figure b

Introduction

The switching of substrate oxidation between fat and carbohydrate in response to physiological conditions of fasting, feeding and exercise in humans has been defined as metabolic flexibility (MetFlex) [1]. Impairments in MetFlex, or metabolic inflexibility (MetInflex), have been described in multiple metabolic disorders including insulin resistance, obesity and type 2 diabetes [2]. The classical approach to assess MetFlex from fasting to insulin-stimulated conditions is during a hyperinsulinaemic–euglycaemic clamp; however, this approach does not provide a physiological context to assess substrate oxidation switching in response to nutrient availability. A more physiological characterisation of MetFlex would include an assessment of acute RQ responses to meal challenges under non-constrained assessment conditions (e.g. under a hood or mask connected to a cart). Currently, the absolute change in RQ (ΔRQ) is used to indicate the switch in substrate oxidation (from fat to carbohydrate in response to insulin), which provides only a partial characterisation of MetFlex. We propose that assessment of the dynamic responses of RQ measured continuously during meals (pre- and postprandial), exercise and sleep will provide a more comprehensive evaluation of the MetFlex phenotype. Modern respiratory exchange chambers (whole-room indirect calorimeters) can provide valid, accurate and high-resolution data [3,4,5,6] to facilitate this type of MetFlex analysis under physiological and controlled conditions [7].

Assessments of the capacity of skeletal muscle mitochondria to oxidise available substrates complement assessments of whole-body substrate oxidation and MetFlex by providing critical mechanistic evaluation of the contribution of mitochondria to whole-body energetics. Previous studies have shown that muscle mitochondrial capacity correlates with MetFlex [8,9,10], and that MetInflex and mitochondrial (dys)function are hallmark features of insulin resistance and type 2 diabetes [11]. Using whole-room indirect calorimetry (WRIC) to assess more physiologically relevant features of MetFlex, however, could establish more robust mechanistic links between muscle mitochondrial capacity and MetFlex in type 2 diabetes. Exercise training elicits numerous metabolic benefits, particularly for individuals with type 2 diabetes. In fact, 12 weeks of combined aerobic and resistance training restored MetFlex in individuals with type 2 diabetes to levels observed in healthy age- and BMI-matched individuals [8]. One caveat, however, is that MetFlex was assessed during a hyperinsulinaemic–euglycaemic clamp (i.e. under insulin-stimulated conditions), which does not represent physiological conditions.

We hypothesise that MetFlex assessed by dynamic RQ responses in WRIC will distinguish metabolic phenotypes and be associated with indices of mitochondrial capacity. The primary goal of our study was to evaluate the ability of 24 h WRIC to assess MetFlex during physiologically relevant conditions of sleeping, fasting (post-absorptive) and post-meal across a range of metabolic phenotypes (individuals with type 2 diabetes, age- and BMI-matched individuals without type 2 diabetes, and normal-weight active individuals) and in response to a 10 week supervised aerobic training intervention in individuals with type 2 diabetes. We then explored relationships among these ‘physiological’ MetFlex features by assessing indices of muscle mitochondrial capacity in vivo and ex vivo.

Methods

Study participants

Twenty-three participants with type 2 diabetes, nine healthy sedentary obese participants without type 2 diabetes (ND) and nine healthy active participants were included in our cross-sectional analysis (Table 1 and electronic supplemental material [ESM] Fig. 1). The type 2 diabetes group underwent an aerobic training intervention (17 completers). The study protocol was reviewed and approved by the Institutional Review Board of AdventHealth (Orlando, USA). All participants provided a written informed consent (ClinicalTrials.gov registration no. NCT01911104).

Table 1 Physical characteristics and EE phenotypes in WRIC and free-living conditions
Full size table

