Male lifespan extension with 17‐α estradiol is linked to a sex‐specific metabolomic response modulated by gonadal hormones in mice

Summary Longevity in mammals is influenced by sex, and lifespan extension in response to anti‐aging interventions is often sex‐specific, although the mechanisms underlying these sexual dimorphisms are largely unknown. Treatment of mice with 17‐α estradiol (17aE2) results in sex‐specific lifespan extension, with an increase in median survival in males of 19% and no survival effect in females. Given the links between lifespan extension and metabolism, we performed untargeted metabolomics analysis of liver, skeletal muscle and plasma from male and female mice treated with 17aE2 for eight months. We find that 17aE2 generates distinct sex‐specific changes in the metabolomic profile of liver and plasma. In males, 17aE2 treatment raised the abundance of several amino acids in the liver, and this was further associated with elevations in metabolites involved in urea cycling, suggesting altered amino acid metabolism. In females, amino acids and urea cycling metabolites were unaffected by 17aE2. 17aE2 also results in male‐specific elevations in a second estrogenic steroid—estriol‐3‐sulfate—suggesting different metabolism of this drug in males and females. To understand the underlying endocrine causes for these sexual dimorphisms, we castrated males and ovariectomized females prior to 17aE2 treatment, and found that virtually all the male‐specific metabolite responses to 17aE2 are inhibited or reduced by male castration. These results suggest novel metabolic pathways linked to male‐specific lifespan extension and show that the male‐specific metabolomic response to 17aE2 depends on the production of testicular hormones in adult life.

severity, and/or frequency of age-associated metabolic dysfunction, frailty, cancers, and some forms of neurological disease, with part of this disparity linked to the underlying effects of sex-specific gonadal hormone production across life (Legato, 2010). Male testicular production of testosterone has been postulated to reduce male lifespan compared to females (Brooks & Garratt, 2017), and castration has been associated with longer lifespan in several mammal species, including humans (Hamilton & Mestler, 1969;Min, Lee & Park, 2012), primates (Kessler et al., 2016), sheep (Jewell, 1997), and rodents (Asdell, Doornenbal, Joshi & Sperling, 1967;Muehlbock, 1959), suggesting a potential common and conserved biological mechanism. At the same time, estrogen production has been linked to female-specific health benefits (Regan & Partridge, 2013), with ovariectomy reported to reduce female rodent lifespan in some instances (Benedusi et al., 2015;Mason, Cargill, Anderson & Carey, 2009).
17-b estradiol is the dominant female sex hormone produced from the ovaries in adult life. Female-specific production of 17-b estradiol contributes to sex differences in metabolism, supported by the observation that circulating 17-b estradiol production declines during menopause, a fall associated with lower glucose homeostasis and elevated visceral adiposity (Mauvais-Jarvis, 2015).
However, other estrogens, and estrogenic actions outside of the classical ER receptors, are increasingly recognized to have potential health and anti-aging benefits, which may mimic or operate outside of the effects of 17-b estradiol on its best-characterized receptors.
This study focuses on a stereoisomer of 17-b estradiol, 17-a estradiol (17aE2). Due to the stereochemistry of the carbon atom 17, 17aE2 has a much weaker binding affinity to the classical estrogen receptors, and in some situations has a greater binding affinity for other estrogen receptors, including the brain ER receptor ER-X (Toran-Allerand, Tinnikov, Singh & Nethrapalli, 2005;Toran-Allerand et al., 2002). 17aE2 has a range of bioactive properties, including inflammatory and antioxidant effects (Moos, Dykens, Nohynek, Rubinchik & Howell, 2009), and an ability to inhibit the activity of 5-alpha reductase enzymes (Schriefers, Wright, Rozman & Hevert, 1991), which convert testosterone to dihydrotestosterone, a more potent activator of the androgen receptor. Treatment of male mice with 17aE2 has been shown to extend median lifespan by 19% (Strong et al., 2016), ameliorate age-associated metabolic and inflammatory dysfunction (Stout et al., 2016), and improve male glucose tolerance across much of adult life (Garratt, Bower, Garcia & Miller, 2017).
The effects of 17aE2 on lifespan and metabolic health are strongly sex-specific. Females accrue no detectable metabolic benefit of 17aE2 treatment, and no life extension has been observed at the two tested concentrations to date Strong et al., 2016). We have previously observed that male-specific increases in glucose tolerance and hepatic mTORC2 signaling with 17aE2 treatment are inhibited in males that are castrated in adulthood, prior to treatment onset . Ovariectomy (OVX) of females prior to treatment with 17aE2 had little effect on these physiological responses, with OVX females showing no improvement in glucose tolerance after 17aE2 treatment. The modulation of sex-specific responsiveness to 17aE2 with male castration, but not female ovariectomy, suggests that the sex-specificity in responsiveness to 17aE2, at least in relation to glucose tolerance, is caused by an interaction with male gonadal hormones, such as production of testosterone.
Untargeted metabolomics provides a powerful approach to understand the underlying metabolic responses that occur after drug treatment or other manipulations. Association of metabolite changes with particular pathways can reveal the biochemical and metabolic processes consistently altered by an intervention, and molecules that are identified as most strongly affected by an intervention are candidates for a role in the intervention response. In this study, we used untargeted metabolomic analysis of liver tissue to identify metabolite responses that occur in mice treated with 17aE2 at the concentration previously shown to extend male but not female lifespan. The liver is a key tissue involved in regulation of glucose, carbohydrate, and protein metabolism, with previous research showing that the metabolite profile of the liver changes in response to lifespan extending manipulations in line with expected alterations in energy utilization (e.g., calorie restriction (Green et al., 2017)). We focused our investigation on liver metabolites that show a sex-specific response to 17aE2 and may underlie the male-specific lifespan response. Comprehensive metabolomics revealed previously unrecognized metabolite changes that were male-specific in response to 17aE2 and were abrogated by male castration. These studies suggest that sex-specific metabolite responses to 17aE2 are modulated by testicular hormones and provide new insight into the mechanisms of sexual dimorphism in lifespan extension.

