Genome wide association study identifies two loci associated with cadmium in erythrocytes among never-smokers

Abstract

Background : Cadmium is a non-essential toxic metal with multiple adverse health effects. Exposure in the general population occurs by smoking and diet. Cadmium in erythrocytes is a valid biomarker of exposure and body burden of cadmium.

Objectives : We aimed to identify genetic variants related to concentrations of cadmium in erythrocytes.

Methods: Erythrocyte cadmium was analyzed in 4432 individuals (1728 never smokers) from the Swedish population-based Malmö Diet and Cancer cohort. Genotyping was performed using the Illumina HumanOmniExpressExome Bead chip with genome-wide coverage. Genome wide analyses were performed in the whole sample and in never smokers.

Results: No single nucleotide polymorphism (SNP) reached a genome-wide significant association with erythrocyte cadmium in the whole sample. However, in never smokers, 14 variants showed genome-wide significant relationships with erythrocyte cadmium after adjusting for age and sex. Thirteen variants were in linkage disequilibrium on chromosome 8q13.3 in the XKR9 and LACTB2 genes. The lead SNP on 8q13.3 was rs12681420 (minor allele G, minor allele frequency [MAF] = 0.46, β: −0.11, P = 3.48 × 10 ⁻¹¹ ), an intron variant within the XKR9 gene. The other significant locus, rs17574271 (minor allele C, MAF = 0.09, β: 0.17, P = 6.18 × 10 ^–9 ), was an intron variant within the DLGAP1 gene at chromosome 18p11.31.

Conclusion: This genome-wide study of never smokers from the general population identified two independent regions related to erythrocyte cadmium. The strongest locus covers the XKR9 and LACTB2 genes, which both could have related functions in cadmium absorption and metabolism. Replication studies are needed to confirm the findings and mechanisms should be further investigated.

Introduction

Cadmium is a non-essential toxic metal with multiple adverse health effects. Epidemiological studies have reported relationships between cadmium exposure and renal dysfunction ( 1–4 ), osteoporosis ( 4–6 ), cancer ( 7–9 ) and cardiovascular disease ( 10–13 ). Smoking is a major source of cadmium exposure and smokers have several fold higher blood cadmium concentrations than non-smokers ( 1 ). Diet accounts for 99% of cadmium exposure in the non-smoking general population ( 14 ). The main food sources of cadmium include vegetables, whole grain, starchy roots and shellfish ( 15 ). Anemia, or low iron stores is associated with increased uptake of cadmium and women have higher concentrations of cadmium than men due to lower iron status ( 16 , 17 ). Cadmium is excreted very slowly from the body and the biological half-life of cadmium is ∼10–40 years ( 1 , 18 ).

After exposure to cadmium from diet and smoking, the concentration of cadmium in the body could be affected by individual differences with respect to absorption, distribution, metabolism and elimination of cadmium. Because diet is the main route of cadmium exposure in non-smokers, it is conceivable that differences in intestinal absorption could cause substantial variations in blood concentrations. There are very few studies of the genetic effect on cadmium concentrations. A study on Swedish twin pairs found that variations in cadmium levels in women were not only explained by environmental exposure, but also by genetic factors ( 18 ). Another genetic linkage analysis of twins reported that the variation in erythrocyte cadmium due to genetic factors was 19%, and linkage for cadmium on chromosomes 2, 18, 20 and X was suggested ( 19 ).

Some studies have also aimed to identify the specific genetic variants associated with cadmium. Two candidate gene studies in non-smoking women from Argentina and Bangladesh found a relationship between urine cadmium and a single nucleotide polymorphism (SNP) in the transferrin receptor ( TFRC ) gene ( 20 , 21 ); erythrocyte cadmium concentration was associated with SNPs in the Zn-transporter genes ( SLC39A8 and SLC39A14 ). These studies suggest that proteins involved in uptake and metabolism of iron and other metal ions could be important for cadmium concentrations in the blood. A recent genome-wide association study (GWAS) of 949 well characterized 70-year-old men and women identified an association between a locus in the CD109 gene and cadmium concentrations in whole blood ( 22 ). However, the mechanism through which this gene effects cadmium concentration in the blood is not clear.

