Abstract
Metal elements are present in the human body, and their levels in the blood have important impacts on health. In this study, 2488 Chinese individuals were included in a genome-wide association study of 21 serum metal levels, with approximately 179,000 East Asian individuals in a bidirectional two-sample Mendelian randomization (MR) analysis, and 628,000 Europeans in a two-sample MR analysis. We identified two single nucleotide polymorphisms (SNPs) rs35691438 and rs671 that were significantly associated with serum copper levels (SCLs). The bidirectional two-sample MR analysis in the East Asian population showed that gamma-glutamyl transpeptidase levels have a causal effect on SCLs. SCLs have causal effects on six outcomes, namely risks of esophageal varix, glaucoma, sleep apnea syndrome, and systemic lupus erythematosus, white blood cell count, and usage of drugs affecting bone structure and mineralization. The two-sample MR analyses in the European population showed causal effects of erythrocyte copper levels on risks of carpal tunnel syndrome and compression fracture. Our results provide original insights into the causal relationship between blood metal levels and multiple human phenotypes.
Similar content being viewed by others
Introduction
Metal elements are present in the human body and can be classified as essential metal elements and toxic metal elements. Essential metal elements, such as calcium, copper, iron, manganese, magnesium, selenium, and zinc, are obtained mainly through the daily diet. They play critical roles in normal physiological activities, including oxidation–reduction reactions, catalysis, metabolic regulation, and intercellular signal transduction1,2. Toxic metal elements, such as arsenic, cadmium, and lead, usually come from environmental exposure3,4. The accumulation of toxic metal elements in the human body can lead to many adverse health effects and even endanger life. In addition to the environmental factors that affect metal levels in the human body, many genetic variants that are significantly related to blood metal levels (BMLs) have been discovered. Numerous genome-wide association studies (GWASs) of European populations found a large number of single nucleotide polymorphisms (SNPs) that were significantly associated with blood levels of calcium, copper, selenium, zinc, iron, magnesium, manganese, cadmium, and mercury5,6,7,8,9. However, as far as we know, only a few GWASs on multiple BMLs in East Asian populations have been reported.
Many epidemiological studies have shown that BMLs are closely related to human diseases10,11,12,13,14,15. However, traditional epidemiological studies are often affected by confounding factors, and therefore determining causality can be difficult. Mendelian randomization (MR) studies can avoid the bias caused by the unavoidable confounding factors in traditional epidemiological studies, thereby allowing more robust and reliable causal inferences16. The application of MR has helped to uncover the causal relationships between BMLs and human disease phenotypes. Genetically higher serum calcium levels were shown to be associated with a higher risk of coronary artery disease and myocardial infarction17. Another study of UK Biobank cohorts found that genetically increased blood levels of both copper and iron were associated with reduced risk of lipid metabolism disorders18. A two-sample MR study based on Europeans showed that bone mineral density of certain skeletal locations was influenced by serum calcium and selenium levels19. Current GWAS-MR studies of BMLs and their causal effects on human diseases focused mainly on European populations. Considering the genetic differences between European and East Asian populations, a systematic GWAS-MR of BMLs in East Asian population is necessary to provide more reliable data for the diagnosis, treatment, and basic scientific research of clinical diseases.
In this study, we conducted a comprehensive GWAS of multiple serum metal levels in the Chinese population and then performed a cross-ethnic MR study to fully understand the causal effects of BMLs on multiple deep phenotypes in humans. For this, we conducted bidirectional two-sample MR analyses using our GWAS summary statistics of serum metal levels and the GWAS summary statistics from BioBank Japan20. In addition, to comprehensively study the causal effects of BMLs on multiple deep phenotypes in different populations, we used the GWAS summary statistics of erythrocytes metal levels from the study of Evans et al.6 and the GWAS meta-analysis summary statistics from UK Biobank and FinnGen to conduct two-sample MR analyses.
Results
Characteristics of subjects
We totally included 2488 subjects in our present study. After the non-parametric test on the concentration of each metal between the two cohorts, only the concentrations of copper and manganese showed no statistical difference (Pcopper = 0.08 and Pmanganese = 0.96). The basic demographic characteristics and the results of the comparison of serum levels for the same metal in two cohorts of the subjects finally included in the GWAS analysis are shown in Table 1.
GWAS and meta-analysis
We performed an independent GWAS for each transformed serum/plasma metal concentration in the Fangchenggang Area Male Health and Examination Survey (FAMHES) and the manganese-exposed workers healthy cohort (MEWHC), respectively. The specific number of subjects included in the GWAS of each metal is shown in Table 1. After imputing and quality control, the GWAS carried out in FAMHES and MEWHC included 7053840 and 4925397 variants, respectively. In FAMHES, two top SNPs (rs78069066, on ADAM1A of chromosome 12, P = 3.58 × 10−9; rs10424895, 0.901 kb away from OR7A5 of chromosome 19, P = 4.42 × 10−8) were found significantly associated (P < 5.0 × 10−8) with serum copper levels (SCLs) (Supplementary Table 1). The Manhattan, Quantile-Quantile and LocusZoom plots are in Supplementary Fig. 1, and the genomic inflation factor (λ) is 1. In MEWHC, one SNP (rs2235321, on TMPRSS6 of chromosome 22, P = 3.20 × 10−9) was found significantly associated with plasma iron levels (Supplementary Table 1). The Manhattan, Quantile-Quantile and LocusZoom plots are in Supplementary Fig. 2, and the genomic inflation factor (λ) is 1. For each metal, more information of SNPs that reached the suggestive threshold (P < 1.0 × 10−5) were in Supplementary Data 1. After stepwise conditional analyses, we didn’t find additional significant variants associated with SCLs in FAMHES and associated with plasma iron levels in MEWHC. For replication, we conducted a GWAS meta-analysis of serum/plasma copper levels using GWAS summary statistics from FAMHES and MEWHC. This meta-analysis was conducted using a fixed-effect inverse variance weighted method of PLINK 1.9. Two top SNPs (rs35691438, 0.84 kb away from CP of chromosome 3, Pmeta = 1.63 × 10−9; rs671, on ALDH2 of chromosome 12, Pmeta = 2.21 × 10−10) were found significantly associated with transformed SCLs (Table 2). We also considered the Bonferroni-corrected threshold of <2.38 × 10−9 (5.0 × 10−8/(21 metals)), and both rs35691438 and rs671 still met the Bonferroni-corrected threshold. The Manhattan, Quantile-Quantile, and LocusZoom plots are in Fig. 1, and the genomic inflation factor (λ) is 1.013. Moreover, we found rs671 and rs78069066 are in extremely strong linkage disequilibrium (LD) with r2 = 0.99. We also conducted a GWAS meta-analysis of serum/plasma manganese level using GWAS summary statistics from FAMHES and MEWHC. However, no significant SNP was found.
a Manhattan plot. The X-axis presents the genomic position, while the Y-axis presents the −log10(P) of SNPs. The solid red line is the genome-wide significance threshold (P = 5.0 × 10−8), and the black dashed line is the genome-wide suggestive threshold (P = 1.0 × 10−5). b The Quantile-Quantile plot shows the deviation of the observed from the expected P-values under the null hypothesis of no association. c LocusZoom plot for gene CP. Purple diamond indicates SNP at the locus with the strongest association evidence (rs35691438) and the r2 (degree of linkage disequilibrium) is 1. Each point represents one SNP. d LocusZoom plot for gene ALDH2. Purple diamond indicates SNP at the locus with the strongest association evidence (rs671) and the r2 (degree of linkage disequilibrium) is 1.
Post-GWAS analysis
Four significant independent SNPs (r2 < 0.6) were identified from the GWAS meta-analysis of SCLs via Functional Mapping and Annotation (FUMA) (Supplementary Table 2). A total of 16 potential causal genes involved in SCLs were identified via positional mapping and eQTL (expression quantitative trait loci) mapping of FUMA (Supplementary Data 2). Through gene-based analysis, we found two genes (CP on chromosome 3, P = 1.94 × 10−6; RPH3A on chromosome 12, P = 2.50 × 10−6) reached the significance threshold (P = 0.05/17322 = 2.887 × 10−6). The number 17322 is the quantity of protein-coding genes which are mapped based on the GWAS meta-analysis summary statistics. More information for the first 30 genes with the lowest P-value of gene-based analysis were in Supplementary Table 3. We didn’t find any significant (P < 0.05/15482 = 3.2296 × 10−6) gene-set via gene-set analysis. The number 15482 is the quantity of gene sets. More information for the first 30 gene-sets with the lowest P-value of gene-set analysis were in Supplementary Table 4. We used the significant genes from gene-based analysis and mapped genes to construct the gene expression heatmap; it shows in Supplementary Fig. 3.
