Molecular Insights From Multiomics Studies of Physical Activity

Regular physical activity can help reduce the incidence of obesity and obesity-associated metabolic diseases (1–3). In contrast, physical inactivity has become increasingly common in the U.S. population (4), representing a significant modifiable risk factor for cardiovascular diseases, obesity, diabetes, cancer, and overall mortality (5,6). Understanding the molecules and pathways involved in physical activity’s cardiometabolic benefits has garnered significant interest (7). This understanding serves as a crucial first step toward developing future therapies that could potentially harness the health benefits of exercise.

Historically, molecular studies of physical activity have focused on specific molecules in individual metabolic tissues, such as muscle, adipose tissue, liver, and pancreas. In the past decade, advancements in multiomics technologies—including metabolomics, proteomics, phosphoproteomics, lipidomics, single-cell RNA sequencing (scRNA-seq), and epigenomics—have offered unparalleled opportunities to extensively investigate the molecular changes associated with physical activity at an organism-wide scale (Fig. 1). Numerous exercise “omics” data sets start to emerge (8,9). These data sets provide a high-dimensional and detailed view of the molecular processes involved in exercise, further refining our understanding of the molecular events related to this complex physiological stimulus. Simultaneously, the vast amount of data generated with these modern techniques have raised new questions and opened up potential avenues for discovery regarding the molecules and molecular mechanisms underlying physical activity’s beneficial effects.

In this article, we will focus on the latest developments using multiomics approaches to explore the molecules of physical activity. We begin with a discussion of how multiomics technologies have provided new insights into the systemic effects of physical activity, including the integrative responses across organs and the molecules and mechanisms mediating tissue communication during exercise. We also discuss multiomics studies that revisit the classical exercise-responsive metabolic tissues, including muscle and adipose, and how these studies have extended our understanding of the responses of nonparenchymal cells, including immune cells, to physical activity. Lastly, we share our perspectives on the future directions for multiomics technologies as applied to exercise research.

One insight from multiomics data sets is the coordinated molecular responses across multiple organ systems in response to physical activity. These responses include specific pathways that are broadly activated across multiple tissues, as well as the increased correlation of metabolic profiles between anatomically segregated cell types and organs. These studies collectively suggest that one effect of physical activity is to reset and to coordinate molecular processes across multiple organ systems.

For instance, several different multiomics data sets have revealed a coordinated activation of the heat shock response of multiple cell types and organs after exercise training. Heat shock proteins (HSPs) are a family of ancient proteins that act as molecular chaperones, with the specific role of preserving cellular homeostasis in response to stress (10). Previously, Salo et al. (11) showed that a specific member of the HSP family, HSP72, is induced in specific tissues such as the liver and skeletal and cardiac muscle after acute exercise. In combining transcriptomic and proteomic analyses for 6-month-old rats of both sexes after 8 weeks of treadmill training, the Molecular Transducers of Physical Activity Consortium (MoTrPAC) consortium found evidence of a consistent activation of the heat shock response across multiple organs—including the cortex, heart, lung, liver, kidney, subcutaneous white adipose tissue, and skeletal muscle (8). The degree to which the levels of heat shock proteins increased was notably influenced by both the duration of exercise training and the sex of the rats. For example, in male rats, an 8-week training regimen resulted in a 16-fold increase in the abundance of HSPA1B and HSPB1 proteins, whereas only a 2-fold increase was observed after 1 week of training. Additionally, the intensity of the exercise-induced heat shock response was less significant in female rats, as the levels of individual HSPs did not increase by more than fourfold even after an 8-week exercise training program (8).

