Early View
Research Article
Open Access

Exercise training modifies skeletal muscle clock gene expression but not 24-hour rhythmicity in substrate metabolism of men with insulin resistance

Jan-Frieder Harmsen

Jan-Frieder Harmsen

Department of Nutrition and Movement Sciences, NUTRIM School for Nutrition and Translational Research in Metabolism, Maastricht University Medical Center, Maastricht, The Netherlands

Search for more papers by this author
Marit Kotte

Marit Kotte

Department of Nutrition and Movement Sciences, NUTRIM School for Nutrition and Translational Research in Metabolism, Maastricht University Medical Center, Maastricht, The Netherlands

Search for more papers by this author
Ivo Habets

Ivo Habets

Department of Nutrition and Movement Sciences, NUTRIM School for Nutrition and Translational Research in Metabolism, Maastricht University Medical Center, Maastricht, The Netherlands

Search for more papers by this author
Frederieke Bosschee

Frederieke Bosschee

Department of Nutrition and Movement Sciences, NUTRIM School for Nutrition and Translational Research in Metabolism, Maastricht University Medical Center, Maastricht, The Netherlands

Search for more papers by this author
Koen Frenken

Koen Frenken

Department of Nutrition and Movement Sciences, NUTRIM School for Nutrition and Translational Research in Metabolism, Maastricht University Medical Center, Maastricht, The Netherlands

Search for more papers by this author
Johanna A. Jorgensen

Johanna A. Jorgensen

Department of Nutrition and Movement Sciences, NUTRIM School for Nutrition and Translational Research in Metabolism, Maastricht University Medical Center, Maastricht, The Netherlands

Search for more papers by this author
Soraya de Kam

Soraya de Kam

Department of Nutrition and Movement Sciences, NUTRIM School for Nutrition and Translational Research in Metabolism, Maastricht University Medical Center, Maastricht, The Netherlands

Search for more papers by this author
Esther Moonen-Kornips

Esther Moonen-Kornips

Department of Nutrition and Movement Sciences, NUTRIM School for Nutrition and Translational Research in Metabolism, Maastricht University Medical Center, Maastricht, The Netherlands

Search for more papers by this author
Jochem Cissen

Jochem Cissen

Department of Nutrition and Movement Sciences, NUTRIM School for Nutrition and Translational Research in Metabolism, Maastricht University Medical Center, Maastricht, The Netherlands

Search for more papers by this author
Daniel Doligkeit

Daniel Doligkeit

Department of Nutrition and Movement Sciences, NUTRIM School for Nutrition and Translational Research in Metabolism, Maastricht University Medical Center, Maastricht, The Netherlands

Search for more papers by this author
Tineke van de Weijer

Tineke van de Weijer

Department of Nutrition and Movement Sciences, NUTRIM School for Nutrition and Translational Research in Metabolism, Maastricht University Medical Center, Maastricht, The Netherlands

Search for more papers by this author
Edmundo Erazo-Tapia

Edmundo Erazo-Tapia

Department of Nutrition and Movement Sciences, NUTRIM School for Nutrition and Translational Research in Metabolism, Maastricht University Medical Center, Maastricht, The Netherlands

Search for more papers by this author
Mijke Buitinga

Mijke Buitinga

Department of Nutrition and Movement Sciences, NUTRIM School for Nutrition and Translational Research in Metabolism, Maastricht University Medical Center, Maastricht, The Netherlands

Search for more papers by this author
Joris Hoeks

Joris Hoeks

Department of Nutrition and Movement Sciences, NUTRIM School for Nutrition and Translational Research in Metabolism, Maastricht University Medical Center, Maastricht, The Netherlands

Search for more papers by this author
Patrick Schrauwen

Corresponding Author

Patrick Schrauwen

Department of Nutrition and Movement Sciences, NUTRIM School for Nutrition and Translational Research in Metabolism, Maastricht University Medical Center, Maastricht, The Netherlands

Corresponding author P. Schrauwen: Department of Nutrition and Movement Sciences, Maastricht University Medical Centre, P.O. BOX 616, 6200MD Maastricht, The Netherlands.  Email: [email protected]

Search for more papers by this author
First published: 05 December 2023
Citations: 1

Handling Editors: Karyn Hamilton & Josiane Broussard

The peer review history is available in the Supporting Information section of this article (https://doi.org/10.1113/JP285523#support-information-section).

Abstract

Twenty-four hour rhythmicity in whole-body substrate metabolism, skeletal muscle clock gene expression and mitochondrial respiration is compromised upon insulin resistance. With exercise training known to ameliorate insulin resistance, our objective was to test if exercise training can reinforce diurnal variation in whole-body and skeletal muscle metabolism in men with insulin resistance. In a single-arm longitudinal design, 10 overweight and obese men with insulin resistance performed 12 weeks of high-intensity interval training recurrently in the afternoon (between 14.00 and 18.00 h) and were tested pre- and post-exercise training, while staying in a metabolic research unit for 2 days under free-living conditions with regular meals. On the second days, indirect calorimetry was performed at 08.00, 13.00, 18.00, 23.00 and 04.00 h, muscle biopsies were taken from the vastus lateralis at 08.30, 13.30 and 23.30 h, and blood was drawn at least bi-hourly over 24 h. Participants did not lose body weight over 12 weeks, but improved body composition and exercise capacity. Exercise training resulted in reduced 24-h plasma glucose levels, but did not modify free fatty acid and triacylglycerol levels. Diurnal variation of muscle clock gene expression was modified by exercise training with period genes showing an interaction (time × exercise) effect and reduced mRNA levels at 13.00 h. Exercise training increased mitochondrial respiration without inducing diurnal variation. Twenty-four-hour substrate metabolism and energy expenditure remained unchanged. Future studies should investigate alternative exercise strategies or types of interventions (e.g. diet or drugs aiming at improving insulin sensitivity) for their capacity to reinforce diurnal variation in substrate metabolism and mitochondrial respiration.

Key points

  • Insulin resistance is associated with blunted 24-h flexibility in whole-body substrate metabolism and skeletal muscle mitochondrial respiration, and disruptions in the skeletal muscle molecular circadian clock.
  • We hypothesized that exercise training modifies 24-h rhythmicity in whole-body substrate metabolism and diurnal variation in skeletal muscle molecular clock and mitochondrial respiration in men with insulin resistance.
  • We found that metabolic inflexibility over 24 h persisted after exercise training, whereas mitochondrial respiration increased independent of time of day.
  • Gene expression of Per13 and Rorα in skeletal muscle changed particularly close to the time of day at which exercise training was performed.
  • These results provide the rationale to further investigate the differential metabolic impact of differently timed exercise to treat metabolic defects of insulin resistance that manifest at a particular time of day.

