Exercise attenuates bed rest-induced increases in insulin resistance while α-klotho increases in 55 to 65 year-old women and men

Abstract

We evaluated the efficacy of combined aerobic, high intensity interval (HIIT) and resistive exercises during Head-Down Bed Rest (HDBR) in preventing metabolic derangement. Twenty-two participants (control: F (5) BMI: 22.1 ± 2.1, M (6) BMI: 25.7 ± 2.2, exercise F (6) BMI: 24.3 ± 3.2, M (5) BMI: 27.5 ± 1.2; Average age: 59 ± 3 years; weight: 70.1 ± 14.2 kg; height: 1.67 ± 0.09 m; BMI: 24.9 ± 3.0 kg/m²)) confined to 6° HDBR for 14-days were randomly assigned to either EX (1 h/day head-down exercise including HIIT every other day) or CONT (passive movement). Oral glucose tolerance tests (OGTT) at baseline, day 9 HDBR (HDT 9) and 13 HDBR (HDT 13) were matched to fasted biomarker analyses and Vastus lateralis biopsies. HDBR increased area under the curve for glucose (p < 0.01, interaction between groups; group × bed rest; p = 0.07) and insulin (p < 0.01). Matsuda index fell in CONT alone, FGF-23 increased in EX correlating with elevations to circulating levels of C reactive protein; a negative correlation existed between FGF-23 and HOMA-IR in CONT. A novel finding was that α-klotho increased in all participants (p < 0.0001), starting higher and rising faster in women, whilst negatively correlating with Matsuda index in CONT alone. Undercarboxylated-osteocalcin ratio fell in CONT (p = 0.05). Neither IGF-1, nor GLUT1/4 protein content changed during HDBR. HDBR promoted insulin resistance with increases in circulating α-klotho; FGF-23 increased only in EX. Differing Matsuda index responses indicate benefits imparted by multimodal exercise despite elevations in α-klotho and FGF-23.

Clinical Trial Registration Number: NCT04964999.

Introduction

Approximately one quarter of the population of the United States of America currently exhibit insulin resistance (IR)¹, a key contributor to the pathogenesis of type 2 diabetes. Progressive and systemic IR stimulates augmented pancreatic β-cell insulin secretion eventually leading to β-cell exhaustion and hyperglycaemia. Furthermore, IR impedes lipid and protein metabolism², contributing to hypercholesterolaemia, dysregulated protein metabolism³ elevated cardiovascular risk and sarcopenia. Muscle tissues constitute the primary corporeal glucose sink; muscle loss during inactivity amplifies IR as resistance reproducibly develops during protracted sedentary behaviour⁴. Comprehension and attenuation of these detrimental physiological consequences is paramount to the continued health of aging individuals.

Bed rest studies impose a severe sedentary environment. Experimentally, they serve as models of the protracted immobility experienced during hospitalization and chronic illness⁵. Six-degree head-down bed rest (HDBR) is an established analogue for spaceflight research⁶, reproducing many of the physiological consequences of spaceflight, including the development of IR^7,8. Most bed rest studies only include young adults⁹, limiting their relevance to older individuals. Therefore, we undertook the study of metabolic dysregulation during 14-days HDBR, commonly used in previous studies¹⁰, in 55–65-year-old men and women, evaluating the efficacy of a multi-modal exercise regime previously reported to limit IR development in young adults⁶. To better understand physiological mechanisms promoting IR, we performed simultaneous quantification of multiple biomarkers including osteocalcin, intact Fibroblast Growth Factor-23 (FGF-23), soluble α-klotho (an activator of FGF-23), muscle glucose transporters GLUT 1 and 4, vitamin D, circulating Insulin Growth Factor-1 (IGF-1), and high-sensitivity C-reactive protein (hsCRP). hsCRP was included as a marker of systemic inflammation, which is tightly linked to insulin resistance, while vitamin D was assessed due to its proposed regulatory role in glucose metabolism and its potential interaction with FGF-23 and klotho signalling. While some biomarkers such as osteocalcin and IGF-1 have known responses to inactivity and exercise, the roles of α-klotho and FGF-23 in this context remain unclear or understudied, highlighting the exploratory nature of their inclusion in this investigation. We previously reported the effects of exercise on strength, body composition, cardiovascular and vascular capacity and mitochondrial function during bed rest^11,12. Here, we report on the effects on IR based on oral glucose tolerance tests conducted prior to and during HDBR. Our hypotheses included confirmation that 14-days HDBR induces insulin resistance in an older population, and that exercise would limit detrimental alterations conducive to IR. Furthermore, we hypothesise that these changes will coincide with reductions in α-klotho and changes to the α-klotho-FGF-23 axis, which is intimately linked to insulin resistance^13,14,15,16; previously α-klotho fell during spaceflight and hind-limb unloading in rodents¹⁷ whilst levels of FGF-23 remain unchanged¹⁸.

