Exercise training remodels inter-organ endocrine networks

Abstract

Background: Exercise induces organism-wide molecular adaptations, partly mediated by humoral factors released in response to acute and chronic physical activity. However, the extent and specificity of endocrine effects from training-induced secreted factors remain unclear.

Methods: Here, we applied systems genetics approaches to quantify inter-organ endocrine networks using multi-tissue transcriptomics and proteomics data collected from endurance-trained rats in The Molecular Transducers of Physical Activity Consortium (MoTrPAC).

Results: Eight weeks of endurance training significantly altered both the magnitude and specificity of endocrine effects across multiple origin-target tissue pairs. Subcutaneous white adipose tissue emerged as a key endocrine regulator impacted by training, while extracellular matrix-derived factors were identified as globally regulated secretory features in trained vs sedentary animals. Notably, secretory Wnt signaling factors were identified as key mediators of exercise-induced endocrine adaptations in multiple tissues.

Conclusion: Our systems genetics framework provides an unprecedented atlas of inter-organ communication significantly remodeled by endurance exercise, serving as a valuable resource for novel exerkine discovery.

Keywords: Endocrine network; Exercise; MoTrPAC; Tissue crosstalk.

Figure 1

Workflow of QENIE and validation of physiological relevance of S_sec. A) Schematic representation of the QENIE workflow and its implementation. B) Number of secretory protein-coding transcripts and proteins identified in each origin tissue after secretome filtering. White bars and numbers within the colored bars indicate the count of secretory-protein coding transcripts or secretory proteins that were significantly different between Control and TR8W. CON, Control; TR, Training. C) S_sec of Lep (WAT-SC-to-HYPOTH) across groups. Fgsea results of HYPOTH transcripts correlating with WAT-SC-Lep across groups are shown. D) S_sec of Il15 (SKM-GN-to-WAT-SC) across groups. Fgsea results of WAT-SC transcripts correlating with SKM-GN-Il15 across groups are shown. E) S_sec of TGF-β2 (WAT-SC-to-WAT-SC) across groups. Fgsea results of WAT-SC proteins correlating with WAT-SC- TGF-β2 across groups are shown. QENIE, Quantitative Endocrine Network Interaction Estimation; S_sec, Secretome score; ADRNL, Adrenal gland; BAT, brown adipose tissue; COLON, colon; CORTEX, cerebral cortex; HEART, heart; HIPPOC, hippocampus; HYPOTH, hypothalamus; KIDNEY, kidney; LIVER, liver; LUNG, lung; PLASMA, plasma; SKM-GN, gastrocnemius (skeletal muscle); SKM-VL, vastus lateralis (skeletal muscle); SMLINT, small intestine; SPLEEN, spleen; VENACV, vena cava; WAT-SC, subcutaneous white adipose tissue. TR, Training; GOBP, Gene Ontology Biological Process; GOCC, Gene Ontology Cellular Component; GOMF, Gene Ontology Molecular Function; NES, Normalized Enrichment Score; BH, Benjamini-Hochberg.

Figure 2

S_sec rank difference in Control vs. TR8W. A) Volcano plot showing results of Wilcoxon signed-rank test comparing gene-to-gene S_sec rank differences between Control and TR8W. B) Network plot showing significantly different Ssec ranking in origin-target pairs in Control vs. TR8W. Red arrows represent significant difference (adjusted p < 0.05). C) Volcano plot showing results of Wilcoxon signed-rank test comparing protein-to-protein S_sec rank differences between Control and TR8W. Significant (adjusted p < 0.05) origin-target pairs are highlighted in black. Color-coded padded boxes represent different origin tissues. Matched origin-target pairs (e.g., CORTEX→CORTEX) are excluded from A and B but were included in the statistical testing.

Figure 3

Top secretory-protein coding transcripts with the largest S_sec rank difference in Control vs. TR8W. A) Bar plots displaying the results of a paired t-test comparing S_sec values between Control and TR8W for the top 20 global transcripts exhibiting the largest S_sec rank differences between the two groups. B) Top 15 transcripts exhibiting the largest S_sec rank difference between the Control and TR8W for WAT-SC-to-SKM-VL, WAT-SC-to-HYPOTH, VENACV-to-HIPPOC, LUNG-to-CORTEX, and SKM-GN-to-HEART. Extracellular score (range: 0–5) obtained from COMPARTMENTS is shown in green numeric next to each feature. Pink triangles represent S_sec ranks in Control, and dark green circles represent S_sec ranks in TR8W. Features highlighted in subsequent figures are colored in red. C) Fgsea results of SKM-VL transcripts correlating with WAT-SC-Sfrp4 across groups. D) Fgsea results of HYPOTH transcripts correlating with WAT-SC-Wnt4 and WAT-SC-Ccn4 across groups. E) Fgsea results of HIPPOC transcripts correlating with VENACV-Wnt7b, VENACV-Wnt16, and VENACV-Mdk across groups. FFgsea results of CORTEX proteins correlating with LUNG-Wif1 across groups.

Figure 4

Feature- and tissue-specific remodeling of Wnt network by exercise training. A) Heatmaps displaying the average difference in S_sec (TR8W – Control) for positive, negative, and dual regulators of Wnt signaling. We curated 13 positive, 19 negative, and 8 dual-role Wnt transcripts from the Gene Ontology database (see “Curation of secretory Wnt signaling lists” in Methods). To assess the effect of endurance training on Wnt signaling for each origin-target tissue pair, we calculated the difference in Ssec between TR8W and Control. Ssec differences were z-scaled and averaged within each Wnt group (positive, negative, dual). Green shading indicates an overall increase in training-induced signaling, while pink indicates a reduction. Rows represent origin tissues, and columns represent target tissues. B) Heatmaps illustrating the average difference in Ssec (TR8W – Control) for secretory Wnt factors, summarized by origin and target tissues. Each Ssec difference was z-scaled before averaging across origin or target tissues. C) Bar graphs showing the normalized S_sec rank difference between TR8W and Control for the top 10 origin-target tissue pairs with the largest rank differences for Rspo1, Wnt3a, Nog, and Mdk. Normalized rank difference was calculated by dividing the absolute S_sec rank difference between TR8W and Control by the total number of secretory genes in the origin tissue. The normalized rank difference ranges from 0 to <1. Dark bars indicate a higher rank in TR8W, while pink bars indicate a higher rank in Control.