Introduction

To be prepared for alternating periods of day and night, feeding and fasting, and activity and recovery, the body contains a central biological clock, located in the suprachiasmatic nucleus (SCN) of the hypothalamus, which sets the circadian rhythm. Over the last decade, it has become clear that peripheral tissues can have their own rhythm, controlled by a cell-autonomous molecular clock, consisting of a transcriptional–translational feedback loop. This includes the transcriptional activators BMAL1 and CLOCK, which induce the expression of their own repressors, CRY and PER, thereby generating approximately 24-h oscillations affecting up to 40% of the genome (Zhang et al., 2014). Given the fact that the core clock machinery controls the expression of more than 2300 genes that regulate metabolism and signalling in skeletal muscle (Pizarro et al., 2013), the importance of proper functioning of the molecular clock for various downstream processes and cellular functions becomes evident. Indeed, genetic mouse models lacking a functional clock due to knockout of essential core-clock components in a whole-body or tissue-specific manner develop hyperglycaemia, hyperinsulinemia and glucose intolerance (Delezie et al., 2012; Dyar et al., 2014; Harfmann et al., 2016; Hodge et al., 2015; Marcheva et al., 2010; Paschos et al., 2012; Perelis et al., 2015; Rakshit et al., 2016; Zhang et al., 2010). Similar associations have been found in humans with Gabriel et al. (2021) and our group recently showing altered 24-h rhythmicity in muscle clock gene expression and ablated rhythmic mitochondrial metabolism in individuals with insulin resistance (Wefers et al., 2020). While young healthy men displayed 24-h rhythmicity in muscle mitochondrial respiration (i.e. lowest at 13.00 h and highest at 23.00 h, based on a tissue sampling frequency of every 5 h), mitochondrial rhythmicity was absent in older men with insulin resistance (Wefers et al., 2020). Under controlled conditions, including three typical meals provided during daytime, both groups of men demonstrated a 24-h rhythm in whole-body substrate metabolism. However, whereas young healthy men switched from carbohydrate oxidation during the day to lipid oxidation during the night, this fed-to-fasted transition was blunted in men with insulin resistance, who maintained high levels of carbohydrate oxidation during the night (Harmsen et al., 2022; Wefers et al., 2020). Although the reliance on nocturnal carbohydrate oxidation during the night might have been driven by consuming a relatively large dinner (e.g. almost half of daily energy intake), the meal distribution was similar for both groups and adjusted for the individuals’ energy demand, thus suggesting that insulin resistance is associated with a reduced capacity to switch from the fed to fasting state.

The impact of lifestyle and drug interventions are often evaluated based on their efficacy to increase insulin sensitivity or other markers in the morning after an overnight fast. With disrupted 24-h rhythmicity evident upon insulin resistance, the further capacity of such interventions to align, restore or amplify the circadian system requires investigation. While light input represents the dominant external ‘Zeitgeber‘ (‘time-giver’, from German) for the oscillators of SCN neurons, timing of food intake is well known to affect the rhythm of peripheral clocks (Eckel-Mahan & Sassone-Corsi, 2013; Gutierrez-Monreal et al., 2020; Opperhuizen et al., 2016). More recently, physical activity and exercise have also been acknowledged as a time cue for the central clock (da Rocha et al., 2022; Schroeder et al., 2012), but also for peripheral tissues, i.e. skeletal muscle (Adamovich et al., 2021; da Rocha et al., 2022; Kemler et al., 2020; Small et al., 2020; Wolff & Esser, 2012; Yamanaka et al., 2008). For example, clock gene expression in human leukocytes responds to an acute bout of endurance exercise in a time of day-dependent manner (Tanaka et al., 2020), but the number of studies in humans is limited in this regard.

Chronic exercise training is well known to ameliorate mitochondrial function and insulin sensitivity, also in individuals with insulin resistance (Meex et al., 2010; Phielix et al., 2010), but the impact on diurnal variation in whole-body and skeletal muscle (mitochondrial) metabolism remains elusive. Therefore, the present single-arm longitudinal trial was set out as a proof-of-concept study to investigate the impact of chronic exercise training on 24-h rhythmicity in whole-body substrate metabolism and diurnal variation in skeletal muscle molecular clock and mitochondrial function in men with insulin resistance.

Methods

Participants

Male, overweight and obese volunteers between the ages of 40 and 75 years were recruited through advertisements in the vicinity of Maastricht and Aachen from January 2020 to September 2022. Participants were non-smokers and generally healthy, based on a medical questionnaire and examination by a physician. Only participants with a habitual bedtime of 23.00 ± 2 h and 7–9 h sleep per day were included. In terms of prior habits of physical activity, participants were only included if they engaged in less than 2 h of structured exercise per week, and if they had stable body weight over the last 3 months before enrolment (±3 kg). Participants were excluded from the study if they performed shift work or travelled across more than one time zone in the 3 months before the study. By means of the Morningness-Eveningness Questionnaire Self-Assessment Version 1.3 we excluded extreme morning or evening types (MEQ-SA; exclusion based on a score ≤30 for extreme evening or ≥70 for extreme morning type). Individuals with habitually high caffeine (i.e. >400 mg caffeine daily) and alcohol consumption (i.e. >20 g alcohol/day) were also excluded. Further exclusion criteria to ensure safety of participants during exercise training sessions were plasma haemoglobin levels <7.8 mmol/l, and cardiac abnormalities revealed through an electrocardiogram judged by a physician. To establish insulin resistance or pre-diabetes by means of an oral glucose tolerance test (75 g of glucose), participants had to fulfil at least one of four criteria to be included in the study: (1) impaired fasting glucose (6.1–6.9 mmol/l); (2) impaired glucose tolerance (7.8–11.1 mmol/l 2 h after glucose consumption); (3) haemoglobin A1C (HbA1c) of ≥5.7%; or (4) low insulin sensitivity defined as a glucose clearance rate ≤360 ml/kg/min. The first two criteria are derived from the World Health Organization recommendations (World Health Organization & International Diabetes Federation, 2006), the third criterion is based on the definition of pre-diabetes from the American Diabetes Association (American Diabetes Association, 2016) and the fourth criterion is used as a measure of insulin sensitivity based on the OGIS model (Mari et al., 2001). The study was conducted in accordance with the principles of the Declaration of Helsinki and approved by the Ethics Committee of the Maastricht University Medical Centre (METC20-030). All participants provided written informed consent before data collection. The study was independently monitored by the Clinical Trial Center Maastricht and was registered at https://clinicaltrials.gov with identifier NCT03733743.

