognosis and their gene-expression profiles suggest that they arise at different stages of normal B-cell development. Whether these specific molecular subtypes are etiologically relevant is unknown, but at least one study of rheumatoid arthritis patients has suggested that severe rheumatoid arthritis is associated specifically with the non-GC subtype of DLBCL [18]. At present, algorithms that can delineate GC from non-GC DLBCL as more stable biomarkers are being developed and would facilitate the evaluation of molecular subtypes in large epidemiologic studies. Such algorithms include the use of antibodies for immunostaining [19,20] and the measurement of DNA methylation profiles where distinct methylation levels of a CpG island proximal to FLJ21062 have been reported for activated B-cell and GC-DLBCL subtypes [21]. Validation of these markers is particularly important in the context of epidemiological studies, given their relative stability compared with gene-expression signatures.

Defining etiologic exposures by other molecular characteristics has also been explored previously, most notably for t(14;18), where insecticides and herbicides were associated with t(14;18)-related NHL [22,23]. Molecularly defined phenotypes, such as by chromosomal translocations and specific gene mutations that characterize specific steps in lymphomagenesis (e.g., BCL2 and BCL6), may also improve outcome definitions for NHL and are worth further exploration. As many of these molecular characteristics are now reported routinely in patient medical records owing to their relevance for selecting appropriate therapeutic options, incorporating these data in epidemiological analyses may become increasingly feasible.

Accurately measuring exposures

Risk factors for NHL have been reviewed extensively [24] and are briefly described here (summarized in Table 1). The most pronounced risk for NHL is from primary or acquired immunodeficiencies. Primary immunodeficiencies carry the highest relative and absolute risk for NHL and up to a quarter of all patients with congenital immunodeficiencies will develop cancer, with half of these cancers being lymphomas [24]. NHL also occurs in higher frequencies among those with autoimmune disorders, notably Sjogren’s syndrome, systemic lupus erythematosus and celiac disease [25]. Lymphomas also arise in organ-transplant recipients on immunosuppressive therapy (heart, kidney and renal), with the resulting NHL risk corresponding to the type of medication and degree of immunosuppression conferred. More modest NHL risks are observed for common medications or recipients of allogeneic blood transfusions and some report decreased risks with atopy [26–28].

Well-established infectious agents associated with NHL include HIV with clinically aggressive NHL, human herpesvirus 8/Kaposi’s sarcoma; herpes virus with primary effusion lymphoma; EBV with Burkitt lymphoma; Campylobacter jejuni with immunoproliferative small intestine disease; Borrelia burgdorferi with primary cutaneous marginal zone lymphoma; Chlamydia psitacci and ocular adnexal lymphoma; and HTLV1 with acute T-cell leukemia/lymphoma [29]. There is increasing evidence supporting the association between H. pylori infection and MALT lymphomas in the gastrointestinal tract and between hepatitis C virus and B-cell NHL [30].

Occupations associated with elevated NHL risk include farmers, livestock workers, printers, dry cleaners, wood workers, teachers, hairdressers and barbers [31,24]. Emerging data from association studies and functional analyses suggest potential relevance of circadian genes in lymphomagenesis [32] and shift work as an emerging new exposure that needs to be considered. Potential environmental exposures include nonarsenical insecticides, tetrachloroethylene and trichloroethylene, 2,3,7,8-tetrachlorodibenzo-para-dioxin and 1,3-butadiene, which are considered ‘probable human carcinogens’ for NHL by the International Agency for Research on Cancer. Growing evidence also implicates dichlorodiphenyldichloroethylene, α-chlordane and organochlorines that persist in the environment and accumulate in fat and blood, including polychlorinated biphenyls, dioxins and furans, as well as solvents, foremost benzene, in NHL risk [33,34].

