You are on page 1of 17

Article

Divergent remodeling of the skeletal muscle


metabolome over 24 h between young, healthy men
and older, metabolically compromised men
Graphical abstract Authors
Jan-Frieder Harmsen,
Michel van Weeghel, Rex Parsons, ...,
Matthijs K.C. Hesselink,
Riekelt H. Houtkooper,
Patrick Schrauwen

Correspondence
r.h.houtkooper@amsterdamumc.nl
(R.H.H.),
p.schrauwen@maastrichtuniversity.nl
(P.S.)

In brief
Harmsen et al. examine the skeletal
muscle metabolome over 24 h comparing
serial tissue samples from young, healthy
men versus older, metabolically
compromised men. Different abundance
levels of most metabolites around the
clock, as well as different temporal
dynamics, are evident between groups.
The night period demonstrates the
Highlights largest disparity.
d The skeletal muscle metabolome vastly changes over 24 h

d Most muscle metabolites are less abundant in older,


metabolically compromised men

d 24 h rhythms differ between young, healthy men versus older,


metabolically compromised men

d The largest differences between groups are identified during


the night

Harmsen et al., 2022, Cell Reports 41, 111786


December 13, 2022 ª 2022 The Author(s).
https://doi.org/10.1016/j.celrep.2022.111786 ll
ll
OPEN ACCESS

Article
Divergent remodeling of the skeletal muscle
metabolome over 24 h between young, healthy men
and older, metabolically compromised men
Jan-Frieder Harmsen,1 Michel van Weeghel,2,3 Rex Parsons,4 Georges E. Janssens,2 Jakob Wefers,1 Dirk van Moorsel,5
Jan Hansen,1 Joris Hoeks,1 Matthijs K.C. Hesselink,1 Riekelt H. Houtkooper,2,* and Patrick Schrauwen1,6,*
1Department of Nutrition and Movement Sciences, NUTRIM School of Nutrition and Translational Research in Metabolism, Maastricht

University Medical Center, 6200 MD Maastricht, the Netherlands


2Laboratory Genetic Metabolic Diseases, Amsterdam Gastroenterology, Endocrinology, and Metabolism, Amsterdam Cardiovascular

Sciences, Amsterdam UMC, University of Amsterdam, Meibergdreef 9, 1105 AZ Amsterdam, the Netherlands
3Core Facility Metabolomics, Amsterdam UMC, University of Amsterdam, Amsterdam, the Netherlands
4Australian Centre for Health Services Innovation and Centre for Healthcare Transformation, School of Public Health and Social Work, Faculty

of Health, Queensland University of Technology, Kelvin Grove, QLD, Australia


5Division of Endocrinology, Department of Internal Medicine, Maastricht University Medical Center, 6200 MD Maastricht, the Netherlands
6Lead contact

*Correspondence: r.h.houtkooper@amsterdamumc.nl (R.H.H.), p.schrauwen@maastrichtuniversity.nl (P.S.)


https://doi.org/10.1016/j.celrep.2022.111786

SUMMARY

24 h whole-body substrate metabolism and the circadian clock within skeletal muscle are both compromised
upon metabolic disease in humans. Here, we assessed the 24 h muscle metabolome by serial muscle sam-
pling performed under 24 h real-life conditions in young, healthy (YH) men versus older, metabolically
compromised (OMC) men. We find that metabolites associated with the initial steps of glycolysis and hexos-
amine biosynthesis are higher in OMC men around the clock, whereas metabolites associated with gluta-
mine-alpha-ketoglutarate, ketone, and redox metabolism are lower in OMC men. The night period shows
the largest number of differently expressed metabolites. Both groups demonstrate 24 h rhythmicity in half
of the metabolome, but rhythmic metabolites only partially overlap. Specific metabolites are only rhythmic
in YH men (adenosine), phase shifted in OMC men (cis-aconitate, flavin adenine dinucleotide [FAD], and uri-
dine diphosphate [UDP]), or have a reduced 24 h amplitude in OMC men (hydroxybutyrate and hippuric acid).
Our data highlight the plasticity of the skeletal muscle metabolome over 24 h and large divergence across the
metabolic health spectrum.

INTRODUCTION hydrocarbon receptor nuclear translocator-like 1 (BMAL1) acting


as transcriptional activators of cryptochrome 1 and 2 (Cry1–2)
In lean individuals, skeletal muscle constitutes about 40% of and period 1, 2, and 3 (Per1–3) genes, whose encoded proteins
human body mass and is the largest contributor to whole-body repress CLOCK:BMAL1 with a recurring periodicity of 24 h.7
energy expenditure.1,2 In response to metabolic demand (resting Accordingly, relevant proportions of the human skeletal muscle
versus physically active) and nutrient status (fasted versus fed), transcriptome (10% of transcripts8), lipidome (12%–20% of
healthy skeletal muscle can rapidly modulate its substrate stor- lipids9,10), and metabolome (4%–7% of metabolites11) display
age and adjust its substrate oxidation to glucose or fat availabil- 24 h rhythmicity. Of note, 24 h rhythms in skeletal muscle meta-
ity.3,4 Over the 24 h cycle, skeletal muscle metabolism switches bolism are not only regulated by the endogenous skeletal muscle
from predominantly carbohydrate oxidation during the fed and clock but are also influenced by the timing of light exposure,
active period to lipid oxidation during the fasted and inactive food intake, and physical activity. Thus, unanticipated occurrence
period.5 This predictable fed-fasting cycle in human skeletal of these factors can alter or blunt 24 h rhythmicity in human meta-
muscle metabolism is reciprocally linked to the endogenous bolism.12,13 Our modern 24 h culture—characterized by eating and
molecular circadian clock present in muscle cells by providing working day and night, reduced sleep, and excessive light expo-
systemic time cues to the clock6 but also receiving reinforce- sure at night—is increasingly being recognized for its negative
ment through the clock.5 impact on metabolic health and associated with an increased
As for other peripheral clocks, the skeletal muscle clock consists risk to develop metabolic diseases such as type 2 diabetes mellitus
of a transcriptional-translational feedback loop with circadian lo- (T2DM).12,13 T2DM is often associated with insulin resistance in
comotor output cycles kaput (CLOCK) and brain and muscle aryl skeletal muscle. Interestingly, Gabriel et al.14 recently found

Cell Reports 41, 111786, December 13, 2022 ª 2022 The Author(s). 1
This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
ll
OPEN ACCESS Article

Table 1. Participant characteristics geted metabolomics analysis was performed to explore mecha-
nisms associated with these large differences in 24 h whole-
Young, healthy Older, metabolically
Parameter (n = 12) compromised (n = 12) body substrate metabolism and within the circadian clock.