Results

Glucose, insulin and resistance indices

HDBR significantly increased the area under the curve (AUC) for both glucose (p < 0.01, effect size 0.55; Fig. 1A) and insulin (p < 0.01, effect size 0.98; Fig. 1B). A close-to significant interactive effect of bed rest * group was identified for glucose AUC alone (p = 0.07, effect size 0.65), post-hoc testing identified increases in CONT alone. These outcomes were no different when analysing the product of glucose AUC and insulin AUC (data not shown). QUICKI index (p = 0.03, effect size − 1.09) was lower in EX than CONT (Fig. 1C). Similarly, HOMA-IR (p = 0.058, effect size 0.89) tended to be greater in EX than CONT (Fig. 1D). In response to a glucose challenge, ISI_0,120 revealed a bed rest effect only (Fig. 1E, p = 0.005, effect size – 0.74) whilst Matsuda index identified the interactive effect of bed rest and group (p = 0.03, effect size 0.97; Fig. 1F), with post-hoc analyses identifying reductions in CONT alone (HDT-2 vs HDT 9, p = 0.054; HDT-2 vs HDT 13, p < 0.01). Correlations between ISI_0,120 and Matsuda index were significant for CONT (r_rm = 0.86, p = 0.0007) but not EX (r_rm = 0.59, p = 0.09). Regarding β-cell function, BIGTT-AIR calculations (exp(β₀ + β₁·ln(Ins₀) + β₂·ln(Ins₃₀) + β₃·ln(Ins₁₂₀) + β₄·ln(Gluc₀) + β₅·ln(Gluc₃₀) + β₆·ln(Gluc₁₂₀) + β₇·Sex + β₈·BMI) did not identify any differences associated with either bed rest or group. A full OGTT dataset is provided in the supplementary data.

Fig. 1

Glucose (A) and insulin (B) area under the curve (AUC), QUICKI (C), HOMA-IR (D), ISI_(0,120) (E), and Matsuda Index (F) for CONT (left) and EX (right) at HDT-2, HDT 9 and HDT 13. Individual participant data are superimposed on the median and interquartile ranges.

FGF-23

Intact FGF-23 concentrations increased in EX at HDT 9 (+ 9.62 pg/ml, p < 0.001) and HDT 15 (+ 8.76 pg/ml, p < 0.01). No changes occurred in CONT (Fig. 2A, bed rest*group interaction, p < 0.01, effect size 1.24), with no role of sex. Neither the incorporation of hsCRP, serum phosphate, 25-hydroxyvitamin D, nor serum iron concentrations into more complex mixed linear models altered the statistical outcome of our comparisons, and always resulted in poorer model fit compared to models including only group and bed rest effects. We identified a positive correlation between FGF-23 and serum hsCRP in EX (p = 0.012, Fig. 2B, C). We did not identify statistically significant correlations between FGF-23 and serum phosphate (Fig. 2D, E), nor 25-hydroxyvitamin D (data not shown), nor serum iron concentrations (data not shown).

Fig. 2

Intact FGF-23 concentrations (A) for CONT (left) and EX (right) at HDT-2, HDT 9 and HDT 15. Individual participant data are superimposed on the median and interquartile ranges, with circles representing women and triangles representing men. Repeated measures correlations between FGF-23 and hsCRP (B and C) and FGF-23 and Serum Phosphate (D and E).

Role of FGF-23 on OGTT outcomes

Incorporation of FGF-23 as an additional fixed effect (along with bed rest and group) in our mixed linear model approach to assess OGTT outcomes (glucose AUC, insulin AUC, HOMA-IR, QUICKI, ISI_(0,120), and Matsuda index) consistently resulted in poorer model fit compared to those without FGF-23. Outcomes also remained unchanged. Repeated measures correlations revealed a negative relationship between intact FGF-23 concentration and HOMA-IR in CONT alone (r_rm = − 0.59, p = 0.04; EX: r_rm = − 0.40, p = 0.20). There were no statistically significant relationships identified between FGF-23 and ISI_(0,120) or Matsuda index.

