Energy expenditure due to gluconeogenesis in pathological conditions of insulin resistance

Abstract

Gluconeogenesis (GNG), the formation of glucose from noncarbohydrate precursors, requires adenosine triphosphate (ATP). Previous studies have estimated the energetic cost of GNG in humans based on theoretical calculations of rates of GNG, moles of oxygen consumption by GNG, and average oxygen consumption. Few human studies have measured the energy expenditure (EE) due to GNG. We estimated EE attributable to GNG in patients with three insulin resistance conditions and high GNG rates (insulin receptor pathogenic variants, lipodystrophy, and type 2 diabetes) and obesity without diabetes. Fractional GNG was measured by incorporation of deuterium from body water into newly formed glucose, endogenous glucose production (EGP) as glucose appearance following administration of [6,6-²H₂]glucose, and total GNG as fractional GNG × EGP. EE was measured by indirect calorimetry and compared with predicted EE from the Mifflin St. Jeor equation. EE attributable to GNG was estimated using linear regression after accounting for age and fat-free mass (FFM). EE in patients with insulin resistance was significantly higher than predicted by the Mifflin St. Jeor equation. GNG correlated with resting EE (REE). EE attributable to GNG in patients with insulin resistance was almost one-third of REE, substantially higher than theorized in healthy subjects. Our findings demonstrate that GNG is a significant contributor to EE in insulin-resistant states. Prediction equations may underestimate caloric needs in patients with insulin resistance. Therefore, targeting caloric needs to account for higher EE due to increased GNG should be considered in energy balance studies in patients with insulin resistance.

NEW & NOTEWORTHY Gluconeogenesis is an energy-requiring process that is upregulated in diabetes, contributing to hyperglycemia. Previous studies have estimated that gluconeogenesis accounts for less than 10% of resting energy expenditure. This study estimates the energy expenditure attributable to gluconeogenesis in uncommon and severe forms of insulin resistance and common, milder forms of insulin resistance. In these populations, gluconeogenesis accounts for almost one-third of resting energy expenditure, substantially higher than previously theorized in the literature.

INTRODUCTION

Gluconeogenesis (GNG), the formation of glucose from noncarbohydrate precursors, requires adenosine triphosphate (ATP) (1). In patients with insulin resistance, insulin-mediated suppression of hepatic glucose production is impaired, and GNG rates are high, contributing to hyperglycemia (2, 3). Because GNG is energetically costly, high GNG may contribute to increased energy expenditure (EE). Few studies have directly measured the EE attributable to GNG in humans (4), although the energy cost of GNG has been estimated based on knowledge of rates of GNG, moles of oxygen consumption by GNG, and average oxygen consumption (1). Furthermore, equations used to predict EE may over or underestimate EE in patients with diabetes (5, 6). Therefore, we estimated the EE attributable to GNG in three pathological conditions of insulin resistance in humans [insulin receptor pathogenic variants (INSR), lipodystrophy, and type 2 diabetes (T2DM)] and obesity without diabetes and compared this with predicted GNG EE using previously published rates of GNG in healthy individuals and the Mifflin St. Jeor equation (7). We hypothesized that EE in patients with high GNG rates would be greater than that predicted in healthy subjects due to excess EE from GNG.

MATERIALS AND METHODS

The Institutional Review Boards of the National Institutes of Health (lipodystrophy, INSR, and T2DM) and Baylor College of Medicine (T2DM and obesity without diabetes) approved all the studies. Patients or legal guardians provided written informed consent before participation; patients < 18 yr provided written assent.