Behavioral risk factors include decreased NHL risk with increasing sunlight exposure [35], alcohol consumption [36], physical activity [37] and high vegetable intake [38]. Moderately increased risks are reported for severe obesity [39] and, potentially, height [37], high fat intake with DLBCL [40] and cigarette smoking with follicular lymphoma [41]. In addition, hair dye use of darker colors and prior to the 1980s appear to be modestly implicated in NHL risk, particularly for follicular lymphoma and chronic lymphocytic leukemia/small lymphocytic lymphoma [42].

We note that several of the listed (behavioral) NHL risk or protective factors have mainly been explored within the context of case–control studies with the methodological drawbacks associated with this study design, including selection and recall bias, and potential disease effects for biological markers. Confirmation of these associations in cohort-based studies will, therefore, be important in verifying hypothesized associations as delineated in Table 1.

Known NHL risk factors explain only a small proportion of NHLs. One explanation for this is that the culpable exposures are not accurately or reliably assessed and misclassification of exposures would bias any resulting risk estimate between an exposure and disease to the null. As it is thought that multiple risk factors each with modest risk estimates exist for NHL (as opposed to a single etiologic risk factor), accurate exposure assessment in evaluating NHL risk even more critical.

The delineated NHL risk factors pose several challenges for pursuing gene–environment interactions, namely the rarity, and varying sensitivity and specificity of some exposure measurements. For example, although autoimmune conditions are an established risk factor for NHL, each condition is extremely rare, making further evaluation and stratification by genotypes for evaluating gene–environment interactions challenging. In addition, treatment of these diseases, such as with TNF-blocking drugs that are suspected to have a role in lymphomagenesis, may interfere with the effect of the autoimmune condition itself on NHL risk. Furthermore, information on medication use and also prevalence of these drugs may differ internationally and further add complexity to analyses of gene–environment interactions in pooled data.

The association between specific occupations and NHL is intriguing, but occupations are not specific to the etiologic agent(s). While aggregating specific occupations may increase the sensitivity of detecting the exposure, the diminished specificity also results in a diminished risk estimate. Identifying specific environmental and occupational exposures has been particularly challenging in NHL research. Studies have used varying levels of detailed questionnaires or occupational histories, and some have integrated more detailed biomarker assessments in their studies. Different study designs also provide varying levels of detail; cohort studies provide long-term histories and temporal variations while case–control studies provide extensive questionnaire data but may be affected by recall bias and reverse causation. Evaluating infectious diseases is also subject to what biological sample the agent is being detected (e.g., tumor specimen vs non-tumor specimen such as serum), what is being measured (e.g., viral DNA vs RNA vs antibodies), and the type of exposure being captured (e.g., chronic or acute infection, current or lifetime exposure and reactivation of latent infection). Biomarker-based assays in the context of case–control studies may also be hampered by probable effects of treatment on assay performance. In pooled biomarker studies the percentage of treated individuals between studies might vary considerably depending upon the mode of recruitment. In addition, identifying confounders or dissecting an exposure from collinear exposures require special attention. Finally, consideration of dose effects of environmental exposures, nonlinear relationships and time of exposure assessment is also required, particularly for pooled studies where exposures may not have been consistently assessed in the different studies. In one example of gene–environment interaction in the etiology of rheumatoid arthritis, effects were observed specifically for HLA shared epitope and smoking when stratified by pack-years of smoking but not by smokers versus never smokers [43].

Other general issues that could pose a challenge to gene–environment analyses across large-scale pooling efforts in consortia are international differences in exposures (e.g., levels of folate fortification). In these instances, biomarkers would be useful in complementing questionnaire-based exposure assessment to clarify potential gene–environment interactions in C1 metabolism in the etiology of NHL. Similarly, varying prevalence of infectious agents and differences in underlying allele frequencies by population need to be carefully understood when harmonizing data in pooled projects among heterogeneous study populations.