α-Klotho

Bed rest augmented concentrations of soluble α-klotho (Fig. 3A, B) with a difference in response between men and women (bed rest*sex interaction, p = 0.038, effect size 0.91); post-hoc analyses identified differences in women between HDT-2 and both HDT 9 (+ 110.0 pg/ml, p < 0.0001) and HDT 15 (+ 150.2 pg/ml, p < 0.0001). In men, α-klotho was only elevated relative to HDT-2 at HDT 15 (HDT 9: + 35.5 pg/ml, p = 0.133, HDT 15: + 73.5 pg/ml, p = 0.003). α-klotho concentrations were negatively correlated with ISI_(0,120) in CONT alone (Fig. 4A, r_m = − 0.602, p = 0.049), EX (Fig. 4B, r_m = − 0.492, p = 0.105); similarly for Matsuda index, only CONT showed a significant correlation with α-klotho levels (Fig. 4C, r_m = − 0.63, p = 0.036), EX (Fig. 4D, r_m = 0.161, p = 0.68). CONT experienced decreases in ISI_(0,120) and in Matsuda Index as α-klotho increased during bed rest. None of the other OGTT or resistance outcomes showed any association with changes to α-klotho. There was also no relationship between α-klotho and FGF-23 concentrations for CONT (r_rm = 0.036, p = 0.87) or EX (r_rm = 0.364, p = 0.088).

Fig. 3

Biomarker changes during HDBR. Soluble α-klotho plasma concentrations (A) for CONT (left) and EX (right) at HDT-2, HDT 9 and HDT 15. For ease of viewing these same data are presented following grouping by sex (B); women (left) and men (right). Individual participant data are superimposed on the median and interquartile ranges, with circles representing women and triangles representing men. Total osteocalcin (C), undercarboxylated osteocalcin (D) and undercarboxylated osteocalcin ratio (E) for CONT (left) and EX (right) at HDT-2, HDT 9 and HDT 13. Individual participant data are superimposed on the median and interquartile ranges.

Fig. 4

Repeated Measures Correlations between α-klotho and Insulin Sensitivity Index ((ISI_(0,120)), A, B) and Matsuda index (C, D), for CONT (left) and EX (right). Individual participant data are provided with circles representing HDT-2 and squares representing HDT 9.

Osteocalcin

Total osteocalcin concentration increased between HDT-2 and HDT 9 (p = 0.005, effect size 0.68, Fig. 3C), without a change in undercarboxylated osteocalcin (Fig. 3D). However, a bed rest*group interaction was identified for the undercarboxylated osteocalcin ratio (p = 0.05, effect size 0.79, Fig. 3E), with post-hoc comparisons revealing a reduction in this ratio at HDT 9 in CONT (− 3.9%, p = 0.016).

Repeated measures correlations revealed a positive relationship between undercarboxylated osteocalcin ratio and Matsuda index score only in CONT (r_rm = 0.73, p = 0.01, Fig. 3F), but it was not associated with any OGTT or insulin resistance/sensitivity factors.

Insulin-like growth factor 1

There was no effect of bed rest or group on changes to Insulin-like Growth Factor 1 between HDT-2 and HDT 9 (Table 1).

Table 1 Median (25th–75th percentile) group values for OGTT outcomes and biochemical analyses performed on participants at HDT-2, HDT 9 and HDT 13.

25-Hydroxyvitamin D

There was no effect of bed rest or group on changes to 25-Hydroxyvitamin D between HDT-2 and HDT 9 (Table 1).

Muscle GLUT1 and GLUT4

There was no change in GLUT1 (bed rest effect p = 0.12; group effect p = 0.91, interactive effect p = 0.37) nor in GLUT4 (bed rest effect p = 0.12; group effect p = 0.66; interactive effect p = 0.88) muscle protein content from the Vastus lateralis between HDT 1 and HDT 14, for either group. In CONT, GLUT1 remained stable (HDT 1: 0.012 ± 0.01 A.U. vs HDT 14: 0.014 ± 0.010 A.U.), as did GLUT 4 (HDT 1: 0.047 ± 0.013 A.U. vs HDT 14: 0.041 ± 0.015 A.U.). In EX, the differences in GLUT1 (HDT 1: 0.009 ± 0.006 A.U. vs HDT 14: 0.016 ± 0.011 A.U.) and GLUT 4 (HDT 1: 0.045 ± 0.014 A.U. vs HDT 14: 0.039 ± 0.012 A.U.) were not statistically different. Regression analyses between GLUT1/GLUT4 and OGTT or insulin resistance/sensitivity factors failed to identify any associations.

Body weight

Bodyweight fell in all participants over the course of bed rest (p < 0.0001, effect size − 0.77), with neither an effect of group (p = 0.47) or interaction between group and bed rest (p = 0.097).

