Heritability of serum sodium concentration: evidence for sex- and ethnic-specific effects

Abstract

Serum sodium concentration is the clinical index of systemic water balance. Although disordered water balance is common and morbid, little is known about genetic effects on serum sodium concentration at the population level. Prior studies addressed only participants of European descent and either failed to demonstrate significant heritability or showed only modest effect. We investigated heritability of serum sodium concentration in large cohorts reflecting a range of races/ethnicities, including the Framingham Heart Study (FHS, non-Hispanic Caucasian), the Heredity and Phenotype Intervention Heart Study (HAPI, Amish Caucasian), the Jackson Heart Study (JHS, African American), the Strong Heart Family Study (SHFS, American Indian), and the Genetics of Kidney Disease in Zuni Indians Study (GKDZI, American Indian). Serum sodium was transformed for the osmotic effect of glucose, and participants with markedly elevated glucose or reduced estimated glomerular filtration rate (eGFR) were excluded. Using a standard variance components method, incorporating covariates of age, glucose, and eGFR, we found heritability to be high in African American and American Indian populations and much more modest in non-Hispanic Caucasian populations. Estimates among females increased after stratification on sex and were suggestive among female participants in FHS (0.18 ± 0.12, P = 0.057) and male participants in JHS (0.24 ± 0.16, P = 0.067) and statistically significant among female participants in JHS (0.44 ± 0.09, P = 1 × 10⁻⁷), SHFS (0.59 ± 0.05, P = 9.4 × 10⁻⁴⁶), and GKDZI (0.46 ± 0.15, P = 1.7 × 10⁻⁴), and male participants in HAPI (0.18 ± 0.12, P = 0.03) and SHFS (0.67 ± 0.07, P = 5.4 × 10⁻²⁶). Exclusion of diuretic users increased heritability among females and was significant in all cohorts where data were available. In aggregate, these data strongly support the heritability of systemic water balance and underscore sex and ethnicity-specific effects.

disorders of systemic water balance are reflected in abnormalities of the serum sodium concentration. Hyponatremia is prevalent among the acutely ill (62), among patients undergoing surgery (1), and among the elderly (2, 20). Hyponatremia complicates a substantial percentage of cases of drug treatment with thiazide diuretics and other agents (13, 35, 56) and is associated with increased mortality in various disease states (e.g., Refs. 6, 30). In some instances, this may be attributable to increased severity of the underlying disease; however, well-controlled studies have found an independent association between hyponatremia and mortality (e.g., Refs. 17, 18, 60). Even subtle abnormalities in water balance cause reversible defects in coordination, balance, and cognition (50, 53). Furthermore, hyponatremia can be extremely challenging to treat (13, 26, 64). Therefore, maintenance of water balance is clinically important.

Although hyponatremia is the most common of the electrolyte abnormalities (61), only limited population-level investigations into a genetic predisposition have been undertaken. Twin- and family-based studies have been performed only in individuals of European descent (5, 37, 39, 44, 48, 66); most such studies were small and in aggregate have yielded equivocal results. Rare point mutations in water-regulatory genes such as those coding for aquaporin-2 and the vasopressin type-2 receptor (AVPR2) are causal for Mendelian forms of nephrogenic diabetes insipidus and the nephrogenic syndrome of inappropriate antidiuresis (12, 15, 38, 47, 51, 63). In contrast to these rare genetic mutations, we observed that a common genetic variant influenced serum sodium concentration at the population level. Specifically, a nonsynonymous polymorphism in the gene coding for the putative osmosensing transient receptor potential channel, subfamily V, member 4 (TRPV4) protein affected channel function and was associated with serum sodium concentration and human hyponatremia (59). We speculate that a range of common genetic variants influences serum sodium concentration and therefore hypothesized that serum sodium concentration would exhibit significant heritability. Unlike prior investigations, we used analytical methods permitting assessment of genetic contributions from all family members; in addition, we tested very large family-based cohorts reflecting a range of races/ethnicities.

Heritability is the fraction of phenotypic variance that can be explained by genetic (or genotypic) variance; the remainder of the phenotypic variance is attributed to environmental variance. “Broad-sense” heritability (abbreviated H²) is the ratio of total genotypic variance to phenotypic variance, where total genotypic variance includes dominant genetic effects, recessive effects, epistatic effects (e.g., gene-gene interactions), and additive genetic effects. Broad-sense heritability is relevant to Mendelian disease states; however, it is less informative for continuously variable phenotypes characteristic of quantitative traits. Such traits (height, for example) exhibit a normal or near-normal distribution within a population and are inferred to represent the sum of the additive effects of variation across multiple genes. This additive genetic variance, when expressed as a fraction of total variance in the phenotype itself, constitutes “narrow-sense” heritability (abbreviated h²) (49). For a quantitative trait, narrow-sense heritability is most reflective of parent-offspring similarity, a biological and intuitive expression of heritability. Heritability of serum sodium concentration, as a quantitative trait, is best measured as h². We sought to assess h² of serum sodium concentration across diverse ethnic/racial groups in support of a genetic influence upon serum sodium concentration at the population level.