Complicating potential downstream prospects for evaluating gene–environment interaction is that many of the exposures implicated in NHL are thought to serve as surrogates for the still unknown etiologic agent. For example, the decreased risk observed with sunlight may be mediated through vitamin D, and alcohol intake may reflect another unidentified aspect of lifestyle that correlates with alcohol consumption. Height is a well-known surrogate for social class and the association between height and NHL may reflect this. The association between diet and NHL may reflect either an association with weight or environmental contaminants in the foods consumed. The association between birth order and NHL is thought to be a proxy for yet to be identified childhood exposure to infectious agents.

The mechanisms of action hypothesized for these exposures also vary and do not necessarily reflect a direct causal association. For example, infection with HIV is thought to indirectly cause lymphoma via disruption of immune surveillance and resulting immunosuppression rather than the virus itself being oncogenic [44]. Similarly, the association between hepatitis C virus and NHL is thought to be attributed to chronic immune stimulation rather than due to the virus itself [45]. It also remains unknown whether the implicated environmental and occupational exposures reflect the immunotoxicity of the agents or are the result of chromosomal damage. Finally, potentially relevant exposures for lymphomagenesis may also be logistically difficult to measure, such as hypothesized in utero or early childhood exposures. Increasing evidence points towards a role of early-life exposure for adult-onset diseases via epigenetic modifications [46,47].

Biomarker-based assays could improve the accuracy of exposure assessment, particularly in prediagnostic specimens from prospective studies. In case–control studies, the use of biomarkers requires careful consideration with regard to their assay performance and resulting measurements and whether they are affected by treatment or disease effects.

In summary, a major limitation to understanding lymphoma etiology is in the precision and accuracy of exposure assessment available for epidemiologic studies. Gene–environment interactions in NHL therefore need to be carefully conducted and subsequently interpreted with regard to the potential limitations of the exposure being evaluated.

Defining which genes to pursue

The role of genetic susceptibility in NHL is well-established, although the effect is recognized to be modest at best. There is a twofold relative risk for NHL among those with a first-degree relative with NHL that translates to an absolute lifetime risk of 3.6% [24]. Unlike breast and ovarian cancer, there are no extensive family histories of NHL and no evidence for a major genetic effect.

Recent advances in genomic technologies give rise to the promise that we will soon come to understand the genetic polymorphisms that play a role in NHL etiology, however modest their effects. Unlike the environmental and other (nongenetic) risk factors described previously, genetic variants are considered discrete exposures and are not necessarily subject to the same misclassification. Published genetic polymorphisms or single nucleotide polymorphisms (SNPs) associated with NHL and its subtypes were recently reviewed [48]. Candidate pathways and regions that have been assessed include: proinflammatory genes, Th1/Th2/Treg genes, innate immunity, oxidative stress, one-carbon metabolism, DNA repair, energy regulation, sex hormone production, detoxification genes, and genes in the MHC region and HLA(Table 1).

Confirmed SNP associations reported to date include a polymorphism in the promoter region of the TNF gene (-308G->A) and the IL-10 -3575T->A polymorphism, both associated with increased NHL risk, and particularly increased DLBCL risk [49,50]. Genome-wide association studies (GWAS) are now underway and have identified variations on chromosome 6p21.33 and in a region near the psoriasis susceptibility region 1 (PSORS1) [51] with follicular lymphoma risk. A GWAS of chronic lymphocytic leukemia identified another six loci at 2q13 (rs17483466), 2q37.1 (rs13397985), 6p25.3 (rs872071, IRF4), 11q24.1 (rs735665), 15q23 (rs7176508) and 19q13.32 (rs11083846, PRKD2) [52]. As these large-scale approaches employ tagging algorithms to identify gene variations for investigation, subsequent efforts to identify causal variants with functional evidence are required. The emerging results from GWAS will result in increased efforts to identify gene–environment interactions in NHL etiology. These ongoing efforts for SNPs will serve as a model for other types of genetic variations where research is still in its infancy for NHL, including evaluation of copy number variants, miRNA variations or epigenetic changes. Goals that can be achieved with evaluation of gene–environment interactions.

