Human Immunology 73 (2012) 1155–1158

Contents lists available at SciVerse ScienceDirect

www.ashi-hla.org

journal homepage: www.elsevier .com/locate /humimm

Genetic markers of immunoglobulin G and susceptibility to breast cancer

Janardan P. Pandey a,⇑, Emily Kistner-Griffin b, Motoki Iwasaki c, Shizhong Bu a, Ray Deepe a, Laurel Black a,Yoshio Kasuga d, Gerson S. Hamada e, Shoichiro Tsugane c

a Department of Microbiology and Immunology, Medical University of South Carolina, Charleston, SC, United Statesb Department of Medicine, Medical University of South Carolina, Charleston, SC, United Statesc Epidemiology and Prevention Division, Research Center for Cancer Prevention and Screening, National Cancer Center, Tokyo, Japand Department of Surgery, Nagano Matsushiro General Hospital, Nagano, Japane Nikkei Disease Prevention Center, São Paulo, Brazil

a r t i c l e i n f o a b s t r a c t

Article history:Received 6 June 2012Accepted 30 July 2012Available online 9 August 2012

0198-8859/$36.00 - see front matter � 2012 Americahttp://dx.doi.org/10.1016/j.humimm.2012.07.340

⇑ Corresponding author. Address: Department of MMedical University of South Carolina, Charleston, SCFax: +1 843 792 4882.

E-mail address: pandeyj@musc.edu (J.P. Pandey).

Immunoglobulin GM allotypes, antigenic determinants of c chains, are encoded by three very closelylinked codominant genes on chromosome 14q32. Particular GM alleles/haplotypes are associated withantibody responses to certain tumor antigens and contribute to the cytotoxicity of breast cancer cells,but their possible role in susceptibility to this malignancy has not been adequately examined. Using amatched case-control design, we genotyped a large (1710 subjects) study population from Japan andBrazil for several GM alleles to determine whether these determinants are associated with susceptibilityto breast cancer. After adjusting for the potential confounders, the GM 3 allele of IgG1 was significantlyassociated with susceptibility to breast cancer in white subjects from Brazil (OR = 2.07, CI 1.16–3.71;p = 0.0147). These data show that Caucasians with the GM 3 allele are over twice as likely to developbreast cancer as those who lack this allele. Since this allele modulates an immune evasion strategy ofcytomegalovirus, the results also shed light on the possible mechanism underlying the reported involve-ment of this virus in the etiology of breast cancer.� 2012 American Society for Histocompatibility and Immunogenetics. Published by Elsevier Inc. All rights

reserved.

1. Introduction

Polymorphic hereditary antigenic determinants expressed onimmunoglobulin c chains are called GM allotypes. Encoded bythree very closely linked and highly homologous codominantgenes on chromosome 14q32, they are expressed on the constantregion of c1, c2, and c3 chains [1]. Linkage disequilibrium in theGM system within a racial group is almost absolute and majorraces are characterized by a unique array of GM haplotypes [2].GM allotypes are associated with many cancers [3–7], but theirrole in susceptibility to breast cancer has not been adequatelyexamined.

Using a matched case-control study design and a large studypopulation from Japan and Brazil, we aimed to determine whetheror not GM alleles are risk factors for the development of breast can-cer. Since the majority of the GM alleles are expressed in the Fc re-gion of c chains, we also wished to investigate whether particularFc (GM), FccRIIa, and FccRIIIa alleles epistatically contribute to therisk of breast cancer.

n Society for Histocompatibility an

icrobiology and Immunology,29425-2230, United States.

2. Materials and methods

2.1. Study subjects

The experimental design, recruitment criteria, and thedemographics of the study population have been described indetail elsewhere [8]. Briefly, breast cancer patients were recruitedbetween 2001 and 2005 at four hospitals in Nagano, Japan, andbetween 2001 and 2006 at eight hospitals in São Paulo, Brazil.Cancer-free controls were matched to case patients by ethnicity,residential area during the study period, and age (within3–5 years). Detailed data were collected on family history of can-cer, menstrual and reproductive history, anthropometric factors,physical activity, and smoking habits. Estrogen and progesteronehormone receptor status was also determined. The study protocolwas approved by the Institutional Review Boards of the respectiveinstitutions. Blood from cases and controls was collected afterinformed consent. The study population consisted of the follow-ing: 258 case-control pairs of Caucasian descent (Brazil), 40case-control pairs of African descent (Brazil), 80 case-control pairsof Japanese descent (Brazil), 80 case-control pairs from the Brazil-ian mulatto population, 397 case-control pairs from Nagano,Japan.

d Immunogenetics. Published by Elsevier Inc. All rights reserved.

