Experiments were approved and performed in accordance with the guidelines of the University of California, San Francisco (UCSF) Institutional Animal Care Committee (IACUC) or the University of California, Davis (UCD) Animal Ethics Committee, the National Institutes of Health Guide for Care and Use of Laboratory Animals, and the recommendations of the International Association for the Study of Pain. The animals used in this study ranged from 1 to 70 weeks of age and included both male and female mice.


The origin of the Esr1fl/fl allele (official allele: Esr1tm1Sakh) on a 129P2 background and used to generate Esr1Nkx2.1-cre mice have been previously described2 and were maintained on a CD-1;129P2 mixed background. Primer sequences used for genotyping are listed in Supplementary Table 1. Esr1Nkx2.1-cre-CAG-Luc-GFP mice were generated by crossing male mice harbouring the CAG-Luc-GFP allele (official allele: L2G85Chco/J) to female mice homozygous for the Esr1fl/fl allele, followed by an additional cross to generate a Esr1fl/fl;Nkx2.1-cre;Luc-GFP colony, which was maintained on a mixed FVB/N, CD-1, 129P2 and C57BL/6 genetic background. Esr1Prodynorphin-cre mice were generated by crossing homozygous Esr1fl/fl females to Prodynorphin-cre (B6;129S-Pdyntm1.1(cre)Mjkr/LowlJ, purchased from Jackson Laboratory) males. Mice were maintained on a 12-h light–dark cycle with ad libitum access to a standard breeder chow diet (PicoLab 5058; LabDiet, 4 kcal% fat, 0.8% Ca2+) and sterile water and housed under controlled and monitored rooms for temperature and humidity. Eighteen-month-old and 24-month-old C57BL/6 female and male mice were obtained through the NIA Aged Rodent Colony Program, available to NIA NIA-funded projects. To create cohorts of OVX females, ovariectomy was performed at 4 months of age, followed by 4 weeks of surgical recovery. All animal procedures were performed in accordance with UCSF and UCD institutional guidelines under the Ingraham and Ambrosi laboratories’ Institutional Animal Care Committee protocol of record.


Parabiosis surgery followed previously described procedures51. In brief, 6-week-old Esr1fl/fl and Esr1Nkx2.1-cre females underwent mirror-image incisions at the left and right flanks through the skin, and incisions were made through the abdominal wall. The peritoneal openings of the adjacent parabionts were sutured together. Additionally, elbow and knee joints from each parabiont were sutured together. The skin incision of each mouse was then stapled. Each mouse was subcutaneously injected with enrofloxacin (Baytril, Bayrer) antibiotic and buprenorphine (Butler Schein) and monitored during surgical recovery. To monitor the health of pairs, their body weights and grooming behaviours were monitored weekly. The pairing was checked by the presence of Evans Blue dye in blood collected 2 h after injection of 200 µl of 0.5% Evans Blue from the submandibular of the uninjected parabiont.

In vivo µCT was performed to determine changes in trabecular bone mass over time using a Scanco Viva CT40 high-speed µCT preclinical scanner. Before scanning, mice were anaesthetized and placed in a 3D-printed nose cone to accommodate surgically paired mice. Distal femurs of each mouse in the parabiont group (opposite to the incision side) were imaged 2 weeks after surgery at baseline (0) and then at 3, 6 and 17 weeks from baseline imaging at intervals that preserve bone mass52. Parabionts were anaesthetized, and a 2 mm region of the contralateral distal femur (opposite to the surgically paired side) was used for assessing the trabecular bone compartment of 1 mm length proximal to the epiphyseal plate and cortical parameters at the diaphysis in an adjacent 0.4 mm region of the femur. Imaging was performed at the UCSF Skeletal Biology and Biomechanics Core supported by the NIAMS Award P30AR075055.

Bone, ocSSC kidney and brain transplants

In brief, long bones were dissected from 4-week-old female or male Esr1fl/fl mice, cleaned of excess tissue and immediately implanted under the skin after creating a horizontal incision and small pocket posterior to the scapula in 8-week-old acceptor Esr1fl/fl or Esr1Nkx2.1-cre female mice. After 6 weeks of incubation, donor femurs were subsequently removed and analysed by ex vivo µCT. In brief, volumetric bone density and volume were measured at the distal femur using a Scanco Medical µCT 50 specimen scanner calibrated to a hydroxyapatite phantom or Bruker SkyScan1276 (Bruker Preclinical Imaging). For Scanco imaging, samples were fixed in 10% phosphate-buffered formalin and scanned in 70% ethanol using a voxel size of 10 mm and an X-ray tube potential of 55 kVp and X-ray intensity of 109 µA. Scanned regions included a 2 mm region of the femur proximal to the epiphyseal plate. For Bruker Skyscan 1276 imaging, a source voltage of 85 kV, a source current of 200 µA, a filter setting of AI 1 mm and a pixel size of 12–20 µm (a set number was used for samples of a specific experiment) at 2,016 × 1,344 were used. Reconstructed samples were analysed using CT Analyser and CTvox software (Bruker). Standard best practices were used to quantify trabecular and cortical bone parameters52. Image acquisition of an implanted femur was captured using an iPhone 13 Pro and then edited in Photoshop CC.

