Altered hormonal and autonomic nerve responses to hypo- and hyperglycaemia are found in overweight and insulin-resistant individuals and may contribute to the development of type 2 diabetes

This study was conducted at the Uppsala University Hospital and the Department of Medical Sciences at Uppsala University. Participants were aged 18–60 years with a BMI of 18.5–50 kg/m². They were recruited by advertisements in newspapers and public spaces, after previously participating in other studies or attending the outpatient obesity unit. Exclusion criteria were as follows: diagnosis of diabetes, endocrine or other diseases that could influence the results or the participants’ ability to participate in the study; use of medication with metabolic side effects; planned or ongoing pregnancy; and significant substance abuse. In the recruitment process, we aimed for a wide distribution of BMI. Participants were recruited across a wide range of BMI, aiming to have a representation of lean (18.5–24.9 kg/m²), overweight or mildly obese (25.0–34.9 kg/m²) and severely obese (35.0–50.0 kg/m²) individuals.

Participants underwent a stepwise hyperinsulinaemic euglycaemic–hypoglycaemic clamp and a stepwise hyperglycaemic clamp on two occasions separated by 1–5 weeks. The order of the two clamps was randomised 1:1 in blocks of four subjects, and this was not blinded. At each visit, anthropometrics were obtained and body composition was assessed using bioimpedance (Tanita body composition analyzer, BC-418; Tanita Corporation, Tokyo, Japan). Heart rate and HRV were measured during both visits with a custom-made single-channel ECG recording system (developed at Biomedical engineering, R&D, University Hospital Umeå, Sweden). Baseline blood samples were drawn, after participants had fasted for at least 10 h overnight at both visits, just before the start of each clamp at approximately 09:00 hours.

The hyperinsulinaemic euglycaemic–hypoglycaemic clamps (henceforth denoted hypoglycaemic clamps), depicted in Fig. 1a, were performed as previously described [26] and modified by our group through previous studies [24]. At 0 min, simultaneous infusions of insulin (56 mU m⁻² min⁻¹ after 10 min priming), potassium chloride (8 mmol/h) and glucose (200 mg/ml) were started. The glucose infusion rate (GIR) was adjusted to initially achieve a glucose level of 5.0 mmol/l followed by a stepwise lowering towards a nadir of 2.7 mmol/l. At 185 min, the insulin infusion was terminated and glucose was infused at a fixed rate of 200 mg kg^-1 h^-1 (recovery phase). The stepwise hyperglycaemic clamp protocol (Fig. 1b) was inspired by previously described two-step hyperglycaemic clamps [27, 28]. A variable glucose infusion (200 mg/ml) was administered to raise plasma glucose above the fasting level in three steps. At 165 min the glucose infusion was terminated to allow glucose levels to normalise. Glucose levels were analysed every 5 min during the clamps with blood drawn from an arterialised vein using a Contour Glucose Meter (Bayer Healthcare, Leverkusen, Germany).

During the clamps, hormonal and haemodynamic measurements were obtained at regular intervals. Single-channel ECG recordings were continuously made throughout each clamp. During hypoglycaemic clamps, participants were also asked to assess their hypoglycaemic symptoms according to the Edinburgh Hypoglycaemia Symptom Scale (ESS) [29, 30]. Further details are provided in electronic supplementary material (ESM) Methods.

Samples were analysed immediately or frozen at −80°C until analysis. Glucagon ELISA, NEFA fluorometric and glycerol colourimetric assays were analysed at the Clinical Diabetes Research Laboratory. All other analyses were performed at the Department of Clinical Chemistry at the Uppsala University Hospital, Sweden using a hexokinase method for glucose and immunoassays for hormones. Details of the analyses are provided in ESM Methods.

ECG recordings were automatically processed and manually inspected. The total spectral power (P_TOT), low-frequency power (P_LF, 0.04–0.15 Hz) and high-frequency power (P_HF, 0.15–0.50 Hz) were calculated. P_HF mainly reflects the parasympathetic part of cardiac autonomic modulation, while P_LF reflects a combination of sympathetic and parasympathetic activity. The ratio P_LF/P_HF is used as a proxy for balance between sympathetic and parasympathetic activity [31]. The HRV analysis was performed using Matlab Software version R2019a (MathWorks, Natick, MA, USA). More details are provided in ESM Methods.