1156 J.P. Pandey et al. / Human Immunology 73 (2012) 1155–1158

2.2. GM genotyping

DNA for genotyping was isolated from peripheral blood using astandard protocol (Qiagen-Kit method). For the determination ofIgG1 markers GM 3 and 17 (arginine to lysine, a G to A substitutionin the CH1 region of the c1 gene), we used a pre-designed Taq-Man� genotyping assay from Applied Biosystems Inc. (Foster City,CA). The probe specific to GM 3 allele was labeled with FAM fluo-rescent at the 50 end and with nonfluorescent quencher at the 30

end. The probe specific to GM 17 allele was labeled with VIC fluo-rescent at the 50 end and with nonfluorescent quencher at the 30

end.GM 23—valine to methionine, a G to A substitution in the CH2

region of the c2 gene—was determined by a nested PCR-RFLPmethod. In brief, a 915 bp region of the c2 gene that incorporatesthe sites for the allelic substitutions was amplified as described byBrusco et al. [9], using the following primers:

50 AAATGTTGTGTCGAGTGCCC 30 and 50 GGCTTGCCGGCCGTGG-CAC 30. A 197 bp segment was further amplified from this 915 bpfragment using the following primers:

50 GCACCACCTGTGGCAGGACC 30 and 50 TTGAACTGCTCCTCCCGTGG 30. After digestion of the amplified product with the restrictionenzyme NlaIII, the following products corresponding to the threegenotypes were obtained:

GM 23+

90 bp, 63 bp, 44 bp

GM 23�

134 bp, 63 bp GM 23+,23� 134 bp, 90 bp, 63 bp, 44 bp

For the determination of IgG3 markers GM 5 and 21, the c3gene containing the allelic sites was amplified [10] using the fol-lowing primers:

50 ACCCAAGGATACCCTTATGATT 30 and 50 GAGGCTCTTCTGCGTGAAGC 30. The amplified product (685 bp) was digested with therestriction enzyme MspA1I. The resulting products correspondingto the three genotypes are as follows:

GM 21

327 bp, 295 bp, 63 bp

GM 5

171 bp, 158 bp, 156 bp, 137 bp, 63 bp GM 5,21 327 bp, 295 bp, 171 bp, 158 bp, 156 bp,

137 bp, 63 bp

In addition to the controls representing the three genotypes forthe marker, we also used a blank (no genomic DNA, only primersand the PCR master mix) control and a DNA molecular weightmarker.

2.3. FccR genotyping

The activating receptors FccRIIa and FccRIIIa are geneticallypolymorphic: a change in the nucleotide at position 497 of FccRIIagene from A to G results in change of amino acid histidine to argi-nine (H/R131); a change in the nucleotide at position 559 of theFccRIIIa gene from T to G results in phenylalanine to valine substi-tution (F/V158). The single nucleotide polymorphisms (SNP)responsible for the allelic variation were previously determinedby the TaqMan� genotyping assays from Applied Biosystems Inc.[8].

Genotyping was done blinded to the case-control and group sta-tus of the subjects.

2.4. Statistical analysis

Genotype frequencies were in Hardy–Weinberg equilibrium inall groups except the mulatto population, which could be due to

population admixture. This group was excluded from further anal-yses. Conditional logistic regression models were constructedwithin each population group. Potential confounders—family his-tory of breast cancer, history of benign breast disease, menopausalstatus and age at menopause, number of births, age at first birth,breast feeding, alcohol drinking, smoking status, moderate physicalactivity in the past 5 years, vitamin supplement use, age at menar-che, body mass index—were considered for inclusion as covariatesin the models of genetic association, and a backwards regressionapproach with an a = 0.10 inclusion level was implemented. Testsof genotype models (2df tests with no assumptions about inheri-tance models) as well as tests of additive effects (1df) were con-structed. In all models, statistical significance was defined asp < 0.05. All reported p values are two-sided.