Study design

After a 2 week period of medication washout, we performed baseline assessments to compare MetFlex and energy expenditure (EE) variables in the three groups of participants (type 2 diabetes, ND, active) ranging in body composition, physical fitness and insulin sensitivity. Then, the type 2 diabetes group completed a 10 week supervised aerobic training programme that consisted of walking on a treadmill (4 days/week, 45 min per session at a heart rate equivalent to ~75% of the maximal volume of O2 consumed [\( \dot{V}{\mathrm{O}}_{2\max } \)]) and repeated the same assessments on the 11th week at least 48 h after the last training session. A more detailed explanation of the exercise intervention is published elsewhere [12] (see also ESM Fig. 2). Multiple features of MetFlex and EE were assessed by two methods of indirect calorimetry (WRIC and metabolic cart). Substrate oxidation switching was assessed by acute changes in RQ. Individual rates of substrate oxidation (fat oxidation rate [FatOx]; carbohydrate oxidation rate [CarbOx]; protein oxidation rate [ProtOx]) in response to glucose and insulin infusions (during hyperinsulinaemic–euglycaemic clamp) and meal loads and during the sleeping period were also evaluated. Together, these latter variables represented MetFlex features (RQ and macronutrient oxidation responses). The assessments occurred over three visits (ESM Fig. 3). At visit 1, in vivo mitochondrial capacity by 31P-magnetic resonance spectroscopy and \( \dot{V}{\mathrm{O}}_{2\max } \) were measured. At visit 2, participants followed a standardised diet (35% fat, 55% carbohydrate and 15% protein; food quotient [FQ] = 0.872) for 48 h prior to admission to the Clinical Research Unit at the Translational Research Institute (TRI-CRU). All meals were prepared and provided by the TRI Metabolic Kitchen. Participants arrived at the research unit the evening before the WRIC assessment, consumed dinner and remained overnight. The next morning, participants entered the whole-room indirect calorimeter for 24 h. At visit 3, participants underwent a single-step hyperinsulinaemic–euglycaemic clamp and resting metabolic rate (RMR) under insulin-stimulated conditions was measured. Blood and urine were collected and a muscle biopsy of the vastus lateralis in fasting conditions was performed. The participants with type 2 diabetes who took part in the 10 week exercise intervention repeated visits 1, 2 and 3 after the intervention. The type 2 diabetes group underwent a washout of glucose-lowering medications 2 weeks prior to the first assessment period and throughout the intervention period and follow-up assessments.

EE and MetFlex assessments

All assessments were performed using indirect calorimetry and carried out under two different physiological conditions.

Twenty-four hour WRIC under normal physiological conditions

During visit 2, whole-room indirect calorimeter was used to assess MetFlex features under normal physiological conditions. The volume of O2 consumed (\( \dot{V}{\mathrm{O}}_2 \)) and the volume of CO2 produced (\( \dot{V}\mathrm{C}{\mathrm{O}}_2 \)) in the chambers were measured using gas analysers (Ultramat/Oxymat 6; SIEMENS, USA). All participants followed a scripted set of activities to assess resting, fasting, meal-related and free-living EE and RQs (ESM Fig. 3). All participants were in energy balance (EB) during the WRIC assessment. Details of the whole-room indirect calorimeter and all assessments, data collection and equations for the WRIC are provided in ESM Methods.

MetFlex assessed under insulin-stimulated conditions

This assessment was performed before and during the steady state of a 2 h hyperinsulinaemic–euglycaemic clamp assessment [12]. Briefly, a primed infusion of insulin (694.4 pmol min−1 m−2) was initially administered with an average 1215 ± 326 pmol/l maintained during the clamp. Plasma glucose was maintained at ~5.4 mmol/l in each participant with a variable infusion of exogenous glucose (20% solution). Before and during the steady state of the clamp, we measured the \( \dot{V}\mathrm{C}{\mathrm{O}}_2 \) and \( \dot{V}{\mathrm{O}}_2 \) using a MAX II Metabolic Cart (AEI Technologies, USA). Urine was collected between baseline and steady state to calculate ProtOx.

MetFlex calculations

For both MetFlex assessments, RQ was calculated as the ratio between \( \dot{V}\mathrm{C}{\mathrm{O}}_2 \) and \( \dot{V}{\mathrm{O}}_2 \). When appropriate, the non-protein RQ (npRQ) was reported following the equation suggested by Jequier et al [13].

ΔRQ and substrate oxidation under insulin-stimulated conditions

The classical measurement of MetFlex was performed as described by Kelley and Mandarino [1]. RQ data were collected during the 30 min before the hyperinsulinaemic–euglycaemic clamp (RQpreclamp) and during the last 30 min of the clamp at steady state (RQinsulin); ΔRQ was the difference between RQpreclamp and RQinsulin. CarbOx and FatOx during clamp (CarbOxclamp and FatOxclamp; respectively) were calculated using Frayn’s Equations [14] and ProtOx by Magnus-Levy’s constant [15].

