Background

The hypothalamus is the central hub controlling energy homeostasis, stress response, temperature, learning, feeding, sleeping, social behavior, sexual behavior, hormone secretion, reproduction, osmoregulation, blood pressure, visceral activities, emotion, and circadian rhythms[1]. The hypothalamic energy-sensing system, especially circuits regulating food intake, is critical for life span extension [2]. Elevated metabolic activity involving many regulatory signaling pathways has been reported in aged hypothalamus, including enhanced mTor signaling [3, 4]. Decreases in gonadotropin-releasing hormone (GnRH), Ghrh, Trh, monoamine neurotransmitters, and blood supply are also hallmarks of aging hypothalamus [5].

Previous studies have shown that 17α-estradiol extends the lifespan of male mice, has beneficial effects on metabolism and inflammation with a similar mechanism as that of rapamycin and acarbose [6–8]. Further investigations revealed certain unique features of 17α-estradiol in life extension distinct to rapamycin and acarbose [9, 10]. Further, 17α-estradiol was shown to target hypothalamic POMC neurons to reduce metabolism by decreasing feeding behavior via anorexigenic pathways [11]. Interestingly, the lifespan extension effect is only found in male animals [12]. The safety of 17α-estradiol is key for translation into clinical treatment, and the side effects on reproduction and potential feminization by 17α-estradiol treatment need to be considered. However, contradictory results have been reported regarding its side effects on reproduction and feminization [6, 13, 14]. A previous study showed that 17α-estradiol elicits similar genomic binding and transcriptional activation through Esr1 (estrogen receptor α, ERα) to that of 17β-estradiol, suggesting a potential side effect on reproduction and feminization [15]. Therefore, the underlying mechanism of lifespan extension and the safety of 17α-estradiol still need further investigation and verification.

In this report, we used single-nucleus transcriptomic sequencing and performed supervised clustering of neurons by neuropeptides, hormones, and their receptors. Supervised clustering provides better resolution in cell cluster screening than is achieved by traditional unsupervised clustering. The effects of 17α-estradiol on metabolism, stress responses, ferroptosis, senescence, inflammation, and synaptic activity were assessed in each neuron subtype, and the most sensitive neurons were ranked. The effects of 17α-estradiol on reversing measures of aging were evaluated by two opposing regulatory networks involved in hypermetabolism, stress, inflammation, and synaptic activity. Several key endocrine factors from serum were checked. Additionally, the potential side effects of 17α-estradiol on certain neurons were also evaluated.

Materials and methods

Animals, treatment and tissues

Nine Norway brown male rats were acquired from Charles River (Beijing). Aged rats were housed one per cage and young rats (1 month old) were housed two per cage. Aged rats were randomly allocated into control and 17α-estradiol-treated groups. Aged rats treated with 17α-estradiol (Catalog #: E834897, Macklin Biochemical, Shanghai, China) were fed freely with regular diet mixed with 17α-estradiol at a dose of 14.4 mg/kg (14.4 ppm), starting at 24 months of age for 6 months according to prior reports [16, 17]. The young rats were fed a regular diet without 17α-estradiol continuously for 3 months until 4 months old. All rats had ad libitum access to food and water throughout the experiments. The rats were then euthanized via CO₂, hypothalami, testes and blood serum were collected for subsequent experiment procedures. All blood samples were collected at 9:00-9:30 a.m to minimize hormone fluctuation between animals. All animal procedures were reviewed and approved by the Institutional Animal Care and Use Committee at Nantong University.

Enzyme immunoassays

Enzyme immunoassays kits for rat Oxt (Catalog #: EIAR-OXT), Corticotropin Releasing Factor (Catalog #: EIAR-CRF), and gonadoliberin-1 (Catalog #: EIAR-GNRH) were obtained from Raybiotech (GA, USA). Enzyme immunoassay kits for rat serum bioavailable testosterone (Catalog #: ml002868), estradiol (Catalog #: ml002891), aldosterone (Catalog #: ml002876), and cortisol (Catalog #: ml002874) were obtained from Enzyme-linked Biotechnology (Shanghai, China). Sera from 3 animals per group were used and each was diluted 10 or 20 times for immunoassays.

Seminiferous tubule inflammation test

8 testes were obtained from each sample group and then subjected to fixation in 4% formalin for at least 1 week. Formalin-fixed paraffin-embedded mouse testis sections of 5 µm thickness were used for HE Staining. At least 30 seminiferous tubules in each slide were examined for inflammation test. Testis with at least 1 inflammatory seminiferous tubule was set as 1, and normal testis was set as 0 for inflammation index calculation.

