Genetic variation of human myokine signaling is dominated by biologic sex and sex hormones

Proteins secreted from skeletal muscle, termed myokines, allow muscle to impact systemic physiology and disease. Myokines play critical roles in a variety of processes, including metabolic homeostasis, exercise improvements, inflammation, cancer and cognitive functions1–6. Despite the clear relevance of these factors in mediating a multitude of physiological outcomes, the genetic architecture, regulation and functions of myokines, as well as degree of conservation of these communication circuits remains inadequately understood. Given that biologic sex controls critical aspects of nearly every physiologic outcome, it is essential to consider when relating specific mechanisms to complex genetic and metabolic interactions. Specifically, many metabolic traits impacted by myokines show striking sex differences arising from hormonal7–10, genetic7,11 or gene-by-sex interactions12,13. In this study, we performed a genetic survey of myokine gene regulation and cross-tissue signaling in humans where sex as a biological variable was emphasized. While expression levels of a majority of myokines and cell proportions within skeletal muscle showed little differences between males and females, nearly all significant cross-tissue enrichments operated in a sex-specific or hormone-dependent fashion; in particular, with estrogens. These sex- and hormone-specific effects were consistent across key metabolic tissues: liver, pancreas, hypothalamus, intestine, heart, visceral and subcutaneous adipose tissue. Skeletal muscle estrogen receptor enrichments across metabolic tissues appeared stronger than androgen receptor and, surprisingly, ~3-fold higher in males compared to females. To define the causal roles of estrogen signaling on myokine gene expression and functions, we generated male and female mice which lack estrogen receptor α (Esr1) specifically in skeletal muscle and integrated global RNA-Sequencing with human data. These analyses highlighted mechanisms of sex-dependent myokine signaling conserved between species, such as myostatin enriched for divergent substrate utilization pathways between sexes. Several other sex-dependent mechanisms of myokine signaling were uncovered, such as muscle-derived TNFα exerting stronger inflammatory signaling in females compared to males and GPX3 as a male-specific link between glycolytic fiber abundance and hepatic inflammation. Collectively, we provide the first genetic survey of human myokines and highlight sex and estrogen receptor signaling as critical variables when assaying myokine functions and how changes in cell composition impact other metabolic organs.

Sequencing with human data. These analyses highlighted mechanisms of sex-dependent myokine 66 signaling conserved between species, such as myostatin enriched for divergent substrate utilization 67 pathways between sexes. Several other sex-dependent mechanisms of myokine signaling were 68 uncovered, such as muscle-derived TNFα exerting stronger inflammatory signaling in females 69 compared to males and GPX3 as a male-specific link between glycolytic fiber abundance and

75
Sex hormones, but not biologic sex show stronger enrichment with myokine expression: Our goal 76 was to perform a comprehensive survey how skeletal muscle communicates with and impacts 77 metabolic organs. We focused these analyses on exploiting natural genetic variation to assay 78 muscle-specific regulation of myokines and changes in cellular composition, then relate these 79 outcomes to consequent cross-tissue signaling mechanisms ( Fig 1A). Initially, we quantified 80 differential expression of genes encoding all known secreted proteins in skeletal muscle from 210 81 3 male and 100 female individuals 14 . While several notable myokines appeared different between 82 sexes (Fig 1B), a striking majority of all muscle secreted proteins (74%) showed no difference in 83 expression between males and females ( Fig 1C, Supplemental table 1). To understand potential 84 sex-effects on the regulation of myokines, gene ontology enrichments were performed in muscle. 85 Specifically, the skeletal muscle genes which showed the strongest correlation with myokines 86 corresponding to each category (male-specific, female-specific or non-sex specific) were used for 87 pathway enrichments. Here, the top 10 pathways which persisted in females were also observed 88 strongly enriched within the non-sex-specific category, whereas pathways enriched for male-89 specific myokines were distinct ( Fig 1D). Notably, the female and shared pathways suggested 90 roles in epigenetics and RNA processing, while male-specific myokine coregulated processes were 91 more enriched in metabolic pathways (ex. NADH metabolism) ( Fig 1D). Further, a majority of 92 myokines showed strong correlation with receptors mediating functions of androgens (androgen 93 receptor -AR), estrogens (ESR1), or both, regardless of sex-specific expression (Fig 1E). We note 94 that expression of hormone receptors themselves were also not significantly different between  hormone receptor correlations, the gene was markedly higher in males compared to females, where 103 ablation of ESR1 uniquely drove expression ( Fig 1H). These data suggest interactions between 104 biologic sex and ESR1 to tightly regulate MSTN in males, where other factors could contribute 105 more in females. This sex-specific regulation of myostatin also showed differences in functional 106 annotations, as the most highly enriched pathways in males showed GO terms related to glycolytic 107 metabolism compared to oxidative phosphorylation in females ( Fig 1I). These observations are 108 consistent with previous studies which note myostatin-dependent increases in muscle mass in 109 males, but not females 15,16 , where estrogen signaling is suggested as a mechanism mediating these 110 differences. These data demonstrate that, expression of most myokines is not different between 111 biologic sexes; however, interactions between sex and hormone receptors likely play important 112 roles in determining myokine regulation and local signaling.

114
Sex dominates cross-tissue pathways enriched for myokines -Given that expression levels of most 115 myokines appeared similar between sexes, we next assessed putative functions across organs. We 116 applied a statistical method developed to infer cross-tissue signaling which occur as a result of 117 genetic variation [17][18][19] . Here, we assayed the distribution of midweight bicorrelation coefficients 118 between myokine expression levels and global gene expression measures from the same 119 individuals in key metabolic tissues including hypothalamus, heart, intestine, pancreas, liver, 120 subcutaneous and visceral adipose tissue. Remarkably, nearly all highly significant correlations 121 between myokines and target organ genes showed sex-specific modes of operation (Fig 2A-H).

