Our sample included 2,853 Mexican individuals, of which 59.3% were women. On average, participants belonged to the sixth decade of life and showed borderline values indicative of hyperglycemia, hypertriglyceridemia or hypercholesterolemia. A large proportion of individuals had obesity, type 2 diabetes or high blood pressure (34.4, 38.6, 29.5%, respectively). Although a greater proportion of women were obese, men showed higher prevalence of hypertension, as well as a worse metabolic profile, as demonstrated by their higher levels of triglycerides, GGT, uric acid and creatinine (Table 1).
Table 1
| All | Women | Men | |
| Mean ± sd (%) | Mean ± sd (%) | Mean ± sd (%) | P |
N | 2853 | 1693 | 1160 | - |
Sex (% women) | 59.34 | - | - | - |
Age (years) | 58.02 ± 10.4 | 58.28 ± 10.33 | 57.63 ± 10.5 | 0.1066 |
Obesity (%) | 34.41 | 39.05 | 27.61 | < 0.001 |
BMI (kg/m2) | 28.66 ± 4.52 | 29.06 ± 4.74 | 28.07 ± 4.09 | < 0.001 |
Waist-hip ratio | 0.94 ± 0.08 | 0.91 ± 0.08 | 0.97 ± 0.06 | < 0.001 |
HTA (%) | 29.50 | 27.95 | 31.83 | 0.0437 |
SBP (mmHg) | 126.55 ± 18.99 | 126.2 ± 19.49 | 127.08 ± 18.21 | 0.2564 |
DBP (mmHg) | 78.77 ± 10.29 | 78.04 ± 10.41 | 79.87 ± 10.01 | < 0.001 |
BP medication (%) | 34.23 | 38.70 | 27.68 | < 0.001 |
Type 2 diabetes (%) | 38.63 | 38.33 | 39.05 | 0.7283 |
Fasting glucose (mmol/l) | 6.65 ± 3.52 | 6.58 ± 3.44 | 6.75 ± 3.64 | 0.2179 |
Fasting insulin (µIU/ml) | 77.81 ± 55.01 | 76.4 ± 47.68 | 79.74 ± 63.65 | 0.3159 |
Hba1c (%) | 7.49 ± 2.33 | 7.54 ± 2.33 | 7.42 ± 2.33 | 0.4487 |
Triglycerides (mmol/l) | 2.26 ± 1.98 | 2.07 ± 1.27 | 2.53 ± 2.68 | < 0.001 |
Total cholesterol (mmol/l) | 5.29 ± 1.14 | 5.37 ± 1.09 | 5.17 ± 1.22 | < 0.001 |
HDL cholesterol (mmol/l) | 1.12 ± 0.33 | 1.21 ± 0.33 | 1 ± 0.29 | < 0.001 |
GGT (IU/l) | 25.49 ± 37.97 | 23.08 ± 28.11 | 29.12 ± 49.06 | 0.0348 |
ALT (IU/l) | 28.27 ± 19.69 | 27.62 ± 18.74 | 29.24 ± 21.02 | 0.2381 |
HS-CRP (mg/dl) | 3.11 ± 5.04 | 3.43 ± 4.96 | 2.67 ± 5.14 | 0.1547 |
Uric acid (mg/dl) | 5.72 ± 5.79 | 5.04 ± 1.45 | 6.68 ± 8.74 | < 0.001 |
Creatinine (mg/dl) | 1.32 ± 3.3 | 1.11 ± 1.89 | 1.62 ± 4.61 | 0.0257 |
HTA is defined as SBP ≥ 140 mmHg and/or DBP ≥ 90 mmHg. Obesity is defined as BMI ≥ 30 kg/m2. Women vs. men comparisons. P value from t-student test or chi square test. |
We started the analyses with 3,912 SNPs within the STK39 genetic region. After quality control, our dataset included 744 SNPs. All of them were grouped into five LD blocks and the following SNPs: rs2044680, rs4668016, rs6749447, rs56088988 and rs10930316 (Supplementary Fig. 1) represented each one. As expected, a pairwise r2 < 0.2 was found between pairs of variants (Supplementary Table 1). As previously mentioned, this unbiased approach was chosen over the evaluation of candidate SNPs from other studies because of differences in patterns of linkage disequilibrium across populations.
After multiple comparison correction, two SNPs, representative of the second and third blocks, showed a statistically significant association with SBP (rs4668016: B(SE)=-3.74(1.22) mmHg, P = 0.0021 and rs6749447: B(SE) = 3.46(0.96) mmHg, P = 0.0003) (Fig. 1 and Supplementary Table 2). We considered a significant P value of 0.003 (0.05/15 comparisons [5 SNPs assessed in the whole sample, in women-only and in men-only]). The effect of both SNPs was mainly seen in women-only (rs4668016: Women B(SE)=-5.31(1.55) mmHg, P = 0.0006 vs. Men B(SE) = 1.045(0.565) mmHg, P = 0.0645 and rs6749447: Women B(SE) = 4.652(1.255) mmHg, P = 0.0002 vs. Men B(SE) = 1.681(1.480) mmHg), P = 0.2563). Although not reaching statistical significance, a fixed-effects meta-analysis showed that 65.4% and 57.3% of the variability in blood pressure effects estimates was due to sex (rs4668016: I2 = 65.4%, P = 0.089 and rs6749447: I2 = 57.3%, P = 0.126). ((Fig. 1 and Supplementary Table 2).