Metabolomics data sets of exercise have also revealed the rewiring of intertissue metabolic coordination by exercise training (Fig. 2). Sato et al. (12) performed undirected condition-specific correlation networks on >1,000 metabolites detected across eight tissues after a single bout of treadmill running in 3-month-old male mice. They showed that this exercise bout significantly increased the correlation of metabolites across several organs, including in the serum, liver, heart, hypothalamus, skeletal muscle, brown adipose tissue, and epididymal and inguinal white adipose tissue. Sato et al. further suggested that this increased metabolite correlation between organs might reflect exercise-dependent shifts in fuel preference during physical activity. For example, during exercise, an increase in the correlation of amino acids was observed between the heart-serum-liver, suggesting an increased amino acid fuel coordination between these two organs. Similarly, exercise increased the correlation between carbohydrates in brown fat and lipids in inguinal white fat, suggesting that coordinated carbohydrate/lipid metabolism serves as an axis for interadipose cross talk in exercise. Another surprising finding from the metabolite correlation network is that the hypothalamus becomes the correlation hub between organs after acute exercise. Some of the underlying correlation signals revealed metabolites involved in amino acid metabolism, with a strong enrichment of a class of poorly studied metabolites, acetylated amino acids, after the exercise perturbation (30% of the total correlated metabolites in amino acid metabolism) compared with the sedentary condition (8%). For example, an increase in correlation of N-acetyl-leucine was observed between hypothalamus-serum-skeletal muscle, suggesting a potential exercise-inducible metabolite shuttle between peripheral tissues and the brain.

Lastly, organism-wide multiomics has also revealed an intimate and much more widespread connection between exercise and circadian rhythms than previously recognized. Recent studies have shown that exercise affects the transcription levels of circadian clock genes across the body in an age-, training regime–, and sex-independent manner but an exercise training frequency–dependent manner. Using scRNA-seq, Sun et al. (13) showed that a 12-month voluntary wheel running intervention in 16-month-old male mice restored crucial circadian rhythm genes, such as transcription factor genes Bmal1 and Dbp, to levels comparable with those seen in 2-month-old young mice (Fig. 2). This restoration occurred in >30 of the 101 cell types identified in 14 major tissues. These cell types include cells that reside in the central nervous system, such as neurons, astrocytes, and oligodendrocytes, as well as in the peripheral tissues, such as macrophages, smooth muscle cells, and endothelial cells. Furthermore, these exercise-inducible changes in circadian clock genes were also observed in young mice. In another study involving 12-week-old male mice exposed to 3 weeks of voluntary wheel running, an enriched circadian rhythm pathway was identified in mesenchymal stem cells (MSCs) across skeletal muscle and subcutaneous and visceral white adipose tissue. After exercise, the gene Dbp was once again increased, regardless of the diet composition, in both chow- and high-fat diet–fed mice (14). Using ATAC-Seq (assay for transposase-accessible chromatin with sequencing) to map differentially accessible regions of the whole genome, the MoTrPAC consortium showed that open chromatin regions were enriched with sequence motifs recognized by key transcription factors in circadian clock programs, namely BHLHE41, NPAS, and BMAL1. This circadian clock gene change was sex independent, as this regulation pattern was found in 6-month-old rats of both sexes subjected to 8-week treadmill training (15). All of these observations show that exercise resets central and peripheral clocks in multiple tissues. Given that global and tissue-specific disruptions of circadian rhythm programs dramatically impair exercise adaptations and benefits (16), these studies add to the growing body of evidence suggesting complex and direct molecular interactions between physical activity and circadian clocks across the organism.

A long-standing hypothesis is that exercise induces the production of circulating signaling molecules, also known as “exerkines” or “exercise factors,” that mediate tissue cross talk and function as molecular effectors of physical activity (17). The first evidence for such hormonal factors dates back to 1961, when blood transfusions from exercised dogs were found to lower blood glucose levels in resting dogs (18). For many years, research focused primarily on secreted polypeptides from skeletal muscle, which have also been referred to as “myokines.” However, recent multiomics studies have expanded both the cell types of origins and the types of molecules that are induced by physical activity.

