Interleukin-6-Related Genotypes, Body Mass Index, and Risk of Multiple Myeloma and Plasmacytoma

Multiple myeloma and plasmacytoma are neoplasms originating from monoclonal neoplastic plasma cell clones (1). Multiple myeloma consists of multifocal lesions, whereas plasmacytoma consists of a single localized lesion. Multiple myeloma, the more common of the two, has one of the poorest survival rates of any cancer. Compared with Whites, African Americans have a 2- to 3-fold increased risk, and Asians have a 50% decreased risk of multiple myeloma and plasmacytoma (2). In addition, there is evidence of an increased risk of multiple myeloma and plasmacytoma among family members (3-7). These findings suggest that genetic factors may play a role in etiology, but specific loci have not yet been identified.

One possible candidate locus is the gene encoding interleukin-6 (IL-6). IL-6 is necessary for the differentiation of immature plasmablasts into mature antibody-producing plasma cells in the bone marrow (8). Myeloma cells in culture do not survive without IL-6 (8). IL-6 antibody inhibits myeloma cell proliferation in vitro and in vivo (8), and high levels of IL-6 have been linked to higher stage (9-11) and poor prognosis (10-12) of both multiple myeloma and plasmacytoma. Furthermore, IL-6 induces polyclonal B cell activation, which can lead to hypergammaglobulinemia (13). Because the cytokine is rapidly cleared from the blood (13), circulating IL-6 levels are determined by the efficiency of expression.

The promoter region of the IL-6 gene contains several polymorphisms. The G allele of the −174 G → C single nucleotide polymorphism (SNP) has been correlated with higher levels of IL-6 (14, 15). This SNP has been examined in relation to multiple myeloma risk with null results (16-19). However, Terry et al. (13) have reported that at least two promoter polymorphisms (−572 G → C and −373 A_nT_n) in addition to the −174 G → C SNP, contribute to regulation of IL-6 levels, and that the effect of each SNP is not independent from the others. To date, these other promoter SNPs have not been examined in association with multiple myeloma.

A functional polymorphism in exon 9 of the IL-6 receptor α (IL-6rα) gene results in an amino acid substitution (Asp³⁵⁸Ala), leading to decreased shedding of membrane-bound IL-6 receptor (20). Because soluble IL-6 receptor binds to IL-6, differences in circulating receptor levels could result in differences in circulating IL-6 levels via sequestration. Thus, this polymorphism could also play a role in heritable multiple myeloma risk by mediating circulating IL-6 levels.

IL-6 is produced by adipocytes, and IL-6 levels are correlated with obesity (21). Genetic susceptibility to obesity may be related to genetic susceptibility to multiple myeloma through heritable elevated IL-6 levels. Obesity has been identified as a risk factor for multiple myeloma in three case-control studies (22-24) and in all but one of four cohort studies (25-28). The negative cohort study (25) was comprised of a relatively young population and thus had limited statistical power to examine associations with an age-related rare cancer, such as multiple myeloma. The magnitude of the positive relative risks ranged from 1.2 to 3.4, comparing obese (≥30 kg/m²) to normal-weight (<25 kg/m²) persons. Obese African-American and White veterans were found to have a risk of multiple myeloma that was 20% greater than nonobese veterans (28), although a smaller cohort study found some heterogeneity by race and sex (23). Overall, the evidence suggests that obesity is an important risk factor for multiple myeloma, although the mechanisms that account for this relationship are unknown.

Both the common IL-6 promoter SNP −174 G → C (29) and an IL-6 receptor SNP described above (30, 31) have been linked to obesity. We conducted a population-based case-control study with two control groups consisting of related and unrelated individuals to determine the role of obesity and of the reported functional IL-6 promoter and receptor polymorphisms in the etiology of multiple myeloma and plasmacytoma.

This study was approved by the University of Southern California (USC) Institutional Review Board in accordance with assurances filed with and approved by the U.S. Department of Health and Human Services. All subjects provided informed consent. The design was a case-control study with incident cases of multiple myeloma and plasmacytoma, also referred to as plasma cell neoplasms, ascertained by the USC Cancer Surveillance Program, the population-based cancer registry for Los Angeles County. Cases were residents of Los Angeles County diagnosed with primary multiple myeloma or plasmacytoma (International Classification of Diseases for Oncology, 3rd Edition 9731-9734) from October 1, 1999 through December 31, 2002 and under age 75 years at diagnosis.