Study Population

This was a cross-sectional analysis of four populations. Patients aged 14–64 yr with lipodystrophy (n = 25) who were taking metreleptin (n = 8) or naïve to metreleptin (n = 17) were studied between 2013 and 2018 for a study of metreleptin independent of food intake (NCT01778556) (7). Patients with INSR aged 15–30 yr (n = 7) were studied in 2016 for a study of the effects of liothyronine on glucose metabolism (NCT02457897) (8). The Baylor cohort comprised adolescents aged 10–17 yr with newly diagnosed T2DM (n = 9) never exposed to glucose- or lipid-lowering medications, studied between 2010 and 2012 at Texas Children’s Hospital (9). Patients with obesity were matched to those with T2DM for age and body mass index (BMI); diabetes was excluded by oral glucose tolerance test. The NIH cohort comprised adolescents aged 14–17 yr within 5 yr of T2DM diagnosis, in whom all glucose-lowering therapy was discontinued 5–7 days before study enrollment. Only patients for whom stable isotope tracers for measurements of GNG were available at baseline were included in this analysis.

Measurement of Gluconeogenesis

Following an overnight fast, [6,6-²H₂]glucose (Cambridge Isotope Laboratories) was used to measure glucose rate of appearance (R_a) using the isotope dilution method using the single pool model (10). A primed, continuous infusion of [6,6-²H₂]glucose was given for 3 h in patients with INSR and lipodystrophy and for 5 h in patients with T2DM and obesity. After equilibration periods, blood samples for isotope enrichment were obtained every 10 min for 30 min at steady state. Enrichment of [6,6-²H₂]glucose was measured using liquid chromatography mass spectrometry in patients with lipodystrophy, INSR, and T2DM-NIH (7) and gas chromatography mass spectrometry in patients with T2DM-Baylor and obesity (9). Fractional GNG (the fraction of glucose derived from GNG) was measured after oral administration of deuterated water (11). Briefly, fractional GNG was calculated based on the average enrichment of deuterium on carbons 1, 3, 4, 5, and 6 and deuterium enrichment in body water. Absolute GNG was calculated as the product of fractional GNG and glucose R_a. Rates of GNG in lean nondiabetic adults (6.7 µmol/kgFFM/min) were obtained from the medical literature (6).

Measurement of Body Composition

Fat mass (FM) and fat-free mass (FFM) were measured by dual energy X-ray absorptiometry (DXA) (7–9, 12) (lipodystrophy and INSR, iDXA, GE Healthcare, Madison WI; T2DM and obesity, QDR 11.2; Hologic Bedford, MA).

Measurement of Energy Expenditure

Resting energy expenditure (REE) was measured in INSR by whole-room indirect calorimetry, and in lipodystrophy, T2DM and obesity without diabetes by ventilated hood connected to a metabolic cart (7–9, 12–14) (ParvoMedics TrueOne2400, Sandy, UT; Deltatrac II Sensormedics, Yorba Linda, CA). Gas analyzers for room calorimeters were calibrated weekly; between analyzer differences were corrected before each study. Monthly propane combustion or gas infusion tests demonstrated no significant differences in accuracy of calorimeters (<1.5% for V̇o₂ and V̇co₂) (13). Calibration of ParvoMedics TrueOne carts was performed before each study and preventative maintenance quarterly. Little intrasubject difference (<3%) was found between room calorimeters and ParvoMedics TrueOne carts (15). ParvoMedics and Deltatrac showed high accuracy versus methanol combustion (16) without differences on intrasubject comparisons (−6 ± 131 kcal/day) (17). REE measurements were performed simultaneously with GNG measurements in most patients (up to 2 days difference in INSR and lipodystrophy).

Statistics and Calculations

Outcomes are reported as means ± SD for normally distributed data and median (25th, 75th percentile) for skewed data. For comparisons between groups, analysis of variance, Wilcoxon rank-sum test, or chi-squared test was used as appropriate. Post hoc pairwise comparisons of means were performed using Tukey’s test. Pearson’s correlations were used to test correlations among metabolic characteristics. P < 0.05 represented statistical significance. All P values are two sided. This was an exploratory analysis of previously collected data, thus no sample size calculations were performed. Analyses were conducted using GraphPad Prism, Version 9.0 (GraphPad Software) and Stata, IC version 16.0 (Stata Corp LP, College Station, TX).