3. Results

The results of the tests of associations between GM genotypesand risk of breast cancer in various population groups are pre-sented in Tables 1–4. Allelic variation at the IgG1 GM 3/17 locuscontributed to the risk of developing breast cancer in white sub-jects from Brazil (Table 1). The association was significant for boththe genotype model, which assumed no particular mode of inher-itance, and the additive model, which assumed that the alleles con-tribute to the risk additively on the logit scale. Compared tosubjects who were homozygous for GM 17, the GM 3 homozygoteswere over twice as likely to develop breast cancer (OR = 2.07, CI1.16–3.71; p = 0.0147). This association would remain significanteven after a conservative correction for multiple testing[p = 0.0441 (0.0147 � 3)]. The GM 5 allele, which is in significantlinkage disequilibrium with the GM 3 allele in whites, was alsoassociated with breast cancer at a borderline significance(OR = 2.14, CI 1.04–4.41; p = 0.0526), which would become non-significant if adjusted for multiple testing. GM genotypes werenot associated with susceptibility to breast cancer in other groups(Tables 2–4).

In our previous investigations involving this study population,we did not find a significant association between any FccRIIa andFccRIIIa genotypes and susceptibility to breast cancer [8]. In thepresent investigation, we tested the possibility that FccR geno-types do contribute to the development of breast cancer, but onlyin the presence of certain Fcc (GM) alleles. An interaction analysisbetween GM and FccR genotypes showed that in the Japanese liv-ing in Brazil, this appears to be the case (Table 5). GM 23 genotypesinteracted epistatically with FccRIIIa (V/F) genotypes and contrib-uted to the risk of breast cancer (p = 0.0390). This apparent associ-ation is driven by the contrasting (protective and risk) effects ofGM 23 genotypes in the presence of FccRIIIa V allele: The OR asso-ciated with the inheritance of one copy of the V allele for individ-uals carrying the heterozygous GM 23+/GM 23� genotype is 0.13(protective), while the OR associated with the inheritance of onecopy of the V allele for individuals carrying the homozygous GM23�/GM 23� genotype is 1.48 (susceptibility). No other interactiveeffects were found (data not shown).

4. Discussion

The results presented here show a distinct association betweenthe GM 3 allele of IgG1 and susceptibility to breast cancer in a Bra-zilian white population. This allele acts additively to increase therisk of breast cancer over two fold compared to the alternativeGM 17 allele at this locus. There are at least two potential mecha-nisms through which GM alleles could be associated with suscep-tibility to breast cancer. The GM 3 allele could itself modulate therisk of breast cancer, possibly through its effect on immunity to self

Table 1Tests of associations between GM genotypes and breast cancer risk in whites (Brazil).

GM genotypes Genotype counts Odds ratiosa (95% CI) P-values

Cases Control 1 Copy 2 Copies Genotype Additive

(3/3, 3/17, 17/17) 92/123/43 73/120/65 1.61 (0.94–2.74) 2.07 (1.16–3.71) 0.0457 0.0147(23+/+,23+/�,23�/�) 48/110/100 35/105/118 1.28 (0.84–1.96) 1.70 (0.93–3.08) 0.1906 0.0687(5/5,5 /21,21/21) 143/93/20 125/103/29 1.80 (0.86–3.79) 2.14 (1.04–4.41) 0.1140 0.0526

a Reference allele (17/17,23�/�,21/21), constructed from the genotype model; regression models included number of births, alcohol consumption, smoking status, and BMIas covariates.

Table 2Tests of associations between GM genotypes and breast cancer risk in blacks (Brazil).

GM genotypes Genotype counts Odds ratiosa (95% CI) P-values

Cases Control 1 Copy 2 Copies Genotype Additive

(3/3, 3/17, 17/17) 2/17/21 4/16/20 0.84 (0.31–2.30) 0.44 (0.07–2.76) 0.6858 0.4363(23+/+,23+/�,23�/�) 0/12/28 2/11/27 0.89 (0.34–2.30) NA 0.9710 0.3978(5/5,5 /21,21/21) 24/13/2 24/13/3 1.95 (0.17–21.95) 2.10 (0.18–24.99) 0.8420 0.6836

a Reference allele (17/17,23�/�,21/21), constructed from the genotype model; no covariates met the p < 0.10 significance level for inclusion in the model.

