Introduction

Cardiovascular disease (CVD) can be ascribed to genetic, environmental, and behavioural risk factors [1, 2]. Elevated systolic blood pressure (SBP) and dietary habits are the two top-ranked modifiable risk factors contributing to the high burden of CVD [1]. An estimated 7.94 million CVD-related deaths and 188 million disability-adjusted life years have been attributed to dietary risk factors [1], with a high sodium intake being a primary culprit [3].

While dietary salt intake is regarded as a behavioural risk factor, its role in CVD development also encompasses genetic susceptibility whereby BP is elevated in response to increased salt intake, which is known as salt sensitivity [4]. Salt sensitivity is frequently observed in Black ethnic groups [4, 5], and has been attributed to genetic factors such as divergent renal sodium handling [4]. In Black individuals excessive salt intake exacerbates this genetic predisposition and promotes the early onset of hypertension and CVD [6].

Excessive salt intake further increases the risk of hypertension and CVD irrespective of ethnicity through mechanisms such as increased sympathetic activity, hormonal responses, oxidative stress, and the gut microbiome [7]. Coinciding metabolic changes related to these mechanisms may result in the up- or downregulation of various metabolites. Sodium reduction has been associated with increased short-chain fatty acid [8], acylcarnitine [9], and organic and amino acid levels [9,10,11]. The metabolomic changes were subsequently associated with improved endothelial function [10] and reduced blood pressure [8, 11].

Ethnic differences in the interaction between diet and the metabolome may play an integral part in the susceptibility, incidence, progression, and severity of hypertension and CVD development. This study (1) aimed to determine the amino acid and acylcarnitine profiles of Black and White populations in relation to dietary salt intake, and (2) investigated the triad of salt intake, SBP, and the associated metabolomics profile.

Methods

The African Prospective study on the Early Detection and Identification of Cardiovascular Disease and Hypertension (African-PREDICT) aims to identify lifestyle behaviours and known and novel biomarkers that may be implicated in the development of hypertension. The African-PREDICT study recruited 1202 participants from communities in and around the Potchefstroom area in the North-West Province of South Africa [12]. This study included 380 Black and 447 White participants with urinary amino acid, acylcarnitine as well as complete 24-hour urinary data (volume ≥300mL and 24-hour urinary creatinine for men: ≤6 and ≥30 mmol/day; and for women: ≤4 and ≥25 mmol/day) (Fig. 1). Inclusion criteria were 20–30 years of age, screening clinic BP <140/90 mmHg, no self-reported chronic diseases or treatment thereof, HIV uninfected, not pregnant or lactating. Despite the BP screening criteria for African-PREDICT 18 participants had clinic SBP ≥140 mmHg with normal DBP, 46 participants had clinic DBP ≥90 mmHg with normal SBP, and 20 participants had clinic BP ≥140/90 when clinic BP was taken on the day of other study measurements. While some participants from the African-PREDICT study have white coat or masked hypertension based on BP reading we did not focus on HT status in this study.

Fig. 1
figure 1

Participant selection for this study. This study included data of 827 participants after excluding those with no metabolomic data or incomplete 24-hour urinary data

Full size image

All participants gave written informed consent. The African-PREDICT study complies with the Declaration of Helsinki, was approved by the Health Research Ethics Committee of the North-West University (NWU-00001-12-A1) and is registered on ClinicalTrials.gov (NCT03292094).

Questionnaire data

Sex and ethnicity data were obtained from a General Health Questionnaire. Data on dietary protein intake was obtained by means of three 24-hour dietary recall interviews using a standardized dietary collection kit (example pictures, packages, measurement tools and food models) and the five-step multiple-pass approach. The data was coded according to the South African Medical Research Council Food Composition Tables and the Food Quantities Manual was used to convert household measures to grams. Nutrient and food analysis of the dietary data was conducted by the South African Medical Research Council at the Biostatistics Unit.

Anthropometric, and physical activity data

Researchers were trained to measure weight (kg) (SECA electronic scales, SECA, Birmingham, UK), height (m) (SECA stadiometer, SECA, Birmingham, UK) and waist circumference (anthropometric nonflexible tape measure, Holtain, Ltd, Crymych, UK). All three variables were measured in triplicate, and the mean height and weight used to calculate body mass index (BMI) using the standard weight (kg)/height (m2) formula. Participants wore a triaxial accelerometer for seven days, during which total energy expenditure (TEE) was measured as an estimate of participants’ physical activity (ActiHeart; CamNtech, Cambridge, UK) and was corrected for body mass.