For kidney transplant assays, primary ocSSCs were isolated as described below from 4-week-old Esr1fl/fl female mice. Approximately 20,000 live cells were resuspended in 2 μl Matrigel (Corning, 356234), and the entire mixture was injected into the renal capsule of 8-week-old recipient mice. Six weeks after transplantation, animals were then euthanized and kidney grafts were processed as described below. Excised kidneys with renal capsule grafts and brain tissue with injected stem cell grafts were scanned using the same settings. For the processing of kidneys after SSC transplants, kidneys were dissected out and cleaned of soft tissue, fixed in 4% paraformaldehyde (PFA) and embedded in OCT for cryosectioning. Sections (5 µm) were subsequently stained using a standard Movat’s pentachrome staining kit (Abcam, ab245884). Bright-field images were taken using a Luminera Infinity-3 and quantified using ImageJ software. Excised kidneys containing renal capsule grafts injected with SSCs were scanned using the same settings as above for bone transplant assays.

For brain transplants, Esr1fl/fl-CAG-Luc-GFP ocSSCs were isolated by FACS as described below and kept on ice in sterile artificial cerebral spinal fluid (Tocris Bioscience, 3525). Approximately 2 h after cell isolation, 400–700 ocSSCs in 1 µl of solution were delivered bilaterally above the third ventricle by the ARC nucleus by stereotaxic injection (anterior–posterior (AP): −1.58; medial–lateral (ML): ±0.3; dorsal–ventral (DV): −5.95 from the skull) into 12–18-week-old Esr1fl/fl or Esr1Nkx2.1-cre females. Six weeks after implant, mice were perfused, and immunohistochemistry was performed on cryosections (20 µm) collected from brains fixed in 4% PFA using standard procedures. GFP staining used a polyclonal chicken anti-GFP antibody (Novus Biologicals, NB100-1614) at 1:2,500. Images were taken using a Keyence B2-X800. Excised medial basal hypothalamic brain tissues from female mice injected with stem cells were scanned using the same settings as described above for bone transplant assays.

Flow cytometry isolation of primary SSCs

Flow cytometry and cell sorting were performed on a FACS Aria II cell sorter (BD Biosciences) and analysed using FlowJo software. Mouse long bones and callus samples from human patients (UC Davis Institutional Review Board (IRB) 1997852; SCRO-1199) were dissected and freed from the surrounding soft tissue, which was then followed by dissociation with mechanical and enzymatic steps as previously described53,54. For isolation of human SSCs, six samples of callus tissue from individuals (age range 14–72  years) with fractured bones were acquired during the process of open reduction internal fixation of fractures at UCD Medical Center. Collection adhered to IRB guidelines (IRB 1997852-3, without restrictions based on the race, sex or age of the donors of the specimens). The isolation of SSCs performed in this study does not meet the criteria for human research according to the IRB assessment. Hence, informed consent was not sought. Only fractures treated with an open approach and direct realignment of fracture fragments were included, and haematoma or callus tissue hindering satisfactory realignment of displaced fragments was excised and preserved for research purposes during the surgery. Excised tissues were promptly placed on ice, and human SSCs were isolated within 5 h after surgery, as detailed below.

In brief, the tissue was placed in collagenase digestion buffer supplemented with DNase and incubated at 37 °C for 60 min under constant agitation. After collagenase digestion and neutralization, undigested materials were gently triturated by repeated pipetting. Total dissociated cells were filtered through a 70-µm nylon mesh and pelleted at 200g at 4 °C for 5 min. Cells were resuspended in ammonium–chloride–potassium lysing buffer to eliminate red blood cells and centrifuged at 200g at 4 °C for 5 min. The pellet was resuspended in 100 µl staining medium (2% FBS/PBS) and stained with antibodies for at least 30 min at 4 °C (antibody information can be found in Supplementary Table 2). Living cells were gated for lack of propidium iodide (1:1,000 diluted stock solution: 1 μg ml–1 in water; mouse cells) signal or DAPI (human cells). Compensation, fluorescence-minus-one control-based gating and FACS isolation were conducted before analysis or sorting using established antibody cocktail combinations. A complete list of antibodies used for FACS purification of SSCs is presented in Supplementary Table 2.

