Article

Temporal dynamics of the multi-omic

response to endurance exercise training

https://doi.org/10.1038/s41586-023-06877-w MoTrPAC Study Group*

Received: 21 September 2022

Accepted: 16 November 2023 Regular exercise promotes whole-body health and prevents disease, but the underlying
Published online: 1 May 2024 molecular mechanisms are incompletely understood1–3. Here, the Molecular
Transducers of Physical Activity Consortium4 profiled the temporal transcriptome,
Open access
proteome, metabolome, lipidome, phosphoproteome, acetylproteome,
Check for updates ubiquitylproteome, epigenome and immunome in whole blood, plasma and 18 solid
tissues in male and female Rattus norvegicus over eight weeks of endurance exercise
training. The resulting data compendium encompasses 9,466 assays across 19 tissues,
25 molecular platforms and 4 training time points. Thousands of shared and tissue-
specific molecular alterations were identified, with sex differences found in multiple
tissues. Temporal multi-omic and multi-tissue analyses revealed expansive biological
insights into the adaptive responses to endurance training, including widespread
regulation of immune, metabolic, stress response and mitochondrial pathways. Many
changes were relevant to human health, including non-alcoholic fatty liver disease,
inflammatory bowel disease, cardiovascular health and tissue injury and recovery.
The data and analyses presented in this study will serve as valuable resources for
understanding and exploring the multi-tissue molecular effects of endurance training
and are provided in a public repository (https://motrpac-data.org/).

Regular exercise provides wide-ranging health benefits, including after the last exercise bout (Fig. 1a). Sex-matched sedentary, untrained
reduced risks of all-cause mortality1,5, cardiometabolic and neuro- rats were used as controls. Training resulted in robust phenotypic
logical diseases, cancer and other pathologies2,6,7. Exercise affects changes (Extended Data Fig. 1a–d), including increased aerobic capacity
nearly all organ systems in either improving health or reducing disease (VO2 max) by 18% and 16% at 8 weeks in males and females, respectively
risk2,3,6,7, with beneficial effects resulting from cellular and molecular (Extended Data Fig. 1a). The percentage of body fat decreased by 5% in
adaptations within and across many tissues and organ systems3. Vari- males at 8 weeks (Extended Data Fig. 1b), without a significant change
ous ‘omic’ platforms (‘omes’) including transcriptomics, epigenomics, in lean mass (Extended Data Fig. 1c). In females, the body fat percent-
proteomics and metabolomics, have been used to study these events. age did not change after 4 or 8 weeks of training, whereas it increased
However, work to date typically covers one or two omes at a single time by 4% in sedentary controls (Extended Data Fig. 1b). Body weight of
point, is biased towards one sex, and often focuses on a single tissue, females increased in all intervention groups, with no change for males
most often skeletal muscle, heart or blood8–12, with few studies consid- (Extended Data Fig. 1d).
ering other tissues13. Accordingly, a comprehensive, organism-wide, Whole blood, plasma and 18 solid tissues were analysed using genom-
multi-omic map of the effects of exercise is needed to understand the ics, proteomics, metabolomics and protein immunoassay technologies,
molecular underpinnings of exercise training-induced adaptations. with most assays performed in a subset of these tissues (Fig. 1b and
To address this need, the Molecular Transducers of Physical Activity Extended Data Fig. 1e,f). Specific details for each omic analysis are
Consortium (MoTrPAC) was established with the goal of building a provided in Extended Data Fig. 2, Methods, Supplementary Discussion
molecular map of the exercise response across a broad range of tis- and Supplementary Table 1. Molecular assays were prioritized on the
sues in animal models and in skeletal muscle, adipose and blood in basis of available tissue quantity and biological relevance, with the
humans4. Here we present the first whole-organism molecular map gastrocnemius, heart, liver and white adipose tissue having the most
of the temporal effects of endurance exercise training in male and diverse set of molecular assays performed, followed by the kidney,
female rats and provide multiple insights enabled by this MoTrPAC lung, brown adipose tissue and hippocampus (Extended Data Fig. 1e).
multi-omic data resource. Altogether, datasets were generated from 9,466 assays across 211 com-
binations of tissues and molecular platforms, resulting in 681,256
non-epigenetic and 14,334,496 epigenetic (reduced-representation
Multi-omic analysis of exercise training bisulfite sequencing (RRBS) and assay for transposase-accessible chro-
Six-month-old male and female Fischer 344 rats were subjected to matin using sequencing (ATAC-seq)) measurements, corresponding to
progressive treadmill endurance exercise training (hereafter referred to 213,689 and 2,799,307 unique non-epigenetic and epigenetic features,
as endurance training) for 1, 2, 4 or 8 weeks, with tissues collected 48 h respectively.

*A list of authors and their affiliations appears at the end of the paper.

174 | Nature | Vol 629 | 2 May 2024

 a Chronic exercise training Rapid tissue removal and freezing
18 organs +
blood or plasma
Incline × Sedentary
duration ×
speed 1

Weeks of training
Cryo-pulverization, aliquoting and
0 1 2 3 4 5 6 7 8 distribution
2
(Sedentary) Weeks of training
4
VO2max and Tissue 48 h rest
NMR tests collection (wash-out) 8 Chemical analysis

b
CORTEX Genomics
HIPPOC
HYPOTH Epigenomics
DNA methylation (METHYL)

BAT Chromatin accessibility (ATAC)

HEART Gene expression (TRNSCRPT)
BLOOD
LUNG
Proteomics
PLASMA
VENACV
Global protein expression (PROT)
LIVER Post-translational modifications
Phosphorylation (PHOSPHO)
SPLEEN
Acetylation (ACETYL)
SMLINT Ubiquitination (UBIQ)
ADRNL
KIDNEY
Immunoassays
OVARY
Multiplexed bead-based
COLON immunoassays (IMMUNO)
TESTES
WAT-SC SKM-VL Metabolomics
SKM-GN
Targeted and untargeted (METAB)

c
METHYL 107 119 328 621 103 109 82 61 Genomics
Proteomics
ATAC 75 442 4 253 1,032 173 237 31 Metabolomics
Immunoassays
TRNSCRPT 1,582 722 90 1,099 566 766 1,742 1,635 276 961 339 4,493 105 16 244 746 2,516 896 107

0.10+
PROT 693 691 649 1,042 318 247 14

Differential proportion of analytes

PHOSPHO 436 694 50 720 272 19 389 0.08

ACETYL 450 2,012

0.06
UBIQ 24 161

0.04
IMMUNO 14 4 12 7 10 12 11 7 9 3 22 10 9 5 6 0 4

METAB 529 568 23 83 298 53 330 254 598 301 220 89 44 46 146 36 52 21 44 0.02

0
BLOOD

HEART

HYPOTH

HIPPOC

SMLINT

COLON

OVARY

TESTES
VENACV

SPLEEN

SKM-GN

SKM-VL

WAT-SC

BAT

LIVER

LUNG

KIDNEY

ADRNL

CORTEX
PLASMA

Circulation Skeletal Adipose Brain Gastro- Gonads

muscle intestinal

Fig. 1 | Summary of the study design and multi-omics dataset. a, Experimental Tissue labels indicate the location, colour code, and abbreviation for each
design and tissue sample processing. Inbred Fischer 344 rats were subjected tissue used throughout this study: ADRNL, adrenal gland; BAT, brown adipose
to a progressive treadmill training protocol. Tissues were collected from male tissue; BLOOD, whole blood, blood RNA; COLON, colon; CORTEX, cerebral
and female animals that remained sedentary or completed 1, 2, 4 or 8 weeks cortex; HEART, heart; HIPPOC, hippocampus; HYPOTH, hypothalamus;
of endurance exercise training. For trained animals, samples were collected KIDNEY, kidney; LIVER, liver; LUNG, lung; OVARY, ovaries; PLASMA, plasma;
48 h after their last exercise bout (red pins). b, Summary of molecular datasets SKM-GN, gastrocnemius (skeletal muscle); SKM-VL, vastus lateralis (skeletal
included in this study. Up to nine data types (omes) were generated for blood, muscle); SMLINT, small intestine; SPLEEN, spleen; TESTES, testes; VENACV,
plasma, and 18 solid tissues, per animal: ACETYL: acetylproteomics; protein vena cava; WAT-SC, subcutaneous white adipose tissue. Icons next to each
site acetylation; ATAC, chromatin accessibility, ATAC-seq data; IMMUNO, tissue label indicate the data types generated for that tissue. c, Number of
multiplexed immunoassays; METAB, metabolomics and lipidomics; METHYL, training-regulated features at 5% FDR. Each cell represents results for a single
DNA methylation, RRBS data; PHOSPHO, phosphoproteomics; protein site tissue and data type. Colours indicate the proportion of measured features
phosphorylation; PROT, global proteomics; protein abundance; TRNSCRPT, that are differential.
transcriptomics, RNA-seq data; UBIQ, ubiquitylome, protein site ubiquitination.

Differential analysis was used to characterize the molecular responses modest: 56% of the per-feature maximum fold changes were between
to endurance training (Methods). We computed the overall signifi- 0.67 and 1.5. Permutation testing showed that permuting the group or
cance of the training response for each feature, denoted as the training sex labels resulted in a significant reduction in the number of selected
P value, where 35,439 features at 5% false discovery rate (FDR) comprise analytes in most tissues (Extended Data Fig. 3a–d and Supplementary
the training-regulated differential features (Fig. 1c and Supplementary Discussion). For transcriptomics, the hypothalamus, cortex, testes and
Table 2). Timewise summary statistics quantify the exercise training vena cava had the smallest proportion of training-regulated genes,
effects for each sex and time point. Training-regulated molecules were whereas the blood, brown and white adipose tissues, adrenal gland
observed in the vast majority of tissues for all omes, including a rela- and colon showed more extensive effects (Fig. 1c). For proteomics, the
tively large proportion of transcriptomics, proteomics, metabolomics gastrocnemius, heart and liver showed substantial differential regula-
and immunoassay features (Fig. 1c). The observed timewise effects were tion in both protein abundance and post-translational modifications

Nature | Vol 629 | 2 May 2024 | 175

Article
a in white adipose tissue (Fig. 2a). We identified pathways enriched by
these tissue-specific training-responsive genes (Extended Data Fig. 4b)
SKM-GN
WAT-SC

