Abstract

Objective

   Menopause is associated with visceral adiposity, hepatic steatosis and
   increased risk for cardiovascular disease. As estrogen replacement
   therapy is not suitable for all postmenopausal women, a need for
   alternative therapeutics and biomarkers has emerged.

Methods

   9-week-old C57BL/6 J female mice were subjected to ovariectomy (OVX) or
   SHAM surgery (n = 10 per group), fed a standard diet and sacrificed 6-
   & 12 weeks post-surgery.

Results

   Increased weight gain, hepatic triglyceride content and changes in
   hepatic gene expression of Cyp17a1, Rgs16, Fitm1 as well as Il18,
   Rares2, Retn, Rbp4 in mesenteric visceral adipose tissue (VAT) were
   observed in OVX vs. SHAM. Liver RNA-sequencing 6-weeks post-surgery
   revealed changes in genes and microRNAs involved in fat metabolism in
   OVX vs. SHAM mice. Energy Homeostasis Associated gene (Enho) coding for
   the hepatokine adropin was significantly reduced in OVX mice livers and
   strongly inversely correlated with weight gain (r = −0.7 p < 0.001) and
   liver triglyceride content (r = −0.4, p = 0.04), with a similar trend
   for serum adropin. In vitro, Enho expression was tripled by
   17β-estradiol in BNL 1 ME liver cells with increased adropin in
   supernatant. Analysis of open-access datasets revealed increased
   hepatic Enho expression in estrogen treated OVX mice and estrogen
   dependent ERα binding to Enho. Treatment of 5-month-old OVX mice with
   Adropin (i.p. 450 nmol/kg/twice daily, n = 4,5 per group) for 6-weeks
   reversed adverse adipokine gene expression signature in VAT, with a
   trended increase in lean body mass and decreased liver TG content with
   upregulation of Rgs16.

Conclusions

   OVX is sufficient to induce deranged metabolism in adult female mice.
   Hepatic adropin is regulated by estrogen, negatively correlated with
   adverse OVX-induced metabolic phenotypes, which were partially reversed
   with adropin treatment. Adropin should be further explored as a
   potential therapeutic target and biomarker for menopause-related
   metabolic derangement.

   Keywords: Menopause, OVX, Fatty Liver, Transcriptome, Estrogen, Adropin

   Abbreviation: OVX, ovariectomy

Graphical abstract

   Image 1
   [41]Open in a new tab

Highlights

     * •
       OVX increased body weight, liver fat & adverse visceral fat
       adipokine signature.
     * •
       OVX altered liver transcriptome & miRNA profile including fat
       metabolism pathways.
     * •
       Enho was downregulated by OVX & inversely correlated with weight
       gain & liver fat.
     * •
       Hepatic adropin expression was upregulated by estrogen in-vitro &
       in-vivo.
     * •
       Adropin treatment partially reversed OVX induced adverse metabolic
       phenotypes.

1. Introduction

   Menopause defined as the permanent cessation of menstruation resulting
   from the loss of ovarian follicular activity [[42]1], is associated
   with a deranged metabolic phenotype including weight gain, visceral
   adiposity, liver steatosis and insulin resistance resulting in a
   significantly increased risk of developing type 2 diabetes (T2D) and
   cardiovascular disease (CVD) in post-versus pre-menopausal women
   [[43]2]. Weight gain results primarily from decreased energy
   expenditure rather than increased food intake [[44]3,[45]4]. There is a
   shift in adipose tissue distribution from a lower body predominant
   subcutaneous fat depot towards a more visceral abdominal distribution
   pattern [[46]5,[47]6]. Increased visceral adiposity is associated with
   adipocyte hypertrophy and secretion of adipokines and pro-inflammatory
   cytokines implicated in insulin resistance and CVD pathogenesis
   [[48][7], [49][8], [50][9]]. Fat oxidative capacity in the liver and
   skeletal muscle is reduced resulting in excessive lipid accumulation
   and increased insulin resistance [[51]10]. As a major metabolic organ
   as well as a target of the metabolic syndrome, the liver appears to
   play an important role in mediation of metabolic changes associated
   with menopause. Known direct liver involvement includes hepatic
   steatosis [[52]11], inflammation [[53]12], decreased fatty acid
   oxidation [[54]13], and increased hepatic glucose production [[55]14].

   Studies in women with natural or iatrogenic premature cessation of
   ovarian function have shown that estrogen depletion and the disruption
   of estrogen signaling are key factors in menopause-related metabolic
   dysfunction [[56]15]. However, the benefits and harms of estrogen
   replacement therapy (ERT) are still a matter of controversy and ERT is
   not suitable for all women [[57]16].

   Historically, recommendations for prevention and treatment of
   cardiometabolic derangement in postmenopausal women were derived from
   studies conducted primarily in men. Moreover, pre-clinical biomedical
   research in animals did not have equal representation of females,
   reportedly due to the higher susceptibility of males to high fat diet
   (HFD)-induced metabolic dysfunction [[58]17]. However, a growing body
   of evidence shows clear sexual dimorphism in metabolic homeostasis with
   distinct sex- and gender-specific cardio-metabolic phenotypes
   [[59]6,[60]18,[61]19]. In addition to sex-hormones, there are sex
   differences in gene expression and networks that impact sex-specific
   metabolic phenotypes [[62]20,[63]21]. With a mean age at menopause of
   50 and life expectancy over 90 years, women in developed countries are
   expected to spend nearly half their lives in the post-menopausal state
   [[64]22]. There is an urgent need to discover new biomarkers and
   therapeutic targets to alleviate metabolic derangement in this rapidly
   growing population.

   Using RNA-sequencing from liver tissue of a mouse model for human
   menopause (ovariectomy, OVX), we revealed new players in OVX-induced
   metabolic abnormalities. To separate the effects of sex hormone
   deficiency from those of aging and obesity, studies were conducted in
   young adult female mice fed a standard diet.