| Untargeted assessment of sex-specific metabolite responses to 17aE2
To uncover the metabolic underpinnings of sex-specific lifespan extension with 17aE2, male and female mice were treated with 17aE2 for eight months starting at 4 months of age, after which liver samples were subjected to comprehensive untargeted metabolomics analysis of polar and nonpolar (lipid) metabolites. Liquid chromatography-mass spectrometry (LC-MS)-based metabolomics detected 10,271 unique spectral features in liver. To identify metabolites that show a change with treatment consistent with lifespan extension (altered in males but not females), we ran two-way ANOVAs for each metabolite, including sex, treatment, and an interaction between sex and treatment as factors. We focused on the interaction as the statistical term of interest, and prioritized metabolites that were altered in males but not females, as these changes were correlated with the male-specific lifespan response.
Across the initial >10,000 spectral features, we identified six metabolites that showed a significant sex by treatment interaction after correction for multiple comparisons (FDR p < 0.05; Table 1).
Four of these showed a change primarily in males but not females, of which one was strongly reduced in livers of 17aE2treated males, while three were strongly increased in males. These metabolites represent sex-specific signals that correlate with the sex-specific lifespan response (and could be used to determine the effects of gonadectomy on treatment responses, see below).
Chemical networking of the four prioritized metabolites revealed two metabolites at m/z 367.1221 and m/z 365.1086 to be chemically related, with distinct clustering (Figure 1a). Manual inspection of extracted ion chromatograms for both related peaks confirmed a stronger elevation in males treated with 17aE2 than in females ( Figure 1b). Manual inspection of fragmentation MS/MS spectra for the two unknown metabolites at m/z 367.1221 and m/z 365.1086 revealed them to be estriol-3-sulfate and 16-oxoestradiol 3-sulfate, respectively ( Figure 1c). Estriol-3-sulfate and 16-oxoestradiol 3-sulfate are two estrogenic compounds of a similar chemical structure and may be metabolized from 17aE2 via 16 alpha hydroxylation (Longcope 1984). These results indicate that 17aE2 undergoes sex-specific metabolism in male mice to secondary estrogenic products.