Measured REE (mREE, kcal/day) was determined by indirect calorimetry. Predicted REE (pREE, kcal/day) was calculated using the Mifflin St. Jeor equation (7). The difference between mREE and pREE is shown in Eq. 1:

$$\text{Measured REE }\left(\text{mREE}\right)-\text{ Predicted REE }(\text{pREE})=\Delta \text{REE}$$ (1)

Linear regression was used to estimate EE attributable to GNG (beta coefficient, β₁) after accounting for age, sex, FFM, FM, and disease/study site (INSR, lipodystrophy, T2DM, and obesity). Because sex, FM, and disease/study site were not significant predictors of REE, they were removed from the final model, yielding: Y_REE = β₀ + β₁ × Absolute GNG + β₂ × Age + β₃ × FFM. Estimated GNG EE was calculated as the product of EE attributable to GNG (β₁) and absolute GNG (Eq. 2):

$$\text{Estimated GNG EE}\left(\frac{\text{kcal}}{\text{day}}\right)=\text{Absolute GNG }\left(\frac{\mu \text{mol}}{\text{min}}\right)\times {\mathrm{\beta }}_1$$ (2)

Predicted GNG EE was calculated as the product of EE attributable to GNG and published GNG rates in lean nondiabetic adults (6) (Eq. 3):

$$\text{Predicted GNG EE}\left(\frac{\text{kcal}}{\text{day}}\right)=6.7 \left(\mu \text{mol }\times \text{ kgFF}{\text{M}}^{-1}\text{mi}{\text{n}}^{-1}\right)\times \text{ FFM }\left(\text{kg}\right) \text{×} {\mathrm{\beta }}_1$$ (3)

Excess GNG EE was calculated as the difference in estimated EE attributable to GNG in patients with INSR, lipodystrophy, T2DM, and obesity (estimated GNG EE) from the estimated EE attributable to GNG in lean nondiabetic adults (predicted GNG EE) (Eq. 4):

$$\text{Excess GNG EE}\left(\frac{\text{kcal}}{\text{day}}\right)=\text{ Estimated GNG EE}-\text{Predicted GNG EE}$$ (4)

RESULTS

Table 1 presents baseline characteristics for 7 patients with INSR, 12 with lipodystrophy, 18 with T2DM (9-NIH, 9-Baylor), and 10 with obesity with complete data. Patients with lipodystrophy were older than those with T2DM-NIH, T2DM-Baylor, and obesity (30.3 ± 16.4 yr vs. 15.7 ± 2.5 yr, 13.9 ± 1.1 yr, and 14.8 ± 2.0 yr; P < 0.01). Patients with INSR and lipodystrophy had lower FM than patients with T2DM-NIH, T2DM-Baylor, and obesity (11.0 ± 13.8 kg and 16.1 ± 10.5 kg vs. 50.4 ± 16 kg, 49.0 ± 11.7 kg, and 40.5 ± 8.8; P < 0.001). Patients with INSR had lower FFM than patients with lipodystrophy, T2DM-NIH, T2DM-Baylor, and obesity (36.5 ± 11.1 kg vs. 60.9 ± 10.9 kg, 56.7 ± 16.7 kg, 59.2 ± 6.6 kg, and 55.2 ± 11.8; P = 0.001). REE correlated with FFM (Fig. 1A, r = 0.67, P < 0.001). There were no between-group differences in REE after accounting for FFM (Fig. 1B, P = 0.3). GNG had a borderline significant correlation with FFM (Fig. 1C, r = 0.27, P = 0.06).

Table 1.