GWAS can be used to study the genetic variation of traits and diseases by genotyping of SNPs throughout the whole genome. The objective of this study was to identify genetic loci related to erythrocyte cadmium by performing a GWAS in all individuals and in never-smokers, respectively, from an ethnically homogenous population-based cohort.

Results

Characteristics of study population

Among the whole sample ( n = 4432) and never smokers ( n = 1728), women had higher levels of both erythrocyte cadmium and blood cadmium level than men. More men were smokers (23.4 versus 21.2%) and more women were never smokers (47.0 versus 29.7%) ( Table 1 ).

Table 1.

Characteristics of study population among men and women

	All ( n = 4432)		Never smoker ( n = 1728)
	Men ( n = 1776)	Women ( n = 2656)	Men ( n= 514)	Women ( n = 1214)
Erythrocyte cadmium (GM, min–max) (µg/l)	0.67 (0.06–10.52)	0.81 (0.08–11.33)	0.37 (0.06–4.08)	0.54 (0.08–5.75)
Blood cadmium a (GM, min–max) (µg/l)	0.29 (0.02-5.07)	0.32 (0.03–4.83)	0.16 (0.02–1.66)	0.22 (0.03–2.31)
Age, mean (SD)	57.5 (5.9)	57.4 (5.9)	57.4 (6.0)	58.4 (5.8)
Smoking status
Current smoker (%)	23.4	21.1
Former smoker (%)	46.9	31.9
Never smoker (%)	29.7	47.0

SD, standard deviation; GM, geometric mean.

^a Calculated from erythrocyte cadmium and hematocrit.

No loci showed a genome-wide significant association (i.e. P <5 × 10 ^–8 ) with erythrocyte cadmium in the whole sample. rs6815218 on chromosome 4 close to the HSPA8P12 gene had the lowest P- value for association with cadmium ( P = 5.4 × 10 ⁻⁶ ). A list of SNPs associated with erythrocyte cadmium at a P- value < 5 × 10 ^–5 is provided in Supplementary Material, Table S1 .

Potential interactions between smoking status (ever versus never smokers) and the SNPs with P < 5 × 10 ^–5 for association with cadmium were tested ( Supplementary Material, Table S1 ). None of the top SNPs were genome-wide significant in ever smokers.

Never smokers

Fourteen variants in two independent regions showed genome-wide significant relationships with erythrocyte cadmium among never smokers ( Table 2 , Fig. 2 ). The lead SNP on 8q13.3 was rs12681420 (minor allele G, MAF = 0.46, β: −0.11, P = 3.48 × 10 ^–11 ), an intron variant within the XKR9 gene. In this region, there were 13 significant SNPs in linkage disequilibrium (LD): 11 of these were in the XKR9 gene, and two in the LACTB2 gene, Table 2 . The other significant locus rs17574271 (minor allele C, MAF = 0.09, β: 0.17, P = 6.18 × 10 ⁻⁹ ) was an intron variant within the DLGAP1 gene at chromosome 18p11.31. Regional significance plots for the significant loci are presented in Supplementary Material, Figure S1 .

Table 2.

Loci showing genome-wide significance with erythrocyte cadmium among never smokers

SNP	CHR	Position	P	MA	MAF	Beta	P_HWE	Gene	r2
rs12681420	8	71593198	3.48 × 10 −11	G	0.4617	−0.11	0.2843	XKR9
rs1477945	8	71615553	9.17 × 10 −11	G	0.4586	−0.107	0.7421	XKR9	1
rs4629903	8	71805132	1.49 × 10 −10	A	0.4604	−0.1061	0.4764	XKR9	1
rs11987501	8	71578518	1.72 × 10 −10	C	0.4615	−0.1058	0.4938	LACTB2	1
rs2290702	8	71646980	1.78 × 10 −10	C	0.4603	−0.1052	0.4429	XKR9	1
rs4532626	8	71944895	2.66 × 10 −10	T	0.4624	−0.1055	0.3379	XKR9	1
rs12675271	8	71602009	2.83 × 10 −10	A	0.4539	0.1049	0.5488	XKR9	1
rs1493199	8	71846504	2.92 × 10 −10	C	0.4614	−0.1044	0.4107	XKR9	1
rs6472551	8	71781849	5.15 × 10 −10	G	0.4548	0.1032	0.567	XKR9	1
rs1389204	8	71928298	6.70 × 10 −10	T	0.4531	0.103	0.6627	XKR9	1
rs13274381	8	71765094	6.82 × 10 −10	A	0.4543	0.1025	0.6432	XKR9	1
rs2639919	8	71875128	1.12 × 10 −9	A	0.4645	0.1013	0.6835	XKR9	1
rs17574271	18	3970124	6.18 × 10 −9	C	0.0884	0.1727	0.1885	DLGAP1	0
rs441890	8	71564667	9.83 × 10 −9	C	0.3989	0.0965	0.483	LACTB2	1