| Modulation of liver metabolome with 17aE2 treatment
To test whether 17aE2 results in coordinated changes in metabolites associated with specific metabolic processes, a subset of metabolites were identified from untargeted metabolomic analysis of liver samples following treatment with 17aE2. Using pure standards, 148 metabolites were identified based on retention time, accurate mass characteristics, and MS/MS fragmentation patterns.
Twenty-five of the 148 metabolites were significantly altered in abundance in the livers of 17aE2-treated males compared to untreated males (uncorrected for multiple comparisons), with the vast majority of metabolites (24/25) found to increase (Table 2).
17aE2 was found to result in a similar (24) number of metabolite changes in female livers (Table S1), although only three fatty acids -eicosatrienoic acid, docosatrienoic acid, and behenic acid-increased in the livers of both sexes with treatment. Approximately 50% of the metabolites significantly altered in abundance in 17aE2-treated females decrease in abundance compared to controls.
T A B L E 1 Metabolites showing a sex-specific response to 17aE2 treatment. Data are split into whether the metabolite was initially untargeted or was identified by standard during analysis. M.Z. and retention time (RT, minutes) are shown for untargeted metabolites. Analysis is then further split by whether the metabolite change is seen in females or males. p-values shown in Table 1  We conducted Quantitative Metabolomic Set Enrichment Analysis to test whether 17aE2 leads to coordinated shifts in metabolites associated with any particular metabolic process. This analysis uses generalized linear models to estimate the association between concentration profiles of different metabolites linked to a particular metabolic process and can detect subtle changes in coordinated sets of metabolites (Xia & Wishart, 2010).
We conducted this analysis separately for each sex, as we were particularly interested in metabolites and metabolomic responses that changed specifically in males, and thus are associated with male-specific lifespan extension. Metabolite responses in each sex were referenced against the "metabolite pathway database," which links groups of metabolites to 88 different pathways.
Metabolite pathway associations with 17aE2 treatment within each sex are shown in Table S2. For males, 17aE2 treatment leads to a significant shift in metabolites enriched in pathways associated with "protein biosynthesis" and the "urea cycle." By contrast, for females, 17aE2 treatment leads to modulation of metabolites associated with "betaine metabolism." This mainly reflects the effect of 17aE2 treatment in females reducing hepatic levels of betaine and dimethylglycine, and increasing S-adenosylhomocysteine (Fig. S1).
In males, pathway enrichment for metabolites associated with "protein biosynthesis" reflected the induction by 17aE2 of a malespecific elevation in many amino acids, including alanine, arginine, glutamine, ornithine, and valine ( Figure 2). Metabolite enrichment for the urea cycle reflected increases in both alanine and glutamate, and direct urea cycle intermediates, including argininosuccinic acid, arginine, and ornithine ( Figure 3).

| Identified metabolites showing a sex-specific response to 17aE2
We further assessed changes in individual known metabolites with 17aE2 treatment, again focusing on metabolites that change specifically in males but not females, as indicated by a significant two-way interaction between sex and treatment, and also showing a nominally significant response to 17aE2 in males. Table 1 shows those known metabolites that show a differential response to showing the most significant sex*17aE2 interaction within the known metabolite dataset is phosphocholine (Sex*Treatment interaction: raw p-value <0.0007; Table 1; Figure 4c), which is increased in males but unaffected in females, and can be produced as a product of serine, glycine and threonine metabolism. Also of interest, and potentially linked to protein metabolism, is the malespecific increase in corticosterone, which in females decreases, though not significantly ( Figure 4E; Table 1). We have previously shown that activation of the serum glucocorticoid receptor 1 (SGK1), a direct target of corticosterone, is also elevated sex-specifically in 17aE2-treated males . The sex-specific metabolite that represents a change specifically in females is betaine, which is decreased in 17aE2-treated females compared to controls (Table 1; Fig. S1).

