Retinoic acid exerts sexually dimorphic effects on muscle energy metabolism and function

The retinol dehydrogenase Rdh10 catalyzes the rate-limiting reaction that converts retinol into retinoic acid (RA), an autacoid that regulates energy balance and reduces adiposity. Skeletal muscle contributes to preventing adiposity, by consuming nearly half the energy of a typical human. We report sexually dimorphic differences in energy metabolism and muscle function in Rdh10+/− mice. Relative to wild-type (WT) controls, Rdh10+/− males fed a high-fat diet decrease reliance on fatty-acid oxidation and experience glucose intolerance and insulin resistance. Running endurance decreases 40%. Rdh10+/− females fed this diet increase fatty acid oxidation and experience neither glucose intolerance nor insulin resistance. Running endurance increases 220%. We therefore assessed RA function in the mixed-fiber type gastrocnemius muscles (GM), which contribute to running, rather than standing, and are similar to human GM. RA levels in Rdh10+/− male GM decrease 38% relative to WT. Rdh10+/− male GM increase expression of Myog and reduce Eif6 mRNAs, which reduce and enhance running endurance, respectively. Cox5A, complex IV activity, and ATP decrease. Increased centralized nuclei reveal existence of muscle malady and/or repair in GM fibers. Comparatively, RA in Rdh10+/− female GM decreases by less than half the male decrease, from a more modest decrease in Rdh10 and an increase in the estrogen-induced retinol dehydrogenase Dhrs9. Myog mRNA decreases. Cox5A, complex IV activity, and ATP increase. Centralized GM nuclei do not increase. We conclude that Rdh10/RA affects whole body energy use and insulin resistance partially through sexual dimorphic effects on skeletal muscle gene expression, structure, and mitochondria activity.