With the challenges in explaining the majority of lymphomas through environmental exposures elusive and few gene variants confirmed to date (although none confirmed as causal), there have been a limited number of evaluations of gene–environment interactions in NHL in the current literature (Table 2). We believe, however, that there are potential clues that can be gained with preliminary exploration at this time.

Linking common mechanisms of action across varied exposures

For example, the SNP close to PSORS1 can be evaluated with respect to autoimmune conditions and allergies, and TNF and IL-10 SNPs can be evaluated with respect to infections, autoimmune conditions or all exposures with hypothesized mode of action being inflammatory pathways. One US-based report evaluated the effects of established and hypothesized NHL risk factors in relation to the TNF G308A or IL-10 T3575A genotypes [53]. The study reported increased DLBCL risk among those with variant TNF or IL-10 alleles and those with an autoimmune condition or who were last-born status or obese. The results of this analysis poses the hypothesis that autoimmune conditions, late birth order and obesity may act partly through a common inflammatory pathway, posing a greater risk to individuals with variant TNF and IL-10 genotypes than those with wild-type alleles. This analysis also reiterates one challenge to conducting gene–environment analyses in defining exposures. The grouping of autoimmune conditions as a single exposure was performed to increase sample size but also probably resulted in misclassification of the exposure. While taking a single well-established autoimmune condition, such as Sjogren’s disease, is preferable, the numbers of individuals with the condition would be too small for further pursuit of gene–environment interactions. Only with large-scale pooled efforts can individual autoimmune conditions be evaluated with sufficient power for gene–environment interactions.

Identifying pathways of importance

Another goal that could be achieved with gene–environment analyses is identifying which pathways (and subsets of these pathways) may be of importance for lymphomagenesis, such as in determining whether organochlorine exposure acts via an immune mechanism or DNA repair mechanism, the former is suggested by Colt et al. in a recent analysis (Table 2)[54]. Gene–environment interactions will also be important for dissecting out intersecting risk factors. A recent GWAS identified IRF4 polymorphisms associated with sun sensitivity/hair color [55]. Another study reported IRF4 polymorphisms associated with NHL [56]. With sun sensitivity associated with NHL, a recent gene–environment analysis conducted by Gathany et al. attempted to delineate whether the association between sun sensitivity and NHL was mediated by the immune gene, IRF4[57].

Identify susceptible subgroups

Similar to the well-characterized gene–environment interaction observed for bladder cancer, such analyses can clarify the role of modest exposures and identify those with elevated risk. In one analysis, no association between smoking and NHL was reported but when assessed among NAT2 intermediate/rapid-acetylators, NHL risk estimates were elevated among current cigarette smokers [58]. Other gene–environment evaluations have been reported for one-carbon metabolism genes and folate intake [59], sunlight exposure and vitamin D gene polymorphisms [60], as well as NAT enzymes and hair dye use (Table 2)[61].

Other considerations in conducting gene–environment interactions include potential geographic specificities, subtype-specificities, other cofactors of interest, age specificity – defining the susceptible population – either geographically, by age, by exposure or by genetic polymorphism.

The aforementioned are examples of evaluating higher order effects only when a main genetic effect or exposure has been identified. As more results emerge from GWAS and other studies, it is likely that an agnostic or data mining approach could be taken, particularly, if the main effects of the gene or environment are muted in the context of a population that includes both susceptible and nonsusceptible individuals.

Data emerging from GWAS clearly point to more complex interactions than historically conducted studies where one genetic variant was evaluated in the context of one exposure. Careful consideration for downstream analyses will be needed, particularly when considering gene–gene interactions and when evaluating genetic variations that influence and even predict phenotypes, diseases and environmental exposures (e.g., polymorphisms for blood type, those that affect or mimic food and satiety responsiveness, those that are linked with known NHL risk factors such as celiac disease and HLA, Crohn’s disease and NOD2/CARD15, those that reflect other phenotypes, such as height, hair color, sun sensitivity, infectious disease susceptibility, alcohol use, tobacco use, obesity and aging) [62].