For mouse SSC lineages, the following antibodies were used: CD90.1 (Thermo Fisher, 47–0900), CD90.2 (Thermo Fisher, 47–0902), CD105 (Thermo Fisher, 13–1051), CD51 (BD Biosciences, 551187), CD200 (Thermo Fisher, MA5-17980), CD45 (BioLegend, 103110), Ter119 (Thermo Fisher, 15–5921), Tie2 (Thermo Fisher, 14–5987), 6C3 (BioLegend, 108312), streptavidin PE-Cy7 (Thermo Fisher, 25–4317), Sca-1 (Thermo Fisher, 56-5981), CD45 (Thermo Fisher, 11–0451), CD31 (Thermo Fisher, 12-0311), CD140a (Thermo Fisher, 17–1401) and CD24 (Thermo Fisher, 47–0242).

For human SSC isolation, the following antibodies were used: CD45 (BioLegend, 304029), CD235a (BioLegend, 306612), CD31 (Thermo Fisher Scientific, 13-0319), CD202b (TIE-2) (BioLegend, 334204), streptavidin APC-AlexaFlour750 (Thermo Fisher, SA1027), CD146 (BioLegend, 342010), PDPN (Thermo Fisher Scientific, 17-9381), CD164 (BioLegend, 324808) and CD73 (BioLegend, 344016).

Cell culturing and differentiation assays of primary mouse and human SSCs

Only freshly sorted primary mouse or human ocSSCs were used in this study. After cell isolation by FACS, primary cells were cultured as described above. Mouse cells were cultured in minimum essential medium-α (MEMα) with 10% FBS and 1% penicillin–streptomycin (Thermo Fisher, 15140–122) and maintained in an incubator at 37 °C with 5% CO2. Human cells were cultured in MEMα (Fisher Scientific, 12561-056) with 10% human platelet-derived lysate (Stem Cell Technologies, 06960) and 1% penicillin–streptomycin solution (Thermo Fisher Scientific, 15140-122). To induce osteogenic differentiation, pre-confluent cells were supplemented with osteogenesis-inducing factors, 100 nM dexamethasone, 0.2 mM l-ascorbic acid 2-phosphate and 10 mM β-glycerophosphate for 14 days.

For testing of candidate factors (mCCN3 (Novus Biologicals, NBP2-35100), hCCN3 (Novus Biologicals, NBP2-35084), mFST (Novus Biologicals, NBP2762685U), hFST (Stem Cell Technologies, 50-197-6487), BAM-22P (Sigma, SCP0057), met-ENK (Sigma, M6638) and hGRP (RayBiotech, 230-00695-10)), indicated concentrations were added to defined medium and changed every second day with fresh medium. Cells were then formalin-fixed and stained with 2% Alizarin Red S (Roth) in distilled water. Wells were washed twice with PBS and once with distilled water. Oil Red O staining was performed by fixing cells with 4% PFA for 15 min at room temperature using an Oil Red O working solution prepared from a 0.5% stock solution in isopropanol and diluted with distilled water at a ratio of 3:2. The working solution was filtered and applied to fixed cells for at least 1 h at room temperature. Cells were washed four times with tap water before evaluation. CFU-F assays were conducted by freshly sorting a defined number of cells of desired cell populations into separate culture dishes containing expansion medium. The medium was changed twice a week. Cells were fixed and stained with crystal violet (Sigma) on day 10 of culturing.

For in vitro osteoclastogenesis assays, bone marrow macrophages were isolated from 3-month-old and 24-month-old C57Bl/6 male mice as previously described55. For osteoclast generation, cells were cultured with 30 ng ml–1 M-CSF and 10 ng ml–1 Rankl (R&D systems) treated with and without recombinant mCCN3 (0.25 M and 2.5 M) (R&D Systems, 1976-NV-050) for 4 days with medium exchange performed daily. On day 4, cells were fixed with 4% PFA, and TRAP staining was performed according to the manufacturer’s protocol (Sigma-Aldrich). TRAP-positive cells with two or more nuclei per well were counted. Bone marrow macrophages were obtained from four mice per treatment group.

HFD, LCD challenges and metabolic parameters

HFD was purchased from Research Diets (D12492, 60 kcal% fat, 0.78% Ca2+). Esr1fl/fl and Esr1Nkx2.1-cre mice were maintained on a HFD for 17 weeks starting from 10 weeks of age. Glucose tolerance tests were conducted after a 6-h fast with glucose administered (i.p., 1.0 g kg–1of body mass). For glucose tolerance tests, mice were subjected to 6 h of fasting (starting at about ZT2) and injected with glucose (i.p., 1 g kg–1). Tail-blood samples were collected at baseline at 15, 30, 45, 90 and 120 min after glucose injection. Blood glucose levels were quantified using a hand-held glucometer (Roche, Accu-Check Compact). For non-fasting triglyceride measurements, whole blood was collected from 27-week-old Esr1fl/fl and Esr1Nkx2.1-cre mice into EDTA-treated tubes (Microvette CB 300 K2E) and placed directly on ice. To isolate plasma, whole blood was spun down at 2,000g for 15 min at 4 °C, and the supernatant was collected. Non-fasting plasma triglycerides levels were then measured using a commercially available kit (Cayman Chemicals, 10010303) as per the manufacturer’s protocols. All plasma samples were stored at −80 °C before analysis. Body composition to determine per cent lean and fat mass was obtained by dual-energy X-ray (DEXA, GE Lunar PIXImus). LCD was purchased from Teklad (TD.95027, 14.7 kcal% fat, 0.01% Ca2+). For lactation studies, a subset of Esr1fl/fl females were switched from standard breeder chow after parturition and maintained on LCD for 12 days before collection of long bones and quantification of bone using parameters described above for Bruker Skyscan 1276 imaging.