The retinol dehydrogenase Rdh10 catalyzes the rate-limiting reaction that converts retinol into retinoic acid (RA), an autacoid that regulates energy balance and reduces adiposity. Skeletal muscle contributes to preventing adiposity, by consuming nearly half the energy of a typical human. We report sexually dimorphic differences in energy metabolism and muscle function in Rdh10+/− mice. Relative to wild-type (WT) controls, Rdh10+/− males fed a high-fat diet decrease reliance on fatty-acid oxidation and experience glucose intolerance and insulin resistance. Running endurance decreases 40%. Rdh10+/− females fed this diet increase fatty acid oxidation and experience neither glucose intolerance nor insulin resistance. Running endurance increases 220%. We therefore assessed RA function in the mixed-fiber type gastrocnemius muscles (GM), which contribute to running, rather than standing, and are similar to human GM. RA levels in Rdh10+/− male GM decrease 38% relative to WT. Rdh10+/− male GM increase expression of Myog and reduce Eif6 mRNAs, which reduce and enhance running endurance, respectively. Cox5A, complex IV activity, and ATP decrease. Increased centralized nuclei reveal existence of muscle malady and/or repair in GM fibers. Comparatively, RA in Rdh10+/− female GM decreases by less than half the male decrease, from a more modest decrease in Rdh10 and an increase in the estrogen-induced retinol dehydrogenase Dhrs9. Myog mRNA decreases. Cox5A, complex IV activity, and ATP increase. Centralized GM nuclei do not increase. We conclude that Rdh10/RA affects whole body energy use and insulin resistance partially through sexual dimorphic effects on skeletal muscle gene expression, structure, and mitochondria activity.
All-trans-retinoic acid (RA) regulates embryonic development, cell proliferation, differentiation, and numerous functions of differentiated cells through complex genomic and nongenomic mechanisms (1)(2)(3)(4). RA biogenesis stems from either retinol (vitamin A) or β-carotene (5)(6)(7)(8). As many as eight retinol dehydrogenases and reductases, of the shortchain dehydrogenase/reductase gene family, catalyze conversion of retinol into retinal, and retinal into retinol (9). At least three retinal dehydrogenases, of the aldehyde dehydrogenase gene family, convert retinal generated from retinol or βcarotene, irreversibly into RA. Cells coexpress multiple retinoid metabolic enzymes, but gene ablations reveal dissimilar phenotypes for each, consistent with distinct contributions to RA function (10,11). For example, knocking out the retinol dehydrogenase Rdh1 increases adiposity in mice fed a low-fat diet, from decreased RA in brown adipocytes, which impairs lipolysis and fatty acid oxidation (12). In contrast, the retinol dehydrogenase Dhrs9 seems crucial to the hair follicle cycle, and its decreased expression correlates with squamous cell carcinoma (13,14). Ablation of the retinol dehydrogenase Rdh10 causes lethality between embryo days 10.5 to 14.5 from impaired nervous system and craniofacial development (15)(16)(17). During postnatal growth, Rdh10 generates RA to support spermatogenesis and organ repair and to regulate energy use (18)(19)(20). Ablation of the retinal reductase Dhrs3 increases embryonic RA amounts by 40% and causes death late in gestation from defects in heart development (21,22). Consistent with partially overlapping contributions to RA homeostasis, reducing expression of one reductase/dehydrogenase can provoke compensation by others. These observations prompt questions about the needs for and functions of so many enzymes dedicated to RA homeostasis.
Establishing precise physiological functions of endogenous RA cannot be attained by feeding vitamin A-deficient diets or pharmacological RA dosing, which prompts RA toxicity. The heterozygote Rdh10-null mouse (Rdh10+/−) experiences a limited decrease in tissue RA (≤25%), allowing reproducible assessment of endogenous RA functions in vivo (20). When fed a high-fat (western) diet, Rdh10+/− mice gain adipose and males suffer liver steatosis, whereas females form adipocytes in bone. To develop further insight into the function of Rdh10 and physiological RA actions, we expanded evaluation of the Rdh10+/− mouse to include whole-body energy metabolism and skeletal muscle function. Skeletal muscle accounts for about half of all energy used, producing not only work but also heat to defend body temperature (23). Exercise, through muscles' bountiful demand for energy, ameliorates obesity and insulin resistance (24,25). Energy metabolism, obesity, metabolic syndrome, and muscle function interweave to affect overall health.
RA induces skeletal muscle development through controlling expression of myogenic regulatory factors, including Myf5, MyoD, and Myog during embryogenesis and in primary myoblasts and cell lines (26)(27)(28)(29). RA induces proliferation of cultured myoblasts and sustains their survival, suggesting a contribution to skeletal muscle regeneration (30,31). Pharmacological dosing with RARγ agonists stimulates skeletal muscle repair and reduces extent of fatty fibrotic tissue lesions in a muscle injury model (32). Although it exerts crucial actions during muscle development and repair, little has been revealed about RA function on muscle performance in vivo.
Here we report sexually dimorphic differences in Rdh10+/− mice fed a high-fat diet (HFD) in the respiratory exchange ratio (RER), running endurance, glucose tolerance, and insulin resistance. Decreased Rdh10 in male gastrocnemius muscle (GM), a mixed-fiber type muscle, associates with TAG accumulation, an increase in centralized nuclei, reduced expression and activity of complex IV components of the electron transport chain, and decreased ATP. Rdh10+/− females have increased complex IV components and activity and a higher concentration of resting ATP than WT. Estrogen increased in Rdh10+/− females, as did expression of the estrogen-induced retinol dehydrogenase Dhrs9 (33). RA decreased in Rdh10+/ − female GM less than half the male decrease, likely due to compensation by Dhrs9. These results provide new insight into the physiological functions of Rdh10, sex differences in RA biosynthesis, and actions related to energy use, insulin resistance, and muscle function.

Rdh10+/− differ by sex from WT in fuel use
Rdh10+/− males fed a purified HFD from weaning endure glucose intolerance and insulin resistance, but females were not assessed (20). Here we found that Rdh10+/− females of the same age and fed a purified HFD diet with the same formulation since weaning did not experience glucose intolerance or insulin resistance (Figs. S1 and S2). Further, the RER (CO 2 exhaled/O 2 consumed) showed that WT males relied more on fatty acid oxidation than WT females during normal activity (Fig. 1A). Rdh10+/− males decreased fatty acid oxidation and increased carbohydrate use during the 12-h dark (feeding) cycle, relative to WT males, but did not differ from WT males during the light cycle. Rdh10+/− females decreased carbohydrate use and increased fat oxidation throughout 24 h, relative to WT females. RER data also are shown as averages (Fig. S3).