Study conditions

Run-in period prior to pre- and post-exercise training laboratory testing

During a 7-day run-in period prior to admission to the research facility, participants were instructed to adhere to a standardized lifestyle that resembled the experimental conditions during the two overnight stays. During this run-in period participants had to refrain from alcohol and caffeinated drinks, sleep only between 23.00 and 07.00 h (±30 min), and only eat three meals per day at 08.00, 13.00 and 18.00 h without snacking in between. Standardized meals with fixed caloric content adjusted to the participants’ needs (described below) were provided for the last 2 days before the visit to the lab. Moreover, volunteers were instructed to refrain from exercise for the last 3 days of the run-in period. Adherence to the activity protocol and sleeping times at home was monitored using wrist-worn actigraphy (Actiwatch Spectrum Plus) and a sleep diary. Mealtimes were documented in a food diary for 1 week, and consumed food items and quantities were noted for the last 3 days before the laboratory visit to ensure consistency between pre- and post-exercise testing. Volunteers followed a similar lifestyle protocol, with the exception of the exercise training bouts for the last week before the post-exercise testing.

Pre- and post-exercise laboratory testing

In general, to ensure comparability to our previous studies, pre- and post-exercise laboratory stays were mostly performed as described previously in the studies by van Moorsel et al. (2016) and Wefers et al. (2020). Volunteers were admitted to the research unit at 12.00 h on test day 1 and stayed for 46 h in total, under standardized conditions. The first test day was used to standardize and monitor meals, physical activity and bedtime. Meals were provided at 13.00 h and 18.00 h in our research facilities. One hour after every meal, participants went for a 15-min low-intensity walk accompanied by a researcher to incorporate some standardized physical activity. Thereafter, participants were instructed to stand for 15 min before they could sit again. Participants stayed in a respiration chamber, which is a small room with a bed, toilet, sink, desk, chair, TV and computer. At 23.00 h, the lights of the respiration chamber were switched off and participants were instructed to try to sleep. During the first night, whole-room indirect calorimetry (Omnical, Maastricht Instruments, Maastricht, The Netherlands) (Schoffelen et al., 1997) was conducted to determine sleeping metabolic rate, as defined as the minimum average energy expenditure over a 3-h interval between 00.00 h and 07.00 h.

On the second test day, an intravenous cannula was placed in the forearm for subsequent blood draws starting at 08.00 h, followed by an indirect calorimetry measurement using a ventilated hood while awake and at rest lying supine to assess resting energy expenditure and substrate oxidation. Thereafter, the first skeletal muscle biopsy was taken (described below). These combined measurements (blood draw, indirect calorimetry and skeletal muscle biopsy) were repeated at 13.00 h and 23.00 h. Next to having 08.00 h as the typical overnight fasted sample, we chose to add 13.00 h and 23.00 h for the following reasons: (1) in young healthy individuals, 13.00 h and 23.00 h are the time points of lowest and highest mitochondrial respiration, respectively (van Moorsel et al., 2016), (2) at these times the largest differences in clock gene expression of Cry1 and Per2 were observed between young healthy and men with insulin resistance (Wefers et al., 2020), and (3) these two time points are close to midday and midnight, respectively, and can therefore indicate the contrast between day and late evening. Further consecutive blood draws and ventilated hood measurements were performed at 18.00 h and 04.00 h to close the full 24-h cycle. Note that we were not able to use data obtained from the respiration chamber, due to the fact that we frequently had to open and enter the chamber, for blood and muscle biopsy sampling. Additional blood samples were taken at least bihourly (10.00, 11.00, 12.00, 14.00, 16.00, 20.00, 22.00, 00.00, 02.00, 06.00 and 08.00 h). The meals on the second test day were provided immediately after the muscle biopsies and were hence delayed by ∼1 h compared to the run-in period and the first test day. After the 23.00 h biopsy, participants returned to the respiration chamber to sleep with lights off. At 04.00 h, the participants were woken up and transferred to another room in a wheelchair and in dim light to conduct the last ventilated hood measurement, after which the participants were transferred back to the respiration chamber and were allowed to sleep until 07.00 h. For the 02.00 and 06.00 h blood draws, a researcher entered the respiration chamber under dim light, so that participants usually woke up but remained lying in bed. After the last blood draw at 08.00 h on the third day, body composition was determined using air-displacement plethysmography (BodPod, Cosmed, Rome, Italy) according to the manufacturer's protocol, as previously reported (Plasqui et al., 2011).

Post-exercise training testing started on the day after the last high-intensity interval training (HIIT) session. Thereby, the time interval between the last exercise stimulus and the first blood draw and muscle biopsy was approximately 40 h to minimize confounding from the acute exercise response. The time course of pre- and post-exercise testing is summarized in Fig. 1A.

Details are in the caption following the image
Figure 1. Design of pre- and post-exercise laboratory testing and the exercise training protocol
A, pre- and post-exercise testing. B, exercise training intervention. HIIT, high-intensity interval training; Wmax, maximal wattage.

Study meals

Two days prior to admission to the laboratory research unit, and during the stay, participants were provided with a diet in energy balance based on Dutch dietary guidelines. Caloric intake for consumption at home was calculated by multiplying the estimated resting metabolic rate, obtained with the Harris–Benedict formula (Harris & Benedict, 1918) multiplied by an activity factor of 1.5. Participants were provided with optional extra snacks to eat with their meals if they were still hungry, up to an activity factor of 1.7. For the first test day in the laboratory, energy requirement was calculated by multiplying the estimated resting metabolic rate with an activity factor of 1.35, because of limited physical activity in the research unit. For the second test day during pre-exercise testing, energy requirement was calculated by multiplying the sleeping metabolic rate of the first study night (measured by whole-room indirect calorimetry) by 1.5. For post-exercise testing, we provided volunteers with the same standardized meals as for pre-exercise testing. For the last 2 days of the run-in period and the 2 days at our research unit, participants received three meals daily. Breakfast accounted for ∼21 energy % (E%), lunch for ∼30 E% and dinner for ∼49 E%. Daily macronutrient composition was ∼52 E% as carbohydrates, ∼31 E% as fat (∼9 E% saturated), and ∼14 E% as protein. Breakfast and lunch were bread-based and hence the proportion of energy from carbohydrates was higher compared to the dinner. No drinks or snacks other than water were provided in-between meals.