snRNA-seq data processing, batch effect correction, and cell subset annotation

Intact hypothalami were cryopreserved in liquid nitrogen from sacrificed rats, and two to three pooled intact hypothalami from each group were homogenized in 500 µL ice-cold homogenization buffer (0.25 M sucrose, 5 mM CaCl₂, 3 mM MgAc₂, 10 mM Tris-HCl [pH 8.0], 1 mM DTT, 0.1 mM EDTA, 1× protease inhibitor, and 1 U/µL RiboLock RNase inhibitor) with Dounce homogenizer. Then, the homogenizer was washed with 700 µL ice-cold nuclei washing buffer (0.04% bovine serum albumin, 0.2 U/µL RiboLock RNase Inhibitor, 500 mM mannitol, 0.1 mM phenylmethanesulfonyl fluoride protease inhibitor in 1× phosphate buffer saline). Next, the homogenates were filtered through a 70-µm cell strainer to collect the nuclear fraction. The nuclear fraction was mixed with an equal volume of 50% iodixanol and added on top of a 30% and 33% iodixanol gradient. This solution was then centrifuged for 20 min at 10 000 ×g at 4 °C. After the myelin layer was removed from the top of the gradient, the nuclei were collected from the 30% and 33% iodixanol interface. The nuclei were resuspended in nuclear wash buffer and resuspension buffer and pelleted for 5 min at 500 ×g at 4 °C. The nuclei were filtered through a 40-µm cell strainer to remove cell debris and large clumps, and the nuclear concentration was manually assessed using trypan blue counterstaining and a hemocytometer. Finally, the nuclei were adjusted to 700–1200 nuclei/µL, and examined with a 10X Chromium platform.

Reverse transcription, cDNA amplification and library preparation were performed according to the protocol from 10X Genomics and Chromium Next GEM Single Cell 3′ Reagent Kits v3.1. Library sequencing was performed on the Illumina HiSeq™ 4000 by Gene Denovo Biotechnology Co., Ltd (Guangzhou, China).

10X Genomics Cell Ranger software (version 3.1.0) was used to convert raw BCL files to FASTQ files, and for alignment and counts quantification. Reads with low-quality barcodes and UMIs were filtered out and then mapped to the reference genome. Reads uniquely mapped to the transcriptome and intersecting an exon at least 50% were considered for UMI counting. Before quantification, the UMI sequences were corrected for sequencing errors, and valid barcodes were identified using the EmptyDrops method. The cell × gene matrices were produced via UMI counting and cell barcodes calling. Cells with an unusually high number of UMIs (≥8000) or mitochondrial gene percent (≥15%) were filtered out. Batch effect correction was performed by SCTransform function built in Seurat V4.4.0.

Pathways, signatures, TFs and TF cofactors, cell communication

Gene sets and pathways were derived from Hallmark gene sets of MSigDB collections, the KEGG pathway database, Reactome pathway database, and WikiPathways database, and some ontology terms derived from the Gene Ontology (GO) resource. Mitochondrial pathways were derived from MitoCarta3.0 [18]. Pathways, gene sets, and signatures were evaluated with the PercentageFeatureSet function built into R package Seurat. TFs and TF cofactors were obtained from AnimalTFDB 3.0 [19]. TFs and TF cofactors were further filtered with mean counts >0.1.

The ligand–receptor pairs were calculated via R package CommPath [20].

Correlation analysis and ROC analysis

Pearson correlation coefficient was calculated with the linkET package (p < 0.05). Fast Wilcoxon rank sum test and auROC analysis was performed with the wilcoxauc function in R package presto. The minimal cell number in either one of the comparing pairs should be no less than 15. Ranks of AUC values were in descending order. A total of 431 pathways from Hallmark, KEGG and PID databases were used for correlation analysis with MitoCarta OXPHOS subunits in neurons and non-neural cells (Figure 2—figure supplement 1). A total of 99 pathways related to synapse activity were derived from GO, including GO cellular components, GO biological processes and GO molecular functions (Table S1).

The division of expression level-dependent clusters in each pathway and their signatures

The quarters of the mixed cell populations from O, O.T and Y hypothalamic neurons were equally divided with R function fivenum from R package stats according to pathway expression levels. The total neurons were then divided into 4 clusters (c1-c4) accordingly. The cell proportion from O.T, O and Y neurons in each cluster were weighted against the total neurons in the 3 groups. The unique markers of each cluster were calculated with FindAllMarkers function of Seurat package. The intersection of the unique markers from 6 pathways were obtained for heatmap plotting. 19 genes highly expressed in c1 were calculated as c1.up.signature via the PercentageFeatureSet function in R package Seurat. 12 genes were highly expressed in c4 and thus were calculated as c4.up.signature. There were no intersected unique markers in cluster c2 and c3 of the 6 selected pathways.

TF and pathway activities

The TF resources were derived from CollecTRI, the pathway resource was from PROGENy, and the enrichment scores of TFs and pathways were performed with the Univariate Linear Model (ulm) method according to the pipeline in R package decoupleR [21].