Rdh10+/− males and females differ in running endurance
Differences in reliance on carbohydrates versus fatty acid oxidation implied possible endurance differences. In run-toexhaustion tests of endurance, WT males ran an average 62 min, whereas Rdh10+/− males ran an average 37 min (40% decrease), resulting in an equivalent 40% decrease in work accomplished (Fig. 1B). WT females ran an average 33 min, i.e., 47% less than WT males. In contrast, Rdh10+/− females ran an average 72 min, 2.2-fold longer than WT females with the same % increase in work performed. Thus, loss of one Rdh10 copy reversed the relative running endurance of males versus females.

Rdh10+/− males and females differ in GM RA concentrations
To examine mechanisms, we focused on GM, because they perform predominantly in running, rather than standing, and have the most extensive overlap of pathways in mouse versus human (34). Rdh10 mRNA in GM in Rdh10+/− mice declined by 53% and 33% in Het males and females, respectively (Fig. 1C). RA decreased 38% and 17% in GM of Rdh10+/− males and females, respectively. Estrogen induces the retinol dehydrogenase Dhrs9 (33). Dhrs9 mRNA in male Rdh10+/− mice did not differ significantly from WT, whereas Dhrs9 mRNA in female Rdh10+/− mice exceeded WT by 1.7-fold. Other mRNAs of the retinoid metabolon, including the retinal dehydrogenases Raldh1, 2, 3, and the RA catabolic enzyme Cyp26b1, were unaffected in both sexes, as was the putative retinol dehydrogenase Dhr7c (Fig. S4). The combined sexspecific differences in Rdh10 and Dhrs9 mRNAs would contribute to the difference between RA concentrations in Rdh10+/− males versus females.
Because estrogen induces Dhrs9, we attempted to quantify estrogen concentrations in female sera. We were unable to detect estrogen in sera of female mice using an LC/MS/MS assay with a 2 fmol lower limit of detection (35). Estrogen concentrations in mice sera occur below limits of detection, unless females are exposed to males, because group-housed female mice enter anestrus. Instead, we quantified uterine weight, which reflects estrogen levels (36,37). Rdh10+/− uteri weighed 39% more than WT relative to total body weight, reflecting an estrogen increase (Fig. 1D). Because of these differential sex differences in response to Rdh10 knockdown, data from each sex were compared with their own littermate WT mice. Rdh10+/− males and females differ in myogenic transcription factor expression mRNA levels of the RA-regulated myogenic transcription factor Myog (myogenin) increased 2.8-fold in Rdh10+/− male GM, relative to WT ( Fig. 2A). In contrast, Rdh10+/− female Myog mRNA decreased by 42%, relative to WT. mRNA of another RA-regulated myogenic transcription factor, Myf5, also increased in Rdh10+/− males (1.8-fold), whereas it decreased 30% in females. The mRNA of a third myogenic transcription factor Myod did not differ between WT and Rdh10+/− GM of either sex (Fig. S5). To verify that mRNA levels represented protein levels, we performed western blots on Myf5 (Fig. 2, B and C). The data reflected adaptations in the mRNA. Rdh10+/− male Myf5 increased threefold, whereas Rdh10+/− female Myf5 decreased 38%. Other factors that regulate skeletal muscle mass and function include myostatin (Mstn), atrogin-1, and MuRF-1 (38)(39)(40). Mice deficient in Mstn have marked increases in skeletal muscle mass, because Mstn suppresses myogenesis by interfering with Myod.