Skeletal muscle biopsies

Three skeletal muscle biopsies were obtained from the middle region of the musculus vastus lateralis according to the Bergström method (Bergström et al., 1967) under local anaesthesia (1% lidocaine, without epinephrine). Each biopsy was taken from a separate incision at least 2 cm from the previous incision, moving from distal to proximal. The first biopsy was randomly taken from the left or right leg, and each subsequent biopsy was taken from the other leg. The order of legs was kept similar between pre- and post-exercise training testing. All biopsies were taken at least 4 h after the last meal to minimize postprandial confounding. Muscle samples were dissected carefully and freed from any visible non-muscle material. One part of the biopsy was immediately placed in ice-cold preservation medium (BIOPS, Oroboros Instruments, Innsbruck, Austria) and used for the preparation of permeabilized muscle fibres and further assessment of mitochondrial oxidative capacity. The remaining part of the muscle biopsy was immediately frozen in melting isopentane and stored in −80°C until further analysis.

High-resolution respirometry

Directly after each biopsy, the collected muscle tissue was partly used to measure ex vivo mitochondrial function using high-resolution respirometry as previously reported (Phielix et al., 2008). In permeabilized muscle fibres, oxygen consumption was measured by high-resolution respirometry using an Oxygraph (Oroboros Instruments). A multisubstrate-uncoupler protocol with malate, octanoylcarnitine, ADP, glutamate, succinate and carbonylcyanide p-trifluoromethoxyphenylhydrazone (FCCP) was performed, as described previously (Hoeks et al., 2010). The integrity of the outer mitochondrial membrane was assessed as a quality control by the addition of cytochrome c upon maximal coupled respiration. All oxygen consumption measurements were performed in quadruplicate and means were taken for statistical analysis.

Indirect calorimetry

Oxygen consumption and carbon dioxide production were measured with an automated respiratory gas analyser using a ventilated hood system (Omnical; IDEE, Maastricht, the Netherlands) and were used to calculate whole-body energy expenditure, respiratory exchange ratio (RER), and glucose and fat oxidation. Substrate oxidation was calculated using the Brouwer equation (Brouwer, 1957) with protein oxidation being estimated as 12.4% of energy expenditure. Sleeping metabolic rate was defined as the lowest 3 h of nocturnal energy expenditure during the first night in the respiration chamber and calculated with the Weir equation (Weir, 1949).

Maximal exercise capacity tests

The pre-exercise laboratory stay and the first maximal exercise capacity test were separated by at least 7 days to ensure recovery from the muscle biopsies. After a 5-min warm-up at 50 W, workload was set at 100 W and increased 25 W every 2.5 min until exhaustion. For participants with a body weight >100 kg, the initial workload progression was set to 100 W, followed by 150 W and thereafter similarly increased by 25 W steps. Peak oxygen consumption ( V ̇ O 2 max ${\dot{V}}_{{{\mathrm{O}}}_{\mathrm{2}}{\mathrm{max}}}$ ) was defined as the highest average value over 30 s. Volunteers were stimulated to exercise until exhaustion, which was monitored by heart rate, respiratory exchange ratio (RER) and the rate of perceived exertion (RPE). In all volunteers, RER was >1.0 when V ̇ O 2 max ${\dot{V}}_{{{\mathrm{O}}}_{\mathrm{2}}{\mathrm{max}}}$ was reached. Maximal power output (maximal wattage, Wmax) was calculated as the workload in the last completed stage and workload relative to the time spent in the last incomplete stage: wattage of last completed stage + (time in seconds)/150 × 25 W. This maximal exercise capacity test was repeated after 2, 5, 8 and 11 weeks of training and served as a training session by itself and to readjust the workload progressively during HIIT sessions. Only for the first and last maximal test, participants wore a spirometry mask for indirect calorimetry to determine peak oxygen consumption from pre- to post-exercise.

Exercise training protocol

As an exercise regimen, volunteers performed high-intensity interval training (HIIT) over a period of 12 weeks with 3 sessions per week. Each supervised training session consisted of 10 intervals of 1 min at 80–90% of an individual's Wmax interspersed by 2 min active rest periods at 20% Wmax. A similar exercise methodology has been prescribed before in obese subjects and was shown to be safe and have beneficial effects on postprandial glucose metabolism (Little et al., 2014). Our participants performed all training sessions deliberately only in the afternoon (between 14.00 and 18.00 h), as current evidence suggests that training in the afternoon or evening is more beneficial for glycaemic control when compared to morning training with respect to type 2 diabetes patients (Savikj et al., 2019) and overweight and obese individuals (Mancilla et al., 2021; Moholdt et al., 2021). The exercise protocol is summarized in Fig. 1B.

Gene transcript quantification

RNA was isolated from 10 mg of muscle material by TRIzol lysis (Qiagen, Hilden, Germany). RNA was further purified by the RNeasy kit from Qiagen (Hilden, Germany). RNA yield was measured using a NanoDrop spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). The high-capacity RNA-to-cDNA kit from Thermo Fisher Scientific was used for transcribing 250 ng of RNA to cDNA. Transcript abundance was determined using a CFX384 Real-Time System (Bio-Rad Laboratories, Veenendaal, the Netherlands). To minimize the variability in reference gene normalization, the geometric mean of two reference genes (RPL26 and RPLP0), which were both stably expressed over time, was used. This geometric mean was used as the internal reference for comparative gene expression analysis over time and between groups (Hellemans et al., 2007).

Blood analysis

Blood samples were collected in serum- and EDTA-containing tubes and centrifuged at 1300 g for 10 min at room temperature and 4°C. Aliquots were frozen in liquid nitrogen and stored at −80°C. Serum triacylglycerol without correction for free glycerol (Roche Diagnostics, Mannheim, Germany), plasma glucose (Horiba, Montpellier, France) and free fatty acid levels were measured colorimetrically using a Cobas Pentra C400 analyser (Horiba).