Bone parameters

Volumetric bone density and BV for mice fed SD and HFD were measured at the right femur using a Scanco Medical µCT 50 specimen scanner calibrated to a hydroxyapatite phantom. In brief, samples were fixed in 10% phosphate-buffered formalin and scanned in 70% ethanol. Scanning was performed using a voxel size of 10 mm and an X-ray tube potential of 55 kVp with an X-ray intensity of 109 µA. Scanned regions included a 2 mm region of the femur proximal to the epiphyseal plate and a 1 mm region of the femoral mid-diaphysis. Scanned femurs were performed with 10 µm resolution at 70 kV, 57 µA, 4 W and an integration time of 700 ms. The analysis threshold for cortical and trabecular bone was 0.8 sigma, 1 support and 260 (lower) and 1,000 (upper) permille. Volumes of interest were evaluated using Scanco evaluation software. Representative 3D images were created using Scanco Medical mCT Ray (v.4.0) software.

Bone histomorphometry analysis

Mice were injected with 20 mg kg–1 calcein (Sigma-Aldrich) 7–9 days before euthanasia and with 15 mg kg–1 of Alizarin (Sigma-Aldrich) 2 days before euthanasia. Bones were fixed in 4% PFA, dehydrated in 30% sucrose and embedded in OCT or embedded in MMA plastic. Standard undecalcified sections (5 mm) were cut using a microtome (Leica CM1950) together with the CryoJane Tape-Transfer System. Mounted sections were imaged with an ECHO REVOLVE R4 using FITC (Calcein) and Texas Red (Alizarin Red) channels. A standard sampling site with an area of 2.5 mm2 was established in the secondary spongiosa of the distal metaphysis. Before histomorphometry analyses, mosaic-tiled images of distal femurs were acquired at ×20 magnification with a Zeiss Axioplan Imager M1 microscope (Carl Zeiss MicroImaging) fitted with a motorized stage. The tiled images were stitched and converted to a single image using Axiovision software (Carl Zeiss MicroImaging) before blinded analyses was performed using two image-analysis software programs: Bioquant OSTEO or ImageJ. The following variables were analysed: %BV/TV, mineral apposition rate, mineral surface/bone surface (MS/BS), bone formation rate/bone surface (BFR/BS), osteoblast number/bone surface (No. Ob/BS), lacunar density (N Lacunae/BA), TRAP-positive osteocytes (%), and osteoclast number/bone surface (No. Oc/BS). Adjacent sections were stained with Alizarin Red for overview bright-field images.

Haematoxylin and eosin, TRAP, silver nitrate staining and osmium staining for BMAT quantification

Femoral or tibial samples were fixed in 4% PFA and demineralized in 10% EDTA for 10–14 days before being embedded in MMA plastic. Sections (5 µm) were cut using a Leica RM2165 and subsequently stained with haematoxylin and eosin (H&E) or stained with TRAP. Photoshop software removed the background in non-tissue areas for images of the proximal tibias. Silver nitrate staining of cortical osteocytes was conducted on sections of undecalcified frozen femurs. In brief, after removal of OCT, sections were incubated in 10% EDTA for 1 h at room temperature. Subsequently, slides with sections were incubated in a silver nitrate–gelatin solution, mixed using 2 parts of 50% w/v silver nitrate and 1 part 2% gelatin in 1% formic acid, at room temperature for 55 min, followed by a 2 min water wash. Sections were then incubated in 5% sodium thiosulfate for 10 min, followed by another 2 min water wash step. Dehydrated sections, protected with a cover slip, were imaged. Images were used for counting osteocytes in defined cortical areas.

Quantification of BMAT followed a published protocol32. In brief, femurs were decalcified in 14% EDTA, pH 7.4 for 2 weeks, followed by incubation with a PBS solution containing 1% osmium tetroxide (Electron Microscopy Sciences 19170) and 2.5% potassium dichromate (Sigma-Aldrich 24–4520) for 48 h. After washing for 2 h with water, osmium-stained bones were embedded in 2% agarose before scanning at 10 µm voxel resolution with a Scanco µCT 40 scanner. Regions of interest were contoured and analysed with a threshold of 400 for BMAT quantification. Specifically, a region of 2 mm immediately above the growth plate in distal metaphysis was used for the quantification of regulated BMAT in femurs.