MATERIAL AND METHODS

Cohorts Investigated

Framingham Heart Study.

The Framingham Heart Study (FHS), begun in 1948, is a population-based investigation designed to identify common factors or characteristics that contribute to cardiovascular disease (http://www.framinghamheartstudy.org/about/history.html). Enrollment included 5,209 men and women age 28–62 yr; participants underwent repeated examination biennially (9, 10). Starting in 1971, 5,124 men and women who were children, or spouses of children, of the original cohort were enrolled to create the FHS Offspring cohort (14, 23). Participants in the Offspring cohort underwent examination approximately every 4 yr (14, 23). Participants are Caucasian of European descent, drawn largely from the town of Framingham, Massachusetts, located 20 miles west of Boston (e.g., Ref. 32). Complete data were available for 1,218 participants (Table 1); 1,198 participants remained after appropriate exclusions (Table 2).

Table 1.

Demographic information for the five cohorts

Cohort	n	Age, yr	Sex	Race/Ethnicity	Sodium, mEq/l	Glucose, mg/dl	Creatinine, mg/dl	Diuretic, %
FHS	1,218	41.0 ± 10.2	614 F	non-Hispanic Caucasian	139.1 ± 2.8	97.4 ± 17.7	1.2 ± 0.3	7.55
			604 M
HAPI	868	43.8 ± 14.0	408 F	non-Hispanic Caucasian	139.3 ± 1.7	93.1 ± 20.5	0.84 ± 0.15	0
			460 M
JHS	1,362	49.5 ± 14.2	903 F	African American	140.5 ± 2.1	91.7 ± 12.8	0.94 ± 0.20	24.45
			459 M
SHFS	3,617	39.9 ± 17.0	2,171 F	American Indian	141.0 ± 2.8	114.1 ± 52.5	0.89 ± 0.67	7.24
			1,446 M
GKDZI	999	37.1 ± 13.6	480 F	American Indian	139.8 ± 2.8	115.6 ± 56.6	0.90 ± 0.81
			519 M

Data from all participants in the cohort from whom serum sodium concentration and other demographic data were available; in subsequent tables, a subset of participants was excluded based upon glucose and estimated glomerular filtration rate (Tables 2 and 3; see methods) and upon diuretic use (Table 3). Cohorts listed include the Offspring Cohort of the Framingham Heart Study (FHS), the Heredity and Phenotype Intervention Heart Study (HAPI), the family component of the Jackson Heart Study (JHS), the Strong Heart Family Study (SHFS), and the Genetics of Kidney Disease in Zuni Indians (GKDZI). The ethnicity of the Framingham Offspring cohort was inferred to be non-Hispanic Caucasian. M, male; F, female.

Table 2.

Heritability (h²) of serum sodium concentration, reported by cohort and by sex (all, F, or M), with and without the inverse normalization procedure to ensure normal distribution of input data

	Heritability of Serum Sodium Concentration (Sodiumcorr), with Exclusions and Covariates
	h2, All	h2, F	h2, M
FHS
Un-norm.	0.07 ± 0.06	0.18 ± 0.12	0.13 ± 0.12
	P = 0.131	P = 0.057	P = 0.133
	n = 1198	n = 608	n = 590
iNorm	0.053 ± 0.06	0.16 ± 0.12	0.11 ± 0.11
	P = 0.18	P = 0.082	P = 0.16
	n = 1198	n = 608	n = 590
HAPI
Un-norm	0.11 ± 0.07	0.24 ± 0.16	0.18 ± 0.12
	P = 0.07	P = 0.09	P = 0.03
	n = 844	n = 398	n = 446
iNorm	0.10 ± 0.06	0.23 ± 0.15	0.19 ± 0.12
	P = 0.08	P = 0.09	P = 0.07
	n = 844	n = 398	n = 446
JHS
Un-norm	0.36 ± 0.06	0.44 ± 0.09	0.24 ± 0.16
	P = 9.19 × 10−11	P = 1.0 × 10−7	P = 0.067
	n = 1331	n = 875	n = 456
iNorm	0.35 ± 0.06	0.42 ± 0.09	0.23 ± 0.16
	P = 1.83 × 10−10	P = 1 × 10−7	P = 0.078
	n = 1331	n = 875	n = 456
SHFS
Un-norm	0.49 ± 0.03	0.59 ± 0.05	0.67 ± 0.07
	P = 5.2 × 10−88	P = 9.4 × 10−46	P = 5.4 × 10−26
	n = 3083	n = 1824	n = 1259
iNorm	0.47 ± 0.03	0.57 ± 0.05	0.67 ± 0.07
	P = 3.5 × 10−81	P = 4.1 × 10−45	P = 3.2 × 10−26
	n = 3083	n = 1824	n = 1259
GKDZI
Un-norm	0.38 ± 0.07	0.46 ± 0.15	0.10 ± 0.15
	P = 1.5 × 10−6	P = 1.7 × 10−4	P = 0.24
	n = 683	n = 333	n = 350
iNorm	0.35 ± 0.10	0.43 ± 0.15	0.13 ± 0.15
	P = 5.8 × 10−6	P = 4.1 × 10−4	P = 0.18
	n = 683	n = 333	n = 350