Table 3Tests of associations between GM genotypes and breast cancer risk in Japanese (Nagano).

GM genotypes Genotype counts Odds ratiosa (95% CI) P-values

Cases Control 1 Copy 2 Copies Genotype Additive

(3/3, 3/17, 17/17) 4/69/324 3/56/338 1.26 (0.80–1.98) 1.39 (0.18–10.58) 0.5859 0.3038(23+/+,23+/�,23�/�) 4/63/330 3/56/338 1.17 (0.74–1.86) 1.36 (0.18–10.39) 0.7776 0.4782(5/5,5 /21,21/21) 8/70/318 5/62/330 1.16 (0.75–1.80) 1.55 (0.40–6.02) 0.6779 0.3871

a Reference allele (17/17,23�/�,21/21), constructed from the genotype model; regression models included family history, number of births, menopausal status, age atmenopause, breast feeding, moderate physical activity in the past five years, and smoking status as covariates.

Table 4Tests of associations between GM genotypes and breast cancer risk in subjects of Japanese descent living in Brazil.

GM genotypes Genotype counts Odds ratiosa (95% CI) P-values

Cases Control 1 Copy 2 Copies Genotype Additive

(3/3, 3/17, 17/17) 1/13/66 0/13/67 0.81 (0.33–1.97) NA 0.8995 0.9933(23+/+,23+/�,23�/�) 0/9/71 0/11/69 0.65 (0.23–1.82) NA 0.4109 0.4109(5/5,5 /21,21/21) 0/14/66 0/16/64 0.81 (0.34–1.93) NA 0.6329 0.6329

a Reference allele (17/17,23�/�,21/21), constructed from the genotype model; regression models included number of births as a covariate.

Table 5Tests of interactions between GM and FccR genotypes in subjects of Japanese descentliving in Brazil.

GM genotypes FccR locus P-valuesa

(3/3, 3/17, 17/17) FccRIIa 0.4871FccRIIIa 0.1853

(23+/+,23+/�,23�/�) FccRIIa 0.6666FccRIIIa 0.0390

(5/5,5/21,21/21) FccRIIa 0.7843FccRIIIa 0.4372

a Reference allele (17/17,23�/�,21/21), constructed from the genotype model;regression models included number of births as a covariate.

J.P. Pandey et al. / Human Immunology 73 (2012) 1155–1158 1157

and non-self antigens that may be relevant to the etiopathogenesisof this malignancy. Alternatively, there may be another locus forsusceptibility to breast cancer on chromosome 14, distinct fromGM, whose alleles are in significant linkage disequilibrium withthose of the GM loci. This putative linkage disequilibrium couldgive rise to the associations observed.

The most relevant among the self-antigens, immunity to whichis influenced by GM alleles, is epidermal growth factor receptor 2

(HER2). This tumor antigen is overexpressed in 25 to 30% of breastcancer patients, and is associated with poor prognosis. ParticularGM alleles, including GM 3 and 5, are associated with natural anti-body responsiveness to this antigen [11].

The most relevant among the non-self factors, immunity towhich is influenced by GM alleles, is human cytomegalovirus(HCMV). Increasing evidence implicates HCMV in the etiopatho-genesis of breast cancer. Elevation in serum HCMV IgG antibodylevels is reported to precede the development of breast cancer[12]. Evidence of viral expression has been found in over 97% ofneoplastic breast epithelium [13]. HCMV is endemic, affecting50–100% of the world population. The question arises: How couldsuch a common virus cause breast cancer in only a subset of thoseinfected? The GM alleles could, at least in part, explain the vast dis-crepancy in HCMV seroprevalence and the prevalence of breastcancer. These determinants modulate certain viral immune eva-sion strategies [14], a property that could explain their involve-ment in susceptibility to breast cancer.