r² LD with rs12681420; rs11987501 correlated with rs441890 ( r² = 1).

CHR, chromosome; SNP, single nucleotide polymorphisms; MA, minor allele; MAF, minor allele frequency; HWE, Hardy-Weinberg equilibrium.

A list of SNPs associated with cadmium in never smokers at a P- value <5 × 10 ^–5 is provided in Supplementary Material, Table S2 . Although not statistically significant in a GWAS setting, it is noteworthy that several more SNPs in the XKR9 and LACTB2 genes on chromosome 8 are associated with cadmium ( Fig. 2 ).

Interactions with sex, anemia and diet

The mean erythrocyte cadmium levels for the genotypes of rs12681420 ( XKR9 gene) and rs17574271 ( DLGAP1 gene), respectively, are shown in Table 3 , in relation to sex, anemia status (hemoglobin < 130 g/l for men and hemoglobin < 120 g/l for women) and dietary intake of vegetables ( ≤0.08 g/kcal/d versus > 0.08 g/kcal/d). There was no significant interaction between genotype and sex with respect to level of erythrocyte cadmium ( P = 0.924 for rs12681420, P = 0.659 for rs17574271). Nor were there any significant interaction with anemia status ( P = 0.134 for rs12681420, p = 0.722 for rs17574271) or vegetable intake ( P = 0.651 for rs12681420, P = 0.676 for rs17574271).

Table 3.

Erythrocyte cadmium concentrations in relation to genotype of the significant SNPs among never smokers ( n = 1728)

SNP	Erythrocyte cadmium (GM, min–max) (µg/l)
	All never smokers ( n = 1728)	Men ( n = 514)	Women ( n = 1214)	Anemia ( n = 40)	No anemia ( n = 1688)	Low vegetable intake ( n = 915)	High vegetable intake ( n = 813)
rs12681420
AA	0.54 (0.08–5.08)	0.41 (0.01–2.53)	0.60 (0.08–5.08)	0.58 (0.22–2.25)	0.54 (0.08–5.08)	0.51 (0.10–2.53)	0.57 (0.08–5.08)
AG	0.48 (0.06–5.75)	0.37 (0.06–4.08)	0.53 (0.11–5.75)	0.75 (0.28–1.93)	0.47 (0.06–5.75)	0.47 (0.06–5.75)	0.48 (0.13–1.80)
GG	0.43 (0.10–5.08)	0.32 (0.10–1.26)	0.49 (0.15–5.08)	0.56 (0.20–0.93)	0.43 (0.10-5.08)	0.40 (0.10–1.77)	0.46 (0.13–5.08)
rs17574271
TT	0.47 (0.06–5.75)	0.36 (0.06–4.18)	0.53 (0.08–5.75)	0.62 (0.28–1.77)	0.47 (0.06–5.75)	0.45 (0.06–5.75)	0.49 (0.08–5.08)
CT/CC	0.56 (0.11–4.84)	0.43 (0.11–3.25)	0.62 (0.21–4.84)	0.71 (0.20–2.25)	0.55 (0.11–4.84)	0.53 (0.11–3.25)	0.58 (0.22–4.84)

GM, geometric mean.

Candidate SNPs from previous studies

Six SNPs that have been associated with cadmium in previous studies ( 20–22 ) were analyzed in our study too. Four of these SNPs were imputed. The results for the six candidate SNPs are presented in Table 4 . None of them were associated with cadmium in our sample (all P values > 0.2).