2. Results

2.1. OVX induced weight gain, pro-inflammatory and adverse adipokine gene
expression in VAT and increased hepatic TG content and related genes

   The study design is depicted in [65]Figure 1A. Baseline body weight did
   not differ between SHAM and OVX mice ([66]Figure 1B), and uterine
   weight was significantly lower in OVX mice, confirming successful
   OVX-induced suppression of endogenous estrogen production
   ([67]Figure 1C). Weight gain in OVX mice was significantly higher than
   in SHAM mice over the course of 12 weeks post-surgery and approached
   forty percent from baseline ([68]Figure 1D). A significant increase in
   gene expression of pro-inflammatory cytokines and adipokines was found
   in VAT of OVX compared to SHAM mice 6-weeks post-surgery
   ([69]Figure 1E). Gene expression of Interleukin 18 (Il18), Retinoic
   acid receptor responder protein 2 (Rarres2), Resistin (Retn) and
   Retinol Binding Protein 4 (Rbp4) were significantly higher in T6/OVX
   vs. T6/SHAM mice, with a similar trend for Interleukin 1 b (Il1b).
   Hepatic triglyceride (TG) content was increased more than 2-fold in OVX
   compared to SHAM mice 12 weeks post-surgery ([70]Figure 1F). These
   findings show that OVX, without additional insult, is sufficient to
   induce a deranged metabolic phenotype relevant to menopause in young
   adult female mice and can thus serve as a model to further investigate
   the molecular mechanisms of OVX-induced metabolic derangement.

Figure 1.

   [71]Figure 1
   [72]Open in a new tab

   OVX induced weight gain, adverse changes in mesenteric visceral adipose
   tissue (VAT) and liver. (A) Study design. (B) Body weight at baseline.
   (C) Uterine weight as percent of body weight. (D) Percent body weight
   change from baseline. (E) Gene expression of pro-inflammatory cytokines
   and adipokines in VAT in SHAM and OVX mice at 6 weeks post-surgery. (F)
   Liver triglyceride (TG) content at 6- and 12-weeks post-surgery. (A–F)
   n = 10 mice/group. (E) determined by qPCR normalized to geometric mean
   of Gapdh, actB and Polr2a. Results are Mean ± SEM; Data analyzed by
   unpaired two-sided t-test (B,E) or one-way ANOVA with Sidak's post-hoc
   correction (C,D,F), ∗P < 0.05; ∗∗P < 0.01; ∗∗∗P < 0.001 vs. SHAM.

2.2. OVX induced transcriptional changes in gene sets for fat metabolism in
the liver

   OVX induced hepatic transcriptional changes were assessed in T6/OVX vs.
   T6/SHAM mice. Principal component analysis (PCA) showed samples cluster
   by gonadal status and not by technical batch ([73]Figure 2A). The
   spread of differently expressed genes (DEGs) is seen in the volcano
   plot ([74]Figure 2B), with a total of 103 DEGs, 35 up and 68
   downregulated. A Heatmap of the RNA-seq data shows the overall
   separation between OVX and SHAM mice, indicating an altered hepatic
   transcriptional response to OVX ([75]Figure 2C). Gene set enrichment
   analysis (GSEA) showed upregulation of genes involved in fatty acid
   metabolism as well as genes involved in TNF-α signaling pathway in OVX
   compared to SHAM mice ([76]Figure 2D, [77]Figure S1). Select
   differentially expressed genes related to fat and energy metabolism
   were chosen for RT-PCR validation as well as for comparison at 12 weeks
   post-surgery ([78]Figure 2E); Cyp17a1 codes for a key enzyme in steroid
   hormone biosynthesis, suppressed in obesity and involved in
   atherosclerosis and dyslipidemia [[79]23]; Cytochrome P450 family 2,
   subfamily b, polypeptide 10 (Cyp2b10) is known to be activated by
   estrogen and calorie restriction through the Constitutive Active
   Receptor (CAR) [[80]24,[81]25]; Fat Storage Inducing Transmembrane
   Protein 1 (Fitm1); G0/G1 Switch 2 (G0s2) and Regulator of G-protein
   signaling 16 (Rgs16).

Figure 2.

   [82]Figure 2
   [83]Open in a new tab

   OVX induced liver transcriptional changes 6-weeks post-surgery. (A).
   Principal component analysis based on top 1000 variably expressed
   genes. (B) Volcano plot; DEGs (differentially expressed genes) marked
   in green (fold change ≥ |2|, FDR ≤0.05). (C) Heatmap with hierarchical
   clustering of DEGs. (D) Top upregulated gene sets by Gene Set
   Enrichment Analysis by normalized enrichment score (NES) and FDR. (E)
   Gene expression of key DEGs in liver in SHAM and OVX mice 6-and-12
   weeks post-surgery determined by qPCR normalized to geometric mean of
   Gapdh, actB and Polr2a. Results are Mean ± SEM. Data analyzed by
   unpaired two-sided t-test; ∗P < 0.05; ∗∗P < 0.01; ∗∗∗P < 0.001;
   ∗∗∗∗P < 0.0001 vs. SHAM. (A–E) n = 10 mice/group.

2.3. OVX repressed hepatic expression of Enho coding for adropin

   We were particularly intrigued by the significant reduction in hepatic
   expression of the Energy Homeostasis Associated (Enho) gene, observed
   in OVX compared to SHAM mice 6- and 12-weeks post-surgery
   ([84]Figure 3A). Enho relative expression was significantly inversely
   correlated with hepatic triglyceride content 12-weeks post-surgery
   ([85]Figure 3B) with a similar trend within the OVX group alone
   (r = −0.58, P = 0.07). Interestingly, we found a positive correlation
   between uterine weight, a reflection of the level of estrogen and
   hepatic Enho expression ([86]Figure 3C). Relative weight gain in OVX
   and SHAM mice at 6 weeks post-surgery also showed a strong inverse
   correlation with Enho liver mRNA expression ([87]Figure 3D), though
   driven only by the inter-group differences.

Figure 3.