A total of 312 potentially eligible cases were identified by the USC Cancer Surveillance Program during the study period; of these, 47 had died before we initiated contact, 15 could not be located, and we were denied permission by the physician of record to contact 10 other patients. The remaining 240 were contacted, and of those, 28 did not speak English, 10 were too ill to participate, 3 were not eligible to participate for various reasons, and 49 refused to participate, resulting in a total of 150 cases who participated in the study [150 / (150 + 49 + 10) = 72% response rate]. Of the participating cases, 16 (10%) were diagnosed with plasmacytoma (osseous or extraosseous) and the remaining 134 (90%) with multiple myeloma.

Two groups of controls were recruited. The first group consisted of relatives of cases (family controls). Cases were asked to enumerate all siblings and cousins, and one control per case was chosen according to a hierarchical algorithm with the following priority: (a) same-sex siblings within 5 years of age, (b) opposite-sex siblings within 5 years of age, (c) same-sex siblings ≥5 years older, (d) opposite-sex siblings ≥5 years older, (e) same-sex siblings ≥5 years younger, (f) opposite sex siblings ≥5 years younger. If no sibling was available, cousins were selected using the same hierarchy. If we were unable to identify an eligible cousin, we recruited either a parent or child of the patient. The priority was maintained even for relatives not residing in the Los Angeles County area. Questionnaires were given by telephone, and blood or buccal samples were collected by mail for relatives residing outside of Los Angeles County. One relative was identified for 127 of the 150 cases, and of these, 10 refused, resulting in a total of 117 family controls for the 150 cases [117/127 = 92% response rate]. Of these, 96 were siblings, 14 were cousins, and 7 were other first-degree relatives (children and parents).

A second group of controls consisted of population controls recruited into a multicenter study of non-Hodgkin's lymphoma that was conducted concurrently with this study. These controls were frequency matched to an expected race, age, and sex distribution of non-Hodgkin's lymphoma cases and were identified by random digit dialing.

Of the 715 eligible controls identified in Los Angeles, 10 (1%) died before contact, 120 (17%) could not be located, 3 (0.4%) were cognitively impaired, 3 (0.4%) had moved out of the area, 3 (0.4%) were too ill to be interviewed, 290 (41%) refused, and 12 (2%) did not participate for miscellaneous reasons. The remaining 274 eligible controls were interviewed for a participation rate of 48% [274/274 completed interview + 290 refused + 3 too ill]. Of those interviewed, 113 (41%) provided a buccal sample, and 144 (51%) provided a blood sample. Only non-Hodgkin's lymphoma controls selected from the Los Angeles County site who provided a blood sample with sufficient DNA (n = 126) were used as controls in this study. Demographic comparisons were made between those used in this study and those whose samples were not; there were no statistically significant differences (32).

Study participation included completion of an extensive questionnaire that included questions on basic demographics and height and weight (1 year before interview). The questions asked of the multiple myeloma patients and their family controls were identical to those asked of the population controls from the non-Hodgkin's lymphoma study. Questionnaires were given by an in-person interview, except for 65 interviews that were conducted with relatives of cases living outside Los Angeles County by telephone. In this study, we examined the following basic demographic variables: gender, age (categorized into four categories: 23-51, 52-60, 61-67, 68-84 years), race/ethnicity [categorized as non-Latino White (White), African American, Latino White (Latino), Asian/Pacific Islander/Other (Asian/Other)], years of education [categorized as <12 years ± vocational school (“vocational/high school”), 13-15 years (“some college”), ≥16 years (“college and above”)], and body mass index (BMI; calculated as kg/m²). BMI was classified into three categories, and BMI < 25 kg/m² was designated as the reference category. Obesity was defined according to WHO guidelines as ≥30 kg/m² and over weight from 25 to 29 kg/m². Subjects with a BMI of <15 or >55 kg/m² were excluded (n = 3); all of these were population controls.

DNA was extracted from cases' and family controls' samples using a QIAamp 96 Blood kit (Qiagen, Valencia, CA; USC Genomics Core) and from blood clots and buffy coats from the population controls' samples at the BBI Biotech Repository (Gaithersburg, MD) using Puregene Autopure DNA extraction kits (Gentra Systems, Minneapolis, MN). DNA was isolated, and at least one SNP was successfully genotyped from 150 of 150 cases, 112 of 117 family controls, and 126 of 126 population controls. (Five relatives of cases had insufficient DNA from self-administered buccal swab collections.) The −174 and −597 IL-6 promoter genotyping assays on the population controls were conducted at the National Cancer Institute genotyping laboratories (32); otherwise, the population control samples were analyzed according to the methods described below, along with all samples from the cases and family controls, at the USC Genomics Core Laboratory.

