Mice were group housed in standard housing under a 12–12 h light–dark cycle with ad libitum access to chow diet and water, with the room temperature kept around 22 °C and humidity kept between 30–80% (not controlled). Mice were single housed for food intake measurement and held at 30 °C for thermoneutral exposure experiments. Mice (aged at least 6 weeks) from the following strains were used for this study: wild-type (WT) C57BL/6J (Jackson stock, 000664), B6.Cg-Gt(ROSA)26Sortm9(CAG-tdTomato)Hze/J (Jackson, 007909, Ai9), Pirt-Cre51, Scn10a-Cre52. Both genders were used for anatomical mapping studies, and male mice were used for in vivo functional experiments. All of the experimental protocols were approved by The Scripps Research Institute Institutional Animal Care and Use Committee and were in accordance with the guidelines from the NIH.
CAV2 was obtained from CNRS Vector Core. rAAVretro-Cre was obtained from Boston Children’s Hospital and Janelia. PHP.S-Cre, PHP.S-DIO-sfGFP (gift from D. Gibbs) was obtained from Janelia. PHP.S-TdTomato (Addgene, 59462), PHP.S-mScarlet (gift from the Deisseroth laboratory and cloned in-house), ROOT-Cre (Addgene, 51904), ROOT-EYFP (cloned in-house), PHP.S-mCherry-flex-DTA (Addgene, 58536), ROOT-mScarlet, AAV9-mScarlet and PHP.S-mScarlet were packaged in-house using published protocol53. AAVs produced in-house were titrated by qPCR before aliquoting into 6–10 μl and flash-frozen for long-term storage.
In vivo selection of ROOT
The plasmids used for in vivo selection were adapted from the previous publication54. rAAV-pUBC-sfGFP-Cap and AAV2/9-REP-AAP was generated from pUBC-mCherry-rAB (Addgene, 115239), pUCmini-iCAP-PHP.S (Addgene, 103006), pAAV2/8 (Addgene, 112864). No in-cis-Lox module or transgenic Cre lines were used given the anatomical separation of DRGs and iWAT. The ROOT capsid library was generated by inserting random heptamers using NNK degenerate primers (Integrated DNA technologies) between the 588 and 589 sites of AAV9 by Gibson assembly as previously described54.
AAV capsid library production
The viral libraries were produced as previously described53,54. In brief, only 10 ng of rAAV-pUBC-sfGFP-Cap library DNA was transfected in HEK293FT (Invitrogen R70007) cells per 150 mm plate, and the virus was collected after 60 h for purification.
DNA recovery and sequencing
The resulting AAV capsid library was injected bilaterally into the iWAT of C57Bl/6J male mice at 109 viral genomes (VGs) per fat pad. rAAV genomes were recovered from T12–L3 DRGs from injected mice two weeks after injection using the DNeasy Blood and Tissue kit (Qiagen). The recovered viral DNA was amplified for two rounds against a fragment containing the heptamer insertion with Q5 high-fidelity polymerase (New England Biolabs). The amplified products were cleaned up and processed for complete amplicon sequencing at Massachusetts General Hospital.
NGS data alignment and processing
Raw FASTQ files from NGS runs were aligned to an AAV9-template DNA fragment containing the 21 bp diversified region between amino acids 588 and 589 using SAMtools (v.1.10). The abundance of each 21 bp sequence in all of the recovered sequences with the heptamer insertion was quantified.
Mice were anaesthetized with isoflurane (4% for induction and 1.5–2% for maintenance), with their skin at the surgical area shaved and hair removed and sterilized using ethanol and iodine. After surgery, the mice were given a subcutaneous injection of flunixin and topical antibiotic ointment for post-operative care.
Retrograde tracer labelling of sensory neurons
For injection into the iWAT, a lateral incision was made on the flank skin of each side. For injection into the eWAT, a lateral incision was made on the lower abdominal wall. For injection into the iBAT, a midline incision was made in the interscapular region. For injection into the skin, an intradermal injection was performed. A Hamilton syringe with a 31G (point type 2) needle was used for all retrograde tracing. A total of 4–5 μl 0.1% CTB-488 or CTB-647 (Invitrogen) in PBS was injected per fat pad (2 μl for iBAT) or the flank skin with 8–15 injections to spread out the tracer. The abdominal wall (for eWAT injection) and skin (for iWAT, eWAT or iBAT injection) were sutured separately. Tissues were taken 3–5 days after injection to allow the dye to reach the DRG soma.