Blood pressure measures

Appropriately sized brachial cuffs were fitted to participants. Clinic BP was measured in duplicate on the right and left arms, with the arm rested at heart level (Dinamap Procare 100 Vital Signs Monitor, GE Medical Systems, Milwaukee, USA). We reported clinic BP as the average of all four measurements.

Biochemical analyses

Participants fasted from 22:00 on the evening prior to the day of the study. Early morning spot urine samples were taken, and a registered nurse collected blood samples from the antebrachial vein using a sterile winged infusion set and syringes. Samples were immediately taken to the onsite laboratory for preparation and storage at −80 °C. Twenty-four-hour urine samples were collected according to the Pan American Health Organization/WHO protocol for population-level sodium determination in 24-hour urine samples [13]. After discarding the first passed urine, participants collected all urinary voids in a 5 L container. Urine was stored in a cool place until a researcher came to collect the 5L container from participants. Twenty-four-hour urinary volume was measured, whereafter samples were aliquoted into cryovials for storage also at −80 °C.

Total cholesterol, high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), triglycerides, creatinine, and gamma-glutamyl transferase (GGT) were measured in serum and glucose in sodium fluoride plasma (Cobas Integra, 400plus, Roche, Basel Switzerland). Insulin was determined from serum using the Electrochemiluminescence method on the E411 (Roche, Basel Switzerland). Serum creatinine was used to determine the estimated glomerular filtration rate (eGFR) using the Chronic Kidney Disease Epidemiology equation (CKD-EPI) (without the race component) [14]. Twenty-four-hour urinary sodium and potassium were measured using ion-selective electrode potentiometry, and creatinine with the Creatinine Jaffé Gen.2 reagent (Cobas Integra, 400plus, Roche, Basel Switzerland). Completeness for 24-hour urine samples was based on urine volume (volume ≥300 mL) and 24-hour urinary creatinine (men: ≤6 and ≥30 mmol/day; women: ≤4 and ≥25 mmol/day) [15].

$$ Estimated\,salt\,intake\left( {g/day} \right) \\ = \frac{{UNa\left( {mmol/L} \right) \ast Uvolume\,(L/24hr) \ast 23\left( {molecular\,weight\,Na} \right)}}{{390}}$$

We have previously published detailed methodology on the metabolomic analyses done in the African-PREDICT study [16]. Amino acids and acylcarnitines were measured from spot urine samples using liquid chromatography-tandem mass spectrometry (LC-MS/MS) (Agilent 6410 LC-MS/MS system with 1200 series LC front-end). We applied a 50% CV filter to the data, and metabolites with more than 50% zero values were removed from the data matrix.

Statistical analyses

Statistical analyses were performed using SPSS version 27 (IBM, Armonk, New York, USA), and figures were created with GraphPad Prism version 8.4.3 (GraphPad software Inc., La Jolla, California, USA). Data for continuous variables that were not normally distributed were logarithmically transformed. Basic characteristics of Black and White adults were compared across low (<5 g/day), moderate (5–10 g/day), and high (>10 g/day) salt intake groups using Welch analysis of variance (ANOVA). The current recommendation of the World Health Organization is less than 5 g salt per day, with most adults consuming almost double [17]. Therefore, increasing increments of 5 grams salt was used to define groups. Data for animal protein intake remained skew despite logarithmic transformation, therefore we performed a nonparametric Kruskal-Wallis H test to compare animal protein intake across salt intake groups. To lower the false discovery rate of statistically significant metabolites as a result of performing multiple comparisons, we calculated the adjusted p values using the Benjamini–Hochberg procedure (reported as q values). We additionally performed Pearson correlations, and partial and linear regression analyses to demonstrate the relationship of estimated salt intake with clinic SBP and significant metabolites (identified using ANOVA). Age, WC, BMI, TEE, total cholesterol, HDL-C, LDL-C, triglycerides, glucose, insulin, eGFR, and GGT were considered as covariates for linear and stepwise multiple regression models. We ultimately included age, WC, TEE, triglycerides, glucose, eGFR, and GGT into backward stepwise multiple regression models based on the strongest bivariate correlations with SBP measures and metabolites. As a sensitivity analyses, we repeated ANOVA with participants stratified according to sex, and included sex as a covariate in linear regression models.

Results

Table 1 demonstrates the characteristics of Black and White adults across low to high salt intake. In Black adults, clinic BPs were similar across salt intake groups (p trend > 0.05). In White participants, the high salt intake group had a higher clinic SBP than those in the low or moderate intake groups (p trend < 0.001). Anthropometric measures (p trend  < 0.001), total (p trend = 0.004) and animal protein intake (p trend = 0.020) increased with an increase in estimated salt intake in White adults only.

