Introduction

Antihypertensive treatment with thiazide-type diuretics has conferred significant reductions in stroke and cardiovascular events by lowering blood pressure (BP) in hypertensive patients.1, 2, 3, 4, 5 However, electrolyte imbalance is a common complication following thiazide treatment.6, 7, 8 In recent years, it has become clear that electrolyte disorders, even if they are chronic and mild, are associated with adverse outcomes.8, 9, 10, 11 The presence of hypokalemia on admission has been associated with higher in-hospital mortality.8 A recent study in patients with myocardial infarction also found increased mortality among patients with hypokalemia, supporting other findings on the negative effects of dyskalemia on outcomes.12

Although thiazide diuretics are associated with an increased risk of hypokalemia, pharmacogenetic markers of thiazide-induced changes in serum potassium are not well studied. When considering dyskalemia, Gitelman’s syndrome and Gordon syndrome are two classic genetic disorders of electrolyte imbalance. Gitelman’s syndrome is an autosomal recessive salt-losing tubulopathy. It is caused by mutations in the SLC12A3 gene encoding the thiazide-sensitive Na-Cl cotransporter (NCC).13, 14, 15, 16 The syndrome is characterized by hypokalemic alkalosis, hypomagnesemia and hypocalciuria that are similar to the side effects of the chronic administration of thiazide diuretics. In contrast, Gordon syndrome (or pseudohypoaldosteronism type 2 (PHA2)) is an autosomal dominant disease. It is caused by mutations in NCC regulators with no-lysine kinase 1 (WNK1) and WNK4.17 The syndrome is characterized by an early onset of hypertension and hyperkalemia. The phenotype is the opposite of Gitelman’s syndrome and is effectively reversed by a low dose of thiazide diuretics. It would be interesting to determine whether the polymorphisms in the SLC12A3 and WNK1 genes are related to the electrolyte imbalance that develops following the administration of thiazide diuretics to hypertensive patients.

Given the clinical recommendation of thiazide diuretics, particularly for elderly hypertensives and people with resistant hypertension, the development of thiazide-induced electrolyte imbalances, such as hypokalemia, should be monitored carefully. However, clinical information regarding which patients are more susceptible to thiazide-induced hypokalemia is lacking and the impact of genetic factors is not known. This study aimed to investigate the potential clinical and genetic predictors of serum potassium decreases following treatment with thiazide diuretics. Because the causes of electrolyte imbalances such as hypokalemia may be complicated clinically, we could only evaluate the direct causal relationship of thiazide-induced hypokalemia through short-term thiazide treatment. Given the increasing interest in personalized medicine, our findings could provide novel rationales for optimizing the use of thiazide diuretics on an individualized basis.

Methods

Study population

A series of consecutive patients with essential hypertension were prospectively included in the study if all the following criteria were fulfilled: (1) age between 25 and 65 years; (2) sitting office systolic BP (SBP) of 140–180 mm Hg and/or sitting diastolic BP (DBP) of 90–110 mm Hg on three different occasions within a period of 3 months, or currently on ⩾1 antihypertensive medications without any diuretics; (3) fasting plasma sugar <126 mg dl−1; and (4) no evidence of secondary hypertension through serial studies including blood chemistry, renal function tests, endocrine examination, abdominal sonogram and/or renal arteriogram, among others, to exclude the possibility of chronic renal disease, renal arterial stenosis, primary aldosteronism, Cushing syndrome, pheochromocytoma, thyroid disorder and coarctation of the aorta. Patients with the following characteristics were excluded: (1) a history of diabetes mellitus; (2) a history of major systemic disease within the past 3 months; (3) body mass index >30 kg m−2; (4) renal dysfunction with a plasma creatinine level >1.7 mg dl−1; (5) liver dysfunction with liver enzymes >2 times the normal upper limit; (6) congestive heart failure with New York Heart Association function class II–IV; and (7) pregnant women.