| Sex-specific metabolomic responses are inhibited by male castration
To test whether the sex-specific metabolomic effects of 17aE2 are generated by the underlying effects of male or female gonadal hormones, we castrated male mice and ovariectomized female mice at three months of age, that is, one month prior to administration of 17aE2, and tested whether this postpubertal gonadectomy altered sex-specific metabolite responses to 17aE2. These animals were produced and evaluated in parallel with the intact (sham-operated) animals described above. All six of the metabolites showing a sex by treatment interaction (i.e., those where FDR < 0.05) across the untargeted dataset showed a response that was dependent on gonadectomy. The four metabolites strongly altered in 17aE2-treated male livers, which include estriol-3-sulfate and 16-oxoestradiol 3-sulfate, show a significantly diminished response to 17aE2 treatment in castrated males ( Figure 5), indicated by the interaction terms within males for "Castration" in Table 1. We also tested whether gonadectomy altered the effects of 17aE2 treatment on known metabolites associated with specific metabolic pathways in either sex. Of the 25 known metabolites altered by 17aE2 treatment in normal (sham)-treated males, only two (4-imidazoleacetic acid and N-acetyl-DL-serine) were also significantly changed, in the same direction, in castrated males treated with 17aE2 (Table 2). For 4-imidazoleacetic acid the response in this metabolite with 17aE2 was significantly diminished, as indicated by the "castration" interaction in Table 1. When considering known metabolite responses to 17aE2 in castrated males, enrichment analysis failed to show a significant association in metabolite responses with any particular biochemical pathway after correction for FDR, and there was no evidence that amino acids or metabolites of the urea cycle were affected by 17aE2 treatment in castrated males (Table S3; Fig. S2). Thus, the male-specific elevations in metabolites associated with protein biosynthesis and the urea cycle with 17aE2 treatment require male-specific production of gonadal hormones.
Ovariectomized females did not show significant enrichment for any specific biochemical pathway after correction for FDR, although it is notable that betaine metabolism, which is altered in sham females, was also modulated in OVX females in the analysis before correction for multiple comparisons by the FDR calculation, and this pathway did produce FDR = 0.07 in the ovariectomized females (Table S3).
2.5 | Sex-specific metabolite responses in plasma are also inhibited by castration To test whether the testosterone dependence of male-specific responses is an effect that is restricted to the liver, or represents a more general sexual dimorphism in metabolomic response, we assessed metabolomic responses to 17aE2 treatment in plasma. As the strongest sex-specific responses in liver were observed for bioactive lipids, we restricted our analysis to this sampling method.
We observed 13 metabolites that both showed a significant sex*17aE2 interaction after correction for FDR and represented responses where the metabolite changed significantly in intact males but not females (thus correlating with the lifespan response). In each case, the significant male response was not observed in castrated males (Table S4), highlighting its dependence on male gonadal hormones, just as in the analysis of male-specific liver metabolic changes.
In addition to these initially unidentified metabolites in plasma, our analysis identified 24 different free fatty acid species (Table 3).
Ten of these were significantly increased in the plasma of 17aE2treated males. Only two showed a shared effect in females. 17aE2treated female mice show a significant reduction in a different subset of these lipids (Table 3). The effects of 17aE2 on plasma free fatty acid abundance are greatly diminished by castration, with only one of 24 fatty acids changing with treatment (Table 3), consistent with the analysis of castration effects on liver metabolite profile.
2.6 | Changes in amino acid abundance with 17aE2 are not observed in skeletal muscle Skeletal muscle is an additional important source of amino acids that can move to the liver during catabolic periods. To test whether the increase in amino acids in livers was correlated with changes in muscle, we assessed whether amino acids levels were increased in quadriceps samples collected at the same time point as livers. We observed no changes in amino acids in response to 17aE2 (Fig. S3).
Furthermore, there was a general lack of metabolite responsiveness in muscle to 17aE2, with enrichment analysis failing to show a coordinated response for metabolites in any included pathway (Table S5).

