Hypertension (HTN) represents a complex heritable disease in which environmental factors may directly affect gene function via epigenetic mechanisms. The aim of this study was to test the hypothesis that dietary salt influences the activity of a histone-modifying enzyme, lysine-specific demethylase 1 (LSD-1), which in turn is associated with salt-sensitivity of blood pressure (BP).
Animal and human studies were performed. Salt-sensitivity of LSD-1 expression was assessed in wild-type (WT) and LSD-1 heterozygote knockout (LSD-1+/−) mice. Clinical relevance was tested by multivariate associations between single-nuclear polymorphisms (SNPs) in the LSD-1 gene and salt-sensitivity of BP, with control of dietary sodium, in a primary African-American hypertensive cohort and two replication hypertensive cohorts (Caucasian and Mexican-American).
LSD-1 expression was modified by dietary salt in WT mice with lower levels associated with liberal salt intake. LSD-1+/− mice expressed lower LSD-1 protein levels than WT mice in kidney tissue. Similar to LSD-1+/− mice, African-American minor allele carriers of two LSD-1 SNPs displayed greater change in systolic BP (SBP) in response to change from low to liberal salt diet (rs671357, P = 0.01; rs587168, P = 0.005). This association was replicated in the Hispanic (rs587168, P = 0.04) but not the Caucasian cohort. Exploratory analyses demonstrated decreased serum aldosterone concentrations in African-American minor allele carriers similar to findings in the LSD-1+/− mice, decreased α-EnaC expression in LSD-1+/− mice, and impaired renovascular responsiveness to salt loading in minor allele carriers.
The results of this translational research study support a role for LSD-1 in the pathogenesis of salt-sensitive HTN.
Essential hypertension (HTN) is a complex heritable disease. The ability to translate gene-based knowledge into improved treatment strategies has been disappointing. In addition to inherent heterogeneity, lack of success may be due to a failure to recognize important effects of environmental factors on gene expression (i.e., epigenetics). Dietary salt is one such factor, yet little is known of its effect on gene expression in blood pressure (BP) regulation.1,–3 This is of particular importance in African-American hypertensives who have a high prevalence of salt-sensitivity and cardiovascular risk.
In an animal model, we recently identified that lysine-specific demethylase 1 (LSD-1) is involved in salt-sensitive HTN.4 LSD-1 acts as a regulator of gene transcription through alteration in histone methylation.5,,,,–10 The heterozygote knockout of LSD-1 (LSD-1+/−) demonstrates (i) decreased LSD-1 mRNA and protein levels in heart and aorta, (ii) salt-sensitivity of BP, (iii) lower plasma and urine aldosterone levels, (iv) and altered aortic vasoactivity.4 This report expands on these results in four ways. First, are LSD-1 levels reduced in the kidney similar to the heart/vasculature? Second, in the setting of reduced aldosterone, what are the renal epithelial sodium channel (ENaC) levels given its role in salt handling? Third, does sodium intake change LSD-1 levels? Finally, to translate the clinical relevance of these findings, we evaluate the relationship of polymorphic variants in the LSD-1 gene to salt-sensitivity in human HTN, focusing initially on African Americans in light of their inherent salt-sensitivity.3,11,,–14
Animals: The interaction of salt intake and LSD-1 was evaluated in wild-type (WT) and LSD-1+/− mice treated with a liberal sodium (HS) (1.6% Na+) or low sodium (LS) (0.02% Na+) diet for 1 week. LSD-1+/− mice were generated using a gene trap approach as previously described.4 The LSD-1 homozygote knockout is embryonically lethal. As previously described, the LSD-1+/− mice compared to WT, have similar systolic BP (SBP) on LS, and significantly higher SBP on HS intake (P < 0.001). LSD-1+/− mice had similar total body and kidney weights but significantly heavier hearts (difference = 50 mg; P = 0.02).4
Animals were housed in an animal facility with 12-h light/dark cycle, at an ambient temperature of 22 ± 1 °C. All studies were performed between 7 and 9 AM. All experimental procedures were approved by the Institutional Animal Care and Use Committee at Harvard Medical School.