A growing number of studies are indicating that exercise-regulated bioactive plasma proteins originate from multiple sources beyond muscle (19–26). For example, we recently used a cell type–specific secretome tagging strategy to identify new exercise-inducible secreted proteins (Fig. 2). Our findings revealed that two intracellular carboxylesterases expressed in the liver generated extracellular proteoforms capable of antiobesity, antidiabetes, and endurance-enhancing effects in mice (26). Additionally, in two distinct studies with use of untargeted proteomic analysis of blood plasma, the induction of a complement protein and a phospholipase in mice subjected to voluntary wheel running was reported. These proteins are primarily expressed in the liver and were observed to reduce brain inflammation and enhance cognitive function, respectively (20,21). The findings of these studies collectively suggest that the liver is an essential contributor to the exerkine pools during exercise. Besides the liver, other tissues such as bone can also act as sources of exercise-induced bioactive plasma proteins. For instance, after a single bout of treadmill running in mice and humans, osteocalcin, a bone-derived circulating hormone, was induced in circulation. Osteocalcin was found to enhance glucose uptake and catabolism and increase fatty acid oxidation in skeletal muscle. Notably, in mice, whole-body loss of osteocalcin or skeletal muscle–specific depletion of Gprc6a, the osteocalcin receptor, led to comparable outcomes of impaired endurance capacity, as demonstrated by a 20–30% reduction in running time and distance on a treadmill (27). More examples of such non-muscle-derived bioactive proteins can be found in a comprehensive review by Chow et al. (17).

Beyond circulating polypeptides, physical activity also dramatically regulates the levels of circulating, bioactive metabolites. Targeted and untargeted metabolomics analyses have now identified hundreds of plasma metabolite changes following physical activity (12,17,22,28,29). For example, malate and succinate, two metabolite intermediates in classical glycolysis and TCA cycle pathways, are dramatically elevated in blood plasma following exercise (22). These metabolites can exert signaling functions, such as regulating muscle remodeling, cell cycle progression, brain function, and lipid oxidation (30–33). Additionally, untargeted metabolomics has been a powerful approach for uncovering novel bioactive metabolites associated with physical activity. For example, in our recent study we identified a lactate–amino acid conjugate, N-lactoyl-phenylalanine (Lac-Phe), to be robustly induced by exercise across multiple species (28). Pharmacological gain- and genetic loss-of-function studies demonstrate that Lac-Phe regulates feeding and obesity. Exercise-inducible Lac-Phe production is likely driven by the dramatically increased circulating lactate during exercise and establishes an unexpected role for lactate-derived metabolite in physical activity (Fig. 2). Another example is the brown adipose–derived linoleic acid metabolite 12,13-diHOME, which promotes skeletal muscle fatty acid uptake (34). The increase in circulating 12,13-diHOME immediately after a single bout of exercise suggests that exercise-induced lipolysis may be responsible for the elevated levels of this metabolite (Fig. 2). These studies collectively suggest that key metabolite effectors of exercise may be commonly derived from the principal metabolic fuels (e.g., lactate or fatty acids) that are mobilized during physical activity.

Traditionally, studies of the exercise response in key metabolic tissues, such as adipose tissue, muscle, and liver, have largely focused on molecular and physiologic changes at the whole organ level. Recent advancements in multiomics techniques, particularly in scRNA-seq, have enabled a high-resolution reexamination of individual cell types’ response in these major exercise-responsive tissues. This has led to an increasing appreciation of nonparenchymal cell involvement in the exercise response in adipose tissue and skeletal muscle.

For example, it is well established that exercise training induces muscle hypertrophy and increased capillary density. However, there has not been a detailed understanding of the cell types contributing to such remodeling. In analyzing scRNA-seq data using pseudotime trajectory analysis in humans subjected to a single bout of cycling exercise, Lovrić et al. (35) demonstrated that the transcriptional features of Pax7+ satellite cells shifted toward Tnni2+ fast-twitch and Tnni1+ slow-twitch myogenic cells. The authors concluded that this satellite-to-myocyte transition explains the cellular sources of increased density of muscle fiber after exercise. This finding was further supported in another single-nuclei sequencing study focusing on chronic exercise training. Here, Wen et al. (36) showed that the number of Pax7+/Myf5+ and Pax7+/Myf5− significantly increased up to 50-fold after 4 weeks of weighted wheel training in mice. They conclude that exercise training increases Pax7+ satellite self-renewal and maturation. Both of these studies underscore the power of single-cell sequencing to define the cellular basis and differentiation trajectories that underlie skeletal muscle adaptations to exercise.