KIDNEY
Number of shared differential genes

HEART

LUNG
LIVER

500 1,000 1,500

and tabulated a subset of highly specific genes to gain insight into
tissue-specific training adaptation (Supplementary Table 4). Focusing
1,391
1,102
on sexually conserved responses revealed tissue-dependent adapta-
718 tions. These included changes related to immune cell recruitment and
698 tissue remodelling in the lung, cofactor and cholesterol biosynthesis
555
292
in the liver, ion flux in the heart, and metabolic processes and striated
249 muscle contraction in the gastrocnemius (Supplementary Discussion).
210 A detailed analysis of white adipose tissue adaptations to exercise train-
168
156
ing is provided elsewhere14. We also observed ‘ome’-specific responses,
132 with unique transcript and protein responses at the gene and pathway
b
118
Cell adhesion molecules
levels (Extended Data Fig. 4c,d, Supplementary Discussion and Sup-
107
Haematopoietic cell lineage plementary Tables 5 and 6).
89
84 Generation of second messenger molecules 2,359 genes had differential features in at least two tissues (Fig. 2a).
73 T cell receptor signalling
Lung and white adipose tissue had the largest set of uniquely shared
69 TH17 cell differentiation
genes (n = 249), with predominantly immune-related pathway
64 Rheumatoid arthritis
51 Lysosome
enrichments (Fig. 2b); expression patterns suggested decreased
47 Viral myocarditis inflammation in the lung and increased immune cell recruitment in
40
Toxoplasmosis white adipose tissue (Supplementary Tables 2 and 3). Heart and gas-
40 Immunoregulatory interactions:
40 lymphoid and non-lymphoid cells trocnemius had the second-largest group of uniquely shared genes,
38 0 2.5 5.0 7.5
with enrichment of mitochondrial metabolism pathways includ-
36 Enrichment FDR (–log10)
ing the mitochondria fusion genes Opa1 and Mfn1 (Supplementary
34
33 c Table 3).
27 HSP90 chaperone cycle for steroid Twenty-two genes were training-regulated in all six tissues, with
26 receptors: ligand present
Protein folding
particular enrichment in heat shock response pathways (Fig. 2c). Exer-
26
26 Chaperone-mediated protein folding cise induces the expression of heat shock proteins (HSPs) in various
23 Cell response: heat stress rodent and human tissues15. A focused analysis of our transcriptomics
23
Blood vessel diameter maintenance and proteomics data revealed HSPs as prominent outliers (Extended
22
21
Negative regulation of signalling:
ligand absent Data Fig. 5a and Supplementary Discussion). Specifically, there was a
21 cGMP–PKG signalling pathway marked, proteomics-driven up-regulation in the abundance of HSPs,
Protein complex assembly:
20
positive regulation including the major HSPs HSPA1B and HSP90AA1 (Extended Data
19 ESR-mediated signalling
16 Fig. 5b,c). Another ubiquitous endurance training response involved
0 1 2 3 4 5
16
Enrichment FDR (–log10) regulation of the kininogenases KNG1 and KNG2 (Supplementary
Table 3). These enzymes are part of the kallikrein–kininogen system
Fig. 2 | Multi-tissue molecular endurance training responses. a, UpSet plot and have been implicated in the hypotensive and insulin-sensitizing
of the training-regulated gene sets associated with each tissue. Bars and dots effects of exercise16,17.
indicating tissue-specific differential genes are coloured by tissue. Pathway
enrichment analysis is shown for selected sets of genes in b,c as indicated by
the arrows. b,c, Significantly enriched pathways (10% FDR) corresponding to Transcription factors and phosphosignalling
genes that are differential in both LUNG and WAT-SC datasets (b) and the 22
We used proteomics and transcriptomics data to infer changes in
genes that are training-regulated in all six tissues considered in a (c). Redundant
transcription factor and phosphosignalling activities in response to
pathways (those with an overlap of 80% or greater with an existing pathway)
were removed. ESR, oestrogen receptor; TH17, T helper 17.
endurance training through transcription factor and PTM enrich-
ment analyses (Methods). We compared the most significantly
enriched transcription factors across tissues (Fig. 3a, Extended
Data Fig. 6a and Supplementary Table 7). In the blood, we observed
(PTMs), with more restricted results in white adipose tissue, lung and enrichment of the haematopoietic-associated transcription fac-
kidney protein abundance. For metabolomics, a large proportion of tors GABPA, ETS1, KLF3 and ZNF143; haematopoietic progenitors
differential metabolites were consistently observed across all tissues, are proposed to be transducers of the health benefits of exercise18.
although the absolute numbers were related to the number of metab- In the heart and skeletal muscle, we observed a cluster of enriched
olomic platforms used (Extended Data Fig. 1e). The vast number of Mef2 family transcription factor motifs (Fig. 3a). MEF2C is a muscle-
differential features over the training time course across tissues and associated transcription factor involved in skeletal, cardiac and
omes highlights the multi-faceted, organism-wide nature of molecular smooth muscle cell differentiation and has been implicated in
adaptations to endurance training. vascular development, formation of the cardiac loop and neuron
differentiation19.
Phosphorylation signatures of key kinases were altered across many
Multi-tissue response to training tissues (Fig. 3b and Supplementary Table 8). This included AKT1 across
To identify tissue-specific and multi-tissue training-responsive gene heart, kidney and lung, mTOR across heart, kidney and white adipose
expression, we considered the six tissues with the deepest molecu- tissue, and MAPK across heart and kidney. The liver showed an increase
lar profiling: gastrocnemius, heart, liver, white adipose tissue, lung in the phosphosignature related to regulators of hepatic regeneration,
and kidney. In sum, 11,407 differential features from these datasets including EGFR1, IGF and HGF (Extended Data Fig. 6b, Supplemen-
were mapped to their cognate gene, for a total of 7,115 unique genes tary Discussion). Increased phosphorylation of STAT3 and PXN, HGF
across the tissues (Fig. 2a, Extended Data Fig. 4a and Supplementary targets involved in cell proliferation, suggest a mechanism for liver
Table 3). Most of the genes with at least one training-responsive fea- regeneration in response to exercise (Extended Data Fig. 6c). In the
ture were tissue-specific (67%), with the greatest number appearing heart, kinases showed bidirectional changes in their predicted basal

176 | Nature | Vol 629 | 2 May 2024

 a b CORTEX HEART KIDNEY LIVER LUNG SKM-GN WAT-SC

SKM-GN
WAT-SC

SPLEEN

SKM-VL
SMLINT

KIDNEY
COLON
BLOOD

HEART
OVARY
ADRNL
Tissue
LUNG Sex
BAT Timepoint

AMPK
Tissue AMPKA2
SP1 AMPKA1
KLF3 CDK1
SP5 CDK5
CDK2

CDK
SP2
KLF10 CDK4
KLF4 CDK9
CDK7
SMAD2
CRE ERK2
ZNF143 ERK1
P38B
ELK1 P38D
ELK4 P38G
FOXO3 TAK1

MAPK
ETV4 MAP3K8
ETS1 P38A
ETV1 MEK1
GABPA JNK1
ELF3 JNK2
ERG MKK4
ETV2 ASK1
MAPKAPK2
AKT-MTOR

EHF
ELF4 MTOR
AKT2
ZNF467 AKT1
RUNX1
RUNX2 EGFR1
TIE2
MITF Leptin
PU.1 RAGE
IRF8 T cell receptor
IRF3 CRH
ERRγ TWEAK
PRDM1 Kit receptor
ZBTB18 Notch
BATF Oxytocin
MEF2B IL-6
MEF2C MAPK signalling
Cell cycle
MEF2A OSM
Pathway

TATA box FSH

MyoD TSH
TBR1 TNF
SIX2 B cell receptor
CLOCK FGF1
FOXM1 Prolactin
HOXA9 CCR7
ASCL2 TGFβ
HEB IL-33
PTF1A GLP1
IL-11
ZNF263 Gastrin
NF1 half site IL-2
HOXA11 IL-7
MAFA PI3K–AKT signalling
HOXB13 TSLP
Tissue enrichment Adjusted P Sex Timepoint NES Adjusted P
z-score Female 1 week 0
<0.1
0.1
≥0.1 Male 2 weeks –10 –5 0 5 10
–4 –2 0 2 4 >0.2
4 weeks
8 weeks

Fig. 3 | Regulatory signalling pathways modulated by endurance training. PTM signature enrichment analysis on phosphoproteomics data. Only kinases
a, Transcription factor motif enrichment analysis of the training-regulated or pathways with a significant difference in at least one tissue, sex or time point
transcripts in each tissue. The heat map shows enrichment z-scores across the (q value < 0.05) are shown. The heat map shows normalized enrichment score
differential genes for the 13 tissues that had at least 300 genes after mapping (NES) as colour; tissue, sex and time point combinations as columns, and
transcript IDs to gene symbols. Transcription factors were hierarchically either kinases or pathways as rows. Kinases are grouped by family; rows are
clustered by their enrichment across tissues. CRE, cAMP response element. hierarchically clustered within each group. FSH, follicle-stimulating hormone;
b, Estimate of activity changes in selected kinases and signalling pathways using TSH, thyroid-stimulating hormone.

activity in response to endurance training (Extended Data Fig. 6d and complete timewise summary statistics using an empirical Bayes graphi-
Supplementary Discussion). Several AGC protein kinases showed a cal clustering approach (Methods). By integrating these results onto a
decrease in predicted activity, including AKT1, whereas tyrosine graph, we summarize the dynamics of the molecular training response
kinases, including SRC and mTOR, were predicted to have increased and identify groups of features with similar responses (Extended Data
activity. The known SRC target phosphorylation sites GJA1 pY265 Fig. 7 and Supplementary Table 10). We performed pathway enrichment
and CDH2 pY820 showed significantly increased phosphorylation in analysis for many graphically defined clusters to characterize putative
response to training (Extended Data Fig. 6e). Notably, phosphorylation underlying biology (Supplementary Table 11).
of GJA1 Y265 has previously been shown to disrupt gap junctions, key We examined biological processes associated with training using
transducers of cardiac electrical conductivity20. This suggests that the pathway enrichment results for up-regulated features at 8 weeks
SRC signalling may regulate extracellular structural remodelling of the of training (Extended Data Fig. 8, Supplementary Table 12 and Supple-
heart to promote physiologically beneficial adaptations. In agreement mentary Discussion). Compared with other tissues, the liver showed
with this hypothesis, gene set enrichment analysis (GSEA) of extra- substantial regulation of chromatin accessibility, including in the
cellular matrix proteins revealed a negative enrichment in response nuclear receptor signalling and cellular senescence pathways. In the
to endurance training, showing decreased abundance of proteins gastrocnemius, terms related to peroxisome proliferator-activated
such as basement membrane proteins (Extended Data Fig. 6f–h and receptors (PPAR) signalling and lipid synthesis and degradation were
Supplementary Table 9). enriched at the protein level, driven by proteins including the lipid
droplet features PLIN2, PLIN4 and PLIN5. At the metabolomic level,
terms related to ether lipid and glycerophospholipid metabolism were
Molecular hubs of exercise adaptation enriched. Together, these enrichments highlight the well-known abil-
To compare the dynamic multi-omic responses to endurance train- ity of endurance training to modulate skeletal muscle lipid composi-
ing across tissues, we clustered the 34,244 differential features with tion, storage, synthesis and metabolism. The blood displayed pathway

Nature | Vol 629 | 2 May 2024 | 177

Article
a ACETYL PHOSPHO PROT TRNSCRPT b 8w_F1_M1 muscle
SKM-GN (no ACETYL) Female Male
1 2 4 8

Normalized value
2
F only up Intersect size 1
100 0
200 –1
Both up –2
No. of analytes
M only up 012 4 8 012 4 8
200 Time trained (weeks)
No change
400
c
Carbon HEART
F only down
600 metabolism SKM-GM
Both down SKM-VL
Amino acid
M only down synthesis
and metabolism
SKM−VL (TRNSCRPT only)
Oxidative phosphorylation
1 2 4 8
Intersect size
F only up Amino acid metabolism
50
100
Both up 150
Fatty acid
metabolism
M only up No. of analytes
100 Cellular response
No change
to heat stress
200
F only down
Mitochondrial
Both down 300 translation
d Longevity
M only down PROT ATAC
Muscle system
PHOSPHO METHYL
Heart Response to mech. stimulus
TRNSCRPT METAB
1 2 4 8 Intersect size Other
50
F only up 100
150 Atp1a2
Both up 200
Ank2
M only up No. of analytes Mef2c Clpb Tbc1d1
Got1
200 Dld Timm13 Map11c3a
No change Mdh1 Sod2
400 Slc2a4 Hdac4 Timm8a1
Asb2
F only down Jun
600 Hspa1b Hspa41
Hsp90aa1
Prkab1 Hspb6
Both down Bag3
800 Camk2g Homer1
Txnip
M only down Fhl1

Fig. 4 | Temporal patterns of the molecular training response. a, Graphical plots of standardized abundances of all 8w_F1_M1 muscle features. The black
representation of training-differential features in the three muscle tissues: line represents the average value across all features. c, Network view of
gastrocnemius (SKM-GN), vastus lateralis (SKM-VL) and heart. Each node significant pathway enrichment results (10% FDR) corresponding to the
represents one of nine possible states (rows) at each of the four training time features in b. Nodes represent pathways; edges represent functionally similar
points (columns). Triangles to the left of row labels map states to symbols used node pairs (set similarity ≥ 0.3). Nodes are included only if they are significantly
in Fig. 5a. Edges represent the path of differential features over the training enriched in at least two of the muscle tissues, as indicated by node colour.
time course (see Extended Data Fig. 7 for a detailed explanation). Each graph Node size is proportional to the number of differential feature sets (for example,
includes the three largest paths of differential features in that tissue, with edges gastrocnemius transcripts) for which the pathway is significantly enriched.
split by data type. Both node and edge size are proportional to the number of High-level biological themes were defined using Louvain community detection
features represented. The node corresponding to features that are up-regulated of the nodes. d, A subnetwork of a larger cluster identified by network clustering
in both sexes at 8 weeks of training (8w_F1_M1) is circled in each graph. b, Line 8w_F1_M1 features from SKM-GN. Mech., mechanical.