| DISCUSSION
Our results document marked sex-specific metabolomic effects of 17aE2, a steroid that leads to substantial male-specific lifespan extension (Strong et al., 2016), in addition to male-specific F I G U R E 2 Elevated hepatic amino acids in males treated with 17aE2. Enrichment analysis (Table S2) showed that metabolites associated with protein biosynthesis were significantly enriched in males treated with 17aE2. This reflects an increase in the abundance of a range of amino acids, an effect that was observed in males but not females. * = p < 0.05 calculated with a Student's t test. N = 7-9 per treatment group, per sex. Error bars show mean AE SEM for standardized abundance values improvements in glucose tolerance  and lowered age-related inflammatory dysfunction (Stout et al., 2016). Although 17aE2 has no detectable lifespan effect in females, our data show that it does produce changes in liver and plasma metabolite profiles in this sex, changes that differ from those seen in males. In males,  In our study, we also observe elevations in metabolites related to the urea cycle with 17aE2, which are produced as a consequence of amino acid catabolism. This could suggest that 17aE2 promotes the utilization of amino acids as a metabolic fuel, potentially in an amino acid specific way, although this hypothesis requires experimental validation. In our study, we observe significant elevations in glucogenic amino acids (e.g., alanine, arginine, glutamine, and valine), which could suggest that amino acid metabolism provides energy via conversion to glucose. It is notable that valine was significantly  The elevations in amino acids and urea cycle products with 17aE2 represent a male-specific phenotype, in that the response is not observed in females. Such sex-specific effects on a metabolomic scale are consistent with the sex-specific effects of 17aE2 on glucose tolerance  and hypothalamic inflammation (Sadagurski, Cady & Miller, 2017  inhibitor (Schriefers et al., 1991), one such mechanism might involve inhibition of testosterone's conversion to dihydrotestosterone, which is a more potent binder of the AR. This could inhibit specific effects of testosterone on metabolism, which include protein anabolism and inhibition of urea cycling (Lam et al., 2017;Rossetti, Steiner & Gordon, 2017), and contribute to the observed elevation of amino acids and urea cycling. Under this hypothesis, inhibition AR expression in mice would be expected to inhibit 17aE2 responses, and other 5alpha reductase inhibitors might induce similar effects to 17aE2.
Another related hypothesis is that the activity, metabolism and/or signaling effects of 17aE2 are dependent on testosterone, and without the presence of testosterone the biological effects of 17aE2 are weaker. Two of the liver metabolites that showed strong, sex-specific responses to 17aE2 were the estrogenic compounds estriol-3-sulfate and 16-oxoestradiol 3-sulfate-products that can be generated from metabolism of estradiol. Estriol is produced from the placenta during pregnancy in females and during nonpregnant states is generated from estradiol or estrone via 16-alpha hydroxylation (Longcope 1984). Most estriol in circulation is conjugated, of which estriol-3-sulfate is the most abundant form (Tanaka et al. 1984). 16-oxoestradiol is an intermediate metabolite in the conversion of estradiol to estriol (Pasqualini & Kincl 1985). Considering that many of the xenobiotic enzymes that metabolize steroids show sexual dimorphism in expression (Waxman & Holloway, 2009), it is possible that 17aE2 may be metabolized into a different form, in either males or females, which could influence either the transport of this steroid, its activation/inactivation, or its ability to bind to estrogen receptors. 17aE2 is readily conjugated into various forms in humans (Hobe et al. 2002), and understanding the metabo-

| EXPERIMEN TAL PROCED URES
UM-HET3 mice were produced as previously described Strong et al., 2008). The mothers of the test mice were CByB6F1/J, JAX stock #100009, whose female parents are BALB/ cByJ and whose male parents are C57BL/6J. The fathers of the test mice were C3D2F1/J, JAX stock #100004, whose mothers are C3H/ HeJ, and whose fathers are DBA/2J. Mice in breeding cages received Purina 5008 mouse chow, and weaned animals were fed Purina 5LG6.
Mice were housed as previously described Strong et al., 2008) in plastic cages with metal tops, using ¼-inch corn-cob bedding (Bed O'Cobs, produced by The Andersons, PO Box 114, Maumee, OH, USA). Mice were given free access to water, using water bottles rather than an automated watering system. Mice were housed in ventilated cages and were transferred to fresh cages every 14 days. Temperature was maintained within the range of 21-23°C.

| Surgical procedures
At three months of age, all animals went through castration, ovariectomy, or a sham procedure. All animals were anaesthetized by injection of 250 mg/kg tribromoethanol, and given a single preoperative injection of the analgesia carprofen, at 5 mg/kg.