Age (years) 22 ± 2 65 ± 9
Mean abundance levels of most metabolites differ
Height (m) 1.80 ± 0.07 1.78 ± 0.05 substantially between OMC and YH men
Body weight (kg) 73 ± 8 96 ± 12 To get an initial impression about global differences in the skeletal
BMI (kg/m2) 22.4 ± 2.0 30.3 ± 2.7 muscle metabolome between YH and OMC men, we first looked at
MEQ-SA score 50 ± 7 57 ± 9 all metabolites with a significant condition effect in the generalized
Fasting plasma 5.3 ± 0.6 5.7 ± 0.4 linear mixed model. This exploratory analysis is equivalent to
glucose (mmol/L) comparing the mean abundance levels around the clock between
Fasting plasma 11.1 ± 4.3 13.8 ± 8.5 groups. Of the 115 analyzed metabolites, 64 metabolites (56%)
insulin (mIU/mL) displayed altered mean abundance levels between groups, and
Data are presented as mean ± SD. MEQ-SA, Morningness-Eveningness the majority of these metabolites had lower levels in OMC men
Questionnaire Self-Assessment Version. Scores of 41 and below indicate than YH men. By assigning all 115 metabolites to their respective
‘‘evening types.’’ Scores of 59 and above indicate ‘‘morning types." superordinate metabolic pathway, we explored overall patterns
Scores between 42 and 58 indicate ‘‘intermediate types.’’ in specific pathways with respect to altered mean abundance
levels of the assigned metabolites between groups (Figure 1A).
From the detected metabolites that are associated with the
altered rhythmicity in clock gene expression of cultured primary tricarboxylic acid (TCA) cycle (Figure 2), cis-aconitate, glutamine,
myotubes from donors with T2DM compared with age- and glutamate, and alpha-ketoglutarate demonstrated lower
BMI-matched normoglycemic controls. Similarly, we demon- mean abundance levels in OMC men (condition: p < 0.001). At
strated that skeletal muscle clock gene expression was altered the mRNA level, Aco2 encoding mitochondrial aconitase 2, which
in in older, metabolically compromised men compared with young, catalyzes the interconversion of citrate to isocitrate via cis-aconi-
healthy men,15 with Per2 and Cry1 having an altered 24 h rhythm tate in the second step of the TCA cycle, also demonstrated lower
and Clock lacking 24 h rhythmicity. Furthermore, 24 h rhythmicity mean expression levels in OMC men (Figure S2B). In contrast, suc-
in whole-body substrate oxidation was blunted in older, metabol- cinate, fumarate, malate, acetyl-CoA, and citrate were not different
ically compromised men compared with young, healthy men.15 between groups, suggesting anaplerotic compensation at the level
Whereas young, healthy men showed a switch from carbohydrate of succinyl-CoA through amino acids in OMC men. Moreover, 2- or
oxidation during the day to lipid oxidation during the night, this 3-hydroxybutyric acid, associated with ketone metabolism, dis-
fed-to-fasted transition in substrate oxidation was severely atten- played lower levels in OMC men (p < 0.001). Creatine and creati-
uated in older, metabolically compromised men. nine were also lower in OMC men (p < 0.01). Associated with the
Since skeletal muscle is the major determinant of whole-body urea cycle, acetylglutamic acid, aspartate, arginine, uric acid,
substrate metabolism, we here aimed to explore how the temporal glutamine, glutamate, and ornithine were lower around the clock
organization of the skeletal muscle metabolome differs between in OMC men (p % 0.001). Several metabolites that are involved
the edges of the metabolic health spectrum. We hypothesized in redox reactions, such as cysteine, glutathione, NADH, NADP+,
that 24 h rhythmicity in the skeletal muscle metabolome is NADPH, and pyroglutamic acid, were less abundant in OMC
disturbed in older, metabolically compromised men compared men (p < 0.05). Nonetheless, the ratios between NAD+/NADH
with young, healthy men. Such temporal changes of muscle me- and ADP/ATP both had higher mean levels in OMC men (p %
tabolites could provide mechanistic insight into how disrupted 0.01; Figure S3).
24 h whole-body substrate metabolism and impaired muscle clock Strikingly, the glycolysis pathway contrasted the otherwise
gene expression emerge. Therefore, we performed semi-targeted consistent global pattern of lower mean metabolite levels in
metabolomics on muscle biopsy samples taken around the clock OMC men (Figure 3). Thus, metabolites associated with the first
from young, healthy and older, metabolically compromised men. steps of glycolysis were higher in OMC men: hexose-phosphate,
dihydroxyacetone-phosphate, D-glyceraldehyde 3-phosphate,
RESULTS and glyceric acid 1,3-biphosphate (p < 0.05). In contrast, metabo-
lites associated with later steps of glycolysis were again lower in
In two separate cross-sectional cohort studies, our group OMC men: 3-phosphoglyceric acid, 2-phosphoglyceric acid,
previously reported how whole-body substrate metabolism and and phosphoenolpyruvate (p % 0.003), suggesting a bottleneck
skeletal muscle metabolism change across 24 h under real-life at the enzymatic conversion of glyceric acid 1,3-biphosphate to
conditions with 3 standardized meals in young, healthy and older, 3-phosphoglyceric acid involving phosphoglycerate kinase 1
metabolically compromised men.15,16 Further comparative partic- (PGK1). Consistent with this suggestion, mRNA levels of Pgk1
ipant characteristics are shown in Table 1. By combining these were lower in OMC men (p < 0.001; Figure 3). However, the end
previously reported data, diurnal differences between young, product of glycolysis, pyruvate, was not different between groups.
healthy (YH) men and older, metabolically compromised (OMC) With respect to glycogenesis, mean abundance levels of uridine
men for the respiratory exchange ratio (RER), core body temper- diphosphate (UDP)-hexose and mean mRNA levels of Gys1
ature, plasma metabolites, and skeletal muscle clock gene encoding glycogen synthase 1 were both lower in OMC men
expression are summarized in Figure S1. The following semi-tar- (Figure S2C), suggesting reduced muscle glycogen synthesis

2 Cell Reports 41, 111786, December 13, 2022


ll
Article OPEN ACCESS

(legend on next page)

Cell Reports 41, 111786, December 13, 2022 3


ll
OPEN ACCESS Article

capacity in OMC men compared with YH men. A relatively minor played 24 h rhythmicity in at least one of the two groups, whereas
branch of glycolysis, the hexosamine biosynthesis pathway, was 45 metabolites were arrhythmic in both (Figure 5A). The peak
also altered in OMC men with its major end product, UDP abundance level occurred at 8 a.m. for most rhythmic metabo-
N-acetylglucosamine (UDP-GlcNAc), being higher in OMC men lites in YH and OMC men, suggesting a superordinate impact
around the clock (p < 0.001). On the contrary, mRNA levels of of the overnight fast and/or continuous lack of physical activity
Gfpt1 encoding the rate-limiting enzyme of this pathway, gluta- during sleep on the muscle metabolome (Figure S4). Notably,
mine-fructose-6-phosphate transaminase 1 (GFAT), were lower most metabolites associated with glycolysis (Figure 3) and the
in OMC men (p < 0.001; Figure S2A). Lastly, mean mRNA levels pentose phosphate pathway (Figure 1A) were arrhythmic,
of Slc2a4 encoding GLUT4, Irs1 encoding insulin receptor sub- suggesting that these pathways are not strictly controlled by
strate 1, and Hk2 encoding hexokinase 2 were all lower in OMC the molecular clock and/or by environmental rhythmicity such
men (Figures 4A–4C), in line with reduced insulin sensitivity in as human behavior (e.g., feeding/fasting and/or activity/rest cy-
OMC men. cles). In contrast, detected metabolites involved in the TCA cycle
In summary, the initial comparison of mean abundance levels were consistently rhythmic either in both groups (cis-aconitate,
around the clock revealed substantial differences in the skeletal alpha-ketoglutarate, glutamate, fumarate, malate) or at least
muscle metabolome between groups with mostly lower abun- for one group (YH men: acetyl-CoA and succinate; OMC men:
dance levels in OMC men, with the exception of those metabo- citrate and transcript levels of Aco2). Most TCA metabolites
lites associated with the initial steps of glycolysis and hexos- shared a similar rhythmic pattern, peaking after the longest
amine biosynthesis metabolism, which were higher in OMC period of fasting and rest at 8 a.m., except for acetyl-CoA
men. Reduced levels of Pgk1 mRNA in OMC men are consistent (peak at 1 p.m.) and citrate (peak at 6 p.m.) (Figure 2).
with a bottleneck in the conversion of the early toward later steps Some metabolites demonstrated 24 h rhythmicity either
in glycolysis in OMC men. only in YH men or only in OMC men
Among the rhythmic metabolites, we first zoomed into those me-
The greatest differences between OMC and YH men are tabolites that were only rhythmic in one of the two groups. Based
present at 4 a.m. on a cosinor-based model (see STAR Methods for details), we
We next explored if the overall differences in metabolite levels found 14 metabolites that were only rhythmic in YH men and
between OMC and YH men could be attributed to specific time 22 metabolites that were only rhythmic in OMC men (see Fig-
points. For this purpose, we compared differences between ure 5B for the complete list of metabolites). For instance, all three
groups over the 5 time points by means of Bonferroni post-hoc branched-chain amino acids (BCAAs)—isoleucine, leucine and
tests. Strikingly, the greatest differences in the muscle metabo- valine—were rhythmic in OMC men, characterized by a peak
lome between groups were found during the night at 4 a.m., with at 4 or 8 a.m. and a consistent trough at 6 p.m. (Figures S5I
33% of detected metabolites (38 of 115) showing differences be- and S5N), suggesting either control by the molecular clock or
tween groups (Figure 1B; Table S1). In the overnight fasted state, by human behavior (e.g., feeding/fasting and/or activity/rest cy-
at 8 a.m., 23 metabolites (20%) were different between groups. cles). However, a similar pattern was observed for BCAAs in YH
In contrast, differences were less pronounced during the post- men that only reached statistical significance for 24 h rhythmicity
prandial daytime, with 16 metabolites (14%) different at 1 p.m. for isoleucine. Therefore, to test which of these metabolites
and 15 metabolites (13%) at 6 p.m. At 11 p.m.—just before par- differed most strongly in rhythmicity between groups, we applied
ticipants went to bed—22 metabolites (19%) were different be- a more stringent analysis using a generalized linear mixed model
tween groups. to further test for interaction effects. This analysis revealed an
Our findings are consistent with the large difference in whole- interaction effect only for a fraction of these 36 metabolites:
body substrate metabolism that we previously observed during 24 h rhythmicity for adenosine was found in YH men with an in-
the night—with YH men switching from carbohydrate oxidation crease at night (peak at 4 a.m.) that differed from the stable
during the day to lipid oxidation during the night, while OMC pattern in OMC men (Figure 5F). Furthermore, 24 h rhythmicity
men maintained carbohydrate oxidation during the night.15 for glycerate, kynurenine, and ophthalmic acid was found only
in OMC men. Glycerate had its peak at 8 a.m. and trough at 11
24 h rhythmicity in the skeletal muscle metabolome is p.m. (Figure 5C), kynurenine had its peak at 11 p.m. and trough
partly altered in OMC men at 1 p.m. (Figure 5D), and ophthalmic acid had its peak at 1 p.m.
Rhythmic versus arrhythmic metabolites and trough at 4 a.m. (Figure 5E). All metabolites that displayed
The serial sampling of muscle tissue over 24 h allowed us to 24 h rhythmicity in only one group without an interaction effect,
analyze 24 h rhythmicity of metabolites. Seventy metabolites dis- and hence no significantly different expression over time, are