All patients were first evaluated at the hypertension clinic of the hospital. A hypertension specialist took a comprehensive history and conducted a physical check-up. Patients underwent a series of tests including office BP, body mass index, waist and hip circumference and blood sampling. The patients were first observed while following their regular cardiovascular medications, if there were any, for 2 weeks. Then, the patients were interviewed again and monitored for their daily lifestyle and eating habits. If their rechecked BP was ⩾140/90 mm Hg, the patients were assigned to take 50 mg of hydrochlorothiazide (HCTZ) once every morning in addition to their usual medications, if they had any, for the next 2 weeks. Blood sampling and sitting office BP were repeated, followed by a comprehensive history and physical check-up. The ethics committee of the hospital approved the study protocol. All the patients agreed to participate after being informed of the nature and purpose of the study. This study was conducted in accordance with the principles of the Declaration of Helsinki and Title 45, US Code of Federal Regulations, Part 46, Protection of Human Subjects, Revised 13 November 2001, effective 31 December 2001.

BP measurement

At each visit, BP was measured on the nondominant upper arm by the same well-trained nurse using a validated automatic digital BP monitor (Omron HEM-780, Lake Forest, IL, USA). This was performed in the morning hours, after sitting for 10 min in a quiet room. Three consecutive BP measurements were performed each time. Each measurement was separated by a 30-s pulse measurement. The recorded BP was the average value of the last two recordings.

Laboratory measurements

Venous blood was collected by phlebotomy from fasting participants who rested supine for ∼10 min, typically between 0730 and 0900 h. Serum sodium and potassium were measured using standard methods at baseline and after 2 weeks of HCTZ treatment. Baseline plasma renin activity was determined using kits from DiaSorin (Stillwater, MN, USA; Gamma-Coat PRA 125I RIA Kit). Baseline plasma aldosterone was measured using a radioimmunoassay (Quest Diagnostics, Madison, NJ, USA).18 Baseline serum angiotensin-converting enzyme activity was measured using the Quantikine Human ACE Immunoassay (R&D Systems, Minneapolis, MN, USA).19 The people performing the above tests were blinded to the clinical data of the study subjects.

Selection of candidate genes

This study evaluated two candidate genes: SLC12A313, 14, 15, 16 and WNK1.17, 20 We searched for single-nucleotide polymorphisms (SNPs) in these genes in the NCBI SNP Database (http://www.ncbi.nlm.nih.gov/SNP/) and selected SNPs with a minor allele frequency of >0.05 as the genotyping markers. We investigated 7 SNP markers in the introns of the SLC12A3 gene (rs2304483, rs2278490, rs13306679, rs2289114, rs4567697, rs2399594 and rs711747) and 13 SNP markers in the introns of the WNK1 gene (rs11064524, rs2107614, rs765250, rs1159744, rs2286007, rs4980973, rs12581940, rs880054, rs956868, rs953361, rs2301880, rs10849582 and rs2286028). Figure 1 shows the linkage disequilibrium plots of SLC12A3 and WNK1. These SNPs had a P-value of Hardy–Weinberg equilibrium >0.01.

Figure 1
figure 1

Haploview-generated linkage disequilibrium (LD) plot of (a) SLC12A3 single-nucleotide polymorphisms (SNPs) and (b) WNK1 SNPs. A full color version of this figure is available at the Hypertension Research journal online.

Full size image

Genotyping

A total of 20 ml of blood was collected from each patient. We isolated the genomic DNA of the patients from peripheral lymphocytes using the phenol/chloroform extraction method. SNP genotyping was performed using high-throughput MALDI-TOF (matrix-assisted laser desorption and ionization-time of flight) mass spectrometry. Briefly, primers and probes were designed using SpectroDESIGNER software (Sequenom, San Diego, CA, USA). Multiplex PCRs were performed, and unincorporated ddNTPs were dephosphorylated using shrimp alkaline phosphatase (Hoffman-LaRoche, Basel, Switzerland), followed by primer extension. The purified primer extension reaction was spotted onto a 384-element silicon chip (Spectro-CHIP, Sequenom) and analyzed using an autoflex MALDI-TOF SpectroREADER mass spectrometer (Sequenom); the resulting spectra were processed with SpectroTYPER (Sequenom). The people performing the genetic study were blinded to the clinical data of the study subjects.