Biomechanical strength testing and stabilized bicortical femoral fracture model

Femurs underwent a three-point bend test using mechanical load frames (Instron E100or EnduraTEC, ELF3230). A span of 7 mm separated the lower supports to support two ends of the specimen. The testing head was aligned at the midpoint between the supports. Femurs were preloaded to a force of 1 N and then loaded at a rate of 0.1 mm s–1. Loading was terminated after mechanical failure, determined by a drop in force to 0.5 N. Force displacement data were collected every 0.01 s. All tests were performed at room temperature using an electromechanical load frame as specified above.

Femurs of anaesthetized mice were exposed following muscle distraction and lateral dislocation of the patella. A 25-gauge needle was inserted between the femoral condyles to provide relative intramedullary fixation before creating a transverse, mid-diaphysis fracture using micro-scissors. Hydrogels were then immediately placed at the fracture site. The patella was relocated, and 6-0 nylon suture (Ethicon) was used to re-approximate the muscles. Mice were euthanized 21 days after surgery, femurs with fracture calluses were dissected and intramedullary pins removed for subsequent analyses. For hydrogel fabrication, eight-arm poly(ethylene glycol) vinyl sulfone (PEG-VS) (10 kDa) (JenKem) was dissolved in HEPES (25 mM, pH 7.2) at a 2× concentration. Murine recombinant CCN3 (R&D Systems) was added to the precursor solution to reach a final concentration of 1 or 2 μg per hydrogel. Hydrogels without growth factors served as controls and were loaded with HEPES of equal volume. GPQ-A (GCRDGPQGIAGQDRCG, GenScript), a protease-cleavable crosslinking peptide, was mixed at a 1:2 volume ratio with PEG-dithiol (PEG-DT) (3.5 kDa) (JenKem) at a 2× concentration in medium (pH 8.3) to permit matrix metalloproteinase-mediated degradation56. The precursor solutions were mixed at a 1:1 volume ratio and pipetted onto a silicone mould for a final volume of 6 μl per hydrogel. Fracture calluses were scanned using a SkyScan1276 (Bruker Preclinical Imaging) with settings described above and analysed by selecting 50 sections in both directions of the fracture site, producing a total area of 100 sections. For analysis of callus mineralization, CTAn software was used to select a region of interest spanning the fracture callus area outside the intramedullary space, excluding cortical bone tissue.

Plasma collection and whole bone assays

In brief, 300 µl of whole blood was collected from the submandibular vein from pre-pubertal Esr1fl/fl and Esr1Nkx2.1-cre female mice into EDTA-treated tubes (Microvette CB 300 K2E) and placed directly on ice. To isolate plasma, whole blood was spun down at 2,000g for 15 min at 4 °C, and the supernatant was collected. The right and left femurs of 10–11-week-old control female and male mice (Esr1fl/fl) were collected and cleaned of soft tissue. Femurs collected from 18-month-old, aged female mice (strain C57BL/6) provided by the National Institutes of Aging were also tested in whole bone assays. The left femur was immediately fixed in 4% PFA and then transferred into PBS at 4 °C for histological assessment to obtain baseline measurements. The right femur was cultured in a 12-well plate containing 1.4 ml primary culture medium (α-MEM; containing l-glutamine and nucleosides; Mediatech), supplemented with 10% FBS (Atlanta Biologicals) and 100 U ml–1 penicillin–streptomycin (Mediatech). The left and right femurs were treated with 15 µl of plasma from Esr1fl/fl and Esr1Nkx2.1-cre females, respectively, or with mCCN3 (14 µl of 0.0125 µg µl–1 of recombinant mCCN3, 1976-NV-050, R&D Systems in 1.4 ml of primary culture medium) or with vehicle (14 µl of 0.9% normal saline in 1.4 ml of primary culture medium), respectively. To assess the degradation of whole bone during culturing, the right tibia or femur was cultured in medium with 14 µl of 0.9% normal saline for 5 days (saline). These were compared with the baseline contralateral femur, which were immediately chilled and fixed in 4% PFA for analysis (baseline). Medium changes, including plasma, mCCN3 or vehicle treatments, were performed daily. Femurs were collected after 5 days of culture, fixed in 4% PFA and then transferred into PBS at 4 °C before µCT imaging. After µCT imaging, femurs were processed for histology as described below.