Figure 1. Global differences within the skeletal muscle metabolome across the metabolic health spectrum
(A) Heatmap depicting abundance levels of muscle metabolites and their associated pathways comparing young healthy (YH) versus older, metabolically
compromised (OMC) men around the clock. The pseudocolor scaling of the metabolite expression is based on the standard deviation (SD) to the mean level
calculated for each metabolite over both groups with lower levels (blue) to higher levels (red) compared with the mean level. Each row represents a single
metabolite, and each column depicts one of the five time points per group.
(B) Overview of the number of metabolites that differed between YH and OMC men at the respective time point (n = 12 subjects per group). The metabolites that
underlie this figure are listed in Table S1.
*Significantly different mean abundance levels between groups (p < 0.05), condition effect in the generalized linear mixed model.

4 Cell Reports 41, 111786, December 13, 2022


ll
Article OPEN ACCESS

Cytosol Pyruvate
Pyruvate

12
Young healthy

Abundance level
8 Older, metabolically compromised

4
time: p = 0.638
condition: p = 0.103
interaction: p = 0.893
0
8:00 13:00 18:00 23:00 4:00

Mitochondrion
Acetyl-CoA
Acetyl-CoA

0.15
Abundance level

0.10

0.05
time: p = 0.009
condition: p = 0.662
interaction: p = 0.210
0.00
8:00 13:00 18:00 23:00 4:00 Citrate
Citric acid

0.4
Abundance level

0.3
Oxaloacetate
0.2
Malate
Malate (n.d.)
0.1 time: p = 0.011
0.5 time: p < 0.001
condition: p = 0.137
interaction: p = 0.601
condition: p = 0.098
Abundance level

0.4 interaction: p = 0.610 0.0


8:00 13:00 18:00 23:00 4:00
0.3
0.2
0.1
0.0
Cis-aconitate
Cis-Aconitate

8:00 13:00 18:00 23:00 4:00 1.5 *


*
Abundance level

1.0

Fumarate
Fumarate

0.6
TCA 0.5
time: p = 0.013
condition: p = 0.002
Abundance level

interaction: p = 0.030
0.0
0.4
cycle 8:00 13:00 18:00 23:00 4:00

0.2
time: p < 0.001

0.0
condition: p = 0.232
interaction: p = 0.545 Isocitrate (n.d.)
8:00 13:00 18:00 23:00 4:00

D-ketoglutarate
Alpha-Ketoglutarate
Succinate
Succinate
15 time: p < 0.001
0.12
Abundance level

condition: p < 0.001


* interaction: p = 0.225
Abundance level

10 *
0.08
5
0.04
time: p = 0.039
condition: p = 0.851 0
interaction: p = 0.164
0.00 8:00 13:00 18:00 23:00 4:00
8:00 13:00 18:00 23:00 4:00
Glutamate
Glutamate Glutamine
Glutamine

20 250
*
*
succinyl-CoA
Abundance level

Abundance level

15 200

(n.d.) 10
150
100
5 time: p = 0.037
50 time: p = 0.188
condition: p < 0.001 condition: p < 0.001
interaction: p = 0.547 interaction: p = 0.953
0 0
8:00 13:00 18:00 23:00 4:00 8:00 13:00 18:00 23:00 4:00

(legend on next page)

Cell Reports 41, 111786, December 13, 2022 5


ll
OPEN ACCESS Article

shown in Figures S5 and S6. On the transcript level, Gfpt1 mRNA while Per3 had a greater amplitude in OMC men (p = 0.017).
was only rhythmic in OMC men, peaking at 8 a.m., following the Rora was phase shifted between groups (p < 0.001), in line
rhythmic pattern in glucose abundance levels (Figure S2A), but with Per2 and Cry1 (Figure S1).
no interaction effect was found compared with YH men. Collectively, by comparing rhythmic features within the pro-
By focusing on those metabolites that showed 24 h rhythmicity portion of the skeletal muscle metabolome that demonstrated
in only one of the two groups, a lack of rhythmicity for adenosine 24 h rhythmicity in both groups, we identified remarkable differ-
and presence of rhythmicity for glycerate, kynurenine and ences between OMC and YH men with respect to mesor, acrop-
ophthalmic acid was identified in OMC men compared with hase, and amplitude. The lower mesor levels in OMC men were in
YH men. line with the global pattern of mostly lower mean abundance
Many metabolites had 24 h rhythmicity in both groups levels in most detected metabolites irrespective of 24 h
but displayed differences in mesor, amplitude, or rhythmicity.
acrophase
Although the overall presence of 24 h rhythmicity in the 2-Hydroxybutyric acid may contribute to shifting
skeletal muscle metabolome seemed to be comparable be- substrate metabolism in skeletal muscle between
tween OMC and YH men, we hypothesized that specific proper- carbohydrate and lipid oxidation around the clock
ties of 24 h rhythmicity are different between groups. Using the Sato et al.20 recently identified 2-hydroxybutyric acid to play
CircaCompare model,17 we estimated differences in mesor an important and time-of-day-dependent signaling role in
(rhythm-adjusted mean abundance level of the metabolite), systemic and local tissue metabolism. In sedentary mice,
amplitude, and acrophase (time at which the peak of the 24 h 2-hydroxybutyric acid displayed synchronized 24 h rhythmicity
rhythm occurs) in those 38 metabolites with 24 h rhythmicity in in serum, liver, and skeletal muscle that peaked at the beginning
both groups (Figure 5B). Most prominently, 21 out of 38 metab- of the active phase.21 In line with these pre-clinical findings, 2- or
olites displayed a lower mesor level in OMC men (Figure 5B), in 3-hydroxybutyric acid peaked at the beginning of the active
line with the global pattern of lower mean abundance levels in phase after the overnight fast in YH and OMC men (Figure 5F).
OMC men as described earlier, whereas only ADP-ribose dis- Since injecting 2-hydroxybutyric acid into mice has been shown
played a higher mesor level in OMC men (p < 0.001). In OMC to induce a sudden and persistent shift from carbohydrate to lipid
men, a phase shift was detected for cis-aconitate (p = 0.013), oxidation,20 we also correlated abundance levels of 2- or
flavin adenine dinucleotide (FAD; p = 0.001), and UDP (p = 3-hydroxybutyric acid to the RER measured at 4 a.m. in both
0.036) (Figures 5I–5K) when compared with YH men, with FAD groups, when the RER showed the largest difference between
showing the largest phase shift. The only two metabolites with groups.15 Interestingly, 2- or 3-hydroxybutyric acid levels and
an altered amplitude were 2- or 3-hydroxybutyric acid (p = the RER were negatively correlated (Spearman r = 0.637, p =
0.022) and hippuric acid (p < 0.001), which both exhibited a 0.001; Figure S8), and this significant correlation persisted
smaller amplitude in OMC versus YH men (Figures 5G and 5H). even when all five time points were considered (Spearman r =
For nine metabolites (asparagine, CDP, fumarate, histidine, 0.274, p = 0.003). A phase shift and reduced 24 h rhythm
itaconic acid, malate, proline, serine, and taurine), no differences amplitude was also found for both (2- or 3-hydroxybutyric acid:
in mesor, amplitude, or phase were detected between p = 0.078 and p = 0.022; RER: p = 0.043 and p = 0.072).
groups, suggesting either control by the molecular clock or
by diurnal human behavior independent of metabolic health DISCUSSION
status.
For those pathways that showed the most differences Diurnal regulation of whole-body metabolism in response to
between groups, we performed complementary analyses of nutrient status and metabolic demand can be reinforced by
transcript levels. Associated with glucose metabolism, Slc2a4, circadian clocks located in peripheral tissues. In the metaboli-
Irs1, Hk2 (Figures 4A–4C), and Gys1 (Figure S2C) were all rhyth- cally compromised state, the endogenous clock within skeletal
mic in both groups with lower mesor levels in OMC men muscle is also compromised,14,15 accompanied by the blunted
(p < 0.001), except for Hk2. With regard to substrate oxidation, ability of whole-body substrate metabolism to switch from car-
mRNA levels of Ppard encoding peroxisome proliferator-acti- bohydrate oxidation during the fed, active daytime period to
vated receptor delta and Pdk4 encoding pyruvate dehydroge- lipid oxidation during the fasted, inactive night-time period.15
nase kinase 4, both known to promote mitochondrial fatty acid Here, we explored how the temporal organization of the skeletal
oxidation over glucose oxidation,18,19 demonstrated 24 h rhyth- muscle metabolome differs between the edges of the metabolic
micity, with Ppard peaking at 6 p.m. and Pdk4 strongly peaking health spectrum. By serial sampling of human skeletal muscle
at 4 a.m., but no differences were found between groups tissue over a 24 h cycle in two distinct cohorts, we provide a
(Figures 4D and 4E). We also analyzed mRNA of additional clock detailed comparative analysis of the muscle metabolome
genes that were not analyzed in our initial study (i.e., Per1, Per3, between YH and OMC men. We demonstrate that the temporal
and Rora; Figure S7). Per1 was unchanged between groups, organization of the muscle metabolome in OMC men diverges