Post-breakfast RQ kinetics, ΔRQ and substrate oxidation

MetFlex was assessed using two different approaches. First, RQ kinetic response after ingesting a high-carbohydrate breakfast (~30% of total daily energy intake; 8.95% protein, 22.03% fat and 69.01% carbohydrate; breakfast FQ 0.919) was assessed. RMR assessments 30 min pre-breakfast, during the 30 min breakfast consumption period and 60 min post-breakfast were included in our analyses. Slopes between the lowest and highest RQ points were analysed by a simple linear regression between time (1 min resolution) and RQ units (RQunits) and slopes and intercepts were compared between groups (ESM Fig. 4a). AUCs were calculated for the entire time frame and individual time points. Second, the change in RQ after ingesting a standard high-carbohydrate breakfast (RQBF) was calculated as the difference between the pre-breakfast RQ and the first 30 min post-breakfast RQ.

Sleep RQ kinetics and oxidation rates

A 5 h assessment period (12:00 hours to 05:00 hours) was utilised for this analysis. Kinetic analysis for sleep was carried out in the same manner as for post-breakfast analysis.

Skeletal muscle mitochondrial capacity assessments

In vivo muscle mitochondrial function

Phosphorus (31P) magnetic resonance spectroscopy (Philips Healthcare, USA) was utilised to measure phosphocreatine (PCr) recovery rate (κ values) in the vastus lateralis as an indicator of mitochondrial capacity in vivo [12].

Skeletal muscle biopsy and measurement of mitochondrial respiration

Percutaneous muscle biopsies were obtained from the vastus lateralis using the Bergstrom technique [16]. Samples were collected after participants had fasted overnight and 72 h after the last training session. Mitochondrial respiration was assessed in fresh saponin-permeabilised myofibres using high-resolution respirometry (Oroboros Instruments, Austria). Measurements were performed at 37°C using two substrate combination protocols (carbohydrates only and carbohydrates plus fatty acids). Additional methodological details are provided in ESM Methods.

Objectively measured free-living physical activity and cardiorespiratory fitness

Total daily free-living physical activity (steps/day) and total daily free-living EE (TDEEFL, kJ/day) variables were measured with a tri-axial accelerometer/temperature sensor (SenseWear Armband, USA) [17]. Physical activity level (PALFL) was calculated as TDEEFL/RMR.

\( \dot{V}{\mathrm{O}}_{2\max } \) was measured during a progressive exercise test on a treadmill following the guidelines of the American College of Sports Medicine (ACSM) (see ESM Methods for further details). A metabolic cart was used to measure breath-by-breath \( \dot{V}{\mathrm{O}}_2 \) and \( \dot{V}{\mathrm{CO}}_2 \) (TrueOne 2400; PARVO Medics, USA).

Body composition assessment

Fat-free mass (FFM, kg) and fat mass percentage were quantified by dual energy x-ray absorptiometry (Lunar iDXA; GE Healthcare, USA). Whole-body and regional body composition were quantified following the protocol of the manufacturer.

Blood analyses

Fasting blood samples were collected for comprehensive metabolic panel, insulin, NEFA and HbA1c. In addition, blood was collected during the clamp procedure to measure insulin, glucose and NEFA and to calculate insulin sensitivity (M value) [12]; urine samples collected during the clamp were used for measurement of urinary nitrogen and glucose.

Statistical analysis

All variables are reported as mean ± SD. One-way ANOVA was utilised to compare variables among the three groups of interest, and differences between groups were confirmed by Tukey’s post hoc tests for multiple comparisons. For EE variables (kJ/day) and substrate oxidation rates (g/min) the comparisons were adjusted to FFM. RQ, oxidation rates or EE differences between resting and post-absorptive conditions (insulin-stimulated or post-breakfast) were analysed by paired sample t test for all groups. MetFlex phenotypes were compared by ANCOVA adjusted to basal measurement (before clamp or before breakfast).