Subtypes of neurons generated by supervised clustering

Subtypes of neurons were clustered by neuropeptides and receptors for neuropeptides or hormones within Neurons with the subset function from R package Seurat (the target gene expression level > 0). A total of 121 neuron subtypes were obtained, comprising 80 neuropeptide-secreting neurons and 41 neurons expressing a unique neuropeptide receptor or hormone receptor. Further groupings may exist within the identified neuron subtypes, and the category of excitatory or inhibitory neuron was not discriminated further. The cell proportion of each neuron subtype was weighted according to the total number of neurons in O.T, O, and Y samples. The mean values ± standard deviation of pathways and signatures were performed for each subtype. The top 10 and the bottom 10 items were calculated, and the top 10 subtypes ranked within each of 16 pathways or signatures were collected for word clouds annotation via R package wordcloud2.

Differential expression and pathway enrichment analysis

DEGs between groups were identified via FindMarkers (test.use = bimod, min.pct = 0.1, logfc. Threshold = 0.25, avg_diff > 0.1 or < −0.1). DEGs were then enriched in redundant GO terms via WebGestalt and filtered with false discovery rate <0.05 [22].

Bidirectional Mendelian randomization (MR) study

The protein quantitative trait locus (pQTL) GWAS summary data of 204 human endocrine-related GWAS summary data with European ancestry were obtained from open-access MRC Integrative Epidemiology Unit (IEU) (Table S4) [23, 24]. Independent genome-wide significant SNPs for exposure OXT (id:prot-a-2159) or GNRH1 (id: prot-a-1233) were used as instrumental variables with genome-wide significance (P < 1 × 10⁻⁵), independence inheritance (r2 < 0.001) without linkage disequilibrium (LD) with each other for MR. For the reverse MR, independent genome-wide significant SNPs from 203 endocrine-related GWAS summary data (P < 1 × 10⁻⁵, r2 < 0.001) without LD with each other were obtained as exposures and human GWAS summary data of OXT (id:prot-a-2159) or GNRH1 (id:prot-a-1233) were used as outcomes. Weak instruments less than 10 were discarded via F-statistics.

MR and reverse MR analysis were conducted with method inverse-variance weighting (IVW), MR Egger, Weighted median, Simple mode, and Weighted mode. The screening criteria: all of the odds ratio (OR) values of the 5 methods should be simultaneously either >1 or <1 and the significant p value of IVW was <0.05. The heterogeneity via IVW method and the horizontal pleiotropy were also evaluated with R package TwoSampleMR [25].

Results

The overall changes in aged hypothalamus with or without long-term 17α-estradiol treatment via snRNA-seq profiling

To investigate the hypothalamus as a potential key target of 17α-estradiol’s effects on life extension, we performed snRNA-seq on the entire hypothalamus of aged and 17α-estradiol-treated aged Norway brown rats, and used the hypothalamus from young adult male rats as a control. We obtained 10 major cell types according to cell markers of the hypothalamus (Figure 1A-B). Notably, the proportions of non-neural cells were changed extensively in O (Figure 1C). For example, Oligo, OPC, and Micro were increased, and Astro, Tany, Fibro, PTC, and Endo were decreased in O compared with Y. The proportion of Oligo, OPC, and Micro were increased in O.T than those in Y. Endo was also increased in O.T. The proportions of Astro, Tany, Epen, and PTC were decreased more in O.T than those in O in comparison with those in Y. The results indicated that 17α-estradiol treatment had extensive effects on the proportions of non-neural cells in hypothalamus.

Figure 1.

Cell communication analysis indicated that the ligand–receptor pairs between neurons and other cell types also changed remarkably, such as those in Endo, Fibro, Tany, and Atro (Figure 1D). The significant ligand–receptor pairs among Neurons were changed in O.T and O, especially in O (Figure 1E). Notably, of the significant ligand–receptor pairs in neurons, Bmp2– Acvr1/Acvr2a/Acvr2b/Bmpr, Gdf11–Acvr2a/Acvr2b, Inhba–Acvr1/Acvr2a/Acvr2b, Nrg1/Nrg2/Nrg4–Erbb4, Rspo1–Lgr5/Lrp6, and Rspo3–Lgr5 were exclusively and significantly increased in Neuron.O, in comparison with neurons of O.T and Y, which suggested enhanced TGF superfamily-mediated signaling activity and canonical Wnt signaling during aging. The significantly-changed ligand–receptor pairs Nlgn1–Nrxn1/Nrxn2, Nlgn2–Nrxn1/Nrxn2/Nrxn3, Nlgn3–Nrxn1/Nrxn2/Nrxn3, Nxph1–Nrxn1/Nrxn2/Nrxn3, Nxph3–Nrxn1/Nrxn2/Nrxn3, Pomc– Oprd1/Oprk1/Oprm1, and Vip–Adcyap1r1/Avpr1a/Vipr2 were exclusively increased in neurons of O.T. These ligand–receptor pairs were related to or belonged to synaptic activity, cellular adhesion, opioid system, and vasodilation, which indicated unique roles of 17α-estradiol in restoring certain physiological functions in aging hypothalamus. The enhanced Pomc signal in O.T neurons is in line with prior report that 17α-estradiol treatment decreased food uptake in mice is potentially correlated with Pomc neurons [11].