Discussion

Historical data confirm the detrimental effects of bed rest for glucose and insulin handling in younger adults⁸, and the protective benefits endowed by exercise⁶. This current study presents the first dataset pertaining to older 55–65-year-olds during HDBR, a population more likely to experience periods of inactivity during changes in health, such as hospitalization or following injury. We aimed to identify important and hitherto unknown metabolic benefits of exercise in this age-group during HDBR. In contrast to our hypothesis based on previous results from animal models during hind-limb unloading and spaceflight, we present novel findings of augmentation of FGF-23 during HDBR in EX only and of α-klotho that was independent of exercise and not associated with the development of insulin resistance.

Glucose, insulin and resistance indices

Our chosen testing timepoints (HDT-2, HDT 9 and HDT 13) allowed for a temporal assessment of OGTT responses throughout HDBR. Previously, impaired glucose handling in younger adults was apparent after only 4-days, and at 6-days, 21-days HDBR⁸, and 60-days HDBR¹⁹ with diminished insulin sensitivity based on QUICKI and Matsuda index. We found HDBR induced increases in glucose and insulin AUC, corroborating previous findings. Despite the subjective appearance of greater incremental increases in these variables over time in CONT, this was not detected with statistical testing, possibly due to our study being underpowered. The close to significant interactive effect of bed rest * group for our glucose AUC results (p = 0.07), and the statistically significant Matsuda index outcomes does suggest a protective role of exercise for insulin sensitivity during HDBR. Throughout HDBR, Matsuda index values fell in CONT whereas those in EX remained stable. Furthermore, CONT exhibited consecutive reductions in Matsuda index from HDT-2 to HDT 9 and from HDT 9 to HDT 13, signifying progressive IR development. However, the lack of difference in ISI_(0,120) changes between HDT-2 and HDT9 in CONT vs EX suggests that per kg of bodyweight, insulin sensitivity did not differ between groups. This appreciable disagreement in sensitivity indices could be linked to changes in participants’ bodyweight during bed rest, as the calculation of Matsuda index does not incorporate bodyweight, nor skewness of the distribution of insulin values²⁰. Additionally, the Matsuda index draws data from all timepoints during OGTT, rather than just prior to and 120-min after ingestion of the OGTT drink; different speeds of glucose clearance prior to 120-min between groups may account for the discrepancy between results. Nevertheless, our findings confirm that 14-days of HDBR in 55–65-year-old adults resulted in increased glucose and insulin AUC, and progressive IR, without any change in calculated β-cell function, whilst also confirming an appreciable benefit of exercise as identified by the Matsuda Index results.

FGF-23

Recent investigations identified the important role FGF-23 plays in energy metabolism; large cohort studies concluded that FGF-23 (both c-terminal FGF-23 and intact FGF-23) are strongly associated with incidence of type-2 diabetes and obesity¹³. Positive correlations between FGF-23 and plasma glucose, insulin and proinsulin were identified, whilst FGF-23 levels prospectively predicted diabetes risk¹³. FGF-23 is also positively correlated with IR in multiple populations including elderly, African-American, obese, and those with cardiovascular disease and therapy-associated lipodystrophy¹⁴. Therefore, FGF-23 may play a central role in metabolic state and may contribute to the development of IR during protracted sedentary behaviours. As the intact pre-cleaved form is biologically active, we chose to study intact FGF-23 and its effects on IR during HDBR.

Intact FGF-23 was unchanged by HDBR in CONT-participants, corroborating previous findings during human dry immersion¹⁸. However, circulating FGF-23 increased in EX-participants by nearly 50%, adding to the rather equivocal discourse on FGF-23 responses in human exercise research^21,22,23. Previously, only one human study identified increases in intact FGF-23 in competitive cyclists during a multistage race²¹, with others showing increases in the c-terminal FGF-23 fragment alone^22,23. Furthermore, increases are associated only with high or maximal intensity regimes.

We did not identify an independent effect of intact FGF-23 on any measured variables or insulin sensitivity/resistance indices. However, we did identify a negative relationship between intact FGF-23 and HOMA-IR in our CONT-participants that mirrored previous findings in obese adolescents²⁴ but contradicted cross-sectional studies^13,14. Our repeated measures approach suggested that CONT-participants exhibited the tendency to conform to this relationship irrespective of absolute intact FGF-23 levels over time; if an individual’s FGF-23 decreased, HOMA-IR increased, and vice-versa. Within CONT, most participants’ insulin AUC increased between HDT-2 and HDT-9. As insulin is a known suppressor of FGF-23²⁵, these relationships may be due to feed-back mechanisms rather than as a consequence of changes to FGF-23 regulation; previously, this positive relationship was identified only in individuals with pre-existing insulin resistance²⁴.