Data for each participant from the RQ kinetic analysis were time-aligned and averaged for each group (active, ND, type 2 diabetes) and assessment (pre- and post-intervention for type 2 diabetes). Regression slopes and intercepts were compared among groups and pre- vs post-intervention for type 2 diabetes by using a multiple regression model that contains a condition variable (group or intervention [18]). Comparisons were adjusted to sex and age where appropriate. Additionally, we used multivariate linear regression analysis to explore associations among MetFlex and EE variables and mitochondrial function. Two statistical packages were utilised to perform the analysis (JMP 13.2.1 [SAS Institute, USA] and Prism 6.07 [GraphPad Software, USA]).

Results

Baseline characteristics, body composition and EE are presented in Table 1 and ESM Tables 1 and 2. Fasting glucose at the screening visit was significantly different from the pre-intervention time point in the type 2 diabetes group due to the 2 week medication washout (Glucosescreen vs Glucose [Table 1] mean difference 2.75 ± 0.65 mmol/l, p < 0.001).

EB inside the whole-room indirect calorimeter was similar across all three groups (p > 0.05) and did not differ significantly from zero (mean EB 13.3 ± 348.5 kJ/day, p > 0.05), which permitted a valid assessment of the EE and substrate oxidation variables (without adaptive thermogenic responses and without adjustments of RQ for EB). The FQ before and during the WRIC stay was similar to the 24 h RQ (24 h FQ 0.872 vs 24 h RQ 0.875 ± 0.030; p > 0.05), ensuring that differences in substrate oxidation switching rates among the groups were not due to adaptation to the standardised diet given in the calorimeter.

Substrate oxidation data are presented in Table 2. All RQ assessments were significantly higher in the ND group compared with the active group and the type 2 diabetes group, except for RQRMR. Substrate oxidation rates expressed in g/h and adjusted to FFM confirmed that both the type 2 diabetes group and the ND group had a higher sleeping CarbOx and a higher resting CarbOx when compared with the active group.

Table 2 RQs and substrate oxidation rates during a 24 h period inside a whole-room indirect calorimeter
Full size table

Insulin-stimulated MetFlex

The ND and active groups displayed significantly increased RQ under insulin-stimulated conditions (ΔRQclamp) by 12.8 ± 2.1% and 12.2 ± 2.0%, respectively, while the type 2 diabetes group did not show increased RQ (Fig. 1a). The switch in substrate oxidation under insulin stimulation in the active group was accompanied by a significant increase in RMR, which was not observed in the ND group nor the type 2 diabetes group (Fig. 1b).

Fig. 1
figure 1

MetFlex assessed under insulin-stimulated conditions during a hyperinsulinaemic–euglycaemic clamp. The following variables were compared between participants in the type 2 diabetes (T2D), ND and active groups: RQ (a); RMR (b); \( \dot{V}{\mathrm{O}}_2 \) (c); \( \dot{V}{\mathrm{CO}}_2 \) (d); Δ resting RQclamp (e); ΔRMRclamp (f); Δ resting \( \dot{V}{\mathrm{O}}_2 \)clamp (g); and Δ resting \( \dot{V}{\mathrm{CO}}_2 \)clamp (h). Data in line graphs are least square means adjusted to FFM (except for RQ). Bars graphs show differences in percentage between insulin-stimulated and basal variables represented in each line graph (e.g. Δ resting RQ = [RQinsulin – RQbasal]/RQbasal × 100). **p<0.01 vs active group and ††p<0.01 vs ND group (Tukey’s multiple comparisons test for significant differences)

Full size image

Post-breakfast and sleeping MetFlex

For the breakfast RQ kinetic analysis, the typical shape of the kinetic response was a drop in the RQ at the beginning of the breakfast followed by a steep increase over about 30 min. This shape was similar across the three groups (ESM Fig. 4b). Breakfast energy intake (active 2770 ± 364 kJ, ND 2569 ± 544 kJ, type 2 diabetes 2778 ± 414 kJ; p > 0.05) and pre-breakfast RQ were not significantly different among the groups (active 0.859 ± 0.028, ND 0.879 ± 0.027, type 2 diabetes 0.869 ± 0.027, p > 0.05). However, the active group had a significantly lower intercept for RQBF (p < 0.001; Fig. 2a) and faster RQBF kinetics (i.e. steeper slope) compared with the type 2 diabetes and ND groups (p < 0.001; Fig. 2a). In addition, the active group had a smaller RQBF AUC adjusted to total movement assessed by radar (counts/min) during breakfast than the type 2 diabetes group (active 79.9 ± 2.2, ND 81.7 ± 2.2, type 2 diabetes 82.3 ± 2.5; p < 0.05), but without differences in peak RQBF or ΔRQBF. Altogether, post-breakfast results suggested that the type 2 diabetes and ND groups had a lower AUC and slower kinetics capability for CarbOx after a meal test, which is indicative of impaired MetFlex (i.e. MetInflex) under physiological conditions.