Retrograde viral labelling of sensory neurons
Injections were performed as described above. For the viral comparison study (Extended Data Fig. 3a), CAV2-Cre (4.2 × 1012 physical particles per ml, 5 μl), PHP.S-Cre (1.6 × 1013 VGs per ml, 3 μl), rAAVretro-Cre (9 × 1012 VGs per ml, 5 μl) was mixed with 0.01% FastGreen (Sigma-Aldrich) and unilaterally injected into iWAT of Ai9 mice. AAV9-TdTomato (2 × 1013 VGs per ml, 2 μl) was mixed with 0.01% FastGreen (Sigma-Aldrich) and unilaterally injected into iWAT of WT mice.
Characterization of ROOT
Injections were performed as described above. For ROOT characterization, AAV9 or ROOT-mScarlet (4 × 1013 VGs per ml, 2 μl) in PBS with 0.001% F-68 and 0.01% FastGreen was administered unilaterally into iWAT of WT mice. Two weeks after the first surgery, 4 μl of 0.1% CTB-647 was injected into the same iWAT fat pad, and tissues were taken 3–5 days after the second injection for quantification.
6-OHDA has been used previously for selective sympathetic denervation55,56. 6-OHDA (Tocris) (12 mg ml−1 in saline with 0.02% ascorbic acid (Sigma-Aldrich)) was prepared fresh before use and kept on ice and in the absence of light. 6-OHDA (8 μl) was injected into each iWAT fat pad. Saline with 0.02% ascorbic acid was used as a control. Tissues were taken 7–9 days after injection.
Intraganglionic DRG injection
Intraganglionic DRG injection was performed according to a previous report57. A midline incision was made on the dorsal skin to expose the dorsal muscles. The muscles along the vertebra were carefully separated to expose DRGs. DRGs at the T13 and L1 vertebral level were exposed, and AAV (~1 × 1013 VGs per ml in PBS with 0.001% F-68 and 0.01% FastGreen) was injected in the ganglion (~200 nl per ganglion) with a pulled glass pipette using the Nanoliter 2020 Injector (World Precision Instrument). Care was taken to avoid damaging the surrounding vasculatures. Dorsal muscle and skin were sutured separately. PHP.S-TdTomato or PHP.S-mScarlet was used for anterograde labelling.
Cre-dependent ablation of iWAT-innervating DRGs
To selectively ablate iWAT-innervating DRGs, ROOT-Cre or ROOT-YFP (4 × 1013 VGs per ml, 2 μl) in PBS with 0.001% F-68 and 0.01% FastGreen was injected into iWAT unilaterally or bilaterally, and PHP.S-mCherry-flex-DTA (1 × 1013 VGs per ml, 200 nl per ganglion) was injected into T13 and L1 DRGs bilaterally as described above. Tissues were extracted, or physiological measurements were performed 3–4 weeks after surgery.
6-OHDA treatment and Cre-dependent sensory ablation
AAVs were injected into iWAT and DRGs to achieve Cre-dependent sensory ablation as described above. Two to three weeks after the first injection, 6-OHDA or saline was injected into the iWAT again. Tissues were extracted 7–9 days after the second injection.
En bloc HYBRiD tissue clearing and immunolabelling of iWAT
To visualize sensory nerves in mouse torso and iWAT samples, tissue samples were cleared using the HYBRiD method as described previously17.
Sample collection and pretreatment
Mice were terminally anaesthetized with isoflurane and intracardially perfused with ice-cold PBS and ice-cold 4% PFA in PBS with 4% sucrose (Electron Microscopy Perfusion Fixative, 1224SK). For torso samples, the skin was carefully removed leaving the iWAT attached to the muscle, the spinal cord was cut from the midline to facilitate clearing and imaging. All of the collected samples were post-fixed in 4% PFA at 4 °C for 1–2 days before being washed in PBS. The torso samples were decalcified in 10% EDTA/15% imidazole (at 4 °C for 7 days), then decolourized in 25% N,N,N′,N′-tetrakis(2-hydroxypropyl)ethylenediamine (Quadrol) (in 1× PBS) (at 37 °C for 4 days).
