Genetically Determined Serum Calcium Levels and Markers of Ventricular Repolarization

Supplemental Digital Content is available in the text.


3
This article is published in its accepted form; it has not been copyedited and has not appeared in an issue of the journal. Preparation for inclusion in an issue of Circulation: Genomic and Precision Medicine involves copyediting, typesetting, proofreading, and author review, which may lead to differences between this accepted version of the manuscript and the final published version.

Introduction
Non-invasive markers of cardiac disease derived from the electrocardiogram (ECG) are associated with major cardiovascular events and reflect underlying abnormalities in cardiac structure and electrical conduction [1][2][3][4] . Abnormal action potential duration and amplification of the spatial dispersion of repolarisation, coupled with early after depolarisations inducing triggered activity is an important mechanism of ventricular arrhythmia, specifically torsades de pointes tachycardia 5,6 . Prolongation of the QT interval, a marker of the time needed for ventricular repolarisation and depolarisation, has consistently been associated with adverse outcomes, including ventricular arrhythmia and sudden cardiac death [7][8][9] . QRS duration (time point from QRS onset to offset) is specific for ventricular depolarisation while the JT interval is specific for ventricular repolarisation spanning the interval from QRS offset to T-wave end.
Multiple factors may influence these ECG markers and thus the potential for arrhythmia, This article is published in its accepted form; it has not been copyedited and has not appeared in an issue of the journal. Preparation for inclusion in an issue of Circulation: Genomic and Precision Medicine involves copyediting, typesetting, proofreading, and author review, which may lead to differences between this accepted version of the manuscript and the final published version.
including mutations in genes encoding ion channels and their accessory proteins (e.g KCNQ1 and KCNE1) and iatrogenic causes due to off target effects by medication (e.g., cancer therapeutics and psychotropics) [10][11][12] .
The different phases of the cardiac action potential are caused by the (inward and outward) movement of different ions across the membrane of the cardiac cells. Serum electrolyte concentrations are associated with alterations in ECG derived indices of cardiac electrophysiological activity. Historically, studies have focused on the effects of electrolytes in clinical populations often with serum electrolyte concentrations significantly outside of the normal range and/or rapid and acute changes in their concentration 13,14 . We recently published the results of a large meta-analysis of cross-sectional data including 153,014 unselected individuals, investigating the association of serum electrolyte levels with ECG-derived indices 15 .
One of the key findings was an association between lower serum calcium and longer QT (2.23 ms per 0.1 mmol/L) and JT (2.27 ms per 0.1 mmol/L) intervals but not with QRS duration. The lack of a calcium-QRS duration association suggested serum calcium specifically affects ventricular repolarisation. However, given the observational and cross-sectional nature of the study, and the limited number of considered confounders, we were unable to determine whether these observations were causal.
Mendelian randomization (MR), in which genetic variants significantly associated with an exposure are used to estimate causal effects of that exposure on outcomes of interest [16][17][18] , has been widely used to assess causality in observational settings. MR overcomes the main limitations of observational studies, notably reverse causation and residual confounding 19 .
Previous genome-wide association studies (GWAS) for serum calcium have identified associated variants, and have been leveraged before in MR studies for cardiovascular disease risk [20][21][22][23] .

5
This article is published in its accepted form; it has not been copyedited and has not appeared in an issue of the journal. Preparation for inclusion in an issue of Circulation: Genomic and Precision Medicine involves copyediting, typesetting, proofreading, and author review, which may lead to differences between this accepted version of the manuscript and the final published version.
However, due to the relatively small sample sizes of these GWAS which limited the number of associations identified, the genetic instruments included in MR analyses explained only a small proportion of the variance of calcium (~0.9%) 24,25 . The release of biochemical data in UK Biobank (UKB) permits the identification of additional genetic variants for serum calcium in larger samples increasing the number of variants and consequently increasing the power of an MR study 26,27 . In this study, we performed a new GWAS on serum total calcium and used the independent lead variants as instrumental variables to assess potential causality of the association between lower serum calcium and prolongation of QT and JT intervals in UKB, including QRS duration as a negative control.

Methods
Anonymized clinical, genotype and ECG data were obtained from UKB 27  and participating studies provided informed consent. Any data generated by this study will be returned to UKB in accordance with researcher obligations, to be made available for further research. Full methods are available in Supplementary Methods, Data Supplement and also summarized in Figure 1.

Calcium GWAS
We identified 208 independent lead variants (201 from novel loci) associated with serum total calcium concentration at genome-wide significance level (P<5x10 -8 ) (Supplementary Table 1, Data Supplement). A Manhattan plot and QQ plot are shown in Supplementary Figures 1 and 2 This article is published in its accepted form; it has not been copyedited and has not appeared in an issue of the journal. Preparation for inclusion in an issue of Circulation: Genomic and Precision Medicine involves copyediting, typesetting, proofreading, and author review, which may lead to differences between this accepted version of the manuscript and the final published version.
respectively in the Data Supplement. The percentage variance of total serum calcium explained by variants included in this MR study was 5.8% (compared with 0.9% for previously reported variants) 24 . Previously reported variants associated with serum calcium showed the same direction of effect and similar effect size estimates (Supplementary Table 2 total serum calcium and albumin-corrected calcium respectively, were included in MR analyses.