Tissue preparation: Mice were euthanized under deep anesthesia with isofluorane, the abdominal cavity opened, and renal cortical tissue excised and placed in liquid nitrogen in preparation for protein analysis.
Western blot analysis: Protein was extracted by homogenizing tissue with RIPA lysis buffer (Santa Cruz Biotechnology, Santa Cruz, CA) as previously described.4 Protein extracts (40µg) were combined with equal volume 2× Laemmli-loading buffer, boiled for 5 min, and size-fractionated by electrophoresis on 7.5% sodium dodecyl sulfate-polyacrylamide gels. Proteins were transferred from gel to a nitrocellulose membrane by electroblotting. Membranes were incubated with 5% nonfat dried milk in TBS-Tween (USB, Cleveland, OH) for 1 h and incubated overnight at 4 °C with primary antibodies. After incubation, samples were washed, incubated with peroxidase-conjugated secondary antibody, and analyzed using enhanced chemiluminescence (Perkin-Elmer Life Sciences, Boston, MA). Blots were subsequently reprobed for β-actin (1:5,000 dilution) and results normalized to correct for loading. Immunoreactive bands were analyzed quantitatively by optical densitometry, and densitometry values represented pixel intensity. Primary antibodies used were from Abcam: rabbit anti-LSD-1 (catalog number ab17721; dilution 1:1,000), from Novus Biologicals (Littleton, CO): rabbit anti-ENaC (dilution 1:200); and from Sigma-Aldrich (St Louis, MO): mouse anti-β-actin (catalog number A5441; dilution 1:5,000).
Western blots included a negative control lane to verify antibodies did not bind nonspecifically. Moreover, a marked decrease (>90%) in LSD-1 levels was observed by western blot after LSD-1 had been knocked down with small interfering RNA (data not shown). No positive controls were included. LSD-1 antibody did not bind to the mineralocorticoid receptor, angiotensin-II receptor, angiotensin-converting enzyme.
Human studies. To augment the association with salt-sensitivity, we chose an African-American hypertensive population from HyperPATH (HyperPATH-AA; n = 63; Table 1) as our primary cohort. Replication cohorts included 422 Caucasian hypertensives from HyperPATH (HyperPATH-C) and 178 Mexican-American hypertensives from HTN-IR. Protocols for subject recruitment and data collection for HyperPATH and HTN-IR populations have been previously described.15,16 The analyses were restricted to subjects with LSD-1 genotype data. The institutional review board of each site approved the study. All subjects gave written informed consent before enrollment. Data presented have not been previously reported except as an abstract.17
HyperPATH subjects and protocol: The HyperPATH consortium represents an intensive phenotyping program to determine the genetic underpinnings of HTN. Its strengths include an enhanced signal-to-noise ratio through rigorous control of factors that influence BP including medication washout, body positioning, diurnal variation, and dietary salt. Subjects were studied in Clinical Research Centers (CRC) of five sites (Brigham and Women's Hospital, Boston, USA; University of Utah Medical Center, Salt Lake City, USA; Vanderbilt University, Nashville, USA; Hospital Broussais, Paris, France and University La Sapienza, Rome, Italy). The HyperPATH inclusion/exclusion criteria and the detailed phenotyping protocol have been previously described.15,18 As per the original study design, HTN was defined as seated diastolic BP ≥100 mm Hg off antihypertensive medication, ≥90 mm Hg on at least one antihypertensive medication, or treatment with up to three antihypertensive medications. Use of more than three medications was exclusionary.
All subjects completed a screening visit consisting of a physical examination, medical history, and laboratory assessment. A history of diabetes mellitus, secondary HTN, renal insufficiency (serum creatinine >1.5 mg/dl), other significant medical illnesses, current tobacco or illicit drug use, or alcohol intake >12 oz/week was exclusionary. Normal laboratory values for electrolytes, thyroid and liver function tests were required. Subjects were between 18 and 65 years of age. Race was self-defined. To minimize interference of medication with renin–angiotensin–aldosterone system (RAAS) activity, all angiotensin-converting enzyme inhibitors, angiotensin receptor blockers, and mineralocorticoid receptor antagonists were discontinued for 3 months. If needed, subjects were placed on amlodipine or hydrochlorothiazide to control BP. At 3 weeks before hormonal and vascular assessment all antihypertensive medications were discontinued.