Femoral samples were cleaned of soft tissue, fixed in 4% PFA and demineralized in 10% EDTA for 10–14 days before embedding in paraffin wax. Sections measuring 5 µm were then cut using a Leica RM2165 and subsequently stained with a Movat’s pentachrome staining kit (Abcam, ab245884). For H&E staining analysis, femurs were collected, fixed in 4% formalin, decalcified in Cal-Rite, dehydrated in 30% sucrose and embedded in OCT. Then, 5 µm standard sections were cut using a microtome (Leica CM1950).

Western blotting

Hepatic protein lysates were prepared as previously described57. Plasma was isolated as described above, and CCN3 protein was enriched by heparin–agarose affinity purification58. Heparin–agarose beads (Sigma, H6508; 200 µl per sample) were washed and equilibrated in PBS with protease inhibitors (Thermo, 78425), mixed with a volume of plasma equivalent to 1 mg of total protein, and incubated overnight at 4 °C with constant rotation. Beads were then washed four times with PBS, and proteins were eluted by boiling the beads for 10 min in Laemmli sample buffer (Bio-Rad, 1610747) containing 50 mM dithiothreitol. Hepatic (10 µg) and affinity-purified plasma proteins were separated by SDS–PAGE and transferred to nitrocellulose membranes (Bio-Rad, 170-4270). Protein loading levels were assessed by Ponceau S staining (Thermo, A40000279). Membranes were de-stained and blocked in TBS-T (0.1% Tween 20) with 5% normal donkey serum (Abcam, AB7475). Blots were then probed overnight at 4 °C with anti-CCN3 antibody (R&D Systems, AF1976; 1:3,000) in TBS-T with 5% serum. After washing in TBS-T, blots were incubated with HRP-conjugated secondary antibody (Invitrogen, A15999; 1:30,000) for 1 h at room temperature, washed in TBS-T, incubated with chemiluminescent substrate (Thermo, 34577) and imaged (Azure Biosystems).

Brain RNAscope and immunohistochemistry

Fluorescent immunohistochemistry was performed using RNAscope (ACD, Multiplex Fluorescent V2) according to the manufacturer’s protocol using the following probes: Ccn3 (ACD, 415341-C2), Esr1 (ACD, 478201), Penk (ACD, 318761) and Kiss1 (ACD, 500141-C1).

Immunohistochemistry was performed using primary antibodies against ERα (EMD Millipore, C1355 polyclonal rabbit, 1:750 dilution), CCN3 (R&D Systems, AF1976 polyclonal goat, 1:1,000 dilution), VIM (Abcam, AB92547-1001 monoclonal rabbit, 1:1,000) and KISS1 (Abcam, ab19028 polyclonal rabbit, 1:200 dilution) diluted in PBS with 0.1% Triton-X100, 5% normal donkey serum and 5% BSA. For detection, sections were labelled with species-appropriate secondary Alexa Fluor-coupled antibodies (Invitrogen, A-21447, A10042 or A-11055; 1:1,000 dilution). Slides were imaged using a Keyence BZ-X800 wide-field fluorescence microscope. Confocal images were acquired at the UCSF Nikon Imaging Center using a Nikon CSU-22 with an EMCCD camera and MicroManager (v.2.0gamma). Images were processed and quantified using ImageJ Fiji (v.1.52i) and the Cell Counter plugin (v.2). Three representative views of each sample were selected. A complete list of all antibodies used in immunohistochemistry analyses is listed in Supplementary Table 2. Cryosections (20 µm) collected from brains fixed in 4% PFA were used for both fluorescent immunohistochemistry and immunostaining.

siRNA and shRNA studies

Mice were secured in a stereotaxic frame (Model 1900, David Kopff Instruments), and 400 nl of Ccn3 or non-targeting siRNA pools (Dharmacon, E-040684-00-0010 or D-001810-10-05, 0.4 mM) were injected bilaterally to the ARC at the following coordinates: AP: Bregma −1.58 mm, ML: Bregma ±0.25 mm, DV: skull −5.9 mm. For shRNA studies, female mice were injected bilaterally (200 nl per side, 2.53 × 1013 GC ml−1) and allowed to recover for 10–14 days before mating with male mice. At 12 days after injection (siRNA) or 12 DPP (shRNA), female mice were euthanized, and brain and bone samples were collected and processed as described above. Isolated femurs were then imaged by µCT as described above for bone transplant assays. Owing to the high bone phenotype of mutant female mice, thresholding and region of interest selection were adjusted between different experiments but kept consistent within each individual experiment.