Delipidation and hydrogel embedding
The torso and iWAT samples were sequentially washed in 50%, 70%, 80%, 95% and 95% tetrahydrofuran in 25% Quadrol (in 1× PBS), 100% dichloromethane (DCM), 100% DCM, 100% DCM, 95%, 95%, 80%, 70% and 50% tetrahydrofuran in 25% Quadrol followed by 1× PBS to wash out any remaining organic solvent. The samples were then incubated in A1P4 hydrogel (1% acrylamide, 0.125% Bis, 4% PFA, 0.025% VA-044 initiator (w/v), in 1× PBS) at 4 °C before being degassed with nitrogen and polymerized (4 h at 37 °C). The samples were then removed from the hydrogel and passively cleared with 20 mM LiOH-Boric buffer pH 8.0 containing 6% SDS at 37 °C until the samples appeared translucent. After clearing, the samples were washed in PBST (0.2% Triton X-100) thoroughly before proceeding to refractive-index matching or immunolabelling.
iWAT samples were incubated with primary antibodies diluted in PBST at room temperature. After incubation with primary antibodies, the samples were washed in PBST and then incubated in secondary antibodies diluted in PBST at room temperature. The samples were washed extensively in PBST before refractive-index matching and mounting. The following antibodies were used: anti-perilipin-1 (Cell Signaling, 9349, 1:400), anti-TH-647 (BioLegend, 818008, 1:300); anti-Ucp1 (Abcam, ab10983, 1:200); anti-rabbit-488 (Jackson Immuno Research 711-546-152, 1;400); anti-rabbit-647 (Jackson Immuno Research 711-606-152, 1:400).
Refractive-index matching and mounting
Cleared or immunolabelled samples were refractive-index matched in EasyIndex (RI = 1.52, Life Canvas) and mounted in spacers (Sunjin Lab) for confocal microscopy imaging or mounted in agarose for light-sheet microscopy imaging.
Histological analysis of whole-mount samples and cryosections
Mice were terminally anaesthetized with isoflurane and intracardially perfused with PBS and 4% PFA. For flank skin samples, skin was shaved, and hair was removed using Nair.
Whole-mount imaging of ganglia and flank skin
Ganglia of interest (DRGs, SChGs and celiac/mesenteric complex) were dissected and mounted in RapiClear (Sunjin Lab) with silicone spacer (Electron Microscopy) for confocal imaging. Flank skin was dissected, post-fixed in PFA and washed three times in PBS before mounting in Fluoromount-G (Invitrogen 00-4958-02) using 0.25 mm iSpacers (Sunjin Lab) for confocal imaging.
Immunolabelling analysis of skin sections
Flank skin was dissected, post-fixed in PFA, dehydrated in 30% sucrose before being embedded in OCT, then sectioned at 25 μm and mounted on gelatin-coated slides. For immunofluorescence analysis, skin tissue sections were blocked with 5% normal donkey serum in PBS with 0.3% Triton X-100. Primary antibodies were prepared in the same blocking solution and incubated overnight (anti-βIII-tubulin (Abcam, ab18207, 1:1,000). The next day, the sections were washed in PBS and then incubated for 2 h at room temperature with secondary antibodies (anti-rabbit-647, Jackson Immuno Research, 711-606-152, 1:500), and stained with DAPI before mounted with ProLong Gold antifade mountant (Invitrogen) for confocal microscopy imaging.
Immunolabelling of DRG sections
T12 to L2 DRGs from mice with CTB-647 injected in iWAT were dissected, post-fixed in PFA, dehydrated in 30% sucrose before being embedded in OCT, then sectioned at 20 μm. DRG sections were stained according to the procedure described above using the following primary antibodies: anti-CGRP (Immunostar, 24112, 1:1,000), anti-neurofilament heavy polypeptide (Abcam, ab4680, 1:1,000). Secondary antibodies and dyes included anti-rabbit-594 (Jackson Immuno Research, 711-586-152, 1:500), anti-chicken-488 (Jackson Immuno Research, 703-546-155, 1:500) and isolectin B4 AlexaFluor 488 (Life Technologies, I21411, 25 µg ml−1).
Haematoxylin and eosin staining of adipose tissues
Mice were terminally anaesthetized with isoflurane. iWAT, eWAT and iBAT were extracted and post-fixed in 4% PFA overnight at 4 °C. The tissues were paraffin-embedded, sectioned at 5 μm and mounted onto glass slides. The sections were then stained with haematoxylin and eosin at Sanford Burnham Prebys histology core and imaged using a bright-field microscope.