Statistical analysis

All data were expressed as the frequency (percentage) or mean±s.d. or median with interquartile ranges. Plasma renin activity and aldosterone levels were natural-log transformed because of their positively skewed distribution. We analyzed categorical variables using the χ2 test or Fisher’s exact test. Comparisons of genotype distribution between groups were also performed using Fisher’s exact test. Relations between two continuous variables were assessed by a bivariate correlation method (Pearson’s correlation). Subsequently, significantly correlated variables were further analyzed using stepwise linear regression for multivariate analysis. The P-value was two sided. A P-value of <0.05 was considered statistically significant. The corrected P (Pc) value was adjusted using Bonferroni’s correction. Statistical analysis was performed using SPSS software (Version 15.0, SPSS, Chicago, IL, USA).

Results

Demographic data on the patients

A total of 92 patients took 50 mg of HCTZ in our study. Of the patients, 2 did not have genotyping data available, and 15 did not have serum potassium levels tested after HCTZ treatment. The final sample included 75 patients who qualified for enrollment in our study. The baseline characteristics of the patients are presented in Table 1. The mean age was 45.3±9.5 years; 54 (72%) patients were male. The mean SBP was 143.7±17.6 mm Hg and the mean DBP was 96.0±9.4 mm Hg (Table 1). The concomitant medications included β-blockers (62.7%), α-blockers (6.7%), calcium channel blockers (30.7%), angiotensin-converting enzyme inhibitors (5.3%) and angiotensin receptor blockers (5.3%) (Table 1).

Table 1 Baseline characteristics of the patients and their response after 2 weeks of HCTZ
Full size table

Response after 2 weeks of treatment with thiazide diuretics

After a 2-week washout period, all the patients received HCTZ for 2 weeks. The patient responses to HCTZ are presented in Table 1. The mean SBP decreased from 141.4±13.8 to 128.1±15.1 mm Hg (net change, −13.3±14.0 mm Hg). The mean DBP dropped from 94.2±7.8 to 86.7±9.8 mm Hg (net change, −7.5±8.9 mm Hg). The mean potassium level declined from 4.3±0.3 to 3.9±0.4 mmol l−1 (net change, −0.3± 0.4 mmol l−1). The mean sodium level was similar before and after 2 weeks of treatment with HCTZ (from 142.7±1.9 to 142.8±1.8 mmol l−1, net change 0.0±2.0 mmol l−1).

Of all the demographic data considered, only baseline potassium was correlated with the decrease in serum potassium after 2 weeks of HCTZ treatment (r=−0.423, P<0.001). The concomitant medications were not correlated with the decrease in serum potassium.

Serum potassium decrease after treatment with thiazide diuretics and the SNPs of the two candidate genes

Of the 13 SNPs in WNK1, rs11064524 (P=0.040), rs4980973 (P=0.0007), rs12581940 (P=0.003), rs880054 (P=0.046), rs953361 (P=0.003) and rs10849582 (P=0.046) were correlated with the decrease in serum potassium (Table 2). After Bonferroni’s correction, only rs4980973 was correlated with the decrease in serum potassium (Pc=0.014). Of the seven SNPs in SLC12A3, none was correlated with the serum potassium decrease in our study (Table 2).

Table 2 Association of individual SLC12A3 and WNK1 gene variants and serum potassium level change after HCTZ
Full size table

WNK1 rs4980973 polymorphism

When comparing the serum potassium change associated with the WNK1 rs4980973 A/A, A/G and G/G polymorphisms, significant differences were observed (A/A vs. A/G vs. G/G=−0.45±0.42 vs. −0.46±0.37 vs. 0.00±0.39 mmol l−1, P=0.0001; Figure 2a). The office SBP change (A/A vs. A/G vs. G/G=−19.35±16.28 vs. −10.47±11.31 vs. −11.52±15.31 mm Hg, P=0.090) and office DBP change (A/A vs. A/G vs. G/G=−8.74±11.04 vs.−6.84±7.67 vs. −6.55±9.54 mm Hg, P=0.725) did not significantly differ among the three groups.