Statistical analysis

A two-way repeated measures ANOVA with time and exercise (condition: pre- vs. post-exercise) and their interaction as fixed effects was applied to test for differences over time from pre to post in 24-h plasma metabolites, energy metabolism, mitochondrial respiration and muscle clock gene expression. In cases of missing time points, a generalized linear mixed model was run instead of the two-way repeated measures ANOVA. A one-way repeated measures ANOVA was performed to test for a time effect in the aforementioned outcomes separately for pre- and post-exercise. Student's paired t test was used to detect differences from pre to post in all other outcomes without repeated measures. For the 24-h plasma metabolites and muscle clock gene expression analyses, complete data were only available for nine participants. Data are presented as means ± SD. The level of significance was set to <0.05 for all analyses. Statistical analyses were performed in the GraphPad Prism 8 software package (GraphPad Software, San Diego, CA, USA).

Results

Exercise training improved body composition and exercise capacity

Ten men with a mean age of 64 ± 6 years who were overweight or obese (BMI: 33.1 ± 4.1 kg/m2) participated in the study. Participant characteristics are summarized in Table 1. Physical activity count derived from the Actiwatch confirmed adherence to the sleep instructions for both 7-day run-in periods, and physical activity did not differ over the run-in period between pre- and post-exercise training testing (Supplemental Fig. 1). An important goal of the study was to isolate the exercise training effect and avoid any confounding from potential weight loss, since it has been shown that weight loss influences muscle clock gene expression (Sardon Puig et al., 2020) and whole-body substrate oxidation (Rabøl et al., 2009). Therefore, we tracked participants’ body weight regularly over the training period. Participants maintained their body weight without deliberate efforts (Pre: 101.7 ± 16.3 kg vs. Post: 101.1 ± 16.2 kg, P = 0.242). However, we observed positive changes in body composition (fat mass percentage: Pre: 36.8 ± 4.9% vs. Post: 35.7 ± 4.9%; fat-free mass percentage: Pre: 63.2 ± 4.9% vs. Post: 64.3 ± 4.9%, P = 0.01). As expected, participants also improved their aerobic capacity: maximal power output increased by 10% (Pre: 1.98 ± 0.31 W/kg body weight vs. Post: 2.17 ± 0.31 W/kg, P = 0.012) and V ̇ O 2 max ${\dot{V}}_{{{\mathrm{O}}}_{\mathrm{2}}{\mathrm{max}}}$ by 6% (Pre: 28.0 ± 4.2 ml/min/kg body weight vs. Post: 29.8 ± 3.7 ml/min/kg, P = 0.042). Individual data on body weight, body composition, maximal power output and V ̇ O 2 max ${\dot{V}}_{{{\mathrm{O}}}_{\mathrm{2}}{\mathrm{max}}}$ are shown in Supplemental Fig. 2.

Table 1. Participant characteristics (n = 10)
Participants characteristics Mean ± SD
Age (years) 64 ± 6
Body mass index (kg/m2) 33.1 ± 4.1
Body fat (%) 36.8 ± 4.9
Systolic blood pressure (mmHg) 148 ± 9
Diastolic blood pressure (mmHg) 93 ± 5
Habitual bedtime (h) 23:13 ± 0:43 – 07:03 ± 0:58
MEQ-SA score 60 ± 4
Fasting plasma glucose (mmol/l) 5.8 ± 0.4
Fasting plasma insulin (μIU/ml) 115 ± 62
90 min plasma insulin (μIU/ml) 844 ± 361
2-h plasma glucose (mmol/l) 7.4 ± 2.0
HbA1c (%) 5.6 ± 0.3
Glucose clearance (ml/kg/min) 323 ± 48
Impaired glucose homeostasis (n)
Impaired fasting glucose, 6.1–6.9 mmol/l 3
Impaired glucose tolerance, 7.8–11.1 mmol/l 3
Elevated HbA1c, 5.7–6.4% 3
Impaired insulin sensitivity, OGIS ≤ 360 ml/min/m2 9
  • MEQ-SA, Morningness-Eveningness Questionnaire Self-Assessment Version. Scores of 41 and below indicate ‘evening types’. Scores of 59 and above indicate ‘morning types’. Scores between 42 and 58 indicate ‘intermediate types’. OGIS, oral glucose insulin sensitivity.

Whole-body substrate oxidation indicates that metabolic inflexibility over 24 h persists after exercise training

To determine changes in 24-h rhythmicity in whole-body substrate oxidation and energy expenditure after exercise training, we performed indirect calorimetry under a ventilated hood in the lying resting position every 5 h from 08.00 h to 04.00 h on the second test day (n = 10). Mostly due to providing meals only during daytime that induce changes in energy expenditure and substrate utilization, the two-way repeated-measures ANOVA revealed a significant time effect for all outcomes (P < 0.001): resting energy expenditure, respiratory exchange ratio (RER), carbohydrate and fat oxidation (Fig. 2A–D). Resting energy expenditure, RER and carbohydrate oxidation were all lowest at 08.00 h and peaked at 23.00 h, while fat oxidation showed a time course opposite to carbohydrate oxidation. As shown previously (Wefers et al., 2020), the expected drop in RER from 23.00 h to 04.00 h, as is observed in young healthy volunteers (van Moorsel et al., 2016), was not observed in our study, indicating the inability of older volunteers with insulin resistance to switch to fat oxidation during the night. As no condition (RER: P = 0.506, resting energy expenditure: P = 0.977) or interaction effects (RER: P = 0.728, resting energy expenditure: P = 0.286) were found, exercise training did not influence whole-body substrate oxidation and energy expenditure measured at five time points over 24 h, and was not able to restore the lack of fed-to-fasting transition in substrate oxidation.

Details are in the caption following the image
Figure 2. 24-h whole-body substrate oxidation and energy expenditure
24-h whole-body substrate oxidation and energy expenditure (n = 10) remain unchanged from pre- (blue dots and line) to post-exercise training (red dots and line). Whole-body resting energy expenditure (A), respiratory exchange ratio (B), carbohydrate oxidation (C), and fat oxidation (D). The dark grey area represents the sleeping period (00.00-07.00 h). Data are presented as means ± SD.

Although volunteers stayed in a respiration chamber, the above measurements had to be conducted with a ventilated hood in another room, in the awake state, to allow sampling of blood and muscle biopsies at the selected time points. However, during the first night of pre- and post-exercise training testing, volunteers stayed continuously in the respiration chamber allowing the measurement of sleeping metabolic rate and nocturnal RER. Sleeping metabolic rate and nocturnal RER did not differ between pre- and post-exercise testing (P = 0.906 and P = 0.729, respectively) and nocturnal RER was similar to the RER measured at 04.00 h (awake) (Supplemental Fig. 3).