Table 4.

Candidate SNPs from previous studies and their association for all subjects, never smokers and ever smokers

SNP	Reference	CHR	Gene	Info score a	P for all	P for never smokers	P for ever smokers
rs233804 a	( 21 )	4	SLC39A8	0.999	0.228	0.093	0.433
rs4872479 a	( 21 )	8	SLC39A14	0.938	0.921	0.239	0.846
rs870215 a	( 21 )	8	SLC39A14	0.993	0.964	0.908	0.942
rs10014145	( 21 )	4	SLC39A8		0.197	0.402	0.795
rs9350504 a	( 22 )	6	CD109	0.952	0.296	0.871	0.636
rs3804141	( 20 )	3	TFRC		0.512	0.632	0.842

^a The SNP was imputed in MDC-CV. Imputation quality assessed by info score.

Discussion

Studies of genetic determinants of high cadmium concentrations could improve our understanding of the routes and mechanisms of cadmium accumulation and could also be useful in identifying individuals with high susceptibility for the toxic effects of cadmium. This GWAS study showed two independent regions associated with cadmium levels in erythrocytes in never smokers from the general population.

The top association with cadmium in never-smokers spans the XKR9 and LACTB2 genes region on chromosome 8q13.3. The XKR9 is a member of the XK, Kell blood group complex subunit-related family, and is highly expressed in the small intestine and also expressed in pancreas, liver, stomach and colon ( 23 ). XKR9 seems to enhance the shedding of aged epithelial cells into the lumen, and it could promote epithelial regeneration and hence integrity and function ( 23 ). XKR family members are localized in the plasma membrane and promote Ca ²⁺ -dependent phospholipid scramblase externalization ( 24 , 25 ). This process involves the transport (‘scramble’) of the negatively charged phospholipids between the inner-leaflet and the outer-leaflet of the bilayer membranes. The increased scramblase activity may enhance apoptosis in cadmium affected tissues ( 23 ). Changes in the phospholipid environment of channels and transporters in the plasma membrane of relevant tissues, such as intestine, by XKR9 SNPs may affect ion and hence possibly cadmium transport. Human phospholipid scramblase has been found to be activated in heavy metal poisoning by binding lead and mercury to bidirectional transbilayer movement of phospholipids ( 26 ).

Two significant SNPs were on LACTB2 , β-Lactamase-like Protein 2, a member of the metallo-beta-lactamase superfamily and the Glyoxalase II family ( 27 ). LACTB2 is a zinc-binding protein, and zinc could be displaced by cadmium and thereby change the function of these proteins. As metallo-enzymes, LACTB2 proteins are expected to have hydrolase activity and metal ion-binding functions ( 28 ). LACTB2 is highly expressed in adrenal gland, testis, kidney, liver and colon in normal tissues ( http://www.gtexportal.org/home/gene/LACTB2 , last accessed February 29, 2016).

One SNP was located at DLGAP1 gene on chromosome 18. DLGAP1 is a member of the Synapse-associated protein 90/Postsynaptic density-95-associated protein family ( http://www.gtexportal.org/home/gene/DLGAP1-AS4 , last accessed February 29, 2016). It may play roles in the molecular organization of synapses and neuronal cell signaling, is highly expressed in brain ( 29 ), and has been found to associate with obsessive-compulsive disorder ( 30 , 31 ). Yet, evidence for neuronal symptoms in chronic cadmium toxicity is scarce. A genetic linkage analysis showed suggestive linkage for cadmium on the p-terminal end of chromosome 18 ( 19 ). The DLGAP1 gene (18p11.31) is located in this region. However, the potential mechanisms for this gene in cadmium metabolisms have to be further searched for.

No SNP reached genome wide significance in the whole cohort including smokers. Smoking is a major source of cadmium and associated with several-fold higher blood concentrations. High concentrations of cadmium can accumulate in tobacco (Nicotiana tabacum L) leaves ( 32 ). Among smokers, it is likely that differences with respect to the amount of tobacco consumption, inhalations patterns and other smoking related factors etc diluted the genetic effects.