MR
In the first round of two-sample MR in East Asian population, we investigated the causal effects of SCLs on 213 deep phenotypes. Firstly, we found 21 exposure-outcome pairs might have causal relationships (Supplementary Table 5). Secondly, for the pairs which had significant horizontal pleiotropy or heterogeneity, we re-performed two-sample MR analyses after excluding the outliers (P-value < 0.05) identified in the Mendelian randomization pleiotropy residual sum and outlier (MR-PRESSO) outlier test. Finally, six exposure-outcome pairs which could draw robust causal conclusions were put into Table 3, including 1 medication usage outcome: drugs affecting bone structure and mineralization (ATC code M05B) (inverse-variance weighted [IVW]: odds ratio [OR] 95% confidence interval [95% CI] 0.887 (0.813 0.967), P = 0.007, Power = 0.83; weighted median [WM]: OR (95% CI) 0.867 (0.771 0.974), P = 0.016, Power = 0.93; MR using the robust adjusted profile score [MR-RAPS]: OR (95% CI) 0.883 (0.809 0.963), P = 0.005, Power = 0.85); and 4 disease outcomes: esophageal varix (IVW: OR (95% CI) 1.998 (1.317 3.032), P = 0.001, Power = 1.00; MR-Egger: OR (95% CI) 6.479 (1.394 30.108), P = 0.041, Power = 1.00; WM: OR (95% CI) 1.773 (1.001 3.141), P = 0.0497, Power = 0.99; MR-RAPS: OR (95% CI) 2.060 (1.372 3.093), P = 0.0005, Power = 1.00); glaucoma (IVW: OR (95% CI) 1.107 (1.031 1.190), P = 0.005, Power = 0.89; WM: OR (95% CI) 1.110 (1.009 1.222), P = 0.032, Power = 0.90; MR-RAPS: OR (95% CI) 1.110 (1.029 1.198), P = 0.007, Power = 0.90); Sleep apnea syndrome (IVW: OR (95% CI) 1.615 (1.212 2.153), P = 0.001, Power = 0.99; WM: OR (95% CI) 1.567 (1.078 2.276), P = 0.019, Power = 0.98; MR-RAPS: OR (95% CI) 1.625 (1.191 2.218), P = 0.002, Power = 0.99); Systemic lupus erythematosus (SLE) (IVW: OR (95% CI) 2.058 (1.422 2.980), P = 0.0001, Power = 1.00; MR-Egger: OR (95% CI) 5.412 (1.344 21.788), P = 0.041, Power = 1.00; WM: OR (95% CI) 1.722 (1.070 2.770), P = 0.025, Power = 0.99; MR-RAPS: OR (95% CI) 2.110 (1.451 3.068), P = 0.00009, Power = 1.00); and 1 biomarker outcome: white blood cell (WBC) count (IVW: β (95% CI) −0.026 (−0.043 −0.008), P = 0.004; WM: β (95% CI) −0.029 (−0.053 −0.005), P = 0.020; MR-RAPS: β (95% CI) −0.027 (−0.046 −0.008), P = 0.005). The results of IVW and MR-RAPS methods for SLE are still significant with the corresponding P-value smaller than the Bonferroni-corrected threshold (0.00023); however, the results of IVW, WM and MR-RAPS methods for the other five exposure-outcome pairs and the results of MR-Egger method for esophageal varix and SLE are suggestively significant with the corresponding P-value smaller than 0.05 but larger than the Bonferroni-corrected threshold (0.00023). Their scatter plots and plots of leave-one-out analyses were displayed in Figs. 2–4. Thirteen exposure-outcome pairs which could not draw robust causal conclusions were put into Supplementary Table 6, and their plots of leave-one-out analysis were in Supplementary Figs. 4−7.
The error bars indicate standard error. a Usage of drugs affecting bone structure and mineralization. b Esophageal varix. c Glaucoma. d Sleep apnea syndrome. MR Mendelian randomization.
The error bars indicate the 95% confidence interval. a Usage of drugs affecting bone structure and mineralization. b Esophageal varix. c Glaucoma. d Sleep apnea syndrome. MR Mendelian randomization.
The error bars in panels a–c indicate standard error. The error bars in panels d–f indicate the 95% confidence interval. a, d Systemic lupus erythematosus. b, e White blood cell count. c, f Gamma-glutamyl transpeptidase levels are exposure factors, and serum copper levels are the outcome. MR Mendelian randomization.
In the second round of two-sample MR in the East Asian population, we investigated the causal effects of 213 deep phenotypes on SCLs. Firstly, we found 16 exposure-outcome pairs might have causal relationships (Supplementary Table 7). Secondly, we re-performed the two-sample MR analysis after excluding the outliers (P-value < 0.05) identified in the MR-PRESSO outlier test for the pairs with significant horizontal pleiotropy or heterogeneity. Finally, one exposure-outcome pair which could draw a robust causal conclusion was put into Table 3. It shows that genetically predicted gamma-glutamyl transpeptidase (GGT) is positively associated with SCLs (IVW: β (95% CI) 0.517 (0.230 0.804), P = 0.0004; WM: β (95% CI) 0.631 (0.194 1.068), P = 0.005; MR-RAPS: β (95% CI) 0.491 (0.210 0.772), P = 0.0006). The results of IVW, WM and MR-RAPS methods for this exposure-outcome pair are suggestively significant with the corresponding P-value smaller than 0.05 but larger than the Bonferroni-corrected threshold (0.00023). Its scatter plot and plot of leave-one-out analysis are displayed in Fig. 4. The information of genetic instrumental variables (IVs) of GGT levels is shown in Supplementary Table 8. Eight exposure-outcome pairs which could not draw robust causal conclusions were put into Supplementary Table 9, and their plots of leave-one-out analysis were in Supplementary Figs. 8, 9.
In the third round of two-sample MR in the European population, we investigated the causal effects of erythrocytes copper levels (ECLs) on 174 deep phenotypes. Firstly, we found 16 exposure-outcome pairs might have causal relationships (Supplementary Table 10). Secondly, for the pairs which have significant horizontal pleiotropy or heterogeneity, we re-performed two-sample MR analyses after excluding the outliers (P-value < 0.05) identified in the MR-PRESSO outlier test. Finally, two exposure-outcome pairs which could draw robust causal conclusions were put into Table 4, including 2 disease outcomes: carpal tunnel syndrome (IVW: OR (95% CI) 1.041 (1.007 1.075), P = 0.018, Power = 0.62; MR-RAPS: OR (95% CI) 1.042 (1.005 1.080), P = 0.024, Power = 0.64); compression fracture (IVW: OR (95% CI) 0.841 (0.748 0.945), P = 0.004, Power = 0.79; WM: OR (95% CI) 0.807 (0.693 0.941), P = 0.006, Power = 0.92; MR-RAPS: OR (95% CI) 0.838 (0.740 0.949), P = 0.005, Power = 0.81). The results of IVW, WM, and MR-RAPS methods for compression fracture are suggestively significant with the corresponding P-value smaller than 0.05 but larger than the Bonferroni-corrected threshold (0.00023). The results of IVW and MR-RAPS methods for carpal tunnel syndrome are suggestively significant with the corresponding P-value smaller than 0.05 but larger than the Bonferroni-corrected threshold (0.00023). Their scatter plots and plots of leave-one-out analyses were in Fig. 5. Fourteen exposure-outcome pairs which could not draw robust causal conclusions were put into Supplementary Table 11, and their plots of leave-one-out analysis were in Supplementary Figs. 10–13.
The error bars in panels a and b indicate standard error. The error bars in panels c and d indicate the 95% confidence interval. a, c Carpal tunnel syndrome. b, d Compression fracture. MR Mendelian randomization.
Discussion
We identified two SNPs (rs35691438 and rs671) that were significantly associated with SCLs in the Chinese population. In the East Asian population, GGT levels had a causal effect on the SCLs, and SCLs have causal effects on six outcomes, namely risks of esophageal varix, glaucoma, sleep apnea syndrome, and SLE, WBC count, and usage of drugs affecting bone structure and mineralization. In the European population, ECLs have causal effects on risks of carpal tunnel syndrome and compression fracture.