All subjects completed two controlled dietary phases: 5–7 days of HS (200 mmol Na+/day) and 7 days of LS (10 mmol Na+/day) diet, prepared by each site's metabolic kitchen. Each diet was isocaloric, contained 100 mmol/day potassium, 20 mmol/day calcium, and caffeine and alcohol-free. On the final day of each diet, subjects were admitted to the CRC and maintained fasting and supine overnight. Measurement of sodium and creatinine excretion in a 24-h urine collection confirmed sodium balance (≥150 mmol for HS and <30 mmol for LS). Hemodynamic and laboratory assessments were obtained in supine posture between 8 AM and 10 AM. BP was measured at 5-min intervals using an automated device (DINAMAP; Critikon, Tampa, FL). Three consecutive readings were averaged for analysis. Salt-sensitivity of BP was recorded as a continuous variable representing the change in SBP in response to dietary salt (ΔSBP). A subset of African-American subjects (n = 22) also completed assessment of effective renal blood flow (RBF) by para-aminohippuric acid clearance method.19
HTN-IR subjects and protocol: The Mexican-American Hypertension cohort consists of Hispanic families (939 individuals from 160 pedigrees) ascertained via a hypertensive proband (HTN defined as sitting BP ≥140/90 mm Hg off medication).16 Probands were recruited through the HTN Clinic at Los Angeles County, University of Southern California Medical Center or the CRC at the University of California, Los Angeles, CA. Subjects were 18–65 years, with mild-moderate HTN and no target organ damage. Antihypertensive medications were discontinued for 2 weeks before study if permissible. Salt-sensitivity of BP protocols included the Weinberger method,20 which involved (i) infusion of 1 l of 0.9% saline over 4 h on day 1; and (ii) 10-mEq sodium diet and three doses of furosemide on day 2. BP was measured at the end of both procedures with the difference (salt loaded—salt depleted) representing measure of salt-sensitivity (SBPsen).
Phenotype(s) examined: The primary phenotype was salt-sensitivity defined as ΔSBP in HyperPATH and SBPsen in HTN-IR as described above. Aldosterone levels, plasma renin activity (PRA) and RBF were examined as secondary phenotypes in an exploratory analysis of underlying mechanisms.
Laboratory analyses: HyperPATH samples were analyzed at a central laboratory. Samples were collected on ice and centrifuged for 20 min. Serum and urine specimens were stored at –20 °C without preservatives until assayed. Sodium and potassium levels were measured by flame photometry. Aldosterone and renin levels were measured by radioimmunoassay techniques.21,22
Genotyping and SNP selection: DNA was extracted from peripheral leukocytes using standard procedures. In HyperPATH, genotyping was conducted using the Illumina Bead Station GoldenGate platform. Tagging single-nucleotide polymorphisms (SNPs) were identified from Haploview using the HapMap23 CEU and YRI populations with an R2 >0.9 and a minor allele frequency >10%. This provided seven tagging-SNPs (Table 2). All SNPs had a completion rate of >95%. Repeat genotyping for 10% of the SNPs demonstrated concordance with the original genotype call.
Linkage disequilibrium constructs in our HyperPATH-AA population revealed that two of seven tagging-SNPs were in linkage disequilibrium with other genotyped SNPs, and five of seven SNPs captured 100% of the common LSD-1 variation with an R2 >0.85. Therefore, five SNPs were analyzed in the HyperPATH-AA cohort. The SNP most significantly associated with salt-sensitivity in HyperPATH-AA (rs587168) was examined in HyperPATH-C and HTN-IR. In the HTN-IR cohort, genotyping was performed using a 7600-SNP Illumina iSelect platform.