   [88]Figure 3
   [89]Open in a new tab

   Adropin is negatively correlated with weight gain and liver TG in OVX
   mice. (A) Enho mRNA in liver in OVX and SHAM mice at 6- and 12-weeks
   post-surgery determined by qPCR normalized to geometric mean of Gapdh,
   actB and Polr2a. (B) Pearson's correlation and linear regression
   between liver triglyceride (TG) content and liver Enho relative
   expression at 12 weeks. (C) Pearson's correlation and linear regression
   between %uterine weight/body weight. (D) Pearson's correlation and
   linear regression between % weight gain from baseline at 6 weeks and
   liver Enho normalized mRNA count from RNA-seq. (A–D) n = 10 mice/group.
   (E) Serum adropin in SHAM and OVX mice, 14- and 28-days post-surgery;
   Results are Mean ± SEM. Data analyzed by unpaired t-test. (F) Pearson's
   correlation and linear regression between serum adropin and %weight
   gain from baseline 28 days post-surgery. (E–F) n = 4–5 mice/group.

   Enho codes for the peptide adropin, a highly conserved hepatokine shown
   to regulate whole body energy homeostasis [[90]26]. Despite the change
   in mRNA levels, no differences in hepatic adropin protein level in
   T6/OVX vs. T6/SHAM or T12/OVX vs. T12/SHAM mice could be detected using
   ELISA ([91]Figure S2). This pattern is consistent with previous reports
   [[92]27], and may be explained by adropin being a secreted peptide.
   Indeed, we found a trend for decreased plasma adropin 28 days after OVX
   which was negatively correlated with weight gain ([93]Figure 3E,F).
   Reduced liver Enho expression and low serum adropin have been
   previously reported in the context of obesity [[94]26], but have never
   been investigated in the context of estrogen deficiency, OVX, menopause
   or disruption of estrogen signaling.

   To understand if Enho expression is directly regulated by E2 and not
   merely a byproduct of changes in body weight or liver fat, we conducted
   in vitro studies in BNL 1 ME murine hepatocytes. We found a 3-fold
   increase in Enho expression in response to E2 administration with an
   increase in adropin levels in the supernatant ([95]Figure 4A,B).

Figure 4.

   [96]Figure 4
   [97]Open in a new tab

   Enho expression is regulated by estrogen status. (A) Enho mRNA in
   vehicle- or E2-treated (10 nM) BNL 1 ME cells at 24 h determined by
   qPCR normalized to geometric mean of Gapdh, actB and Polr2a expression;
   Adropin level measured in cell supernatant (n = 9 replicates/group) (B)
   ENHO mRNA in MCF7 cell line with or without E2 ([98]GPL4133 at
   [99]GSE42619, n = 3 replicates/group). (C) ENHO mRNA in primary human
   hepatocytes ([100]GSE17251, data obtained from 2 female and 4 males at
   2 time points). (D) Enho mRNA in liver from OVX vs SHAM mice (RNA-Seq,
   [101]GSE112947 n = 4 replicates per group) (E) Enho mRNA in liver from
   OVX mice on HFD treated with VEH (43% DMSO, 15% ethanol, and 42%
   saline) vs E2, (RNA-seq, [102]GSE92968 n = 3 replicates/group). Results
   are Mean ± SEM. Data analyzed by unpaired two-sided t-test; ∗P < 0.05;
   ∗∗P < 0.01; vs. vehicle (A-B, E); vs. male (C); vs. SHAM (D).

   To further validate our findings of direct regulation of hepatic Enho
   expression by E2, we analyzed relevant publicly available datasets from
   the Gene Expression Omnibus (GEO). In human cells, E2-treated MCF-7
   breast cancer cells showed a significant increase in ENHO
   ([103]Figure 4C). In primary human hepatocytes higher ENHO expression
   was found in cells obtained from females compared to males
   ([104]Figure 4D), compatible with the sexually dimorphic expression
   pattern observed in mice [[105]28]. Consistent with our own
   experimental data, we found decreased liver Enho expression in OVX vs
   SHAM mice both on chow or HFD ([106]Figure 4E) and a marked increase in
   Enho expression following E2 treatment in OVX mice on HFD
   ([107]Figure 4F).

   To gain insight into the potential mechanisms of estrogen regulation of
   Enho we analyzed data from liver ERα knockout (KO) and ERα ChIP-Seq
   experiments. In mouse livers, Enho expression was reduced in ERα KO vs.
   WT mice ([108]Figure 5A) with a significant peek for ERα binding at
   Enho following E2 treatment ([109]Figure 5B). We found a significant
   ERα binding peak in the promotor region of ENHO in pooled human livers
   (three males and three female, age 44–73 years, [110]Figure 5C). We
   could not detect an estrogen responsive element (ERE) near the Enho
   promoter. Enrichment analysis on Enho co-expressed genes in our RNA-seq
   data using ChEA [[111]29] reveled enrichment for targets of retinoid X
   receptor (RXR) (data not shown), known to form a heterodimer with liver
   X receptor α (LXR). Notably, LXR was previously reported to negatively
   regulate adropin in vitro and in vivo [[112]26].

Figure 5.

   [113]Figure 5
   [114]Open in a new tab

   Liver Enho expression is regulated via ERα. (A) Enho expression in
   female mouse livers in wild-type (WT) and ERα-knockout (ERαKO) mice
   (n = 3 per group, [115]GPL1261, [116]GSE95283). Results are Mean ± SEM
   analyzed by unpaired t-test; ∗P < 0.05 vs. WT mice. (B) ChIP-Seq data
   for mouse liver ERα binding sites with E2 vs control (GEO
   [117]GSE52351, n = 5 mice/group). (C) ChIP-Seq data for human liver ERα
   binding sites (GEO [118]GSE158856, n = 6 samples).