The −174 G → C SNP was determined using the fluorogenic 5′-nuclease assay (Taqman assay; ref. 33). The Taqman assays were done using a Taqman PCR Core Reagent kit (Applied Biosystems, Foster City, CA) according to manufacturer's instructions. The oligonucleotide primers for amplification of the polymorphic region were GC112 forward (5′-CAATGACGACCTAAGCTGCACTT-3′) and GC112 reverse (5′-GGGCTGATTGGAAACCTTATTAAGAT-3′). In addition, the fluorogenic oligonucleotide MGB probes used to detect each of the alleles were GC112F (5′-CTTTAGCATCGCAAGAC-3′) labeled with 6-FAM to detect the C allele and GC112V (5′-CTTTAGCATGGCAAGAC-3′) labeled with VIC to detect the G allele (Applied Biosystems). PCR amplification using ∼10 ng of genomic DNA was done with an initial step of 95°C for 10 minutes followed by 50 cycles of 95°C for 25 seconds and 60°C for 1 minute. The fluorescence profile of each well was measured in an ABI 7900HT Sequence Detection System, and the results were analyzed with Sequence Detection Software (Applied Biosystems). Experimental samples were compared with 12 controls to identify the three genotypes at each locus (C/C, C/G, and G/G). Any samples that were outside the variables defined by the controls were identified as noninformative and were retested. The IL-6 receptor SNP in exon 9 (30) was determined using RFLP. Genomic DNA was amplified using 5′-GCTTTTGACAGCACCAGCTAAG-3′ and 5′-GGACCCATCTCACCTCAGAAC-3′ as primers and digested with HindIII to identify the Ala (A) and Asp (D) coding alleles.

The −597 G → A, −572 G → C, and the variable number of tandem repeats (VNTR) −373 A_nT_n SNPs were determined by direct sequencing (USC Genomics Core). The region of the gene containing the three SNPs was amplified by PCR using primers GC108 forward (5′-AGCAGCCAACCTCCTCTAAG-3′) and GC108 reverse (5′-TTGGCATGTCTTGACAAAGAG-3′). DNA sequencing was done using primer GC108seq (5′-AGCCAACCTCCTCTAAGTGG-3′), ∼10 to 20 ng of purified PCR product, fluorescently labeled dideoxynucleotide triphosphates (ABI Dye Terminator Sequencing kit, Applied Biosystems), and capillary electrophoresis (ABI 3730xl).

χ² tests were used to compare the distribution of demographic variables between cases and controls. Hardy-Weinberg equilibrium was evaluated in family and population controls separately, and in African-American and White controls separately, using a one-degree-of-freedom χ² test for the SNPs, and a three-degree-of-freedom χ² test for the VNTR. Linkage disequilibrium between polymorphic sites was estimated using the squared correlation coefficient, calculated using the EM algorithm²³ as implemented in PROC ALLELE (SAS version 9.1) (34).

To assess factors influencing risk of plasma cell neoplasms, odds ratios (OR) were calculated using conditional (for cases versus family controls) and unconditional (for cases versus population controls) multivariable logistic regression (SAS version 9.1) (35, 36). (Because not all cases had eligible relatives, the case numbers are ∼25% smaller in the case-family control comparisons than in the case-population control comparisons).

Homozygotes for the minor allele were collapsed with heterozygotes due to sparse data. For the VNTR, classification based on exact numbers of repeats was not feasible due to small numbers of subjects with each genotype. Based on Terry et al.'s suggestion that the number of “A” alleles might be the key factor in determining the effect on IL-6 transcription (13), we categorized the genotypes accordingly using the most common genotype as the reference group. Due to small numbers, the genetic analyses were limited to assessment of individual SNPs only; no haplotype or gene-gene interaction analyses were done. For each assay, some samples could not be genotyped; thus, the number of subjects in each case/control stratum varies again due to the small sample size, risk by genotype was not assessed for individual racial/ethnic groups. The effect of BMI on multiple myeloma and plasmacytoma risk was assessed using the lowest BMI group (<25 kg/m²) as the reference group. All models were adjusted for age (continuous), gender, education [≤12 years (high school ± vocational school), 13-15 years (some college), ≥16 years (completed college/professional/graduate school), fit as an ordinal variable], race/ethnicity (African American versus non–African American), and, for genetic analyses, BMI (<25, 25-29.9, ≥30 kg/m²).

The association between obesity and IL-6 genotype was assessed using unconditional multivariable logistic regression, designating obese subjects (BMI ≥ 30 kg/m²) as “cases” and nonobese subjects as “controls,” among cases, family controls, and population controls separately. The model included age, gender, education, and race/ethnicity, as described above for other analyses.