We found that SNP rs671 was significantly associated with SCLs. Many studies have reported associations between rs671 and other phenotypes, such as alcohol consumption (drinkers vs non-drinkers), coffee consumption, serum alpha1-antitrypsin levels, esophageal cancer, gout21,22,23,24,25. SNP rs671 is a non-synonymous SNP within the exonic region of ALDH2, which is widely expressed in many human tissues, especially fat and liver. The encoded ALDH2 protein belongs to the aldehyde dehydrogenase family and is involved in the synthesis of acetate from ethanol in the alcohol metabolism pathway. Though there are some researchers reported links between SCLs and alcohol consumption26,27,28,29,30, it still needs further study to clarify the intricate relationship, such as the potential association between rs671 and SCLs may be mediated by alcohol drinking behavior, between SCLs and alcohol consumption behavior on various metabolic conditions31,32,33. For rs35691438, which was significantly associated with SCLs in our study, we found no reports of associations between it and other phenotypes. The CP gene, which is close to rs35691438, is highly expressed mainly in the liver. CP encodes ceruloplasmin, a metalloprotein that binds most of the copper in plasma and is involved in the peroxidation of Fe (II) transferrin to Fe (III) transferrin34,35. Mutations in CP cause aceruloplasminemia, which results in iron accumulation and tissue damage and is associated with diabetes and neurologic abnormalities. To date, a study reported 2 SNPs (rs1175550 and rs2769264) on chromosome 1 significantly (P < 5.0 × 10−8) associated with ECLs6, whereas another study found no SNPs that were associated with whole blood copper levels (BCLs) at the significance threshold (P < 5.0 × 10−8)9. The finding of Evans et al.6 has not been replicated in other studies. In our GWAS, we originally planned to test two SNPs (rs1175550 and rs2769264) on chromosome 1. However, because rs1175550 had a minor allele frequency (MAF) of < 0.01 in the East Asian population and was excluded according to the quality control standards, rs1175550 was not included in the GWAS. Although rs2769264 was found in our SCLs GWAS meta-analysis results, it was not significant (P = 0.4646); the significance threshold was P < 1.0 × 10−5. We found that the MAF of rs1175550 varied widely between the East Asian and European populations in the 1000 Genomes project (http://www.1000genomes.org); 0.2% in East Asians and 21.9% in Europeans. The MAF of rs2769264 also varied widely; 28.6% in East Asians and 16.1% in Europeans. These inter-ethnic discrepancies in the MAFs and the differences between disparate biological samples (serum copper and erythrocytes copper) may explain the highly significant association of rs2769264 in Europeans and no significant association in East Asians36. Moreover, rs2235321, a synonymous SNP on the transmembrane serine protease 6 (TMPRSS6) gene of chromosome 22, was significantly associated with plasma iron levels in MEWHC. A previous GWAS reported five SNPs (rs8177240, rs1800562, rs1799945, rs7385804, rs855791) that were significantly associated with serum iron levels7. We checked these five SNPs in the results of our GWAS of the MEWHC and found that only rs855791 (in TMPRSS6 in strong LD (r2 = 0.90) with rs2235321) was significant (P = 1.19 × 10−8). The number of individuals (687) in our study is quite different from the number (48972) in the study of Benyamin et al.7, which may explain why the other four SNPs were not significant in our study.
In the first round of two-sample MR analyses conducted in the East Asian population, we found that genetically higher SCLs had causal effects on lower trends for the usage of drugs affecting bone structure and mineralization (ATC code M05B), fewer counts of WBC. The genetically increased SCLs also have causal effects on higher risks of esophageal varix, glaucoma, sleep apnea syndrome, and SLE. No direct study of the relationship between SCLs and the usage of M05B Drugs has been reported so far. M05B Drugs include mainly M05BA (bisphosphonates), M05BB (bisphosphonates, combinations), and M05BC (bone morphogenetic proteins). The bisphosphonates and bone morphogenetic proteins are both conducive to bone formation37,38,39. Animal studies have shown that copper deficiency decreased bone strength40, and that the released copper stimulated bone formation41. Our results are indirectly consistent with the results of these studies.
Esophageal varix is related to cirrhosis and portal hypertension42. Until now, seldom studies have directly focused on the link between SCLs and esophageal varix. However, numerous studies have shown that the accumulation of copper can lead to serious liver diseases, including hepatitis and cirrhosis, whether in Wilson’s disease or a non-Wilson’s disease43,44,45,46,47,48. One epidemiological study found that SCLs were increased in patients with liver cirrhosis and variceal bleeding and suggested that variceal bleeding in patients with liver cirrhosis was associated with important imbalances in blood concentrations of zinc, copper, and magnesium49. Although the specific biological mechanism between increased SCLs and the higher risk of esophageal varix is unclear, our MR finding is consistent with what was reported by Dumea et al.49 and further showed the possible causal relationship between them. Therefore, we speculate that esophageal varix may be triggered by cirrhosis, which is caused by the toxicity of copper. An in-depth study of the mechanism needs to be carried out in future work.
Glaucoma is a complex disease with the main risk factor being intraocular pressure. However, previous studies suggested that the further pathogenic factors in glaucoma should include oxidative stress. Oxidative stress can destroy neurons and alter the structure of the trabecular meshwork, which is the principal out channel of aqueous humor, while intraocular pressure continuously increases50,51. A study showed that unbalanced copper levels could lead to oxidative stress through the reduction of glutathione or the Fenton reaction. In a state of oxidative stress, large numbers of reactive oxygen species are generated, which induces subsequent lipid peroxidation and DNA damage52. An early epidemiological study reported that higher mean SCLs were seen in the patients with glaucoma53. Another study showed that the SCLs increased in females with primary open-angle glaucoma compared with that of males with or without primary open-angle glaucoma54. Therefore, copper ions may mediate the occurrence of glaucoma via oxidative stress. Our results are consistent with those of previous studies.
Sleep apnea is a condition in which patients have repeated bouts of shallow breathing or apnea while sleeping. It is related to oxidative stress, high blood pressure, hypercoagulation, and inflammation, and can cause episodic hypoxemia55,56,57. Although copper is closely related to the oxidative stress state of the human body, no previous epidemiological studies have reported associations between SCLs and sleep apnea syndrome. Only one study compared complaints and disease symptoms of obstructive sleep apnea syndrome in patients with Wilson’s disease, which resulted in copper overload, and healthy controls; no significant difference was found between them58. Conversely, we found a strong causal effect of SCLs on sleep apnea syndrome, which suggests this relationship is well worth exploring in future studies.
SLE is an inflammatory autoimmune disease in which the immune system destroys healthy cells and tissues throughout the body59,60. Higher SCLs have been found in patients with SLE61,62,63. However, another study reported that there was no significant difference in SCLs between SLE patients and healthy controls64. The results of traditional epidemiological studies on the relationship between SCLs and SLE disease activity are controversial. Yilmaz et al. found that there was a positive correlation between SCLs and SLE disease activity63, but Sahebari et al. reported that there was a negative correlation between SCLs and SLE disease activity64. Ferns et al. proposed that copper had both pro- and anti-oxidative in vivo effect65. With regard to the pathogenic effect of reactive oxygen species on SLE, previous studies generally believed that reactive oxygen species is an important pathogenic factor, but recent studies believe that insufficient production of reactive oxygen species will lead to the occurrence of SLE66,67. There is a sharp contrast between these research results. Our research provides clues to the current controversial epidemiological results from a genetic point of view, and further corresponding research is needed in the future.
Although no epidemiological research on the association between SCLs and WBC count has been reported, we found some definitive evidence of such an association. Serum copper and zinc concentrations were found to be higher in users of proton pump inhibitors than in controls, and WBC counts may reduce with long‑term proton pump inhibitor therapy68. Copper ions have been shown to aid in the regulation of disulfiram cytotoxicity in myeloid leukemia cells. Copper can help patients with high WBC counts by restoring the sensitivity of high-density leukemic cells to disulfiram and improving the efficacy of anti-leukemic therapies that include disulfiram69. Further research on this relationship between SCLs and WBC count is well worth exploring.
To fully understand whether these 213 deep phenotypes have causal effects on SCLs, we conducted the second round of two-sample MR analyses in the East Asian population. We found that genetically high GGT levels have a causal effect on elevated SCLs. GGT is a hepatic enzyme that aids in the production of the antioxidant glutathione at the outer surface of the plasma membrane70. In clinical laboratories, serum GGT has long been used as a marker for excessive alcohol intake and obesity, particularly in nonalcoholic fatty liver disease, alcoholic liver disease, and chronic hepatitis71. Indeed, serum GGT is a very early and sensitive indicator of oxidative stress and inflammation72. One epidemiological study suggested that serum GGT levels were positively correlated with SCLs73. Our study is the first to describe the causal relationship between them. Although both copper and GGT are involved in oxidative stress and inflammation74,75, the mechanism of the causal relationship between them has not been reported so far. The specific biological mechanism is well worth discovering.
