Hydrogen sulphide is endogenously produced in the mammalian vasculature from the amino acids cysteine and homocysteine by the pyridoxal-5′-phosphate-dependent enzyme cystathionine-γ-lyase (CSE; EC (reviewed in Whiteman and Moore [1]). Vascular H2S acts as an endothelium- and KATP-channel-dependent vasodilator [1] and plasma and serum H2S levels in rodents and healthy humans are reported to be in the range 40–80 µmol/l [1]. Higher levels of H2S have been observed in animal models of endotoxic and haemorrhagic shock, in which pharmacological inhibition of CSE using d,l-propargylgycline significantly increased blood pressure and survival (reviewed in Li et al. [2]). Conversely, lower H2S levels were reported in animal models of hypertension, in which the administration of H2S ‘donors’ decreased blood pressure [1]. These observations are consistent with the role of H2S as an endogenous vasodilator.

Recently, a role for H2S in the aetiology of diabetes has been suggested. Pancreatic synthesis of H2S is markedly elevated in the streptozotocin (STZ) rat [3], in which biphasic effects on beta cells have been observed; at low concentrations, H2S inhibited insulin release through KATP-dependent/Ca2+-independent mechanisms [4], whereas higher levels induced beta cell death through endoplasmic-reticular-stress-dependent pathways [5]. However, plasma H2S levels were lower after STZ treatment [3]. Similarly, in the NOD mouse, plasma levels of H2S and the responsiveness of vascular tissue to endothelium-dependent vasodilators, such as acetylcholine, declined and aortic synthesis of H2S were also progressively reduced as diabetic pathology increased [6].

Since there is increased risk of hypertension and cardiovascular disease and loss of vascular responsiveness to endogenous vasodilators in individuals with type 2 diabetes mellitus and obesity, we investigated whether there would also be a loss of vasodilatory H2S in type 2 diabetes and obesity.



Participants were recruited from advertisements in the local community. Individuals were excluded from the study if they had suffered a stroke or myocardial infarction, or had uncontrolled hypertension (>160/90 mmHg). Participants with type 2 diabetes were also excluded if they were treated with insulin. Additional exclusion criteria are listed in the Electronic Supplementary Material (ESM). Three study groups were recruited: (1) lean men (BMI < 25.0 kg/m2) (n = 11); (2) age-matched overweight men (BMI > 25.0 kg/m2) (n = 16); and (3) age-matched men with type 2 diabetes who were BMI matched (to within ±2 kg/m2) to the overweight non-diabetic men (n = 11). Participants’ characteristics are shown in Table 1. All assessments were performed in a temperature-controlled laboratory (22.0 ± 0.5°C) with the participants having fasted from 22:00 hours the previous evening and having taken no medication on the morning (09:00 hours) of the tests. All patients provided written informed consent. The study was approved by the Devon and Torbay Ethics Committee and was performed in accordance with principles of the Declaration of Helsinki.

Table 1 Participant demographics

Measurement of plasma H2S

Peripheral blood samples (10 ml; EDTA anti-coagulated) were collected as described previously [1, 3, 4, 6]. H2S levels were determined in triplicate as described previously [16] and concentrations calculated against a calibration curve of sodium sulphide (1.06–100 µmol/l; Sigma-Aldrich, MO, USA). Briefly, 250 μl zinc acetate (1% wt/vol. in water) was injected into 200 μl plasma in tightly sealed Eppendorf vials followed by injection of 133 μl N,N-dimethyl-p-phenylenediamine sulphate (20 mmol/l in 7.2 mol/l HCl; Sigma-Aldrich) and 133 μl FeCl3 (30 mmol/l in 1.2 mol/l HCl; Sigma-Aldrich) [1, 3, 4, 6]. After incubation in the dark for 30 min, samples were centrifuged at 5,000 g for 10 min and the absorbance of the supernatant fraction at 670 nm determined. The intra-assay coefficient of variation was 0.1–3.76% over the concentration range studied (n = 12). No significant interference was observed in the presence of physiological concentrations of sulphite (SO 2−3 ), sulphate (SO 2−4 ), thiosulphate (S2O 2−3 ), reduced glutathione, oxidised glutathione, cysteine, cystine, homocysteine, homocystine, cystathionine, methionine, nitrite (NO 2 ) or nitrate (NO 3 ), (p = NS, Kruskal–Wallis; n = 4). At physiological pH, H2S (pK a 7.04) dissociates to the hydrosulphide anion (HS) and the sulphide anion (S2−) [1]. Therefore, we use the term H2S to refer to the sum of these species (H2S, HS and S2−) present at physiological pH [1, 3, 4].