Figure 2. Differences in metabolites associated with the tricarboxylic acid (TCA) cycle
Abundance levels of metabolites are shown for YH (blue dots and line) and OMC men (red dots and line). Data are presented as mean ± SEM; n = 12 subjects per
group. *Significantly different between groups (p < 0.05), condition effect in the generalized linear mixed model followed up by Bonferroni post-hoc tests at each
time point; n.d. refers to metabolites that were not detected.

6 Cell Reports 41, 111786, December 13, 2022


ll
Article OPEN ACCESS

Figure 3. Differences in metabolites involved in glycolysis


Abundance levels of metabolites are shown for YH (blue dots and line) and OMC men (red dots and line). Pgk1 mRNA expression, encoding for phosphoglycerate
kinase 1, which converts glyceric acid 1,3-biphosphate to 3-phosphoglyceric acid, is shown to complement the metabolomics data.
Data are presented as mean ± SEM; n = 12 subjects per group. *Significantly different between groups (p < 0.05), condition effect in the generalized linear mixed
model followed up by Bonferroni post-hoc tests at each time point.

from YH men, with the greatest differences present at 4 a.m. have previously shown15 (Figure S1). Interestingly, 24 h rhyth-
We found that some metabolites associated with glycolysis micity of adenosine was markedly different in YH men compared
and hexosamine biosynthesis metabolism demonstrated with the stable pattern in OMC men. Adenosine can modify circa-
higher levels in OMC men around the clock, whereas metabo- dian clockwork through adenosine A1/A2A receptor signaling and
lites associated with the urea cycle and glutamine-alpha- downstream Ca2+ release regulating the clock genes Per1 and
ketoglutarate, ketone, and redox metabolism were lower in Per2.22 We here demonstrate that adenosine follows a 24 h
OMC men. rhythm in skeletal muscle from YH donors with highest levels at
Beyond classifying differences between cohorts in a dichoto- the beginning of the inactive period (11 p.m. and 4 a.m.). Since
mous manner (i.e., lower versus higher levels), the great advan- interstitial adenosine levels within skeletal muscle rise in response
tage of serial muscle sampling around the clock is that temporal to contractions/exercise,23 potential reductions in physical activity
organization patterns evolve. Thereby, we could classify muscle in OMC men could explain the lack of 24 h rhythmicity in adeno-
metabolites as being rhythmic over 24 h and do so separately in sine, although in the current study, YH and OMC men were on
each group. Based on this analysis, we identified several metab- the same fixed activity schedule during the day of muscle sam-
olites that were only rhythmic in YH men (14 metabolites). These pling. It remains to be investigated why glycerate, kynurenine,
are of particular interest since their lack of rhythmicity in OMC and ophthalmic acid were the only metabolites with the presence
men may contribute to the impaired 24 h rhythmicity in substrate of a 24 h rhythm in OMC men that strongly differed from the stable
metabolism and/or skeletal muscle clock gene expression, as we pattern in YH men.

Cell Reports 41, 111786, December 13, 2022 7


ll
OPEN ACCESS Article

A B C

D E

Figure 4. Transcript levels of key regulators of glucose and lipid metabolism show 24 h rhythms and differ across the metabolic health
spectrum
Additional mRNA expression analysis between YH (blue dots and line) and OMC men (red dots and line): (A) Slc2a4 encoding GLUT4; (B) Irs1 encoding insulin
receptor substrate 1; (C) Hk2 encoding hexokinase 2; (D) Ppard encoding peroxisome proliferator-activated receptor delta; and (E) Pdk4 encoding pyruvate
dehydrogenase kinase 4. Note that all five genes showed 24 h rhythmicity in their expression pattern; no differences in rhythmicity were found except for Slc2a4
and Irs1, which had a lower mesor in OMC men.
Data are presented as mean ± SEM; n = 12 subjects per group. *Significantly different between groups (p < 0.05), p values are derived from the generalized linear
mixed model followed up by Bonferroni post-hoc tests at each time point.

From the 38 metabolites that displayed 24 h rhythmicity in both this, we found that transcript levels of Pgk1, which encodes
groups, only 9 metabolites did not differ in their rhythmic proper- the enzyme PGK1 converting glyceric acid 1,3-biphosphate to
ties, whereas the vast majority (21 metabolites) displayed lower 3-phosphoglyceric acid, were lower in OMC men. Acute sleep
mesor levels in OMC men, in line with the global pattern of loss has been shown to reduce levels of PGK1 and other glyco-
lower mean abundance levels of most metabolites independent lytic enzymes and increase muscle protein breakdown.26 The
of 24 h rhythmicity. Moreover, we found phase shifts for cis-aco- lower levels of most amino acids and the changes within the
nitate, FAD, and UDP. In particular, the identification of FAD glycolysis pathway in OMC men may therefore mimic the skel-
being phase shifted is intriguing since the connection between etal muscle metabolic profile after acute sleep loss, suggesting
redox homeostasis, involving FAD, FADH2, NAD+, NADH, that circadian misalignment plays a role in both metabolic
NADP+, and NADPH, and the circadian clock is established.13 phenotypes. Despite the differences in glycolytic intermediates,
Previously, FAD has been shown to display circadian rhythmicity pyruvate levels did not differ between OMC and YH men. How-
in mouse liver.24 Furthermore, through an FAD binding pocket, ever, it should be noted that pyruvate is not solely dependent
FAD can stabilize cryptochrome proteins CRY1 and CRY2, on glycolysis but can also be formed by the degradation of amino
with FAD levels usually increasing right before the accumulation acids such as alanine, serine, cysteine, glycine, threonine, and
of CRY proteins. FAD treatment can even lengthen the circadian tryptophan. Consistently, we found lower levels of cysteine
period of cell cultures in a dose-dependent manner.24 As CRY and tryptophan in OMC men, suggesting that a lower glycolytic
proteins inhibit glucagon-inducible genes and thereby gluconeo- output of pyruvate could be compensated by a greater degree
genesis,25 FAD can indirectly influence glycolytic pathways. of amino acid degradation to maintain pyruvate supply to the
Thus, our finding of a phase shift in FAD, paralleled by the TCA cycle.
phase shift in Cry1 gene expression for OMC muscle, suggests Next to changes in the glycolysis and TCA pathway, upregula-
a changed temporal organization of glycolytic pathways tion of the hexosamine synthesis pathway (HSP)27 and defi-
compared with YH muscle, which could contribute to the greater ciency in scavengers of reactive oxygen species (e.g., gluta-
abundance in metabolites of the first steps of glycolysis thione)28 have both been previously suggested as contributing
(Figure 3). factors to reduced metabolic health. In line with this notion, our
In this regard, we found contrasting differences between findings in human skeletal muscle highlight that higher hexos-
groups with respect to the initial versus later steps of the glycol- amine and lower glutathione levels persist throughout the day
ysis pathway. While OMC men demonstrated higher mean abun- in OMC men. Enhanced flux through HSP requires sustained
dance levels of most glycolytic metabolites from glucose until high glucose concentrations, which is why this pathway has
glyceric acid 1,3-biphosphate, lower levels were found for also been proposed to function as a cellular nutrient sensor.27
3-phosphoglyceric acid until phosphoenolpyruvate. In line with Originating out of the conversion of glucose and glutamine,

8 Cell Reports 41, 111786, December 13, 2022


ll
Article OPEN ACCESS

A B

C D E

F G H

I J K

(legend on next page)