Sodium_corr is serum sodium concentration corrected (i.e., transformed) for the osmotic effects of glucose using simultaneously obtained serum glucose concentration, and calculated according to the formula of Katz (see methods). Exclusion criteria were: missing data; glucose ≥150 mg/dl or estimated glomerular filtration rate (eGFR) [estimated via the 4-component modification of diet in renal disease (MDRD) formula] >2 SD below the population mean when stratified by sex. Covariates are described in methods and included: age; glucose; and eGFR. For SHFS, Recruitment Center was also included as a covariate. iNorm indicates heritability data calculated after application of the inverse normalization procedure to ensure normal distribution of the input corrected serum sodium concentration (see methods).

Heredity and Phenotype Intervention Heart Study.

The Heredity and Phenotype Intervention (HAPI) Heart Study, conducted in 868 healthy Amish adults from large families (40), was designed to assess the cardiovascular response to four short-term interventions impacting cardiovascular risk factors to better gauge the genetic and environmental factors influencing these responses. There were 184 sibships (ranging from two to 10 members), 633 sib pairs, and 327 parent-offspring pairs (40). The genetic underpinnings of complex phenotypes may be more readily apparent in a genetically homogeneous population; of the 25,000 Old Order Amish in Lancaster County, Pennsylvania, nearly all trace their ancestry to ∼200 founding individuals (reviewed in Ref. 40). Complete data were available for 868 participants (Table 1); 844 participants remained after appropriate exclusions (Table 2).

Jackson Heart Study.

The Jackson Heart Study (JHS) is a long-term observational study designed to identify factors, both genetic and environmental, that contribute to cardiovascular and related disorders in African American participants in Mississippi (57). The JHS Family Study includes 1,499 members of 291 families, from whom phenotypic data were obtained (67). There were 2,656 parent-offspring pairs and 1,196 sib pairs in the Family Study (67). Complete data were available for 1,362 participants (Table 1); 1,331 participants remained after appropriate exclusions (Table 2).

Strong Heart Family Study.

The Strong Heart Study (SHS) was started in 1988 to investigate determinants of cardiovascular disease in a geographically diverse group of resident American Indian tribal members recruited from three geographic regions (Arizona, Oklahoma, and North and South Dakota) and 13 distinct tribes (28). The Strong Heart Family Study (SHFS), a component of the SHS, was started in 1996 to aid in the identification of genes impacting cardiovascular risk (46). Data used for the present study were collected during Phase IV (2001–2003) of the SHS. More than 3,600 participants aged 14–93 yr were recruited from over 90 multigenerational extended families from 13 American Indian tribes (42). Complete data were available for 3,617 participants (Table 1); 3,083 participants remained after appropriate exclusions (Table 2).

Genetics of Kidney Disease in Zuni Indians Study.

The Genetics of Kidney Disease in Zuni Indians (GKDZI) Study was designed to identify genes and environmental factors impacting susceptibility to renal disease in Zuni Indians, a small endogamous tribe located in rural New Mexico (36). Members of extended families were recruited following identification of a proband with kidney disease and at least one sibling with kidney disease. Complete data were available for 999 participants (Table 1); 683 participants remained after appropriate exclusions and provided consent for detailed heritability analysis (Table 2).

These studies were approved by the Institutional Review Board (IRB) of the Portland V. A. Medical Center and/or by the IRBs of the institutions of the investigators for each study, or were deemed exempt under Code of Federal Regulations, Title 45 - Public Welfare, Department of Health and Human Services; Part 46 - Protection of Human Participants; Paragraph 46.101(b)(4) - i.e., Exemption 4.

Statistical Methods

Heritability.