Future efforts

New clues are now emerging from large consortial efforts that allow sufficient statistical power (and adequate sample sizes) for evaluating NHL by their distinct subtypes. A combination of approaches will be required to fully dissect the interactions between genes and the environment. Although case–control studies are adequate for assessing a stable biomarker, such as genetic variations, exploring other gene–environment interactions will require prospective studies, particularly for exposures such as persistent organochlorines, which would require measurements from multiple biospecimens collected over time. Behavioral risk factors are another exposure for which prospective studies may be desirable for evaluating interactions to avoid the recall bias and temporal ambiguity that potentially hamper interpretation in case–control data. Further evaluation of interactions by NHL subtypes and by molecularly defined subtypes may also shed important light on lymphomagenesis.

At present, there is an immediate need to refine exposure measurements for known risk factors and identify causal variants for already established genetic associations. Pursuing gene–environment interactions with these few well-characterized factors can shed important light on the mechanism of lymphomagenesis, reveal key biomarkers that could be used as alternate end points for early detection and treatment [62] and identify potential susceptible groups of individuals. Especially for rare NHL subtypes, the integration of intermediate phenotypes/internal exposure markers will be essential. For example, strong correlations of EBV antiviral capsid antigen (anti-VCA) antibody levels have been found in first-degree relatives from families with EBV-related lymphomas, suggesting a genetic control of anti-VCA IgG titers [63]. As high anti-VCA IgG levels are considered a marker of EBV reactivation and are associated with EBV-induced lymphoproliferation, it could potentially be useful to explore environmental and genetic factors and their potential interaction in driving elevated anti-VCA IgG titers to clarify mechanisms underlying EBV-related lymphomagenesis. Other examples of using intermediate markers include detection of H. pylori CagA biomarker levels in gastric carcinoma and MALT lymphoma, vitamin D nuclear receptor levels in prostate cancer and NHL, or evaluation of markers that reflect inflammatory potential in across several inflammation-driven cancer sites.

Large consortial efforts to evaluate these interactions by NHL subtypes will be necessary to attain adequate sample size and power for evaluating interactions. However, while consortial efforts will be important for identifying key exposures of interest in lymphomagenesis, the level of detail for pooled exposures at the moment will be broad. In such pooled efforts, the least common denominator of exposure assessment is typically used to allow the inclusion of the maximum numbers of study participants. Individual studies with detailed levels of exposures will therefore remain critical to delve into more in-depth analyses of targeted exposures.

Unraveling the interactions between environmental factors and genetic polymorphisms in NHL risk will be essential in understanding lymphomagenesis, there are clearly a number of challenges to address before we are successful.

Expert commentary

At present, the rising NHL incidence in the last half of the 20th Century and the recent plateau in rates remain unexplained. Known risk factors for NHL account for only a small proportion of all NHL cases. Research in identifying genetic variations associated with NHL is therefore important for uncovering potential mechanisms of action for lymphomagenesis, for clarifying known or hypothesized risk factors, and for identifying new exposures relevant for NHL etiology. Genome-wide association studies in NHL, being conducted in consortial settings, will be critical in identifying genes and subsequently confirming causal SNPs associated with NHL risk. These studies will be important for providing a basis for investigating gene–environment interactions. Only through large pooling or consortial efforts will there be sufficient sample sizes and power to evaluate gene–environment interactions for NHL and its heterogeneous subtypes.

Five-year view

In 5 years we will have a better understanding of the pathways, genes and genetic variations within them that play a role in NHL risk, based on ongoing GWAS and subsequent studies that will follow to identify and confirm causal SNPs. Our understanding of hypothesized exposures that affect NHL risk should also improve based on ongoing pooled analyses and biomarker-based projects in nested case–control studies within ongoing cohort studies. The conduct of these parallel efforts in consortia are critical and will provide a platform for evaluating gene–environment interactions, particularly within the heterogeneous NHL subtypes for which sufficient sample sizes and statistical power are difficult to attain in individual studies. These data will provide clues regarding exposures important for lymphomagenesis and clarify the role of known and suspected lymphomagens. These data should also reveal and further clarify biological mechanisms of interest and relevance for lymphomagenesis.