Changes in inflammation, vitamin D and iron status, and phosphate levels can all alter FGF-23 control^26,27,28,29. We assessed these relationships in our participants. We found no effect of circulating hsCRP or iron availability on FGF-23 throughout HDBR. We did identify a positive correlation between intact FGF-23 and hsCRP in EX, but as acute inflammation simultaneously increases cleavage rates so that levels of intact FGF-23 remain unchanged, we would not expect to see such a clear augmentation in FGF-23 levels^27,28,29. The absolute difference in inflammatory markers between HDT-2 and HDT 9 was also small. Vitamin D supplementation of > 2000 IU/day can increase circulating FGF-23 concentrations³⁰. However, we found no effect of 1200 IU/day vitamin D-3 supplementation on circulating 25-hydroxyvitamin D levels, nor of measured 25-hydroxyvitamin D levels on those of intact FGF-23. Serum phosphate increased in both groups and was not associated with changes to intact FGF-23.

Summarising our investigations on FGF-23, exercise and development of IR, we believe it reasonable to postulate that exercise limited progression of IR during HDBR independently of its induction of intact FGF-23, most likely due to the intensity of the personalized exercise prescription in EX participants.

α-Klotho

α-Klotho is an obligatory cofactor expressed on cell membranes for the normal binding and functioning of FGF-23, with the soluble circulating form thought to provide this essential cofactor to tissues that do not normally express the klotho gene³¹. Mainly expressed in the renal tubule, its relationship with FGF-23 is classically attributed to that of phosphate and calcium homeostasis³¹. However, it also acts upon multiple pathogenic cardiometabolic pathways including endothelial dysfunction, vascular calcification, and atherosclerosis. It is also implicated as an inhibitor of insulin and IGF-1 signalling in cell and animals models³², corroborating cross-sectional human studies that identified positive correlations between α-klotho concentrations and insulin resistance measured by HOMA-IR¹⁵. Finally, it is believed that klotho plays a key and positive role in the aging process; knock-out of the klotho gene leads to rapid aging in mice³³, whilst exogenous administration can attenuate age-related cognitive decline in primates³⁴.

We found elevations to soluble α-klotho following HDBR across the whole cohort that contradicted our hypothesis. These increases were independent of exercise intervention. Furthermore, the change in α-klotho was far greater in women than in men, where differences achieved statistical significance by only HDT 13. Our findings are opposite to changes in rodents, where α-klotho levels were found to fall following both hind-limb unloading and spaceflight¹⁷. We also found a stark difference in baseline α-klotho concentrations between men and women, contradicting previous cross-sectional investigations in adults between 18 and 85 years of age³⁵ but in keeping with recent findings¹⁶.

To the best of our knowledge, this is the first report of bed rest induced changes to soluble α-klotho and represents a novel and important finding. It may signify an attempt to minimise the bed rest-induced muscle losses in opposition to FGF-23^14,16. The lack of relationship between changes to α-klotho and FGF-23 concentrations over the bed rest intervention confirms that FGF-23 did not rise in our exercise participants due to a ‘resistance’ induced by diminished klotho availability³⁶, further supporting our notion that FGF-23 changes in EX were stimulated by the exercise intervention.

Regarding the previously reported actions of α-klotho on insulin resistance^15,32, we identified negative correlations between α-klotho and both insulin sensitivity indicies calculated in this study in CONT, with higher α-klotho and poorer insulin sensitivity during bed rest. Recently, it was identified that α-klotho can exacerbate insulin resistance through a number of mechanisms including prevention of GLUT4 translocation and interference with intracellular insulin signalling mediators³². Furthermore, insulin itself increases serum klotho levels by enhancing cleavage of transmembrane klotho, resulting in a feedback loop whereby klotho dampens insulin’s sensitivity. Therefore, bed rest induced insulin resistance may, in part, be linked to elevated circulating α-klotho. As this relationship was disrupted in our EX-participants, the protective effects of exercise may include mechanisms that limit the intracellular effect of α-klotho, but exploration of these relationships remains outside the scope of our current study.