2.4. Adropin treatment partially reversed OVX induced adverse metabolic
changes

   To test the hypothesis that reduced liver adropin expression plays a
   causative role in OVX induced metabolic derangement, we performed a
   small proof of concept intervention study using previously reported
   therapeutic doses of adropin, administered for 6-weeks following OVX.
   As expected, adropin treated mice showed a marked increase in adropin
   serum level ([119]Figure 6A). Surprisingly, neither vehicle nor adropin
   treated OVX mice gained total body weight relative to baseline
   ([120]Figure 6B), possibly due to the significant stress of a prolonged
   twice daily injection regimen. Nonetheless, the adropin treated mice
   exhibited a trend towards increased lean mass ([121]Figure 6C) and
   reduced liver fat ([122]Figure 6D–F) with increased expression of Rgs16
   and reduced expression of the fat transporter cluster of
   differentiation 36 (Cd36) ([123]Figure 6G). Importantly, the adverse
   adipokine pattern in VAT seen with OVX vs SHAM was partially reversed
   with a significant reduction in gene expression of Il1b and Retn and
   trend for Il18, Rars2, Rbp4 in adropin vs. vehicle treated OVX mice
   ([124]Figure 6H). Stearoyl-CoA desaturase-1 (Scd1), a key fat storage
   gene was downregulated in both visceral and subcutaneous adipose
   tissues in adropin treated mice ([125]Figure 6G,H). Scd1 was previously
   shown to be downregulated by adropin in the liver [[126]26]. We found
   no significant change in blood glucose levels including following a
   glucose tolerance test (GTT) in adropin vs vehicle treated mice
   ([127]Figure S3).

Figure 6.

   [128]Figure 6
   [129]Open in a new tab

   Adropin treatment partially reversed OVX induced metabolic derangement.
   (A) Serum adropin determined upon sacrifice in adropin and vehicle
   treated OVX mice 2 h after the last dose. (B) Percent body weight
   change. (C) Body composition at 4 weeks, p-value for change in lean
   mass from ANCOVA with total body weight. (D) Representative slides from
   liver stained with oil-red-o and quantification of lipid droplet area.
   (F) Liver triglyceride (TG) content. (F) Gene expression of fat
   metabolism genes in liver. (G) Gene expression of pro-inflammatory
   cytokines and adipokines in VAT and SCAT; (F,G,H) determined by qPCR
   normalized to geometric mean of Gapdh, actB and Polr2a. n = 4
   mice/group; Results are Mean ± SEM; Data analyzed by unpaired one-sided
   t-test; ∗P < 0.05; ∗∗P < 0.01 vs. vehicle.

2.5. OVX-induced changes in miRNAs involved in fatty acid metabolism

   As miRNAs play a role in post-transcriptional regulation of mRNA, we
   performed miRNA-seq analysis on the same samples used for the RNA-seq.
   The relative abundance of specific miRNAs is presented in
   [130]Figure 3A. We found 48 miRNAs to be differentially expressed in
   OVX vs. SHAM mice, with 24-up and 24 down-regulated ([131]Figure 7A).
   Samples clustered by gonadal status ([132]Figure 7B). Pathway
   enrichment analysis of the predicted targets of the differently
   expressed miRNAs revealed enrichment for the KEGG pathways fatty acid
   biosynthesis and fatty acid metabolism ([133]Figure 7C). We used
   negative co-expression analysis between previously described miRNA-mRNA
   pairs to create a miRNA-mRNA network ([134]Figure S4B). 3 of the miRNAs
   with the most statistically significant differential expression were
   chosen for validation using RT-PCR ([135]Figure 7D). Liver expression
   of miRNA-802 and miRNA-455 has been reported to be sexually dimorphic
   in young adult mice with higher miR-802 and lower miR-455 expression in
   females vs. males [[136]30]. It seems plausible that the opposite
   expression pattern we observed in OVX vs. SHAM results from the loss of
   female sex hormones. miRNA-200a is known to be over-expressed in
   nonalcoholic fatty liver disease and was previously shown to be
   up-regulated in OVX mice [[137]31]. Of note, miRNA-29–3p family has
   been previously reported to regulate hepatic Enho expression in-vivo in
   insulin resistant male mice [[138]32], but we found no significant
   correlation between the two transcripts in OVX and/or SHAM mice
   ([139]Figure S4A).

Figure 7.

   [140]Figure 7
   [141]Open in a new tab

   OVX induced changes in liver miRNA signature 6-weeks post-surgery. (A)
   Volcano-plot for differently expressed miRNAs (B) PCA analysis of miRNA
   expression pattern. Top up (C) and downregulated (D) KEGG pathways from
   targets of differentially expressed miRNA. (D) miRNA expression of
   select differently expressed miRNAs 6 W post-surgery, determined by
   qPCR, normalized to the geometric mean expression of U6 and SnoRNA202.
   (E) Pearson's correlation and linear regression between liver Enho
   normalized mRNA count from RNA-seq at 6 W and liver miR-29 b-3p
   normalized miRNA count from miR-SEQ. Results are Mean ± SEM (n = 10
   mice/group); Data analyzed by unpaired two-sided t-test (C); ∗P < 0.05,
   ∗∗∗∗P < 0.0001 vs. SHAM mice.

3. Discussion

   This study demonstrates that OVX without additional insults is
   sufficient to induce a deranged metabolic phenotype, characteristic of
   the post-menopausal state including weight gain, increased liver fat
   and increased gene expression of pro-inflammatory cytokines and
   unfavorable adipokines in visceral abdominal adipose tissue. While OVX
   in rodents is often used to study human menopause-related metabolic
   dysfunction [[142]33], most studies have used older-age, high-fat diet
   or genetic interventions in addition to OVX to accentuate the deranged
   phenotype. Our experimental design highlights the net effects of
   sex-hormone deficiency.