HCMV has evolved a large repertoire of immune evasion strat-egies. One strategy involves generating two proteins—encoded bygenes TRL11/IRL11 and UL119-UL118—that have functional proper-ties of FccR [15], which may enable the virus to evade host immu-

1158 J.P. Pandey et al. / Human Immunology 73 (2012) 1155–1158

nosurveillance by evading the effector consequences of antibodybinding, such as antibody-dependent cellular cytotoxicity (ADCC),complement-dependent neutralization, and phagocytosis. Interest-ingly, HCMV TRL11/IRL11-encoded FccR has significantly higheraffinity for IgG1 proteins expressing the GM 3+,1�,2� allotypesthan for those expressing the allelic GM 17+,1+,2+ allotypes [14].It follows that subjects with the GM 3+,1�,2� allotypes would bemore likely to have their Fc domains scavenged, thereby reducingtheir immunological competence to eliminate the virus and virally-infected cells through ADCC and other Fc-mediated effector mech-anisms. Consequently, these subjects would be more likely to besusceptible to HCMV-spurred diseases. Our finding of a significantassociation between the GM 3 allele and susceptibility to breastcancer is consistent with this prediction. An additional mechanismcould involve the influence of GM allotypes on humoral immunityto HCMV. These determinants influence antibody responsivenessto HCMV epitopes in patients with scleroderma, an autoimmunedisease with suspected HCMV etiology [16]. Whether this relation-ship also holds for breast cancer needs to be investigated.

As mentioned above, the association observed here can also beexplained by postulating the presence of a breast cancer suscepti-bility locus whose alleles are in significant linkage disequilibriumwith those of the GM loci. It is interesting to note that breast cancermetastasis-related genes have been localized in chromosome14q32, the region that also harbors GM genes [17]. Additionally,genome-wide association studies (GWAS) have identified breastcancer risk genes on this chromosome in several different popula-tion groups. Such GWAS [18], however, cannot evaluate theinvolvement of GM alleles, as these determinants are not includedin the current genotyping platforms. To our knowledge, they arenot being tagged by any SNPs that are included in the platforms.Furthermore, GM alleles cannot even be imputed because theywere not genotyped in the HapMap panel [19]. IgG gene segmentsharboring GM genes are highly homologous and apparently notamenable to the high throughput genotyping technology; thisattribute may have contributed to their exclusion from the geno-typing panels.

Although we did not find a strong interactive/epistatic effect ofFcc (GM) and FccR alleles on breast cancer in this study, there is asound biological rationale for extending this investigation at a lar-ger scale in future studies: Particular alleles at these loci jointlycontribute to the cytotoxicity of breast cancer cells by ADCC [20],a major host immunosurveillance mechanism against tumors aswell as the leading mechanism underlying the clinical efficacy oftherapeutic antibodies such as trastuzumab, which targets HER2.Detection of epistasis requires a large sample size, and the presentinvestigation was most likely underpowered.

We found a significant association between GM alleles andbreast cancer in white subjects from Brazil but not in other groups.The lack of association in the Japanese population is consistent witha previous report in this group [3]. The reasons for these racialdifferences in disease associations are not clear. Linkage disequilib-rium between GM alleles in the Japanese is different from that inwhites or blacks, resulting in distinct arrays of GM haplotypes invarious groups. It follows that linkage disequilibrium between anyputative risk-conferring genes for breast cancer and GM allelesmight also be different in these groups, contributing to theethnic differences in genetic associations. Multiple genetic andnon-genetic factors probably contribute to the risk of breast cancer,and racial differences in these factors may contribute to the differ-

ences in the observed associations. To our knowledge, this is the firstreport implicating GM genes in susceptibility to breast cancer. Itneeds to be replicated in a large multiethnic study population.

Acknowledgments

This work was supported in part by a Grant from the US Depart-ment of Defense (W81XWH-08-1-0373) and by a Grant-in-Aid forResearch on Risk of Chemical Substances from the Ministry ofHealth, Labor and Welfare of Japan, and Grants-in-Aid for ScientificResearch on Priority Areas (17015049).

References

[1] Jefferis R, Lefranc MP. Human immunoglobulin allotypes: possibleimplications for immunogenicity. MAbs 2009;1:332–8.

[2] Dugoujon JM, Hazout S, Loirat F, Mourrieras B, Crouau-Roy B, Sanchez-MazasA. GM haplotype diversity of 82 populations over the world suggests acentrifugal model of human migrations. Am J Phys Anthropol2004;125:175–92.