Fig. 2
figure 2

MetFlex assessed by RQ kinetics analysis inside a whole-room indirect calorimeter. (a) Average RQ kinetic responses of the three groups from the lowest to the highest RQ during the breakfast period (vertical dashed lines mark the start and the end of breakfast; before and after this time participants were supine). Slopes and intercepts are the constants of simple regression equations between RQ and time (min). Slopes: ND 0.002655 RQunits/min, type 2 diabetes 0.003323 RQunits/min, active 0.004478 RQunits/min; F=57.857, p<0.0001. Intercepts: ND 0.7365, type 2 diabetes 0.7050, active 0.6225; p<0.0001. (b) Average RQsleep responses of the three groups between 12:00 hours and 05:00 hours. Slopes: ND −3.891×10−5 RQunits/min, type 2 diabetes −3.573×10−5 RQunits/min, active 1.840×10−5 RQunits/min; F=22.2, p<0.001. Intercepts: ND 0.866, type 2 diabetes 0.858, active 0.822; p<0.001. F values are effect sizes testing parallelism between regression lines. The key in (a) also applies to (b). BF, breakfast starts; End, breakfast finishes; T2D, participants with type 2 diabetes

Full size image

The active group had a positive slope for sleep RQ (RQsleep); conversely, the ND and type 2 diabetes groups exhibited negative RQsleep kinetics, such that their RQsleep slopes were almost twice as steep as those in the active group (p < 0.001; Fig. 2b). The active group also had a lower RQsleep intercept compared with the other two groups (p < 0.001; Fig. 2b). Interestingly, the RQs almost converged throughout the sleeping period, whereby the three groups had closer RQs at the end of the sleeping period (04:30 hours to 05:00 hours) than at the start of the sleeping period (00:00 hours to 00:30 hours); however, the RQsleep remained significantly different among the ND and type 2 diabetes groups compared with the active group (active 0.824 ± 0.012, ND 0.843 ± 0.015, type 2 diabetes 0.851 ± 0.01; p < 0.001; Fig. 2b). In addition, the RQsleep of the active group was the closest to the dinner FQ (RQsleep 0.822 vs dinner FQ 0.834), which suggests an enhanced ability to oxidise the carbohydrates from the dinner meal compared with the ND and type 2 diabetes groups. Taken together, these results indicate a defective switching from carbohydrate to fat oxidation under post-absorptive conditions (e.g. during sleep) in obese individuals with and without type 2 diabetes. Mean 24 h RQ responses are shown in ESM Fig. 4c.

Associations between MetFlex and muscle mitochondrial capacity

In vivo muscle mitochondrial capacity (κ value) was highest in the active group (active 0.0405 ± 0.0102, ND 0.0234 ± 0.0103, type 2 diabetes 0.0205 ± 0.008; p < 0.001). The active group also exhibited significantly higher ex vivo muscle mitochondrial capacity than the type 2 diabetes and ND groups (Fig. 3a). In vivo and ex vivo mitochondrial capacity were correlated (PRE κ value&PI + II, R2 = 0.60, p < 0.001; κ value&PI + IIFAO, R2 = 0.55, p < 0.001; κ value&maximal carbohydrate-supported electron transfer system [ETS], R2 = 0.54, p < 0.001). Furthermore, impaired sleeping fat oxidation (i.e. higher RQ) was negatively correlated with ex vivo muscle mitochondrial respiratory capacities (Fig. 3b–d). RQsleep kinetics, in which a positive slope indicates higher sleeping fat oxidation, were positively correlated with in vivo (PCr recovery rate) and ex vivo muscle mitochondrial capacities (ESM Figs 5a–d, 6a–f).