Table 1 Characteristics of Black and White adults across low (<5g salt/day), medium (5–10 g salt/day) and high salt intake (>10 g salt/day) groups
Full size table

When we compared thirty-five amino acids and acylcarnitines across salt intake groups, we found no significant differences in metabolite levels in Black adults (all p values > 0.05). In White adults however, we demonstrated lower GABA (p = 0.001), citrulline (p = 0.001), glutamic acid (p = 0.004), serine (p = 0.006), proline (p = 0.008), aspartic acid (p = 0.023), arginine (p = 0.026), dimethylglycine (p = 0.037) and glycine (p = 0.049) with increased salt intake. After using the Benjamini–Hochberg procedure, GABA (q = 0.020), citrulline (q = 0.020) and glutamic acid (q = 0.046) remained significantly different between the salt intake groups (Table 2). Additionally, White adults in the high salt intake group had attenuated levels of proline (q = 0.042) and serine (q = 0.042) compared to those in the low intake group.

Table 2 Significant metabolites across salt intake groups
Full size table

Relationship between estimated salt intake and metabolites

We subsequently performed Pearson (Fig. 2) and partial correlations to determine relationships between the identified metabolites and estimated salt intake in the Black and White groups. In Black adults, estimated salt intake correlated negatively with proline (r = −0.107; p = 0.037). In White adults we confirmed inverse relationships of estimated salt intake with GABA (r = −0.140; p = 0.004), citrulline (r = −0.134; p = 0.011), glutamic acid (r = −0.098; p = 0.038) and serine (r = −0.107; p = 0.024), and a borderline significant correlation with proline (r = 0.086; p = 0.069). The negative correlations of estimated salt intake with GABA and citrulline persisted after partial adjustment for either age, WC, TEE, or animal protein intake. The correlations of estimated salt intake with serine lost significance when adjusting for animal protein intake, and with glutamic acid when adjusting for TEE or animal protein intake. With partial adjustment for clinic SBP, all relationships between estimated salt intake and metabolites became non-significant (Supplementary Table 1).

Fig. 2
figure 2

Pearson correlations between urinary metabolites, estimated salt intake and clinic SBP. Pearson correlations are presented for (A) Black adults () and (B) White adults () from the African-PREDICT study

Full size image

Relationship between systolic blood pressure and estimated salt intake

With Pearson analysis we observed positive correlations between estimated salt intake and clinic SBP in Black (r = 0.101; p = 0.049) and White adults (r = 0.230; p < 0.001) (Fig. 2). In Black adults the relationship between estimated salt intake and SBP became non-significant after adjusting for age, WC, and TEE (Adj. R2 = 0.026; Std. β = 0.079; p = 0.18). In White adults, however, the positive relationship between estimated salt intake and clinic SBP remained robust after adjusting for age, WC, TEE, glucose, triglycerides, GGT, eGFR, and animal protein intake (Model 4) (Adj. R2 = 0.35; Std. β = 0.142; p = 0.001) (Fig. 3). The relationship between estimated salt intake and clinic SBP was also independent of GABA, citrulline, glutamic acid, serine and proline.

Fig. 3
figure 3

Multiple regression analyses with clinic SBP as the dependent variable and identified metabolites as well as estimated salt intake as the main independent variables. Model 1: adjusted for waist circumference and total energy expenditure. Model 2: adjusted for waist, circumference, total energy expenditure, glucose and triglycerides, λ-glutamyl transferase. Model 3: adjusted for waist, circumference, total energy expenditure, glucose, triglycerides, λ-glutamyl transferase and estimated glomerular filtration rate. Model 4: adjusted for waist, circumference, total energy expenditure, glucose and triglycerides, λ-glutamyl transferase, estimated glomerular filtration rate and animal protein intake