Cell Reports 41, 111786, December 13, 2022 9


ll
OPEN ACCESS Article

the end product of HSP, UDP-GlcNAc, is the donor of in metabolically compromised volunteers and may have a stron-
N-acetylglucosamine to the hydroxyl group of serine and threo- ger impact on nocturnal metabolism by inducing a more pro-
nine residues in specific proteins, initiating the process known as nounced fasting state during the night. More studies are needed
O-GlcNAcylation. Muscle-specific overexpression of the rate- to investigate if such interventions exert their beneficial metabolic
limiting enzyme of the HSP, glutamine-fructose-6-phosphate effects by influencing the nocturnal skeletal muscle metabolome.
aminotransferase (GFPT1), results in insulin resistance in In an attempt to determine the confounding effect of aging, we
mice.29 On the transcript level, Gfpt1 mRNA levels were lower compared our data with recently published human skeletal mus-
in OMC men despite higher levels of UDP-GlcNAc. While such cle metabolomics data from our group,38 for which we used the
protein O-GlcNAcylation is thus known to induce changes in same metabolomics approach and platform ensuring compara-
gene expression in a way that facilitates the development of bility to our current data in terms of identification of metabolites.
T2DM,27 it can also directly regulate the transcriptional activity In that study, we compared the skeletal muscle metabolome in
and stability of BMAL1, CLOCK, and PER2 with consequences the overnight fasted state (i.e., one single time point) between
for entrainment and the period length of the circadian clock.30– YH men (23 ± 2 years, BMI: 22.7 ± 3.4 kg/m2) and healthy, older
32
Concurrently, depleted levels of glutathione throughout the adults (71 ± 4 years, BMI: 25.8 ± 3.3 kg/m2). By comparing the
day render the skeletal muscle susceptible to mitochondrial log10 p value of the difference in metabolites between those
damage from oxidative stress.28 Taken together, the upregu- healthy, older and healthy, young men with the difference in me-
lated HSP and affected glutathione metabolism, evident in tabolites between our OMC and YH men at the 8 a.m. time point,
OMC men around the clock, may contribute to the disturbed we were able to identify muscle metabolites that change due to
24 h rhythmicity in the muscle molecular clock and mitochondrial aging, while others seem to change rather due to the metabolic
function as observed in OMC men.15 However, our observational impairment (Figure S9). Overall, linear regression analysis on
study does not allow us to determine cause and effect. There- these data was indicative of a negative correlation between the
fore, future pre-clinical studies are needed to test these two comparisons (Pearson r = 0.33, p < 0.001; Figure S9A),
hypotheses. suggesting rather opposite changes within the fasted muscle
The skeletal muscle metabolome of OMC men differed the metabolome between metabolically healthy versus metaboli-
most from YH men during the night at 4 a.m., coinciding with cally compromised aging. On the one hand, higher levels in
the time of the designated metabolic switch from carbohydrate glycerate, taurine, dihydoxyacetone-phosphate, and ophthalmic
to lipid oxidation that was present in YH men but blunted in acid and lower levels in hydroxyphenyllactic acid, NADH,
OMC men.15 At this time, the largely disrupted muscle metabo- NADP+, 2- or 3-hydroxybutyric acid, UDP-hexose, creatinine,
lome might fail to create the necessary conditions to facilitate tryptophan, beta-alanine, and aspartate were consistent in
the metabolic switch. Complementing mechanistic animal both older groups, suggesting modulation with aging (Fig-
work,20 we provide some evidence for a link between 2- or ure S9B). However, most metabolites demonstrated opposite
3-hydroxybutyric acid and human substrate metabolism. The directional changes between the two comparisons. In particular,
reduced levels of hydroxybutyric acid in OMC men upon over- metabolites associated with glycolysis, like, e.g., glucose, ADP-
night fasting could be the result of attenuated ketogenesis in ribose, and UDP-GlcNAc, tended to be lower in healthy, older
the liver and hence a reduced systemic supply of ketone bodies, adults while higher in OMC men (Figure S9B). In addition,
further illustrating that OMC men do not enter a fully fasted state. oxidized glutathione, ornithine, alpha-ketoglutarate, and lysine
In support of this, the increase of plasma free fatty acid levels in (Figure S9B) were higher in healthy, older adults while being
response to the overnight fast was also markedly attenuated in lower in OMC men. Therefore, the direction of change in these
OMC men compared with YH men (Figure S1E).15 Taken metabolites is likely to be linked to the metabolically compro-
together, the metabolic impact of the overnight fast on whole- mised state.
body substrate oxidation and on the skeletal muscle metabolome An important finding of our study, with practical implications, is
is blunted in metabolically compromised individuals. Therefore, that a single time point (i.e., overnight fasted), as is routinely used
the fasted night/sleep period is a time window of opportunity in metabolic cross-sectional studies,39 would have revealed only
for targeted intervention. In this context, early time-restricted 23 metabolites as different between groups (i.e., at 8 a.m.),
eating (TRE) and afternoon to evening exercise could be prom- whereas serial muscle sampling around the clock revealed 33
ising interventional tools to influence nocturnal skeletal muscle additional metabolites to be different, underscoring the gain in
metabolism. Both early TRE33,34 and afternoon to evening information through around-the-clock sampling. In particular,
exercise35–37 have been shown to improve metabolic outcomes none of the glycolytic metabolites would have been identified

Figure 5. 24 h rhythms differ across the metabolic health spectrum


The comparison of 24 h rhythms in skeletal muscle metabolites between YH (blue dots and line) and OMC men (red dots and line). Out of 115 detected me-
tabolites, 70 metabolites displayed 24 h rhythmicity in at least one of the two groups, whereas 52 were arrhythmic in both groups (A). Out of these rhythmic
metabolites, 38 metabolites were rhythmic in both groups, 15 were only rhythmic in YH men, and 17 were only rhythmic in OMC men (B). For the metabolites that
are underscored in (B), an interaction effect was found between groups: glycerate (C); kynurenine (D); ophthalmic acid (E); adenosine (F); 2- or 3-hydroxybutyric
acid (G); hippuric acid (H); cis-aconitate (I); FAD (J); and UDP (K). The solid line represents the fitted cosinor-based model and is only displayed for metabolites and
groups where the data proved to be rhythmic. p values in (G)–(K) are derived from the nonlinear mixed-effects regression model CircaCompare;17 see STAR
Methods for more information. To better visualize 24 h rhythmicity, the x axis indicating clock time has been extended beyond the 5 time points of muscle
sampling (8–4 a.m.) to allow double plotting. Data are presented as mean ± SEM; n = 12 subjects per group.

10 Cell Reports 41, 111786, December 13, 2022


ll
Article OPEN ACCESS

to be different between groups if only the overnight fasted time tions in metabolites associated with the initial steps of glycolysis
point had been taken into account. Moreover, for skeletal muscle as well as with HSP that demonstrated higher levels in OMC
metabolites with 24 h rhythmicity, past and future reports about men. Our findings underscore the relevance of taking timing for
decreased or increased levels in metabolomics studies with sin- targeted intervention into account. Changes within the temporal
gle time point analyses should be interpreted with caution since organization of the skeletal muscle metabolome could contribute
these changes could also be explained by phase shifts in the to the disturbed molecular clock in metabolically compromised
respective metabolites. Our analysis of the skeletal muscle me- individuals and their blunted ability for metabolic switching be-
tabolome around the clock in two distinct cohorts with respect tween carbohydrate and lipid oxidation over the 24 h cycle.
to metabolic health status can be used as an informative
resource for future studies to estimate if 24 h rhythmicity has STAR+METHODS
to be taken into account for sampling and analyzing the metab-
olite of interest in humans. Detailed methods are provided in the online version of this paper
and include the following:
Limitations of the study
The following limitations of the current study should be taken into d KEY RESOURCES TABLE
account for the interpretation of findings and for designing d RESOURCE AVAILABILITY
follow-up trials: first, differences between the groups at both B Lead contact

sides of the metabolic health spectrum can be attributed to dif- B Materials availability

ferences in metabolic state but also to differences in age, which B Data and code availability

by itself is also known to affect metabolic health. Since no 24 h d EXPERIMENTAL MODEL AND SUBJECT DETAILS
data are available in age-matched controls, confounding due B Participants

to aging cannot be ruled out. Secondly, our metabolomics B Study protocol overview

analysis was performed on skeletal muscle biopsies, which is a d METHOD DETAILS


naturally diverse sample type, in particular in terms of muscle B Metabolomics analysis

fiber type and intracellular heterogeneity (e.g., cytosolic versus B Gene transcript quantification

mitochondrial subcellular location cannot be distinguished), d QUANTIFICATION AND STATISTICAL ANALYSIS