Statistical analysis. All tests for association were performed using SAS 9.1 (SAS Institute, Cary, NC). Hardy–Weinberg Equilibrium was assessed using the χ2-test. The D' and R2 measures of linkage disequilibrium were estimated using Haploview. In the HyperPATH cohorts, a mixed effect linear regression model (PROC MIXED) was used to test associations between LSD-1 genotype and phenotypes, accounting for relatedness, and adjusted for age, gender, body mass index and study site. Logarithmic transformations were performed for serum aldosterone and PRA to meet normality assumptions. Wilcoxon signed ranks test was used for paired comparison of RBF on HS vs. LS diet within genotype. In the HTN-IR cohort, the PROC GENMOD procedure in SAS was used for analyzing genotype–phenotype association used the generalized estimating equation, accounting for familial correlations and adjusted for gender, age, and body mass index.
For the primary phenotype (ΔSBP), P = 0.01 was considered statistically significant in the HyperPATH-AA cohort to account for testing five SNPs. Given that the AA population could be highly admixed, we tested for population stratification. Using the control SNPs available in this cohort (N = 801), the genomic inflation factor was 1.007. Statistical significance was indicated by P < 0.05 for all secondary analyses and replication analyses in the HyperPATH-C and HTN-IR cohorts, as only one SNP was tested. For animal studies, data were analyzed by Student's t-test using Graphpad Prism, with P < 0.05 considered significant.
Relationship between LSD-1 and dietary sodium
LSD-1 protein levels were significantly higher on LS compared to HS diet in WT mouse kidney tissue (Figure 1). In contrast, LSD-1 protein levels did not respond to salt modulation in LSD-1+/− mice (Figure 1). A similar change in LSD-1 protein was seen in mouse heart tissue (data not shown).
Effect of dietary sodium on lysine-specific demethylase 1 (LSD-1) protein levels in mouse kidney. #LSD-1 protein levels were significantly higher on low salt (LS) compared to liberal salt (HS) diet in kidney tissue of wild-type (WT) mice (N = 3–4; P < 0.01). *In LSD-1 heterozygote knockout (LSD-1+/−) mice, LSD-1 protein levels in the kidney were significantly lower compared to WT mice on both diets (N = 3–4; P < 0.05). LSD-1 protein levels in the LSD-1+/− mice did not change significantly between diets. Error bars = SEM.
Baseline characteristics of human subjects
Characteristics of the primary HyperPATH-AA cohort are shown in Table 1. Baseline characteristics (age, body mass index, gender) did not differ greatly between study cohorts.
Relationship between salt-sensitivity of BP and LSD-1 genotype in human hypertension
All SNPs were in Hardy–Weinberg Equilibrium. In the HyperPATH-AA cohort, a significant association was observed for ΔSBP in two of five SNPs (rs671357, P = 0.01; rs587168, P < 0.005), whereby minor allele carriers displayed a greater ΔSBP (Figure 2). To replicate this finding, we tested SNP rs587168 in the HyperPATH-C and HTN-IR cohorts. In the HTN-IR cohort, minor allele homozygotes of rs587168 had significantly greater SBPsen (11.11 ± 2.58 (mean ± SEM) mm Hg in minor allele homozygotes vs. 5.49 ± 1.23 mm Hg in major allele carriers, P = 0.04). No association with ΔSBP was observed in Caucasians (14 ± 1 mm Hg in minor allele carriers vs. 15 ± 1 mm Hg in major allele homozygotes, P = 0.25).
Effect of lysine-specific demethylase 1 (LSD-1) genetic variants on the systolic blood pressure (SBP) response to dietary salt in African-American hypertensive subjects. #African-American hypertensive minor allele carriers of LSD-1 rs587168 (N = 26) compared with major allele homozygotes (N = 28) had significantly greater increase in SBP in response to change from low to liberal salt diet (P < 0.01). Data shown are mean values adjusted for age, gender, and body mass index (BMI) differences. Error bars = SEM. A significant association of the minor allele with salt-sensitivity of BP was also seen for rs671357 (data not shown; P = 0.01) in the African Americans.