CCN3, S961 and naloxone in vivo treatments

For recombinant CCN3, 10-week-old control female and 13-week-old control male mice (Esr1fl/fl) were injected daily with recombinant mCCN3 (i.p., 7.48 µg kg–1; R&D Systems, 1976-NV-050) or vehicle control (i.p., 0.9% normal saline) for 21 days. Seven days before euthanasia, mice were first injected with 20 mg kg–1 calcein (Sigma-Aldrich) and then 15 m kg–1 of Alizarin (Sigma-Aldrich) 2 days before euthanasia. Right femurs were cleaned of soft tissue, fixed in 4% PFA and then stored in PBS at 4 °C before imaging. Isolated femurs were imaged by µCT scanning as described above for bone transplant assays. After imaging, femurs were then processed for H&E histology and dynamic histomorphometry as described above. For S961 treatment, 10-week-old Esr1fl/fl and Esr1Nkx2.1-cre female mice were infused with continuous S961 (70 nM per osmotic pump), an insulin peptide receptor antagonist, as adapted from a previous study31 and obtained from the Novo Nordisk Compound Sharing Program (NNC0069-0961) or vehicle solution through an ALZET mini osmotic pump (Model 1004), which was exchanged once after 4 weeks for a total infusion period of 8 weeks. S961 was reconstituted in 0.9% normal saline. The mini osmotic pumps were implanted into an interscapular subcutaneous pocket under isoflurane anaesthesia and exchanged once. For naloxone treatments, 10-week-old Esr1fl/fl and Esr1Nkx2.1-cre female mice were infused with continuous naloxone, a non-selective opioid antagonist (0.5 mg per 24 h over 4 weeks, Tocris, 0599/100) or vehicle solution through an ALZET mini osmotic pump (Model 1004). Naloxone was reconstituted in the vehicle solution containing 0.9% normal saline. The mini osmotic pump was implanted into an interscapular subcutaneous pocket under isoflurane anaesthesia.

Hepatic viral transduction of mCCN3

Ectopic hepatocyte expression of mCCN3 protein was achieved using AAVdj viral vectors with high liver tropism59 encoding mouse CCN3 under the control of the constitutive cytomegalovirus immediate-early enhancer/chicken β-actin promoter (CAG) promoter (Vector Biolabs, AAV-265951). Viral vectors were first diluted in sterile saline, and control (Esr1fl/fl) mice were injected retro-orbitally with 100 µl of 0.5 × 1010, 3 × 1010 or 15  ×1010 GC per mouse titres of AAVdj-CAG-CCN3 or the negative control. After 2 weeks, one animal from each group was euthanized, and its liver was checked for expression of CCN3 using an anti-CCN3 antibody as listed in Supplementary Table 2. After 5 weeks, mice were euthanized, and bone, plasma and liver were collected. Liver samples were divided and processed separately for analysis of Ccn3 mRNA expression by qPCR or CCN3 protein expression by immunohistochemistry. Total liver RNA was isolated by phenol–chloroform extraction and purified using a RNeasy Mini kit (Qiagen, 74104). qPCR was performed as described below. For immunohistochemistry, liver samples were drop-fixed in 4% PFA, cryosectioned (10 µm) and antibody stained as described above. Isolated femurs were cleaned and then imaged as described above for bone transplant assays.

RNA isolation, qPCR and bulk RNA-seq

Microdissected ARC or medial basal hypothalamic tissue was obtained from control and mutant female mice (1–27 weeks of age) using the optic chiasm as a reference point, and a 2 mm block of tissue containing the hypothalamus was isolated with a matrix slicer. For ARC, total RNA was purified using a RNA Mini kit (Invitrogen). For qPCR, cDNA was synthesized using an Applied Biosystems High-Capacity cDNA Reverse Transcription kit. Expression analysis was performed using SYBR Green. Values were normalized to either 36b4 or mCyclo. Sequences for primer pairs are provided in Supplementary Table 1.

For bulk RNA-seq analyses, barcoded sequencing libraries were prepared using a NEBNext Ultra II RNA Library Prep kit for Illumina from RNA samples after a quality check, and sequencing was performed on Illumina’s NovaSeq 6000, S4 flow cell. Novogene carried out these steps. For all tissue samples, sequencing-generated reads were aligned to the mouse transcriptome (mm10) using Kallisto in gene mode60. Differential gene expression was evaluated using the likelihood-ratio test by Sleuth (qval < 0.05)61. All heatmaps were generated with the top 50 female/male-biased genes obtained from 27-week-old mice and were generated in R62.