Exercise training increases mitochondrial oxidative capacity independent of time of day

We recently showed for young healthy men that mitochondrial respiration shows 24-h rhythmicity, which is absent in older individuals with insulin resistance (Wefers et al., 2020). Here, we performed high-resolution respirometry in freshly isolated permeabilized muscle fibres sampled at 08.00, 13.00 and 23.00 h (n = 10). We confirmed that in volunteers with insulin resistance, no time effect was found for any of the mitochondrial respiration states before exercise training (state 3 MO: P = 0.191, state 3 MOG: P = 0.274, state 3 MOGS: P = 0.146, state uncoupled: P = 0.903; where M is malate, O is octanoylcarnitine, G is glutamate and S is succinate), indicating lack of 24-h rhythmicity, and exercise training did not change this (state 3 MO: P = 0.869, state 3 MOG: P = 0.688, state 3 MOGS: P = 0.993, state uncoupled: P = 0.14). However, exercise training did have an overall effect on all mitochondrial respiration states, as shown by a condition effect (P < 0.001, Fig. 3A–D), suggesting that exercise training enhanced mitochondrial respiration irrespective of time of day. Bonferroni post hoc tests revealed that, particularly at 13.00 and 23.00 h, all mitochondrial respiration states increased (P < 0.05), while only state U respiration was also increased at 08.00 h (P = 0.034, Fig. 3D). No interaction effects were found (state 3 MO: P = 0.328, state 3 MOG: P = 0.286, state 3 MOGS: P = 0.226, state uncoupled: P = 0.221), suggesting that exercise training did not induce diurnal variation in mitochondrial respiration.

Details are in the caption following the image
Figure 3. Mitochondrial oxidative capacity in skeletal muscle
Mitchondrial oxidative capacity in skeletal muscle (n = 10) increases from pre- (blue dots and line) to post-exercise training (red dots and line). ADP-stimulated respiration of permeabilized muscle fibres fuelled with the lipid substrate octanoylcarinitine (state 3 MO; A); addition of complex I substrates (state 3 MOG; B); addition of substrates for parallel electron input into complex I and II (state 3 MOGS; C); and maximal uncoupled respiration after FCCP (state U) titration (D). M, malate; O, octanoylcarnitine; G, glutamate; S, succinate. Data depict oxygen consumption per mg wet weight per second and are shown as means ± SD.*P < 0.05 based on Bonferroni post hoc tests.

Muscle clock gene analysis reveals period genes as most responsive to exercise training

To determine if exercise training can modify the skeletal muscle clock of older men with insulin resistance, we analysed mRNA levels of clock genes known to oscillate over 24 h (Gutierrez-Monreal et al., 2020; Harmsen et al., 2022) in skeletal muscle biopsies collected at 08.00, 13.00, and 23.00 h on the second test day (n = 9). Two-way repeated-measures ANOVAs revealed a significant time effect for mRNA levels of all clock genes (P < 0.001, for Cry1: P = 0.02), except for Rorα (P = 0.445), for which muscle sampling at 04.00 h is crucial to detect 24-h rhythmicity, as we previously showed (Harmsen et al., 2022). mRNA levels of Bmal1 showed a robust difference between day and late evening, which was unaffected by exercise training (Fig. 4A). A condition effect was found for Clock (P = 0.017, Fig. 4B) and Rorα (P = 0.04, Fig. 4D), both having lower mRNA levels after exercise training. An interaction effect was found for Per1 (P = 0.013, Fig. 4E) and Per3 (P = 0.006, Fig. 4G), whereas Per2 (P = 0.088, Fig. 4F) and Reverbα (P = 0.053, Fig. 4C) showed a trend for interaction. Bonferroni post hoc tests revealed that all three period genes had or tended to have lower mRNA levels after exercise training particularly at 13.00 h (Per1: P = 0.068, Per2: P = 0.026, Per3: P = 0.055). Similarly, gene expression of Rorα was lower at 13.00 h after exercise training (P = 0.038).

Details are in the caption following the image
Figure 4. Core molecular clock gene expression in skeletal muscle
Changes in clock gene expression in skeletal muscle (n = 9) from pre- (blue dots and line) to post-exercise training (red dots and line). mRNA expression of Bmal1 (A), Clock (B), Reverbα (C), Rorα (D), Per1 (E), Per2 (F), Per3 (G), Cry1 (H). Data are normalized to the geometric mean of two housekeeping genes. Data are presented as means ± SD. *P < 0.05 based on Bonferroni post hoc test.

Exercise training reduced 24-h plasma glucose but did not influence 24-h plasma free fatty acids and triacylglycerol

We collected 16 blood samples over 24 h on the second test day to test for changes in circulating levels of metabolites after exercise training (n = 9). Glucose levels peaked in the postprandial periods and decreased after all meals (Fig. 5A). We detected a significant condition effect for 24-h glucose levels (P = 0.039) with slightly reduced levels after exercise training. For insulin, we only assessed the directly pre-prandial and postprandial levels for the three meals, since insulin levels even between young healthy and older men with insulin resistance did not differ at other times of the day (Wefers et al., 2020). Insulin levels did not differ from pre- to post-exercise training (condition effect: P = 0.501, interaction effect: P = 0.643; Fig. 5B). Free fatty acid (FFA) levels were high under overnight-fasted conditions in the morning and dropped after breakfast (09.00 h) and lunch (14.00 h) (Fig. 5C). During the night FFA increased again, peaking at 02.00 h in line with the fasting-induced mobilization of FFA from adipose tissue to provide energy. Of note, however, the increase in FFA at night is low when compared to young healthy volunteers (Wefers et al., 2020). Triacylglycerol levels increased over the course of the day and progressively more after each meal, and kept increasing until 23.00 h – 4 h after the last meal (Fig. 5D). Over the sleeping period (00.00 until 07.00 h), triacylglycerol levels dropped again back to fasting levels the next morning. For both FFA and triacylglycerol levels, neither a condition (FFA: P = 0.696, triacylglycerol: P = 0.962) nor an interaction effect (FFA: P = 0.967, triacylglycerol: P = 0.696) was found. For visual clarity, individual time courses of these four plasma metabolites are separately shown in Supplemental Fig. 4.