Full size image

Relationship between systolic blood pressure and metabolites

In the following analyses we determined whether clinic SBP was related to metabolites that differed significantly across salt intake groups. Only in White, and not Black adults, did we observe significant negative correlations between SBP and amino acids (GABA, serine, and proline) after multivariable adjusted analyses (Fig. 3). In the White group only, inverse association of clinic SBP with GABA (Adj. R2 = 0.34; Std. β = −0.133; p = 0.003), serine (Adj. R2 = 0.33; Std. β = −0.109; p = 0.014) and proline (Adj. R2 = 0.33; Std. β = −0.109; p = 0.014) persisted with linear regression analyses after adjusting for age, WC, TEE, glucose, GGT, triglycerides, and eGFR, and a borderline significant association with glutamic acid (Adj. R2 = 0.33; Std. β = −0.083; p = 0.063). These negative associations were substantiated with backward stepwise multiple regression analyses for GABA (Adj. R2 = 0.35; Std. β = −0.135; p = 0.002), serine (Adj. R2 = 0.34; Std. β = −0.110; p = 0.011), proline (Adj. R2 = 0.34; Std. β = −0.111; p = 0.010) and glutamic acid (Adj. R2 = 0.34; Std. β = −0.085; p = 0.049). When NaCl or animal protein intake was added to linear regression models, the negative association of clinic SBP with GABA (Adj. R2 = 0.36; Std. β = −0.114; p = 0.012) (Adj. R2 = 0.35; Std. β = −0.120; p = 0.009), serine (Adj. R2 = 0.35; Std. β = −0.091; p = 0.038) (Adj. R2 = 0.34; Std. β = −0.092; p = 0.040) and proline (Adj. R2 = 0.35; Std. β = −0.096; p = 0.030) (Adj. R2 = 0.34; Std. β = −0.91; p = 0.047) remained robust.

Sensitivity analyses for sex

When stratified according to sex, we found that the reduction of metabolite levels with increased salt intake in the White group was particularly evident in women (Supplementary Table 2). Although dimethylglycine, serine, proline, glutamic acid, GABA and citrulline had a p trend <0.05 across increasing salt intake groups in White women, all q values were >0.05. Notably sex also had a significant confounding effect on the relationships observed between clinic SBP, estimated salt intake and metabolites (GABA, serine, and proline) (Fig. 3).

Discussion

In this study we investigated the triad of salt intake, clinic SBP and the associated metabolomics profile. We found that in White, but not Black adults, a high salt intake associated with a unique metabolomics profile consisting of lower levels of GABA, glutamic acid, serine, and proline (Fig. 4)––which in turn were related to elevated SBP. Furthermore, in White adults only, estimate salt intake associate positively with clinic SBP after multiple adjustments. Although the relationship between estimated salt intake and clinic SBP was independent of identified metabolites we still postulate that lower levels of these amino acids––related to increased salt intake––may play a contributing role in pathophysiological mechanisms whereby excessive salt intake increases SBP in White individuals.

Fig. 4
figure 4

Amino acids affected by dietary salt intake in white adults. This figure illustrates the amino acids that were shown to be affected by dietary salt intake in the white participants, and their link with physiological mechanisms associated with increased BP

Full size image

Metabolomics profile of Black adults in relation to dietary salt intake

Despite Black ethnicity being associated with salt sensitivity we did not observe significant differences in clinic SBP, amino acids or acylcarnitines across low to high salt intake groups. Additionally, the observed correlations of estimated salt intake with clinic SBP, and GABA were non-significant after adjusting for WC and TEE expenditure, suggesting that the latter relationship are dependent on central adiposity and physical activity. Given that WC, BMI and TEE were similar across salt intake groups, it may explain the lack of a significant increase in SBP with higher salt intake. Using targeted metabolomics, this study only measured amino acids and acylcarnitines. The likelihood of salt intake influencing other metabolite classes such as organic compounds, phospholipids, sphingolipids, advanced glycation end-products and others––all linked to arterial stiffness [18]––cannot be disregarded. An untargeted metabolomics approach in future studies may help identify other metabolite classes affected by salt intake in Black ethnic groups. Arterial stiffness and early vascular aging are hallmark cardiovascular risk factors associated with Black ethnicity, and can precede marked BP changes [19, 20]. Our research group have previously demonstrated independent associations of arterial stiffness with β-alanine, 1-methyl-histidine, L-proline, threonic acid and malonic acid in Black but not White boys (6–8 years of age) [21]. Further investigations into possible relationships between amino acids, acylcarnitines and arterial stiffness are required.

Metabolomics profile of White adults in relation to dietary salt intake

Glutamic acid, serine and proline are glucogenic acids that can be converted to glucose via gluconeogenesis [22]––a metabolic pathway associated with reduced insulin sensitivity and type 2 diabetes mellitus (T2DM) [23]. Given the health status of this study cohort, it is an alluring prospect that the early changes in the glucogenic metabolites associated with a high salt intake, could reflect the initial stages of a harmful continuum ranging from increased gluconeogenesis to hyperinsulinemia, insulin resistance, hyperglycaemia, and ultimately the development of T2DM. Considering the higher BMI and WC with excessive salt intake in the White group, central obesity may play an important role. Although not the focus of this article, we did indeed find negative correlations between glucogenic metabolites (glutamic acid, serine, and proline) and glucose in our cohort (Supplementary Table 4). There is also some evidence that the role of gluconeogenesis in the aetiology of insulin resistance may be ethnic-specific [24]. Chung et al., showed that Black women had a 10% lower rate of gluconeogenesis despite similar age, BMI, and whole-body insulin sensitivity than their White counterparts. The authors suggest that while increased gluconeogenesis may be a hallmark of insulin resistance in Whites, it may not be the case for Black ethnic groups [24].