Gene set enrichment analysis (GSEA) based on DEGs also verified the expression profiles in stress responses and synaptic activity in neurons (Figure 1F). ROC analysis of significantly differently expressed pathways related to neural synapse manually selected from Gene Ontology databases showed that most top-ranked pathways related to synaptic activity according to area under the curve (AUC) values were downregulated in aged neurons, and that 17α-estradiol treatment reversed this pattern (Figure 1G, Table S1).

Together, the findings suggested that 17α-estradiol extensively reshaped the cell populations, cellular communication, neuropeptide secretion, and neural synapse activity in aged hypothalamus, which was distinct from that of the young hypothalamus and aged hypothalamus.

The two opposing signaling networks in regulating metabolism and synapse activity, which can be balanced effectively by 17α-estradiol

In order to monitor the metabolism and neural status interfered by 17α-estradiol, energy metabolism pathway MitoCarta OXPHOS subunits was used to calculate the positively-or negatively-correlated pathways in hypothalamic neurons (Figure 2—figure supplement 1). Energy metabolism and synapse activity was thus revealed to be two opposing regulatory signaling networks in hypothalamic neurons and 17α-estradiol strongly balanced these two opposing signaling networks (Figure 2A). Underlying these opposing signaling pathways were two category opposing transcriptional factors (TFs) (Figure 2B). For example, Calr, Clu, Peg3, Prnp, Ndufa13, Actb, Ywhab, Nfe2l1, Mtdh, Npm1, Bex2, Aft4, and Maged1 were positively correlated with pathways in OXPHOS subunits, lysosome, protein export, mTorc1 signaling, unfolded protein response (UPR) in O, O.T and Y neurons, and were negatively correlated with pathways in ubiquitin mediated proteolysis, endocytosis, tight junction, focal adhesion, axon guidance, and MAPK signaling. Additionally, TFs Myt1l, Ctnnd2, Tenm4, Camta1, Med12l, Rere, Csrnp3, Erbb4, Jazf1, Dscam, Klf12, and Kdm4c had opposite correlation patterns with these selected pathways in O, O.T and Y neurons. These TFs may take conserved roles in regulating the two opposing biological processes in neurons of the hypothalamus.

Figure 2.

Then, we tried to establish gene signatures to represent these two opposing signaling networks to display the cell status of aging and evaluate the effect interfered by 17α-estradiol. To accomplish this, 4 quarters (c1-c4) according to the expression level of each of the 6 selected pathways from the two opposing signaling networks were evenly divided in quantity among the mixed neurons from O, O.T and Y and the shared unique markers in each quarter were calculated (Figure 2C, D). From the distribution patterns we revealed that the neuron proportion decreased in O from c1 to c4 in metabolism pathways (MitoCarta OXPHOS subunits and Hallmark mTorc1 signaling), and this trend was reversed in the opposing signaling pathways (GOBP synapse organization and KEGG Mapk signaling pathway) (Figure 2C). But in Y this trend was on the contrary, implying the expression level of 4 quarters (c1-c4) from the two opposing signaling networks can be used to monitor the aging status. 17α-estradiol treatment alleviated this trend or even reversed the trend in O. We then screened the shared unique markers of each quarter from the 6 selected pathways to try to establish the signatures representing the two opposing signaling networks. Unique markers in c1 (19 genes, c1.up.signature) and c4 (12 genes, c4.up.signature) were obtained, but c2 and c3 didn’t have unique markers shared by the 6 pathways (Figure 2D). Accordingly, the 19 genes in c1.up.signature displayed conversed correlation pattern with the 12 genes in c4.up.signature, indicating the two opposing signatures are capable of reflecting the two opposing signaling networks in hypothalamic neurons (Figure 2E). But the two opposing signaling networks balanced by 17α-estradiol in non-neural cell types were not so striking as that in neurons, displaying variable effects on non-neurons (Figure 2—figure supplement 2). Therefore, in this report we mainly focused on hypothalamic neurons and its response to aging and 17α-estradiol.

Supervised clustering revealed distinct responses of different subtypes of hypothalamic neurons to aging and 17α-estradiol

The hypothalamus has numerous neuron subtypes that release various neuropeptides and hormones to regulate fundamental body functions. To discriminate the changes during aging and the effect of 17α-estradiol on each neuron subtype, we performed supervised clustering by neuropeptides, hormones or their receptors (Figure 3A). Most of the top 10 neuron subtypes in unfolded protein response (UPR), ferroptosis, mTorc1 signaling, insulin signaling pathway, immune response pathways, OXPHOS subunits, and senescence signature during aging were attenuated by 17α-estradiol treatment (Figure 3B, Figure 3—figure supplement 1). For example, Serpine1-secreting neurons had very high expression levels of UPR, ferroptosis signature, mTorc1 signaling, Tnfa signaling via NFKB, and OXPHOS subunit during aging that were attenuated by 17α-estradiol.

Figure 3.