| Castration and sham castration
After surgical preparation, an incision was made in the caudal end of each scrotal sac, the testicle was pulled through the incision by gentle traction, and the blood vessels, vas deferens, and deferential vessels were clamped and sutured. The incision was closed with tissue adhesive. For sham surgery, the testicles were exteriorized and then replaced in the scrotum, without being ligated or excised.

| Ovariectomy or sham ovariectomy
After surgical preparation, an incision was made on the left side perpendicular to the vertebral column approximately midway between the iliac crest and the last rib. The ovarian fat pad was grasped and exteriorized. The pedicle under the ovarian blood vessels and fat pad under the ovary were grasped and crushed, the pedicle cut on the ovary side and the ovary removed, and the blood vessels tied with absorbable suture. The abdominal wall was closed with absorbable suture, and skin was closed with staples. The procedure was then repeated on the opposite side. For sham ovariectomy, animals underwent the same surgical procedure, but the ovary and fat pad were exteriorized and replaced without being excised.

| Diets
At four months of age, animals in different sibling groups were randomly allocated to control or 17aE2 treatment. Animals in the control group remained on the 5LG6 diet, while animals allocated to 17aE2 had their diet switched to a food containing this drug. For measurement of small lipids, identical preparation protocols were used with the exception of 80:20 ethanol:water for extraction solvent, followed by solid-phase extraction of lipid metabolites using a Strata-X polymeric 10 mg/ml 96-well SPE plate, as previously described (Watrous et al., 2017).

| LC-MS-based metabolomics
LC-MS/MS-based metabolomics analysis was performed using a Thermo QExactive orbitrap mass spectrometer coupled to a Thermo 4.9 | Statistical analysis 4.9.1 | Analysis of untargeted metabolites We first focused on metabolite responses that occur in mice where gonadal hormone production was unaltered, that is, mice that had only been through sham surgery, as data on lifespan responses to 17aE2 were conducted in male and female mice without surgical manipulation (Strong et al., 2016).
For each metabolite, we fitted a two-way ANOVA model testing the effects of sex, treatment, and the interaction between sex and treatment with data from sham-operated males and females. The interaction term in the model was our primary interest as it indicates that the effect of 17aE2 treatment was modified by sex. By scanning through all interactions, we were able to identify metabolites that reveal significant treatment by sex interaction. Because multiple two-way interaction effects were tested across the dataset, we adjusted raw p-values with False Discovery Rate (FDR) method (Benjamini & Hochberg, 1995).
For those metabolites that showed a significant sex by treatment interaction, we subsequently tested whether male castration or female ovariectomy influenced the treatment response within either sex. In this analysis, data were split according to sex, then tested for an effect of surgery (e.g., within males, castration or sham castration; within females, ovariectomy or sham ovariectomy), treatment, and an interaction between surgery and treatment. An interaction between surgery and treatment indicates that the response to 17aE2 depends on whether animals had been gonadectomized (and gonadal hormones are removed) or not.

| Analysis of known metabolites
For analysis of metabolites that were identified against a standard reference library, we conducted Quantitative Metabolomic Set Enrichment Analysis using Metaboanalyst 3.0 (Xia & Wishart, 2010) after matching metabolites to their respective HMDB IDs. Data for control and 17aE2 animals within each sex and surgery group were uploaded to the metaboanalyst 3.0 website separately. We conducted quantitative enrichment analysis by uploading a concentration table that showed abundance values for each metabolite for each individual, with animals differentiated according to whether they were control or 17aE2. Data were log-transformed, and then, enrichment analysis was conducted against the "pathway-associated metabolite sets" library. Enrichment for a specific pathway was considered statistically significant when the FDR p-value was lower than 0.05.
We also looked within the known metabolite data to test whether there were changes in individual metabolites that differed depending on sex, and whether sex-specific responses in these instances were altered by surgical removal of gonads in either males or females. These statistical tests were conducted as part of the larger analysis of untargeted metabolite responses, but we present sex by treatment interaction responses that are uncorrected for FDR.