On the basis of our robust first and second round MR results for the East Asians, we inferred that serum copper might have a mediation effect on the causal effects of the GGT levels on the risk of esophageal varix, glaucoma, sleep apnea syndrome, and SLE, WBC count, and the possibility of using drugs affecting bone structure and mineralization. Serum GGT levels have been reported to be associated with the severity of obstructive sleep apnea syndrome76,77. A recent case-control study found that the risk of SLE increased with increased GGT levels78. Another early study reported that an increase of urinary GGT activity was seen in patients with SLE79. A study of Indian children and another study of Korean adults both showed that GGT was positively associated with WBC count80,81. These studies showed associations between GGT levels and sleep apnea syndrome, SLE, and WBC count. We did not find relevant literature for the other diseases and traits. Currently, there are no MR studies that have directly investigated causality between GGT levels and these six phenotypes. Because of the limitation of data acquisition, we did not conduct MR analysis on these possible relationships, and did not deeply explore whether serum copper might play a mediating role or determine the extent of the possible mediating effect of serum copper in these processes. Further independent studies aimed at solving these problems are needed.
To further study the causal effects of BCLs on these deep phenotypes of Europeans, we conducted a third round of two-sample MR analyses in European populations. We found that genetically high ECLs had causal effects on the higher risk of carpal tunnel syndrome and lower risk of compression fracture.
The most common nerve entrapment syndrome is carpal tunnel syndrome, and it has been studied extensively82. Carpal tunnel syndrome is produced when the median nerve in the wrist was compressed while passing through a space-constrained osteofibrous canal83. Although no epidemiological study of the association between ECLs and carpal tunnel syndrome has been reported so far, our results provide valuable evidence that is worth exploring. The causal effect of copper on carpal tunnel syndrome be based on the neurotoxicity of copper ions and the oxidative stress induced by copper. To corroborate this relationship, independent cohort and biological function studies are needed.
So far, we have not found epidemiological studies of ECLs and compression fracture. However, many studies have shown that copper plays a vital role in maintaining bone health. Copper eliminates bone-free radicals generated by osteoclast activity as a cofactor in antioxidant enzymes84, and copper has been suggested to suppress osteoclastic resorption directly85. Conversely, a national cross-sectional study in the United States suggested that high SCLs were strongly related to increased total fracture86. The differences among these studies can be attributed to the different ethnicities, sample types, and other possible confounding factors that were not controlled. Our finding that genetically high ECLs were protective against the risk of compression fracture provides a reference for subsequent epidemiological studies and a more in-depth analysis of the biological mechanisms involved.
Interestingly, we found that two diseases, food allergy, and urticaria, both appeared in the European and East Asian MR results where robust conclusions could not be drawn. Although our results did not show a robust causal relationship between these diseases and BCLs, our research gives suggestive clues on the causal relationship between them. Moreover, we found that SLE and the usage of drugs affecting bone structure and mineralization both appeared in the robust results of the East Asian MR and results where we could not draw robust conclusions of the European MR analyses. We also found that compression fracture appeared in the robust results of European MR and results where we could not draw robust conclusions of East Asian MR analyses. Because we used genetic IVs related to SCLs for the East Asian MR analyses and genetic IVs related to ECLs for the European MR analyses, whether differences of significance for the same disease in different populations can be attributed to ethnic differences needs to be studied further.
We noted that another BCLs MR analysis from UK Biobank indicated that genetically increased ECLs were causally associated with a reduced risk of iron deficiency anemia, whereas this association was not significant in our study18. We can suggest two possible explanations for this difference. (1) We used the GWAS meta-analysis summary statistics of UK Biobank and FinnGen for iron deficiency anemia, which increased the number of cases compared with the number for the single GWAS summary statistic from UK Biobank or FinnGen. However, because using the GWAS meta-analysis summary statistics may also greatly increase the overall number, the proportion of cases in the total number for a specific disease may decrease. Such decreased proportions will lead to weaker statistical power, which may have led to our failure to detect a causal effect between iron deficiency anemia and ECLs. (2) In their study, Zhou et al. used only two IVs that were associated with ECLs, and then conducted a PheWAS-MR (phenome-wide association study and MR) analysis. After the PheWAS, some phenotypes unrelated to the selected genetic variants were excluded, and their MR analysis did not include them. However, we included 12 SNPs as IVs in the MR analysis, which increased the statistical power. This may explain the difference to a certain extent. Our significantly robust results were not referred to in their paper.
To the best of our knowledge, ours is the first reported systematic GWAS of 21 BMLs using two independent prospective East Asian cohorts. Compared with the GWAS of 11 BMLs conducted in the European population by Ng et al.9, we have included more metals and discovered loci for SCLs. To our knowledge, this may also be the first study to use bidirectional two-sample MR analysis with hypothesis-free investigations of the causal effects between BCLs and 213 deep phenotypes in the East Asian population. We also used two-sample MR analysis with hypothesis-free investigations of the causal effects of BCLs on 174 deep phenotypes in the European population. This approach provides a more comprehensive understanding of the impact of BCLs on the health of East Asians and Europeans. Moreover, the impacts of potential weak instrumental variable bias, heterogeneity, and pleiotropic effects on the results were also considered. Possible biases caused by horizontal pleiotropy were fully taken into account in our analyses process. We used the MR-PRESSO outlier test and other robust methods for the MR analysis, and no horizontal pleiotropy was detected in the results. The PhenoScanner database or the Ensembl Project would not necessarily distinguish between vertical and horizontal pleiotropy, and the former would not bias the MR analysis87. Furthermore, because the biological mechanisms and functions of numerous genetic variants are unknown, we did not use these variants to assess the IVs associated with potential confounding factors. The genetic IVs we used explain 11.09% of the variance in SCLs and 17.06% of the variance in ECLs, which increased the statistical power to identify significant associations88. F statistics were considered to avoid weak instrument bias, and I2, which evaluates the heterogeneity scale of IVs, was used to reduce the impact of potential heterogeneity. The presence of horizontal pleiotropy was examined or corrected using MR-Egger tests and MR-PRESSO. We used the leave-one-out sensitivity analysis to ensure no results were impacted by specific variants.
Our study has several limitations. Although we conducted the GWAS on 21 serum and plasma metals, because of differences in the concentrations of the same metals in serum and plasma, we only did the meta-analyses of the serum copper and serum manganese. The GWASs of the remaining 19 serum/plasma metals need to be combined with other appropriate independent cohorts for further replication in the future. Because relatively small numbers of cases were used in the GWAS, it was difficult to find significant SNPs. Despite this, we found loci for SCLs; however, more research is needed to verify our conclusions. In the two-sample MR analysis, we did not find genetic IVs of ECLs or SCLs in the East Asian and European populations, respectively, which limited our comprehensive study of the causal effects of BCLs on 213 deep phenotypes in these populations. For the same reason, we separately included the genetic IVs related to the SCLs of East Asians and the ECLs of Europeans, which had an impact on the horizontal comparison of differences in the same exposure between ethnicities. Although approximately 800,000 individuals were included in the MR analyses, the number of cases of specific diseases was still small, which reduced the statistical power. Moreover, because access to data was limited, our MR results were not verified in large independent databases.
Methods
Subjects
To conduct comprehensive GWASs of 22 serum/plasma metal levels in the Chinese population, we used the serum metal levels data obtained from FAMHES in Guangxi Fangchenggang and plasma metal levels data obtained from MEWHC. FAMHES included 4303 Chinese males aged 17–88 years old at baseline. Their questionnaires and physical examinations at baseline were completed in the Medical Center of Fangchenggang First People’s Hospital between September 2009 and December 200989. MEWHC included 825 Chinese with their follow-up completed in 2017. More detailed information about FAMHES, MEWHC were introduced elsewhere90,91. Our study was approved by the Ethics and Human Subject Committee of Guangxi Medical University. All the subjects provided written informed consent before participating in the study.
Phenotypes
All blood samples are stored at −80 °C until analysis. We used an inductively coupled plasma mass spectrometry (ICP-MS, Perkin Elmer NexION 350xX, USA) to measure the serum metal concentrations of FAMHES and plasma metal concentrations of MEWHC, respectively. Our previous research also introduced the corresponding methods89,92. Specifically, each serum/plasma sample was diluted 20 times by using acidic solution containing 0.01% TritonTM X-100, 0.5% n-butyl alcohol and 1% nitric acid. We used ClinChek® human serum controls for trace elements (order no.8880 and order no.8881, Recipe Chemicals, Germany) and Standard Reference Material®1640a (Trace Elements in Natural Water, National Institute of Standards and Technology, USA) as standard reagent for quality assurance in serum measurement. Moreover, we used ClinChek® human plasma controls for trace elements (order no.8,885; Recipe Chemicals, Germany) and Standard Reference Material®1640a for quality assurance in plasma measurement. The following 22 metals (aluminum, antimony, arsenic, barium, cadmium, calcium, cobalt, copper, chromium, iron, lead, magnesium, manganese, molybdenum, nickel, rubidium, selenium, strontium, titanium, tin, vanadium, and zinc) were measured in both cohorts. The samples with concentrations less than the limit of detection were given a value equal to the limit of detection divided by √2 as concentrations. More than 20% of samples in FAMHES did not reach the limit of detection for antimony levels. Therefore, in the subsequent analysis, we included the remaining 21 metals, excluding antimony.