Fig. 3
figure 3

Ex vivo muscle mitochondrial capacity and MetFlex. (a) Differences in ex vivo muscle mitochondrial respiratory states among active (normal-weight), ND (non-diabetic with obesity) and type 2 diabetes (T2D) groups. (b–d) Correlations between RQsleep and selected mitochondrial respiratory states (PI+II, R2=0.10, p<0.05; PI+IIFAO, R2=0.196, p<0.01; ETS, R2=0.237, p<0.01). Mitochondrial respiration consists of leak (LI), complex I-supported OXPHOS (PI), complex I+II-supported OXPHOS (PI+II) and maximal ETS, with carbohydrates as substrates. FAO indicates respiration supported by palmitoyl carnitine. *p<0.05 for type 2 diabetes or ND groups vs active group (Tukey’s test)

Full size image

Intervention effects

Seventeen participants with type 2 diabetes completed the aerobic training intervention with no significant changes in fasting blood glucose levels (pre- vs post-intervention 10.64 ± 3.02 vs 10.93 ± 3.58 mmol/l; p > 0.05) or HbA1c (pre- vs post-intervention 52.5 ± 12.3 mmol/mol [6.95 ± 1.13%] vs 55.1 ± 15.0 mmol/mol [7.19 ± 1.39%], p > 0.05) (ESM Table 3). \( \dot{V}{\mathrm{O}}_{2\max } \) significantly increased by 11.2 ± 7.2% (ESM Table 3). Mean RQ (24 h, RMR, sleep), substrate oxidation rates (ESM Table 4) and total daily EE (ESM Table 3) were not significantly affected by the training intervention.

MetFlex changes following the aerobic training intervention

We did not observe any significant changes in ΔRQclamp (pre- vs post-intervention 0.044 ± 0.066 vs 0.038 ± 0.047; p > 0.05) or ΔnpRQclamp (pre- vs post-intervention 0.055 ± 0.086 vs 0.046 ± 0.059; p > 0.05) under insulin-stimulated conditions following training.

The post-breakfast RQ slope was significantly increased in the type 2 diabetes group (p = 0.013; Fig. 4a) following the exercise intervention, as was the RQsleep intercept (p < 0.0001; Fig. 4b). Both adaptations suggest an improvement in CarbOx in response to aerobic training. In support, carbohydrate-supported ETS O2 flux ex vivo also increased, indicating enhanced maximal muscle mitochondrial respiratory capacity following aerobic training (pre- vs post-intervention maximal ETS capacity 87.7 vs 105.6 pmol s−1 mg−1, p < 0.01; Fig. 4c). Training-induced changes in ex vivo muscle mitochondrial respiratory capacity supported by carbohydrate-derived substrates were not significantly correlated with either the changes in RQBF kinetics or with AUC RQsleep (Fig. 4d,e).

Fig. 4
figure 4

Responses in MetFlex and skeletal muscle mitochondrial energetics ex vivo following a 10 week aerobic training programme in individual with type 2 diabetes. (a, b) RQ kinetics (slopes and intercepts) during breakfast (a) and sleep (b) (n=15). Solid and dashed lines are mean simple regression lines between RQ values and time for pre- and post-intervention data, respectively (breakfast pre- vs post-intervention slope 0.003323 vs 0.003944 RQunits/min, F=6.47, p=0.013; breakfast pre- vs post-intervention intercept 0.690 vs 0.675, p<0.0001; sleep pre- vs post-intervention slope −9.459×10−6 vs −3.391×10−6 RQunits/min, F=0.56, p=0.457; sleep pre- vs post-intervention intercept 0.848 vs 0.857, p<0.0001). (c) Changes in ex vivo muscle mitochondrial respiratory capacity (complex I+II-supported OXPHOS [PI+II]] and ETS). (d, e) Correlations between pre- to post-intervention changes in ex vivo muscle mitochondrial capacity and metabolic flexibility. Mitochondrial respiration consisting of PI+II (R2=0.19, p=0.08) and maximal ETS capacity (R2=0.18, p=0.09) with carbohydrates as substrates. *p<0.05 for pre- vs post-intervention. PRE, pre-intervention; POST, post-intervention