Mounted iWAT samples, whole-mount ganglia, stained skin sections were imaged with Olympus FV3000 confocal microscope using one of the following objectives: ×4/0.28 NA, air (XLFluor, Olympus); ×10/0.6 NA, water immersion (XLUMPlanFI, Olympus). Images were acquired using Fluoview (v.126.96.36.199).
Torso or iWAT samples were mounted using 1% agarose/EasyIndex. Mounted samples were imaged inside the SmartSPIM chamber filled with EasyIndex and sealed with mineral oil on the top. Images were acquired using a ×3.6/0.2 NA objective (LifeCanvas), with a 1.79 μm, 1.79 μm, 4 μm xyz voxel size. Image acquisition was completed with bilateral illumination along the central plane of symmetry within the sample.
Freshly fixed liver samples were imaged using the Leica M165 FC stereomicroscope and FLIR BFS-U3-51S51M-C camera. Images were acquired using SpinView (v.188.8.131.52).
Slides stained with haematoxylin and eosin were imaged using the Keyence BZ-X710 microscope with a ×40/0.6 NA objective (CFI S Plan Fluor ELWD ADM, Nikon).
RNA preparation and RT–qPCR analysis
Adipose tissues were dissected between 12:00 and 14:00 and flash-frozen in liquid nitrogen. Total RNA was extracted from frozen tissue using TRIzol (Invitrogen) and RNeasy Mini kits (Qiagen). For RT–qPCR analysis, total RNA was reverse-transcribed using the Maxima H Minus First Strand cDNA Synthesis Kit (Thermo Fisher Scientific). The resultant cDNA was mixed with primers (Integrated DNA Technology) and SyGreen Blue Mix (Genesee Scientific, 17-507) for RT–qPCR using the CFX384 real-time PCR system (BioRad). Normalized mRNA expression was calculated using ΔΔCt method, using Tbp (encoding TATA-box-binding protein) mRNA as the reference gene. Statistical analysis was performed on ΔΔCt. The primer sequences (forward and reverse sequence, 5′ to 3′, respectively) were as follows: Tbp (CCTTGTACCCTTCACCAATGAC and ACAGCCAAGATTCACGGTAGA); Ucp1 (AGGCTTCCAGTACCATTAGGT and CTGAGTGAGGCAAAGCTGATTT); Elovl3 (TTCTCACGCGGGTTAAAAATGG and GAGCAACAGATAGACGACCAC); Cidea (ATCACAACTGGCCTGGTTACG and TACTACCCGGTGTCCATTTCT); Fasn (GGAGGTGGTGATAGCCGGTAT and TGGGTAATCCATAGAGCCCAG); Acacb (CGCTCACCAACAGTAAGGTGG and GCTTGGCAGGGAGTTCCTC); ChREBPβ (TCTGCAGATCGCGTGGAG and CTTGTCCCGGCATAGCAAC).
RNA library preparation and sequencing
RNA library preparation and sequencing was performed at Scripps Genomics core. Total RNA samples were prepared into RNA-seq libraries using the NEBNext Ultra Directional RNA Library Prep Kit for Illumina according to the manufacturer’s recommended protocol. In brief, 1 μg total RNA was poly(A)-selected for each sample, converted to double-stranded cDNA followed by fragmentation and ligation of sequencing adapters. The library was then PCR-amplified for 8 cycles using barcoded PCR primers, purified and size-selected using AMPure XP Beads before loading onto an Illumina NextSeq 2000 for 100 bp single-read sequencing.
Sequenced reads were aligned to the GRCm39 reference genome (Ensembl, v.104; and gene counts were quantified using Salmon (v.1.5.1)59. Differential gene expression analysis and P-value calculation were performed by DESeq2 (v.1.32.0)58. Gene Ontology enrichment analysis was performed using Metascape59 with the Gene Prioritization by Evidence Counting setting.
Behavioural and physiological assays
Mice were acclimated for 1 h in von Frey chambers. The 50% mechanical threshold was measured with calibrated von Frey filaments (0.07, 0.16, 0.4, 0.6, 1.0, 1.4, 2.0, 4.0 and 6.0 g) using the up–down method60.
Two-temperature choice assay
The two-temperature choice test apparatus was set up as previously described61. Lanes were evenly divided between two different temperature plates set at 30 °C and 18 °C, and individual mice were placed in one of the lanes. The mice were given 10 min to acclimatize and were then tracked for 1 h using the EthoVision tracking system (Noldus Information Technology) in a dark room with infrared lighting. The total time spent in each temperature zone was analysed.