Details are in the caption following the image
Figure 5. 24-h plasma metabolites
Changes in 24-h plasma metabolites (n = 9) from pre- (blue circles and line) to post-exercise training (red circles and line). 24-h plasma levels of glucose slightly decreased after exercise training (A), whereas pre- and postprandial insulin levels (B), 24-h free fatty acids (C) and triacylglycerol (D) remained unchanged. The dark grey area represents the sleeping period (00.00–07.00 h). Data are presented as means ± SD. The respective individual data are shown in Supplemental Fig. 4.

Discussion

We previously showed that, in contrast to young healthy men, older men with insulin resistance have blunted metabolic flexibility over 24 h (i.e. reduced switch to fat oxidation during the night), display no 24-h rhythmicity in mitochondrial respiration and have altered skeletal muscle clock gene expression (Harmsen et al., 2022; Wefers et al., 2020). In the present study, we found that high-intensity interval training over 12 weeks without weight loss neither influences whole-body substrate oxidation and energy expenditure measured at five time points over 24 h nor induces diurnal variation in mitochondrial respiration. However, the skeletal muscle clock was modified by exercise training, with the period genes and Rorα having lower mRNA levels at 13.00 h, close to the time of day at which the regular exercise training was performed, and Clock mRNA levels were also altered.

To the best of our knowledge, there has only been one study which found an impact of chronic exercise training on the skeletal muscle clock in older volunteers with insulin resistance (Erickson et al., 2020). However, that study combined exercise training (e.g. cycling at 85% of maximal heart rate for 60 min for 5 days per week for 12 weeks) with a dietary intervention, and participants lost body weight and visceral fat mass, possibly underlying the increase in Bmal mRNA levels and PER2 protein abundance, as weight loss has been shown to modify the expression of other clock genes (Sardon Puig et al., 2020). Moreover, it has been shown that Bmal1 mRNA levels are acutely increased by 1.6-fold at 4 h and by 3.5-fold at 8 h after an acute endurance exercise bout in endurance-trained individuals (Popov et al., 2018). Due to the oscillatory nature of the core clock translational–transcriptional feedback loop, conclusions based on a single time point analysis should be treated with caution. Even our approach of having three time points does not allow assertions about whether exercise training induced changes in mesor, phase or amplitude of certain clock genes, which is a limitation of our study. However, interaction effects (time × intervention) could be detected, which were particularly found for the period gene family. Previously, directly comparing clock gene expression in skeletal muscle from young healthy versus older men with insulin resistance revealed phase shifts for Cry1, Per2 and Rorα, as well as absent rhythmicity for Clock and increased amplitude for Per3 in older men with insulin resistance (Harmsen et al., 2022). The present study suggests that chronic HIIT does not adjust the expression patterns of these clock genes towards the profile of young healthy men, but instead modifies the expression of Per1–3 and Rorα particularly close to the time of day at which HIIT sessions were performed. This finding underscores one of the key functions of molecular clocks, to anticipate time of day changes in environmental conditions (Bass, 2012; Hodge et al., 2015), which in the present study arise from the recurring metabolic demand of the timed exercise bout. It is also in line with the temporal specificity of HIIT to increase exercise capacity with greater improvements at the time of day at which HIIT is performed (Hill et al., 1998).

Through different clock mutant mouse models, Adamovich et al. (2021) showed that Per2 and Bmal1 are indispensable for diurnal variation of exercise capacity to occur (e.g. greater capacity during the late versus early active period). Taken together, modification of gene expression specifically of the Per family in response to a recurring timed exercise bout could underlie the temporal specificity of chronic exercise training. It has been demonstrated that muscle contractions acutely influence Per2 gene expression through a calcium-dependent pathway in mice (Kemler et al., 2020; Small et al., 2020). In humans, combining publicly available muscle transcript data from different exercise studies across different populations in a meta-analysis approach revealed that Per2 and Cry1 mRNA levels are acutely increased after both resistance and aerobic exercise (Pillon et al., 2020; Small et al., 2020). Our study expands the current knowledge on the responsiveness of the molecular clock to exercise training, previously limited to the acute perspective, by showing that also chronic exercise training modifies clock gene expression in human skeletal muscle. Particularly, the Per gene family plays a role in metabolic control (Sahar & Sassone-Corsi, 2012) and is known to respond to environmental cues providing a common mechanism to induce phase shifts of the molecular clock (Martin et al., 2023). For example, an acute exercise bout has been shown to increase mRNA levels of Per2 in both mice (Ezagouri et al., 2019; Saracino et al., 2019; Small et al., 2020) and humans (Pillon et al., 2020; Zambon et al., 2003). Subsequently increased PER2 protein levels can phase-shift the core clock mechanism by suppressing the BMAL1–CLOCK heterodimer and its positive reinforcement of the transcription–translation feedback loop. Thereby, recurring exercise bouts at the same time of day, as performed in the present study in the afternoon (14.00–18.00 h), are likely to shift the molecular clock in the long term, as suggested by the detected interaction effects in the present study. In support, Small et al. (2020) found that electric pulse stimulation of myotubes at any time point induced a phase-shift in Per2 oscillation, e.g. lowest Per2 mRNA levels over 24 h always coincided with the initial time of stimulation. In line, we also found reduced mRNA levels of Per genes post-exercise training close to the time of regularly performed HIIT sessions.

With respect to Rorα, pooling the currently available evidence in human skeletal muscle (Pillon et al., 2020) suggests that prolonged inactivity increases its mRNA levels, while HIIT training generally reduces its expression, in line with our findings. Retinoic acid receptor-related orphan receptors (ROR) have been suggested to be downstream mediators by which clock proteins regulate the expression of metabolic genes in a circadian manner (Jetten et al., 2013). The ROR agonist nobiletin has been shown to amplify the molecular clock of pancreatic islet cells from type 2 diabetes patients associated with increased insulin secretion (Petrenko et al., 2020). Moreover, nobiletin administration enhanced mitochondrial respiration in skeletal muscle and promoted healthy ageing in metabolically challenged mice (Nohara et al., 2019). In the present study, we found modification of Rorα expression close to the time of day at which exercise training was performed, and it is tempting to suggest that this may facilitate the beneficial metabolic adaptations known to occur with recurring exercise. However, more studies with a longer sampling duration and higher sampling frequency are needed to investigate the true importance of these time-specific changes in Per1–3 and Rorα after exercise training.