both between and within individuals over time. In particular, it re-
SUPPLEMENTAL INFORMATION
mains to be investigated if differences in metabolite abundance
levels or rhythmicity properties were due to changes in subcellu- Supplemental information can be found online at https://doi.org/10.1016/j.
lar metabolite pools. Thirdly, our study was restricted to male celrep.2022.111786.
participants; how temporal dynamics within the skeletal muscle
metabolome are influenced by sex, and how the metabolome ACKNOWLEDGMENTS
changes across the metabolic health spectrum in females, re-
quires future investigation. In general, it should also be high- This work is partly financed by the Netherlands Organisation for Scientific
Research (TOP 40-00812-98-14047 to P.S.). We acknowledge support from
lighted that our human trial was designed to mimic real-life con-
the Netherlands Cardiovascular Research Initiative: an initiative with the sup-
ditions with three typical meals and some physical activity port of the Dutch Heart Foundation (CVON2014-02 ENERGISE). Work in the
throughout daytime. As a consequence, we suspect that the Houtkooper group is financially supported by a grant from the Velux Stiftung
high proportion of rhythmic metabolites evident irrespective of (no. 1063).
the metabolic health status is likely linked to the superordinate
impact of the overnight fast and/or continuous lack of physical AUTHOR CONTRIBUTIONS
activity during sleep. To better isolate the role of the endogenous
J.-F.H., J.W., D.v.M., J. Hansen, J. Hoeks, R.H.H., and P.S. designed the
circadian clock on 24 h rhythms within the muscle metabolome,
research; J.W., D.v.M., and J. Hansen collected data; J.-F.H., M.v.W., and
future trials should make use of ‘‘constant routine’’ protocols that R.P. analyzed data; M.v.W. and R.H.H. provided the analytic tools; R.P. devel-
keep confounding factors such as light, food intake, physical ac- oped software necessary to analysis data; and all authors wrote the
tivity, and wakefulness constant over 24 h, as previously done by manuscript.
Perrin et al.8 when studying the 24 h muscle transcriptome with
participants in a continuous semi-recumbent position. Lastly, DECLARATION OF INTERESTS
our findings are observational; hence, cause and effect relation-
The authors declare no competing interests.
ships cannot be determined.
Received: March 25, 2022
Conclusions Revised: July 28, 2022
In conclusion, through serial sampling of skeletal muscle tissue Accepted: November 15, 2022
around the clock, we found large divergence in the skeletal mus- Published: December 13, 2022
cle metabolome between YH and OMC men that would have
REFERENCES
been mostly missed by a single time point analysis. Indeed,
the muscle metabolome was most different at night. With the 1. Periasamy, M., Herrera, J.L., and Reis, F.C.G. (2017). Skeletal muscle
global pattern of lower mean abundance levels in most metabo- Thermogenesis and its role in whole body energy metabolism. Diabetes
lites throughout the 24 h cycle in OMC men, we identified excep- Metab. J. 41, 327–336.

Cell Reports 41, 111786, December 13, 2022 11


ll
OPEN ACCESS Article
2. Zurlo, F., Larson, K., Bogardus, C., and Ravussin, E. (1990). Skeletal mus- , N., Bakke, S.S., Boekschoten, M.V., Kersten, S., Kase,
19. Feng, Y.Z., Nikolic
cle metabolism is a major determinant of resting energy expenditure. E.T., Rustan, A.C., and Thoresen, G.H. (2014). PPARd activation in human
J. Clin. Invest. 86, 1423–1427. myotubes increases mitochondrial fatty acid oxidative capacity and re-
3. McCarthy, J.J., and Esser, K.A. (2010). Anabolic and catabolic pathways duces glucose utilization by a switch in substrate preference. Arch. Phys-
regulating skeletal muscle mass. Curr. Opin. Clin. Nutr. Metab. Care 13, iol. Biochem. 120, 12–21.
230–235. 20. Sato, S., Dyar, K.A., Treebak, J.T., Jepsen, S.L., Ehrlich, A.M., Ashcroft,
4. Smith, R.L., Soeters, M.R., Wu €st, R.C.I., and Houtkooper, R.H. (2018). S.P., Trost, K., Kunzke, T., Prade, V.M., Small, L., et al. (2022). Atlas of ex-
Metabolic flexibility as an Adaptation to energy resources and require- ercise metabolism reveals time-dependent signatures of metabolic ho-
ments in health and disease. Endocr. Rev. 39, 489–517. meostasis. Cell Metab. 34, 329–345.e8. https://doi.org/10.1016/j.cmet.
2021.12.016.
5. Hodge, B.A., Wen, Y., Riley, L.A., Zhang, X., England, J.H., Harfmann,
B.D., Schroder, E.A., and Esser, K.A. (2015). The endogenous molecular 21. Dyar, K.A., Lutter, D., Artati, A., Ceglia, N.J., Liu, Y., Armenta, D., Jastroch,
clock orchestrates the temporal separation of substrate metabolism in M., Schneider, S., de Mateo, S., Cervantes, M., et al. (2018). Atlas of circa-
skeletal muscle. Skelet. Muscle 5, 17. dian metabolism reveals system-wide coordination and communication
between clocks. Cell 174, 1571–1585.e11.
6. Sardon Puig, L., Pillon, N.J., Näslund, E., Krook, A., and Zierath, J.R.
(2020). Influence of obesity, weight loss, and free fatty acids on skeletal 22. Jagannath, A., Varga, N., Dallmann, R., Rando, G., Gosselin, P., Ebrahim-
muscle clock gene expression. Am. J. Physiol. Endocrinol. Metab. 318, jee, F., Taylor, L., Mosneagu, D., Stefaniak, J., Walsh, S., et al. (2021).
E1–E10. Adenosine integrates light and sleep signalling for the regulation of circa-
dian timing in mice. Nat. Commun. 12, 2113.
7. Gutierrez-Monreal, M.A., Harmsen, J.-F., Schrauwen, P., and Esser, K.A.
(2020). Ticking for metabolic health: the skeletal-muscle clocks. Obesity 23. Langberg, H., Bjørn, C., Boushel, R., Hellsten, Y., and Kjaer, M. (2002). Ex-
28, S46–S54. ercise-induced increase in interstitial bradykinin and adenosine concen-
trations in skeletal muscle and peritendinous tissue in humans.
8. Perrin, L., Loizides-Mangold, U., Chanon, S., Gobet, C., Hulo, N., Iseneg-
J. Physiol. 542, 977–983.
ger, L., Weger, B.D., Migliavacca, E., Charpagne, A., Betts, J.A., et al.
(2018). Transcriptomic analyses reveal rhythmic and CLOCK-driven path- ek, L.J. (2017). FAD regulates
24. Hirano, A., Braas, D., Fu, Y.-H., and Ptác
ways in human skeletal muscle. Elife 7, e34114. CRYPTOCHROME protein stability and circadian clock in mice. Cell
Rep. 19, 255–266.
9. Held, N.M., Wefers, J., van Weeghel, M., Daemen, S., Hansen, J., Vaz,
F.M., van Moorsel, D., Hesselink, M.K.C., Houtkooper, R.H., and Schrau- 25. Zhang, E.E., Liu, Y., Dentin, R., Pongsawakul, P.Y., Liu, A.C., Hirota, T.,
wen, P. (2020). Skeletal muscle in healthy humans exhibits a day-night Nusinow, D.A., Sun, X., Landais, S., Kodama, Y., et al. (2010). Crypto-
rhythm in lipid metabolism. Mol. Metab. 37, 100989. chrome mediates circadian regulation of cAMP signaling and hepatic
gluconeogenesis. Nat. Med. 16, 1152–1156.
10. Loizides-Mangold, U., Perrin, L., Vandereycken, B., Betts, J.A., Walhin,
J.P., Templeman, I., Chanon, S., Weger, B.D., Durand, C., Robert, M., 26. Cedernaes, J., Schönke, M., Westholm, J.O., Mi, J., Chibalin, A., Voisin,
et al. (2017). Lipidomics reveals diurnal lipid oscillations in human skeletal S., Osler, M., Vogel, H., Hörnaeus, K., Dickson, S.L., et al. (2018). Acute
muscle persisting in cellular myotubes cultured in vitro. Proc. Natl. Acad. sleep loss results in tissue-specific alterations in genome-wide DNA
Sci. USA 114, E8565–E8574. methylation state and metabolic fuel utilization in humans. Sci. Adv. 4,
eaar8590.
11. Lundell, L.S., Parr, E.B., Devlin, B.L., Ingerslev, L.R., Altıntasx, A., Sato, S.,
Sassone-Corsi, P., Barrès, R., Zierath, J.R., and Hawley, J.A. (2020). Time- 27. Buse, M.G. (2006). Insulin resistance, and the complications of diabetes:
restricted feeding alters lipid and amino acid metabolite rhythmicity current status. American Journal of Physiology-Endocrinology and Meta-
without perturbing clock gene expression. Nat. Commun. 11, 5142. bolism 290, E1–E8.
12. Mason, I.C., Qian, J., Adler, G.K., and Scheer, F.A.J.L. (2020). Impact of 28. Nguyen, D., Samson, S.L., Reddy, V.T., Gonzalez, E.V., and Sekhar, R.V.
circadian disruption on glucose metabolism: implications for type 2 dia- (2013). Impaired mitochondrial fatty acid oxidation and insulin resistance
betes. Diabetologia 63, 462–472. in aging: novel protective role of glutathione. Aging Cell 12, 415–425.
13. Reinke, H., and Asher, G. (2019). Crosstalk between metabolism and 29. Cooksey, R.C., and McClain, D.A. (2002). Transgenic mice overexpressing
circadian clocks. Nat. Rev. Mol. Cell Biol. 20, 227–241. the rate-limiting enzyme for hexosamine synthesis in skeletal muscle or
14. Gabriel, B.M., Altıntasx, A., Smith, J.A.B., Sardon-Puig, L., Zhang, X., adipose tissue exhibit total body insulin resistance. Ann. N. Y. Acad. Sci.
Basse, A.L., Laker, R.C., Gao, H., Liu, Z., Dollet, L., et al. (2021). Disrupted 967, 102–111.
circadian oscillations in type 2 diabetes are linked to altered rhythmic mito- 30. Ma, Y.-T., Luo, H., Guan, W.J., Zhang, H., Chen, C., Wang, Z., and Li, J.D.
chondrial metabolism in skeletal muscle. Sci. Adv. 7, eabi9654. (2013). O-GlcNAcylation of BMAL1 regulates circadian rhythms in NIH3T3
15. Wefers, J., Connell, N.J., Fealy, C.E., Andriessen, C., de Wit, V., van Moor- fibroblasts. Biochem. Biophys. Res. Commun. 431, 382–387.
sel, D., Moonen-Kornips, E., Jörgensen, J.A., Hesselink, M.K.C., Havekes, 31. Kaasik, K., Kivimäe, S., Allen, J.J., Chalkley, R.J., Huang, Y., Baer, K., Kis-
B., et al. (2020). Day-night rhythm of skeletal muscle metabolism is sel, H., Burlingame, A.L., Shokat, K.M., Ptác ek, L.J., and Fu, Y.H. (2013).
disturbed in older, metabolically compromised individuals. Mol. Metab. Glucose sensor O-GlcNAcylation coordinates with phosphorylation to
41, 101050. https://doi.org/10.1016/j.molmet.2020.101050. regulate circadian clock. Cell Metab. 17, 291–302.
16. van Moorsel, D., Hansen, J., Havekes, B., Scheer, F.A.J.L., Jörgensen, 32. Li, M.-D., Ruan, H.B., Hughes, M.E., Lee, J.S., Singh, J.P., Jones, S.P., Ni-
J.A., Hoeks, J., Schrauwen-Hinderling, V.B., Duez, H., Lefebvre, P., tabach, M.N., and Yang, X. (2013). O-GlcNAc signaling entrains the circa-
Schaper, N.C., et al. (2016). Demonstration of a day-night rhythm in human dian clock by inhibiting BMAL1/CLOCK ubiquitination. Cell Metab. 17,
skeletal muscle oxidative capacity. Mol. Metab. 5, 635–645. 303–310.
17. Parsons, R., Parsons, R., Garner, N., Oster, H., and Rawashdeh, O. (2020). 33. Hutchison, A.T., Regmi, P., Manoogian, E.N.C., Fleischer, J.G., Wittert,
CircaCompare: a method to estimate and statistically support differences G.A., Panda, S., and Heilbronn, L.K. (2019). Time-restricted feeding im-
in mesor, amplitude and phase, between circadian rhythms. Bioinformat- proves glucose tolerance in men at risk for type 2 diabetes: a randomized
ics 36, 1208–1212. crossover trial. Obesity 27, 724–732.
18. Pettersen, I.K.N., Tusubira, D., Ashrafi, H., Dyrstad, S.E., Hansen, L., Liu, 34. Sutton, E.F., Beyl, R., Early, K.S., Cefalu, W.T., Ravussin, E., and Peterson,
X.Z., Nilsson, L.I.H., Løvsletten, N.G., Berge, K., Wergedahl, H., et al. C.M. (2018). Early time-restricted feeding improves insulin sensitivity,
(2019). Upregulated PDK4 expression is a sensitive marker of increased blood pressure, and oxidative stress even without weight loss in men
fatty acid oxidation. Mitochondrion 49, 97–110. with prediabetes. Cell Metab. 27, 1212–1221.e3.