Osteocalcin

Despite no change to the total undercarboxylated osteocalcin levels in our participants, the undercarboxylated osteocalcin ratio did fall in CONT whilst remaining stable in EX. CONT also exhibited simultaneous reductions in Matsuda index, mimicking similar associations in diabetic and non-diabetic individuals³⁷. Differences between our groups may be due to the exercise intervention; whilst extreme sedentary behaviours reduced osteocalcin in cosmonauts³⁸, others have found little change or increases in undercarboxylated osteocalcin during bed rest³⁹. This may be in part due to improvements in diet and vitamin K intake during controlled bed rest studies that reduced undercarboxylated osteocalcin levels during the acclimatization phase, following which they remained at a lower level³⁹; this may have also occurred to all of our participants. However, robust evidence confirms the beneficial effects of exercise on increasing levels of osteocalcin; given that all other study factors were tightly controlled between CONT and EX, exercise remains a possible protective mechanism through which EX participants maintained their undercarboxylated osteocalcin ratio.

Skeletal muscle contributions

Our evidence of decreasing insulin sensitivity in CONT, identified via attenuated Matsuda index, supports the hypothesis of beneficial effects of countermeasure exercises during HDBR. One strength of the Matsuda index lies in its incorporation of whole-body insulin sensitivity during the OGTT, rather than assuming equal peripheral and hepatic insulin sensitivity as with the HOMA-IR and Quicki indices⁴⁰. There was a significant reduction in thigh muscle volume measured by MRI experienced in CONT¹², where an average of 6.7% volume loss occurred over HDBR, and might better account for potentially disproportionate changes in hepatic vs peripheral IR. Muscle atrophy could account for the CONT changes to insulin resistance as it is the primary site of glucose disposal. Additionally, and rather unexpectedly, we found unaltered levels of GLUT4 protein content in the Vastus lateralis (VL) in the same participants. The lack of change across our entire cohort is contrary to multiple reports of attenuated gastrocnemius and/or VL GLUT4 muscle content following periods of inactivity and bed-rest⁴¹. Within EX, the witnessed maintenance rather than rise in protein content may be attributable to total exercise loading during HDBR, and pre-HDBR habitual exercise. Previously, increases to GLUT4 protein content following 1 h training per day at 75% VO2 max occurred in otherwise sedentary men and women⁴². Our participants lived active lifestyles prior to HDBR and we witnessed maintenance of GLUT4 protein content and of VO2 max⁴³; increases could require supra-normal exercise loading. Also, the 23-h of otherwise sedentary behaviour may not have been completely counteracted by our exercise intervention. Regarding GLUT4 bioavailability, our inability to assess localization to the muscle cell membrane and to measure muscle capillary density prevented us from commenting on aspects of glucose handling that change independently of GLUT4 protein content.

Conclusion

Fourteen days of HDBR in 55–65-year-old men and women resulted in the development of insulin resistance with amelioration mediated by a high-intensity exercise countermeasure, as manifested by lack of deterioration of the Matsuda index and no change to calculated β-cell function. Benefits imparted by exercise occurred despite increases to FGF-23 and α-klotho, both of which positively correlate with insulin resistance in multiple human cross-sectional studies. Further work is required to elicit the temporal and mechanistic relationships between changes to the α-klotho -FGF-23 axis and development of insulin resistance during periods of sudden and protracted sedentary behaviours. These data highlight the usefulness of the inclusion of an exercise protocol to bed rest to minimize insulin resistance with implications for the health of astronauts during future spaceflight.

Methods

Study protocol

Twenty-two healthy habitually active 55–65-year-old adults (minimum 2.5 h of moderate-to-vigorous aerobic activity per week) were recruited for this 14-day, 6° HDBR study (control: F (5) BMI: 22.1 ± 2.1, M (6) BMI: 25.7 ± 2.2, exercise F (6) BMI: 24.3 ± 3.2, M (5) BMI: 27.5 ± 1.2; Average age: 59 ± 3 years; weight: 70.1 ± 14.2 kg; height: 1.67 ± 0.09 m; BMI: 24.9 ± 3.0 kg/m²). Participants were eligible if they were aged 55–65, physically and mentally healthy, postmenopausal for at least one year (for women), with a BMI of 20–35 kg/m² and height between 158 and 195 cm, and passed psychological and medical screening. Exclusion criteria included ECG abnormalities, anemia, positive status for HIV or hepatitis B/C, abnormal bone density, special diets, sedentary behavior, substance use (including tobacco, marijuana, and specific drugs), implanted devices, and recent participation in other studies. Participants also had to test negative for COVID-19 prior to study start, and allocation followed CSA guidelines, with no blinding among research staff. The study proceeded as previously reported^11,43; briefly, participants were admitted to the Centre of Innovative Medicine, McGill University Health Centre, and randomly assigned following pairing to either the exercise (EX; 6 women) or control (CONT; 5 women) arm. EX performed approximately 1-h of head-down exercise per day, split into three sessions breaking up sedentary time including upper- and lower-body resistive exercises, continual aerobic, progressive aerobic, or high-intensity interval training (HIIT). HIIT was individually prescribed for each participant to achieve 80–90% heart-rate reserve during 11 30-s intervals interspersed by 90 s of active recovery on alternating days over the 14-days of HDBR. Resistance exercises were performed in a supine position using cable-based machines that combined both upper and lower body movements. Images and detailed equipment setups have been previously described in our methodological paper⁴⁴. Exercise intensities and weights were increased based on the borg scale and the capacity of the participant to complete the set. CONT received passive movement physiotherapy and massage. Both groups received 1200 IU/day vitamin D3 oral supplementation throughout the study. Diet was also controlled and adjusted to the total energy expenditure of each participant measured using indirect calorimeter. The study was performed in accordance with the declaration of Helsinki, following ethical approval from McGill University Research Ethics board (IRB00010120) and University of Waterloo Research Ethics Board (ORE No. 40420). All participants provided signed informed consent and were informed of their ability to withdraw from the study with no penalty. The study was registered in ClinicalTrials.gov, NCT04964999.