Heritability was estimated using the variance components method as implemented in the Sequential Oligogenic Linkage Analysis Routines (SOLAR) statistical package (3). This method is the standard approach for assessment of heritability in complex pedigrees (65). According to the classical quantitative genetics theory, the total phenotypic variance can be partitioned into its genetic and environmental components. Therefore the phenotypic variance (σ_P²) is represented by the equation

$${\sigma }_{\text{P}}^2={\sigma }_{\text{G}}^2+{\sigma }_{\text{E}}^2$$

where σ_G² is the variance due to additive genetic factors and σ_E² is the variance resulting from individual-specific environmental effects. The heritability of a phenotypes refers to the ratio of variance contributed by additive genetic factors to the total phenotypic variance (σ_G²/σ_P²). P values for the heritability estimates were obtained by likelihood ratio tests by comparing a model in which additive heritability was estimated with one in which the parameter was fixed at zero. In these studies, multigenerational families with extended complex pedigrees with all possible familial combinations were utilized. The complexity of these pedigrees is illustrated by the number and variety of pairwise relationships (Supplementary Table S1).¹ SOLAR utilizes data for all relative pairs, as feasible, irrespective of genealogical distance (v.i.). For the HAPI Heart Study cohort, these details were previously reported (40). Specifically, among the 868 participants, there were 633 sibling pairs, 327 parent-offspring pairs, 443 avuncular (aunt/uncle-niece/nephew) pairs, and 191 first cousin pairs. It was further reported that 425 of these subjects could be combined into a single pedigree when nonparticipating relatives were included (40). Therefore, although these 425 participants are not all “related” within the present study, they are related within the population as a whole (i.e., through one or more common ancestors not enrolled in the study and from whom phenotypic information is unavailable). This is consistent with the genetic architecture of this founder population.

Calculation of heritability of sodium concentration.

For estimation of heritability of serum sodium concentration, an effort was made to minimize the effect of nongenetic factors potentially impacting measured serum sodium concentration and/or systemic water balance. Low glomerular filtration rate (GFR) leads to an inability to appropriately excrete water, irrespective of appropriate central osmosensing. Elevated serum glucose concentration obligates intravascular water influx, isotonically suppressing the serum sodium concentration. For each cohort, participants with estimated GFR (eGFR) >2 SD units below the population mean [stratified for sex, and as calculated via the four-component modification of diet in renal disease (MDRD) formula (31)] were excluded, as were participants with simultaneous serum glucose concentration ≥150 mg/dl. For all participants, serum sodium concentration was transformed for serum glucose concentration in accordance with the formula of Katz, i.e., sodium (in meq/l) + (1.6*{[glucose (in mg/dl) − 100]/100}), to account for the intravascular water movement (i.e., isotonic suppression of serum sodium concentration) obligated by excess glucose (24, 54). Stratification on sex was performed because of the sex-specific association noted between serum sodium concentration and a polymorphism in the gene coding for the putative osmosensing protein, TRPV4 (59). Covariates for the analysis included age, glucose concentration, and eGFR. For SHFS, the additional covariate of recruitment center was included to account for potential population heterogeneity. Although distribution of transformed serum sodium concentration was grossly Gaussian by inspection in all cohorts, it failed strict tests of normality. Because this can impact estimates of heritability, a separate analysis was performed for each cohort using the inverse normalization procedure to ensure a normal distribution of the glucose-corrected serum sodium concentration. To achieve this, observations were ranked and replaced by expected value for the rank from a standard normal distribution, and this was used to calculate heritability. The effect of this normalization intervention was modest in all cohorts (see Table 2).

For direct comparison with earlier observations of others, heritability was also calculated using the raw (as-reported) serum sodium concentration on all participants, and then on participants stratified on sex (not shown). These data are briefly discussed in results.

Influence of diuretic use on heritability estimates.

An additional analysis was conducted wherein participants taking diuretic-type medications at the time of the determination of serum sodium concentration were identified. Although other medications may impact serum sodium concentration, diuretics, especially thiazide-type diuretics, are among the most common culprits in drug-associated hyponatremia. Because their usage was substantial across these cohorts addressing primarily cardiovascular outcomes and because these data were available for most cohorts (in contrast to other medication data), we restricted this additional analysis to diuretic use. For one element of the analysis, diuretic use was considered as an additional covariate; for another, diuretic use was an exclusion criterion (Table 3). Additional covariates potentially affecting serum sodium concentration (e.g., presence of congestive heart failure; diagnosis of acute or chronic liver disease; or the use of anticonvulsant, antipsychotic, or antidepressant medications) were either unavailable or were not uniformly available or uniformly coded across cohorts and could not be analyzed.

Table 3.