Core body temperature measurement
Core temperature was measured using a thermocouple rectal probe (World Precision Instruments). Room temperature core body temperature was taken when the mice in thermoneutral condition moved to room temperature for more than 24 h.
Blood pressure and heart rate measurement
Blood pressure and heart rate were measured using the tail-cuff method with the CODA High Throughput Non-Invasive Blood Pressure System (Kent Scientific) as previously described62.
Targeted detection of norepinephrine in bulk iWAT
Frozen iWAT tissues were lysed in 4× weight of 0.1 mol l−1 perchloric acid, centrifuged and run through a 30 kDa filtration tube (Millipore). The filtrates were analysed on an Agilent 6470 Triple Quadrupole (QQQ) liquid chromatography–mass spectrometry (LC–MS) system using electrospray ionization (ESI) in positive mode. The AJS ESI source parameters were set as follows: the gas temperature was set at 250 °C with a gas flow of 12 l min−1 and the nebulizer pressure at 25 p.s.i. The sheath gas temperature was set to 300 °C with the sheath gas flow set at 12 l min−1. The capillary voltage was set to 3,500 V. Separation of metabolites was conducted on an Agilent Eclipse Plus C18 LC column (3.5 μm, 4.6 × 100 mm, 959961-902). Mobile phases were as follows: buffer A, water with 0.1% formic acid; buffer B, acetonitrile with 0.1% formic acid. The LC gradient started at 5% B from 0 to 2 min. The gradient was then increased linearly to 5% A/95% B from 2 to 23 min. From 23 to 28 min, the gradient was maintained at 5% A/95% B; and from 28 to 29 min, the gradient went back to the starting concentration of 5% B. The flow rate was maintained at 0.7 ml min−1 throughout the run. Multiple reaction monitoring was performed for noradrenaline, looking at the transition of m/z = 170.1 as the precursor ion to the m/z = 152 fragment. The dwell time was 200, the fragmentor set at 60, the collision energy set at 4 and the cell accelerator voltage set at 4.
Mice received bilateral sensory ablation were fed a high-fat diet (HFD) (D12492, Research Diets) under thermoneutrality at 30 °C. HFD feeding was started at 1 month after surgery (aged around 11–12 weeks old). Body weight was measured every week. Fasting glucose was taken and glucose tolerance tests were performed at 9 weeks on HFD, and plasma insulin was measured at 13 weeks on HFD.
Glucose tolerance test
Mice were fasted for 4 h (08:30–12:30) before intraperitoneal injection of glucose (1 g kg−1). Blood glucose levels were measured at the indicated time point using the OneTouch Ultra 2 blood glucose meter.
Plasma insulin measurement
Fasting blood samples were retro-orbitally obtained 3 h into the light cycle after fasting for 14 h. Plasma was separated in heparin-treated tubes (Microvette CB 300LH). Plasma insulin was measured with an ELISA kit (Crystal Chem, 90080) and read using the BioTek Cytation 5 Imaging Reader.
Whole-tissue protein lysate was extracted in RIPA buffer (G-Biosciences) containing Halt proteinase inhibitor (Thermo Fisher Scientific, 78430) and phosphatase inhibitor (Thermo Fisher Scientific, 78420). Protein lysate was determined using the bicinchoninic acid assay (Pierce). Protein lysates were denatured in Laemmli buffer (BioRad), resolved by 4–12% Mini-PROTEAN TGX SDS–PAGE (BioRad) and transferred to a polyvinylidene difluoride membrane. The membrane was incubated with primary antibodies diluted in EveryBlot blocking buffer (BioRad) overnight at 4 °C and then incubated with secondary antibody anti-rabbit HRP (Jackson Immuno Research, 711-036-152, 1:10,000) or anti-mouse HRP (Jackson Immuno Research, 715-036-150, 1:10,000) diluted in EveryBlot at room temperature. The results were visualized using SuperSignal West Pico PLUS Chemiluminescent Substrate (Invitrogen). The following antibodies were used for immunoblotting: anti-p-HSL(Ser660) (Cell Signaling, 45804, 1:1,000), anti-HSL (Cell Signaling, 4107, 1:1,000), anti-α-tubulin (Abcam, 7291, 1:10,000). Specifically, HSL was blotted after stripping the p-HSL membrane.
Imaging analysis and quantification
All of the images were analysed using ImageJ. 3D volume images were rendered in Imaris.