We also observed an increase in animal protein intake with a higher salt intake. This is in line with a positive correlation reported between animal protein intake and sodium from a recent study that evaluated the nutritional contents of 971 foods from the South African Food Composition Database [25]. While it is known that dietary protein intake affects endogenous metabolite levels, it was beyond the scope of this article. We did, however, find that animal protein intake influenced the relationship of salt intake with glutamic acid, serine, and proline. Still, the marked reduction in glutamic acid, serine, and proline with excessive salt intake, be it due to the high sodium content associated with animal protein intake or not, remains of interest.

Metabolite markers and blood pressure

GABA and glutamic acid

Glutamic acid (glutamate) and GABA are major excitatory and inhibitory neurotransmitters in the central nervous system that play a role in BP regulation [26]. In White adults, lower levels of sympathoinhibitory GABA, associated with a high salt intake, may increase clinic SBP by means of increased sympathetic activity. A recent review by Ngo et al., additionally highlighted not only the anti-hypertensive properties of GABA, but also the antioxidative, anti-inflammatory, anti-diabetic, hepatic––and reno-protective effects thereof [27].

Although a positive relationship between SBP and glutamic acid could be expected based on the notion that glutamic acid/glutamate promotes sympathetic outflow, we observed an inverse relationship with unadjusted analyses. Accordingly negative relationships between SBP and glutamic acid were also reported for findings from the INTERMAP study [28] and the Twins UK registry [29]. This relationship in our study, however, was significantly influenced by WC and physical activity. Stemler et al. suggested that the inverse association between SBP and glutamate may be attributed to glutamate being a substrate for arginine, a precursor of nitric oxide (NO) which promotes vasodilation. Glutamate itself, however, can also increase the bioavailability of NO by activating neuronal nitric oxide synthase (nNOS) [30]. The inhibition of nNOS in healthy men (24 ± 4 years) has been reported to increase mean arterial pressure and systemic vascular resistance [31]. Conversely, there is also evidence that nNOS activation by glutamate may promote oxidative stress and vasoconstriction [30].

Serine and proline

White adults from our study with a habitual daily salt intake >10g/day had significantly lower levels of serine and proline compared with those consuming <5g salt/day. The inverse relationships of serine and proline with SBP in White adults from this study again correspond with findings from the INTERMAP study mentioned earlier [28]. Higher levels of serine are also associated with a lower risk of developing hypertension [32, 33]. Elevated L-serine may play an antihypertensive role by promoting endothelium-dependent vasodilation. In vitro, L-serine was shown to exhibit antioxidative and cytoprotective properties by promoting nuclear factor erythroid 2 related factors or activity, heme-oxygenase-1 expression and NO production [34]. Lower serine levels, associated with increased salt intake, may therefore increase SBP by means of impaired endothelial vasodilation in White adults.

In contrast with our findings and that of the INTERMAP study, Teymoori et al., reported a greater risk of incident hypertension associated with increased proline intake [35]. Proline plays an important role in maintaining healthy collagen turnover and stability [36]. It is possible that alterations in proline metabolism may contribute to adverse collagen turnover which promotes arterial stiffness and concurrently increases BP.

Strengths and limitations

Our study cohort of apparently healthy Black and White adults (aged 20–30 years) is larger than previously reported studies that have focused on metabolomics profiling and salt intake [8,9,10,11]. Since this study only recruited participants from the North West Province of South Africa the results may not be reflective of the South-African population in its entirety. We made use of a targeted metabolomics approach to measure predetermined metabolite groups (amino acids and acylcarnitines). Future studies may use untargeted metabolomics to determine whether other metabolite groups such as organic acids, not included in our study, are affected by salt intake. While multiple 24-hour urine samples are recommended to determine individual salt intake, the use of a single 24-hour urine sample to determine habitual population salt intake is sufficient [37].

Conclusion

In White adults, early metabolomic changes related to salt intake may contribute to an increased risk of hypertension and concurrently early CVD development. Our results suggest that the manner in which excessive salt intake promotes elevated SBP in White compared to Black adults may be mediated by different pathophysiological mechanisms.