enrichments related to translation and organelle biogenesis and main- each muscle tissue converged to a sex-consistent response by week
tenance. Paired with the transcription factor analysis (Fig. 3a), this 8 (Fig. 4a). Because of the large number of muscle features that were
suggests increased haematopoietic cellular mobilization in the blood. up-regulated in both sexes at week 8, we further examined the cor-
Less studied tissues in the context of exercise training, including the responding multi-omic set of analytes (Fig. 4b). Pathway enrichment
adrenal gland, spleen, cortex, hippocampus and colon, also showed analysis of the genes associated with these differential features dem-
regulation of diverse pathways (Supplementary Discussion). onstrated a sex- and muscle-consistent endurance training response
To identify the main temporal or sex-associated responses in each that reflected up-regulation of mitochondrial metabolism, biogen-
tissue, we summarized the graphical cluster sizes by tissue and time esis and translation, and cellular response to heat stress (Fig. 4c and
(Extended Data Fig. 7a). We observed that the small intestine and Supplementary Table 11).
plasma had more changes at weeks 1 and 2 of training. Conversely, many We used a network connectivity analysis to study up-regulated fea-
up-regulated features in brown adipose tissue and down-regulated tures in the gastrocnemius at week 8 (Extended Data Fig. 9a,b, Methods
features in white adipose tissue were observed only at week 8. The and Supplementary Discussion). Mapping features to genes revealed
largest proportion of opposite effects between males and females overlaps between transcriptomic, chromatin accessibility, and pro-
was observed at week 1 in the adrenal gland. Other tissues, including teomic assays, but no overlaps with methylation. Three molecular
the blood, heart, lung, kidney and skeletal muscle (gastrocnemius and interaction networks were compared (Methods), and BioGRID21 was
vastus lateralis), had relatively consistent numbers of up-regulated and used for further clustering analysis, which identified three clusters
down-regulated features. (Extended Data Fig. 9c and Supplementary Table 13). The largest cluster
We next focused on characterizing shared molecular responses in was significantly enriched for multiple muscle adaptation processes
the three striated muscles (gastrocnemius, vastus lateralis and heart). (Fig. 4d and Supplementary Table 14). This analysis illustrates the direct
The three largest graphical clustering paths of differential features in linkage among pathways and putative central regulators, emphasizing

178 | Nature | Vol 629 | 2 May 2024

a Down-regulated Up-regulated b BAT TRNSCRPT WAT−SC TRNSCRPT
8w_F0_M1 8w_F0_M1
Sex-consistent Female-specific Male-specific
Female Male Female Male
Antigen processing and presentation

Normalized value

Normalized value
2 2
B cell receptor signalling pathway 1 1
0 0
C-type lectin receptor signalling pathway –1 –1
–2 –2
Chemokine signalling pathway
012 4 8 012 4 8 012 4 8 012 4 8
Complement and coagulation cascades Time trained (weeks) Time trained (weeks)
Cytosolic DNA-sensing pathway
FcεRI signalling pathway
c BAT WAT-SC
FcγR-mediated phagocytosis
CD45+ cells
Haematopoietic cell lineage
# # B cells
Intestinal immune network for IgA production
# # CD4 T cells
Leukocyte transendothelial migration # # CD8 T cells
NK cell mediated cytotoxicity # NK cells
Neutrophil extracellular trap formation # # Monoctye/macrophage
NOD-like receptor signalling pathway # # Dendritic cells

Platelet activation Neutrophils

Eosinophils (Ccr3)
RIGI-like receptor signalling pathway
# Mast cells
T cell receptor signalling pathway
Platelets
TH1 and TH2 cell differentiation # # Erythrocytes
TH17 cell differentiation Lymphatic tissue
Toll-like receptor signalling pathway Proliferation (Ki67)

–1 0 1 –1 0 1
BLOOD
HEART
SPLEEN
SKM-GN
SKM-VL
WAT-SC
BAT
LUNG
ADRNL
CORTEX
SMLINT
COLON

Pearson r between marker and TRNSCRPT

8w_F0_M1 centroid

Fig. 5 | Training-induced immune responses. a, Enrichment analysis results of of features in b at the transcript level. A pink dot indicates that the marker is
the training-differential transcripts at 8 weeks in Kyoto Encyclopedia of Genes also one of the differential features plotted in b. A pound sign indicates that
and Genomes (KEGG) immune system pathways (10% FDR). NK, natural killer. the distribution of Pearson correlations for a set of at least two markers is
b, Line plots of standardized abundances of selected training-differential significantly different from 0 (two-sided one-sample t-test, 5% FDR). When only
transcripts. Brown and white adipose tissue show male-specific up-regulation one marker is used to define a category on the y axis, the gene name is provided
at week 8 (8w_F0_M1). The small intestine (SMLINT) shows down-regulation in in parentheses. In box plots, the centre line represents median, box bounds
females and partial down-regulation in males at week 8 (8w_F-1_M0 or 8w_F-1_M-1). represent 25th and 75th percentiles, whiskers represent minimum and maximum
c, Box plots of the sample-level Pearson correlation between markers of excluding outliers and blue dots represent outliers.
immune cell types, lymphatic tissue or cell proliferation and the average value

the importance of multi-omic data in identifying interconnected net- diseases25. Overall, these results support a high concordance of our data
works and understanding skeletal muscle remodelling. from rats with human studies and their relevance to human disease.

Connection to human diseases and traits Sex-specific responses to exercise

To systematically evaluate the translational value of our data, we inte- Many tissues showed sex differences in their training responses
grated our results with extant exercise studies and disease ontology (Extended Data Fig. 10), with 58% of the 8-week training-regulated fea-
(DO) annotations (Methods). First, we compared our vastus later- tures demonstrating sex-differentiated responses. Opposite responses
alis transcriptomics results to a meta-analysis of long-term training between the sexes were observed in adrenal gland transcripts, lung
gene-expression changes in human skeletal muscle tissue8, demon- phosphosites and chromatin accessibility features, white adipose tis-
strating a significant and direction-consistent overlap (Extended Data sue transcripts and liver acetylsites. In addition, proinflammatory
Fig. 9d–g and Supplementary Discussion). We also identified a signifi- cytokines exhibited sex-associated changes across tissues (Extended
cant overlap between differential transcripts in the gastrocnemius of Data Fig. 11a,b and Supplementary Table 16). Most female-specific
female rats trained for 8 weeks and differentially expressed genes iden- cytokines were differentially regulated between weeks 1 and 2 of train-
tified in the soleus in a study of sedentary and exercise-trained female ing, whereas most male-specific cytokines were differentially regulated
rats selectively bred for high or low exercise capacity22 (Extended Data between weeks 4 and 8 (Extended Data Fig. 11c).
Fig. 9h). Similarly, adaptations from high-intensity interval training in We observed extensive transcriptional remodelling of the adrenal
humans23 significantly overlapped with the proteomics response in rats gland, with more than 4,000 differential genes. Notably, the largest
(Extended Data Fig. 9i), particularly for female rats trained for 8 weeks graphical path of training-regulated features was negatively corre-
(Extended Data Fig. 9j). Finally, we performed DO enrichment analysis lated between males and females, with sustained down-regulation
using the DOSE R package24 (Supplementary Table 15 and Methods). in females and transient up-regulation at 1 week in males (Extended
Down-regulated genes from white adipose tissue, kidney and liver Data Fig. 11d). The genes in this path were also associated with ster-
were enriched for several disease terms, suggesting a link between the oid hormone synthesis pathways and metabolism, particularly those
exercise response and type 2 diabetes, cardiovascular disease, obesity pertaining to mitochondrial function (Supplementary Table 11). Fur-
and kidney disease (5% FDR; Extended Data Fig. 9k and Supplemen- ther, transcription factor motif enrichment analysis of the transcripts
tary Discussion), which are all epidemiologically related co-occurring in this path showed enrichment of 14 transcription factors (5% FDR;

Nature | Vol 629 | 2 May 2024 | 179

Article
Supplementary Table 17), including the metabolism-regulating factors subcategories of KEGG pathways (10% FDR; Extended Data Fig. 13a and
PPARγ, PPARα and oestrogen-related receptor gamma (ERRγ). The Supplementary Table 11). The liver showed the greatest number of
gene-expression levels of several significantly enriched transcription significantly enriched metabolite classes, followed by the heart, lung
factors themselves followed the same trajectory as this path (Extended and hippocampus (Fig. 6a and Supplementary Discussion). Inspection
Data Fig. 11e). of individual metabolites and acylcarnitine groups revealed changes
In the rat lung, we observed decreased phosphosignalling activity associated with functional alterations in response to training (Extended
with training primarily in males (Fig. 3b). Among these, the PRKACA Data Fig. 13b–d and Supplementary Discussion). Of particular inter-
phosphorylation signature showed the largest sex difference at 1 est, trimethylamine-N-oxide has been associated with cardiovascular
and 2 weeks (Extended Data Fig. 11f–h and Supplementary Table 8). disease32. We observed up-regulation of 1-methylhistidine, a marker of
PRKACA is a kinase that is involved in signalling within multiple cellular muscle protein turnover, in the kidney at 1, 2 and 4 weeks, which may
pathways. However, four PRKACA substrates followed this pattern indicate muscle breakdown and clearance through the kidney during
and were associated with cellular structures (such as cytoskeleton early training time points. Cortisol levels were increased as expected
and cell–cell junctions): DSP, MYLK, STMN1 and SYNE1 (Extended from the physiological stress of training, and we observed a substantial
Data Fig. 11i). The phosphorylation of these proteins suggests a increase in the kidney, again probably owing to renal clearance33. The
sex-dependent role of PRKACA in mediating changes in lung structure liver showed up-regulation of 1-methylnicotinamide, which may have
or mechanical function with training. This is supported as DSP and a role in inflammation34, at 8 weeks.
MYLK have essential roles in alveolar and epithelial cell remodelling in The heart showed enrichment of various carbohydrate metabolism
the lung26,27. subcategories across many omes (Extended Data Fig. 13a), and remark-
Immune pathway enrichment analysis of training-regulated tran- ably, all enzymes within the glycolysis–gluconeogenesis pathway
scripts at 8 weeks showed limited enrichment in muscle (heart, gas- showed a consistent increase in abundance, except for GPI, FBP2 and
trocnemius and vastus lateralis) and brain (cortex, hippocampus, DLAT (Extended Data Fig. 13e). Oxidative phosphorylation was enriched
hypothalamus), down-regulation in the lung and small intestine, and in most tissues and is consistent with the joint analyses of the muscle tis-
strong up-regulation in brown and white adipose tissue in males only sues (Fig. 4c), suggesting potential changes in mitochondria biogenesis.
(Fig. 5a, Extended Data Fig. 12a and Supplementary Table 11). Many of the We estimated proportional mitochondrial changes to endurance train-
same immune pathways (Supplementary Table 18) and immune-related ing using mitochondrial RNA-sequencing (RNA-seq) reads (Extended
transcription factors (Supplementary Table 19) were enriched in both Data Fig. 14a–c) and changes of mitochondrial functions through
adipose tissues in males. Furthermore, correlation between the tran- GSEA using gene expression, protein abundance and protein PTMs
script expression profiles of male-specific up-regulated features in (Fig. 6b, Extended Data Fig. 14d and Supplementary Tables 22–25).
the adipose tissues and immune cell markers from external cell-typing Increased mitochondrial biogenesis was observed in skeletal muscle,
assays revealed a strong positive correlation for many immune cell heart and liver across these analyses. Moreover, sex-specific mito-
types, including B, T and natural killer cells, and low correlation with chondrial changes were observed in the adrenal gland, as described
platelets, erythrocytes and lymphatic tissue (Fig. 5b,c, Methods and above, and in the colon, lung and kidney. These results highlight
Supplementary Table 20). These patterns suggest recruitment of a highly adaptive and pervasive mitochondrial response to endur-
peripheral immune cells or proliferation of tissue-resident immune ance training; a more in-depth analysis of this response is provided
cells as opposed to non-biological variation in blood or lymph content. elsewhere35.
Correlations at the protein level were not as marked (Extended Data In the liver, we observed substantial regulation of metabolic path-
Fig. 12b,c). Complementary analyses using CIBERTSORTx produced ways across the proteome, acetylome and lipidome (Fig. 6a,b and
similar results (Extended Data Fig. 12d,e). In summary, our data sug- Extended Data Fig. 13a). For example, there was significant enrichment
gest an important role of immune cell activity in the adaptation of male in 12 metabolite classes belonging to ‘lipids and lipid-related com-
adipose tissue to endurance training. pounds’ (Fig. 6a and Supplementary Table 26). We therefore focused
The small intestine was among the tissues with the highest enrich- on the large group of features that increased in abundance over time
ment in immune-related pathways (Extended Data Fig. 12a), with for both sexes (Fig. 6c). Most of these liver features corresponded to
down-regulation of transcripts at 8 weeks, and a more robust response protein abundance and protein acetylation changes in the mitochon-
in females (Fig. 5b). This transcript set was significantly enriched with drial, amino acid and lipid metabolic pathways (Fig. 6d and Supplemen-
pathways related to gut inflammation (Supplementary Table 11). We tary Table 27). We also observed an increase in phosphatidylcholines
observed positive associations between these transcripts and markers and a concomitant decrease in triacylglycerols (Fig. 6e). Finally, there
of several immune cell types, including B, T, natural killer and dendritic was increased abundance and acetylation of proteins from the peroxi-
cells, suggesting decreased abundance (Fig. 5c and Supplementary some, an organelle with key functions in lipid metabolism (Extended
Discussion). Endurance training also decreased the expression of tran- Data Fig. 14e). To our knowledge, these extensive changes in protein
scripts with genetic risk loci for inflammatory bowel disease (IBD), acetylation in response to endurance training have not been described
including major histocompatability complex class II28, a finding that previously. Together, these molecular adaptations may constitute
also emerged through the DO enrichment analysis (Supplementary part of the mechanisms underlying exercise-mediated improvements
Table 15). Endurance training is suggested to reduce systemic inflam- in liver health, particularly protection against excessive intrahepatic
mation, in part by increasing gut microbial diversity and gut barrier lipid storage and steatosis36.
integrity29. In accordance, we observed decreases in Cxcr3 and Il1a with
training (Extended Data Fig. 12f), both of which are implicated in the
pathogenesis of IBD30,31. Together, these data suggest that endurance Discussion
training improves gut homeostasis, potentially conferring systemic Mapping the molecular exercise responses across a whole organism
anti-inflammatory effects. is critical for understanding the beneficial effects of exercise. Previ-
ous studies are limited to a few tissues, a narrow temporal range, or
a single sex. Substantially expanding on the current work in the field,
Multi-tissue changes in mitochondria and lipids we used 25 distinct molecular platforms in as many as 19 tissues to
We summarized the organism-wide metabolic changes for metabo- study the temporal changes to endurance exercise training in male and
lomic datasets using RefMet metabolite classes (Fig. 6a and Supple- female rats. Accordingly, we identified thousands of training-induced
mentary Table 21) and for non-metabolomics datasets using metabolic changes within and across tissues, including temporal and sex-biased