Rdh10 affects GM fiber types in males
Differential myosin heavy chain (Myh) expression designates skeletal muscle fibers as type I (Myh7), type IIa (Myh2), or type IIb (Myh4) (42). Type I (slow-twitch) fibers have abundant mitochondria and promote endurance. These fibers rely mainly on fatty acid oxidation. Type IIa fibers rely on both oxidative and glycolytic pathways. Type IIb fibers (fast-twitch or "sprint") have few mitochondria and rely mainly on glycolysis. GM consist of mixed fiber types, with 54% type IIb fibers, and the rest distributed among types I and IIa/d. The mRNA of Myh7, an indicator of type I fibers, increased approximately threefold in Rdh10+/− male GM (Fig. 3, A and B). No changes occurred in Rdh10+/− Myh2 (type IIa fibers) or Myh4 (IIb fibers) mRNA in male GM. Rdh10+/− female GM Myh7 mRNA remained the same as WT, as did Myh2 and Myh4 (Fig. 3, C and D). Quantification of fiber numbers verified a 1.7-fold increase in Rdh10+/− male type I fibers and no change in female type I fibers (Fig. 3E). Thus, Rdh10+/− males had decreased endurance despite compensation by increased slow-twitch oxidative fibers. females differ from each other and WT in GM RA, fuel use, running endurance, and myogenic transcription factor expression. A, respiratory exchange ratio (CO 2 /O 2 ) during ambient temperature and ad lib feeding a high-fat diet (HFD). Data are means ± SEM (n = 6-10 mice/genotype/sex). Data were analyzed by ordinary two-way ANOVA from 75 equally spaced observations over 24 h for each sex/genotype to assess genotypic differences. B, running time and work done during the run-to-exhaustion test (n = 7-13). The last panel shows work done relative to female WT average of 1. Red bars denote averages. C, Rdh10 mRNA (n = 7-14), RA content (n = 12), Dhrs9 mRNA (n = 7-13) in GM. D, weight of uteri relative to total body weight (BW) (n = 5-9). Symbols in all seven figures: WT male, light blue circles; Rdh10+/− (Het) male, dark blue squares; WT female, light pink circles; Rdh10+/− (Het) female, dark purple squares.
This ineffective compensation by increasing slow-twitch fibers would not correct the phenotype, because the defective mitochondria in these fibers would not improve fuel use and energy production.

Rdh10+/− males and females differ from each other and WT in muscle fiber characteristics
Normally, nuclei in muscle localize peripherally to fibers. Nuclei migration to the centers of fibers accompanies muscle dysfunction or regeneration (43). Figure 4A illustrates the difference between peripheral and centralized nuclei. Centralized nuclei increased 2.4-fold in Rdh10+/− male GM, but decreased 26% in Rdh10+/− female GM (Fig. 4B). Crosssectional areas (CSA) of Rdh10+/− GM fibers adapted to Rdh10 knockdown according to sex (Fig. 4C). Numbers of larger male fibers increased 10.7%, whereas numbers of larger female fibers decreased 12.6%. The range of Het male CSA shifted to larger ones relative to WT; the range of Het female CSA shifted to smaller ones (Fig. 4D). Direct comparison of male and female WT CSA revealed extensive similarity, indicating CSA normally do not diverge substantially according to sex (Fig. 4E). All four curves (WT and Rdh10+/−, male and female) had similar if not identical areas, consistent with shifts in fiber size distributions in Rdh10+/−, rather than muscle growth or reduction. Although Rdh10+/− female GM did not change in fiber type, nor show an increase in centralized nuclei, homogenization for the same length of time disrupted both Rdh10+/− male and female GM much faster than WT (Fig. S8). This indicates that the connective tissue surrounding the muscle fibers is weaker in Rdh10+/− mice, suggesting impaired tissue integrity.

Rdh10+/− males experience disturbances in GM lipid metabolism
Muscle weights relative to whole body weights also exhibited sexually dimorphic differences. Rdh10+/− male GM weighed 10% more than WT, whereas Rdh10+/− female GM weights did not differ significantly from WT (Fig. 5, A and B). The weight increase of Rdh10+/− males was not caused by an increase in protein content, as no marked differences occurred among GM protein amounts in either sex (Fig. S9). Oil red O staining indicated greater TAG in Rdh10+/− males than other groups. This was confirmed by quantification of TAG by biochemical assay, which revealed a 1.7-fold increase in Rdh10+/− male GM, but no increase in Rdh10+/− female GM relative to WT (Fig. 5C). The decrease in Rdh10 triggered changes in mRNA of genes related to fatty acid biosynthesis, lipolysis, and mitochondria long-chain fatty acid import. Fatty acid synthase (Fasn), hormone sensitive lipase (Hsl), and carnitine palmitoyltransferase 1b (Cpt1b) mRNA decreased 30 to 40% in Rdh10+/− male GM (Fig. 5D). Rdh10+/− female GM had a 50% increase in Fasn mRNA and a 36% decrease in Cpt1b mRNA, with no change in Hsl mRNA (Fig. 5D). FASN protein reflected changes in mRNA, showing a 55% decrease in Rdh10+/− males and a 1.9-fold increase in Rdh10+/− females (Fig. 5, E-G). Rdh10+/− male and female GM mRNA did not change for Atgl, CD36, Hk2, Mcad, Pgcla, and Ppara (Fig. S10). Insulin induces the elongation and initiation factor Eif6, which stimulates lipogenesis in the liver and adipose, including transcription of Fasn (44). Eif6 +/− mice also endure reduced exercise endurance (34). Eif6 mRNA decreased 26% in Rdh10+/− male GM, with no decrease in Rdh10+/− females, consistent with Het male insulin resistance, the decrease in FASN, and the decline in running endurance (Fig. 5H).