Genotyping, imputation, and quality control
In our study, a total of 2020 subjects were chosen from FAMHES and genotyped using the Illumina Omini one platform93. We excluded individuals with (1) individuals call rate <0.95, (2) genotypic and phenotypic sex mismatch, (3) heterozygosity rate deviating more than three standard deviations from the mean, (4) kinship coefficient > 0.0884, (5) missingness of phenotypes or covariates. We also excluded SNPs with (1) call rate <0.95, (2) MAF < 0.05, (3) Hardy–Weinberg equilibrium P-value (PHWE) < 1.0 × 10−6. Then we used liftOverPlink (https://github.com/sritchie73/liftOverPlink) to convert genome position from hg18 to hg19. Snpflip (https://github.com/biocore-ntnu/snpflip) was used to fix the problem with reverse and ambiguous strand SNPs. Before phasing and imputation, we redid the same quality control procedure mentioned above. Eagle v2.4.1 (https://alkesgroup.broadinstitute.org/Eagle/) and Minimac4 (https://github.com/statgen/Minimac4) were used to phase and impute genotyped data of autosomal chromosomes, respectively. In the procedure of phasing and imputation, we chose 1000 G Phase3 v5 EAS (GRCh37/hg19) as a reference panel. After phasing and imputation, SNPs with (1) MAF < 0.01, (2) minor allele count < 20, (3) PHWE < 1.0 × 10−6, (4) imputation quality (Rsq) < 0.7 were excluded. A total of 805 subjects were included from MEWHC and genotyped using Infinium Asian Screening Array-24 + v1.0 Kit (ASA). Because of the lack of X chromosome genotype data, we did not compare whether the genotypic and the phenotypic sex match. The remaining procedure of quality control was the same as what was done in FAMHES. Finally, before conducting GWAS, we removed the outliers whose metal concentrations were more than 4.5 inter-quartile ranges away from the median of each metal concentration. The rank-based inverse normal transformation was used to transform each metal concentration to approximate normality94,95,96.
GWAS and meta-analysis
For each metal, we conducted a GWAS respectively in both cohorts using a generalized linear model adjusting age, sex, body mass index, ethnicity, first ten principal components, alcoholic drinking, and smoking status. PLINK 2.0 (www.cog-genomics.org/plink/2.0/) was selected to perform the GWAS under assumptions of dosage additive effects97. P = 5.0 × 10−8 was set as the genome-wide significance threshold, and P = 1.0 × 10−5 was selected as the genome-wide suggestive threshold. Stepwise conditional analysis was conducted to find the additional independent significant signal. For replication, we planned to perform meta-analyses using GWAS summary statistics of the same metal from both FAMHES and MEWHC. However, considering the difference between serum and plasma metal concentrations, we conducted the non-parametric test on the concentrations of each metal between the two cohorts via IBM SPSS Statistics 25 (https://www.ibm.com/support/pages/downloading-ibm-spss-statistics-25). Finally, since there was no statistically significant difference in copper concentrations and manganese concentrations between the two cohorts, we respectively conducted GWAS meta-analysis on the copper and manganese levels. Because of the significant statistical difference between the concentrations of other same metals in the two cohorts, the meta-analysis was not performed in the remaining metals. The Bonferroni-corrected threshold of < 2.38 × 10−9 (5.0 × 10−8/(21 metals)) in the context of meta-analysis was considered. Since we did not find any SNP significantly associated with manganese levels in the results of GWAS meta-analysis on manganese levels, and find two SNPs significantly associated with copper levels in the results of GWAS meta-analysis on copper levels, we only conducted the subsequent analysis of blood copper.
Post-GWAS analysis
FUMA is an internet-based platform developed to annotate, prioritize, visualize, and interpret GWAS results98,99. We conducted annotation, functional mapping, gene-based analysis, gene-set analysis for the results of GWAS meta-analysis of SCLs via SNP2GENE function module of FUMA100,101,102, and then we used both mapped genes and the significant gene from gene-based analysis to construct gene expression heatmap via GENE2FUNC function module of FUMA to explore their expression states in various tissues103,104.
MR
To investigate the causal relationships between SCLs and 213 deep phenotypes (152 diseases, 38 biomarkers, and 23 medication usages) from BioBank Japan in the East Asian population20, we performed bidirectional MR analysis. Detailed information about the 213 deep phenotypes is shown in Supplementary Data 3, and their GWAS summary statistics can be downloaded from the website (https://pheweb.jp/downloads).
In the first round two-sample MR analyses, we set SCLs as exposure and 213 deep phenotypes as outcomes. To identify the genetic IVs, we firstly selected genetic variants which were associated (P < 1.0 × 10−5) with SCLs from GWAS meta-analysis as candidate IVs105. Secondly, we filtered out those SNPs in strong LD since strong linkage disequilibrium will bias the subsequent analysis. In our study, we conducted LD clumping process (window size = 10,000 kb, r2 < 0.001) using 1000G Phase3 v5 EAS (GRCh37/hg19) as a reference panel, and SNPs with r2 < 0.001, MAF > 0.01 were retained. Thirdly, we evaluated the heterogeneity of each candidate IV using I2 and P-value for Cochran’s Q statistic. F-statistic was considered to avoid weak IV bias. In this step, we included SNPs with I2 < 50, P-value for Cochran’s Q statistic > 0.05, and F-statistic >10106. Fourthly, in the harmonizing process of two-sample MR analyses, both ambiguous and palindrome SNPs would be corrected or directly excluded. Since there was no need, we did not use any proxy SNPs. Finally, a total of 11 SNPs were included as IVs in the first round two-sample MR analyses. The total variance explained by the selected 11 SNPs was calculated using the following equation (Eq. 1):
where \({{R}}_{{{{{{\rm{total}}}}}}}^{2}\) referred to the total variance explained by the selected 11 SNPs, fn referred to the effect allele frequency, n referred to the number of the selected SNPs (n = 11), and βn referred to the additive effect94. All the 11 SNPs totally explained 11.09% of the transformed SCLs variance (Supplementary Table 12).
To further research whether the 213 deep phenotypes from BioBank Japan have causal effects on SCLs, we conducted the second round two-sample MR analyses, which is an inverse MR analysis comparing with the first round. We selected genetic variants which were associated (P < 1.0 × 10−8) with each phenotype from the GWAS summary statistics in BioBank Japan as candidate IVs. The rest of the process of selecting genetic IVs is the same as that done in the first round.
To comprehensively determine the causal effects of BCLs on these 213 deep phenotypes in humans, we further used the GWAS meta-analyses summary statistics of UK Biobank and FinnGen to perform the third round of two-sample MR analyses. The GWAS meta-analysis summary statistics could be downloaded at the website (https://pheweb.jp/downloads). Because of lacking serum/plasma copper levels GWAS summary statistics for European populations, we finally used the SNPs which were independently (r2 < 0.05) associated (P < 1.0 × 10−5) with ECLs as IVs6. Except that the reference panel used in the LD clumping process is 1000G Phase3 v5 EUR (GRCh37/hg19), the rest process of selecting genetic IVs is the same as that done in the first round. The details of the 12 SNPs, which were finally included as genetic IVs, are shown in Supplementary Table 13.
Because of the lack of GWAS summary statistics for the 38 biomarkers, we only conducted the two-sample MR analyses between ECLs and 174 deep phenotypes (151 diseases and 23 medication usage) in the end. More information about the 174 deep phenotypes is shown in Supplementary Data 4.