180 | Nature | Vol 629 | 2 May 2024

 a Metabolite enrichment by RefMet classes b PROT ACETYL

HYPOTH
SKM-GN
PLASMA

CORTEX
VENACV

WAT-SC
SPLEEN

HIPPOC
SKM-VL

TESTES
SMLINT
KIDNEY

COLON

SKM-GN

CORTEX
OVARY

WAT-SC
ADRNL
HEART

KIDNEY
LUNG
LIVER

HEART

HEART
LUNG
LIVER

LIVER
BAT
Amino acids and peptides
Monosaccharides Group
Short-chain acids Sex
TCA acids Amino acids or acid-like
and derivatives Timepoint
Bile acids
Cardiolipins Energy Mitochondrial central Sex
Ceramides Lipids and lipid-related dogma Female
Diacylglycerols compounds Male
Eicosanoids Nucleotides and Protein import sorting and
Fatty acids homeostasis Timepoint
derivatives
Fatty amides 1 week
Fatty esters Other
2 weeks
Glycerophosphocholines OXPHOS
4 weeks
Glycerophosphoethanolamines
8 weeks
Glycerophosphoglycerols
Glycerophosphoinositols NES
Glycerophosphoserines
Glycosphingolipids 10
Octadecanoids Adjusted P
Sphingoid bases
≤ 0.01 Metabolism
Sphingomyelins 5
Sterol esters 0.02
Triacylglycerols
0.03
Nicotinamides 0
Purines 0.04
Pyrimidines 0.05
Small molecule transport
Benzoic acids > 0.05 –5
Imidazoles Not tested Signalling
Indoles Mitochondrial dynamics and
surveillance –10
c LIVER:1w_F0_M1−>2w_F0_M1−> d Amino acid Mitochondria e
4w_F0_M1−>8w_F1_M1 metabolism metabolism
Up-regulated
Female Male Female
Male
2
Normalized value

Down-regulated Liver 8 week logFC

1 Peroxisome and lipid Lipids and lipid-related compounds 2
Female
metabolism Complement Cardiolipins 1
0 Male
Ceramides 0
–1 Diacylglycerols
Fatty esters –1
–2 Glycerophosphocholines –2
Glycerophosphoethanolamines
012 4 8 012 4 8 Glycerophosphoglycerols Sex
TRNSCRPT PROT ACETYL Glycerophosphoserines Female
Time trained (weeks) Glycerophosphoinositols Male
Sterol esters
Triacylglycerols

Fig. 6 | Training-induced changes in metabolism. a, RefMet metabolite class sexes, with a later response in females (LIVER: 1w_F0_M1 − >2w_F0_M1 − >4w_
enrichment calculated using GSEA with the −log10 training P value. Significant F0_M1 − >8w_F1_M1). The black line represents the average value across all
chemical class enrichments (5% FDR) are shown as black circles with size is features. d, Network view of pathway enrichment results corresponding to
proportional to FDR. Small grey circles are chemical class enrichments that features in c. Nodes indicate significantly enriched pathways (10% FDR); edges
were not significant, and blank cells were not tested owing to low numbers connect nodes if there is a similarity score of at least 0.375 between the gene
of detected metabolites. TCA, tricarboxylic acid cycle. b, GSEA results using sets driving each pathway enrichment. Node colours indicate omes in which
the MitoCarta MitoPathways gene set database and proteomics (PROT) or the enrichment was observed. e, log 2 fold changes (logFC) relative to sedentary
acetylome (ACETYL) timewise summary statistics for training. NESs are shown controls for metabolites within the ‘Lipids and lipid related compounds’
for significant pathways (10% FDR). Mitochondrial pathways shown as rows are category in the 8-week liver. Heat map colour represents fold change (red,
grouped using the parental group in the MitoPathways hierarchy. OXPHOS, positive; blue, negative). Compounds are grouped into columns based on
oxidative phosphorylation. c, Line plots of standardized abundances of liver category (coloured bars).
training-differential features across all data types that are up-regulated in both

responses, in mRNA transcripts, proteins, post-translational modifi- excluding acute responses. Our assays were performed on bulk tissue
cations and metabolites. Each omic dataset provides unique insights and do not cover single-cell platforms. Our resource has limited omic
into exercise adaptation, where a holistic understanding requires characterization for certain tissues, and additional platforms with
multi-omic analysis. This work illustrates how mining our data resource emerging biological relevance were not utilized, including microbiome
can both recapitulate expected mechanisms and provide novel profiling. Moreover, our results are hypothesis-generating and require
biological insights. biological validation; supporting this, we have established a publicly
This work can be leveraged to deepen our understanding of exercise- accessible tissue bank from this study.
related improvement of health and disease management. The global This MoTrPAC resource provides future opportunities to enhance
heat shock response to exercise may confer cytoprotective effects, and refine the molecular map of the endurance training response.
including in pathologies related to tissue damage and injury recovery37. We expect that this dataset will remain an ongoing platform to trans-
Increased acetylation of liver mitochondrial enzymes and regulation of late tissue- and sex-specific molecular changes in rats to humans.
lipid metabolism may link exercise to protection against non-alcoholic MoTrPAC has made extensive efforts to facilitate access, exploration
fatty liver disease and steatohepatitis36. Similarly, exercise-mediated and interpretation of this resource. We developed the MoTrPAC Data
modulation of cytokines, receptors and transcripts linked to intes- Hub to easily explore and download data (https://motrpac-data.org/),
tinal inflammation or IBD may be associated with improved gut software packages to provide reproducible source code and facilitate
health. These examples highlight unique training responses illumi- data retrieval and analysis in R (MotrpacRatTraining6mo and Motrpa-
nated by a multi-omics approach that can be leveraged for future cRatTraining6moData38,39), and visualization tools for data explora-
hypothesis-driven research on how exercise improves whole-body tion (https://data-viz.motrpac-data.org). Altogether, this multi-omic
and tissue-specific health. resource serves as a broadly useful reference for studying the milieu
We note limitations in our experimental design, datasets and analyses of molecular changes in endurance training adaptation and provides
(Supplementary Discussion). In short, samples were collected 48 h new opportunities to understand the effects of exercise on health
after the last exercise bout to capture sustained alterations, thereby and disease.