Demographic characteristics of participants are shown in Table 1. Family controls were similar to cases with respect to education and race/ethnicity distribution but had a higher percentage of women. Population controls had a similar gender distribution to that of the cases, but they had lower educational achievement. Cases tended to be slightly older (median age = 61 years) than controls. Cases and family controls were more likely to be obese compared with population controls.

There were no differences between genotype distributions for patients with multiple myeloma and those with plasmacytoma; thus, these cases were considered together as a group (plasma cell neoplasms). All genotypes, except those for Asians/others (n = 15 family controls and 5 population controls), were in Hardy-Weinberg equilibrium for both control groups. The lack of Hardy-Weinberg equilibrium in Asians/others was probably due to the small numbers of subjects. The IL-6 promoter SNPs −174 and −597 were in strong linkage disequilibrium among population and family controls with r² values ranging from 0.85 to 1.00 in Whites and African Americans. The −572 SNP was not in linkage disequilibrium with these two SNPs, with r² values ranging from 0.006 to 0.05. Allele frequencies of the three promoter SNPs −174, −597, and −572 were similar in the two control groups (Table 2), and similar to those published for U.S. and European populations (37-39). The −373 A_nT_n polymorphic alleles were also distributed similarly in the two control groups (Table 2); no published data currently exist on ethnic-specific distributions for comparison.

The variant allele of the IL-6 promoter SNP −572 was associated with a roughly 2-fold increased risk of plasma cell neoplasms when cases were compared with family [OR, 1.8; 95% confidence interval (95% CI), 0.7-4.7] or population controls (OR, 2.4; 95% CI, 1.2-4.7; Table 2). Because Asian/other controls were not in Hardy-Weinberg equilibrium, we repeated the analyses excluding these case and controls; the association was unchanged. The −373 A_nT_n 9A/9A genotype was associated with a 60% to 70% decreased risk compared with the most common genotype; however, few individuals carried this genotype (6 cases, 11 family controls, and 12 population controls; Table 2). When Asians/others were excluded from the analysis, the ORs were unchanged, but the CIs did not exclude 1.0 (data not shown). None of the other SNPs were associated with risk of plasma cell neoplasms overall or with Asians/others removed from each comparison.

Increasing BMI was modestly associated with increasing risk of plasma cell neoplasms (Table 3). The effects were stronger in the case-population control comparison (P_trend = 0.08). Relative to the lowest weight category, being overweight or obese increased risk by 60% and 80%, respectively. Risks of similar magnitude were observed for both African Americans and Whites when examined separately, but the CIs were considerably wider (for African Americans, <25 versus ≥30 kg/m²: OR, 1.8, 95% CI, 0.5-7.0; for Whites, <25 versus ≥30 kg/m²: OR, 2.0; 95% CI, 0.8-5.2). When cases were compared with family controls, a 30% and 40% nonsignificant increase in risk was observed for overweight and obese subjects, respectively, relative to the lowest weight category.

The IL-6rα SNP was strongly associated with obesity among population controls but not among family controls or cases (Table 4). Population controls who carried the variant allele had a ≥5-fold increased statistically significant risk of being obese compared with noncarriers (95% CI, 1.7-17.4; Table 4). This association was based on only seven obese population controls who carried the D allele. None of the other SNPs were associated with obesity in any group.

Results were unchanged when analyses were restricted to cases of multiple myeloma only. The number of patients with plasmacytoma (n = 16) was too small to analyze separately.

This study confirms several previously reported findings. First, we found no association between risk of multiple myeloma and the well-studied −174 G → C IL-6 promoter SNP, consistent with other studies (16-19). In our study population, as in others (13, 37), the IL-6 promoter SNPs −174 G → C and −597 G → A were in strong linkage disequilibrium. In addition, we found the expected association between plasma cell neoplasms and weight, with risk increasing with increasing BMI (22, 23, 26, 27).

We found a previously unreported association between the variant allele of the IL-6 promoter polymorphism −572 and an increased risk. In other studies unrelated to multiple myeloma, the −572 polymorphism was a predictor of bone mineral density (40, 41), metabolic syndrome (37), and other conditions (42, 43). Studies of genotype-phenotype correlation regarding the −572 SNP, as with the −174 SNP, have produced varied results. Two studies reported that the −572 C allele was associated with higher serum levels of IL-6 (38). However, some studies have found no correlation between the −572 locus and IL-6 serum levels (14, 29). Ferrari et al. (40) found higher C-reactive protein levels, which reflect IL-6 induction, in postmenopausal women with the C allele.