Demographic, clinical, and anthropometric measurements

Clinical Values	INSR (n = 7)	Lipodystrophy (n = 12)	T2DM-Baylor (n = 9)	T2DM-NIH (n = 9)	Obesity (n = 10)	P
Age, yr	21.9 ± 6.0	30.3 ± 16.4	15.7 ± 2.5	13.9 ± 1.1	14.8 ± 2.0	<0.001e,f,g
Sex (male/female)	4/3	5/7	0/9	3/6	2/8	0.09
Weight, kg	48.6 ± 23.4	73.8 ± 17.7	107 ± 31.4	113.9 ± 15.6	97.2 ± 17.4	<0.001b,c,d,e,f
BMI, kg/m2	19.7 ± 8.4	24.5 ± 5.4	39.0 ± 5.9	41.4 ± 7.9	36.5 ± 5.3	<0.001b,c,d,e,f,g
Fat mass, kg	11.0 ± 13.8	16.1 ± 10.5	50.4 ± 16	49.0 ± 11.7	40.5 ± 8.8	<0.001b,c,d,e,f,g
Body fat, %	18.6 ± 10.7	20.0 ± 11.0	46.8 ± 4.8	44.9 ± 5.2	41.7 ± 5.3	<0.001b,c,d,e,f,g
Fat-free mass, kg	36.5 ± 11.1	60.9 ± 10.9	56.7 ± 16.7	59.2 ± 6.6	55.2 ± 11.8	0.001a,b,c,d
Adjusted REE, kcal/day*	1843 ± 425	1802 ± 230	2010 ± 325	1903 ± 115	1756 ± 166	0.3
RQ	0.82 ± 0.01	0.88 ± 0.05	0.80 ± 0.04	0.80 ± 0.04	0.85 ± 0.05	0.003a,e,f
Fractional GNG, %	0.77 ± 0.09	0.52 ± 0.11	0.59 ± 0.06	0.68 ± 0.1	0.55 ± 0.12	<0.001a,c,e,f,h
Endogenous leptin level, ng/dL	2.8 [2.7, 12.0]	3.8 [1.4, 8.9]				0.9
Patients on insulin, %	57	50	None	None	None	1.0
Insulin dose, units/day†	900 [195,1500]	219 [137.5, 320]	None	None	None	0.3
No. of noninsulin diabetes medications	1.0 ± 0	1.2 ± 0.6	None	1.0 ± 0	None	0.5

BMI, body mass index; GNG, gluconeogenesis; INSR, insulin receptor pathogenic variants; NIH, National Institutes of Health; REE, resting energy expenditure; RQ, respiratory quotient; T2DM, type 2 diabetes mellitus.

^a P < 0.05 for INSR vs. lipodystrophy; ^bP < 0.05 for INSR vs. T2DM-Baylor; ^cP < 0.05 for INSR vs. T2DM-NIH; ^dP < 0.05 for INSR vs. obesity; ^eP < 0.05 for lipodystrophy vs. T2DM-Baylor; ^fP < 0.05 for lipodystrophy vs. T2DM-NIH; ^gP < 0.05 for lipodystrophy vs. obesity; ^hP < 0.05 for T2DM-NIH vs. obesity.

^*Adjusted for fat-free mass.

^†Among insulin users only.

Figure 1.

Energy expenditure attributable to gluconeogenesis (GNG) and fat-free mass (FFM). A: FFM correlated with resting energy expenditure (REE); however, there were no between-group differences in FFM-adjusted REE (B). C: there was a trend toward significant correlation between FFM and GNG. D: after adjusting for age and FFM, absolute GNG correlated with REE among patients with insulin receptor pathogenic variants (INSR, squares), lipodystrophy (LD, circles), type 2 diabetes (T2DM, triangles, diamonds), and obesity (x-sign). P < 0.05 represented statistical significance.

Gluconeogenesis Correlated with REE

Figure 1D shows the adjusted relationship between GNG and REE. GNG correlated with REE (P = 0.0002). Table 2 shows multivariate linear regression estimates (Y_REE = β₀ + β₁ × Absolute GNG + β₂ × Age + β₃ × FFM; adjusted R² = 0.56). The estimated EE attributable to GNG (β₁) after accounting for age and FFM was 0.99 kcal/day per 1 µmol/min increase in GNG (P < 0.001). Sensitivity analyses removing patients with INSR (due to lower FFM) did not significantly alter regression estimates.

Table 2.