   OVX led to increased gene expression of Il18, Rares, Retn and RBP4 in
   mesenteric VAT. Il18 is a pro-inflammatory cytokine from the Il1 family
   that is increased in visceral adipocytes of obese humans [[143]34] and
   involved in accelerated atherosclerosis [[144]35]. Increased serum IL18
   level was previously reported in post-compared to pre-menopausal women
   [[145]36]. Il18 has been shown to be downregulated by estrogen in
   multiple tissues [[146]37]. Rares2 codes for the adipokine chimerin, a
   regulator of adipocyte development [[147]38] specific to mesenteric fat
   [[148]39]. Lower serum chimerin level was reported in pre-compared to
   post-menopausal women [[149]40]. Retn codes for the adipokine resistin,
   a driver of insulin resistance in adipose tissue [[150]41] by inducing
   pro-inflammatory cytokines production from macrophages residing in
   adipose tissue [[151]42]. Resistin expression in VAT was shown to be
   downregulated by estrogen in-vitro and in-vivo [[152]43], and increased
   serum resistin was observed in post-menopausal women with the metabolic
   syndrome compared to those without [[153]44]. Rbp4 induces inflammation
   and increased insulin resistance in mice [[154]45]. In humans RBP4
   expression is downregulated by estrogen in-vivo [[155]46] and exhibits
   a sexually dimorphic expression pattern in adipose tissue [[156]47].

   The liver is a target of the metabolic syndrome but also may play a
   causative role in menopause-related metabolic derangement. Thus, we
   focused on transcriptional changes in the liver induced by OVX that
   occurred prior to the changes observed in liver fat. To our knowledge,
   this is the first report of the OVX-induced transcriptional landscape,
   integrating data from same-sample RNA-seq and miRNA-seq, thus
   minimizing the effects of sample handling on the data and allowing for
   establishing co-expression networks. As expected, we discovered
   significant changes, primarily in gene sets and miRNAs involved in
   lipid and fatty acid metabolism.

   In the liver, OVX induced expression of several genes known to regulate
   liver fat content: Fitm1 is a target for Peroxisome
   proliferator-activated receptor gamma (Pparγ), involved in lipid
   droplet formation and steatohepatitis [[157]48]; G0s2 is a master
   regulator of lipid and energy metabolism [[158]49] and Rgs16 provides a
   signaling mechanism for glucose production to inhibit G-protein-coupled
   receptor-stimulated fatty acid oxidation in hepatocytes [[159]50], with
   a sexually dimorphic expression pattern [[160]28]. Recently, Rgs16 and
   G0s2 have been shown to regulate hepatic substrate oxidation as well as
   inflammation and fibrosis in livers of obese male mice [[161]51].

   Of particular interest is our discovery of reduced liver Enho
   expression in OVX compared to SHAM mice, and its strong inverse
   correlation with adverse menopause-related clinically important
   phenotypes, namely weight gain and hepatic steatosis. Enho codes for
   adropin, a conserved 76 amino-acid peptide, highly expressed in the
   brain. In the liver, adropin is considered a hepatokine, with amino
   acids 1–33 serving as signal peptide and Adropin^34-76 as the secreted
   portion [[162]26]. Adropin has both paracrine effects on hepatocytes as
   well as endocrine effects on muscle and adipocytes. The cell surface
   receptor for adropin has not yet been characterized, but some studies
   suggest its effects are mediated via GPR19 [[163]52,[164]53]. Adropin
   has been implicated in the regulation of glucose and lipid homeostasis
   mostly in studies conducted in male mice. Adropin-over-expressing male
   and female mice display milder hepatosteatosis and insulin resistance
   compared to WT mice when fed HFD [[165]26]. Consistently, male and
   female adropin knock-out mice exhibit hepatic steatosis, increased
   serum TG, insulin resistance and increased relative fat mass [[166]54].
   Importantly, numerous studies in male mice have investigated the
   therapeutic potential of Adropin in ameliorating HFD-induced metabolic
   derangement and have demonstrated beneficial effects in key metabolic
   organs. Systemic administration of Adropin^34-76 to male WT mice on HFD
   attenuated liver steatosis and hyperinsulinemia [[167]26], suppressed
   hepatic gluconeogenesis in vivo and in vitro [[168]55] and decreased
   liver fatty acid uptake via downregulation of Cd36 [[169]56],
   consistent with our results. In another study [[170]57], adropin
   treatment ameliorated NASH progression by up regulating the expression
   of Glutamate-Cysteine Ligase Catalytic and Modifier Subunits (Gclc,
   Gclm) as well as Glutathione Peroxidase 1 (Gpx1) via the factor Nuclear
   factor erythroid 2-related factor 2 (Nrf2). Other metabolically
   beneficial effects of adropin have been reported in male mice in
   skeletal muscle and adipocytes [[171]58,[172]59]. In contrast to the
   reports regarding the positive effects of adropin on glucose
   homeostasis [[173]60], in our model, as in post-menopausal women,
   estrogen depletion does not lead to overt hyperglycemia, which may
   explain why we found no significant effects of adropin on glucose
   levels.

   In humans, reduced serum adropin has been reported in a wide range of
   conditions of cardiometabolic dysregulation [[174]61]. Specific
   polymorphisms in the ENHO gene in humans have been described leading to
   lower serum adropin levels and an increased severity of glucose
   homeostasis impairment and fat metabolism disorders [[175]62]. In
   women, lower levels of serum adropin were observed in women with
   polycystic ovary syndrome, gestational diabetes and endometrial cancer
   [[176][63], [177][64], [178][65]]. Interestingly, unlike in men,
   pre-menopausal women exhibited no association between serum adropin and
   BMI [[179]54] or LDL cholesterol [[180]66]; Though serum adropin was
   reported to be affected by changes in dietary fat content in young
   women [[181]67]. To our knowledge, thus far, only one brief report
   related serum adropin to menopausal status, where a differential
   response to meal composition on serum adropin was observed in pre but
   not post-menopausal women [[182]68].