To further determine the most affected neurons in response to aging and 17α-estradiol, we calculated the frequency of each neuron subtype by combining the top 10 neurons in each of the 16 pathways and signatures (Figure 3C, Table S2). Serpine1-secreting neurons were enriched exclusively in aged hypothalamus, suggesting this type of neuron was an effective target of 17α-estradiol. Additionally, Mlnr, Nmb, Pomc, Ednra, Serpine3, Gast, and Pcsk6 neurons were all affected by 17α-estradiol (Figure 3D). Galp, Glp1r, Serping1, Sstr2, Sstr3, and Vip neurons were enriched in O, O.T and Y hypothalamus, suggesting a lack of effect of 17α-estradiol on these neuron types. The unique neurons in O.T (Crh, Serpinb9, Cckbr, Pth2r, Kiss1, Prpr, Hcrtr1, Npb, Nppa, Nxph3, and Npff) may be the consequence of side effects or compensatory effects by 17α-estradiol treatment.

We then evaluated the most and the least affected neurons by aging via ranking the neuron subtypes according to the levels of c1.up.signature and c4.up.signature (Figure 3E). Serpine1, Galp, Calca neurons were among the most affected neuron subtypes in aged hypothalamus which was similar to the combined ranks of several pathways (Figure 3—figure supplement 1). Npr3, Crh, Ar, Esr1, and Esr2 neurons were among the least affected neuron subtypes in aged hypothalamus (Figure 3E). The neuron subtypes ranked by c1.up.signature and c4.up.signature in O.T indicated that neurons Gast, Npb, Nppa, and Crh were affected most by the treatment of 17α-estradiol (Figure 3F). And neurons Oxtr, Glp1r, Gnrh1, and Crh also showed lowest levels of c4.up.signature, indicative of aging phenotype. Meanwhile, neurons Glp2r, Ar and Esr1 were among the top ranked neurons by the level of c4.up.signature indicating higher synapse activities and relatively less stress status in these types of neurons in O.T.

These results indicated that supervised clustering of each subtype of neurons facilitated the visualization of different subtypes of neurons in response to aging and medical treatments. Additionally, it also strongly suggested that Crh, Oxtr, Glp1r, Gnrh1, Glp2r, Ar, and Esr1 neuron subtypes were most sensitive to long-term 17α-estradiol treatment, which were associated with appetite, glucose metabolism, stress, and sex hormone secretion and signaling.

The potential side effects on hypothalamic–pituitary–adrenal (HPA) axis by long-term 17α-estradiol treatment in the males

To further investigate the potential side effect or compensatory effect of 17α-estradiol treatment, we performed stricter screening by intersection of the top 10 ranks in UPR, OXPHOS subunits, and ferroptosis signature (Figure 4A). Crh and Nppa were the only two intersected neuron subtypes with this strict screening. 17α-estradiol elevated several key metabolic pathways in Crh neurons comparing with that in Y and O (Figure 4B). 17α-estradiol treatment also increased c1-up-signature, and at the same time lowered many pathways associated with synapse activity and c4-up-signature in Crh neurons of O.T, indicative of senescent phenotype in Crh neuron. But in Nppa and Nppc neurons, decreased c1-up-signature in O.T implied lesser extent of senescent phenotype in these neurons than Crh neurons. The senescence of Crh neurons was also verified by the increased number of DEGs of mitochondria-expressed genes in O.T and lowered the number of DEGs in pathway adherens junction (Figure 4C). The status of Crh neurons in O.T may be associated with the elevated TF activities of Esr2, Usf2, Hdac5, Creb3l1, Tfam, Preb, Pou3f2, and Hoxb5 (Figure 4D).

Figure 4.

Notably, the HPA axis was altered by 17α-estradiol treatment as evidenced by the elevated cortisol in O.T comparing with O (p = 0.078) (Figure 4E). The correlation of the elevated cortisol production and the increased senescence in Crh neurons by 17α-estradiol treatment needs further investigation. Additionally, as a key knot of renin-angiotensin-aldosterone system, the significantly increased serum aldosterone in O.T and its potential role in sodium reabsorption and cardiovascular fitness also needs further detailed investigation (Figure 4E).

In summary, 17α-estradiol treatment changed the activity of HPA axis in male BN rats and also brought increased potential side effects to Crh neuron.

17α-estradiol increased Oxt neuron proportion and secretion and its possible role in mediating the effect of 17α-estradiol on endocrine system