Systemic blood pressure

Supine blood pressure was measured from the left arm using a semi-automatic blood pressure recorder (Critikon Dinamap, Deltona, FL, USA). Five measurements were obtained at 1 min intervals and the mean of the final three readings was taken to be the representative blood pressure.

Microvascular assessment

Maximum hyperaemia was assessed by heating a small area of skin to 42–44°C [7], which induces maximum hyperaemia. This was achieved by attaching a small brass heater (area 0.76 cm2) to the dorsum of the foot for 30 min. The resultant maximum hyperaemic response was assessed by single-point laser Doppler fluximetry (LDF; Perflux Pf2: Perimed, Järfälla Sweden), eight equally spaced measurements were made within the heated area and the mean used to represent maximum hyperaemia, arbitrarily expressed as volts. Day-to-day intra-individual variation with this technique is 6.6% (mean ± standard deviation: 1.81 ± 0.12 V) in one participant assessed over five separate occasions. Minimum vascular resistance (MVR) of the skin microcirculation was calculated by dividing mean arterial pressure (MAP) by maximum hyperaemia (MVR = MAP/MH, mmHg/V) [7].

Insulin sensitivity assessments

Central insulin sensitivity was calculated by HOMA insulin sensitivity, based on fasting insulin and glucose levels [8]. Peripheral insulin sensitivity was assessed using a 15 min continuous insulin tolerance test (ITT) [7]. A bolus of 0.1 U kg−1 of insulin (Human Actrapid; Novo Nordisk Pharmaceuticals, Crawley, UK) was administered into a vein at the site of the anticubital fossa. Blood samples for the determination of blood glucose were taken each minute until the blood glucose level reached 3 mmol/l or 15 min duration was reached [7]. The glucose samples were measured in duplicate (YSI2300 Stat Plus; Yellowsprings Instruments, Yellowsprings, OH, USA). The slope of the blood glucose curve was used to represent insulin sensitivity [7].

Biochemical variables

All blood samples were obtained after assessment of microcirculatory function. HbA1c (normal range 4.0–6.0%) was determined using an in-house high performance liquid chromatography method (Hewlett Packard 1050 HPLC) [7]. Fasting glucose was determined using a glucose oxidase method (GOD-PAP; Roche Modular analyser, Roche Diagnostics, Lewes, UK) [7]. Triacylglycerol, lipids and cholesterol were determined using a colorimetric assay (using a Vitros Analyser; Johnson & Johnson Clinical Diagnosis, Amersham, UK) [7]. An electrochemiluminescence assay was used for the determination of fasting plasma insulin (0.05% cross-reactivity with human proinsulin) (Cobas; Roche, Basel, Switzerland) [7].

Statistical analysis

Statistical analysis was performed by the Statistical Package for Social Sciences (SPSS) version 15.0 (Chicago, IL, USA). The Mann–Whitney test was used for between-group analyses. To enable the examination of associations between H2S levels and diabetes and obesity, data from the three participant groups were merged (n = 38) to generate a continuum of data. An adiposity measure (waist circumference, waist:hip ratio or BMI) and diabetes status were initially entered into a forced regression model. To determine whether the relationship between plasma H2S with obesity and diabetes status was influenced by their related milieu, potential confounders—such as systolic blood pressure, fasting glucose, HbA1c, minimum vascular resistance, insulin sensitivity as assessed by both HOMA and insulin tolerance test—were each entered into separate regression models. Data were checked to ensure the assumptions required for regression analysis were fulfilled. This study was powered to detect a between-individual difference of 1.4 standard deviations, and a large effect size (f 2 = 0.420) with three predictors in the linear regression model at the 5% level with 90% power. p values less than 0.05 were considered significant.