All three rounds of two-sample MR analyses followed the flow chart (Fig. 6). Firstly, we used IVW, MR-Egger, and WM methods of the TwoSampleMR package to identify exposure-outcome pairs that might have causal relationships107,108,109,110. Significant (P < 0.05) difference between y-intercept of MR-Egger and 0 represents that there is significant horizontal pleiotropy. P-value of Cochran’s Q statistics of IVW and MR-Egger < 0.05 represents that there is significant heterogeneity. If there was significant horizontal pleiotropy or heterogeneity, the MR-PRESSO outlier test would be used to detect the outlier, which usually leads to horizontal pleiotropy111. The NbDistribution (number of elements to simulate to form the null distribution to compute empirical P-values) in the MR-PRESSO test was set to 20,000. After removing the outliers, we re-performed two-sample MR analyses with IVW, MR-Egger, and WM methods. Based on the results of the leave-one-out sensitivity analysis, horizontal pleiotropy tested via MR-Egger, and heterogeneity tested via IVW and MR-Egger, we divide the results into the section where robust conclusions can be drawn and the section where robust conclusions cannot be drawn. Then we performed MR-RAPS to re-estimate causality in exposure-outcome pairs of robust conclusions section to eliminate the estimation bias of weak IVs112. In addition, we also considered the corresponding Bonferroni-corrected threshold < 0.00023 (0.05/213 outcomes) and the suggestively significant threshold (0.00023 ≤ P ≤ 0.05) in MR analysis. Finally, we calculated the statistical power of IVW, MR-Egger, MR-RAPS, and WM methods using the methods which are based on the non-centrality parameter113,114. Because of the lacking variance of the exposure variable and outcome variable, we didn’t calculate the statistical power for quantitative trait outcomes. All two-sample MR analyses were performed in R version 4.0.5 using the TwoSampleMR package109. A two-sided P value < 0.05 was used to determine statistical significance.
IVs instrumental variables, MR Mendelian randomization, IVW inverse-variance weighted, WM weighted median, MR-PRESSO Mendelian randomization pleiotropy residual sum and outlier.
Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.
Data availability
The GWAS summary statistics of 213 deep phenotypes can be downloaded from the website (https://pheweb.jp/downloads). The GWAS summary statistics of serum/plasma metal levels have been provided to the NHGRI-EBI GWAS Catalog and the study accession numbers are GCST90100517, GCST90100518, GCST90100519, GCST90100520, GCST90100521, GCST90100522, GCST90100523, GCST90100524, GCST90100525, GCST90100526, GCST90100527, GCST90100528, GCST90100529, GCST90100530, GCST90100531, GCST90100532, GCST90100533, GCST90100534, GCST90100535, GCST90100536, GCST90100537, GCST90100538, GCST90100539, GCST90100540, GCST90100541, GCST90100542, GCST90100543, GCST90100544, GCST90100545, GCST90100546, GCST90100547, GCST90100548, GCST90100549, GCST90100550, GCST90100551, GCST90100552, GCST90100553, GCST90100554, GCST90100555, GCST90100556, GCST90100557, GCST90100558. The source data for graphs and charts can be downloaded from Figshare (https://figshare.com) through the https://doi.org/10.6084/m9.figshare.17696942. The datasets generated and analyzed during the current study are available in the Genome variation Map (GVM) of National Genomics Data Center (NGDC) (Accession Number: GVM000052). Please contact the corresponding author for more information if necessary.
Code availability
The liftOverPlink (https://github.com/sritchie73/liftOverPlink); Snpflip (https://github.com/biocore-ntnu/snpflip); Eagle v2.4.1 (https://alkesgroup.broadinstitute.org/Eagle/); Minimac4 (https://github.com/statgen/Minimac4); PLINK 2.0 (www.cog-genomics.org/plink/2.0/); IBM SPSS Statistics 25 (https://www.ibm.com/support/pages/downloading-ibm-spss-statistics-25); FUMA (https://fuma.ctglab.nl); R (https://www.r-project.org).
References
Kim, J. K. et al. Elucidating the role of metal ions in carbonic anhydrase catalysis. Nat. Commun. 11, 4557 (2020).
Zoroddu, M. A. et al. The essential metals for humans: A brief overview. J. Inorg. Biochem. 195, 120–129 (2019).
Wang, Y. X. et al. Associations of urinary metal levels with serum hormones, spermatozoa apoptosis, and sperm DNA damage in a Chinese population. Environ. Int. 94, 177–188 (2016).
Meeker, J. D. et al. Environmental exposure to metals and male reproductive hormones: circulating testosterone is inversely associated with blood molybdenum. Fertil. Steril. 93, 130–140 (2010).
O’Seaghdha, C. M. et al. Meta-analysis of genome-wide association studies identifies six new Loci for serum calcium concentrations. PLoS Genet. 9, e1003796 (2013).
Evans, D. M. et al. Genome-wide association study identifies loci affecting blood copper, selenium, and zinc. Hum. Mol. Genet. 22, 3998–4006 (2013).
Benyamin, B. et al. Novel loci affecting iron homeostasis and their effects in individuals at risk for hemochromatosis. Nat. Commun. 5, 4926 (2014).
Meyer, T. E. et al. Genome-wide association studies of serum magnesium, potassium, and sodium concentrations identify six Loci influencing serum magnesium levels. PLoS Genet. https://doi.org/10.1371/journal.pgen.1001045 (2010).
Ng, E. et al. Genome-wide association study of toxic metals and trace elements reveals novel associations. Hum. Mol. Genet. 24, 4739–4745 (2015).
Yarmohammadi, H. et al. Serum calcium and risk of sudden cardiac arrest in the general population. Mayo Clin. Proc. 92, 1479–1485 (2017).
Fang, A. P. et al. Serum copper and zinc levels at diagnosis and hepatocellular carcinoma survival in the Guangdong Liver Cancer Cohort. Int. J. Cancer 144, 2823–2832 (2019).
Vaugier, C. et al. Serum iron protects from renal postischemic injury. J. Am. Soc. Nephrol. 28, 3605–3615 (2017).
Kunutsor, S. K., Whitehouse, M. R., Blom, A. W. & Laukkanen, J. A. Low serum magnesium levels are associated with increased risk of fractures: A long-term prospective cohort study. Eur. J. Epidemiol. 32, 593–603 (2017).
Xu, C. et al. Serum nickel is associated with craniosynostosis risk: Evidence from humans and mice. Environ. Int. 146, 106289 (2021).
Nie, X. et al. Lead and cadmium exposure, higher thyroid antibodies and thyroid dysfunction in Chinese women. Environ. Pollut. 230, 320–328 (2017).
Davey Smith, G. & Hemani, G. Mendelian randomization: Genetic anchors for causal inference in epidemiological studies. Hum. Mol. Genet. 23, R89–R98 (2014).
Larsson, S. C., Burgess, S. & Michaëlsson, K. Association of genetic variants related to serum calcium levels with coronary artery disease and myocardial infarction. Jama 318, 371–380 (2017).
Zhou, J. et al. The causal effects of blood iron and copper on lipid metabolism diseases: Evidence from phenome-wide Mendelian randomization study. Nutrients https://doi.org/10.3390/nu12103174 (2020).
Qu, Z. et al. Relationship between serum nutritional factors and bone mineral density: A Mendelian randomization study. J. Clin. Endocrinol. Metab. 106, e2434–e2443 (2021).
Sakaue, S. et al. A cross-population atlas of genetic associations for 220 human phenotypes. Nat. Genet. 53, 1415–1424 (2021).
Nakayama, A. et al. Subtype-specific gout susceptibility loci and enrichment of selection pressure on ABCG2 and ALDH2 identified by subtype genome-wide meta-analyses of clinically defined gout patients. Ann. Rheum. Dis. 79, 657–665 (2020).
Matoba, N. et al. GWAS of 165,084 Japanese individuals identified nine loci associated with dietary habits. Nat. Hum. Behav. 4, 308–316 (2020).
Jorgenson, E. et al. Genetic contributors to variation in alcohol consumption vary by race/ethnicity in a large multi-ethnic genome-wide association study. Mol. Psychiatry 22, 1359–1367 (2017).
Setoh, K. et al. Three missense variants of metabolic syndrome-related genes are associated with alpha-1 antitrypsin levels. Nat. Commun. 6, 7754 (2015).
Cui, R. et al. Functional variants in ADH1B and ALDH2 coupled with alcohol and smoking synergistically enhance esophageal cancer risk. Gastroenterology 137, 1768–1775 (2009).
Zhang, H. Q. et al. Serum zinc, copper, and zinc/copper in healthy residents of Jinan. Biol. Trace Elem. Res. 131, 25–32 (2009).
Schuhmacher, M., Domingo, J. L. & Corbella, J. Zinc and copper levels in serum and urine: Relationship to biological, habitual and environmental factors. Sci. Total Environ. 148, 67–72 (1994).
Halsted, C. H. Nutrition and alcoholic liver disease. Semin Liver Dis. 24, 289–304 (2004).
Shibazaki, S., Uchiyama, S., Tsuda, K. & Taniuchi, N. Copper deficiency caused by excessive alcohol consumption. BMJ Case Rep. https://doi.org/10.1136/bcr-2017-220921 (2017).