Full size image

Discussion

The overall goal of this study was to comprehensively assess MetFlex and muscle mitochondrial capacity in individuals with type 2 diabetes within a physiological framework of fasting, feeding, sleep and exercise. Specifically, we used a high-carbohydrate meal challenge and overnight substrate oxidation as physiological paradigms to illustrate the concept of substrate switching (e.g. from fat to carbohydrates). The novel contribution of our methodology to the concept of MetFlex in type 2 diabetes was the analysis of RQ data as a continuous variable in addition to, and compared with, the ΔRQ (either under insulin-stimulated conditions or the physiological conditions within the whole-room indirect calorimeter). By using this approach, we uncover two key findings in individuals with type 2 diabetes: (1) a slow transition from fat to carbohydrate oxidation after a meal, which is in support of previous findings under hyperinsulinaemic conditions of a clamp; and (2) a greater reliance on carbohydrate (vs fat) oxidation during sleep that is further enhanced with a training intervention and is a novel and somewhat contrary finding to previous reports of training-induced enhancements in FatOx.

The new features of MetFlex described in this paper provide a comprehensive assessment of substrate competition in type 2 diabetes by elucidating oxidation responses under non-restrictive and non-invasive conditions in the whole-room indirect calorimeter. Ours is the first study to assess dynamic aspects of substrate oxidation in response to the physiological challenges of fasting, feeding and sleeping in individuals across a spectrum of metabolic health. Specifically, kinetic analysis of RQ at breakfast derived from WRIC assesses the body’s natural ability to metabolise a high-carbohydrate meal following an overnight fast without the artificial constraints of insulin infusion (e.g. during a hyperinsulinaemic–euglycaemic clamp). One of our initial findings under these physiological conditions is that we do not observe the differences in MetFlex (ΔRQBF) between obese individuals without (ND) and with type 2 diabetes and normal-weight active (active) individuals that have been previously reported under insulin-stimulated conditions [1, 19]. We do, however, observe a faster RQ response (from the lowest to the highest RQBF) and a lower RQBF AUC in the active group compared with the type 2 diabetes and ND groups. Collectively, these results demonstrate that the dynamic responses in substrate oxidation to a high-carbohydrate meal are impaired in individuals with type 2 diabetes compared with healthy active individuals. We initially described our RQ kinetic analysis method in 2016 [20]. Since then, others have utilised the ‘after-meal-slope’ approach to describe MetFlex differences in animal studies [21], RQ responses to dietary and training interventions in humans [3] and daily RQ variability [7]. Our study validates and extends the kinetic analysis approach from previous studies [3, 7, 21] by comparing three different metabolic phenotypes (normal-weight active, obese without type 2 diabetes, obese with type 2 diabetes). This highlights the potential of WRIC data to provide a comprehensive and dynamic view of oxidative metabolism in response to meal tests across a spectrum of metabolic health.

In line with our results from the post-breakfast analysis, our RQsleep kinetic analysis confirms the inability to modulate substrate switching in individuals with obesity and type 2 diabetes, which has been described in limb balance studies performed with tracers in catheterised muscle [19]. The RQsleep kinetics showed that the active individuals had higher FatOx than the ND and type 2 diabetes groups. A deeper analysis of these data indicated that the active group showed faster oxidation of the dinner meal than the ND and type 2 diabetes groups since the RQ at 12:00 hours was closer to the dinner FQ than in the other two groups. This suggests an enhanced MetFlex in the active individuals who may reap metabolic benefits from a larger overnight period of high FatOx compared with obese individuals with and without type 2 diabetes. The positive RQsleep slope in the active group is likely a consequence of a lower RQ at the beginning of the sleeping period and a greater ability to oxidise the substrates provided by the dinner meal (dinner FQ was higher than RQ). The negative RQsleep slopes observed in the type 2 diabetes and ND groups are indicative of the opposite phenomenon. Additionally, the higher RQsleep in the type 2 diabetes group at the beginning of the sleeping period and throughout the night may be reflective of a preference to oxidise carbohydrate in the context of high glucose availability. This mechanism was recently confirmed in individuals with normal glucose tolerance who exhibited high glucose oxidation rates under experimentally induced systemic hyperglycaemia [22]. In our study, the removal of glucose-lowering medications 2 weeks prior to baseline assessments and throughout the training intervention resulted in ambient hyperglycaemia in the type 2 diabetes group. The aerobic training intervention significantly improved the RQBF slope, which became steeper compared with pre-training slope. This adaptation indicates an improved capacity to acutely and more quickly metabolise large amounts of carbohydrates consumed at breakfast. This finding is in agreement with previous reports of a positive effect of training on features of MetFlex, such as ΔRQ under insulin-stimulated conditions [8, 23, 24] and in vitro skeletal muscle palmitate oxidation [25] in individuals with type 2 diabetes and insulin resistance. While we observed a significant improvement in MetFlex after training, this was only evident when assessed by the meal challenge and not under the insulin-stimulated conditions of the hyperinsulinaemic–euglycaemic clamp. The route of glucose administration influences the substrate metabolism [26]; therefore, it is possible that this discrepancy may be partially explained by distinct exercise-induced adaptations in oral (breakfast) vs parenteral (clamp) routes of glucose metabolism. Other factors to be considered are the ambient hyperglycaemic conditions as a result of the medication washout and the overall degree of insulin resistance among the participants. Additionally, the change in RQsleep intercept following training indicated an enhancement of carbohydrate oxidation overnight. While initially viewed as contrary to the anticipated training-induced increase in fat oxidation during sleep, in the context of chronic ambient hyperglycaemia induced by medication washout in this study, these findings suggest an improved ability to oxidise substrate based on its availability (i.e. high glucose availability).