We then continued the analysis only with the SNPs that showed significant association with SBP. When assessing the effect of the two SBP-associated SNPs with BMI levels, we found a nominal association of rs6749447 in the whole sample without stratifying by sex (B(SE) = 0.54(0.22) kg/m2, P = 0.0127) (Fig. 1 and Supplementary Table 3). Risk allele G of rs6749447 variant showed a frequency of 0.423 in the sample (Supplementary Table 2).
As seen with SBP, the effect of rs6749447 over BMI was also stronger in women compared with men (Women: B(SE) = 0.81(0.3) kg/m2, P = 0.0067; Men: B(SE) = 0.19(0.31) kg/m2, P = 0.540), although the heterogeneity test did not reach statistical significance (I2 = 51.6%, P = 0.151). After Bonferroni correction, the effect in women was statistically significant. We considered a significant P value of 0.008 (0.05/6 comparisons [2 SNPs assessed in the whole sample, in women-only and in men-only]) (Fig. 1 and Supplementary Table 3).
The effect of rs6749447 on SBP was stronger in postmenopausal women (age ≥ 55 years) (B = 7.216 mmHg, P = 3.78x10− 05) than in the whole female sample, premenopausal (age < 50 years) or menopausal women aged between 50 and 55 (B = 4.351 mmHg, P = 5.55x10− 04; B=-1.74 mmHg, P = 0.487 and B = 3.046 mmHg, P = 0.1806, respectively) (Supplementary Table 4).
The prevalence of the risk homozygote individuals (GG genotype) was 18.3%. No difference was found between the prevalence of the risk homozygote women and men (Women: GG frequency = 18.28%, Men: GG frequency = 18.31%, chi square P value = 1) (Supplementary Table 5).
In order to assess between a direct or a BMI-mediated effect of rs6749447 variant on systolic blood pressure and given the observed sex-related differences in the effect estimates, we performed a path analysis in both men and women separately. We stratified these analyses based on our previous findings showing genotype effects mainly in women but not in men. After checking the model information in women (CFI = 1.0; TLI = 1.0 and RMSEA 90% CI (0,0.07), we found that besides having a direct effect on both SBP and BMI in women-only, rs6749447 showed an indirect BMI-dependent effect on SBP (B(SE) = 0.010 mmHg (0.004), P = 0.024). That is, 12% of the effect of rs6749447 on SBP is mediated by BMI in women. Interestingly, we found that rs6749447 also showed an indirect BMI-dependent effect on T2D (B(SE) = 0.007(0.004), P = 0.047). Regarding to the model considering men only, the model fit parameters were CFI = 0.99, TLI = 0.95 and RMSEA 90% CI (0,0.08); The rs6749447 variant did not show any direct effect on either SBP or in BMI (SBP: B(SE) = 0.030 mmHg (0.031), P = 0.329; BMI: B(SE) = 0.016 kg/m2 (0.033), P = 0.624) (Figs. 1 and 2).
To further highlight the pleiotropic effects of SPAK in humans, we extracted the association P values of rs6749447 and surrounding ± 50 kb variants from the Common Metabolic Diseases Knowledge Portal. We found that besides renal traits, this genetic locus shows an enrichment of associations with anthropometric as well as sleep-wake patterns, such as excessive daytime sleepiness, short sleep duration or frequent insomnia symptoms (P < 5x10− 04) (Supplementary table 6).
When examining the expression level of STK39 in subcutaneous adipose tissue, we found it was comparable with that of PPARG and LIPE genes, which are known to be highly expressed genes in adipose tissue. By examining the expression of the 7 genes surrounding 1Mb of STK39 locus, only NOSTRIN was found to be expressed at comparable levels. In contrast, CERS6, G6PC2 and DHRS9 showed low expression levels. Regarding the genes taking part in the STK39 signaling cascade, WNK1, OSR1 and SLC12A2 genes also showed expression in subcutaneous adipose tissue. Interestingly, STK39 gene expression was higher in women than in men (P = 0.0008). We did not observe the same sex dimorphic gene expression above in either WNK1, OSR1 or SLC12A12 genes (Fig. 3, Supplementary Tables 7 and 8).
Finally, we found an increase in STK39 gene expression in adipose tissue of hypertensive individuals (B = 0.279, P = 0.0488). No statistically significant association was found between the STK39 expression and overweight/obesity status, nor between non-risk and risk rs6749447 carriers, assuming a recessive model (Supplementary Table 8).