Nature | Vol 629 | 2 May 2024 | 181

Article
34. Zhang, W. et al. Nicotinamide N-methyltransferase ameliorates renal fibrosis by its
Online content metabolite 1-methylnicotinamide inhibiting the TGF-β1/Smad3 pathway. FASEB J. 36,
e22084 (2022).
Any methods, additional references, Nature Portfolio reporting summa- 35. Amar, D. et al. The mitochondrial multi-omic response to exercise training across tissues.
ries, source data, extended data, supplementary information, acknowl- Prepint at BioRxiv https://doi.org/10.1101/2023.01.13.523698 (2023).
36. Thyfault, J. P. & Rector, R. S. Exercise combats hepatic steatosis: potential mechanisms
edgements, peer review information; details of author contributions
and clinical implications. Diabetes 69, 517–524 (2020).
and competing interests; and statements of data and code availability 37. Dornbos, D. et al. Preischemic exercise reduces brain damage by ameliorating metabolic
are available at https://doi.org/10.1038/s41586-023-06877-w. disorder in ischemia/reperfusion injury. J. Neurosci. Res. 91, 818–827 (2013).
38. Gay, N. R., Amar, D., Beltran, P. M. J. & MoTrPAC Study Group. MotrpacRatTraining6mo:
Analysis of the MoTrPAC endurance exercise training study in 6-month-old. R package
1. Blair, S. N. et al. Physical fitness and all-cause mortality. A prospective study of healthy version 1.6.3 https://motrpac.github.io/MotrpacRatTraining6mo/ (2023).
men and women. JAMA 262, 2395–2401 (1989). 39. Gay, N. R. & MoTrPAC Study Group. MotrpacRatTraining6moData: Data for analysis of the
2. Booth, F. W., Roberts, C. K. & Laye, M. J. Lack of exercise is a major cause of chronic MoTrPAC endurance exercise training study in 6-month-old rats. R package version 1.9.0
diseases. Compr. Physiol. 2, 1143–1211 (2012). https://motrpac.github.io/MotrpacRatTraining6moData/ (2023).
3. Neufer, P. D. et al. Understanding the cellular and molecular mechanisms of physical
activity-induced health benefits. Cell Metab. 22, 4–11 (2015). Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
4. Sanford, J. A. et al. Molecular Transducers of Physical Activity Consortium (MoTrPAC): published maps and institutional affiliations.
mapping the dynamic responses to exercise. Cell 181, 1464–1474 (2020).
5. Nocon, M. et al. Association of physical activity with all-cause and cardiovascular
Open Access This article is licensed under a Creative Commons Attribution
mortality: a systematic review and meta-analysis. Eur. J. Cardiovasc. Prev. Rehabil. 15,
4.0 International License, which permits use, sharing, adaptation, distribution
239–246 (2008).
and reproduction in any medium or format, as long as you give appropriate
6. Moore, S. C. et al. Association of leisure-time physical activity with risk of 26 types of
credit to the original author(s) and the source, provide a link to the Creative Commons licence,
cancer in 1.44 million adults. JAMA. Intern. Med. 176, 816–825 (2016).
and indicate if changes were made. The images or other third party material in this article are
7. Pedersen, B. K. & Saltin, B. Exercise as medicine — evidence for prescribing exercise as
included in the article’s Creative Commons licence, unless indicated otherwise in a credit line
therapy in 26 different chronic diseases. Scand. J. Med. Sci. Sports 25, 1–72 (2015).
to the material. If material is not included in the article’s Creative Commons licence and your
8. Amar, D. et al. Time trajectories in the transcriptomic response to exercise - a meta-analysis.
intended use is not permitted by statutory regulation or exceeds the permitted use, you will
Nat. Commun. 12, 3471 (2021).
need to obtain permission directly from the copyright holder. To view a copy of this licence,
9. Gibb, A. A. et al. Exercise-induced changes in glucose metabolism promote physiological
visit http://creativecommons.org/licenses/by/4.0/.
cardiac growth. Circulation 136, 2144–2157 (2017).
10. Lindholm, M. E. et al. An integrative analysis reveals coordinated reprogramming of the
epigenome and the transcriptome in human skeletal muscle after training. Epigenetics 9, © The Author(s) 2024
1557–1569 (2014).
11. Overmyer, K. A. et al. Maximal oxidative capacity during exercise is associated with
skeletal muscle fuel selection and dynamic changes in mitochondrial protein acetylation.
Cell Metab. 21, 468–478 (2015). MoTrPAC Study Group
12. Pillon, N. J. et al. Transcriptomic profiling of skeletal muscle adaptations to exercise and
inactivity. Nat. Commun. 11, 470 (2020). Primary authors
13. Sato, S. et al. Atlas of exercise metabolism reveals time-dependent signatures of Lead Analysts
metabolic homeostasis. Cell Metab. 34, 329–345.e8 (2022). David Amar1,75, Nicole R. Gay2,75, Pierre M. Jean-Beltran3,75
14. Many, G. M. Sexual dimorphism and the multi-omic response to exercise training in rat
subcutaneous white adipose tissue. Nat. Metab. https://doi.org/10.1038/s42255-023-
00959-9 (2024). Lead Data Generators
15. Henstridge, D. C., Febbraio, M. A. & Hargreaves, M. Heat shock proteins and exercise Dam Bae4, Surendra Dasari5, Courtney Dennis6, Charles R. Evans7, David A. Gaul8,
adaptations. Our knowledge thus far and the road still ahead. J. Appl. Physiol. 120, Olga Ilkayeva9,10, Anna A. Ivanova11, Maureen T. Kachman12, Hasmik Keshishian3, Ian R. Lanza13,
683–691 (2016). Ana C. Lira4, Michael J. Muehlbauer10, Venugopalan D. Nair14, Paul D. Piehowski15,
16. Dumke, C. L., Kim, J., Arias, E. B. & Cartee, G. D. Role of kallikrein–kininogen system in Jessica L. Rooney16, Kevin S. Smith17, Cynthia L. Stowe18 & Bingqing Zhao2
insulin-stimulated glucose transport after muscle contractions. J. Appl. Physiol. 92,
657–664 (2002). Analysts
17. Vettor, R. et al. Effect of exercise on plasma kallikrein and muscular phospholipase A2 Natalie M. Clark3, David Jimenez-Morales1, Malene E. Lindholm1, Gina M. Many19,
activity in rats. Mol. Cell. Endocrinol. 45, 65–70 (1986). James A. Sanford19, Gregory R. Smith14, Nikolai G. Vetr17, Tiantian Zhang20,
18. De Lisio, M. & Parise, G. Exercise and hematopoietic stem and progenitor cells: Bingqing Zhao2, Jose J. Almagro Armenteros2, Julian Avila-Pacheco6, Nasim Bararpour2,
protection, quantity, and function. Exerc. Sport Sci. Rev. 41, 116–122 (2013). Yongchao Ge14, Zhenxin Hou20, Anna A. Ivanova11, Shruti Marwaha1, David M. Presby21,
19. Cho, E.-G. et al. MEF2C enhances dopaminergic neuron differentiation of human Archana Natarajan Raja1, Evan M. Savage8, Alec Steep22, Yifei Sun23,24, Si Wu2 & Jimmy Zhen1
embryonic stem cells in a parkinsonian rat model. PLoS ONE 6, e24027 (2011).
20. Lin, R., Warn-Cramer, B. J., Kurata, W. E. & Lau, A. F. v-Src phosphorylation of connexin 43 Animal Study Leadership
on Tyr247 and Tyr265 disrupts gap junctional communication. J. Cell Biol. 154, 815–827 Sue C. Bodine4,25,76 ✉, Karyn A. Esser26,76 ✉, Laurie J. Goodyear21 & Simon Schenk27,76 ✉
(2001).
21. Oughtred, R. et al. The BioGRID database: a comprehensive biomedical resource of
curated protein, genetic, and chemical interactions. Protein Sci.30, 187–200 (2021). Manuscript Writing Group Leads
22. Bye, A. et al. Gene expression profiling of skeletal muscle in exercise-trained and Nicole R. Gay2,75, Pierre M. Jean-Beltran3,75, David Amar1,75,
sedentary rats with inborn high and low VO2max. Physiol. Genomics 35, 213–221 (2008).
23. Hostrup, M. et al. High-intensity interval training remodels the proteome and acetylome Manuscript Writing Group
of human skeletal muscle. eLife 11, e69802 (2022). ✉
Malene E. Lindholm1, Gina M. Many19, Simon Schenk27,76, Stephen B. Montgomery2,17,28,76 ,
24. Yu, G., Wang, L.-G., Yan, G.-R. & He, Q.-Y. DOSE: an R/Bioconductor package for disease Jose J. Almagro Armenteros2, Julian Avila-Pacheco6, Nasim Bararpour2, Sue C. Bodine4,25,76,
ontology semantic and enrichment analysis. Bioinformatics 31, 608–609 (2015). Karyn A. Esser26,76, Facundo M. Fernández8, Zhenxin Hou20, David Jimenez-Morales1,
25. Aguilar, D. Heart failure, diabetes mellitus, and chronic kidney disease: a clinical Archana Natarajan Raja1, James A. Sanford19, Stuart C. Sealfon14, Gregory R. Smith14,
conundrum. Circ. Heart Fail. 9, e003316 (2016). Michael P. Snyder2,76 ✉, Nikolai G. Vetr17, Bingqing Zhao2 & Tiantian Zhang20
26. van Moorsel, C. H. M. Desmoplakin: an important player in aging lung disease. Am. J.
Respir. Crit. Care Med. 202, 1201–1202 (2020). Senior Leadership
27. Wang, T. et al. Myosin light chain kinase (MYLK) coding polymorphisms modulate human Joshua N. Adkins19, Euan Ashley1, Sue C. Bodine4,25,76, Charles F. Burant7, Steven A. Carr3,76 ✉,
lung endothelial cell barrier responses via altered tyrosine phosphorylation, spatial Clary B. Clish6, Gary Cutter29, Karyn A. Esser26,76, Facundo M. Fernández8, Robert E. Gerszten30,
localization, and lamellipodial protrusions. Pulm. Circ. 8, 2045894018764171 (2018). Laurie J. Goodyear21, William E. Kraus9,10, Ian R. Lanza13, Jun Z. Li22, Michael E. Miller31,
28. Jostins, L. et al. Host–microbe interactions have shaped the genetic architecture of Stephen B. Montgomery2,17,28,76, K. Sreekumaran Nair32, Christopher Newgard10,
inflammatory bowel disease. Nature 491, 119–124 (2012). Eric A. Ortlund20, Wei-Jun Qian19, Simon Schenk27,76, Stuart C. Sealfon14, Michael P. Snyder2,76,
29. Clarke, S. F. et al. Exercise and associated dietary extremes impact on gut microbial Russell Tracy16, Martin J. Walsh23,24 & Matthew T. Wheeler1,76 ✉
diversity. Gut 63, 1913–1920 (2014).
30. Lammers, K. M. et al. Gliadin induces an increase in intestinal permeability and zonulin
release by binding to the chemokine receptor CXCR3. Gastroenterology 135, 194–204.e3 Co-corresponding Authors
(2008). Sue C. Bodine4,25,76, Steven A. Carr3,76, Karyn A. Esser26,76, Stephen B. Montgomery2,17,28,76,
31. Scarpa, M. et al. The epithelial danger signal IL-1α is a potent activator of fibroblasts and Simon Schenk27,76, Michael P. Snyder2,76, Matthew T. Wheeler1,76
reactivator of intestinal inflammation. Am. J. Pathol. 185, 1624–1637 (2015).
32. Wang, Z. et al. Gut flora metabolism of phosphatidylcholine promotes cardiovascular Department of Medicine, Stanford University, Stanford, CA, USA. 2Department of Genetics,
1

disease. Nature 472, 57–63 (2011). Stanford University, Stanford, CA, USA. 3Proteomics Platform, Broad Institute of MIT and
33. Daly, W., Seegers, C., Timmerman, S. & Hackney, A. C. Peak cortisol response to Harvard, Cambridge, MA, USA. 4Department of Internal Medicine, University of Iowa, Iowa
exhausting exercise: effect of blood sampling schedule. Med. Sportiva 8, 17–20 City, IA, USA. 5Department of Quantitative Health Sciences, Mayo Clinic, Rochester, MN,
(2004). USA. 6Metabolomics Platform, Broad Institute of MIT and Harvard, Cambridge, MA, USA.

