Association of Common Variants in TNFRSF13B, TNFSF13, and ANXA3 with Serum Levels of Non-Albumin Protein and Immunoglobulin Isotypes in Japanese

We performed a genome-wide association study (GWAS) on levels of serum total protein (TP), albumin (ALB), and non-albumin protein (NAP). We analyzed SNPs on autosomal chromosomes using data from 9,103 Japanese individuals, followed by a replication study of 1,600 additional individuals. We confirmed the previously- reported association of GCKR on chromosome 2p23.3 with serum ALB (rs1260326, P meta = 3.1×10−9), and additionally identified the significant genome-wide association of rs4985726 in TNFRSF13B on 17p11.2 with both TP and NAP (P meta = 1.2×10−14 and 7.1×10−24, respectively). For NAP, rs3803800 and rs11552708 in TNFSF13 on 17p13.1 (P meta = 7.2×10−15 and 7.5×10−10, respectively) as well as rs10007186 on 4q21.2 near ANXA3 (P meta = 1.3×10−9) also indicated significant associations. Interestingly, TNFRSF13B and TNFSF13 encode a tumor necrosis factor (TNF) receptor and its ligand, which together constitute an important receptor-ligand axis for B-cell homeostasis and immunoglobulin production. Furthermore, three SNPs, rs4985726, rs3803800, and rs11552708 in TNFRSF13B and TNFSF13, were indicated to be associated with serum levels of IgG (P<2.3×10−3) and IgM (P<0.018), while rs3803800 and rs11552708 were associated with IgA (P<0.013). Rs10007186 in 4q21.2 was associated with serum levels of IgA (P = 0.036), IgM (P = 0.019), and IgE (P = 4.9×10−4). Our results should add interesting knowledge about the regulation of major serum components.


Introduction
Serum proteins possess various biological functions such as hormones, enzymes, antibodies, and clotting agents, and some serve as valuable biomarkers that reflect several disease conditions.Major components of serum proteins are ALB (approximately 60%), globulins (mainly as c-globulins, approximately 30%), and fibrinogens.Total serum protein levels range from 6.5 to 8.5 g/dl and show significant inter-individual variation.These variations are found to be influenced by environmental factors.However, genetic factors are also known to affect their levels although the range of genetic effects varies by the reports from 20% to 77% [1].Genome-wide association studies (GWAS) recently demonstrated that serum levels of several proteins can be strongly influenced by common genetic variants through either cis or trans effects [2][3][4].
We previously reported the GWAS results for hematological and biochemical traits, including TP and ALB, in the Japanese population [5].An associated SNP for TP, rs4273077 (Pvalue = 4.5610 210 ), is located in an intron of TNFRSF13B (Tumor Necrosis Factor Receptor Superfamily member 13B), which encodes TACI (transmembrane activator and calciummodulator and cytophilin interactor), one of three TNF-receptor family members (BAFF-R, TACI, and BCMA) [6].However, since rs4273077 showed no significant association with the serum ALB level (P = 0.089), we suspected that this SNP would have genetic effects primarily on the levels of the non-albumin fraction.TACI is expressed mainly in activated B cells and binds with a high affinity to two TNF ligands; APRIL (a Proliferation-Inducing Ligand, encoded by TNFSF13), and BAFF (B Cell-Activating Factor, encoded by TNFSF13B) [7].TACI is implicated in B-cell homeostasis (including B-cell survival, activation, and differentiation), immunoglobulin production, and antibody class switching [8][9][10].Hence, the association of variants in TNFRSF13B with TP is likely to reflect the immunoglobulin serum levels.
The aim of this study is to identify the genetic variations associated with serum levels of non-albumin proteins (NAP), particularly those of immunoglobulins by GWAS of Japanese subjects.

GWAS of Total Protein (TP), Albumin (ALB), and Nonalbumin Protein (NAP)
We conducted a GWAS using genotyping data and clinical information on 9,103 individuals who had been collected in the BioBank Japan Project [11] (Table 1, Table S1).Genotyping was performed using Illumina Human610-Quad BeadChip (Illumina, CA, USA).After applying stringent quality control (QC) filters for selection of individuals and SNPs (Materials and Methods), we additionally performed whole-genome imputation analysis using the data of HapMap Phase II East Asian populations, and we obtained the information of 2,178,644 SNPs on autosomal chromosomes with minor allele frequencies (MAF) of $0.01 and Rsq of $0.7.We then evaluated the association of the SNPs with the adjusted Z scores of serum levels of total protein (TP), albumin (ALB), and non-albumin protein (NAP).A Quantile-quantile (Q-Q) plot for each trait indicated low possibility of population stratification (inflation factors (l GC ) for TP, ALB and NAP were 1.04, 1.02 and 1.02, respectively) (Figure S2).
Since TNFSF13 encodes APRIL, a ligand of TACI encoded by TNFRSF13B, this ligand-receptor interaction is likely to play a critical role in regulation of the serum NAP levels.However, we did not find any synergistic effects between SNPs in the receptor and ligand on NAP levels.
When we examined the genetic contribution of these variations for the traits, the combinations of the SNPs indicated above could explain nearly 0.5%, 2.3%, and 0.3% of variations in serum TP, NAP, and ALB, respectively.