Exploratory mechanistic studies in humans and mice
To identify potential mechanisms, we assessed physiologic pathways involved in salt homeostasis. We previously reported that LSD-1+/− have lower serum aldosterone and 24-h aldosterone excretion compared with WT mice 4. We now assessed whether a target of aldosterone linked to sodium reabsorption, renal α-ENaC, may be dysfunctional. Consistent with lower aldosterone, LSD-1+/− mice had lower α-ENaC expression compared to WT (Figure 3). Concordant with animal data, HyperPATH-AA minor allele carriers of rs587168 had lower serum aldosterone on HS diet compared to major allele homozygotes (Figure 4; P = 0.003). No significant differences were observed in serum aldosterone on LS or PRA on either diet between genotype, although there was a trend toward lower PRA in the minor allele (Table 1). We also assessed for correlation between SBP and PRA, angiotensin-II, or aldosterone by genotype in the HyperPATH-AA cohort. As might be anticipated given the small sample size, we found none except a marginal correlation (P = 0.05) between aldosterone and SBP on HS diet.
Aldosterone levels by lysine-specific demethylase 1 (LSD-1) genotype. #African-American hypertensive minor allele carriers of rs587168 (N = 26) had significantly lower serum aldosterone levels compared with major allele homozygotes (N = 28) on liberal salt diet (P < 0.01). Error bars = 95% confidence interval.
Renal epithelial sodium channel (ENaC) expression on liberal salt diet. #Lysine-specific demethylase 1 (LSD-1) heterozygote knockout (LSD-1+/−) compared with wild-type (WT) mice, had significantly lower expression of the α-subunit of ENaC (N = 6–8 mice in each group; P < 0.01) on liberal salt diet. Error bars = SEM.
We then explored the relationship between salt and RBF by LSD-1 genotype. Minor allele carriers for rs587168 (n = 12) had no change in RBF (1.99 ± 22.7 (mean ± SEM) cc/min/1.73 m2) in response to change from LS to HS diet whereas major allele homozygotes (n = 10) had an increase in RBF (18.5 ± 24.3 cc/min/1.73 m2) as would be expected in response to increased salt intake, although not statistically significant.
We report that dietary salt modifies LSD-1 expression and that genetic alteration in LSD-1 (likely “loss of function,” although a dominant negative effect cannot be ruled out) is associated with salt-sensitivity of BP in both an animal model and human HTN. This finding is important because it provides evidence for an environmental factor (dietary salt) affecting expression of a gene known to be an epigenetic regulator. This layer of environmental regulation would be critical to acknowledge when investigating candidate genes implicated in BP regulation, and could in part explain difficulties in attempting to reproduce gene association studies. It could also explain inconsistencies in the literature regarding the variable BP response to salt intake and its role in human HTN. Finally, these findings, together with our previous report,4 suggest that the LSD-1+/− mouse may serve as a model for salt-sensitivity in human HTN (Table 3); however, the mechanisms responsible for this effect are yet to be determined.
The contribution of dietary salt to the pathogenesis of HTN has been an area of longstanding interest. Salt-sensitivity of BP is commonly present in essential HTN, particularly in African Americans.3,11,,–14 The pathophysiologic mechanisms leading to salt-sensitivity are complex and only partially understood. Our results, from animal studies and two hypertensive cohorts of different ethnicities, support LSD-1's role as an epigenetic mediator of dietary sodium's effect on BP. Interestingly, we did not observe this in the Caucasian cohort, despite an identical study protocol. This is unlikely to be explained solely by a lower minor allele frequency in the Caucasians (18% vs. 28%) given the larger sample size, and may suggest a relatively intact dietary salt-LSD-1 relationship and/or compensatory mechanisms operating to dampen penetrance in Caucasians.