182 | Nature | Vol 629 | 2 May 2024

7
Department of Internal Medicine, University of Michigan, Ann Arbor, MI, USA. 8School of Michael J. Muehlbauer10, Charles C. Mundorff3, Daniel Nachun17, K. Sreekumaran Nair32,
Chemistry and Biochemistry, Georgia Institute of Technology, Atlanta, GA, USA. 9Department Venugopalan D. Nair14, Michael D. Nestor19, Christopher Newgard10, German Nudelman14,
of Medicine, Duke University, Durham, NC, USA. 10Duke Molecular Physiology Institute, Duke Eric A. Ortlund20, Cadence Pearce3, Vladislav A. Petyuk19, Paul D. Piehowski15, Hanna Pincas14,
University, Durham, NC, USA. 11Emory Integrated Metabolomics and Lipidomics Core, Emory Wei-Jun Qian19, Irene Ramos14, Alexander (Sasha) Raskind12, Stas Rirak14, Jeremy M. Robbins30,
University, Atlanta, GA, USA. 12BRCF Metabolomics Core, University of Michigan, Ann Arbor, Aliza B. Rubenstein14, Frederique Ruf-Zamojski14, Tyler J. Sagendorf19, James A. Sanford19,
MI, USA. 13Division of Endocrinology, Nutrition, and Metabolism, Mayo Clinic, Rochester, MN, Evan M. Savage8, Stuart C. Sealfon14, Nitish Seenarine14, Gregory R. Smith14, Kevin S. Smith17,
USA. 14Department of Neurology, Icahn School of Medicine at Mount Sinai, New York, NY, Michael P. Snyder2,76, Tanu Soni12, Alec Steep22, Yifei Sun23,24, Karan Uppal62, Sindhu Vangeti14,
USA. 15Environmental Molecular Sciences Division, Pacific Northwest National Laboratory, Mital Vasoya14, Nikolai G. Vetr17, Alexandria Vornholt14, Martin J. Walsh23,24, Si Wu2, Xuechen Yu14,
Richland, WA, USA. 16Department of Pathology and Laboratory Medicine, University of Elena Zaslavsky14, Navid Zebarjadi2, Tiantian Zhang20 & Bingqing Zhao2
Vermont, Burlington, VT, USA. 17Department of Pathology, Stanford University, Stanford,
CA, USA. 18Department of Biostatistics and Data Science, Wake Forest University School of Clinical Sites
Medicine, Winston-Salem, NC, USA. 19Biological Sciences Division, Pacific Northwest National Marcas Bamman63, Bryan C. Bergman64, Daniel H. Bessesen64, Thomas W. Buford65,
Laboratory, Richland, WA, USA. 20Department of Biochemistry, Emory University, Atlanta, Toby L. Chambers66, Paul M. Coen67, Dan Cooper68, Gary Cutter29, Fadia Haddad68,
GA, USA. 21Section on Integrative Physiology and Metabolism, Joslin Diabetes Center, Boston, Kishore Gadde69, Bret H. Goodpaster67, Melissa Harris69, Kim M. Huffman9,10,
MA, USA. 22Department of Human Genetics, University of Michigan, Ann Arbor, MI, USA. Catherine M. Jankowski70, Neil M. Johannsen69, Wendy M. Kohrt64, William E. Kraus9,10,
23
Department of Pharmacological Sciences, Icahn School of Medicine at Mount Sinai, New Bridget Lester66, Edward L. Melanson64, Kerrie L. Moreau64, Nicolas Musi71,
York, NY, USA. 24Department of Genetics and Genomic Sciences, Icahn School of Medicine Robert L. Newton Jr72, Shlomit Radom-Aizik68, Megan E. Ramaker10, Tuomo Rankinen69,
at Mount Sinai, New York, NY, USA. 25Aging and Metabolism Research Program, Oklahoma Blake B. Rasmussen73, Eric Ravussin69, Irene E. Schauer64, Robert S. Schwartz64,
Medical Research Foundation, Oklahoma City, OK, USA. 26Department of Physiology and Lauren M. Sparks67, Anna Thalacker-Mercer63, Scott Trappe66, Todd A. Trappe66 & Elena Volpi74
Aging, University of Florida, Gainesville, FL, USA. 27Department of Orthopaedic Surgery,
School of Medicine, University of California, San Diego, La Jolla, CA, USA. 28Department of
Biomedical Data Science, Stanford University, Stanford, CA, USA. 29Department of Biostatistics,
33
Department of Statistics, Stanford University, Stanford, CA, USA. 34Department of Biomedical
University of Alabama at Birmingham, Birmingham, AL, USA. 30Division of Cardiovascular Data Sciences, Stanford University, Stanford, CA, USA. 35Department of Aging and Geriatric
Medicine, Beth Israel Deaconess Medical Center, Boston, MA, USA. 31Division of Public Health Research, University of Florida, Gainesville, FL, USA. 36Section on Gerontology and Geriatric
Sciences, Wake Forest University School of Medicine, Winston-Salem, NC, USA. 32Department Medicine, Wake Forest University School of Medicine, Winston-Salem, NC, USA. 37Department
of Medicine, Mayo Clinic, Rochester, MN, USA. of Health and Exercise Science, Wake Forest University School of Medicine, Winston-Salem,
NC, USA. 38National Institute on Aging, National Institutes of Health, Bethesda, MD, USA.
39
National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of
MoTrPAC Study Group
Health, Bethesda, MD, USA. 40Applied Physiology and Kinesiology, University of Florida,
Bioinformatics Center
Gainesville, FL, USA. 41Department of Biomedical Sciences, University of Missouri, Columbia,
David Amar1,75, Euan Ashley1, Karen P. Dalton1, Trevor Hastie33, Steven G. Hershman1,
MO, USA. 42Department of Medical Pharmacology and Physiology, University of Missouri,
David Jimenez-Morales1, Malene E. Lindholm1, Shruti Marwaha1, Archana Natarajan Raja1,
Columbia, MO, USA. 43Department of Nutrition and Exercise Physiology, University of
Mihir Samdarshi1, Christopher Teng1, Rob Tibshirani33,34, Matthew T. Wheeler1,76 & Jimmy Zhen1
Missouri, Columbia, MO, USA. 44Dalton Cardiovascular Research Center, University of
Missouri, Columbia, MO, USA. 45Department of Kinesiology and Health Education, University
Biospecimens Repository of Texas, Austin, TX, USA. 46Department of Medicine, Division of Endocrinology and Diabetes,
Elaine Cornell16, Nicole Gagne16, Sandy May16, Jessica L. Rooney16 & Russell Tracy16 University of California, Los Angeles, CA, USA. 47Center for Public Health Genomics,
University of Virginia School of Medicine, Charlottesville, VA, USA. 48Section on Clinical,
Administrative Coordinating Center Behavioral, and Outcomes Research, Joslin Diabetes Center, Boston, MA, USA. 49Department
Brian Bouverat35, Christiaan Leeuwenburgh35, Ching-ju Lu35 & Marco Pahor26 of Molecular Physiology and Biophysics, Vanderbilt University, Nashville, TN, USA.
50
Department of Health Sciences, Stetson University, Deland, FL, USA. 51Department of
Medicine, University of Missouri, Columbia, MO, USA. 52NextGen Precision Health, University
Data Management, Analysis, and Quality Control Center
of Missouri, Columbia, MO, USA. 53Cell Biology and Physiology, Internal Medicine, University
Fang-Chi Hsu18, Michael E. Miller31, Scott Rushing18, Cynthia L. Stowe18 & Michael P. Walkup18
of Kansas Medical Center, Kansas City, KS, USA. 54Center for Skeletal Muscle Research at
Robert M. Berne Cardiovascular Research Center, University of Virginia School of Medicine,
Exercise Intervention Core Charlottesville, VA, USA. 55Department of Medicine, University of Virginia School of Medicine,
Barbara Nicklas36 & W. Jack Rejeski37 Charlottesville, VA, USA. 56Department of Pharmacology, University of Virginia School of
Medicine, Charlottesville, VA, USA. 57Department of Molecular Physiology and Biological
NIH Physics, University of Virginia School of Medicine, Charlottesville, VA, USA. 58Fralin Biomedical
John P. Williams38 & Ashley Xia39 Research Institute, Center for Exercise Medicine Research at Virginia Tech Carilion, Roanoke,
VA, USA. 59Department of Human Nutrition, Foods, and Exercise, College of Agriculture and Life
Sciences, Virginia Tech, Blacksburg, VA, USA. 60Department of Computational and Systems
Preclinical Animal Study Sites
Biology, University of Pittsburgh, Pittsburgh, PA, USA. 61Petit Institute of Bioengineering and
Brent G. Albertson21, Dam Bae4, Elisabeth R. Barton40, Sue C. Bodine4,25,76,
Biosciences, Georgia Institute of Technology, Atlanta, GA, USA. 62Department of Medicine,
Frank W. Booth41,42,43,44, Tiziana Caputo21, Michael Cicha4, Luis Gustavo Oliveira De Sousa4,
Emory University, Atlanta, GA, USA. 63Department of Cell, Developmental, and Integrative
Karyn A. Esser26,76, Roger Farrar45, Laurie J. Goodyear21, Andrea L. Hevener46,
Biology, University of Alabama at Birmingham, Birmingham, AL, USA. 64Department of Medicine,
Michael F. Hirshman21, Bailey E. Jackson4, Benjamin G. Ke47, Kyle S. Kramer4, Sarah J. Lessard48,
University of Colorado Anschutz Medical Campus, Aurora, CO, USA. 65Department of Medicine,
Ana C. Lira4, Nathan S. Makarewicz21, Andrea G. Marshall4,49, Pasquale Nigro21, Scott Powers50,
University of Alabama at Birmingham, Birmingham, AL, USA. 66Human Performance Laboratory,
David M. Presby21, Krithika Ramachandran21, R. Scott Rector43,51,52, Collyn Z-T. Richards4,
Ball State University, Muncie, IN, USA. 67Translational Research Institute, AdventHealth, Orlando,
Simon Schenk27,76, John Thyfault53, Zhen Yan54,55,56,57,58,59 & Chongzhi Zang47
FL, USA. 68Department of Pediatrics, University of California, Irvine, CA, USA. 69Pennington
Biomedical Research Center, Baton Rouge, LA, USA. 70College of Nursing, University of
Chemical Analysis Sites Colorado Anschutz Medical Campus, Aurora, CO, USA. 71Department of Medicine, Cedars-Sinai
Joshua N. Adkins19, Jose J. Almagro Armenteros2, Mary Anne S. Amper14, Julian Avila-Pacheco6, Medical Center, Los Angeles, CA, USA. 72Population and Public Health, Pennington Biomedical
Ali Tugrul Balci60, Nasim Bararpour2, Charles F. Burant7, Steven A. Carr3,76, Clarisa Chavez2, Research Center, Baton Rouge, LA, USA. 73Biochemistry and Structural Biology, Center for
Maria Chikina60, Roxanne Chiu2, Natalie M. Clark3, Clary B. Clish6, Surendra Dasari5, Metabolic Health, Barshop Institute for Longevity and Aging Studies, University of Texas Health
Courtney Dennis6, Charles R. Evans7, Facundo M. Fernández8, David A. Gaul8, Science Center, San Antonio, TX, USA. 74Barshop Institute for Longevity and Aging Studies,
Nicole R. Gay2,75, Yongchao Ge14, Robert E. Gerszten30, Marina A. Gritsenko19, University of Texas Health Science Center, San Antonio, TX, USA. 75These authors contributed
Kristy Guevara14, Joshua R. Hansen19, Krista M. Hennig2, Zhenxin Hou20, Chia-Jui Hung2, equally: David Amar, Nicole R. Gay, Pierre M. Jean-Beltran. 76These authors jointly supervised this
Chelsea Hutchinson-Bunch19, Olga Ilkayeva9,10, Anna A. Ivanova11, Pierre M. Jean-Beltran3,75, work: Sue C. Bodine, Steven A. Carr, Karyn A. Esser, Stephen B. Montgomery, Simon Schenk,
Christopher A. Jin2, Maureen T. Kachman12, Hasmik Keshishian3, Ian R. Lanza13, Jun Z. Li22, Michael P. Snyder, Matthew T. Wheeler. ✉e-mail: [email protected]; [email protected];
Xueyun Liu11, Kristal M. Maner-Smith11, D. R. Mani3, Gina M. Many19, Nada Marjanovic14, [email protected]; [email protected]; [email protected]; [email protected];
Matthew E. Monroe19, Stephen B. Montgomery2,17,28,76, Ronald J. Moore19, Samuel G. Moore61, [email protected]

Nature | Vol 629 | 2 May 2024 | 183

Article
Methods ATAC-seq, RRBS and proteomics are available in the following Github
repositories: https://github.com/MoTrPAC/motrpac-rna-seq-pipeline41,
All methods are included in the Supplementary Information. https://github.com/MoTrPAC/motrpac-atac-seq-pipeline42, https://
github.com/MoTrPAC/motrpac-rrbs-pipeline43 and https://github.
Reporting summary com/MoTrPAC/motrpac-proteomics-pipeline44. Normalization and
Further information on research design is available in the Nature quality control scripts are available at https://github.com/MoTrPAC/
Portfolio Reporting Summary linked to this article. MotrpacRatTraining6moQCRep45.