Multivariate linear regression estimates

Variable	β Coefficient [95% CI]	P
Adjusted R2 = 0.56
β0: Intercept	672.4 [340, 1004]	<0.001
β1: Absolute gluconeogenesis, µmol/min	0.99 [0.5, 1.5]	<0.001
β2: Age, yr	−7.9 [−15.2, −0.5]	0.04
β3: Fat-free mass, kg	14.8 [9.5, 20.0]	<0.001

Multivariate linear regression was used to estimate energy expenditure (EE) attributable to gluconeogenesis (beta coefficient, β₁). We performed forward selection of covariates including age, sex, fat-free mass, fat mass, and disease type/study site (INSR, lipodystrophy, T2DM-Baylor, T2DM-NIH, and obesity). Potential covariates that were not significant in univariate analyses (sex, fat mass, and disease type/study site) were removed from the final model. P < 0.05 represented statistical significance. P values are two sided. CI, confidence interval; INSR, insulin receptor pathogenic variants; T2DM, type 2 diabetes mellitus.

Gluconeogenesis Accounts for 22%–39% of REE

Figure 2 shows rates of GNG, estimated GNG EE, predicted GNG EE, and excess GNG EE. After adjustment for FFM, GNG was highest in patients with severe insulin resistance (INSR: 646 ± 139 µmol/min, lipodystrophy: 632 ± 204 µmol/min) versus obesity (396 ± 80 µmol/min) (P = 0.001). Estimated GNG EE was higher in INSR than in obesity (591 ± 129 kcal/day vs. 400 ± 105 kcal/day, P = 0.02). By contrast, predicted GNG EE was lowest in INSR versus all other groups (P = 0.002). Excess GNG EE was higher in INSR and lipodystrophy than in obesity (P = 0.003). GNG EE across all groups accounted for 29% (541 ± 167 kcal/day) of REE, ranging from 22% in obesity to 39% in INSR.

Figure 2.

Measures of gluconeogenesis (GNG) and energy expenditure. Absolute GNG (A) adjusted for fat-free mass was the highest in patients with severe insulin resistance [insulin receptor pathogenic variants (INSR, squares) and lipodystrophy (LD, circles)] vs. mild insulin resistance [type 2 diabetes (T2DM, triangles and diamonds) and obesity (x-sign)]. Estimated GNG energy expenditure (EE; B) was higher in INSR vs. obesity. Predicted GNG EE (product of the EE attributable to GNG and published GNG rates in lean nondiabetic adults) (C) was the lowest in INSR vs. all other groups. Mean excess GNG EE (D) equaled 12% of REE and mean GNG EE (E) equaled 29% of REE. These results suggest that the higher the rates of GNG, the more significant the proportion of resting EE accounted for by GNG. The Mifflin St. Jeor equation (F) underestimated REE. Between-group comparisons between GNG rates, GNG EE, predicted GNG EE, and excess GNG EE were made using analysis of variance (ANOVA), with post hoc pairwise by Tukey’s test. Difference in mean REE between measured (indirect calorimetry) and Mifflin St. Jeor equation [Men: 10 × weight (kg) + 6.25 × height (cm) − 5 × age (yr) + 5; Women: 10 × weight (kg) + 6.25 × height (cm) − 5 × age (yr) − 161] was made using paired t test. P < 0.05 represented statistical significance. P values are two sided. *Indicates statistical significance.

Predicted REE by Mifflin St. Jeor Equation Underestimated REE Primarily in Patients with Severe Insulin Resistance

The Mifflin St. Jeor equation correlated with estimated REE (r = 0.70, P < 0.001) but underestimated REE (Fig. 2F) in 64% (30/47) of patients, primarily in those with severe insulin resistance due to INSR or lipodystrophy in whom the underestimate was ∼250 kcal/day.

DISCUSSION

This study demonstrates that, in a pooled population ranging from mild to severe insulin resistance, GNG correlates with REE. Previous studies suggested that GNG accounts for 5%–8% of REE in healthy humans based on theoretical calculations of rates of GNG, moles of oxygen consumption by GNG, and average oxygen consumption (1). Our study of patients with high GNG rates found that GNG accounted for a higher proportion of REE, equal to 29% of REE, and 12% above predicted in lean nondiabetic adults.