This was an exploratory study, and no formal power calculation for sample size selection was performed. The primary endpoint was the glucagon levels during experiments, including hypo- and hyperglycaemia. As estimated from previous studies, there is more than 80% power to detect 25% differences in glucagon levels between groups with 15 participants each, assuming an inter-subject CV of 0.20 and a two-sided α of 0.05; the same estimate also applies for cortisol levels [24, 25, 32‐35]. In the analyses, participants were allocated to two equally large groups, lean (LO) and overweight/obese (HI) with cut-off at the median BMI. Mann–Whitney U tests were consistently used for groupwise comparisons of continuous variables and Fisher’s exact test for categorical variables. p values <0.05 were considered significant.

Spearman’s Rank correlation analyses and multilinear regression analyses were performed on pooled data for all participants between neuroendocrine response variables (hormones and HRV indices) during the hypoglycaemic and hyperglycaemic phases and candidate predictors.

To describe hormonal responses vs achieved glucose levels, scatterplots of hormone levels at the end of each clamp stage vs the mean glucose for the preceding 20 min were used for visual presentation, and linear mixed models were used for statistical inference. All statistical methods are described in detail in ESM Methods.

All analyses were performed in SPSS for Mac version 25 (IBM Corp., Armonk, NY, USA). Figures and graphs were made using GraphPad Prism version 8.3.0 (GraphPad Software, San Diego, CA, USA).

All study procedures were performed in accordance with the Declaration of Helsinki. The local Research Ethics Committee of Uppsala gave their approval (DNR 2017/550). The participants received oral and written information about the study and signed an informed consent form.

The main finding from the hypoglycaemic clamps was an augmented responsiveness of the cortisol axis to hypoglycaemia among overweight and insulin-resistant participants compared with lean and more insulin-sensitive participants. This appears to be of central origin, involving hypothalamic and pituitary functions, since ACTH and cortisol responses were similarly elevated in the HI vs LO group.

While the elevated hypoglycaemic symptom scores in HI vs LO did not reach significance, there were significant, BMI-independent, associations between symptoms and the cortisol axis response, suggesting a causal connection between perceived glucopenia and the augmented cortisol axis response in overweight individuals. Undoubtedly, the anatomical bridge for this connection would be within the CNS. Therefore, an increased CNS sensing of hypoglycaemia in obese individuals is possible but obviously not proven. This would be expected to raise the glycaemic ‘setpoint’ for cortisol axis responses but the magnitude of this shift could not be exactly defined due to the limited sample size and experimental design. However, as visualised in Fig. 5(c,e), a physiological cut-off for hypoglycaemia of 3 mmol/l appears to be shifted to about 3.3 mmol/l in the HI vs LO group with respect to ACTH and cortisol responses. Moreover, the fact that the enhanced cortisol axis response was inversely associated with the M value, independent of BMI, points to a potential role in the development of insulin resistance. This may be further amplified by an increased local generation of cortisol in adipose tissue, and elevated tissue cortisol can be hypothesised to play a role in type 2 diabetes development [36]. Naturally, longitudinal studies of larger cohorts are needed to further support this hypothesis.

We evaluated sympathetic and parasympathetic nerve activity by HRV assessments. Given the increased symptoms of hypoglycaemia, an augmented sympathetic response in the overweight group might have been surmised. However, both obesity and, more clearly, insulin resistance was associated with a less dynamic ANS response to hypoglycaemia, characterised by less parasympathetic inhibition. Insulin resistance but not obesity was also associated with less sympathetic activation. Of interest, our group has previously reported that gastric bypass surgery was followed by an attenuated ANS response to hypoglycaemia [24]. We also reported that visceral adiposity [33] as well as insulin resistance [37] was associated with an increased sympathetic/parasympathetic ratio under normoglycaemic conditions. It should be acknowledged that HRV has limitations as a marker of ANS activity and has been questioned [38]. Although P_HF supposedly reflects parasympathetic activity, the P_LF represents both sympathetic and parasympathetic activity, and the P_LF/P_HF ratio is utilised to reflect their relative contributions. Moreover, the importance and exact peripheral mechanisms of the two branches of the ANS with regards to glucose regulation is still uncertain [39‐41]. The role of parasympathetic activity is particularly controversial. Apart from a proposed stimulation of glucagon secretion [40], our research group reported an unexpected rapid increase in insulin sensitivity following infusion of atropine [42], suggesting a paradoxical short-term effect by cholinergic pathways of the parasympathetic system to reduce peripheral glucose uptake. Thus, the attenuated inhibition of parasympathetic activity observed in overweight and insulin-resistant individuals during hypoglycaemia may potentially enhance the defence against hypoglycaemia and the maintenance of elevated everyday glucose levels. Catecholamine levels would be of interest in this context but they were presently not possible to analyse and were not considered as critical, since HRV assessments did not suggest any substantial group difference in sympathetic activity.