Oral glucose tolerance testing

Oral glucose tolerance testing was performed two days prior to the start of HDBR (HDT-2), and on days 9 and 13 of HDBR (HDT 9 and HDT 13). Testing was performed first thing in the morning in the fasted state. Participants ate a standardized meal the preceding evening (minimum 10 h between meal and OGTT). In EX, the time between last exercise session and the start of OGTT was at least 12 h. A venous catheter was placed in the antecubital fossa and blood drawn immediately prior to testing. For HDT-2 and HDT 9 only, this draw also included samples used for biochemical analyses presented in this paper. Participants drank 75 g of an oral glucose solution (Trutol, Thermo Fisher Scientific, MA, USA) with subsequent venous sampling timed from the moment the drink was finished. Additional venous blood was drawn at 30, 60, 90, 120 and 180 min for measurement of glucose and insulin concentrations performed at the biochemistry laboratory of the MUHC using standard clinical-processing techniques. Finger-prick glucose measurements were also obtained at these timepoints. Throughout testing, participants remained supine or in HDBR and did not perform any activity.

Calculation of OGTT-derived indices

Glucose and insulin area under the curve (AUC) were calculated according to the trapezoidal rule. Two measurements were taken prior to consumption of the glucose drink, averaged together, and used as the baseline. All positive areas above the baseline level were summed to calculate positive incremental AUC. For glucose AUC calculations, glucose concentrations from finger-prick measurements were substituted for venous concentrations when unavailable (8 samples total). The static HOMA-IR and QUICKI indices were calculated from the pre-test blood draw alone, whereas the dynamic Matsuda index was calculated with data obtained at all OGTT timepoints, as previously published⁴⁰. We also calculated insulin sensitivity index (ISI_0,120) as previously validated by Gutt et al⁴⁵, a measurement that correlates with whole-body glucose uptake obtained during hyperglycaemic-euglycaemic clamp experiments. Briefly, measurements of glucose (mg/l) and insulin (mU/l) concentrations prior to and 120 min after finishing the OGTT drink were incorporated into the following series of equations in order to estimate insulin sensitivity using data available to us:

• 1. ${\text{m}} = \frac{{\left( {75,000\;{\text{mg }} + { }\left( {0\;\min \;{\text{glucose}}_{{\left( {{\text{mg}}/{\text{l}}} \right)}} - 120\;\min \;{\text{glucose}}_{{\left( {{\text{mg}}/{\text{l}}} \right)}} } \right)*0.19*{\text{bodyweight}}_{{\left( {{\text{kg}}} \right)}} } \right)}}{{120\;{\text{min}}}}$

• 2. ${\text{M}} = \frac{{\text{m}}}{{{\raise0.7ex\hbox{${(0\;\min \;{\text{glucose}}_{{\left( {{\text{mg}}/{\text{l}}} \right)}} + 120\;\min \;{\text{glucose}}_{{\left( {{\text{mg}}/{\text{l}}} \right)}} )}$} \!\mathord{\left/ {\vphantom {{(0\;\min \;{\text{glucose}}_{{\left( {{\text{mg}}/{\text{l}}} \right)}} + 120\;\min \;{\text{glucose}}_{{\left( {{\text{mg}}/{\text{l}}} \right)}} )} 2}}\right.\kern-0pt} \!\lower0.7ex\hbox{$2$}}}}$