scRNA-seq has also revealed increased immune cell recruitment and immune pathway activation in muscle and adipose tissues following exercise. For example, in human skeletal muscle, it was recently demonstrated that both the lymphocyte cell population (including T, B, and natural killer [NK] cells) and the monocyte cell population exhibited a substantial increase, from 4 to 9% and 2% to 4%, respectively, following repeated bouts of all-out cycling sprints (35). In a separate study of trained and untrained human subjects following acute bouts of bilateral knee extension, a Pathway-Level Information ExtractoR (PLIER) deconvolution analysis of microarray data showed an increased proportion of neutrophils, identified by marker genes such as S100A8 and S100A9, in the skeletal muscle of both (37,38). Lastly, single-cell transcriptomic analysis of white and brown adipose tissues in male rats undergoing an 8-week treadmill training regimen revealed a significant increase in the number of T, B, and NK cells, as indicated by increased transcript abundance of canonical immune cell markers and enriched immune pathways such as the B and T cell receptor signaling pathway, which was consistent with the findings in skeletal muscle (8). The cellular origins of the increased immune cell recruitment and activation in exercise, as well as the functional consequences of perturbing this process for exercise adaptations, remain unclear at this time.

MSCs in adipose tissue and muscle are another nonparenchymal cell type that displays an unexpectedly strong response to physical activity. These stromal cells have the unique ability to self-renew and differentiate into various cell types, depending on their location within the tissue. Due to their low tissue abundance, e.g., <1% in adipose tissue (39), MSCs have not been previously studied in the context of exercise. A recent study with use of single-cell transcriptomic analysis of skeletal muscle and subcutaneous and visceral adipose tissues highlighted a central role of MSCs in tissue adaptation to 3 weeks of voluntary running in mice (14). Among 22 cell types identified within muscle and fat, MSCs had the largest number of exercise-induced differentially expressed genes. Pathway enrichment analysis of these genes revealed the downregulation of extracellular matrix remodeling, highlighting a previously unknown cellular contributor to local tissue remodeling. Similarly, in our own work using a cell type–selective secretome profiling approach (40,41), we showed that Pdgfra+ MSCs exhibited the most robust secretome response across all of the cell types studied in response to 1-week treadmill training in mice (26). These cells could be found in adipose tissue and muscle, as well as additional nonmetabolic tissues such as the lung. Following exercise, Pdgfra+ MSCs released various molecules into the bloodstream, including TIMP3, which regulates extracellular remodeling; F13A, which is involved in coagulation; and LOXL1, which contributes to fibrosis. These findings suggest that MSCs play a systemic regulatory role in the modulation of whole-body physiology after exercise.

The use of multiomics approaches for analyzing molecular changes during physical activity has revealed a more complex exercise landscape than previously understood. This comprehensive and unbiased strategy has unveiled surprising results of exercise-induced effects, including the dramatic exercise response of nonparenchymal cells, the integrative response across tissues/organ systems, and the intricate regulation of multiorigin exerkines. Despite the wealth of data that has been collected, as well as the new insights obtained, many open questions remain. Below, we share our perspectives on four fertile and exciting future research directions that are now addressable using these high-dimensional tools.

Blood lactate levels are one of the most common blood measurements during exercise testing, especially in athletes. By both abundance and magnitude, lactate is one of the most dramatically regulated molecules in exercise: its levels can be elevated >10-fold to >20 mmol/L after vigorous sprinting. Beyond its primary metabolic role as an interorgan metabolic fuel during exercise (e.g., the “lactate shuttle”), more recent studies have identified even more unexpected functions for lactate in the context of the response to physical activity. For instance, we recently showed that the role of lactate also extends to lactate-derived metabolites such as Lac-Phe, an exercise-inducible signaling metabolite that suppresses food intake and obesity (28). In addition, extracellular lactate can directly induce the secretion of intracellular carboxylesterase proteins from hepatocytes, which function as bioactive circulating plasma proteins that regulate energy metabolism (26). That some of these roles have only been identified, or further expanded upon, in the past few years suggests that we are only in the early stages of understanding the diversity of lactate functions in exercise physiology. One possibility for speculation is that lactate elevation functions as a key driver for numerous secondary downstream molecular events in exercise. Such a hypothesis has not been rigorously tested but may provide a unifying framework for understanding and organizing the thousands of molecular changes that occur in response to physical activity. We propose future studies should prioritize exploring other potential roles of lactate in exercise biology, as well as a consideration of how lactate might function as a coordinating signal between the molecular responses to physical activity.