12 Cell Reports 41, 111786, December 13, 2022


ll
Article OPEN ACCESS

35. Mancilla, R., Brouwers, B., Schrauwen-Hinderling, V.B., Hesselink, tion are positively associated with NAD+ abundance in humans. Nat. Ag-
M.K.C., Hoeks, J., and Schrauwen, P. (2021). Exercise training elicits su- ing 2, 254–263. https://doi.org/10.1038/s43587-022-00174-3.
perior metabolic effects when performed in the afternoon compared to
morning in metabolically compromised humans. Physiol. Rep. 8, e14669. 39. Belhaj, M.R., Lawler, N.G., and Hoffman, N.J. (2021). Metabolomics and
lipidomics: expanding the molecular landscape of exercise biology. Me-
36. Moholdt, T., Parr, E.B., Devlin, B.L., Debik, J., Giskeødegård, G., and Haw-
tabolites 11, 151.
ley, J.A. (2021). The effect of morning vs evening exercise training on gly-
caemic control and serum metabolites in overweight/obese men: a rand- 40. Mari, A., Pacini, G., Murphy, E., Ludvik, B., and Nolan, J.J. (2001). A
omised trial. Diabetologia 64, 2061–2076. https://doi.org/10.1007/ model-based method for assessing insulin sensitivity from the oral glucose
s00125-021-05477-5. tolerance test. Diabetes Care 24, 539–548.
37. Savikj, M., Gabriel, B.M., Alm, P.S., Smith, J., Caidahl, K., Björnholm, M.,
41. Hellemans, J., Mortier, G., De Paepe, A., Speleman, F., and Vandesom-
Fritz, T., Krook, A., Zierath, J.R., and Wallberg-Henriksson, H. (2019). Af-
pele, J. (2007). qBase relative quantification framework and software for
ternoon exercise is more efficacious than morning exercise at improving
management and automated analysis of real-time quantitative PCR
blood glucose levels in individuals with type 2 diabetes: a randomised
data. Genome Biol. 8, R19.
crossover trial. Diabetologia 62, 233–237.
38. Janssens, G.E., Grevendonk, L., Perez, R.Z., Schomakers, B.V., de Vogel- 42. Gu, Z., Eils, R., and Schlesner, M. (2016). Complex heatmaps reveal pat-
van den Bosch, J., Geurts, J.M.W., van Weeghel, M., Schrauwen, P., terns and correlations in multidimensional genomic data. Bioinformatics
Houtkooper, R.H., and Hoeks, J. (2022). Healthy aging and muscle func- 32, 2847–2849.

Cell Reports 41, 111786, December 13, 2022 13


ll
OPEN ACCESS Article

STAR+METHODS

KEY RESOURCES TABLE

REAGENT or RESOURCE SOURCE IDENTIFIER


Biological samples
Human muscle samples Van Moorsel et al.16 and Wefers et al.15 N/A
Critical commercial assays
RNeasy tissue kit Qiagen 74104
High capacity RNA -to- cDNA kit ThermoFischer 4387406
HOT FIREPol Probe qPCR mix Bioconnect 08-14-00001
plus (ROX)
Sensimix sybr Hi-ROX Bioline QT 605-05
Deposited data
Raw and analyzed metabolomics https://www.ebi.ac.uk/metabolights/ MTBLS6049
data in MetaboLights MTBLS6049/descriptors
Oligonucleotides
Primers for gene expression This paper N/A
analysis, see Table S2
Software and algorithms
R (v4.1.2) CRAN https://cran.r-project.org/
circacompare (0.1.1) CRAN https://cran.r-project.org/web/
packages/circacompare/index.html
ComplexHeatmap (2.13.1) CRAN https://cran.r-project.org/
ggplot2 (3.3.6) CRAN https://cran.r-project.org/
Biorad CFX manager 3.1 for Biorad https://www.bio-rad.com/
gene expression analysis

RESOURCE AVAILABILITY

Lead contact
Further information and requests for resources should be directed to and will be fulfilled by the lead contact, Patrick Schrauwen (p.
schrauwen@maastrichtuniversity.nl).

Materials availability
This study did not generate new unique reagents.