The −572 SNP lies between the −174 and −597 SNPs, which are in linkage disequilibrium, but the −572 SNP is not in linkage disequilibrium with either of the other two. This implies that the −572 SNP possibly developed from a more recent mutation, occurring on a specific allele of the other two SNPs. The −572 locus does not have strong homology to any known transcription factor binding site; however, there is a glucocorticoid receptor element located at position −557 to −552 (13). The influence of the glucocorticoid receptor was apparent in a study by Ferrari et al. (40). In the presence of dexamethasone stimulation, −572 C alleles were associated with a larger increase in IL-1β- and tumor necrosis factor-α-induced IL-6 secretion, compared with that obtained from persons carrying −572 G alleles (40). Thus, it is possible that steroid binding plays a role in regulating IL-6 secretion. In addition, because the −572 SNP has been associated with bone density variation in two studies, it is possible that this locus is contributing to regulation of osteoclast activity in this disease.

We also found evidence of decreased risk associated with at least one of the −373 A_nT_n VNTR genotypes, albeit based on small numbers. Although Terry et al. (13) found that the VNTR alleles influenced IL-6 transcription, the relationship was not straightforward because the effect depended on other SNPs in the promoter. Thus, the significance of this finding cannot adequately be evaluated without knowledge of the entire haplotype, which was not possible in this study due to its small sample size.

We found that obesity increased risk by 40% to 80% similar to the magnitude reported in other studies (22). However, obesity did not seem to be consistently associated with any of the genotypes that we examined, in contrast to previous reports (30). Although we found a strong association between the IL-6 rα and risk of obesity among population controls, as reported elsewhere (30), this observation was based on only seven obese subjects positive for the IL-6rα D allele. Because of this, and because we did not observe a similar relationship among either cases or family controls, we conclude that chance is a likely explanation.

It is possible that we did not examine the relevant genotypes. At least one other IL-6 receptor VNTR has been associated with obesity and possibly IL-6 levels (31). In addition, because IL-6 produced by adipocytes (21) accounts for up to 30% of circulating IL-6 (21, 44), non-heritable obesity could produce elevated IL-6 levels independent of IL-6 genotype.

Other genetic pathways may be involved. Insulin-like growth factor-1 promotes myeloma cell proliferation independent of IL-6 (45) and is related to prognosis (46). Promoter SNPs in the gene coding for insulin-like growth factor-1 are correlated with plasma levels of these proteins (47) and also have been linked with increasing amounts of body fat (48). Genes encoding other proteins in the lipid-inflammation pathway, such as tumor necrosis factor-α, are also potential candidates.

The case sample may not have been representative of most cases because 15% of eligible cases died before we could obtain samples, and because the age range of the study subjects was skewed toward younger patients (median age for cases in the study was 61 versus 71 years old for Los Angeles County cases overall diagnosed during the same time period). Furthermore, because high IL-6 levels are known to be related to poor prognosis, it is possible that the surviving cases represent a genetically skewed sample, and that the true risk estimate for the high-secreting IL-6 alleles is even higher.

IL-6 alleles may also be related to survival in general, which could affect prevalence rates in available living controls. There are conflicting reports of decreased life span associated with either the G or C allele of the −174 IL-6 promoter polymorphism, depending on the source population (49-51). Tan et al. evaluated longevity and IL-6 promoter haplotypes in a Danish population of younger (<70 years old) and older (93 years old) individuals and found a deficit of a haplotype containing the −572 G allele and the −373 A₈T₁₂ allele in the older individuals (52). Because our subjects were relatively young (<75 years old), differential survival by genotype probably did not appreciably affect our results.

This study is limited by the relatively small sample size, which precluded examination of ethnic-specific genetic associations, gene-gene interactions, haplotype, and genotype-BMI interaction. The low participation rate among population controls could potentially introduce selection bias, if the exposures of interest are related to determinants of participation. However, comparisons of cases with both family controls and unrelated population controls produced remarkably similar effect estimates. Furthermore, both control groups had nearly identical genotype distributions, which were also similar to those observed in control groups from other studies, providing evidence that they were generally representative.

In summary, we present evidence to suggest that the −572 IL-6 promoter SNP may predict risk of multiple myeloma and plasmacytoma. The −373 A_nT_n VNTR also seemed to be associated with risk, but the result was based on small numbers. High BMI may be a risk factor for plasma cell neoplasms, but the mechanism seems to be independent of these genotypes.

We thank Anna Gilmore and Roger Mathison for giving input into questionnaire design and doing the interviews and DNA collection for all cases and controls involved in this study, Audrey Cairns for providing clerical support, and Lonn Irish and Michael Stagner (Information Management Systems, Silver Spring, MD) for providing database management and analysis support.