Figure 2
figure 2

Serum potassium (K) level change was significantly different when comparing (a) WNK1 rs4980973 A/A (n=18), A/G (n=35) and G/G polymorphisms (n=22) (A/A vs. A/G vs. G/G=−0.45±0.42 vs. −0.46±0.37 vs. 0.00±0.39 mmol l−1, P=0.0001), and (b) WNK1 rs4980973 (A/A and A/G) (n=53) and G/G polymorphisms (n=22) (A/A and A/G vs. G/G=−0.46).

Full size image

When comparing the serum potassium changes associated with the WNK1 rs4980973 (A/A and A/G) and G/G polymorphisms, significant differences were observed (A/A and A/G vs. G/G=−0.46±0.38 vs. −0.00±0.39 mmol l−1, P=0.00006; Figure 2b). The office SBP change (A/A and A/G vs. G/G=−9.07±9.00 vs. −8.19±10.50 mm Hg, P=0.740) and office DBP change (A/A and A/G vs. G/G=−7.83±9.36 vs. −6.42±10.87 mm Hg, P=0.605) were not significantly different between the two groups.

Predictors of the decrease in serum potassium after treatment with thiazide diuretics

Finally, we included the serum potassium level and WNK1 rs4980973 in a multivariate stepwise regression analysis. Using multivariate stepwise linear regression analysis, the baseline potassium level (β=−0.587, 95% confidence interval=−0.875 to −0.299, P=0.0001) and WNK1 rs4980973 (A/A and A/G vs. G/G, β=−0.418, 95% confidence interval=−0.598 to −0.237, P=0.00002) were independent predictors of the decrease in serum potassium following HCTZ treatment (Table 3).

Table 3 Clinical and genetic correlates of serum potassium level change: stepwise linear regression model R2=0.377
Full size table

In this model, the baseline potassium level explained 14.9% and the WNK1 rs4980973 polymorphism explained 22.8% of the total variance in the decrease in serum potassium levels. Overall, 37.7% of the change in serum potassium levels could be explained by our model (Table 3).

Discussion

This study was designed to assess potential determinants of a serum potassium decrease following the short-term use of thiazide diuretics. The findings indicated that baseline potassium levels and the WNK1 rs4980973 polymorphism could predict the decrease in serum potassium following the use of thiazide diuretics. To our knowledge, ours is the first study to show that the WNK1 polymorphism can predict the decrease in serum potassium following the use of thiazide diuretics in nondiabetic hypertensive patients.

The use of thiazide diuretics increases urinary potassium excretion and lowers serum potassium levels. Because most potassium in the body is intracellular, the serum potassium concentration is tightly regulated. Therefore, frank hypokalemia (a serum potassium <3.5 mmol l−1) occurs only in cases of severe potassium depletion.21, 22, 23 However, hypokalemia is common in patients using thiazide diuretics.8, 9, 10, 11 A cross-sectional analysis of 22 239 patients who presented to an emergency room in a university hospital in Switzerland8 found that the use of thiazide diuretics was associated with hypokalemia (odds ratio, 2.18, P<0.0001). In a population-based Rotterdam study, Liamis et al.6 reported that thiazide diuretics increased the risk of hypokalemia (odds ratio, 7.68; 95% confidence interval, 4.92–11.98).