In recent years, enormous investments have been made in genotyping efforts, resulting in a rich resource of genetic data for subsequent analyses of gene–environment interactions. Similar investments, however, are also needed for improving exposure assessment and for further understanding NHL classifications. Exposure assessment will require further validation of instruments for questionnaire-based exposure assessment and biomarker-based tests of internal exposures. For the latter, cross-study validations are needed and multiplex assays that measure a number of exposures with limited sample materials would be beneficial. For example, determining immunological fingerprints to clarify the role of past infections in lymphomagenesis may employ technologies such as multiplex serology (e.g., Luminex^® technology) in large-scale epidemiological studies. Recently, data on 15 H. pylori proteins in the context of chronic atrophic gastritis and gastric cancer risk were published [64,65], and similar approaches may improve internal exposure assessment for NHL and facilitate gene–environment analysis in H. pylori-associated MALT lymphoma. Detailed phenotyping of NHL subtypes that incorporate genetic, epigenetic and other molecular data may also improve the detection of potential interactions. Defining and validating immunological, biochemical and/or molecular markers that characterize early stages of lymphoproliferative disease/lymphoproliferative precursors will improve our understanding of lymphomagenesis and potentially serve as surrogates for disease for evaluation of gene–environment interactions.

Table 1. Exposures and genes found associated with non-Hodgkin lymphoma and potential interactions.

Risk factors [24,31]		Subtype-specificity	Genes [48,50–52]	Subtype-specificity	Potential interactions
Established	>↑ Age ↑ Male gender ↑ Family history ↑ Inherited immunodeficiencies ↑ Autoimmune conditions • Sjogren’s syndrome • Celiac disease • Systemic lupus erythematosus • Severe arthritis ↑ Infectious diseases • HTLV1 • Helicobacter pylori • EBV • HHV-8 • Hepatitis C • HIV/AIDS ↑ Benzene	B-cell lymphomas DLBCL, MZ T-cell lymphomas B-cell lymphomas ATL MALT BL PEL CLL	Confirmed SNPs rs6457327 (6p21.33, near PSORS1) ↑ TNF G-308A ↑ IL10 T-3575A rs17483466 (2q13) rs13397985 (2q37.1) rs872071 (6p25.3, IRF4) rs735665 (11q24.1) rs7176508 (15q23) rs11083846 (19q13.32, PRKD2)	Follicular DLBCL DLBCL CLL CLL CLL CLL CLL CLL	rs6457327 (PSORS1) – autoimmune conditions Immune genes (including TNF, IL-10 and HLA) – infections, autoimmune conditions, occupational exposures, sunlight, obesity Oxidative stress Nutrient intake – one-carbon metabolism Detoxification genes – occupational exposures DNA repair – sunlight
Hypothesized	↑ Obesity ↓ Atopy (asthma and allergies) ↑ Mild infections (infectious mononucleosis, early birth order and transfusions) ↓ Alcohol ↓ Sunlight ↑ Serum dioxin ↑ Smoking ↑ Hair dyes ↑ Occupational exposures (serum polychlorinated biphenyl level, occupational pesticides and solvent exposures) Diet: ↑ nitrates, nitrosamines, meat, protein, ↓ vegetables, fruits, fish and green tea	DLBCL B-cell lymphomas Follicular Follicular, CLL/SLL	Candidate pathways/regions: • Proinflammatory genes • Th1/Th2/Treg genes • Innate immunity • Oxidative stress • One-carbon metabolism • DNA repair • Energy regulation • Sex hormone production • Detoxification/bioactivation • HLA region		Gender – sex hormone Obesity – energy regulation

EBV: Epstein–Barr virus; HHV:Human herpesvirus; HTLV1: Human T-cell lymphotrophic virus type 1; TCE: Tetrachloroethylene and trichloroethylene.