Previous studies of polymorphisms in Zinc-transporter genes among non-smoking women from Argentina have reported relationships between cadmium in erythrocytes and polymorphisms in the SLC39A8 and the SLC39A14 genes ( 20 , 21 ). We imputed these SNPs and none of them were significantly associated with cadmium in erythrocytes. However, there was a directly genotyped SNP (rs7664683) on the Zinc-transporter gene SLC39A8 with a low P -value ( P = 3.18 × 10 ⁻⁷ , Supplementary Material, Table S2 ). Even though the association between rs7664683 and rs233804 is quite weak ( r² = 0.081) in our study, this gives some support to the results by Rentschler et al. ( 21 ). The SNP at locus 6q14 (rs9350504), which was GWAS significant in the study by Ng et al. ( 22 ), was imputed by us, but was not nominally significant in our study.

The population-based cohort of an ethnically homogenous population with comparatively low cadmium concentrations is a major strength of the study. Because smoking is a major source of cadmium, and smoking to some extent is genetically determined, another major strength is that the analysis could be performed in never smokers. The laboratory analysis of cadmium showed high precision and the concentrations could be quantified also at very low levels. A limitation is that cadmium was measured in erythrocytes and data from urine and whole blood could potentially provide additional information. However, almost all cadmium in blood is concentrated to the erythrocytes, and the concentration in plasma is very low. Cadmium in erythrocytes is a valid marker of cadmium exposure, especially in never smokers in whom exposure is relatively constant.

We identified 14 SNPs at two independent loci, one in the XKR9 and LACTB2 genes and another in the DLGAB1 gene, related to cadmium in erythrocytes among never smokers from the general population. Further replication studies are needed and mechanisms should be investigated.

Materials and Methods

Selection and description of participants

Between 1991 and 1996, individuals living in the city of Malmö, Sweden, were invited to participate in the Malmö Diet and Cancer (MDC) study. Between 1991 and 1994, a random subsample (the cardiovascular cohort (MDC-CV) (n = 6103), was invited to take part in a sub-study of the epidemiology of carotid artery disease ( 33 ), Supplementary Material, Figure S2 . The age range was 46–68 years and 60% were women. Smoking was assessed in a questionnaire. Due to the long half-time of cadmium in the body, the subjects were categorized into never-smokers versus ever-smokers (current, former and occasional smokers) for the analyses in this study.

Intake of vegetables, adjusted for total energy intake, was used as a proxy of dietary cadmium. Dietary information in MDC was collected using a modified diet history method, combining a 168-item quantitative diet history questionnaire, a 7-day menu book and 1-h dietary interview. Energy (kcal) and vegetable intake (g/d) were computed from the MDC Food and Nutrient database, based on the PC Kost2-93 food database of the Swedish National Food Administration ( 34 , 35 ). Vegetable intake was divided into two groups using mean of vegetable intake/energy (0.08 g/kcal) as cutoff.

Hemoglobin and hematocrit was analysed consecutively in venous blood within 2 h from blood sampling using a SYSMEX K1000 fully automated assay (Sysmex Europe, Norderstedt, Germany).

Cadmium in blood

The erythrocytes were stored in −80 °C from the baseline examination until the analysis in 2013. Erythrocyte cadmium was analyzed using inductively coupled plasma mass spectrometry (Agilent 7700x ICP-MS; Agilent Technologies, Santa Clara, CA, USA). The reaction system was operated in the helium collision cell mode to eliminate interference from isobaric polyatomic species via kinetic energy discrimination ( 32 ). The imprecision was 9.6%, calculated as the coefficient of variation for 50 duplicate samples (mean 0.43 µg/l). The limit of detection (LOD) was 0.02 µg/l, calculated as three times the standard deviation of the blank. None of the samples were below the LOD. All samples were analyzed in three different rounds with external quality control (QC) samples included. Two QC samples were used (Seronorm Trace Elements Whole Blood l-1, Lot no. 1103128, and Seronorm Trace Elements Whole Blood l-2, Lot No.1103129; Sero AS, Billingstad, Norway). The results from all rounds versus recommended limits were 0.34 ± 0.02 µg/l ( n = 70) versus 0.32–0.40 µg/l, and 5.7 ± 0.18 µg/l (n = 70) versus 5.4–6.2 µg/l. Furthermore, an inter-lab comparison showed good agreement ( 13 ). The blood cadmium concentrations were calculated as erythrocyte concentrations of cadmium × hematocrit.