Aging and 17α-estradiol treatment also changed the proportions of various neuron subtypes among O, O.T and Y (Figure 5A, B, Table S3). The proportions of Grp, Pmch, Npb, Serpinb9, Sstr2, Agrp, Sstr3, Mlnr, and Hcrt neurons ranked in the top 10 in O, and Oxt, Vip, Avp, Calca, Glp2r, Tacr1, Trh, Serping1, Npff, and Npy1r were in the top 10 in O.T. Galp, Calcrl, Ednra, Oxt, Serpinh1, Pomc, Cck, Crh, Tacr1, and Kiss1 neurons were the bottom 10 neurons in O, and Oxtr, Galp, Agrp, Serpinb9, Npvf, Serpinh1, Ednrb, Agt, Gipr, and Pomc neurons were the bottom 10 in O.T. Agrp, Pomc, Oxt, Oxtr, Gipr, and Glp2r neurons are well-known for their roles in regulating food intake and energy homeostasis. Agrp neurons are activated upon hunger and Pomc neurons are activated by satiety in the ARC of the hypothalamus. 17α-estradiol treatment effectively elevated the expression profiles of the c4.up.signature and synapse activities in neuron subtypes Agrp, Pomc, Oxt and Glp2r in O.T in comparison with O, which may relieve the adverse effects of reduced cell populations in Pomc and Agrp neurons in aging hypothalamus (Figure 5C), which indicated a potential role of 17α-estradiol in appetite control, as has been previously reported [11]. Additionally, 17α-estradiol treatment resulted in elevated proportions of Vip, Avp, Npff, Calca, and Tacr1 neurons, which ranked in the top 10, and decreased proportions of Agt neurons, all of which are associated with blood pressure regulation (Figure 5—figure supplement 1). Notably, the proportions of Oxt neurons and Glp2r neurons, two types of neurons with anorexigenic effects [26, 27], were increased in O.T. In addition to the increased number of Oxt-positive neurons, the expression level of Oxt was also increased in O.T. The elevated synapse activities were also evidenced by the increased DEGs in enriched pathway synaptic membrane in Oxt neurons (Figure 5D). More importantly, serum level of Oxt was significantly elevated in O.T in comparison with O (p=0.04), though its level was still lower than that of Y (Figure 5E). Unfortunately, the top TF activities in O.T and O were much different from those in Y (Figure 5F). The elevated Hopx and Xbp1 may be associated with the response to 17α-estradiol treatment.

Figure 5.

Due to the intricate regulatory networks between various endocrine factors, the elucidation of the causal effect of Oxt on other endocrine factors is thus very complicate through traditional methods. MR analysis with variant SNPs as genetic tools has much advantageous in such task. We thus performed bidirectional MR analysis of the GWAS summary data of human plasma OXT and 203 endocrine-related and hypothalamus-related factors, most of which are protein quantitative trait loci (pQTL) data from IEU (Table S4). As an exposure, OXT was revealed to be the significant causal effect on POMC/beta-endorphin (id:prot-a-2327, id:prot-a-2325), glucagon (id:prot-a-1181), GNRH1/Progonadoliberin-1 (id:prot-a-1233) and total testosterone levels (id:ebi−a−GCST90012112, id:ieu−b−4864) (Figure 5G). NPW and CBLN1 were found to be negatively associated with OXT. The significance of these significant associations was not found in the reverse MR analysis (Figure 5—figure supplement 2A, B). Unfortunately, we were unable to screen significant associations between OXT and estradiol levels (id:ebi-a-GCST90012105, id:ebi-a-GCST90020092, id:ebi-a-GCST90020091, id:ieu-b-4872, id:ieu-b-4873, id:ukb-e-30800_AFR, id:ukb-e-30800_CSA). Interestingly, QRFP, IGF1, AGRP, TAC4, GRP, CLU, BNF, PCSK7, PACAP, ANP, TAC3, CRH, INSL6, and PRL displayed significant associations with OXT in both MR and reverse MR, indicative of their complicate causal effects (Figure 5—figure supplement 2A, B). The results suggested that elevated Oxt levels induced by 17α-estradiol may have positive associations with endocrine factors governing feeding behavior, glucose metabolism, male reproduction, and sex hormones. Therefore, OXT was a potential mediator of 17α-estradiol.

17α-estradiol activated HPG axis and the elevated Gnrh also took important roles in mediating the effect of 17α-estradiol on other endocrine factors

Since Gnrh and sex hormone expressed neuron subtypes were sensitive to 17α-estradiol treatment (Figure 3F, Figure 3—figure supplement 1), we then checked their expression profile with representative pathways of two opposing signaling networks of metabolism and synapse, including c1-up-signature and c4-up-signature (Figure 6A). But both c1-up-signature and c4-up-signature were not up-regulated in Gnrh1 neuron in O.T, which was similar to Esr2 neuron. Only in Esr1 neuron, the c1-up-signature was up-regulated. Meanwhile, both Ar and Esr neurons also displayed high level of c4-up-signature, indicative of relatively healthy status. The data implied that most of the 4 types of sex hormone-related neurons didn’t experience severe aging in O.T. However, from these expression profile, it’s hard to define the exact physiological status of these subtype of neurons, especially neural endocrine activities. Therefore, we further performed enzyme immunoassays of hormones from the serum of O, O.T and Y. 17α-estradiol treatment significantly increased the plasma level of Gnrh in comparison with Y (p = 0.0099) and O (p = 0.096) (Figure 6B). More interestingly, testosterone level in serum was significantly increased in O.T in comparison with O (p = 0.018) and Y (p = 0.052). Meanwhile, the serum level of estradiol was significantly increased in O comparing with Y (p = 0.011) and significantly decreased in O.T in comparison with O (p = 0.019), suggesting 17α-estradiol treatment strongly altered the homeostasis of testosterone and estradiol. Additionally, most testes from 30-month-old male BN rats experienced severe age-related inflammation and epithelial collapse of seminiferous tubules (Figure 6C). The testes without inflammation in O.T displayed normal morphology. 17α-estradiol treatment slightly decreased the testis inflammation in O.T in comparison with that in O (p = 0.15), indicating a positive role of 17α-estradiol treatment in male reproductive system. Additionally, the elevated transcriptional factors as Sf1, Pparg, Litaf, Nupr1, Rxrg, E2f2, and Zfp42 may be involved in the transcriptional regulation by 17α-estradiol in O.T (Figure 6D). More importantly, both the activities of androgen and estrogen pathways were decreased in Gnrh1 neurons in O.T comparing with O and distinct to those in Ar, Esr1, and Esr2 neurons (Figure 6E). These signaling is important for the feedback controlling of sex hormone secretion in Gnrh neurons. This result may also reflect the strong effect of 17α-estradiol on Gnrh neurons.