Table 2. Published literature on gene–environment interaction.

Gene family	Exposure	Summary of main results	Ref.
Immune genes	Organochlorine	Relationship between organochlorine exposure and NHL risk may be modified by immune genes (IFNG, IL-4, IL-16, IL-8 and IL-10)	[54]
	Autoimmune conditions	Increased DLBCL risk among persons with autoimmune conditions with variant TNF G-308A or IL-10 T-3575A allele	[53]
	Sun sensitivity	No joint effect between IRF4 and sun sensitivity in mediating NHL risk	[57]
		No joint effect between immune genes (IL-4, IL-10, IL-12A, IL-12B and TNF) and sun sensitivity	[60]
	Obesity	Increased DLBCL risk among obese persons with variant TNF G-308A or IL-10 T-3575A allele	[53]
	NSAIDs	IL-1 haplotype associated with increased B-NHL risk among nonregular NSAID users	[66]
	Birth order	Increased DLBCL risk among last born persons with variant TNF G-308A or IL-10 T-3575A allele	[53]
Energy homeostasis	BMI	Associations between leptin gene variants and follicular lymphoma independent of BMI	[67]
		Genetic interactions between LEP-G2548A and LEPR Q223R polymorphisms evidence for increased NHL risk	[68]
Oxidative stress	Vegetable/fruit intake, antioxidant intake	Two single nucleotide polymorphisms in NOS1 modified the association between total vegetable and fruit intake and risk of NHL	[69]
		NHL risk modified among those with: higher total vegetable intake and GSTM3 polymorphism; GSTP1 and higher cruciferous vegetable intake; OGG1 and higher daily zinc intake; XRCC3 and higher green leafy vegetable intake	[38]
Detoxification GST, CYP450	Tobacco	Increased DLBCL risk in nonsmokers was associated with variant allele for GSTP1 and CYP1A1; decreased risk observed for those with variant CYP1B1 allele	[70]
NAT1, NAT2	Tobacco	Compared with nonsmokers, NHL risk estimates for current cigarette smoking were increased only among NAT2 intermediate/rapid-acetylators	[58]
		Compared with nonsmokers, increased risk of NHL for current smokers without the NAT1*10 genotype but not among smokers with the NAT1*10 genotype	[70]
One-carbon metabolism	Folate, methyl donor nutrients, co-factors for C1-metabolism	Protective effect of vitamin B6 on NHL risk limited to FPGS, MTHFS. Protective association of methionine limited to FTHFD, MTHFR and MTRR	[59]
Vitamin D receptor	Sun	Decreased NHL risk with time in the midday sun dependent upon vitamin D receptor gene	[60]

DLBCL: Diffuse large B-cell lymphoma; FPGS: Folylpolygutamate synthetase; FTHFD: 10-formyl-THF dehydrogenase; MTHFS: Methenyltetrahydrofolate synthetase; MTHFR: Methylenetetrahydrofolate reductase; MTRR: Methionine synthase reductase; NAT: N-acetyl transferase; NHL: Non-Hodgkin lymphoma.

Table 3. Potential avenues for NHL research in gene–environment interactions and their respective challenges.