Glycogen and glucose adaptations differ sexually in
The fasting glycogen content of Rdh10+/− GM did not differ from WT in either sex (Fig. 6, A and B). Gys mRNA, which encodes the rate-limiting enzyme in glycogen synthesis, decreased 30% in Rdh10+/− males, but did not change in females (Fig. 6C). Pygm, which encodes the rate-limiting enzyme for muscle glycogenolysis, did not change in Rdh10+/− mice of either sex. The lack of a decrease in muscle glycogen despite the decrease in Gys likely reflects 16 h fasting, which would deplete muscle glycogen.
Skeletal muscle relies for glucose uptake on glucose transporters Glut1 and Glut4. Glut1 mRNA decreased 36% in Rdh10+/− male GM, but did not change in Rdh10+/− female GM (Fig. 6D). Glut4 mRNA did not change in Rdh10+/− mice of either sex. The intensity of Glut1 immunostaining decreased a modest 5%, which could have a profound influence on glucose uptake, because the mice were fasted 16 h (Fig. 6E). The mRNA of two key genes that encode enzymes essential to gluconeogenesis, Pck1 (phosphoenolpyruvate carboxykinase 1) and G6pc3 (glucose-6-phosphatase catalytic subunit 3) decreased by 73 and 27%, respectively, in male GM, but did not change in female GM (Fig. 6F). The average amount of glucose in fasted male GM decreased by 34%, reflecting the decrease in Glut1 and gluconeogenesis genes and the glucose intolerance and insulin resistance of male Rdh10+/− mice (Fig. 6G).

Rdh10+/− males suffer reduced GM mitochondria function; females gain function
We assessed mRNA of genes associated with mitochondria function to determine the extent of electron transport chain activity. mRNAs of complex IV components (cytochrome C oxidase) Cox5a and Cox8b declined by 46% and 39%, respectively, in Rdh10+/− male GM (Fig. 7A). Cox5a and Cox8b mRNA increased from 20 to 36% in Rdh10+/− female GM, but data were shy of statistical significance at a p 0.07. The mRNA amounts of other electron transfer chain components in Rdh10+/− GM did not differ from WT, indicating no decrease in total mitochondria (Fig. S11). These included Ndufs2 (complex I), Sdhb (complex II), Uqcrc2 (complex III), Cycs (cytochrome c), and Apt5al (complex V). Cox5a protein expression in GM lysates decreased in Rdh10+/− male GM by an average 28%, followed by a complex IV (cytochrome C oxidase) activity decrease of 13% and a fasting ATP decrease of 60% (Fig. 7, B and C). In contrast, Cox5a protein expression increased by 40% in Rdh10+/− female GM, accompanied by a 40% increase in complex IV activity, and >3-fold increase in fasting ATP (Fig. 7, B and D).