Heritability (h²) of glucose-transformed serum sodium concentration by cohort and sex, when diuretic use is incorporated as either an additional covariate or as an additional exclusion criterion

	Diuretic Use Incorporated as a Covariate			Diuretic Use Incorporated as an Exclusion
Cohort	h2, All	h2, F	h2, M	h2, All	h2, F	h2, M
FHS	0.07 ± 0.06	0.14 ± 0.12	0.13 ± 0.12	0.07 ± 0.07	0.29 ± 0.14	0.19 ± 0.12
	P = 0.13	P = 0.11	P = 0.13	P = 0.15	P = 0.017	P = 0.054
	n = 1198	n = 608	n = 490	n = 1106	n = 557	n = 549
HAPI	N/A	N/A	N/A	N/A	N/A	N/A
JHS	0.36 ± 0.06	0.47 ± 0.07	0.24 ± 0.16	0.29 ± 0.08	0.49 ± 0.12	0.01 ± 0.22
	P = 4.4 × 10−9	P = 1.0 × 10−7	P = 0.067	P = 8.5 × 10−6	P = 2.8 × 10−6	P = 0.48
	n = 1331	n = 875	n = 456	n = 998	n = 671	n = 327
SHFS	0.48 ± 0.03	0.59 ± 0.05	0.67 ± 0.07	0.51 ± 0.03	0.66 ± 0.05	0.65 ± 0.07
	P = 2.5 × 10−80	P = 1.3 × 10−45	P = 6.4 × 10−26	P = 3.3 × 10−86	P = 4.3 × 10−50	P = 1.1 × 10−23
	n = 3083	n = 1824	n = 1259	n = 2915	n = 1710	n = 1205
GKDZI3	N/A	N/A	N/A	N/A	N/A	N/A

Serum sodium concentration corrected for serum glucose concentration (sodium_corr) in accordance with formula of Katz (see Table 2). Exclusion criteria included missing data, glucose ≥150 mg/dl, or eGFR (estimated via the 4-component MDRD formula) >2 SD below the population mean when stratified by sex (as in Table 2). Covariates are as in Table 2. The use of diuretics was incorporated as an additional covariate (left side of table) or as an additional exclusion criterion (right side of table), and effect upon h² was determined. No HAPI participants were using diuretics. Data were not available from GKDZI participants.

In all cohorts, serum sodium concentration had been measured once per participant. This was not felt to be an obstacle because serum sodium concentration is among the most tightly regulated of physiological variables. Although the phenotypic variable may fluctuate subtly over time, the large cohort size permits dissection of the genetic contribution as in other biochemical variables (e.g., Refs. 25, 39, 58). Substantial intraindividual variability of the parameter reduces sensitivity but, provided sample size is sufficient, needn't preclude detection of significance (e.g., Refs. 25, 39).

RESULTS

Crude Heritability of Serum Sodium Concentration in Multiple Cohorts

Unique to this investigation, we sought cohorts of diverse races/ethnicities. Two of the investigated cohorts are non-Hispanic Caucasian (Offspring Cohort of the FHS and the HAPI), one is African American (JHS), and two are American Indian (SHFS and the GKDZ). One non-Hispanic Caucasian population is a founder population (HAPI, Amish). Demographic data for each of the five cohorts are depicted in Table 1. Mean serum sodium concentration (mean ± SD) ranged from 139.1 ± 2.8 (FHS) to 141.0 ± 2.8 (SHFS). Although not a goal of this study, it's important to point out that direct comparison of serum sodium concentration between the cohorts was not undertaken because of our inability to control for the effects of nongenetic but cohort-specific factors (e.g., different assaying laboratories, geographic environments, etc.).

Heritability (i.e., h²) of serum sodium concentration was determined in each cohort. For complex multigenerational pedigrees, additive genetic effects are best estimated using a linear mixed model (65); we used the SOLAR statistical package to achieve this objective. An effort was made to control for factors potentially confounding the serum sodium concentration and rendering it less reflective of systemic water balance. Because chronic kidney disease can impair free water excretion (e.g., Ref. 1), participants with advanced chronic kidney disease were excluded. In our preliminary analysis, an effect upon serum sodium concentration was not evident at eGFR >30 ml/min [as calculated via the four-component MDRD formula (31); data not shown]. Because mean eGFR differs among male and female participants, we were unable to impose an eGFR threshold for each cohort. We instead excluded all participants with calculated eGFR >2 SD units below the population mean when stratified by sex; this threshold effectively excluded all participants with eGFR <30 ml/min from all cohorts, and stratification precluded introduction of a bias by sex.

Glucose is osmotically active in the plasma, and excessive glucose obligates intravascular movement of water (i.e., “translocational hyponatremia”; Ref. 1); the net effect of hyperglycemia is the isotonic suppression of measured serum sodium concentration (54). All participants with glucose >150 mg/dl (at the time that serum sodium concentration was determined) were excluded from analysis. In the remaining participants, serum sodium concentration was transformed to account for the independent osmotic effect of glucose concentration (see methods; Ref. 24). Heritability was then calculated on this variable (Sodium_corr, Table 2), using the covariates of glucose, eGFR, and age; we had earlier detected an independent association between age and serum sodium concentration (59).