Participant demographics are shown in Table 1. Median plasma H2S levels (25th, 75th percentiles) in lean controls, overweight volunteers and men with type 2 diabetes were 38.9 (29.7, 45.1) µmol/l, 22.0 (18.6, 26.7) µmol/l and 10.5 (4.8, 22.0) µmol/l, respectively (Fig. 1). Median plasma H2S levels were significantly lower in patients with type 2 diabetes compared with age-matched controls (p = 0.001) and overweight men (p = 0.008), presumably because of the increased adiposity of the type 2 diabetes group. Median plasma H2S levels in overweight volunteers were also significantly lower than in lean age-matched controls (p = 0.003). In the univariate analysis, plasma H2S was negatively correlated with systolic (r = −0.580, p < 0.001) and diastolic (r = −0.527, p < 0.001) blood pressures, glycaemic control (fasting glucose, r = −0.494, p = 0.001; Hb1Ac, r = −0.423, p = 0.006) and insulin sensitivity (peripheral insulin [ITT], r = −0.497, p = 0.004; central insulin [HOMA], r = 0.566, p < 0.001). Microvascular function tests further showed significant associations with plasma H2S (maximum hyperaemia, r = 0.402, p = 0.008; minimum vascular resistance, r = −0.436, p = 0.004). However, the strongest correlations were observed with measurements of adiposity i.e. waist circumference (r = −0.650, p < 0.001), waist:hip ratio (r = −0.657, p < 0.001), BMI (r = −0.609, p < 0.001) and, to a lesser extent, hip circumference (r = −0.358, p = 0.017). Regression analysis suggested that adiposity, as assessed by waist circumference, but not diabetes status was a significant independent predictor of plasma H2S levels (waist circumference: standardised beta: −0.543, p = 0.001; diabetes status: standardised beta: −0.257, p = 0.077). The observed independent relationship between waist circumference and plasma H2S remained after adjusting for the potential confounders: systolic blood pressure, minimum vascular resistance, age, central insulin sensitivity (HOMA) and total cholesterol:HDL ratio (standardised beta range of −0.432 to −0.541, p value range 0.01–0.001; ESM Table 1). Similar trends were observed between plasma H2S and other measures of adiposity (i.e. waist:hip ratio and BMI; ESM Tables 2 and 3, respectively). Waist:hip ratio, but not BMI or waist circumference, was an independent predictor of plasma H2S levels when adjusting for peripheral insulin sensitivity (waist:hip ratio: −0.407, p = 0.029; waist circumference: −0.283, p = 0.109; BMI: −0.183, p = 0.329), suggesting that the relationship between adiposity and plasma H2S may be related to peripheral insulin sensitivity. However, diabetes status was not a significant independent predictor of plasma H2S in any of these models in this study.

Fig. 1
figure 1

Plasma H2S levels in lean healthy and overweight volunteers and patients with type 2 diabetes mellitus. Plasma (EDTA) was collected from participants and each sample analysed in triplicate for H2S as described in the Methods. Data are expressed as median (25th and 75th percentiles). All samples were age matched. Overweight volunteers were matched by BMI to type 2 diabetic patients. The Mann–Whitney test was used to determine differences between sample groups with significance set at p < 0.05. Lean controls vs overweight (p = 0.003), lean controls vs type 2 diabetic patients (p = 0.001), overweight vs type 2 diabetic patients (p = 0.008). Outlying data points are shown by crosses


It is well established that diabetes is associated with the clustering of central obesity, dyslipidaemia, raised blood pressure, hyperglycaemia and increased risk of cardiovascular disease [9]. However, the precise molecular mechanisms underpinning these clinical associations are currently not known. H2S is emerging as important vasodilatory intermediate, inducing endothelium-dependent and KATP-channel-dependent vasorelaxation in vitro and in vivo. Our current observations that plasma H2S levels negatively correlated with systolic and diastolic blood pressure, microvascular dysfunction, glycaemic control and insulin sensitivity are the first in man and are consistent with the observations in animal models of diabetes and hypertension [1, 3, 6].

A potential role for CSE-derived H2S in adipose fat metabolism has also been proposed in rats [10]. In our study, the major determinant of plasma H2S levels was adiposity. This relationship was not accounted for by diabetes status, glycaemic control, blood pressure, microvascular function, lipids or central insulin sensitivity. However, when adjusting for peripheral insulin sensitivity, the independent relationships of plasma H2S levels with BMI and waist circumference—but not waist:hip ratio—were lost, suggesting that peripheral insulin sensitivity may be contributing to the relationship between adiposity and plasma H2S.

A limitation of the current study is the potential for a type 2 error. However, the fact that H2S was consistently associated with adiposity as assessed by BMI, waist circumference and waist:hip ratio, endorses the proposal that an increase in adiposity is associated with a reduction in plasma levels of H2S. However, further longitudinal studies with increased sample sizes are required to substantiate the precise relationship between plasma levels of H2S and diabetes.

In summary, our study has shown that adiposity is a major determinant of plasma levels of H2S. The loss of the vasodilatory gas H2S could represent a novel mechanism for mediating cardiovascular complications in obesity and type 2 diabetes. Whether this loss occurs because of reduced H2S synthesis, enhanced enzymatic removal or increased consumption by oxidants known to be elevated in the diabetic milieu requires further attention.