Figure 6.

To decipher the potential effects of elevated serum Gnrh levels on endocrine system, we performed bidirectional MR analysis of the GWAS summary data of human GNRH1 (id: prot-a-1233) and 203 endocrine-related factors with genetic variants SNPs. The strong causal effects of GNRH1 on GAL/Galanin (id:prot−a−1166), POMC/Beta−endorphin (id:prot−a−2327, id:prot−a−2325), Adrenomedullin (id:prot−a−48), BDNF (id:prot−a−242), and LPR (id:prot−a−1724), which are involved in feeding, energy homeostasis, osmotic regulation, and neuron plasticity (Figure 6F). Notably, CRH/Corticotropin (id:prot−a−2326), PRLH/Prolactin−releasing peptide (id:prot−a−2376), NPW/Neuropeptide W (id:prot−a−2082), Glucagon (id:prot−a−1181), Chromogranin−A (id:prot−a−538) displayed bidirectional significance, indicating close and complicated causal effects between GNRH1 and these endocrine factors (Figure 6G, Figure 6—figure supplement 1A, B). These results also suggested that the role of 17α-estradiol treatment in feeding, energy homeostasis, reproduction, osmotic regulation, stress response, and neuron plasticity may be mediated at least by elevated Gnrh secretion.

Discussion

The most striking role of 17α-estradiol treatment revealed in this study showed that HPG axis was substantially improved. The underlying molecular mechanism is still unknown. But prior report indicated that 17α-estradiol is able to bind ESR1 [15]. In this report, 17α-estradiol treatment significantly decreased serum estradiol and elevated serum testosterone. Combining these information, we propose 17α-estradiol may have similar roles of estrogen antagonists or the roles of aromatase inhibitors, which may prevent the conversion of androgens to estrogens [28, 29]. These roles can remove the feedback inhibition by estrogen on hypothalamus and pituitary to secret Gnrh, FSH, and LH [30].

The testosterone levels declined gradually from the third decade of life in men [31]. Ageing of the gonadotropic axis, especially aged males with low serum testosterone, is usually associated with many aging symptoms as the age-related loss of skeletal muscle mass, muscle strength, muscle power, low bone mineral density, frailty, impaired physical performance, mobility limitation, increased risk of diabetes, higher all-cause mortality and cardiovascular mortality, cognitive decline, and increased risk of Alzheimer’s disease [32]. Therefore, testosterone supplementation to older men is benefit. Additionally, Gnrh supplement alleviates aging-impaired neurogenesis and thus decelerates aging [33]. 17α-estradiol treatment didn’t lead to feminization and didn’t affect the sperm parameter, fertility in male animals [6, 13]. Therefore, the increased Gnrh levels and serum testosterone by 17α-estradiol treatment is likely benefit to older males especially to those with syndrome of late-onset hypogonadism.

Similar aging syndrome to aged males with low serum testosterone, postmenopausal women with low estrogen also experience increased mortality, cardiovascular disease, osteoporosis fracture, urogenital atrophy and dementia, which are also benefit from hormone therapy [34]. But why 17α-estradiol treatment doesn’t display positive life extension roles in aged females according to prior report [12]? The underlying key point may be the inhibition of estrogens by 17α-estradiol treatment. This is also evidenced by the inability of 17α-estradiol to improve female fertility [35]. However, due to lacking parallel data in aged female BN rats treated with 17α-estradiol, the exact answer to this question needs more investigation in the females in the future.

Another striking effect of 17α-estradiol to hypothalamus neurons was to decrease the overall energy metabolism in aged male BN rats. The nutrient-sensing network, mediated by MTORC1 complex, is a central regulator of mRNA and ribosome biogenesis, protein synthesis, glucose, autophagy, lipid metabolism, mitochondrial biosynthesis, and proteasomal activity [36]. Downregulation of the nutrient-sensing network can increase lifespan and healthspan [37]. 17α-estradiol treatment attenuated nutrient-sensing network activity in most hypothalamic neurons, which promoted lifespan extension. Additionally, synaptic aging leads to a decline of synaptic activity and estrogen improves the performance of aging synapses [38]. In line with this, our data showed that the benefit of improved synaptic activity by 17α-estradiol treatment was prominent in aging male hypothalamus.