[3] Nakao Y, Matsumoto H, Miyazaki T, Watanabe S, Mukojima T, Kawashima R,et al. Immunoglobulin G heavy-chain allotypes as possible genetic markers forhuman cancer. J Natl Cancer Inst 1981;67:47–50.

[4] Ilic V, Milosevic-Jovcic N, Markovic D, Petrovic S, Stefanovic G. A biased Gmhaplotype and Gm paraprotein allotype in multiple myeloma suggests a rolefor the Gm system in myeloma development. Int J Immunogenet2007;34:119–25.

[5] Pandey JP, Johnson AH, Fudenberg HH, Amos DB, Gutterman JU, Hersh EM. HLAantigens and immunoglobulin allotypes in patients with malignant melanoma.Hum Immunol 1981;2:185–90.

[6] Morell A, Scherz R, Käser H, Skvaril F. Evidence for an association betweenuncommon Gm phenotypes and neuroblastoma. Lancet 1977;1:23–4.

[7] Pandey JP, Ebbesen P, Bülow S, Svendsen LB, Fudenberg HH. IgG heavy-chain(Gm) allotypes in familial polyposis coli. Am J Hum Genet 1986;39:133–6.

[8] Iwasaki M, Shimada N, Kasuga Y, Yokoyama S, Onuma H, Nishimura H, et al.Fragment c gamma receptor gene polymorphisms and breast cancer risk incase-control studies in Japanese, Japanese Brazilians, and non-JapaneseBrazilians. Breast Cancer Res Treat 2011;126:497–505.

[9] Brusco A, de Lange GG, Boccazzi C, Carbonara AO. Molecular characterization ofG2m(n+) and G2m(n�) allotypes. Immunogenetics 1995;42:414–7.

[10] Balbín M, Grubb A, de Lange GG, Grubb R. DNA sequences specific forCaucasian G3m(b) and (g) allotypes: allotyping at the genomic level.Immunogenetics 1994;39:187–93.

[11] Pandey JP, Namboodiri AM, Kurtenkov O, Nietert PJ. Genetic regulation ofantibody responses to human epidermal growth factor receptor 2 (HER-2) inbreast cancer. Hum Immunol 2010;71:1124–7.

[12] Cox B, Richardson A, Graham P, Gislefoss RE, Jellum E, Rollag H. Breast cancer,cytomegalovirus and Epstein-Barr virus: a nested case-control study. Br JCancer 2010;102:1665–9.

[13] Harkins LE, Matlaf LA, Soroceanu L, Klemm K, Britt WJ, Wang W, et al.Detection of human cytomegalovirus in normal and neoplastic breastepithelium. Herpesviridae 2010;1:8.

[14] Namboodiri AM, Pandey JP. The human cytomegalovirus TRL11/IRL11-encoded FccR binds differentially to allelic variants of immunoglobulin G1.Arch Virol 2011;156:907–10.

[15] Atalay R, Zimmermann A, Wagner M, Borst E, Benz C, Messerle M, et al.Identification and expression of human cytomegalovirus transcription unitscoding for two distinct Fcgamma receptor homologs. J Virol 2002;76:8596–608.

[16] Pandey JP, Immunoglobulin GM. Genes and IgG antibodies to cytomegalovirusin patients with systemic sclerosis. Clin Exp Rheumatol 2004;22:S35–7.

[17] O’Connell P, Fischbach K, Hilsenbeck S, Mohsin SK, Fuqua SA, Clark GM, et al.Loss of heterozygosity at D14S62 and metastatic potential of breast cancer. JNatl Cancer Inst 1999;91:1391–7.

[18] Figueroa JD, Garcia-Closas M, Humphreys M, Platte R, Hopper JL, Southey MC,Van’t Veer LJ, et al. Associations of common variants at 1p11.2 and 14q24.1(RAD51L1) with breast cancer risk and heterogeneity by tumor subtype:findings from the Breast Cancer Association Consortium. Hum Mol Genet2011;20:4693–706.

[19] Pandey JP. Genomewide association studies and assessment of the risk ofdisease. N Engl J Med 2010;363:2076–7.

[20] Namboodiri AM, Pandey JP. Differential inhibition of trastuzumab andcetuximab induced cytotoxicity of cancer cells by IgG1 expressing differentGM allotypes. Clin Exp Immunol 2011;166:361–5.