These novel aspects of the MetFlex assessment under physiological conditions were significantly perturbed in the individuals with type 2 diabetes, possibly conferring a strong construct validity to these new variables (RQ slopes, AUC, RQsleep slopes) to describe MetFlex in metabolic disease states. In addition, several variables of these WRIC continuous data were correlated with features of muscle mitochondrial capacity. Impaired MetFlex under insulin-stimulated conditions has been previously correlated with defective muscle mitochondrial capacity [6, 8,9,10]. Our new assessments of MetFlex described herein confirm these previous findings [6] since a higher RQsleep was negatively associated with oxidative phosphorylation (OXPHOS) and ETS capacity; however, the mechanistic relationship between the dynamic RQ variables and muscle mitochondrial capacity will need to be resolved in future studies.

Our method described herein for analysing continuous data from WRIC assessments complements data obtained under insulin-stimulated conditions as it integrates new quantitative measures (slopes, Δ values, AUCs) under physiological conditions. Additionally, using slopes to analyse substrate oxidation switching enables statistical analysis of individual results. For example, we cannot determine whether an individual ΔRQ is significantly different following an intervention (i.e. there are no statistical approaches that allow evaluation of differences from a single number); however, we can compare two regression lines for a single person to determine whether the slopes and intercepts of the function differ significantly [27].

Strengths and limitations

A critical strength of our study is the highly controlled conditions inside the WRIC with tight EB and a 48 h diet standardisation prior to the in-chamber assessments [5]. Both strategies removed the impact of antecedent diet on RQ, meaning that our results are not related to dietary adaptation or energy imbalance [28]. The advanced technological capabilities of our push–pull system in the whole-room indirect calorimeter permitted analysis of dynamic data with high resolution and rapid responses.

Certain aspects of our study could have been improved. Some of the group differences and correlations between muscle mitochondrial capacity and MetFlex variables were not statistically significant and would have benefited from a larger sample size to confirm the observed trends. Metformin could affect the results from the training intervention, so the results would have been strengthened by the inclusion of a control group and a training group that did not cease medication use. It is important to consider that the active group of participants included only men, and even though we adjusted for sex in our analyses, this could have partially influenced our results [29].

Conclusions

Through employing 24 h WRIC, our study revealed differences in the dynamic responses of substrate switching during meal challenges and sleep, further defining significant MetInflex in type 2 diabetes. Our results show that RQ kinetic analyses are valid tools to more comprehensively investigate impaired MetFlex, account for individual dynamic responses, and correlate with metabolic disease states and muscle-specific mitochondrial capacity. The training intervention revealed that in individuals with type 2 diabetes, there was significant improvement in the dynamic aspects of MetFlex after a meal challenge. Collectively, our novel analyses and results emphasise the importance of analysing continuous RQ data vs single time points and highlights the capabilities of the WRIC to perform complete whole-body studies of human metabolism.