Plate-based Smart-Seq2 scRNA-seq of ocSSCs

Single ocSSCs from 4-week-old Esr1fl/fl and Esr1Nkx2.1-cre long bones were isolated by FACS using processing and flow cytometry protocol as described above. Single-cell suspensions pooled from four mice per group was used, and individual cells were captured in separate wells of a 96-well plate containing 4 µl lysis buffer (1 U μl–1 RNase inhibitor (Clontech, 2313B)), 0.1% Triton (Thermo Fisher Scientific, 85111), 2.5 mM dNTP (Invitrogen, 10297–018), 2.5 μM oligo dT30VN (IDT, custom: 5′–AAGCAGTGGTATCAACGCAGAGTACT30VN-3′) and 1:600,000 External RNA Controls Consortium ExFold RNA Spike-In Mix 2 (ERCC; Invitrogen, 4456739) in nuclease-free water (Thermo Fisher Scientific, 10977023) according to a modified Smart-Seq2 protocol54,63. Two 96-well plates per phenotype with a single ocSSC per well were sorted and processed. Plates were spun down and kept at −80 °C until cDNA synthesis, which was conducted using oligo-dT-primed reverse transcription with SMARTScribe reverse transcriptase (Clontech, 639538) and a locked nucleic acid containing template-switching oligonucleotide (Exiqon, custom: 5′-AAGCAGTGGTATCAACGCAGAGTACATrGrG+G-3′). PCR amplification was conducted using KAPA HiFi HotStart ReadyMix (Kapa Biosystems, KK2602) with in situ PCR primers (IDT, custom: 5′-AAGCAGTGGTATCA-ACGCAGAGT-3′). Amplified cDNA was then purified with Agencourt AMPure XP beads (Beckman Coulter, A63882). After quantification, cDNA from each well was normalized to the desired concentration range (0.05–0.16 ng μl–1) by dilution and consolidated into a 384-well plate. Subsequently, this new plate was used for library preparation (Nextera XT kit; Illumina, FC-131–1096) using a semi-automated pipeline. The barcoded libraries of each well were pooled, cleaned-up and size-selected using two rounds (0.35× and 0.75×) of Agencourt AMPure XP beads (Beckman Coulter), as recommended by the Nextera XT protocol (Illumina). A high-sensitivity fragment analyser run was used to assess fragment distribution and concentrations. Pooled libraries were sequenced on NovaSeq6000 (Illumina) to obtain 1–2 million 2× 151 base-pair paired-end reads per cell.

scRNA-seq data processing

Sequenced data were demultiplexed using bcl2fastq2 (v.2.18; Illumina). Raw reads were further processed using a skewer for 3′ quality trimming, 3′ adaptor trimming and removal of degenerate reads. Trimmed reads were then mapped to the mouse genome (v.M20) using STAR (v.2.4), and counts for gene and transcript reads were calculated using RSEM (v.1.2.21). Data were explored, and plots were generated using Scanpy (v.1.9). To select high-quality cells only, we excluded cells with fewer than 450 genes, and genes detected in fewer than three cells were excluded. Cells with a mitochondrial gene content higher than 5%, ERCC content higher than 30% and ribosomal gene content higher than 5% were excluded as well. Scrublet was then used to detect and remove residual duplicates. A total of 264 high-quality cells (122 control and 142 mutant mouse cells) were included in the final analysis. Raw counts per million (CPM) values were mean-normalized and log-normalized, and then data were scaled to a maximum value of 10. Combat batch correction was applied to account for potential biases through minor differences in cell processing. Principal component (PC) ‘elbow’ heuristics were used to determine the number of PCs for clustering analysis with UMAP and Leiden algorithm (leidenalg). Differential gene expression between Esr1fl/fl (WT) and Esr1Nkx2.1-cre (mutant), as well as Leiden clusters, was calculated using Wilcoxon rank-sum test. EnrichR was used to explore enrichment for pathways and ontologies of DEGs between WT and mutant groups64.


Statistical tests, excluding RNA-seq analyses, were performed using Prism 10 (GraphPad). A description of the test and results are provided in Supplementary Table 3. Multiple comparisons correction for one-way, two-way and repeated-measures ANOVA were performed using the post hoc tests as indicated in each figure legend. For all panels in the main figures and Extended Data figures, N indicates biological sample sizes used and n indicates technical replicates in cell culture assays. Outliers were identified using Grubbs test (α = 0.05). Unless otherwise noted, data are presented as mean ± s.e.m. or box plots, in which whiskers represent minimum and maximum values, edges of the box are 25th and 75th percentiles, and the centre line indicates the mean. Sample sizes are based on previous work from our laboratories; however, no specific statistical calculation was performed to determine sample size. For AAVdj-CCN3 and siRNA injections, mice of identical genotypes were drawn at random from littermate pools to receive functional or control virus injections. Experimenters were blinded to the type of AAV received and genotype of the mice under study for all subsequent µCT and dynamic histomorphometry analyses. All raw data and processed data files for the bulk RNA-seq and scRNA-seq are publicly available at the Gene Expression Omnibus (GEO) under sample accession numbers GSE248882 and GSE241478, respectively. A list of reagents used in this study is provided in Supplementary Table 1. Graphics in Figs. 1b, 2b (kidney), 5c and 6j (mammary, bone and calcium) were reproduced or adapted from BioRender ( Graphics in Figs. 1g, 2b (mouse) and 6j (dam and litter) were reproduced or adapted from Mind the Graph ( under a Creative Commons licence CC BY-SA 4.0.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Source link