Quantification of TH+ DRG nerves in the iWAT
Regions of 40 μm × 40 μm × 40 μm (x,y,z) were randomly selected and maximally projected over z using customized ImageJ scripts in the whole stacks of intraganglionically labelled iWAT with TH staining. Only regions containing positive TdTomato (DRG) signals were retained. The thickness of TdTomato-positive fibres was measured using the ImageJ straight line function across two different places in the view and averaged. Fibres with widths of less than 2.5 μm (arbitrary cut-off) were considered to be thin fibres. If the TH-647 channel showed overlap with the TdTomato signal, the view was considered to be positive. We quantified 13 images from 3 biological replicates, and a total of 77 views containing the TdTomato positive thin fibres were quantified for TH positivity.
Quantification of nerve density in flank skin and iWAT
Regions of 80 μm × 80 μm × 20 μm (x,y,z) were randomly selected and maximally projected over z using customized ImageJ scripts in the whole stacks of flank skin or iWAT from mice received intraganglionic labelling. Areas containing nerve fibres were automatically segmented using auto thresholding in ImageJ. Nerve density was calculated as AreaNerve/AreaTotal. Only views containing nerve signals were retained for quantification. We quantified 466 views from 39 images from 3 biological replicates for flank skin, and 468 views from 31 images from 2 biological replicates for iWAT.
Quantification of intraepidermal nerve fibres
Quantification of intraepidermal nerve fibres with antibodies against βIII-tubulin was determined according to the number of nerve fibres crossing the basement membrane in the cross-sections of flank skin. Scoring was performed in a blinded manner on all of the images and post hoc registered to the condition. Ipsilateral and contralateral flank skin from 4 mice (10–15 sections per tissue) was quantified.
Quantification of ROOT/AAV9 labelling in ganglia
DRGs and SChGs were taken from mice injected with AAV (ROOT or AAV9) and CTB for whole-mount imaging as described above. T11–L3 DRGs and T12 SChGs were quantified for cell numbers labelled by AAV and CTB. The percentage of AAV and CTB double-positive cells in all AAV labelled cells were quantified in T13 and L1 DRGs as intraganglionic surgery was performed on T13 and L1 DRGs for later in vivo experiments.
Quantification of CTB/ROOT/AAV9-labelled DRG soma sizes
Soma sizes were quantified manually in full stack images from DRG whole-mount imaging using Fiji. We noticed that injecting CTB conjugated to different fluorophores (that is, CTB-488 versus CTB-647) into the iWAT may result in a slight difference in the distribution of the soma diameter of labelled DRG neurons; we therefore exclusively used CTB-647 when quantifying CTB-labelled DRG soma sizes.
Quantification of liver fluorescence intensity
For liver fluorescence determination in the characterization of ROOT, the same views were acquired with two different exposure times. The images with a longer exposure time were used to determine the shape of the tissue, and the images with a shorter exposure time were used to quantify intensity to avoid over-saturation. The region of interest of the liver and the background were drawn manually in ImageJ, and the liver fluorescence intensity was defined as Mean Intensityliver − Mean Intensitybackground.
No statistical methods were used to calculate the sample size. The sample size was determined on the basis of previous studies and literature in the field using similar experimental paradigms. No data were excluded, except for mice with deteriorating health issues after surgery or during the experiment and mice with missed viral targeting assessed by histology. Data were collected in a blinded manner and post hoc registered to the condition and analysed accordingly to prevent any bias.
Statistics and reproducibility
All non-RNA-seq analysis was performed using GraphPad Prism 9 (v.9.3.1) with the indicated statistical tests. Paired two-tailed Student’s t-tests were performed to compare one animal’s ipsilateral and contralateral sides. The individual animal’s left or right sides were randomly assigned for unilateral treatment. Mice were randomly assigned for bilateral treatment. Sample sizes for each experiment are reported in the figure legends. All in vivo experiments were performed at least twice or grouped from two independent cohorts with the same conclusion, except HFD treatment was performed in one cohort. For representative images, the numbers of experimental repetitions were as follows: Figs. 1b–d (four), 1e,g,h,j (three), 2a–b,d (two), 4a (four), 4c,d (two) and Extended Data Figs. 1a (three), 1b (four), 1c (two), 2a (six), 2f (three), 2h–i,k (two), 3a,d,e (three), 3b (two), 4a,h (two); 5b (two), 7d (three), 7e (two).
Further information on research design is available in the Nature Research Reporting Summary linked to this article.