Discussion
We have reported that eliminating one Rdh10 copy in vivo (Rdh10+/−) reduced RA modestly in adult liver and white adipose (≤25%) and increased diet-induced obesity in both sexes (20). Only males endured liver steatosis when fed a HFD, and only females sustained increased bone marrow adipocyte formation, regardless of dietary fat. Here we report the impact of marginally reduced RA in muscle, showing an outcome likely influenced by increased estrogen in Rdh10+/− females, associated with an attenuated decrease in RA relative to males. These data provide further insight into RA modulation of estrogen effects and, reciprocally, estrogen modulation of RA effects. Assessing the whole body Rdh10 heterozygote permitted assessment of overall Rdh10/RA impact on adiposity, insulin resistance, and carbohydrate versus fat use as fuels.
Male C57/Bl6 mice have larger body weights and more muscle mass compared with age-matched females (20). This difference in lean body mass, along with a greater reliance on fatty acid oxidation (RER), enables longer running times in WT males. The modest but significant decrease in Rdh10+/− male reliance on fatty acid oxidation and increased reliance on carbohydrates (RER), would affect running endurance. Decreases in male GM Cpt1b and Hsl mRNA would limit fattyacid oxidation and prompt the increase in TAG. Decreases in Gys and Glut1 would reduce glucose uptake and storage, contributing to whole-body glucose intolerance, as muscle accounts for >70% of insulin-stimulated glucose use (45). Decreases in Pck1 and G6pc3 would contribute to the decrease in Rdh10+/− male GM glucose. These decreases in muscle energy availability, verified by glucose intolerance and insulin resistance, plus impairment of complex IV amount and activity, would hinder ATP production in male GM and reduce energy for running endurance. Impaired energy use would prompt insulin resistance, which explains the decrease in Fasn, an insulin-regulated gene. The selective increase in type I (oxidative) fibers in Rdh10+/− males would not fully compensate for impaired endurance because the fibers included defective mitochondria.
The increased dependence on fatty acid oxidation in Rdh10+/− females (RER) and the increase in complex IV activity, with more efficient glucose uptake, produced more ATP and greater running endurance. These data show that RA has a complex impact on postnatal skeletal muscle function that alters ATP generation. These sexually dimorphic differences in muscle contribute to the whole-body phenotype of Rdh10+/− and very likely are influenced by sex hormones.
The surprising increase in female muscle performance with a decrease in RA seems to challenge concepts of RA function, because decreased RA usually diminishes cell, organ, and physiological performance (46). Then again, this outcome expands insight into interactions between RA and estrogen and is consistent with reports that estrogen enhances RA signaling, and its biosynthesis, whereas RA retards estrogen action. RA restricts estrogen receptor α action in the estrogen-sensitive human breast cancer cell line MCF7/BUS (47), inhibits estrogen-induced proliferation of MCF-7 cells (48), and promotes ERα degradation and reduces its transcriptional activity in breast cancer cells (49). RA promotes homeostatic maintenance of the mouse uterus, preventing estrogen-promoted proliferation (50)(51)(52). In contrast, estrogen sensitizes human breast cancer cells to RA by inducing RARβ (53). Estrogen also induces expression of stimulated by retinoic acid gene 8 (Stra8), showing that it can compensate for RA loss (54). Estrogen amplifies the effects of retinoids on ameliorating obesity in female mice (55). Estrogen enhances muscle mass and strength in vivo and promotes muscle regeneration (56)(57)(58). Notably, estrogen action in muscle increases insulin sensitivity, prevents lipid accumulation, and promotes metabolic health (59). Our cumulative data and these references support an estrogen contribution to improving the phenotype of Rdh10 +/− females, relative to males. In contrast to estrogen actions on RA signaling, androgens repress RAR mRNA in rat prostate (60), whereas RA suppresses androgen receptor expression in rat testis (61). The decreased performance by Rdh10+/− males could reflect a contribution of androgens to decreasing RA signaling and amplifying repression of RAR.
A multifaceted process controls skeletal muscle development, regeneration, and function (62). Ablating the myogenic transcription Myog after muscle development produces a mouse with enhanced performance in treadmill running, as a result of improved fuel metabolism (63), whereas ablating Eif6 reduces running performance (34). The decreased running performance and restricted energy use of Rdh10+/− males reflect the increase and decrease, respectively, in expression of these genes. The juxtaposed increase in running performance and energy production of Rdh10+/− females reflects the decrease in Myog expression. Myogenic factors, Myf5 and MyoD commit stem cells to a skeletal muscle lineage. RA effects on Myog, Myf5, and MyoD have been studied during embryogenesis and in cell lines (64), but postnatal effects of endogenous RA have not received the same attention. Cumulative data in cell lines and chick limb buds indicate concentration-dependent RA effects on myogenesis (65)(66)(67). These results may reflect retinoid-related hormesis: i.e., RA effects depend on concentrations. Beneficial effects occur as concentrations increase to an optimum, which presumably occurs at the low nM RA that occurs in tissues (68). As concentrations increase beyond an optimum, beneficial effects subside and pharmacological effects ensue; ultimately toxic effects prevail (69).
Hormesis also may contribute to sexual dimorphism. Estrogens versus androgens exert different effects on RA functions apparently by increasing or decreasing RAR expression, respectively, which would impact the RA doseresponse curves based on sex. The sex-hormone/RA nexus would help explain how modest alterations in RA have contrary sex-related effects. Moreover, RA effects on myogenic differentiation depend on the exact cellular context, as location-specific interactions with the local chromatin environment affect transcriptional activity (70), as well as expression of distinctive retinoid-binding proteins (71).