40. Gay, N. R., Amar, D. & MoTrPAC Study Group. Visualization of graphical analysis results:
Data availability Temporal dynamics of the multi-omic response to endurance exercise training across
MoTrPAC data are publicly available via http://motrpac-data.org/ tissues. Zenodo https://doi.org/10.5281/zenodo.7703294 (2023).
41. Raja, A. et al. MoTrPAC/motrpac-rna-seq-pipeline. GitHub https://github.com/MoTrPAC/
data-access. Data access inquiries should be sent to motrpac-helpdesk@ motrpac-rna-seq-pipeline (2023).
lists.stanford.edu. Additional resources can be found at http://motrpac. 42. Gay, N. R., Raja, A. & MoTrPAC Study Group. MoTrPAC/motrpac-atac-seq-pipeline. GitHub
org and https://motrpac-data.org/. Interactive data visualizations are https://github.com/MoTrPAC/motrpac-atac-seq-pipeline (2023).
43. Akre, S., Raja, A., Samdarshi, M. & MoTrPAC Study Group. MoTrPAC/motrpac-rrbs-pipeline.
provided through a website (https://data-viz.motrpac-data.org) and GitHub https://github.com/MoTrPAC/motrpac-rrbs-pipeline (2023).
HTML reports summarizing the multi-omic graphical analysis results 44. Jimenez-Morales, D., Samdarshi, M., Hershman, S. & MoTrPAC Study Group. MoTrPAC/
in each tissue40. Processed data and analysis results are additionally motrpac-proteomics-pipeline. GitHub https://github.com/MoTrPAC/motrpac-proteomics-
pipeline (2023).
available in the MotrpacRatTraining6moData R package39 (https:// 45. Amar, D., Samdarshi, M., Raja, A. & Gay, N. R. MoTrPAC/MotrpacRatTraining6moQCRep.
github.com/MoTrPAC/MotrpacRatTraining6moData). Raw and pro- GitHub https://github.com/MoTrPAC/MotrpacRatTraining6moQCRep (2023).
cessed data for were deposited in the appropriate public reposito- 46. McCarron, A. et al. Phenotypic characterization and comparison of cystic fibrosis rat
models generated using CRISPR/Cas9 gene editing. Am. J. Pathol. 190, 977–993 (2020).
ries as follows. RNA-seq, ATAC-seq and RRBS data were deposited at
the Sequence Read Archive under accession PRJNA908279 and at the
Acknowledgements Funding: The MoTrPAC Study is supported by NIH grants U24OD026629
Gene Expression Omnibus under accession GSE242358; multiplexed (Bioinformatics Center), U24DK112349, U24DK112342, U24DK112340, U24DK112341,
immunoassays were deposited at IMMPORT under accession SDY2193; U24DK112326, U24DK112331, U24DK112348 (Chemical Analysis Sites), U01AR071133,
U01AR071130, U01AR071124, U01AR071128, U01AR071150, U01AR071160, U01AR071158 (Clinical
metabolomics data were deposited at Metabolomics Workbench under
Centers), U24AR071113 (Consortium Coordinating Center), U01AG055133, U01AG055137 and
project ID PR001020; and proteomics data were deposited at MassIVE U01AG055135 (PASS/Animal Sites). This work was also supported by other funding sources:
under accessions MSV000092911, MSV000092922, MSV000092923, NHGRI Institutional Training Grant in Genome Science 5T32HG000044 (N.R.G.), National
Science Foundation Graduate Research Fellowship Grant No. NSF 1445197 (N.R.G.), National
MSV000092924, MSV000092925 and MSV000092931. We used the
Heart, Lung, and Blood Institute of the National Institute of Health F32 postdoctoral fellowship
following external datasets: release 96 of the Ensembl R. norvegicus award F32HL154711 (P.M.J.B.), the Knut and Alice Wallenberg Foundation (M.E.L.), National
(rn6) genome (https://ftp.ensembl.org/pub/release-96/fasta/rattus_ Science Foundation Major Research Instrumentation (MRI) CHE-1726528 (F.M.F.), National
Institute on Aging P30AG044271 and P30AG003319 (N.M.), and NORC at the University of
norvegicus/dna/) and gene annotation (https://ftp.ensembl.org/pub/
Chicago grant no. P30DK07247 (E.R.). Parts of this work were performed in the Environmental
release-96/gtf/rattus_norvegicus/Rattus_norvegicus.Rnor_6.0.96.gtf. Molecular Science Laboratory, a US Department of Energy national scientific user facility at
gz); RefSeq protein database (https://ftp.ncbi.nlm.nih.gov/refseq/R_ Pacific Northwest National Laboratory in Richland, WA. The views expressed are those of the
authors and do not necessarily reflect those of the NIH or the US Department of Health and
norvegicus/, downloaded 11/2018); the NCBI gene2refseq mapping files Human Services. Some figures were created using Biorender.com. Fig. 1b was modified with
(https://ftp.ncbi.nlm.nih.gov/gene/DATA/gene2refseq.gz, accessed 18 permission from ref. 46.
December 2020); RGD rat gene annotation (https://download.rgd.mcw.
Author contributions All authors reviewed and revised the manuscript. Detailed author
edu/data_release/RAT/GENES_RAT.txt, accessed 12 November 2021); contributions are provided in the Supplementary Information.
BioGRID v4.2.193 (https://downloads.thebiogrid.org/File/BioGRID/
Release-Archive/BIOGRID-4.2.193/BIOGRID-ORGANISM-4.2.193.tab3. Competing interests S.C.B. has equity in Emmyon, Inc. G.R.C. sits on data and safety monitoring
boards for AI Therapeutics, AMO Pharma, Astra-Zeneca, Avexis Pharmaceuticals, Biolinerx,
zip); STRING v11.5 (https://stringdb-downloads.org/download/protein.
Brainstorm Cell Therapeutics, Bristol Meyers Squibb/Celgene, CSL Behring, Galmed
physical.links.v11.5/10116.protein.physical.links.v11.5.txt.gz); GEN- Pharmaceuticals, Green Valley Pharma, Horizon Pharmaceuticals, Immunic, Mapi
CODE release 39 metadata and annotation files (https://ftp.ebi.ac.uk/ Pharmaceuticals, Merck, Mitsubishi Tanabe Pharma Holdings, Opko Biologics, Prothena
Biosciences, Novartis, Regeneron, Sanofi-Aventis, Reata Pharmaceuticals, NHLBI (protocol
pub/databases/gencode/Gencode_human/release_39/, accessed 20 review committee), University of Texas Southwestern, University of Pennsylvania, Visioneering
January 2022); MatrisomeDB (https://doi.org/10.1093/nar/gkac1009); Technologies, Inc.; serves on consulting or advisory boards for Alexion, Antisense Therapeutics,
MitoPathways database available through MitoCarta (https://personal. Biogen, Clinical Trial Solutions LLC, Genzyme, Genentech, GW Pharmaceuticals, Immunic,
Klein-Buendel Incorporated, Merck/Serono, Novartis, Osmotica Pharmaceuticals, Perception
broadinstitute.org/scalvo/MitoCarta3.0/); PTMSigDB v1.9.0 PTM set Neurosciences, Protalix Biotherapeutics, Recursion/Cerexis Pharmaceuticals, Regeneron,
database (https://doi.org/10.1074/mcp.TIR118.000943); UniProt Roche, SAB Biotherapeutics; and is the president of Pythagoras Inc., a private consulting
human proteome FASTA for canonical protein sequences (UniProtKB company. S.A.C. is a member of the scientific advisory boards of Kymera, PrognomiQ, PTM
BioLabs, and Seer. M.P.S. is a cofounder and scientific advisor to Personalis, Qbio, January AI,
query “reviewed:true AND proteome:up000005640”, download Filtricine, SensOmics, Protos, Fodsel, Rthm, Marble and scientific advisor to Genapsys, Swaz,
date 3 March 2021); the CIBERSORT LM22 leukocyte gene signature Jupiter. S.B.M. is a consultant for BioMarin, MyOme and Tenaya Therapeutics. D.A. is currently
matrix (https://doi.org/10.1007/978-1-4939-7493-1_12); published employed at Insitro, South San Francisco, CA. N.R.G. is currently employed at 23andMe,
Sunnyvale, CA. P.M.J.B. is currently employed at Pfizer, Cambridge, MA. Insitro, 23andMe and
results from Amar et al.8, Bye et al.22 and Hostrup et al.23; and GTEx v8 Pfizer had no involvement in the work presented here.
gene-expression data (dbGaP Accession phs000424.v8.p2). Details are
provided in the Supplementary Information, Methods. Additional information
Supplementary information The online version contains supplementary material available at
https://doi.org/10.1038/s41586-023-06877-w.
Correspondence and requests for materials should be addressed to Sue C. Bodine,
Code availability Karyn A. Esser, Simon Schenk, Stephen B. Montgomery, Michael P. Snyder, Steven A. Carr or
Code for reproducing the main analyses is provided in the Motrpa- Matthew T. Wheeler.
Peer review information Nature thanks Atul Deshmukh, Jorge Ruas and the other, anonymous,
cRatTraining6mo R package38 (https://motrpac.github.io/Motrpa- reviewer(s) for their contribution to the peer review of this work. Peer review reports are available.
cRatTraining6mo/). MoTrPAC data processing pipelines for RNA-seq, Reprints and permissions information is available at http://www.nature.com/reprints.
Extended Data Fig. 1 | See next page for caption.
Article
Extended Data Fig. 1 | Animal phenotyping and data availability. capacity through a VO2max test until exhaustion. Data are reported in
a-d) Clinical measurements before and after the training intervention in ml/(kg.min) for all individual rats and time points. b) Body fat percentage.
untrained control rats (SED), 4-week trained rats (4w), and 8-week trained rats c) Percent lean mass. (b-c) were assessed through nuclear magnetic resonance
(8w). Data are displayed pre and post for each individual rat (connected by a spectroscopy. d) Body weight (in grams). e) Description of available datasets.
line), with males in blue and females in pink. Filled symbols (n = 5 per sex and Colored cells indicate that data are available for that tissue and assay. Individual
time point) represent rats used for all omics analyses, whereas the rat utilized panels and platforms are shown for metabolomics and the multiplexed
for proteomics only (n = 1 per sex and time point) is represented by a non-filled immunoassays. f) Detailed availability of sample-level data across assays. Each
symbol. Significant results by ANOVA of the overall group effect (#, p < 0.05; column represents an individual animal, ordered by training group and colored
##, p < 0.01) and interaction between group and time (§, p < 0.05; §§ p < 0.01) are by sex. Gray cells indicate that data were generated for that animal and assay;
indicated. Significant within-group differential responses from a Bonferroni black cells indicate that data were not generated. Rows are ordered by ome and
post hoc test are indicated (*, q-value < 0.05; **, q-value < 0.01). a) Aerobic colored by assay and tissue.
Extended Data Fig. 2 | See next page for caption.
Article
Extended Data Fig. 2 | Quality control metrics for omics data. a) Proteomics fraction of reads in peaks (FRiP, bottom) across ATAC-seq samples per tissue.
multiplexing design using TMT11 reagents for isobaric tagging and a pooled h) Distributions of RNA integrity numbers (RIN, top) and median 5′ to 3′ bias
reference sample. The diagram describes processing of a single tissue. Following (bottom) across samples in each tissue in the RNA-Seq data. i) Percent
multiplexing, peptides were used for protein abundance analysis, serial PTM methylation of CpG, CHG and CHH sites in the RRBS data. For boxplots in
enriched for phosphosite and optional acetylsite quantification, or ubiquitylsite (h,i): center line represents median; box bounds represent 25th and 75th
quantification through enrichment of lysine-diglycine ubiquitin remnants. percentiles; whiskers represent minimum and maximum excluding outliers;
b) Total number of fully quantified proteins per plex in each global proteome filled dots represent outliers. j) Number of wells across multiplexed
dataset. c-e) The total number of fully quantified phosphosites (c), acetylsites immunoassays with fewer than 20 beads. Measurements from these 182 wells
(d), and ubiquitylsites (e) per plex in each dataset. f) Distributions of coefficients were excluded from downstream analysis. k) 2D density plot of targeted
of variation (CVs) calculated from metabolomics features identified in pooled analytes’ mean fluorescence intensity (MFI) versus corresponding CHEX4 MFI
samples and analyzed periodically throughout liquid chromatography-mass from the same well for each multiplexed immunoassay measurement, where
spectrometry runs. CVs were aggregated and plotted separately for named and CHEX4 is a measure of non-specific binding.
unnamed metabolites. g) Transcription start site (TSS) enrichment (top) and
Extended Data Fig. 3 | Permutation tests. a-b) Permutation tests of groups considered sexually dimorphic if for at least one time point the z-score (absolute)
within males (a) and females (b). For each sex, the original group labels were difference between males and females was greater than 3. c) Counts of sexually
shuffled to minimize the number of animal pairs that remain in the same group. dimorphic genes among the IHW-selected genes of the original data. d) Counts
Only the group labels were shuffled and all other covariates remained as in the of sexually dimorphic genes among the 5% FDR selected genes within each
original data. For each permuted dataset, the differential abundance pipeline permuted dataset. Each boxplot in (a-d) represents the differential abundance
was rerun and the number of transcripts that were selected at 5% FDR adjustment analysis results over 100 permutations of the transcriptomics data in a specific
were re-counted. c-d) Permutation tests of sex within groups. For each group tissue. Center line represents median; box bounds represent 25th and 75th
and each sex, half of the animals were selected randomly and their sex was percentiles; whiskers represent minimum and maximum excluding outliers;
swapped. Only the sex labels were shuffled and all other covariates remained open circles represent outliers. Added points represent the results of the true
as in the original data. For each permutation the differential analysis pipeline data labels, and their shape corresponds to the empirical p-value (●: p > 0.05;
was rerun and the timewise summary statistics were extracted. A gene was ×: 0.01 < p < 0.05; *: p ≤ 0.01).
Article

Extended Data Fig. 4 | Correlations between proteins and transcripts KEGG and Reactome pathways were queried, and redundant pathways were
throughout endurance training. a) Number of tissues in which each gene, removed (i.e., those with an overlap of 80% or greater with an existing pathway).
including features mapped to genes from all omes, is training-regulated. Only c) Heatmaps showing the Pearson correlation between the TRNSCRPT and PROT
differential features from the subset of tissues with deep molecular profiling timewise summary statistics (z- and t-scores, respectively) (top, gene-level) and
(lung, gastrocnemius, subcutaneous white adipose, kidney, liver, and heart) pathway-level enrichment results (Gene Set Enrichment Analysis normalized
and the subset of omes that were profiled in all six of these tissues (DNA enrichment scores) (bottom, pathway-level). d) Scatter plots of pathway GSEA
methylation, chromatin accessibility, transcriptomics, global proteomics, NES of the TRNSCRPT and PROT datasets in the seven tissues for which these
phosphoproteomics, multiplexed immunoassays) were considered. Numbers data were acquired. Pathways showing high discordance or agreement across
above each bar indicate the number of genes that are differential in exactly the TRNSCRPT and PROT and with functional relevance or general interest were
number of tissues indicated on the x-axis. b) Pathways significantly enriched by highlighted.
tissue-specific training-regulated genes represented in Fig. 2a (q-value < 0.1).
Extended Data Fig. 5 | Heat shock response. a) Scatter plots of the protein points indicate those with a large differential response at the protein level.
t-scores (PROT) versus the transcript z-scores (TRNSCRPT) by gene at 8 weeks b-c) Line plots showing protein b) and transcript (c) log 2 fold-changes relative
of training (8 W) relative to sedentary controls. Data are shown for the seven to the untrained controls for a subset of heat shock proteins with increased
tissues for which both proteomics and transcriptomics was acquired. Red abundance during exercise training. Each line represents a protein in a single
points indicate genes associated with the heat shock response, and the labeled tissue.
Article

Extended Data Fig. 6 | See next page for caption.