   Known regulators of hepatic adropin are feeding state, nutrients and
   clock signals [[183]60]. In mice, a short term (2 day) high-fat-diet
   leads to increased hepatic Enho expression, whereas a prolonged HFD
   (1–3 months) as well as in genetic mouse models of obesity caused by
   melanocortin receptor or leptin deficiency, liver Enho is downregulated
   [[184]26,[185]66]. In HepG2 cells, adding cholesterol to the culture
   was shown to downregulate ENHO. Treatment with an LXRα agonist (GW3965)
   downregulated ENHO in HepG2 cells, while antisense RNA targeting LXRα
   blocked this effect [[186]26,[187]66]. Moreover, lean, chow-fed, male
   B6 mice treated with an I.V. GW3965 injection exhibited a rapid
   reduction in hepatic Enho expression 4 h after injection
   [[188]26,[189]66]. Some evidence also points to Enho regulation by the
   circadian clock; Treatment with a Rev-erb agonist was shown to suppress
   hepatic Enho in vitro and in-vivo while the RORα/γ inverse agonist
   (SR1001) suppressed Enho in-vitro [[190]26].

   To our knowledge this is the first report of Enho direct regulation by
   sex-hormones and its association with OVX-related metabolic
   derangement. Our findings of Enho regulation by estrogen are supported
   by data we extracted and analyzed from available in vivo and in vitro
   studies in mice and humans. Results from these analyses showed
   increased hepatic Enho expression in estrogen-treated vs. untreated OVX
   HFD mice, upregulation of ENHO in response to 17β-estradiol in a human
   breast cancer cell line, higher expression in hepatocytes from women
   vs. men and a genomic ERα binding site by ENHO in liver. The mechanisms
   of hepatic Enho regulation by estrogen are elusive and remain to be
   investigated. LXRα has been shown to directly downregulate hepatic
   Enho. As ERα is known to interact with liver LXRα and inhibit its
   function [[191]69], a plausible mechanism for Enho regulation by
   estrogen involves ERα repression of LXRα-mediated inhibition of Enho
   transcription. In a recently available dataset, we found reduced
   hepatic Enho expression in OVX vs SHAM both in ERα^f/f as well as in
   liver specific ERα knockout (LERKO) mice [[192]70], suggesting
   additional non-ERα or extra hepatic regulation by estrogen. This is
   further supported by the reported increase in hepatic Enho expression
   following E2 treatment to LERKO mice [[193]71].

   Importantly, using a small proof-of-concept intervention study, we were
   able to show that adropin reversed some of the metabolic derangement
   induced by OVX including a trended increase in lean body, a decrease in
   VAT gene expression of Il18, Il1b, Rares, Retn and reduced liver fat
   with marked upregulation of Rgs16 gene expression. Larger studies with
   varying doses and regimens of adropin treatment as well as additional
   metabolic insults such as HFD seem warranted.

   This study has several strengths. By using young adult female OVX mice
   on a normal diet, we avoided the confounding effects of aging and
   obesity. Our initial unbiased approach using NGS of the hepatic
   transcriptome enabled the discovery of a new and potentially druggable
   player in OVX-induced metabolic abnormalities. Same-sample mRNA and
   miRNA profiling allowed for direct interaction analysis. We employed a
   broad-spectrum of experimental methodologies spanning from studies
   in-vivo in the OVX mouse model, in-vitro in a murine hepatocyte cell
   line model, as well as bioinformatic analyses of our own data and
   in-silico analysis of relevant open-access data. Importantly, our
   results may have therapeutic implications, as administration of
   synthetic adropin^36-74 has been shown to improve multiple metabolic
   parameters, albeit in mostly in male mice. Indeed, in proof-of-concept
   intervention study we showed that adropin treatment partially reversed
   deranged phenotypes induced by OVX.

   This study is not without limitations. In-vivo studies were performed
   after an overnight fast. As Enho liver expression is suppressed by
   fasting, our results may have underestimated the magnitude of
   difference in Enho and adropin between OVX and SHAM mice. Additionally,
   the intervention study may have been too small and short to show more
   significant effects; conversely, prolonged twice daily injections may
   have put too much stress on the mice.

   In conclusion, this study indicates that changes in adropin liver
   expression may contribute to OVX-induced metabolic derangement and
   provides a basis for further intervention studies in OVX animals. In a
   search for new therapies and biomarkers beyond estrogens to improve
   women's health in the postmenopausal phase, the association between
   perimenopausal serum adropin level and the changes in metabolic
   parameters over time in post-menopausal women warrants further
   evaluation. Finally, although our study was aimed at searching for new
   targets to alleviate metabolic dysfunction associated with menopause,
   estrogens and ERα play a role in hepatic steatosis and glucose
   metabolism also in males and thus our results may be relevant to both
   sexes [[194]71].

4. Methods

4.1. Animals & experimental protocol

   Forty 8-week-old C57BL/6 J female mice were purchased from Harlan
   (Rehovot, Israel), housed in constant temperature rooms with 12-hour
   light/dark cycles and allowed free access to Harlan Teklad chow (2018S)
   and water. 9-week-old mice were randomly subdivided into 2 groups of
   bilateral ovariectomy (OVX) and 2 groups of sham surgery (SHAM), (10
   animals/group). To allow metabolic changes to occur a group of each OVX
   and SHAM animals was sacrificed 6- and 12-weeks post-surgery (T6/OVX,
   T6/SHAM, T12/OVX, T12/SHAM, respectively, [195]Figure 1A). Body weight
   was determined weekly and uterine weight was determined upon sacrifice
   to confirm successful OVX. For adropin serum measurement study, five
   mice underwent OVX and four mice SHAM surgery at 3 months-of-age. Blood
   from facial vein was drawn at 2- and 4-weeks post-surgery. For adropin
   treatment study, eight 5-month-old C57BL/6 J female mice underwent OVX
   and after one week were randomized to receive twice daily
   intraperitoneal adropin treatment (450 nmol/kg, china peptide company)
   or vehicle (PBS/0.1% BSA) for 6 weeks. After 4 weeks of treatment, body
   composition was determined by EchoMRI-100HTM (EchoMedical Systems LLC,
   Houston, TX, USA). All mice were fasted overnight and killed by CO[2]
   inhalation. Liver and mesenteric visceral adipose tissue (VAT), the fat
   pad most analogous to human intra-abdominal VAT [[196]72], were
   immediately harvested and kept frozen at −80ᵒC. The study was approved
   by The National Council for Animal Experimentation and the Hebrew
   University ethics committee (approval numbers IL-10-12-113,
   MD-19-15747-4). All studies were conducted in accordance with ARRIVE
   guidelines.