• 3. ${\text{ISI}}_{0,120} = \frac{{{\text{MCR}}}}{{\log {\raise0.7ex\hbox{${(0\min {\text{insulin}}_{{\left( {mU/l} \right)}} + 120\min {\text{insulin}}_{{\left( {mU/l} \right)}} )}$} \!\mathord{\left/ {\vphantom {{(0\min {\text{insulin}}_{{\left( {mU/l} \right)}} + 120\min {\text{insulin}}_{{\left( {mU/l} \right)}} )} 2}}\right.\kern-0pt} \!\lower0.7ex\hbox{$2$}}}}$

Finally, we assessed pancreatic β-cell function using the BIGTT-AIR_(0,60,120) index that incorporates sex and BMI alongside glucose and insulin levels prior to, 60-min, and 120-min after ingestion of the OGTT drink. This index has previously been correlated closely to intravenous assessments of β-cell function^46,47.

Biomarker testing

We performed specific biomarker investigations at HDT-2 and HDT 9, and on the morning of the last day of HDBR prior to any upright posture (HDT 15) for FGF-23 and α-klotho alone. We adhered to all manufacturer’s protocols for ELISA analyses of serum intact FGF-23 (CY-4000 kit, Kainos Laboratories Inc., Tokyo, Japan), plasma human soluble α-klotho (No. 27998, IBL-America, Minneapolis, USA), uncarboxylated osteocalcin (LEGEND MAX™ Human ucOCN 446707 Kit, The BioLegend Inc., San Diego, USA) and intact osteocalcin (ALPCO, 43-OSNHU-E01 kit, New Hampshire, USA). For α-klotho testing, plasma samples were tested following the first thaw. Phosphate was measured as orthophosphate, total serum iron, 25-hydroxyvitamin D and 1,25-dihydrovitamin D, IGF-1 and hsCRP were conducted at the MUHC Biochemistry Laboratory.

Muscle GLUT1 and GLUT4 quantification

Biopsies of the right Vastus lateralis (VL) muscle were performed on day 1 and 14 of HDBR (HDT 1 and HDT 14) under local anesthesia and samples placed into liquid nitrogen and stored at − 80 °C until analysed. In EX participants, biopsies were obtained prior to any exercise that day, and at least 12 h after the previous exercise session. 30–50 mg of muscle tissue was homogenized in 10 volumes of an extraction buffer (tris base 50 mM, NaCl 150 mM, triton X-100 1%, sodium deoxycolate 0.5%, SDS 0.1% and 10 ug/ml of a protease inhibitor cocktail (Sigma P8340)). The homogenate was centrifuged at 15,000g for 15 min at 4 °C. Protein content in the supernatant was determined using the Bradford method. Aliquots of supernatant were mixed with Laemli buffer and subsequently boiled at 95 °C for 5 min. Approximately 20 µg of proteins were loaded onto gradient (4–15%) stain-free polyacrylamide gels (Mini PROTEAN® TGX Stain-Free TM Gels, Biorad), separated by electrophoresis, and then transferred to polyvinylidene fluoride membranes (PVDF, Biorad). Following blocking in PBS + 1% Tween® 20 + 5% BSA for 1 h at room temperature, samples were incubated with primary antibodies specific for GLUT1 (ab115730; dilution 1:1000, Abcam, Cambridge, UK) and GLUT4 (ab33780; dilution 1:1000, Abcam, Cambridge, UK) for 1-h. Following washing and incubation with HRP-conjugated secondary antibodies, signals were detected using enhanced chemiluminescence substrate (Clarity ECL substrate #170-5060, Biorad) and the ChemiDocTM Touch Imaging System. Images were analyzed with ImageLab software (Biorad). Enhanced chemiluminescence signals were normalized to the intensity of the stain-free blot image.

Statistical testing

The research and analysis were performed in an un-blinded manner. All statistical testing was performed using R. Outcomes were compared with linear mixed models to assess fixed effects of group (EX vs CONT) and HDBR timepoint, in addition to the random effect of individual participant. Due to the limited number of individuals in each group and losses of data within datasets, sex was not included as an independent variable for most of our analyses; for FGF-23 and α-klotho, sex was included as we obtained complete datasets for all timepoints. When statistically significant differences were identified, effect size was calculated using Cohen’s d analysis. Post-hoc testing was employed using pairwise comparisons of estimated marginal means with Holm correction. When assessing for the potential impact of biomarkers on OGTT outcomes, the particular biomarker of interest was included as an additional fixed effect within the model. Repeated measures correlations with ‘rmcorr’ investigated responses within each group without violating assumptions of interdependence. In all circumstances, statistical significance was reported when p ≤ 0.05.