Despite specific exercise responses being conserved across many cell types—such as upregulated heat shock response and changes in circadian clock programs—many other cellular responses are tissue and cell type specific and can be opposite depending on the tissue/cell type. For example, exercise training leads to the recruitment and activation of immune cells in both white adipose tissue and skeletal muscle, while decreasing the number of immune cells in the small intestine of rats (8). This contrast in regulation patterns between these different tissues indicates that the immune effects of exercise are specific to each tissue. Additionally, our recent studies profiling secreted proteins released by different cell types revealed bidirectional regulation of the same proteins in different cell types by exercise. For example, secretion of a superoxide dismutase was suppressed from pancreatic β-cells but induced from myocytes and Pdgfra+ MSCs in mice following 1 week of treadmill training (26). Taken together, these studies show that many cellular responses to exercise are dependent on the specific cell types under study and include many cell types that have not classically been studied as part of the exercise response. Therefore, higher-resolution approaches that can be applied in a cell type–specific or organ-specific manner become increasingly important for understanding these bidirectional responses to physical activity, which may become masked in studying more “bulk” samples. Importantly, future studies in this area should also be focused on capturing unusual cell types for an understanding of the diversity of cellular responses to exercise beyond those of classical metabolic tissues.

The majority of studies on candidate, exercise-inducible circulating molecular effectors have relied heavily on gain-of-function experiments. There are limited data available from loss-of-function models. Consequently, whether candidate molecules are truly necessary for a particular physiologic effect of physical activity has remained a largely unanswered question. The lack of loss-of-function studies is due, in part, to technical challenges. For instance, interpretation of genetic manipulation of secreted proteins can be complicated due to concurrent ablation of intracellular pools, which may have additional or secondary functions. Neutralizing antibodies are typically not readily available for most circulating polypeptides. For exercise-inducible metabolites, robust methods for their genetic manipulation can be difficult due to multiple biosynthetic pathways and/or undefined molecular pathways for their efflux from cells. The development of new tools in this area, such as neutralizing nanobodies or engineered enzymes for manipulation of specific metabolite pathways, may enable the key functional experiments that demonstrate physiologic roles for exercise-inducible circulating effectors in metabolism and physiology.

A crucial and exciting future direction of exercise research involves developing “exercise mimetics” that can pharmacologically imitate a wide range of health benefits associated with physical activity. Most past efforts toward such molecules have primarily targeted pathways that improve exercise performance. For example, erythropoietin increases hematocrit and VO_2max for endurance performance; similarly, testosterone and other androgen receptor agonists increase lean mass and strength. However, performance is not health. Salutary effects such as reduced cardiovascular mortality, increased bone strength, and improved brain health have yet to be conferred by therapeutic capture of exercise-regulated molecules or pathways. As multiomics techniques advance and we acquire more knowledge about the nonmuscle/nonperformance benefits, we may discover new molecules and opportunities to truly translate the concept of “exercise as medicine” into reality.

Exercise induces profound molecular changes on an organism-wide scale. In previous studies investigators have focused on characterizing individual exercise-regulated molecules and specific organs such as muscle and fat. Recent technical advances in multiomics technologies have allowed for deep mapping of molecular changes across the entire organism and at high resolution. These data sets have revealed thousands of molecular changes that are dependent on the exercise subject (such as species, age, and sex), exercise modality (acute vs. chronic, voluntary vs. involuntary, intensity and frequency), and a variety of environmental factors (such timing of exercise). They have also provided important insights into exercise physiology that were previously not well understood, such as the role of exercise to coordinate molecular processes across multiple organ systems, or the dynamic responses of nonparenchymal cells to physical activity. We anticipate that in providing high-resolution insights into physical activity, these multiomics technologies will provide new insights that serve as a foundation for therapeutic “capture” of the benefits of physical activity for human health.

Acknowledgments. The authors thank members of the J.Z.L. and K.J. Svensson (Stanford University) laboratories for helpful discussions. The authors apologize for not citing the many other excellent articles in this area due to space limitations.

Funding. This work was supported by the National Institutes of Health (DK105203 and DK124265) and the Wu Tsai Human Performance Alliance.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Prior Presentation. Parts of this study were presented in abstract form at the 83rd Scientific Sessions of the American Diabetes Association, San Diego, CA, 23–26 June 2023.