4.2. Gene and miRNA expression analysis

   RNA was extracted from liver and mesenteric VAT using BIO-TRI RNA
   (Bio-Lab), reverse transcribed into cDNA (qScript cDNA Synthesis Kit,
   Quanta Biosciences), and analyzed with SYBR Green-based quantitative
   RT-PCR. Relative mRNA expression was determined by the comparative CT
   method. For each sample, the mean cycle threshold (CT) for each gene
   (run in triplicate) was normalized to the geometric mean of the mean CT
   of the 3 reference genes (Gapdh, actB and Polr2a) using the formula:
   2^-(gene of interest CT− reference CT). The resulting ΔCT for each gene
   was used to calculate relative gene expression changes between samples.
   Genes with CT values > 35 were considered not expressed.
   [197]Supplementary Table 1A provides all the primer sequences used in
   this study. miRNA-specific cDNA was generated with TaqManTM microRNA
   Reverse Transcription Kit (4366596, Applied Biosystem). Relative miRNA
   expression was determined by the comparative CT method and normalized
   to the geometric mean expression of U6 (TaqManTM microRNA assay: ID
   001973) and SnoRNA202 (TaqManTM microRNA).

4.3. RNA-sequencing, gene expression profiling and pathway enrichment
analysis

   RNA was extracted from liver tissue samples of T6/OVX and T6/SHAM mice
   with QIAGEN miRNeasy Mini Kit (cat. 217004, Qiagen). All RNA samples
   had an RNA Integrity Number (RIN) of more than 8.1 confirmed by Agilent
   TapeStation system. RNA libraries were generated with 1ug of RNA as
   input using the TruSeq RNA Sample Prep kit (Illumina) and
   poly(A)-enriched according to the TrueSeq protocol. Single-Read 75
   reads were sequenced on 4 lanes of an Illumina NextSeq 500. Output was
   ∼29 million reads per sample. Poly-A/T stretches, and Illumina adapters
   were trimmed using cutadapt [[198]73]. Resulting reads shorter than
   30bp were discarded. Reads were mapped to the Mus musculus GRCm38
   reference genome using STAR [[199]74] supplied with gene annotations
   downloaded from Ensembl release 88 with EndToEnd option and
   outFilterMismatchNoverLmax set to 0.04. Expression levels for each gene
   were quantified using htseq-count [[200]75]. Analysis of differentially
   expressed genes between SHAM and OVX mice was performed using DESeq2
   [[201]76] with the betaPrior, cooksCutoff and independentFiltering
   parameters set to False. Principle component analysis (PCA) was
   performed with DESeq2 package. Heatmaps were created with Morpheus
   ([202]https://software.broadinstitute.org/morpheus). Gene set
   enrichment analysis (GSEA) was performed using GSEA and MSigDB
   [[203]77]. Raw P values were adjusted for multiple testing using the
   procedure of Benjamini and Hochberg's false discovery rate (FDR
   q < 0.05)

4.4. Micro-RNA sequencing

   micro-RNA (miRNA) sequencing was performed on same liver samples used
   for RNA-seq. Libraries were prepared using TruSeq Small RNA. SR60 reads
   were sequenced on 1 lane of an Illumina HiSeq 2500 (v4). Output was ∼9
   million reads per sample. Low quality bases were trimmed from the reads
   using Burrows-Wheeler Alignment [[204]78], adapters were trimmed using
   cutadapt [[205]73]. Reads were mapped to the mouse genome using Bowtie
   [[206]79]. Resulting reads shorter than 17bp and longer than 25bp, were
   discarded. Reads were collapsed into tags for quantification. BLAT was
   run on tags against 3 blast databases (hairpin, mature and nt). Tags
   that were not mature miRNA were filtered out, and remained reads were
   merged by miRNA id. Normalization and differential expression analysis
   were performed by DESeq2 with the betaPrior, cooksCutoff and
   independentFiltering parameters set to False. Raw P values were
   adjusted for multiple testing using Benjamini and Hochberg method. GO
   terms and KEGG Pathway enrichment analysis for DE miRNAs were performed
   using mirPath v.3 [[207]80]. Experimentally validated miRNA targets
   were downloaded from DIANA TarBase v8 [[208]81] and visual miRNA-mRNA
   network were created using cytoscape 3.8 [[209]82] based on a negative
   significant Pearson's correlation (r  ≤  −0.5, p ≤ 0.05). Pathway
   analysis for miRNA targets was performed using DIANA-miRPath v3. 0
   [[210]80], using tarbase as interaction database and Fisher's exact
   test for KEGG pathway enrichment analysis with FDR correction.

4.5. In vitro studies in the BNL 1 ME murine liver epithelial cell line model

   Murine BNL 1 ME liver cells [[211]83] (ATCC®TIB-75) were cultured in
   Dulbecco's modified eagle medium (DMEM/10% FBS) with a medium change
   every 3 days. All experiments were conducted in hormone-free medium
   (charcoal stripped fetal bovine serum; Biological industries,
   04-201-1 A). Cells were plated at density of 9x10^5 cells/well and
   treated with 17β-estradiol (E2) (Calbiochem, #3301) at 10 nM for 24 h.
   Experiments were performed in triplicates and repeated 3 times. Adropin
   in cell supernatant was measured using ELISA kit (Phoenix
   Pharmaceuticals, EK-032-35) according to manufacturer's instructions.

4.6. Liver triglyceride content

   Liver triglyceride content was determined by a Triglyceride
   colorimetric assay (No. 10010303, Cayman Chemical). Triglyceride
   content was normalized to protein content determined by Bio-Rad Protein
   Assay (Cat. #500-0006, Bio-Rad).

   Liver slides (3 per mouse) were stained using oil-red-o (ab150678,
   abcam) per manufacturer's instructions. The entire slides were
   photographed and scanned on Olympus VS200 and automatically analyzed
   using ImageJ (1.8.0_172) as previously described for quantification of
   lipid droplet area [[212]84].