Extended Data Fig. 6 | Regulatory signaling pathways modulated by clustering. e) (top) Log 2 fold-change of GJA1 and CDH2 protein abundance in
endurance training. a) Heatmap of differences in TF motif enrichment in the heart. No significant response to exercise training was observed for these
training-regulated genes across tissues. Each value reflects the average proteins (F-test; q-value > 0.05). (bottom) Log 2 fold-changes for selected Src
difference in motif enrichment for shared transcription factors. Tissues are kinase phosphosite targets, GJA1 pY265 and CDH2 pY820, in the heart. These
clustered with complete linkage hierarchical clustering. b) (left) Filtered PTM- phosphosites show a significant response to exercise training (F-test, 5% FDR).
SEA results for the liver showing kinases and signaling pathways with increased Error bars=SEM. f) Gene Set Enrichment Analysis (GSEA) results from the heart
activity. (right) Heatmap showing t-scores for phosphosites within the HGF global proteome dataset using the matrisome gene set database. Heatmap
signaling pathway. c) Hypothetical model of HGF signaling effects during shows NES as color and enrichment p-value as dot size. Rows are clustered
exercise training. Phosphorylation of STAT3 and PXN is known to modulate cell using hierarchical clustering. g) Log 2 fold-change for basement membrane
growth and cell migration, respectively. Error bars=SEM. d) Filtered PTM-SEA proteins in heart. Proteins showing a significant response to exercise training
results for the heart showing selected kinases with significant enrichments in are highlighted in orange (F-test; 5% FDR). Error bars=SEM. h) Log 2 protein fold-
at least one time point. Heatmap shows the NES as color and enrichment p-value change of NTN1 protein abundance in heart. A significant response to exercise
as dot size. Kinases are grouped by kinase family and sorted by hierarchical training was observed for these proteins (F-test; 5% FDR). Error bars=SEM.
Article

Extended Data Fig. 7 | Graphical representation of differential results. graphical representation of the differential analysis results. Top: the z-scores
a) Number of training-regulated features assigned to groups of graphical states of four features. A positive score corresponds to up-regulation (red), and a
across tissues and time. Red points indicate features that are up-regulated in at negative score corresponds to down regulation (blue). Bottom: the assignment
least one sex (e.g., only in males: F0_M1; only in females: F1_M0; in both sexes: of features to node sets and full path sets (edge sets are not shown for
F1_M1), and blue points indicate features down-regulated in at least one sex conciseness but can be easily inferred from the full paths). Node labels follow
(only in males: F0_M-1; only in females: F-1_M0; in both sexes: F-1_M-1). Green the [time]_F[x]_M[y] format where [time] shows the animal sacrifice week and
points indicate features that are up-regulated in males and down-regulated in can take one of (1w, 2w, 4w, or 8w), and [x] and [y] are one of (−1,0,1), corresponding
females or vice versa (F-1_M1 and F1_M-1, respectively). Point size is proportional to down-regulation, no effect, and up-regulation, respectively. c) Graphical
to the number of features. Point opacity is proportional to the within-tissue representation of the feature sets. Columns are training time points, and rows
fraction of features represented by that point. Features can be represented in are the differential abundance states. Node and edge sizes are proportional to
multiple points. The number of omes profiled in each tissue is provided in the number of features that are assigned to each set.
parentheses next to the tissue abbreviation. b) A schematic example of the
Extended Data Fig. 8 | Key pathway enrichments per tissue. Key pathway omes for which the pathway was significantly enriched (q-value < 0.01) (lighter
enrichments for features that are up-regulated in both sexes at 8 weeks of gray: 1 ome; darker gray: 2 omes; black: 3 omes). Pathways were selected from
training in each tissue. For display purposes, enrichment q-values were floored Supplementary Table 10.
to 1e-10 (Enrichment FDR (−log10) = 10). Bars are colored by the number of
Article

Extended Data Fig. 9 | See next page for caption.

Extended Data Fig. 9 | Associations with signatures of human health and between the rat gene sets from (f) and the high-heterogeneity human
complex traits. a) Jaccard coefficients between gene sets identified by meta-analysis genes (I2 > 75%). h) -log10 overlap p-values (Fisher’s exact test),
different omes in 8-week gastrocnemius up-regulated features (“X” marks comparing rat female gastrocnemius and vastus lateralis week-8 differential
overlap p > 0.05). b) Network connectivity p-values (Pathways, Biogrid, and transcripts from this study (p < 0.01) and the differential genes from the rat
string) among the gastrocnemius week-8 multi-omic genes and with the female soleus data of Bye et al. (p < 0.01). HCR: high capacity runners, LCR: low
single-omic genes. c) Proportion of features from each ome represented in the capacity runners. i) A comparison of rat gastrocnemius differential proteins
gastrocnemius response clusters, identified by the network clustering analysis. from this study (p < 0.01) and the human endurance training proteomics
d-g) Overlap between our rat vastus lateralis differential expression results results of Hostrup et al. (p < 0.01) using Fisher’s exact test. Left: -log10 overlap
and the meta-analysis of human long-term exercise studies by Amar et al. d-e) p-values. Right: -log10 sex concordance p-values. j) Statistics of the overlapping
Spearman correlation (d) and its significance (e) between the meta-analysis fold- proteins from (i), week-8 female comparison (y: rat z-scores, x: human t-scores).
changes and the log2 fold-changes foreach sex and time point. f) GSEA results. k) DOSE disease enrichment results of the white adipose, kidney, and liver gene
Genes were ranked by meta-analysis (−log10 p-value*log 2 fold-change) and the rat sets. DOSE was applied only on diseases that are relevant for each tissue. The
training-differential, sex-consistent gene sets were tested for enrichment at the network shows the results for the sex-consistent down-regulated features at
bottom of the ranking (negative scores) or the top (positive scores). g) Overlap week-8.
Article

Extended Data Fig. 10 | See next page for caption.

Extended Data Fig. 10 | Characterization of the extent of sex difference in with a label that provides the ome and the number of differential features
the endurance training response. The extent of sex differences in the training represented. Right plot for each tissue: 2D density plot of Δ AUC against the
response were characterized in two ways: first, by correlating log 2 fold-changes correlation between the male and female log2 fold-changes for each training-
between males and females for each training-differential feature; second, by differential feature used to simultaneously evaluate sex differences in the
calculating the difference between the area under the log 2 fold-change curve for direction and magnitude of the training response. Points at the top-center of
each training-differential feature, including a (0,0) point (Δ AUC, males - females). these 2D density plots represent features with high similarity between males
The first approach characterizes differences in direction of effect while the and females in terms of both direction and magnitude; features on the right and
second approach characterizes differences in magnitude. Left plot for each left sides of the plots represent features with greater magnitudes of response
tissue: density line plots of correlations from the first approach. Densities or in males and females, respectively.
correlations corresponding to features in each ome are plotted separately,
Article

Extended Data Fig. 11 | See next page for caption.

Extended Data Fig. 11 | Sex differences in the endurance training response. across all features. The closer a colored line is to this average, the darker it is
a) Heatmap of the training response of immunoassay analytes across tissues. (distance calculated using sum of squares). e) Line plots of transcript-level log 2
Gray indicates no data. Bars indicate the number of training-regulated analytes fold-changes corresponding to six transcription factors (TFs) whose motifs are
in each tissue (top) and the number of tissues in which the analyte is training- significantly enriched by transcripts in (d). TF motif enrichment q-values are
regulated (right, 5% FDR). b) Training-differential cytokines across tissues. provided in the legend (error bars = SEM). f) Male versus female NES from
5, 24, and 9 cytokines were annotated as anti-, pro-, and pro/anti- inflammatory, PTM-SEA in the lung. Anticorrelated points corresponding to PRKACA NES are
respectively. Bars indicate the number of annotated cytokines in each category in dark red. g) Line plots of standardized abundances of training-differential
that are differential (5% FDR). c) Counts of early vs. (1- or 2-week) vs. late (4- or phosphosites that follow the largest graphical edges of phosphosites in the lung
8-week) differential cytokines, according to states assigned by the graphical (1w_F1_M-1 − >2w_F1_M-1 − >4w_F0_M-1). h) Top ten kinases with the greatest
analysis, including all tissues. Cytokines with both early and late responses in over-representation of substrates (proteins) corresponding to training-
the same tissue were excluded. d) Line plots of standardized abundances of differential phosphosites in (g). MeanRank scores by library are shown, as
training-differential features that follow the largest graphical path in the adrenal reported by KEA3. i) Line plots showing phosphosite-level log 2 fold-changes
gland (i.e., 1w_F-1_M1 − >2w_F-1_M0 − >4w_F-1_M0 − >8w_F-1_M0 according to of PRKACA phosphosite substrates identified in the lung as differential with
our graphical analysis notation). The black line represents the average value disparate sex responses (error bars = SEM).
Article

Extended Data Fig. 12 | Assessment of immune responses to endurance differential features plotted in (b). # indicates when the distribution of Pearson
training. a) Heatmap of the number and percent of KEGG and Reactome correlations for a set of at least two markers is significantly different from 0
immune pathways significantly enriched by training-regulated features at 8 (two-sided one-sample t-test, 5% BY FDR). When only one marker is used to
weeks. b) Line plots of standardized abundances of training-differential define a category on the y-axis, the gene name is provided in parentheses.
proteins in white adipose tissue up-regulated only in males at 8 weeks. Black d) Trajectories of mean absolute signal of various immune cell types in
line shows average across all features. c) Boxplots of the sample-level Pearson BAT or WAT-SC following deconvolution of bulk RNA-Seq with CIBERSORTx
correlation between markers of immune cell types, lymphatic tissue, or cell (error bars = SEM). e) Immune cell type enrichment analysis results of training-
proliferation and the average value of features in (b) at the protein level. Center differentially expressed transcripts. Points represent significant enrichments
line represents median; box bounds represent 25th and 75th percentiles; (5% FDR, one-sided Mann-Whitney U test). f) Line plots showing the log 2 fold-
whiskers represent minimum and maximum excluding outliers; filled dots changes for Cxcr3 and Il1a transcripts in the small intestine (error bars = SEM).
represent outliers. A pink point indicates that the marker is also one of the
Extended Data Fig. 13 | Metabolic effects of endurance training. a) Significant error bars = SEM). d) Volcano plots showing abundance changes (log 2 fold-
enrichments for relevant categories of KEGG metabolism pathways from changes; logFC) and significance (-log10 nominal p-values) for acyl-carnitines.
features that are up- or down- regulated in both sexes at 8 weeks (8w_F1_M1 and Features are colored based on the carnitine chain length. e) Protein abundance
8w_F-1_M-1 nodes, respectively). Triangles point in the direction of the response changes in the glycolysis and gluconeogenesis pathway in the heart tissue after
(up or down). Points are colored by ome. b) Log 2 fold-change of metabolites 8 weeks of training. Line plots show the log 2 fold-changes over the training time
regulated across many tissues (F-Test, 5% FDR, error bars=SEM). c) Log 2 course (error bars = SEM). Red and blue boxes indicate a statistically significant
fold-change of training-regulated metabolites: 1-methylhistidine in the kidney, (F-test, 5% FDR) increase and decrease in abundance, respectively, for both
cortisol in the kidney, and 1-methylnicotinamide in the liver (F-Test, 5% FDR, males and females at 8 weeks.
Article

Extended Data Fig. 14 | Mitochondria and peroxisome adaptations to change throughout the training time course (F-test, 5% FDR). Center line
endurance training. a) Boxplots showing the percent of mitochondrial represents median; box bounds represent 25th and 75th percentiles; whiskers
genome reads across samples in each tissue that map to the mitochondrial represent minimum and maximum excluding outliers; blue dots represent
genome (% MT reads). b) Comparison of % MT reads between untrained outliers. d) GSEA using the MitoCarta MitoPathways gene set database and
controls and animals trained for 8 weeks. Plot shows tissues with a statistically transcriptome (TRNSCRPT) or phosphoproteome (PHOSPHO) differential
significant change after 8 weeks in at least one sex (red asterisk, two-sided analysis results. NES are shown for significant pathways (10% FDR) for all
Dunnett’s test, 10% FDR). For boxplots in (b,c): center line represents median; tissues, sexes, and time points within the heatmap. Mitochondria pathways
box bounds represent 25th and 75th percentiles; whiskers represent minimum (rows) are grouped using the parental group in the MitoPathways hierarchy.
and maximum excluding outliers; filled dots represent outliers. c) Boxplots e) Protein abundance and protein acetylation level changes in the peroxisome
showing the percent of mitochondrial genome reads across tissue, sex, and KEGG pathway in the liver tissue after 8 weeks of training. Red boxes indicate an
time points. Center line represents median; box bounds represent 25th and increase in abundance for both males and females, while red circles indicate an
75th percentiles; whiskers represent minimum and maximum excluding increase in at least one acetylsite within the protein (8w_F1_M1 cluster).
outliers; open circles represent outliers. Red asterisks indicate a significant