Study design	Recommended for	Specific challenges
Case–control study	Pooled data analyses of rare exposures, rare subtypes (i.e., InterLymph consortium)	Definition of exposure – Exposure assessment at one point in time – For some time-sensitive exposures, further confirmation in cohorts may be required – May not be adequate for some biological markers that are affected by disease or treatment
Nested case–control studies within cohort studies	Prevalent suspected risk factors that are affected by disease onset, treatment or recall bias (i.e., alcohol, diet and biomarkers [vitamin D, IgE])	Lymphoma classification
Exposure-specific cohorts Occupationally exposed cohorts (e.g., benzene, pesticides, chronic antigen-stimulating agents and UV-exposure)	Better definition of exposure biomarkers, internal and effective dose markers; (i.e., chromosomal aberrations (t14;18)(q32;q21), DNA or protein adducts, END, TRECs)	Repeated blood sampling Lymphoma classification
Infectious disease cohorts with repeat blood sampling (e.g., Helicobacter pylori, hepatitis C virus, HIV, EBV)	Identification of biologically effective dose; (i.e., CagA level, free plasma EBV DNA and EBV latency types) Identification of early markers for (pre)carcinogenesis/lymphogenesis (i.e., inflammatory index and type II mixed cryoglobulinemia)	Repeated blood sampling Lymphoma classification
Allergy/atopy cohorts	Definition of temporal sequence between allergy/atopy and NHL	Lymphoma classification
Autoimmune condition cohorts (e.g., Sjogren’s syndrome, celiac disease and systemic lupus erythematosus)	Identification of early markers/intermediate steps for lymphomagenesis Assessment of treatment effect on NHL risk; (i.e., TNF-blocker and corticosteroids)	Sample size
Immunodeficiency patient cohorts	Joint effect between common exposures and high-risk genetic background (i.e., diet, smoking and solvent exposure)	Heterogeneity in phenotypes Sample size
Solid-organ transplantation cohorts	Joint effects between intense immunosuppression and common genetic risk factors Identification of viral reactivation markers	Heterogeneity in phenotypes Sample size
Benign B-cell lymphoproliferative condition cohorts Monoclonal B-cell lymphocytosis, IgM monoclonal gammopathy of undetermined significance, t(14;18)(q32;q21) cohort	Identification of early markers/intermediate steps for lymphomagenesis (i.e., 13q14 deletions and mutated IGHV genes)	Few studies available Repeated blood sampling

EBV: Epstein–Barr virus; END: Endoreduplication; NHL: Non-Hodgkin lymphoma; TREC: T-cell receptor excision DNA circle.

Box 1. Challenges in evaluating gene–environment interactions in non-Hodgkin lymphoma in defining disease outcome, exposures and genes.

Defining the disease outcome

• • Need to consider subtype specificity (e.g., Helicobacter pylori and gastric mucosa-associated lymphoid tissue lymphoma)

• • Needs large consortial efforts because non-Hodgkin lymphoma is a rare tumor

• • No biomarker available for early detection (e.g., intermediate outcome and preclinical disease)

• • Potential heterogeneity even within subtypes (e.g., molecular subtypes)

Defining the exposures

• • Current known exposures do not explain the majority of lymphomas and cannot explain previous rise in incidence rates

• • Consortia provide broad definitions; still require detailed exposure measurements within studies

• • No biomarker – precise measure of exposure

Identifying genes

• • Modest associations to date

• • No causal variations identified

• • Need to consider other genetic factors: epigenetics, microRNA, other regulatory factors

• • Modest role for family history; no major gene

Key issues

• • The rising non-Hodgkin lymphoma (NHL) incidence in the last half of the 20th Century and the recent plateau in rates remain unexplained.

• • Although there are some striking and known risk factors for NHL, these risk factors account for only a small proportion of all NHL cases.

• • Research in identifying genetic variations associated with NHL are important for uncovering potential mechanisms of action for lymphomagenesis and identifying additional exposures that may be relevant for NHL risk.

• • Genome-wide association studies in NHL being conducted in consortial settings will be critical in identifying genes and subsequently confirming causal single nucleotide polymorphisms associated with NHL risk.

• • Evaluation of gene–environment interactions requires careful consideration of the appropriate study design, particularly for accurate exposure assessment.

• • Investigating gene–environment interactions may reveal which biological mechanisms of action are relevant and shared across exposures and whether particularly vulnerable populations exist.

• • Consortial efforts will play a critical role in evaluating gene–environment interactions, particularly for conducting evaluations that are sufficiently powered to determine if heterogeneity exists by NHL subtypes.

Financial & competing interests disclosure

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

No writing assistance was utilized in the production of this manuscript.