4.7. Adropin serum measurement

   Serum adropin was measured in duplicates using an EIA kit (EK-032-35,
   Phoenix Pharmaceuticals) as per manufacturer's instructions.

4.8. Analysis of open access datasets

   Expression data for OVX vs SHAM mouse liver on chow and HFD were
   downloaded from GEO [213]GSE112947. Specific experimental methods were
   previously published [[214]85]. Briefly, sixteen female C57BL/6 J were
   maintained on a chow (Ralston Purina Company) or HF/HS diet (Research
   Diets D12266B) at 8 weeks of age until 16 weeks of age. At 6 weeks of
   age the mice were ovariectomized or sham operated. Frozen liver tissues
   were RNA-Seq expression profiling using Illumina HiSeq 2000. We used
   DESeq2 for differential expression analysis.

   Expression data for OVX HFD mouse liver with E2 or vehicle treatment
   were downloaded from GEO [215]GSE92968. The specific experimental
   methods were previously published [[216]86]. Briefly, female C57BL/6 J
   mice were ovariectomized at 8 weeks of age, switched to a high-fat diet
   (Harlan TD.88137) and divided randomly to vehicle or E2 (0.72 mg,
   60-day release E2 pellets, Innovative Research of America) for six
   weeks till sacrifice. Three liver RNA samples from each treatment group
   were used for RNA-Seq. Gene expression values quantified from BAM files
   were calculated using StrandNGS and DESeq normalization. Normalized
   data was directly available to us for download. We assessed the data
   for quality control using DSEeq package for R. We used unpaired t-test
   for Enho expression analysis in E2-vs. vehicle-treated mice.

   Expression data for mouse ERαKO liver were downloaded from GEO
   [217]GSE95283. The specific experimental methods were previously
   published [[218]87]. Briefly, WT and global ERαKO female mice (lacking
   exon 3 of Esr1) were placed on regular ad-libitum diet (NIH-31, Harlan
   Laboratory) and sacrificed at age 20 weeks. Liver tissue was harvested
   for RNA extraction and hybridization on [219]GPL1261 Affymetrix Mouse
   Genome 430 2.0 Array. We assessed raw data for quality control and log2
   transformed using the R packages GEOquery, limma and umap. Probe ID
   1428739_at was used to obtain normalized Enho expression. Enho
   expression in WT vs ERαKO samples was compared using an unpaired
   t-test.

   Expression data in MCF7 cell line were downloaded from GEO
   [220]GSE42619. The specific experimental methods were previously
   published [[221]88]. Briefly, MCF-7 cells were cultured for 48 h in
   medium with 10% charcoal-stripped FBS, and then were treated with or
   without E2 for 24 h before collection (n = 3 replicates/per group).
   Gene expression analysis was performed using Agilent-014850 Whole Human
   Genome Microarray. We assessed raw data for quality control and log2
   transformed using the R packages GEOquery, limma and umap. Probe IDs
   9947 was used to obtain ENHO expression. Unpaired t-test was used to
   compare ENHO expression in between each two pairs of groups.

   Expression data in primary human hepatocytes were downloaded from GEO
   [222]GSE17251. The specific experimental methods were previously
   published [[223]89]. Briefly, primary hepatocytes were isolated from
   surgical liver biopsies obtained from six individual donors, 2 females
   and 4 males. Cells were incubated in fresh medium dissolved in DMSO for
   6 and 24 h, followed by RNA isolation and gene expression profiling
   using Affymetrix Human Genome U133 Plus 2.0 Array. We assessed raw data
   for quality control and log2 transformed using the R packages GEOquery,
   limma and umap. Probe ID 228403_at was used to obtain normalized ENHO
   expression. ENHO expression in male vs female samples at both time
   points was compared using an unpaired t-test.

   Data for mouse liver ERα ChIP-Seq were downloaded from GEO
   [224]GSE52351. The specific experimental methods were previously
   published [[225]90]. Briefly, liver fragments were harvested from
   8-week-old female C57/BL6 OVX mice and treated ex vivo with 10 nM E2 or
   ethanol as vehicle for 45 min in DMEM (n = 5 mice/condition). ChIP-Seq
   performed with anti-ERa HC-20 antibody (Santa Cruz Biotechnology) and
   10 μL of anti-ERa Ab-10 (NeoMarkers). ERα binding peaks were called
   using Model-based Analysis of ChIP-Seq (MACS) [[226]91]. Data for human
   liver ERα ChIP-Seq were downloaded from GEO [227]GSE158856. The
   specific experimental methods were previously published [[228]92].
   Briefly, six human liver tissues (three males and three female, age
   44–73 yrs) were pooled. ChIP-Seq was performed using a commercial
   service (Active Motif, Carlsbad, CA, USA) with an anti-ESR1 antibody
   (sc-543, Santa Cruz Biotech, Dallas, TX, USA). Peaks were called using
   Homer v4.10 [[229]93]. We used the resulting bedgraph files from both
   data sets to display significant peaks using the IGV browser for
   windows ([230]http://software.broadinstitute.org/software/igv/) on top
   of the mouse (mm9) and human (hg38) genomes, respectively.

4.9. Statistical analysis

   Statistical significance was calculated using GraphPad Prism for
   Windows (GraphPad Software, La Jolla California USA). Outliers were
   automatically removed using the ROUT method with Q = 1%. Unpaired
   Student's t-test was used to compare means of 2 groups. To assess
   differences between multiple means one-way ANOVA followed by Sidak
   post-hoc correction was performed. Correlation coefficients were
   calculated with Pearson's correlation analysis. Differences of P < 0.05
   were considered significant.

Author contributions

   JS & IG – planned and conducted experiments, acquired, and analyzed
   data, wrote and prepared manuscript; ECK, OY, & NH - conducted
   experiments; HB - analyzed data; NL, NO & JT conducted experiments and
   analyzed data. RDP – designed research, wrote and prepared manuscript.

Acknowledgments