Data and code availability


d The raw metabolomics data is available via MetaboLights under the identifier MTBLS6049.
d This paper does not report original code.
d Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Participants
Skeletal muscle tissue was obtained around the clock in two previously published studies.15,16 The design of these studies was
essentially identical, apart from the study population, i.e. young healthy men16 vs. older metabolically-compromised men.15 To
establish volunteers with compromised metabolic health, we selected participants with reduced glucose tolerance and insulin sensi-
tivity based on an oral glucose tolerance test and a glucose clearance rate <360 mL/kg/min established insulin resistance.40 The
participant characteristics of the two study populations are listed in Table 1. Both studies were approved by the Ethics Committee
of the Maastricht University Medical Center. All participants provided written informed consent. Both studies were registered at
https://clinicaltrials.gov with identifiers NCT02261168 and NCT03733743, respectively.

e1 Cell Reports 41, 111786, December 13, 2022


ll
Article OPEN ACCESS

Study protocol overview


Detailed descriptions of the study procedures can be found elsewhere (see15,16). In short, after seven days of adherence to a
standardized lifestyle with respect to eating and sleeping times at home, YH and OMC men were admitted to our research unit
for 44 h in total, under standardized conditions mimicking real-life conditions with scheduled physical activity, meals and normal
room lighting conditions. After spending the first day (last meal at 6PM) and night in a respiration chamber, the main test day began
with participants waking up at 7AM. At 8AM, 1PM, 6PM, 11PM and 4AM, indirect calorimetry was performed in a recumbent
position and directly after each measurement, a muscle biopsy was taken from the vastus lateralis muscle. Three standardized meals
were provided immediately after the first three muscle biopsies, i.e. at 9AM, 2PM and 7PM. Energy requirements were calcu-
lated by multiplying the sleeping metabolic rate (assessed during the first night spent in a respiration chamber) with an activity factor
of 1.5. Breakfast accounted for 21 energy% (E%), lunch for 30 E%, and dinner for 49 E%. Daily macronutrient composition was 55 E
% as carbohydrates (3 E% fibers), 31 E% as fat (9 E% saturated), and 14 E% as protein. 2 days preceding the admission to the
research unit, meals were provided in a standardized manner at similar clock times for consumption at home.

METHOD DETAILS

Metabolomics analysis
Corresponding to Janssens et al. (2022), the following internal standards diluted in water were added to approximately 5 mg
of freeze-dried muscle tissue: adenosine-15N5-monophosphate (5 nmol), adenosine-15N5-triphosphate (5 nmol), D4-alanine
(0.5 nmol), D7-arginine (0.5 nmol), D3-aspartic acid (0.5 nmol), D3-carnitine (0.5 nmol), D4-citric acid (0.5 nmol), 13C1-citrulline
(0.5 nmol), 13C6-fructose-1,6-diphosphate (1 nmol), guanosine-15N5-monophosphate (5 nmol), guanosine-15N5-triphosphate
(5 nmol), 13C6-glucose (10 nmol), 13C6-glucose-6-phosphate (1 nmol), D3-glutamic acid (0.5 nmol), D5-glutamine (0.5 nmol), D5-gluta-
thione (1 nmol), 13C6-isoleucine (0.5 nmol), D3-lactic acid (1 nmol), D3-leucine (0.5 nmol), D4-lysine (0.5 nmol), D3-methionine
(0.5 nmol), D6-ornithine (0.5 nmol), D5-phenylalanine (0.5 nmol), D7-proline (0.5 nmol), 13C3-pyruvate (0.5 nmol), D3-serine
(0.5 nmol), D6-succinic acid (0.5 nmol), D5-tryptophan (0.5 nmol), D4-tyrosine (0.5 nmol), D8-valine (0.5 nmol). Subsequently, 1:1
(v/v) methanol:water (to a total volume of 1000 mL) and a 5 mm stainless-steel bead were added to the samples and homogenized
for 5 min at a frequency of 30 times/sec using a TissueLyser II (Qiagen, Hilden, Germany). To each sample 1000 mL of chloroform was
added, thoroughly mixed and centrifuged for 10 min at 18.000g. The top ‘‘polar’’ layer was transferred to a new 1.5 mL tube and dried
using a vacuum concentrator at 60 C. The residues were reconstituted in 100 mL 3:2 (v/v) methanol:water. Metabolites were analyzed
using a Waters Acquity ultra-high performance liquid chromatography system coupled to a Bruker Impact II Ultra-High Resolution
Qq-Time-Of-Flight mass spectrometer. Samples were kept at 12 C during analysis and 5 mL of each sample was injected. Chromato-
graphic separation was achieved using a Merck Millipore SeQuant ZIC-cHILIC column (PEEK 100 3 2.1 mm, 3 mm particle size). Col-
umn temperature was held at 30 C. Mobile phase consisted of (A) 1:9 (v/v) acetonitrile:water and (B) 9:1 (v/v) acetonitrile:water, both
containing 5 mmol/L ammonium acetate. Using a flow rate of 0.25 mL/min, the LC gradient consisted of: 100% B for 0-2 min, reach
0% B at 28 min, 0% B for 28-30 min, reach 100% B at 31 min, 100% B for 31-32 min. Column re-equilibration is achieved at a flow rate
of 0.4 mL/min at 100% B for 32-35 min. MS data were acquired using full scan negative and positive ionization mode over the range of
m/z 50-1200. Data were analyzed using Bruker TASQ software version 2.1.22.3. All reported metabolite intensities were normalized
to their corresponding internal standards based on comparable retention time and responses in the MS as well as dry tissue weight.
Metabolite identification has been based on a combination of accurate mass, (relative) retention times and fragmentation spectra,
compared to the analysis of a library of standards (Sigma-Aldrich MSMLS). General repeatability of metabolite analysis was assessed
for each metabolite using repeated measurements of a pooled sample. Additionally, all peak integrations were manually checked for
quality in each sample, as large natural variance may skew pooled sample results.

Gene transcript quantification


RNA was previously isolated from 10 mg of muscle material by TRIzol lysis (Qiagen, Hilden, Germany). RNA was further purified by the
RNeasy kit from Qiagen (Hilden, Germany). RNA yield was measured using a NanoDrop spectrophotometer (Thermo Fisher Scien-
tific, Waltham, USA). The high-capacity RNA-to-cDNA kit from Applied Biosystems (Foster City, USA) was used for transcribing
100 ng of RNA to cDNA. Transcript abundance was determined using a CFX384 Real-Time System (Biorad-Laboratories, Veenen-
daal, NL). To minimize the variability in reference gene normalization, the geometric mean of two reference genes (RPL26 and RPL0),
which were both stably expressed over time, was used. This geometric mean was used as the internal reference for comparative
gene expression analysis over time and between groups.41 Primers used in this study are listed in Table S2.

QUANTIFICATION AND STATISTICAL ANALYSIS

In total, 115 metabolites were subjected to statistical analyses over time between groups. If certain metabolites were below the
detection limit at single time points for individual subjects, these were treated as missing values. To assess differences in rhythmic
parameters – mesor, amplitude and acrophase – between groups, we applied nonlinear mixed-effects regression using the
R-package for CircaCompare17 with a random-effect for the mesor (k) and a pre-defined period of 24-h. Before the model is fit to

Cell Reports 41, 111786, December 13, 2022 e2


ll
OPEN ACCESS Article

compare the two groups, each group is independently assessed for the presence of rhythmicity, defined as non-zero parameter es-
timate (p < 0.05) for amplitude (⍺) when fitting the following model:

Yb = k + a cosðt  4Þ
In this equation, Y represents the response or outcome variable, k represents the mesor, a rhythm-adjusted mean, a represents
amplitude, t refers to the time in hours, and 4 is the phase. If rhythmicity was observed in both groups for a given metabolite, differ-
ences in mesor, amplitude and acrophase for that metabolite between OMC and YH were modeled.
To determine whether there was an effect of time on metabolite expression and whether this was different between groups, we
conducted a generalized linear mixed model with time (categorical), condition (YH vs. OMC) and their interaction (time x condition)
as fixed effects with random intercepts for subjects. In this model, the interaction effect represents different expression of the respec-
tive metabolite over time between groups (e.g. an increasing vs. a decreasing abundance pattern over time between groups or
consistently different peak time points between groups). This model takes missing values into account, which occurred due to
missing samples or poor muscle tissue quality at a few time points (out of 60 muscle samples per group 5 time points for 12 sub-
jects– 5 samples were missing for OMC and 2 samples for YH). In the absence of interaction, we interpreted the main effects for con-
dition and time, and in the case of statistical significance, we followed up the respective main effects with pairwise comparisons per
time point using Bonferroni post-hoc tests. To better visualize 24h-rhythmicity in figures, the x axis indicating clock time has been
extended beyond the 5 time points of muscle sampling (8AM–4AM) to allow double-plotting. For metabolites that did not demon-
strate 24h-rhythmicity in both groups, the generalized linear mixed model as described above was conducted to compare changes
in metabolites over time between the two groups. mRNA levels were analyzed in the same manner as metabolites. The Spearman
rank correlation was used to establish the relationship between specific metabolites and the RER.
Data are presented as mean ± SEM (standard error of the mean). The level of significance was set to <0.05 for all analyses. Sta-
tistical analyses were performed using the R programming language (version 4.0.4; http://www.r-project.org) and further processed
with the GraphPad Prism 8 software package (GraphPad Software, San Diego, CA, USA). The heatmap (Figure 1) was generated
using the ComplexHeatmap R-package.42

e3 Cell Reports 41, 111786, December 13, 2022

You might also like