Replication Study
To validate the GWAS results, we performed a replication study using an independent set of ,1,600 subjects from BioBank Japan [11] (Table 1).For each trait, we selected marker SNPs for the replication analysis at each locus that indicated the genome-wide significant level of 5.0610 28 (rs4985726 in TNFRSF13B, rs3803800 in TNFSF13, rs1260326 in GCKR, and rs10007186 on 4q21.2).In addition, the two SNPs that remained significant after accounting for the effect of each marker SNP at two loci (rs11552708 in TNFSF13 and rs3817588 in GCKR) were also further investigated.
SNPs rs4985726 in the TNFRSF13B locus as well as rs3803800 and rs11552708 in the TNFSF13 locus revealed significant associations with both TP and NAP (Table 2).The association of rs1260326 in GCKR with serum ALB was also replicated (P = 0.029; Table 2).Meta-analyses combining the GWAS and the replication study yielded stronger associations of these SNPs than the GWAS alone (Table 2 and Figure 2A, B, C, and E).Rs10007186 near ANXA3 revealed a suggestive association in the replication study (P = 0.065), and meta-analyses indicated that the association was unlikely to be false positive (P = 1.3610 29 ) (Table 2 and Figure 2D).

Association of the SNPs Identified in the GWAS of NAP with Serum Immunoglobulin Isotypes
Immunoglobulin isotypes constitute the major components of NAP.Hence, we further examined the NAP-associated SNPs in the GWAS (TNFRSF13B, TNFSF13, and ANXA3) for the association with various serum immunoglobulins using the samples in BioBank Japan [11] (IgG: n = 1,794, IgA: n = 1,675, IgM: n = 1,649, and IgE: n = 549; Table 1).

Discussion
On the basis of the information of 10,716 Japanese individuals, we identified one genetic locus (TNFRSF13B) on chromosome 17p11.2associated with both TP and NAP, two loci (TNFSF13 on 17p13.1 and a region near ANXA3 on 4q21.2) associated with NAP, and one locus (GCKR) on 2p23.3 associated with ALB at the level of genome-wide significance.
The marker SNP rs4985726 shows association with TP and NAP is located in an intron of TNFRSF13B on chromosome 17p11.2.A possible mechanism for its association with these traits could be explained by its strong LD with rs34562254 (D9 = 1, r 2 = 0.97), which exhibits a missense variation (C.T, Pro251Leu) located in the intracellular domain of the receptor molecule.The in silico prediction of the amino acid substitution by rs34562254 in the PolyPhen-2 and SNPinfo database [12,13] suggested a ''probably damaging'' effect on the protein structure.
The SNPs in TNFSF13 (encoding APRIL) that identified as being associated with NAP are missense variants; rs3803800 (A.G, Asn96Ser), and rs11552708 (G.A, Gly67Arg).APRIL was first described as having a promoter function for tumor-cell proliferation and survival [14].APRIL is cleaved in the Golgi apparatus by furin at its 104Arg/105Ala site [15], and interestingly, rs3803800 is closely located to this cleavage site.Hence, this SNP might affect the cleavage affinity.Another possibility is the effect on splicing, because both SNPs are predicted to be located within binding sites of splicing regulatory elements [13].However, further investigation should be required to address these possibilities.
The SNP rs4985726 in TNFRSF13B as well as rs3803800 and rs11552708, in TNFSF13 also revealed significant associations with serum levels of IgG, IgA, and IgM.It is notable that the two genes encode a TNF-receptor and ligand axis that plays important roles for mediating antibody class switching and regulating immunoglobulin production [8,9].Furthermore, knockout mice of either TNFRSF13B or TNFSF13 presented a common phenotype of the IgA deficiency with impaired antibody response to T cellindependent antigens [16].In addition, germ-line mutations in TNFRSF13B were reported in cases of common variable immunodeficiency (CVID; MIM # 607594) and selective IgA deficiency (IGAD; MIM # 137100) [17].The combination of these significant statistical and biological evidences would suggest that the association of these two loci with NAP reflect at least their associations with regulation of serum immunoglobulin levels.It is also known that immunoglobulins are the major components of NAP, which provides compelling evidence for our results.The facts that both SNPs rs3803800 [A] and rs11552708 [G] in TNFSF13 were reported to be associated with the susceptibility to the Systemic Lupus Erythematosus (SLE) in the Japanese population and that high serum APRIL was detected in the sera of individuals with the rs3803800 [A]-rs11552708 [G] haplotype [18] further support the significance of these SNPs in the regulation of immunoglobulin production.In this study, we observed that possession of two copies of SLE-risk alleles was associated with higher serum levels of NAP, IgG, IgA, and IgM (Figure S3), providing a good example of genetic loci that influence both quantitative traits and susceptibility to complex diseases.
Rs10007186, which was associated with NAP (P meta = 1.3610 29 ) is located about 57.4 kb downstream of ANXA3 encoding annexin A3, a member of annexin family of calcium- dependent phospholipid-binding proteins [19].Annexin A3 was found to be translocated into phagosomes in dendritic cells [20], which are antigen-presenting cells that serve as messengers between the innate and adaptive immune response, and play a key role in allergic, inflammatory, and autoimmune conditions.In addition, annexin A3 was also found to be associated with neutrophil granule membranes [21], where it can play a regulatory role in calcium-dependent granule secretions that contribute to acute inflammation and chronic tissue destruction.The association of rs10007186 with IgA, IgM, and IgE, would suggest additional biological roles of annexin A3 in the immune response.
We also confirmed the association of SNPs in GCKR with serum ALB levels (rs1260326, P meta = 3.1610 29 ).Rs1260326 is a missense variant (T.C, Leu446Pro) and predicted to cause a damaging effect on the protein structure.GCKR is a locus frequently associated with several metabolic traits [4,[22][23][24] and rs1260326 has been reported to be associated with serum triglycerides [4].As a conclusion, the present study identified genetic loci that influence the inter-individual variation in serum levels of TP, ALB, and NAP.The loci associated with NAP encompass genes encoding a TNF-receptor and its ligand, which are implicated in biological roles in the immune system, and their associations with immunoglobulin isotypes were demonstrated here.Our results should add novel insight toward understanding the genetic background contributing to the regulation of the serum levels of NAP and its major components.