In this report, we showed that neuron populations in appetite control, specifically Agrp, Pomc, Oxt, Oxtr, Gipr, and Glp2r neurons, were extensively changed. The Oxt neuron proportion was elevated the most by 17α-estradiol in all detected neuron subtypes (Figure 5A). Oxt plays versatile roles in social behavior, stress response, satiety, energy balance, reproduction, and inflammation [39]. Most Oxt neurons are produced by the PVN and supraoptic nuclei (SON) in hypothalamus with very high plasticity during development and intricate circuits [40, 41]. PVN, ARC, and ventromedial hypothalamic nucleus compose the neural hub in the hypothalamus that integrates peripheral, nutritional, and metabolic signals to control food intake behavior and energy balance [42]. Many effects of Oxt are sex-specific [43]. For example, females are less sensitive to exogenous Oxt than males in social recognition [44]. More interestingly, Oxt injection, which was delivered via improved blood–brain barrier penetration nanoparticles, decreased body mass, and increased social investigation and Oxt-positive cell number in the SON of hypothalamus preferentially in male rats [45]. Intracerebroventricular Oxt injection in rats showed that food intake was reduced in both sexes with a more pronounced effect in males [46].Therefore, we propose that the role of Oxt in systemic aging and feeding behavior is another potential contributing factor in mediating the sex-biased effects of 17α-estradiol.

17α-estradiol treatment apparently induced elevated aging of HPA axis. One evidence was the increased senescence of Crh neurons. The other evidence was the elevated serum cortisol, which is a hallmark of aging HPA axis [47, 48]. Therefore, more attentions should be paid to the potential side effects of 17α-estradiol especially in its clinical application.

Together, the findings suggest that HPG axis and neurons involved in appetite and energy balance were improved by 17α-estradiol treatment, which may be closely related to the life-extension roles of 17α-estradiol in aged males. Additionally, supervised clustering by neuropeptides, hormones and their receptors is proved to be a useful strategy to investigate the pharmacological, pathological and physiological processes in each subtype of neurons in the hypothalamus.

Abbreviations

Ar: androgen receptor; ARC: arcuate nucleus; Astro: astrocyte; AUC: area under the curve; Cga: glycoprotein hormones, alpha Polypeptide; Crh: corticotropin releasing hormone; DEG: differentially-expressed gene; Endo: endothelial cell; Esr1: estrogen receptor 1; Esr2: estrogen receptor 2; Fibro: fibroblast; Glp2r: glucagon-like peptide 2 receptor; GnRH: gonadotropin releasing hormone; GSEA: gene set enrichment analysis; HPA: hypothalamic–pituitary–adrenal; HPG: hypothalamic-pituitary-gonadal; IEU: MRC Integrative Epidemiology Unit; IVW: inverse-variance weighting; KEGG: Kyoto Encyclopedia of Genes and Genomes; Micro: microglia; MR: Mendelian randomization; MTORC1: mechanistic target of rapamycin kinase 1; Nppa: natriuretic peptide A; Oligo: oligodendrocyte; OPC: oligodendrocyte precursor cell; OR: odds ratio; OXPHOS: oxidative phosphorylation; PID: Pathway Interaction Database; POMC: proopiomelanocortin; pQTL: protein quantitative trait loci; PTC: pars tuberalis cell; PVN: paraventricular nucleus; ROC: receiver operating characteristics; snRNA-seq: single-nucleus transcriptomic sequencing; Tany: tanycyte; TF: transcription factor; UPR: unfolded protein response.

Acknowledgements

We appreciate Dr. Qinghua Wang and Hongyun Shi from animal facility of Nantong University in helping with the animal experiments. Special thanks to Professor Ken-ichiro Fukuchi from University of Illinois College of Medicine for constructive comments and suggestions in manuscript preparation.

Funding

This work was supported by the National Natural Science Foundation of China grant 31271448 (YL), 82171621 (LL), 82172566 (ZY), 82150107 (RL) and the National High Level Hospital Clinical Research Funding (2022-PUMCH-A-231) (LL).

Author contributions

Conceptualization, YL, LL; Methodology, YL, GW, XX; Animal operation: YL, LL; EIA assays: YL, LY; Inflammation test: YL, ZY, LL, YS. Validation, LL, JY, ZY; Formal Analysis, YL, JY; Resources, GW, YS, ZY, ZM, LX; Visualization, YL; Writing – Original Draft, YL, JY; Writing – Review & Editing, LL, YS; Supervision, YS, ZY, YL; Project Administration, YL, ZY, LL; Funding Acquisition, ZY, LL. All authors have read and approved the final manuscript.

Declaration of competing interest

The authors have no conflict of interests to declare.

Availability of data and materials

All data are available at GEO accession number GSE248413.

Competing interests

All authors declare no competing interests.