The regional Ethics Committee (LU51/90) approved the study and all participants provided written informed consent. The study complies with the Declaration of Helsinki.

Genotyping

Blood samples were drawn at the baseline examination and stored at −80°C in a biobank before analysis. Genotyping was performed using The Illumina HumanOmniExpressExome Bead Chip, the iScan system and the Autocall calling algorithm (Illumina, San Diego, CA, USA). The array includes >700 000 single nucleotide variations with genome-wide coverage of common genetic variation ( http://www.illumina.com/ , last accessed February 29, 2016). All procedures followed the standard Protocol. The additional specific exome part of the chip was not included in the present analysis.

Genotyping QC

Individual level QC was performed by exclusion of individuals with (i) a call rate <95%; (ii) an inbreeding coefficient <−0.2 or 0.2; (iii) disconcordant sex in self-report versus genetically determined sex; (iv) a second degree relatedness or higher within the sample, based on identity by descent sharing calculations; and (v) individuals that were population outliers based on inspection of the first two principal components. In marker level QC, we excluded variants with a call rate <95%, variants on sex chromosomes and mitochondrial DNA, as well as variants showing an extreme deviation from Hardy-Weinberg equilibrium ( P < 1 × 10 ^–6 ). We also used a MAF limit of 0.01. The number of individuals remaining after QC was 4432 and 1728 of them were never smokers. A total of 658 884 SNPs were included in the analysis. The quantile-quantile (QQ) plot shows no population stratification ( Fig. 1 , median genomic inflation factor = 0.998).

Figure 1.

QQ plot of the whole sample ( n = 4432).

Figure 2.

Manhattan plot (−log10[ P ] genome-wide association plot) of the GWAS on erythrocyte cadmium in 1728 never smoker individuals in the MDC-CV cohort.

Candidate SNPs for cadmium ( Table 4 ) that were not directly genotyped in our study were looked up in an imputed data set. Imputation was performed using Impute 2 software ( 36 ) on a genome-wide basis with the 1000 Genomes Integrated Phase I release version 3 all populations (ancestry) panels. The Info score for these imputed SNPs are >0.93.

Statistical analysis

Linear regression models, assuming an additive effect of each allele, were used to test the association between genetic variants and log transformed erythrocyte cadmium adjusting for age and sex. Log transformed erythrocyte cadmium was treated as a linear variable; the association analysis thus gives a beta coefficient and 95% CI for the change in log transformed erythrocyte cadmium level per each risk allele. A P -value below 5 × 10 ^–8 was considered as genome-wide significant, corresponding to a Bonferroni correction for one million tests. Version 1.07 of PLINK software was used for association analyses and QC. Manhattan plots and QQ plots were drawn with the R software version 3.1.2. Regional significance plots were drawn using LocusZoom ( http://locuszoom.sph.umich.edu/locuszoom/ , last accessed February 29, 2016) ( 37 ). LD was assessed using the Pearson’s correlation coefficient ( r² ) and checked using SNP Annotation and Proxy Search SNAP in 1000 Genomes with CEU population panel ( https://www.broadinstitute.org/mpg/snap/ldsearch.php , last accessed February 29, 2016) ( 38 ).

Erythrocyte cadmium concentrations in never smokers were presented in relation to genotype of the top SNPs and sex, smoking, anemia and intake of vegetables, respectively, and potential interactions were explored using interaction terms in the linear regression models. For this analysis, among the SNPs that were in LD on the same loci, the SNP with the lowest p value was used. Statistics softwares IBM SPSS Statistics (version 22; IBM Sweden AB, Stockholm, Sweden) and Stata software version 12.0 (Stata Corp, College Station, TX, USA) were used for analysis.

Supplementary Material

Supplementary Material is available at HMG online.

Acknowledgements

We are grateful to Peter Almgren for help with QC and genetic data management.

Conflict of Interest statement . None declared.

Funding

This work was supported by the Swedish Research Council, the Swedish Heart–Lung Foundation and the Swedish governmental research council FORTE.