Although the exercise intervention applied in the present study, e.g. cycling HIIT performed three times per week for 12 weeks in the afternoon, was able to improve aerobic capacity and mitochondrial function, it was not sufficient to either improve 24-h rhythmicity in whole-body substrate oxidation or induce diurnal variation in mitochondrial respiration in older individuals with insulin resistance. Nonetheless, the possibility that these potentially favourable metabolic adaptations could be achieved by other modes of exercise should not be discarded. The following adjustments could be tested in the future: (1) the HIIT setting (e.g. daily frequency, amount, and duration of intervals), (2) the type of exercise (e.g. alternatively long-lasting high volume training (HVT) at light to moderate intensity), (3) different activity modes (e.g. resistance training, increasing every-day life physical activity by breaking sedentary time with deliberate standing, walking or stairclimbing), and (4) the timing of exercise (e.g. morning or evening). In the present study, we chose HIIT over HVT for its time efficiency, thereby increasing the feasibility for participants to combine the training with their work schedule. In addition, recent data suggested that HIIT at a relative exercise intensity of ∼90% of the maximal power output is superior over HVT to increase mitochondrial respiration in younger individuals (Granata et al., 2018). In contrast to our HIIT protocol, a more intense, glycogen-lowering approach has been shown to induce an acute drop in RER (i.e. increased fat oxidation) the next day in obese volunteers (Schrauwen et al., 1998). Furthermore, based on the aforementioned meta-analysis from Pillon et al. (2020), resistance training exerts a greater acute impact on transcription of clock genes compared to endurance training (Small et al., 2020). In mice, Reverbα even has a differential response to different types of exercise with opposite acute regulation between HIIT versus endurance and resistance exercise (da Rocha et al., 2022).

Next to the type of exercise, recent evidence also suggests that the time of day at which regular exercise is performed also matters for improving metabolic health: training in the afternoon or evening was shown to be more beneficial for glycaemic control when compared to morning training in type 2 diabetes patients (Savikj et al., 2019) and overweight and obese individuals (Mancilla et al., 2021; Moholdt et al., 2021), and similar differences have been observed in epidemiological studies (Qian et al., 2023; van der Velde et al., 2022). In the present study, HIIT sessions were only performed in the afternoon (i.e. 14.00–18.00 h) to maximize metabolic benefits and thereby induce diurnal variation in metabolic outcomes. Therefore, our study cannot conclude whether morning exercise would have resulted in different outcomes. Future studies could explore if differently timed exercise training (e.g. morning or evening) exerts a differential temporal impact on whole-body and skeletal muscle (mitochondrial) metabolism.

Next to exercise training, candidate interventions to improve 24-h whole-body substrate oxidation and/or induce diurnal variation in mitochondrial respiration are diet and drug interventions aiming for weight loss and/or improving glucose control. In this context, 2 weeks of time-restricted eating (i.e. 10 h eating period per day) in type 2 diabetes patients reduced 24-h carbohydrate oxidation but did not influence nocturnal RER or mitochondrial respiration in the morning, while achieving minor weight loss and lower 24-h blood glucose (Andriessen et al., 2022). A 2-week treatment with a sodium-glucose cotransporter 2 (SGLT2) inhibitor in volunteers with insulin resistance resulted in higher 24-h and nocturnal fat oxidation and lower 24-h blood glucose without weight loss, and increased mitochondrial respiration in the morning (Veelen et al., 2022). In type 2 diabetes patients, SGLT2 inhibitor treatment even induced an enhanced day-to-night-time difference in the RER, suggesting improved 24-h metabolic flexibility (Op den Kamp et al., 2021). These studies suggest that stronger interventions, with larger effects on blood glucose might be necessary to ameliorate 24-h whole-body substrate oxidation, particularly the stimulation of the fed-to-fasting transition. With respect to diurnal variation in mitochondrial respiration, no serial muscle sampling has been performed in these diet or drug studies, and therefore future studies are required on this matter.

As limitations of our study, it should be noted that through our in vivo design in humans we cannot separate whether exercise training modified skeletal muscle clock gene expression due to local factors within the peripheral skeletal muscle (e.g. muscle contractions, calcium signalling and/or energy status) or central factors stemming from the central SCN clock (e.g. changes in body temperature and/or hormone levels). It is important to note that in the current study we aimed to examine the effects of chronic exercise training and hence aimed to minimize confounding of the acute exercise response. Therefore, muscle biopsies were taken at least 40 h after the last exercise bout. This may explain the lack of more pronounced changes in gene expression in our study, and future studies should explore if the effects of an acute bout of exercise on clock gene expression are altered after a prolonged period of exercise training. Another limitation of our study is the limited sample size – partly due to the COVID-19 pandemic affecting participant recruitment – and therefore the statistical power of the study may have been limited for some of the exploratory outcome measures. Also, although we aimed to cover measurement over the full 24-h period, the number of time points was limited by the invasiveness of the measurements (i.e. muscle biopsies) or by technical limitations (indirect calorimetry via ventilated hood versus respiration chamber). Also, our study was restricted to males only, and studies in females are needed to reach more general conclusions. We performed a single-arm study, with volunteers being their own control pre- and post-exercise. Due to the lack of a non-exercise control group, we cannot fully exclude that some of the observed changes were due to other factors than the exercise training per se. Furthermore, it should also be highlighted that our participants were tested under conditions with typical meals and some physical activity, so that any detected 24-h rhythms or diurnal variation are rather the result of behavioural rhythms (e.g. feeding–fasting and/or activity–rest cycles) than being driven by the endogenous circadian clock.

Biography

  • image

    Jan-Frieder Harmsen is a PhD Candidate at the Department of Nutrition and Movement Science at Maastricht University. Since shift work has been associated with a greater risk for metabolic disorders, he strives to better understand the impact of the biological clock on metabolic health and its relevance in the aetiology of type 2 diabetes. In this context, he studies how the benefits of lifestyle interventions can potentially be maximized by optimizing the timing of administration. In the future, he will investigate how time cues, such as light, can be exploited at the workplace to prevent metabolic disorders in the long-term.

Data availability statement

The underlying data and analysis in this manuscript are available upon reasonable request to the authors.

Competing interests

The authors declare no conflict of interest.

Author contributions

Conception or design of the work: all authors. Acquisition, analysis or interpretation of data for the work: all authors. Drafting the work or revising it critically for important intellectual content: all authors. All authors have read and approved the final version of this manuscript and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. All persons designated as authors qualify for authorship, and all those who qualify for authorship are listed.

Funding

This study was funded through an EFSD award supported by the EFSD/Lilly European Diabetes Research Programme.

Additional information