Hyperglycemia is associated with increased risk of microvascular and macrovascular complications (18). Medications to reduce glucose, such as insulin, sulfonylureas, and thiazolidinediones, are mainstays of treatment (19), and decrease GNG in human (20) and rodent studies (21–23), which may contribute to weight gain associated with these medications. Conversely, increased GNG with sodium glucose tranporter type 2 inhibitors may contribute to weight loss associated with these drugs (24, 25). However, changes in GNG-associated EE with these drugs may lead to compensatory changes in energy intake, thus reducing any effects on body weight change.

Understanding the physiological parameters contributing to variation in REE may provide insight into developing therapeutic targets to reduce obesity. Equations to predict EE, such as the Mifflin St. Jeor, have been developed based on known parameters contributing to REE variation, including body size and age (26). In our study of patients with insulin resistance, we found that the Mifflin St. Jeor equation underestimated REE. Our findings have research implications, particularly in studies using controlled diets, as prediction equations may underestimate caloric needs in patients with insulin resistance. We speculate that underestimation of REE using prediction equations in patients with high GNG is due to the energetic cost of GNG. However, this speculation is countered by the observation that adjusted REE was not higher in the most insulin-resistant populations in our study. This may be due to the lack of a control population with normal GNG or due to the partial correlation of GNG with FFM. Alternatively, other components of energy expenditure (e.g., thyroid hormone) may be downregulated in patients with severe insulin resistance to compensate for the increased energetic cost of GNG (27).

The strength of this study is the use of obese patients without diabetes as well as mild (i.e., T2DM) and severe (i.e., INSR and lipodystrophy) forms of insulin resistance, and high GNG rates, which allows for a clearer estimation of EE attributable to GNG than healthy populations where GNG rates are typically low. This study has several limitations. The small sample size did not permit subgroup analyses to estimate relationships of GNG with REE separately in each patient group; however, INSR and lipodystrophy are rare diseases. The Mifflin St. Jeor equation used to predict REE was validated in a different population and may not be generalizable to our study population, as REE was underestimated in our study. The patient populations were heterogenous with respect to age; GNG and EE in adult cohorts with type 2 diabetes and obesity without diabetes were not available, and no data was available in lean subjects. Published GNG rates in healthy adults may not be applicable to adolescents, although most adolescents in our study were postpubertal and at adult body size. REE and body composition were measured by two different metabolic carts and a metabolic chamber and by two DXA scanners. Within-subject comparability of these metabolic carts and chamber has shown <2% differences in EE measurement between devices (15–17). Finally, this analysis can only distinguish the mathematical contribution of GNG to EE. Future studies measuring EE after manipulation of GNG are needed to verify the effect.

In conclusion, this study demonstrates that the energetic cost of GNG is substantial in patients with insulin resistance and high GNG rates. In clinical practice, treatments for diabetes that inhibit GNG can be expected to reduce REE and may promote positive energy balance. Research studies should tailor their use of prediction equations to estimate caloric needs in patients with disease states that increase GNG and hence REE.

GRANTS

This work was supported by the intramural research program of the National Institute of Diabetes and Digestive and Kidney Diseases, United States Department of Agriculture Research Service under Cooperative Agreement Number 58-3092-5-001, and the NIH Medical Research Scholars Program, a public-private partnership supported jointly by the NIH and contributions to the Foundation for the NIH from the Doris Duke Charitable Foundation, Genentech, the American Association for Dental Research, the Colgate-Palmolive Company, and other private donors.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

E.Q. and R.J. Brown conceived and designed research; S.C., S.T.C., and R.J. Brychta performed experiments; E.Q. analyzed data; E.Q., S.C., S.T.C., R.J. Brown, K.Y.C., and R.J. Brychta interpreted results of experiments; E.Q. prepared figures; E.Q. drafted manuscript; E.Q., S.C., S.T.C., R.J. Brown, K.Y.C., and R.J. Brychta edited and revised manuscript; E.Q., S.C., S.T.C., R.J. Brown, K.Y.C., and R.J. Brychta approved final version of manuscript.