Conclusions
The present data complement insight into the antiobesity effects of RA by revealing that endogenous RA promotes skeletal muscle fuel use, mitochondrial function, and ATP biosynthesis in males, but restricts the same in females. These male and female specific actions also affect glucose tolerance and insulin resistance. Contrasting interactions between estrogens and androgens on RA biosynthesis and function suggest mechanisms for these dimorphic effects. This sexually dimorphic insight into RA in muscle contributes to understanding differences between males, premenopausal females, and postmenopausal females with respect to regulation of metabolic health. Given the complexity of RA actions in vivo and its hormesis-dependent effects, partial ablation of Rdh10 should continue to provide opportunity to distinguish physiological from pharmacological and/or toxic actions of RA.

Mice and diets
Mouse experimental protocols were approved by the University of California-Berkeley Animal Care and Use Committee. Rdh10-floxed mice were bred with mice expressing CMV-Cre (B6.C-Tg (CMV-Cre) 1Cgn/J) to generate a wholebody knockout of Rdh10 as described (20). Homozygous Rdh10 knockout causes embryonic lethality, therefore Rdh10 heterozygotes (Rdh10+/−) were studied. CMV-Cre+ littermates served as controls (WT). Mice were backcrossed into the C57BL/6J background >12 generations. Mice were fed an AIN93G purified diet containing 4 IU/g vitamin A with 50% calories as lard (HFD) since weaning (Research Diet, Inc, D17092103) and were fasted overnight prior to tissue collection to promote fatty acid use, unless noted otherwise. Mice were 4 to 5 months old. Female mice were synchronized in anestrous, because they were group-housed in the absence of males-the Lee-Boot effect (72)(73)(74) Respiratory exchange ratio (RER) RERs were determined with a Columbus Instruments Comprehensive Laboratory Animal Monitoring System (CLAMS). Mice were acclimatized to chamber cages 24 h before values were recorded. The CLAMS was housed in a temperature controlled (23 C), 12-h light/dark cycle room, with lights on at 7 AM. Mice had access ad libitum to food and water. Activities were measured simultaneously and showed no differences based on sex or genotype.   Figure 2, B and C. C, quantification of male GM data in B relative to WT, relative complex IV activity (n = 6-10), and ATP concentration (n = 6 each genotype). D, quantification of female GM data in B relative to WT, relative complex 4 activity (n = 7-8), and ATP concentration (n = 6-8). ab109911). Glucose was quantified with a Promega Kit (Glucose-Glo Assay).

Run-to-exhaustion test
We used a motor-driven treadmill (Columbus Instruments, Exer-6M Open Treadmill) to assess running endurance. To encourage mice to run, tactile stimuli were provided by a manual light tap with a small paint brush to the hindquarters when the animal reached the back of the treadmill. On days 0 and 1, mice were placed on the treadmill to acclimate for 10 min with a speed of 10 m/min at 0% gradient. On day 2, running capacity was determined by placing mice on the treadmill without treadmill motion for 2 min for acclimation. Next, mice were subjected to a 5 min warm-up period at 10 m/ min. Speed was then increased by 2 m/min each 2 min until a maximum of 20 m/min. Mice were then run to exhaustion, defined as sitting at the back wall for 10 s despite continual tapping with the brush. Work (Joules) was calculated from body weight (kg) × 9.8 × distance run in m.