Rahelić, D., Kujundzić, M., Romić, Z., Brkić, K. & Petrovecki, M. Serum concentration of zinc, copper, manganese, and magnesium in patients with liver cirrhosis. Coll. Antropol. 30, 523–528 (2006).
Collins, J. F., Prohaska, J. R. & Knutson, M. D. Metabolic crossroads of iron and copper. Nutr. Rev. 68, 133–147 (2010).
Aigner, E. et al. Copper availability contributes to iron perturbations in human nonalcoholic fatty liver disease. Gastroenterology 135, 680–688 (2008).
Grochowski, C. et al. Serum iron, magnesium, copper, and manganese levels in alcoholism: A systematic review. Molecules https://doi.org/10.3390/molecules24071361 (2019).
Hellman, N. E. & Gitlin, J. D. Ceruloplasmin metabolism and function. Annu. Rev. Nutr. 22, 439–458 (2002).
Mukhopadhyay, C. K., Attieh, Z. K. & Fox, P. L. Role of ceruloplasmin in cellular iron uptake. Science 279, 714–717 (1998).
Vinayagamoorthy, N. et al. Association of common variants in the calcium-sensing receptor gene with serum calcium levels in East Asians. J. Hum. Genet. 60, 407–412 (2015).
von Knoch, F. et al. Effects of bisphosphonates on proliferation and osteoblast differentiation of human bone marrow stromal cells. Biomaterials 26, 6941–6949 (2005).
Hodsman, A. B. et al. Parathyroid hormone and teriparatide for the treatment of osteoporosis: A review of the evidence and suggested guidelines for its use. Endocr. Rev. 26, 688–703 (2005).
Bem, R., Jirkovská, A., Fejfarová, V., Skibová, J. & Jude, E. B. Intranasal calcitonin in the treatment of acute Charcot neuroosteoarthropathy: A randomized controlled trial. Diabetes Care 29, 1392–1394 (2006).
Jonas, J. et al. Impaired mechanical strength of bone in experimental copper deficiency. Ann. Nutr. Metab. 37, 245–252 (1993).
Prinz, C., Elhensheri, M., Rychly, J. & Neumann, H. G. Antimicrobial and bone-forming activity of a copper coated implant in a rabbit model. J. Biomater. Appl. 32, 139–149 (2017).
Kovacs, T. O. G. & Jensen, D. M. Varices: Esophageal, gastric, and rectal. Clin. Liver Dis. 23, 625–642 (2019).
Lutsenko, S. Modifying factors and phenotypic diversity in Wilson’s disease. Ann. N. Y Acad. Sci. 1315, 56–63 (2014).
Rosencrantz, R. & Schilsky, M. Wilson disease: Pathogenesis and clinical considerations in diagnosis and treatment. Semin Liver Dis. 31, 245–259 (2011).
Strausak, D., Mercer, J. F., Dieter, H. H., Stremmel, W. & Multhaup, G. Copper in disorders with neurological symptoms: Alzheimer’s, Menkes, and Wilson diseases. Brain Res. Bull. 55, 175–185 (2001).
Tanner, M. S. et al. Increased hepatic copper concentration in Indian childhood cirrhosis. Lancet 1, 1203–1205 (1979).
Baker, A. et al. Copper-associated liver disease in childhood. J. Hepatol. 23, 538–543 (1995).
Scheinberg, I. H. & Sternlieb, I. Wilson disease and idiopathic copper toxicosis. Am. J. Clin. Nutr. 63, 842s–845s (1996).
Dumea, M., Cimpoesu, D. C. & Nechifor, M. Divalent cations profile in patients with liver cirrhosis and variceal bleeding. Rev. Med Chir. Soc. Med Nat. Iasi 119, 1166–1173 (2015).
Chrysostomou, V., Rezania, F., Trounce, I. A. & Crowston, J. G. Oxidative stress and mitochondrial dysfunction in glaucoma. Curr. Opin. Pharm. 13, 12–15 (2013).
Awai-Kasaoka, N. et al. Oxidative stress response signaling pathways in trabecular meshwork cells and their effects on cell viability. Mol. Vis. 19, 1332–1340 (2013).
Jomova, K. & Valko, M. Advances in metal-induced oxidative stress and human disease. Toxicology 283, 65–87 (2011).
Akyol, N., Değer, O., Keha, E. E. & Kiliç, S. Aqueous humour and serum zinc and copper concentrations of patients with glaucoma and cataract. Br. J. Ophthalmol. 74, 661–662 (1990).
Fick, A., Jünemann, A., Michalke, B., Lucio, M. & Hohberger, B. Levels of serum trace elements in patients with primary open-angle glaucoma. J. Trace Elem. Med. Biol. 53, 129–134 (2019).
Marin, J. M. et al. Association between treated and untreated obstructive sleep apnea and risk of hypertension. Jama 307, 2169–2176 (2012).
Lavie, L. Oxidative stress in obstructive sleep apnea and intermittent hypoxia–revisited–the bad ugly and good: Implications to the heart and brain. Sleep. Med. Rev. 20, 27–45 (2015).
Phillips, C. L. et al. Effects of continuous positive airway pressure on coagulability in obstructive sleep apnoea: A randomised, placebo-controlled crossover study. Thorax 67, 639–644 (2012).
Cochen De Cock, V., Lacombe, S., Woimant, F. & Poujois, A. Sleep disorders in Wilson’s disease. Sleep. Med. 83, 299–303 (2021).
Kiriakidou, M. & Ching, C. L. Systemic lupus erythematosus. Ann. Intern Med. 172, Itc81–itc96 (2020).
Kamen, D. L. Environmental influences on systemic lupus erythematosus expression. Rheum. Dis. Clin. North Am. 40, 401–412, vii (2014).
González, S. et al. Microbiota and oxidant-antioxidant balance in systemic lupus erythematosus. Nutr. Hosp. 34, 934–941 (2017).
Tóth, C. N. et al. Elemental analysis of whole and protein separated blood serum of patients with systemic lupus erythematosus and Sjögren’s syndrome. Biol. Trace Elem. Res. 179, 14–22 (2017).
Yilmaz, A., Sari, R. A., Gundogdu, M., Kose, N. & Dag, E. Trace elements and some extracellular antioxidant proteins levels in serum of patients with systemic lupus erythematosus. Clin. Rheumatol. 24, 331–335 (2005).
Sahebari, M. et al. Association between serum trace element concentrations and the disease activity of systemic lupus erythematosus. Lupus 23, 793–801 (2014).
Ferns, G. A., Lamb, D. J. & Taylor, A. The possible role of copper ions in atherogenesis: The Blue Janus. Atherosclerosis 133, 139–152 (1997).
Urbonaviciute, V., Luo, H., Sjöwall, C., Bengtsson, A. & Holmdahl, R. Low production of reactive oxygen species drives systemic lupus erythematosus. Trends Mol. Med. 25, 826–835 (2019).
Zhong, J. et al. Association of NOX2 subunits genetic variants with autoimmune diseases. Free Radic. Biol. Med. 125, 72–80 (2018).
Kaczmarczyk, O. et al. Effect of long-term proton pump inhibitor therapy on complete blood count parameters and selected trace elements: a pilot study. Pol. Arch. Intern. Med. 130, 179–186 (2020).
Navrátilová, J. et al. Copper ions regulate cytotoxicity of disulfiram to myeloid leukemia cells. Int. J. Mol. Med. 24, 661–670 (2009).
Karp, D. R., Shimooku, K. & Lipsky, P. E. Expression of gamma-glutamyl transpeptidase protects ramos B cells from oxidation-induced cell death. J. Biol. Chem. 276, 3798–3804 (2001).
Lippi, G., Danese, E. & Montagnana, M. Serum gamma-glutamyltransferase and alanine aminotransferase levels are correlated with hematocrit in a general population of outpatients. Scand. J. Clin. Lab Invest. 73, 95–96 (2013).
Lim, J. S. et al. Is serum gamma-glutamyltransferase inversely associated with serum antioxidants as a marker of oxidative stress? Free Radic. Biol. Med. 37, 1018–1023 (2004).
Peng, Y. F., Wang, C. F. & Pan, G. G. Relation of serum γ-glutamyl transferase activity with copper in an adult population. Clin. Chem. Lab Med. 55, 1907–1911 (2017).
Vivoli, G., Bergomi, M., Rovesti, S., Pinotti, M. & Caselgrandi, E. Zinc, copper, and zinc- or copper-dependent enzymes in human hypertension. Biol. Trace Elem. Res. 49, 97–106 (1995).
Whitfield, J. B. Gamma glutamyl transferase. Crit. Rev. Clin. Lab Sci. 38, 263–355 (2001).
Kanbay, A. et al. Serum gamma-glutamyl transferase activity is an independent predictor for cardiovascular disease in obstructive sleep apnea syndrome. Respir. Med. 105, 637–642 (2011).