122
This sex specificity also appeared more pronounced for positive correlations between myokines 123 and target tissue genes, as compared to negative (Fig 2-H). Further, among these high significant 124 cross-tissue circuits, myokine hormone receptor enrichment was strongly dependent on the 125 category (ex. significant only in females) rather than target tissue (Fig 2A-H). This observation 126 further suggests that hormone receptor levels (ESR1 or AR) in muscle are a stronger determinant 127 4 of myokine expression compared to biologic sex; however, sex is suggested to dominate 128 coregulated signaling processes across organs via myokines. Therefore, to gauge the relative 129 impact of muscle steroid hormone receptors across organs, the number of significant correlations 130 between ESR1, AR or both were quantified for each tissue. Remarkably, ESR1 specifically as 131 showing an order of magnitude stronger enrichment across metabolic tissues in compared to AR 132 or both, where the number of significantly correlated cross-tissue male ESR1 genes ( Fig 2I) were 133 three-fold higher than females (Fig 2J). Because both sex and ESR1 signaling appeared critical in 134 the regulation of myokine functions, we binned significant cross-tissue enrichments into categories 135 taking into consideration whether myokines were driven by ESR1 in muscle, and/or showing a 136 sex-specific mode of inter-organ enrichment. This analysis suggested that a majority of myokines 137 were either driven by ESR1 and signaled robustly across sexes (Fig 2K, yellow) or signaled 138 differently between sexes, but regulated independent of ESR1 (Fig 2K, red). These categories   Table 2). Similar to myokine expression, no differences were observed 156 between sexes in terms of cell composition, with the exception of modest higher glycolytic fiber 157 in males, compared to elevated oxidative fiber levels in females ( Fig 3B). Additionally, no an individual muscle cell type differed between sexes ( Fig 3C). Generally, differences in skeletal 161 muscle cell abundance corresponded to changes in liver and visceral adipose tissue pathways in 162 males, compared to pancreas in females ( Fig 3C). In contrast to general myokine enrichments, 163 cell proportions showed stronger correlations with AR when compared to ESR1 across both sexes; 164 however, the most abundant cell types were significantly enriched for both steroid hormone 165 receptors ( Fig 3D). Next, to uncover potential direct mechanisms linking changes in cell 166 composition to peripheral tissues, we analyzed associated myokines and adopted and adjusted 167 regression-based mediation approach. Despite few differences between sexes in terms of myokine 168 expression and cell composition, specific myokines highly correlating with individual cell type 169 were markedly different between males and females with the exception of one, APOD in slow-170 twitch fibers (Fig 3E). To determine if variation in cell compositions corresponding to sex-specific 171 tissue signaling via myokines was predicted to be causal, we implemented adjusted regression 172 mediation analyses 23,24 for glycolytic fiber composition. Because male glycolytic fiber type was 173 5 selectively enriched for liver pathways such as immune cell activation and regulated exocytosis 174 (Fig 3F), the top-genes driving these enrichments were used to determine causality. The top-175 correlated muscle secreted protein with male glycolytic fiber type levels was secreted glutathione 176 peroxidase 3 (GPX3). Here, adjusting regressions between glycolytic fiber and liver pathways on 177 GPX3 significantly reduced the overall significance across tissues (Fig 3G), suggesting GPX3 as 178 a mediator of this communication. These data point to a potential mechanism whereby muscle 179 fiber abundance could buffer free radical generation in the liver, thereby feeding back on 180 inflammation. This analysis appeared additionally sensitive to inferring non-dependent 181 relationships between muscle cell types, top-ranked myokines and cross-tissue processes. For 182 example, female glycolytic fibers were strongly enriched for pancreatic protein synthesis 183 pathways; however, when adjusted for the top-ranked myokine CES4A, no changes in regression 184 significance were observed (Fig 3F-G). These analyses show that male GPX3 is a likely 185 mechanism whereby fast-twitch muscle signals to liver; however, the same cell type in females 186 drive pancreas protein synthesis independent of CES4A. In summary, we show that cell 187 composition is strongly conserved between sexes, but cross-tissue signaling of altered composition 188 differs entirely. We further suggest putative myokines and mechanisms, as well as highlight the 189 key regulatory roles of estrogen in both sexes.

191
Conclusions and limitations -Here we provide a population survey of skeletal muscle myokine 192 regulation and putative functions using genetic variation and multi-tissue gene expression data. 193 We find that in general, expression of myokines do not significantly differ between sexes; (ex. myokine-target gene) as well as coregulated processes across organs. It is important to note 266 that we exclusively rely on these empirical pvalues when surveying broad correlation structures, 267 whereas much more stringent and appropriate thresholds (ex. p<1e-6 for Fig 3G) were applied 268 when inferring direct interactions.

302
Conservation of gene between mice and humans -To find which myokines and pathways were 303 conserved between mice and humans, all orthologous genes were accessed from MGI vertebrate 304 homology datasets, which have been compiled from the Alliance for Genome Resources 26 and 305 intersected at the gene level.

307
We acknowledge the following funding sources for supporting these studies: LMV, CV, CJ and LV, CV, TMM, ZQ and MMS accessed raw data, performed analyses and drafted the manuscript.

316
CJ and AH provided critical insight into data use and interpretation, as well as guided the study.

317
All authors read and approved this manuscript.    Here, comparisons of the three most common methods (DCQ, NNLS and porportionsInAdmixture) were plotted for each pseudosc-proportion, where proportionsInAdmixture method captured the largest relative number of cell types