Study Cohorts
For the GWAS, 9,103 subjects derived from 10 disease cohorts (colorectal cancer, breast cancer, prostate cancer, lung cancer, gastric cancer, diabetes mellitus, peripheral artery disease, atrial fibrillation, ischemic stroke, and myocardial infarction) were selected, and for the replication study, we used data from .1600independent individuals selected from the BioBank Japan Project [11] (Table 1 and Table S1).For immunoglobulin isotypes analyses, the data from ,1,600 additional individuals in BioBank a M6S.D: mean value6standard deviation of each protein is indicated in g/dl except for IgE, which is indicated as IU/ml.b Age and body mass index (BMI) are indicated as mean values6standard deviation.*Log-transformed values were applied in the analysis.Abbreviations: GWAS: genome-wide association study, TP: total protein, ALB: albumin, NAP: non-albumin protein.doi:10.1371/journal.pone.0032683.t001

Figure 1 .
Figure 1.Manhattan plots for the GWAS of (A) TP, (B) NAP, and (C) ALB.SNPs were plotted based on their physical chromosomal positions (horizontal axis) together with their -log 10 (P-values) in the GWAS (vertical axis).The black horizontal line shows the genome-wide significance threshold of P = 5.0610 28 .The SNPs for which P-values were smaller than 1.0610 215 are indicated at the upper limit of the plots.doi:10.1371/journal.pone.0032683.g001

Figure 2 .
Figure 2. Regional plots for the associations of the SNPs in the GWAS stage of TP, ALB and NAP.SNPs plotted with their -log 10 (P-values) in the GWAS based on their physical chromosomal positions.Genotyped SNPs are indicated as circles, while imputed SNPs are indicated as triangles.The color scheme indicated the linkage disequilibrium displayed as r 2 values between all SNPs and the top-ranked SNP in each plot.The tested trait, chromosomal locus, and the top-ranked SNPs (in purple color) in the GWAS and combined analyses together with their P-values are shown in each plot.The blue lines represent the recombination rates estimated based on HapMap Phase II database.The plots were drawn using Locus Zoom software.doi:10.1371/journal.pone.0032683.g002

2 . 02 a
The effect of the minor alleles on the standardized values.b P-values for the associations of SNPs with each normalized immunoglobulin isotype obtained by using a linear regression model.Abbreviations: s.e: standard error, %EV: percentage of the explanatory variance.doi:10.1371/journal.pone.0032683.t003

Table 1 .
Characteristics of the examined proteins.

Table 2 .
Summary results of the GWAS and the replication study of TP, ALB, and NAP.The effect of the minor allele on the normalized values based on an additive genetic model.For the GWAS and replication analysis, P-values were obtained by linear regression test model, for the Meta analysis by inverse-variance method.*SNPs obtained by whole-genome imputation analysis.Abbreviations: GWAS: genome-wide association study, MAF: minor allele frequency, TP: total protein, ALB: albumin, NAP: non-albumin protein, s.e: standard error.doi:10.1371/journal.pone.0032683.t002 b c

Table 3 .
Association of the SNPs in the GWAS of the NAP with immunoglobulin isotypes.