Heritability under this model was modest for the two Caucasian cohorts and although not reaching statistical significance, was suggestive in the HAPI cohort (h² = 0.11 ± 0.07, P = 0.07). In contrast, heritability was high among the African American (0.36 ± 0.06, P = 9.2 × 10⁻¹¹) and American Indian cohorts (0.49 ± 0.03, P = 5.2 × 10⁻⁸⁸ for SHFS; 0.38 ± 0.07, P = 1.5 × 10⁻⁶ for GKDZI). We sought to account for the effect of sex, based upon the sex-specific genetic association we observed with this phenotype in another context (59). Stratification on sex specifically addresses female-to-female heritability and male-to-male heritability. (Given the generally complex nature of the pedigrees studied, the former would include not only mother-daughter dyads, but also aunt-niece, grandmother-granddaughter, etc.) Importantly, heritability among females increased substantially across all cohorts following stratification on sex (Table 2). Heritability was suggestive in females in FHS (0.18 ± 0.12; P = 0.057) and HAPI (0.24 ± 0.16; P = 0.09), and significant in all African American and American Indian female cohorts, achieving a maximum of 0.59 ± 0.05 (P = 9.4 × 10⁻⁴⁶) in SHFS. Heritability among men achieved statistical significance in HAPI (0.18 ± 0.12; P = 0.03), but not in the JHS and GKDZI cohorts. In SHFS, male heritability was the highest of all groups at 0.67 ± 0.07 (P = 5.4 × 10⁻²⁶). Of note, we were not able to perform a global analysis of heritability across all cohorts because of cohort-specific restrictions on the distribution of unaggregated data.

The distribution of serum sodium concentration (and of corrected serum sodium concentration) was grossly Gaussian however, it failed strict tests of normality. Because a nonnormal distribution of the phenotypic variable across the population can impact the estimate of heritability, a separate analysis was performed for each cohort using the inverse normalization procedure to ensure a normal distribution of the transformed serum sodium concentrations (see methods). The effect of this intervention upon heritability estimates was negligible in each cohort (Table 2). The only substantive change affected the significance level for heritability (but not the heritability level itself) for males in the HAPI cohort; it went from significant (P = 0.03) to suggestive (P = 0.07). We inferred that the modest deviation from strict normality in the phenotype distribution did not substantively affect our assessments of heritability.

To assess the impact of our model refinements, an additional analysis was performed using only the raw serum sodium concentration (uncorrected for glucose), without excluding participants with markedly elevated serum glucose concentrations or reduced eGFR. Data were similar (not shown) with several notable exceptions. With this simplistic model, heritability was present among females in the FHS cohort (0.22 ± 0.12, P = 0.021), whereas it was only suggestive in the more physiologically based model (see above) and only suggestive (rather than significant) in males in HAPI (0.19 ± 0.11, P = 0.06). There was a major impact in the GKDZI cohort. Although total and female heritability were little affected, heritability among males when only the raw serum sodium concentration was considered was significant (0.27 ± 0.12, P = 0.005), whereas it had not even been suggestive in the more refined model (0.10 ± 0.15, P = 0.24). Together, these data suggested that an approach informed by a physiological understanding of biological variables potentially impacting serum sodium concentration independent of a direct genetic effect influenced the analytical outcome. That is, the genetics of serum sodium concentration cannot be accurately interpreted without considering other meaningful variables.

Effect of Diuretic Usage

Medication usage may also affect systemic water balance, and diuretic-type drugs are among the most frequently implicated (33). We performed a separate analysis such that diuretic usage was incorporated into the model as either an additional covariate or as an additional exclusion criterion (Table 3). Depending upon cohort, diuretic usage varied from 0 to 24.6% (Table 1). Incorporation of diuretic use as a covariate modestly reduced heritability estimates among FHS females and among all SHFS participants, SHFS females, and SHFS males (Table 3, left); however, the effect of excluding diuretic users was substantial in some cohorts. When diuretic users were excluded from the FHS cohort, heritability among female participants was increased to 0.29 ± 0.14, achieving statistical significance (P = 0.017; Table 3, right), and was increased (and newly suggestive of significance, P = 0.054) in male participants. Excluding the substantial number of diuretic users from the JHS cohort increased the strong heritability in female participants; however, heritability was reduced across all JHS participants and markedly diminished (from 0.24 to 0.01, and no longer suggestive) among JHS males. Exclusion of diuretic users from the SHFS further increased the already-high heritability estimates in all and female participants. Of note, no participants in the HAPI cohort were using diuretics. Because women may be predisposed to the hyponatremic effect of thiazide diuretics (e.g., Ref. 55), a separate analysis was performed to assess for an interaction between sex and diuretic use as an additional covariate. In SHFS, a significant interaction (P = 0.027) was observed using the raw (untransformed) serum sodium concentration; however, this effect was not significant when the more refined Katz-corrected serum sodium concentration was used (as in Table 2). We speculate that the known effect of thiazide diuretics upon serum glucose level confounded the interaction analysis when the untransformed serum sodium concentration was used. Similarly, in JHS, with its higher prevalence of diuretic use and greater male-female disparity in heritability, there was again no significant sex-diuretic interaction detected (data not shown).