We found no differences in glucagon levels between groups during hypoglycaemia, nor were they associated with measures of obesity or insulin resistance. However, there were group differences in achieved glucose and insulin levels that may have underestimated differences in hormone responses. When adjusting for glucose levels in linear mixed models, both obesity and insulin resistance were indeed associated with significantly higher glucagon levels during hypoglycaemia. This is in concordance with a previous study showing augmented glucagon, ACTH and noradrenaline (norepinephrine) responses in obese individuals exposed to hypoglycaemia [7]. The more pronounced differences in their study may be due to use of the less specific RIA technique for glucagon measurement [32], greater BMI difference between groups and a younger and all-male study population.

The attenuation of the growth hormone response during hypoglycaemia in overweight individuals did not reach significance in our study, but has been observed in previous studies [18, 19].

During the hyperglycaemic clamps, overweight and insulin-resistant participants displayed less suppression of glucagon than lean participants. This is in accordance with some previous studies [43‐45] and may contribute to the development and progression of insulin resistance and potentially type 2 diabetes.

The cortisol responses to hyperglycaemia were highly variable between individuals and there were no consistent differences between groups. The initial decline in both groups is most likely explained by diurnal variations and the subsequent rise could represent a glucose-mediated stress response. A rise in cortisol levels after meals or an oral glucose load is well-established in previous work [46, 47].

We found markedly lower growth hormone levels during hyperglycaemia in overweight participants but glucose-mediated inhibition was similar to that in lean participants, which is somewhat different to findings of a previous study [48].

The P_LF/P_HF ratio rose during hyperglycaemia but was significantly lower in the overweight group indicative of a decreased sympathetic relative to parasympathetic activity. This resembles findings during hypoglycaemic conditions. Thus, the sympathetic response to acute hyperglycaemia and hyperinsulinaemia appears to be impaired in overweight insulin-resistant individuals, perhaps because of adaptation to slightly elevated glucose levels.

There are some limitations to this study. First, as previously discussed, the minutely higher glucose and insulin levels achieved during clamps in overweight compared with lean participants may have affected insulin-antagonistic responses. Notably, this would mainly underestimate the differences in hormone levels found and, for completeness, we also adjusted for actual glucose levels in regression analyses. Second, while the elevated hormonal responses associated with obesity and insulin resistance are compatible with a hypothesised upward shift in glycaemic setpoint, our current data do not allow more than a very rough quantification of this shift. We plan analyses of data from pooled cohorts to address this. Third, the design may be underpowered to detect hypothetical effect sizes of interest. Fourth, several of the neuroendocrine alterations reported in this study were small in magnitude and the clinical implications need to be confirmed. Fifth, participants were recruited based on BMI rather than insulin resistance and the associations with neuroendocrine responses should be interpreted in the light of this. However, recruiting participants based on measures of insulin resistance or dysglycaemia (e.g. following OGTTs) would have markedly hampered feasibility. Finally, no adjustment for multiple comparisons was made. This was due to the exploratory nature of this work and also to biological interdependencies of the measured neurohormonal responses.

Overall, these findings are hypothesis-generating and need to be confirmed in larger studies.

This study demonstrates that overweight insulin-resistant individuals compared with lean individuals have increased central activation of the cortisol axis during hypoglycaemia, associated with more pronounced hypoglycaemic symptoms. This suggests an increased CNS-mediated response to hypoglycaemia. The finding that insulin resistance, more than obesity, is associated with the cortisol axis response is compatible with a causal role of the neurohormonal responses for the development of dysglycaemia and potentially type 2 diabetes. Furthermore, an attenuated suppression during hyperglycaemia and, to a lesser extent, an augmented glucagon rise during hypoglycaemia seem to be features of both insulin resistance and obesity. The anatomical sites, such as brain, pancreas or both, involved in this dysregulation remain to be elucidated in onward studies. By contrast, there is an attenuation of autonomic nerve responses to glucose fluctuations in overweight and insulin-resistant individuals that may reflect a less dynamic sympathetic and parasympathetic regulation, which in the long term may potentially contribute to dysglycaemia.

Taken together, altered insulin-antagonistic responses, including the cortisol axis, glucagon and ANS, in obese insulin-resistant individuals may contribute to the development of long-term dysglycaemia and, hypothetically, also type 2 diabetes. These perturbations may involve glucoregulatory functions of the brain shifting the ‘glycaemic setpoint’ for glucose-regulating hormones upwards. Our ongoing and future work will include individuals with type 2 diabetes and also the use of neuroimaging techniques to assess regional brain responses to hypo- and hyperglycaemia during diabetes development.