The underlying molecular pathways linking LSD-1 to salt-sensitive HTN remain to be elucidated but several possibilities exist. First, our exploratory studies suggest LSD-1's effect is not due to dysfunctional RAAS activity. Increased angiotensin-II/RAAS activation have been shown to play a role in several types of BP salt-sensitivity.1,15,19 However, data presented herein could not be explained by an activated RAAS, but rather volume expansion—induced thru an alternate mechanism—followed by “reactive” RAAS suppression. In the face of inappropriate salt retention, the RAAS should be quiescent. Indeed, LSD-1+/− mice displayed lower aldosterone levels than WT littermates.4 Similarly, LSD-1 minor allele carriers in the African-American population had significantly lower circulating aldosterone. PRA was also reduced, but not significant perhaps due to assay sensitivity and small sample size. Second, aldosterone regulates the α-subunit of ENaC24 to modify renal tubular sodium resorption. As would be anticipated with reduced aldosterone, LSD-1+/− mice displayed reduced α-ENaC expression. Whether this is due to decreased synthesis or increased catabolism is unknown. Third, our preliminary analyses suggest a possible abnormality in the renovascular response to dietary salt. Normally, when dietary sodium is increased, RBF increases to promote sodium excretion and maintain volume homeostasis. African-American minor allele carriers displayed no change in RBF, whereas African-American major allele homozygotes displayed an expected increase in RBF with HS intake. Thus, LSD-1 may influence renal vascular function by affecting the relationship between dietary salt, the RAAS and vascular function. Sodium excretion studies would be useful to confirm this possibility. Fourth, soluble guanylate cyclase, which seems to mediate in part the vascular phenotype in the LSD-1+/− mice is another potential candidate.4 Fifth, protein kinase C activity has been suggested by Metzger et al.25 to prevent LSD-1 from demethylating the H3K4 residue. Studies are needed to determine whether these systems interact with LSD-1 to induce volume expansion in vivo.
This study has several limitations. We assume, because of similar phenotypes between human minor allele carriers and LSD-1+/− mice, the minor allele is associated with decreased LSD-1 levels and/or function in humans, although this was not specifically documented. The sample sizes of our human cohorts are small by genome-wide association study standards; however this is, a candidate gene study defined by work in the mouse. In addition, our strict phenotyping protocol reduces signal variance from potent confounders of vascular and hormonal responses (medication, posture, dietary salt, diurnal variation) enabling greater power to detect small changes with smaller sample sizes. This is further supported by findings in separate populations and concordance in a specific animal model. It is important to emphasize that LSD-1 tagging-SNPs identified as associated with salt-sensitive HTN are not necessarily causal. Similarly, LSD-1 is an epigenetic regulator of gene transcription. We have not identified other pathways (e.g., natriuretic peptide system, adrenergic system) that may be transcriptionally modified by LSD-1 and also mediate LSD-1's role in the salt-sensitivity phenotype.
In summary, our findings suggest an interaction between dietary salt, LSD-1 and BP. Genetic variation in LSD-1 is associated with salt-sensitivity of BP. Further characterization of how LSD-1 affects genes involved in volume/pressure homeostasis is needed, but available data suggest that these genes are likely involved in vascular adaptation to changes in sodium intake. Thus, manipulation of LSD-1 activity may provide entrée to a novel paradigm for understanding the mechanisms underlying and/or treatment of salt-sensitive HTN.
We gratefully acknowledge the support of the dietary, nursing, administrative, and laboratory staff of the Clinical Research Centers in which these studies were performed. We also acknowledge the investigators, fellows, nurses and research coordinators at each of the study sites for the HyperPATH and HTN-IR study groups for their contribution to the study of these subjects. This research was supported by the following grants: The National Institutes of Health (NIH) grants HL47651, HL59424, DK63214, HL094452, HL67974, a Specialized Center of Research (SCOR) in Molecular Genetics of Hypertension (P50HL055000), P50-HL55005 (HTN-IR cohort), M01-RR000425 (Cedars-Sinai Medical Center), M01-RR000043 (University of Southern California), i2b2, a National Center for Biomedical Computing (NCBC) grant (LM008748) Diabetes Research Center P30DK063491 and Harvard Catalyst 5 UL1 RR025758. B.C. was supported by T32 HL007609 (NIH) and 5 KL2 RR025757 (Harvard Catalyst/NIH). J.S.W. was supported by NIH grant K23 HL084236. M.O.G. was supported by R01 DK079888 and the Winnick Clinical Scholars Award. L.H.P. was supported by KL2 RR025757 (NIH) and 0735609T (AHA Scientist Development Grant). J.I.R. was supported by the Cedars-Sinai Board of Governors' Chair in Medical Genetics.
The authors declared no conflict of interest.