Primary analysis -IVW
Study characteristics for individuals included in each ECG cohort specific GWAS and subsequently combined in the meta-analysis, and the calcium GWAS are shown in Table 1 This article is published in its accepted form; it has not been copyedited and has not appeared in an issue of the journal. Preparation for inclusion in an issue of Circulation: Genomic and Precision Medicine involves copyediting, typesetting, proofreading, and author review, which may lead to differences between this accepted version of the manuscript and the final published version.

Sensitivity analyses
Genetically-determined lower serum calcium concentrations were consistently associated with longer QT and JT intervals across sensitivity analyses using weighted median estimator, MR-Egger and MR-PRESSO methods, with similar or stronger effect sizes as using the IVW model (

Discussion
This study utilises MR to demonstrate the causal inverse relationship between serum calcium concentration and longer QT and JT intervals in UKB, a large middle-aged European ancestry population. This inverse relationship was consistent across all sensitivity analyses. These results along with the absence of a clinically relevant association with QRS duration due to its very small effect size, collectively suggest that a genetically-predicted lower serum calcium is a Previous randomised control and crossover trials estimated an increase in serum total calcium of 0.07 -0.13 mmol/L approximately 4 hours after ingestion of calcium carbonate (500mg) 30,31 . Thus, the results of this study suggest oral calcium supplementation could temporarily decrease the QT interval by 2.11 -3.91 ms. As the effect of oral calcium supplementation on serum total calcium concentration is small, we would expect no direct 9 This article is published in its accepted form; it has not been copyedited and has not appeared in an issue of the journal. Preparation for inclusion in an issue of Circulation: Genomic and Precision Medicine involves copyediting, typesetting, proofreading, and author review, which may lead to differences between this accepted version of the manuscript and the final published version. clinical benefits. However, the results of this study suggest further research into the effects of serum calcium concentration on arrhythmogenesis is warranted and calcium variants could be considered for inclusion in genetic risk score models for risk prediction. This may be of particular importance in patient sub-groups such as endocrinology disorders affecting calcium homeostasis, concurrent use of medication which prolong the QT interval, and in the context of other co-morbidities where a substrate exists for ventricular arrhythmia such as ischemic heart disease, cardiomyopathies or channelopathies 12,32 .
Although an inverse relationship between calcium and markers of ventricular repolarisation were identified in this study, associations between higher serum calcium concentrations and increased cardiovascular disease risk including myocardial infarction, stroke and cardiovascular mortality have been reported in individual epidemiological studies, metaanalyses and some randomised control trials [33][34][35]  Significant associations have not been observed with heart failure (as an endpoint after myocardial infarction) or stroke risk 22,23 . Additionally, despite calcium supplementation being common in the general population with the intention to reduce the risk of fractures, an association between life-long calcium levels and risk of fracture was not observed in a previous MR study 38 . However, these studies may have been limited by the low variation of calcium explained by variants included in the MR analyses, despite having large sample sizes for testing these clinical outcomes 39 .
Despite showing evidence for a causal association between lower serum calcium and longer QT and JT intervals, this study does not provide information on the biological mechanisms involved, which remain uncertain. In animal models, the duration of phase II of the cardiac action potential is determined by the inactivation of voltage-gated long-lasting (L-type) calcium channels, which are dependent on calcium entering these channels and their release from the sarcoplasmic reticulum 28,40 . Higher extracellular calcium concentrations increase L-type calcium channel inactivation which in turn reduces phase II of the action potential and the inverse is present in lower calcium concentration states, as identified in a more recent in-silico theoretical study using a human ventricular myocyte model 41 . These mechanisms could explain the associations observed in our study between serum calcium and ventricular repolarisation.

Strengths and limitations
The present study performed a new serum calcium GWAS to increase the number of genetic instrumental variables and to increase the variance explained to perform a more statistically powerful MR analysis. Furthermore, two-sample MR studies assume the two samples (exposure This article is published in its accepted form; it has not been copyedited and has not appeared in an issue of the journal. Preparation for inclusion in an issue of Circulation: Genomic and Precision Medicine involves copyediting, typesetting, proofreading, and author review, which may lead to differences between this accepted version of the manuscript and the final published version. and outcome) were performed in different individuals from the same source population. By design, we performed the new calcium GWAS in individuals not contributing to the QT/JT/QRS intervals GWASs ensuring this assumption was met.
UKB is a densely phenotyped cohort and participants are generally healthy compared with the general UK population. Additionally, this study was conducted only in individuals of European ancestry due to a limited sample size available for other ancestries. Therefore, these results may not be extrapolated to population groups of non-European ancestry or within highrisk clinical cohorts such as post-myocardial infarction or channelopathies showing a mendelian pattern of inheritance.

Conclusion
In summary, this MR study indicates that genetically-determined lower serum calcium concentrations are causally associated with longer ventricular repolarization time in a middleaged population where serum calcium concentration exposure is likely stable and chronic. This article is published in its accepted form; it has not been copyedited and has not appeared in an issue of the journal. Preparation for inclusion in an issue of Circulation: Genomic and Precision Medicine involves copyediting, typesetting, proofreading, and author review, which may lead to differences between this accepted version of the manuscript and the final published version.   This article is published in its accepted form; it has not been copyedited and has not appeared in an issue of the journal. Preparation for inclusion in an issue of Circulation: Genomic and Precision Medicine involves copyediting, typesetting, proofreading, and author review, which may lead to differences between this accepted version of the manuscript and the final published version.