DISCUSSION

Using data from five large family-based cohort studies reflecting diverse ethnicities, we establish the heritability of serum sodium concentration. Heritability was strongest among female and male American Indian participants in the SHFS and male American Indian participants in GKDZI (i.e., up to 67% in American Indian males; Table 2); it was also high among female African American participants in the JHS. Consistent with earlier reports, a much more modest effect was observed in the non-Hispanic Caucasian populations. In all cohorts, the heritability estimates among females increased after stratification on sex. In the JHS, FHS, and GKDZI cohorts, heritability among female participants substantially exceeded that among males. This relationship may also have been present in the HAPI cohort; however, statistical significance was not achieved. For the most part, heritability estimates increased after excluding or accounting for potential confounding factors blunting the degree to which serum sodium concentration reflected systemic water balance (e.g., low GFR, elevated serum glucose). In addition, exclusion of only diuretic users generally enhanced heritability estimates.

Previously, heritability of serum sodium concentration had only been investigated in individuals of European descent (i.e., the population in which we observed the weakest heritability). Two twin studies [one Australian (66) and one Swedish (44)] suggested mainly environmental influences on serum sodium concentration; however, there were only 200 or so pairwise comparisons in each of these small investigations. A larger Danish study of elderly twins (age 73–95 yr) detected correlation in serum sodium concentration between monozygotic twins exceeding that of dizygotic twins (i.e., consistent with heritability); however, fully 44% of participants had a serum sodium concentration falling below the population reference range (5). Consistent with our findings in this one ethnicity, family-based studies examining an array of quantitative traits in participants of European descent have shown insignificant (39, 48) or modest (37) heritability of serum sodium concentration.

The recent study of Meyer et al. (39) warrants additional comment. This group reported that heritability of serum sodium concentration in the FHS was not significant, estimated at 0.04 ± 0.06 (mean ± SE, P = 0.27) (39). Notably, this analysis did not stratify on sex; we, too, failed to detect heritability in this population prior to stratification on sex (Table 2). In addition, the Meyer et al. analysis excluded participants “on hypertension treatment” (39). The present analysis, in contrast, excluded only those participants treated with diuretic-type medications, the antihypertensive class most commonly implicated in disordered water balance. In aggregate, these data suggest that serum sodium concentration is modestly heritable across populations of European descent; however, an understanding of the physiology of water balance is needed to exclude or account for confounding influences.

Far more striking is our demonstration of high heritability of serum sodium concentration in other ethnicities, particularly female African Americans and both male and female American Indians. Ethnic differences in the heritability of a trait may be a consequence of nongenetic factors, including disparities in exposure to environmental factors. Apart from the use of certain medications (which we attempted to control for), the only major known environmental factor impacting serum sodium concentration is access to water. Given the ubiquity and low cost of water as a beverage, it seems unlikely that access to water contributed to the differences we observed. In addition, the two Caucasian populations, although exposed to vastly different environmental influences, yielded similar heritability estimates. Therefore, it seems reasonable to postulate a genetic basis for these differences. We are not aware of literature describing ethnic-specific differences in systemic water balance; however, minor allele frequency of a genetic polymorphism impacting systemic water balance clearly varied by ethnicity and was more prevalent among African Americans (59). The effect of this allele vis-à-vis water balance was more pronounced in men (59), suggesting that additional sources of genetic variation likely confer the greater heritability of serum sodium concentration among women in the present investigation.

In general, exclusion of participants taking diuretic-type medications strengthened the estimates for heritability of systemic water balance. Use of thiazide and, to a lesser extent, loop diuretics causes a predisposition to aberrant water balance (33); hence, serum sodium concentration values obtained from diuretic users may be less reflective of genetic influences upon systemic water balance. It is, however, conceivable that there is a genetic basis for susceptibility to thiazide-inducible hyponatremia (for example) and that exclusion of such participants could weaken the heritability estimate. Exclusion of diuretic users weakened the heritability estimate in only male African American participants, from 0.24 ± 0.16 (P = 0.07) to 0.01 ± 0.22 (P = 0.48); however, heritability was only suggestive in the former instance. In addition, some participants potentially received diuretic medications for treatment of edematous states associated with decreased effective circulating volume (e.g., cirrhosis, nephrotic syndrome, or congestive heart failure), which independently predispose to hyponatremia (1).