Koseoglu, H. I. et al. Serum levels of gamma-glutamyl transferase are associated with cardiovascular disease in obstructive sleep apnea syndrome. Ann. Saudi Med. 33, 584–590 (2013).
Zhang, W. et al. Association between gamma-glutamyl transferase, total bilirubin and systemic lupus erythematosus in Chinese women. Front. Immunol. 12, 682400 (2021).
Delektorskaya, L., Janushkevich, T. & Okunev, D. The significance of the assays of urinary enzymes activity in patients with systemic lupus erythematosus. Z. Med. Lab Diagn. 31, 375–379 (1990).
Lee, Y. J. et al. Association of serum gamma-glutamyltransferase with C-reactive protein levels and white blood cell count in Korean adults. Clin. Chem. Lab Med. 46, 1410–1415 (2008).
Ortega, E., Koska, J., Salbe, A. D., Tataranni, P. A. & Bunt, J. C. Serum gamma-glutamyl transpeptidase is a determinant of insulin resistance independently of adiposity in Pima Indian children. J. Clin. Endocrinol. Metab. 91, 1419–1422 (2006).
Olney, R. K. Carpal tunnel syndrome: Complex issues with a “simple” condition. Neurology 56, 1431–1432 (2001).
Padua, L. et al. Carpal tunnel syndrome: Clinical features, diagnosis, and management. Lancet Neurol. 15, 1273–1284 (2016).
Kubiak, K., Klimczak, A., Dziki, Ł., Modranka, R. & Malinowska, K. Influence of copper (II) complex on the activity of selected oxidative enzymes. Pol. Merkur Lekarski 28, 22–25 (2010).
Li, B. B. & Yu, S. F. In vitro study of the effects of copper ion on osteoclastic resorption in various dental mineralized tissues. Zhonghua Kou Qiang Yi Xue Za Zhi 42, 110–113 (2007).
Qu, X. et al. Serum copper levels are associated with bone mineral density and total fracture. J. Orthop. Transl. 14, 34–44 (2018).
Wu, F., Huang, Y., Hu, J. & Shao, Z. Mendelian randomization study of inflammatory bowel disease and bone mineral density. BMC Med. 18, 312 (2020).
Pierce, B. L. & Burgess, S. Efficient design for Mendelian randomization studies: Subsample and 2-sample instrumental variable estimators. Am. J. Epidemiol. 178, 1177–1184 (2013).
Li, L. et al. Non-linear association of serum molybdenum and linear association of serum zinc with nonalcoholic fatty liver disease: Multiple-exposure and Mendelian randomization approach. Sci. Total Environ. 720, 137655 (2020).
He, M. et al. A genome wide association study of genetic loci that influence tumour biomarkers cancer antigen 19-9, carcinoembryonic antigen and α fetoprotein and their associations with cancer risk. Gut 63, 143–151 (2014).
Lv, Y. et al. Rationale, design and baseline results of the Guangxi manganese-exposed workers healthy cohort (GXMEWHC) study. BMJ Open 4, e005070 (2014).
Ge, X. et al. Plasma metals and serum bilirubin levels in workers from manganese-exposed workers healthy cohort (MEWHC). Environ. Pollut. 258, 113683 (2020).
Yang, X. et al. Genome-wide association study for serum complement C3 and C4 levels in healthy Chinese subjects. PLoS Genet. 8, e1002916 (2012).
Kanai, M. et al. Genetic analysis of quantitative traits in the Japanese population links cell types to complex human diseases. Nat. Genet. 50, 390–400 (2018).
Hagenbeek, F. A. et al. Heritability estimates for 361 blood metabolites across 40 genome-wide association studies. Nat. Commun. 11, 39 (2020).
Beasley, T. M., Erickson, S. & Allison, D. B. Rank-based inverse normal transformations are increasingly used, but are they merited? Behav. Genet. 39, 580–595 (2009).
Chang, C. C. et al. Second-generation PLINK: Rising to the challenge of larger and richer datasets. Gigascience 4, 7 (2015).
Watanabe, K., Taskesen, E., van Bochoven, A. & Posthuma, D. Functional mapping and annotation of genetic associations with FUMA. Nat. Commun. 8, 1826 (2017).
Watanabe, K., Umićević Mirkov, M., de Leeuw, C. A., van den Heuvel, M. P. & Posthuma, D. Genetic mapping of cell type specificity for complex traits. Nat. Commun. 10, 3222 (2019).
Wang, K., Li, M. & Hakonarson, H. ANNOVAR: functional annotation of genetic variants from high-throughput sequencing data. Nucleic Acids Res. 38, e164 (2010).
Kircher, M. et al. A general framework for estimating the relative pathogenicity of human genetic variants. Nat. Genet. 46, 310–315 (2014).
de Leeuw, C. A., Mooij, J. M., Heskes, T. & Posthuma, D. MAGMA: Generalized gene-set analysis of GWAS data. PLoS Comput Biol. 11, e1004219 (2015).
Human genomics. The Genotype-Tissue Expression (GTEx) pilot analysis: Multitissue gene regulation in humans. Science 348, 648–660 (2015).
Battle, A., Brown, C. D., Engelhardt, B. E. & Montgomery, S. B. Genetic effects on gene expression across human tissues. Nature 550, 204–213 (2017).
Sanna, S. et al. Causal relationships among the gut microbiome, short-chain fatty acids and metabolic diseases. Nat. Genet. 51, 600–605 (2019).
Burgess, S., Butterworth, A. & Thompson, S. G. Mendelian randomization analysis with multiple genetic variants using summarized data. Genet. Epidemiol. 37, 658–665 (2013).
Bowden, J., Davey Smith, G. & Burgess, S. Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression. Int. J. Epidemiol. 44, 512–525 (2015).
Bowden, J., Davey Smith, G., Haycock, P. C. & Burgess, S. Consistent estimation in Mendelian randomization with some invalid instruments using a weighted median estimator. Genet. Epidemiol. 40, 304–314 (2016).
Hemani, G. et al. The MR-Base platform supports systematic causal inference across the human phenome. Elife https://doi.org/10.7554/eLife.34408 (2018).
Hemani, G., Tilling, K. & Davey Smith, G. Orienting the causal relationship between imprecisely measured traits using GWAS summary data. PLoS Genet. 13, e1007081 (2017).
Verbanck, M., Chen, C. Y., Neale, B. & Do, R. Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases. Nat. Genet. 50, 693–698 (2018).
Zhao, Q., Wang, J., Hemani, G., Bowden, J. & Small, D. S. Statistical inference in two-sample summary-data Mendelian randomization using robust adjusted profile score. The Annals of Statistics 48, 1742–1769, 1728 (2020).
Brion, M. J., Shakhbazov, K. & Visscher, P. M. Calculating statistical power in Mendelian randomization studies. Int. J. Epidemiol. 42, 1497–1501 (2013).
Li, X. et al. MR-PheWAS: Exploring the causal effect of SUA level on multiple disease outcomes by using genetic instruments in UK Biobank. Ann. Rheum. Dis. 77, 1039–1047 (2018).
Acknowledgements
Our study was funded by the National Key R&D Program of China (Grant No. 2020YFE0201600), the National Natural Science Foundation of China (Grant No. 82073504), and the Guangxi Natural Science Fund for Innovation Research Team (Grant No. 2017GXNSFGA198003). We thank Margaret Biswas, PhD, from Liwen Bianji (Edanz) (www.liwenbianji.cn/) for editing the English text of a draft of this manuscript.
Author information
Authors and Affiliations
Contributions
Conceptualization, data curation, and funding acquisition: X.Y. and Z.M. Original draft and formal analysis: W.Y. and L.L. Methodology: L.H. and X.F. Software: H.C. and X.G. Visualization: Y.B. and F.W. Review and editing: C.L. Validation: X.C.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Peer review
Peer review information
Communications Biology thanks the anonymous reviewers for their contribution to the peer review of this work. Primary Handling Editors: Eirini Marouli and Caitlin Karniski.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Yang, W., Li, L., Feng, X. et al. Genome-wide association and Mendelian randomization study of blood copper levels and 213 deep phenotypes in humans. Commun Biol 5, 405 (2022). https://doi.org/10.1038/s42003-022-03351-7
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s42003-022-03351-7
This article is cited by
-
A genome-wide association study provides insights into the genetic etiology of 57 essential and non-essential trace elements in humans
Communications Biology (2024)
-
Sex-Specific Associations of Five Serum Essential Metal Elements with Thyroid Nodules in Euthyroid Adults: a Cross‑sectional Study
Biological Trace Element Research (2023)
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.