Moreover, accumulating evidence indicates that thiazide-induced hypokalemia is associated with adverse outcomes. In particular, there is concern about hypokalemic arrhythmias from thiazide use.7, 24 Variation in the serum potassium level has also been implicated in the development and progression of coronary heart disease and myocardial infarction.24, 25 Arampatzis et al.8 reported that the presence of hypokalemia on admission was associated with higher in-hospital mortality (odds ratio, 1.89; P<0.0001).8 Furthermore, thiazide-induced hypokalemia is associated with increased blood glucose values and an increased risk of new-onset diabetes mellitus.26, 27

The WNK1 gene is located on chromosome 12 (12p13.3). It encodes a lysine-deficient protein kinase that regulates NCC in the luminal membrane of the distal convoluted tubule.17 Turner et al.20 investigated the change in BP in 585 adults with essential hypertension (30 to 59.9 years of age; 50% blacks; 47% women) after 4 weeks of HCTZ treatment (25 mg daily). After adjusting for ethnicity, sex, age and waist-to-hip ratio, three SNPs in WNK1 (rs2107614, rs2277869 and rs1159744) predicted the ambulatory BP response, accounting for 2 to 4% of the variation in the systolic and diastolic BP response (P<0.05). Polymorphisms in genes regulating renal sodium transport, in particular WNK1, predict interindividual differences in antihypertensive responses to HCTZ. In the current study, we found that the WNK1 rs4980973 polymorphism could predict the change in serum potassium following use of thiazide diuretics, accounting for 22.8% of the total variance in the decrease in serum potassium levels. Because rs4980973 is located in the intron of the WNK1 gene, the polymorphism of this SNP may be involved in RNA splicing and stability. However, further replication studies or functional studies are needed to clarify the role of WNK1 in the regulation of potassium levels in patients taking thiazide diuretics.

The SLC12A3 gene is located on chromosome 16 (16q13). It encodes the NCC in the luminal membrane of the distal convoluted tubule.13, 14 Mutations in the NCC have been reported to be the cause of Gitelman’s syndrome,13, 14, 15, 16 an autosomal recessive renal tubular disorder characterized by hypokalemia, metabolic alkalosis, hypomagnesemia and hypocalciuria. Matayoshi et al.28 reported that the SLC12A3 polymorphism could also predict the effects of thiazide diuretics. Comparison of the polymorphism prevalence between responders and nonresponders showed significant differences in SLC12A3 C1784T (C allele vs. T allele, odds ratio=3.81, confidence interval=1.25–11.63, P=0.016). However, the association between SLC12A3 and thiazide-induced hypokalemia is still unknown. In our study, we found that SLC12A3 polymorphisms were not related to the change in serum potassium level following the use of thiazide diuretics.

Study limitations

Although the findings are interesting, the current study has some limitations worth noting. First, HCTZ was given for only 2 weeks. However, we still found a change in serum potassium in these patients using thiazide diuretics in the short term. Second, the urine potassium level was not determined in this study. However, the daily intake was recommended to be maintained within a normal level in each patient before and during the study. Furthermore, we requested that the patients keep their daily lifestyle and eating habits the same before and during the 2 weeks of HCTZ treatment. Third, we analyzed the data of the 75 patients who completed the study. However, 15 patients had missing data. Fourth, some of our patients were on concomitant medications. However, we did not alter the concomitant medications before or after treatment with HCTZ. We found that the concomitant use of these medications (including angiotensin-converting enzyme inhibitors and angiotensin receptor blockers) was not correlated with the decrease in serum potassium after HCTZ treatment. Therefore, we did not exclude these cases from the study. Fifth, we enrolled patients aged 25 to 65 years (average 45.3±9.5 years). The results of our study may therefore not apply to the aging population. Finally, our sample size was small and is limited to nondiabetic hypertensive patients in a Han Chinese population in Taiwan. Further replication or functional studies are needed to confirm the current findings.

Conclusion

Baseline potassium levels and the WNK1 rs4980973 polymorphism independently predicted the decrease in serum potassium after short-term HCTZ treatment in nondiabetic hypertensive patients. Our findings provide additional information that may aid individualizing the use of thiazide diuretics and identifying hypertensive individuals at greater risk of a decrease in serum potassium.