Histology and immunofluorescence
Entire GM were dissected from legs, snap-frozen in liquid nitrogen, and stored at −80 C. Cryosections (10 μm each) were harvested from frozen muscles with a Cryostat (Leica CM3050) and processed for histological and immunofluorescence analysis. Sections were stained with hematoxylin (Sigma GHS3) and eosin Y (RICCA 2845-32), washed with distilled water, and mounted in Shur Mount (General Data LC-W). For CSA quantification, all fibers in each field were quantified from four mice per group with four sections for each mouse, except for one of the female WT and one female Het for which three sections were analyzed. In total 1454 to 1635 fiber sections were quantified for CSA. Oil red O was used to stain lipids (75). Sections were mounted in Shur Mount and glycogen was visualized by Periodic Acid-Schiff staining (PAS) (MilliporeSigma 395B-1KT) following manufacturer's instructions. Slides were imaged with a Zeiss AxioObserverZ1 microscope at 40× magnification, using QImaging 5MPx a MicroPublisher color camera (Biological Imaging Facility, UC-Berkeley).
For immunofluorescence imaging, sections were warmed to room temperature for 40 min and washed three times with 0.1% PBST. After blocking with 5% NGS/PBST (normal goat serum in 1% Tween 20/phosphate-buffer) 1 h at room temperature, sections were incubated in 5% NGS/PBST with a mixture of three primary antibodies against Myh7 (IgG2b), Myh2 (IgG1), and Myh4 (IgM), obtained from DSHB at the University of Iowa. After three 5 min washes in PBST, sections were incubated 1 h at room temperature with a mixture of three goat anti-mouse secondary antibodies against IgG2b (Alexa 350), IgG1 (Alexa 488), and IgM (Alexa 555) in 5% NGS/PBST, followed by three washes in PBST. For analysis of centralized nuclei, sections were first incubated in the primary anti-Desmin monoclonal antibody overnight at 4 C. For nuclear staining, sections were incubated in 300 nM DAPI (4 0 6diamidino-2-phenylindole, Thermo Fisher Scientific, D1306) working solution for 5 min at room temperature, followed by washing in PBS. Sections were mounted with Prolong Diamond Anti-fade mountant (Thermo Fisher P36965) and sealed with colorless nail polish. Washing and blocking for IHC staining were the same as for immunofluorescence. Slides were incubated overnight at 4 C with primary glucose transporter 1 antibodies (Glut1), washed with 0.1% PBST, and incubated with goat-anti-rabbit IgG HRP. Stained slides were examined at 40× magnification with oil immersion using Zeiss Axiovert M1 fluorescent microscope equipped with Hamamatsu Orca CCD camera and CoolLED pE-300 white LED illuminator, and Zeiss Z1 AxioObserver inverted microscope equipped with a 40× immersion lens, X-Cite 120 LED fluorescence source and Qimaging Retiga SRV CCD camera (Biological Imaging Facility, UC Berkeley). To quantify Glut1 intensity, the reciprocal of brightness in the blue channel-measure was used in Image J.

RNA extraction and qPCR
Total RNA was extracted from tissues with TRIzol reagent (Life 15596026). The RNA concentration was measured with a Nanodrop spectrophotometer and reverse-transcribed with iScript (Bio-Rad 1708841), followed by qPCR reactions using Prime Time Gene Expression Master Mix (Integrated DNA Technologies 1055771). qPCR data were normalized to the housekeeping genes Gusb and Tbp. Each Rdh10+/− mRNA was normalized to its WT littermates. We used predesigned primers purchased from Integrated DNA technology. Primer details are provided in Supporting information (Table S1).

Western blots
Immunoblotting was performed as described (76). Muscles were thawed on ice and homogenized with a TissueLyser II (Qiagen Inc) in 1 ml cold RIPA buffer (Thermo Fisher 89900) supplemented with protease inhibitor (Thermo Fisher A32965) and phosphatase inhibitor (Roche 4906845001). Tissue lysates were denatured and centrifuged. Protein was quantified in the supernatant with the BCA protein assay (Thermo Fisher 23225), subjected to SDS-PAGE, and transferred onto PVDF membranes (Bio-Rad 1620177). Membranes were blocked with 5% fat-free milk, incubated with primary antibodies overnight (4 C), and with secondary antibodies 1 h at room temperature. Membranes were scanned and signal intensities were quantified using a LiCOR Odyssey imager. Protein loaded per lane was determined with Coomassie blue staining (Thermo Fisher 46-5034) and analyzed with NIH Image J. Data were normalized to total protein loaded per lane. Units are arbitrary (AU), normalized to WT males set to 1, unless noted otherwise.

RA quantification
RA was quantified by LC/MS/MS as published, except methanol was used instead of ethanol, and homogenates were centrifuged at 1200g for 5 min to remove particulates (77). LC was done as published (78).

Statistics
All data are present in this manuscript. Data were plotted as means ± SD, unless noted otherwise. Two-tailed unpaired ttests were used to analyze results between two groups. Welch's corrections were applied to groups without equivalent SD. Significantly different from WT: *p < 0.05, ** p < 0.01, ***p < 0.001, unless noted otherwise; n = number of mice per genotype/sex.

Data availability
All data are present in this manuscript.
Supporting information-This article contains supporting information.