The higher heritability for serum sodium concentration in female participants observed in most of the present cohorts, and suggested in the twin study of Bathum et al. (5), is of interest. The AVPR2 gene is on the X-chromosome, and mutations in AVPR2 causes hereditary dysnatremias (15, 38, 47, 51, 63); disease severity in heterozygous females is modified by skewed X-inactivation (4, 11, 45). Therefore, common genetic variants that alter AVPR2 expression, or expression of other X-linked genes, could influence water balance differentially in males and females due to the hemizygous state in males and nonrandom X-inactivation in females. Alternatively, sex steroids may impact the regulation of systemic water balance. The putative central osmosensing channels TRPV1 and TRPV4 likely influence water balance in animal models (34, 41, 55a) and in humans (59); function and/or expression of these channels (22, 68, 69), and of related TRP channels (7, 8, 19, 21, 27, 29), is regulated by sex steroids. It is conceivable that an interaction between TRPV4, or a related osmoregulating TRP channel, and the female-specific sex steroid profile may impact systemic water balance.

Because heritability reflects the ratio of genotypic variation to phenotypic variation and not the absolute amount of the former, estimates of heritability may be inflated by increased genetic variation or decreased environmental variation in the population under study (49). Conversely, heritability may be underestimated when there is reduced genetic diversity (i.e., in founder populations arising shortly after a genetic “bottleneck” and in endogamous populations marrying exclusively within the group or ethnicity) or in the setting of increased environmental diversity (49). Among populations in the present report with potentially reduced genetic diversity, the Amish founder population and the endogamous Zuni, it is expected that there is both decreased genotypic variation and decreased environmental variation; the latter is a consequence of geographical restriction of these populations and, potentially, similar social situations among individuals within each cohort. Therefore, it seems unlikely that the present estimates of heritability in these populations with reduced genetic diversity dramatically over- or underestimate the true heritability.

Multiple studies have shown an association between hyponatremia and poor clinical outcome. In congestive heart failure, the presence of hyponatremia is an independent predictor of mortality (reviewed in Ref. 16). Although more profound hyponatremia may be a manifestation of more severe underlying disease, pharmacological correction of hyponatremia in patients with congestive heart failure may be associated with a reduction in mortality (52). Studies such as these have generally focused upon the hypervolemic hyponatremia accompanying one of the principal edematous states (1); the present analysis, in contrast, emphasizes the heritability of serum sodium concentration in comparatively healthy participants.

Together, these data strongly support the heritability of serum sodium concentration on a population-wide basis across diverse ethnicities. Identification of genetic polymorphisms influencing, or even merely associated with, water balance might aid in the prediction or earlier recognition of spontaneous and/or iatrogenic hyponatremia.

GRANTS

This work was supported by grants from the National Institutes of Health, the Department of Veterans Affairs, and the American Heart Association (to D. M. Cohen). A portion of this work was supported by the National Heart, Lung, and Blood Institute's (NHLBI) FHS (contract no. N01-HC-25195). The FHS is conducted and supported by the NHLBI in collaboration with Boston University. The HAPI Heart Study is supported by NHLBI Grant HL-72515-01. The SHS is supported by the following NHLBI grants: U01 HL-041654, U01 HL-041642, U01 HL-041652, U01 HL-065520, and U01 HL-065521. GKDZI receives funding from National Institute of Diabetes and Digestive and Kidney Diseases (DK-066660-03; DK-57300-05), University of New Mexico Clinical Research Center (5M01RR-00997), National Institute of Environmental Health Sciences (P30 ES-012072), and Dialysis Clinic. The JHS is supported by the NHLBI and the National Center on Minority Health and Health Disparities.

DISCLAIMERS

The opinions expressed in this paper are those of the authors and do not necessarily reflect the views of the Indian Health Service. This manuscript was not prepared in collaboration with investigators of the Framingham Heart Study and does not necessarily reflect the opinions or views of the Framingham Heart Study, Boston University, or NHLBI.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

Author contributions: B.W., V.S.V., Y.-P.C.C., Y.F., Z.C., H.A.T., J.G.W., T.G., V.O.S., J.G.U., M.F.F., R.H., A.R.S., A.G.C., S.M., P.G.Z., J.W.M., S.A.C., and D.M.C. analyzed data; B.W., V.S.V., Y.-P.C.C., Y.F., Z.C., H.A.T., J.G.W., T.G., V.O.S., J.G.U., M.F.F., R.H., A.R.S., A.G.C., S.M., P.G.Z., J.W.M., S.A.C., and D.M.C. interpreted results of experiments; B.W., V.S.V., and D.M.C. prepared figures; B.W., V.S.V., Y.-P.C.C., Y.F., A.R.S., J.W.M., S.A.C., and D.M.C. edited and revised manuscript; B.W., V.S.V., Y.-P.C.C., Y.F., Z.C., H.A.T., J.G.W., T.G., V.O.S., J.G.U., M.F.F., R.H., A.R.S., A.G.C., S.M., P.G.Z., J.W.M., S.A.C., and D.M.C. approved final version of manuscript; D.M.C. conception and design of research; D.M.C. drafted manuscript.