The serotonin transporter sustains human brown fat thermogenesis

In vivo crossover research study in healthy volunteers

Fifteen healthy volunteers aged 18–35 years (Table 1) were hired to a double-blind placebo-controlled randomised crossover research study. Inclusion requirements were as follows: BMI 18.5–25 kg m⁻²; weight modification of <5% in preceding 6 months; no intense or persistent medical conditions; on no routine medications; alcohol consumption ≤14 systems each week; no claustrophobia or contraindication to MRI scanning; no pregnancy or breastfeeding in female individuals; typical screening blood tests (complete blood count, glucose, kidney, liver and thyroid function). Approval was obtained from the South East Scotland Research Ethics Committee (principles number 18/SS/0104) and notified approval was obtained from each research study individual. Volunteers received the SSRI sertraline at a dosage of 50 mg daily for 7 days or placebo in randomised order (Extended Data Fig. 3a). Sertraline was selected as the SSRI for this research study due to the fact that it shows very little inhibition of the noradrenaline transporter⁶⁹ and attains high concentrations in fat⁷⁰. On the seventh day of tablet administration, research study individuals participated in the Edinburgh Clinical Research Facility (ECRF) in the Royal Infirmary of Edinburgh at 08:00 hours following an over night quick. Volunteers were advised to prevent alcohol for 7 days and workout for 2 days prior to each see and to use similar basic light similar clothes at each see. At each research study see, measurements of height, weight, body fat mass by bioimpedance (utilizing an Omron BF302 screen) and high blood pressure were carried out. Fasting blood samples were obtained for measurement of glucose, insulin, NEFAs and noradrenaline. At time T = 0 minutes, individuals were positioned in a room at 23–24 °C (warm room) for 1 h (Extended Data Fig. 3a). At T + 60 minutes, volunteers were relocated to a room cooled to 16–17 °C (cold room) for 3 h (approximately T + 240 minutes) to trigger their BAT. Participants were examined every 15 minutes for indications or signs of shivering. Blood samples were obtained every 30 minutes from T = 0 to T + 120 minutes, EE was determined two times each hour over the exact same duration. Thermal imaging was carried out of the upper body area periodically from T = 0 to T + 180 minutes to determine supraclavicular and lower sternal skin temperature levels. At T + 180 minutes, individuals received an intravenous injection of 75 MBq of ¹⁸F-FDG. At T + 240 minutes, individuals went through a FAMILY PET–MRI scan to measure BAT mass and activity. Following the scan, volunteers returned home. Female individuals participated in for their 2nd see 4 or 8 weeks after their preliminary check out to guarantee they remained in similar stages of their menstruation. Male individuals attended their 2nd see a minimum of 3 weeks after the very first.

In vivo study in normal weight and obese healthy volunteers

Twenty healthy volunteers aged 18–40 years (Table 2) were recruited to a case-control study. Inclusion criteria were as follows: BMI 18.5–25 kg m⁻² (normal weight group) or 30–55 kg m⁻² (obese group); weight change of <5% in preceding 6 months; no acute or chronic medical conditions known to alter BAT activity or preclude an abdominal adipose tissue biopsy; on no regular medications known to alter serotonin concentrations or BAT activity; alcohol intake ≤14 units per week; no pregnancy or breastfeeding in female individuals; no allergy to local anaesthetic; normal screening blood tests (full blood count, glucose, kidney, liver and thyroid function). Approval was obtained from the Edinburgh Medical School Research Ethics Committee (ethics number 18-HV-049) and informed consent was obtained from each participant. Subjects attended the ECRF at 0800 hours following an overnight fast. Volunteers were instructed to avoid alcohol and exercise for 2 days before the study visit. Measurements of height, weight, body fat mass by bioimpedance and blood pressure were performed. Blood samples were obtained for measurement of glucose, insulin, NEFAs, serotonin and noradrenaline. At T = 0 min participants were placed in a room at 23–24 °C (warm room) for 2 h (Extended Data Fig. 3b). At T + 120 min, volunteers were moved to a room cooled to 16–17 °C (cold room) for 2 h (up to T + 240 min). Participants were checked every 15 min for signs or symptoms of shivering. Blood samples were obtained every 30 min from T = 0 to T + 240 min, EE was measured once each hour over the same period. Thermal imaging was performed of the upper body region intermittently from T = 0 to T + 240 min to measure supraclavicular and lower sternal skin temperatures. At T + 105 and T + 225 min (following 105 min of warm and cold exposure), an abdominal subcutaneous adipose tissue biopsy was performed. Following the final blood sample, volunteers were given lunch and allowed home.

PET–MR scanning protocol and analysis

All participants received 75 MBq of ¹⁸F-FDG 1 h before being placed supine on a Siemens mMR scanner (Siemens Healthineers). Following initial localization, a standard MRAC_GRAPPA scan was acquired for each bed position, used to calculate a standard umap for attenuation correction. A three-dimensional T1-weighted Dixon VIBE acquisition was used to generate images at 1.34 and 2.56 ms (repetition time (TR) = 4.02) to calculate a fat fraction map (FFM) as outlined below. ¹⁸F-FDG uptake by BAT was quantified using Analyze v.12.0 (AnalyzeDirect). We developed a semi-automated process to standardize the PET/MRI analyses. Following registration, the MRI fat and water images were used to generate a FFM of each participant using the following equation:

A median spatial filter was applied to the FFM, and any voxels below 50% fat fraction (that is not adipose tissue) were removed to generate a fat fraction region of interest (ROI). The registered PET images were loaded and thresholded using the BARCIST criteria (a standard uptake value (SUV)_lean of ≥1.2 g ml⁻¹ for each participant⁹). Any remaining ROIs that encompassed the brain were manually removed. Voxels meeting both the FFM and PET thresholds were analysed. Cervical, supraclavicular and axillary depots were tagged together (classified as supraclavicular or SCV for simplicity), whereas paraspinal BAT was tagged separately. Finally, image erosion with a (3 × 3 × 1) structuring element was applied to the BAT ROIs to remove PET blooming and boundary artefacts. The effect of sertraline on ¹⁸F-FDG uptake by BAT was similar irrespective of whether image erosion was performed (see Source data for Fig. 3 and Extended Data Fig. 4). Data are presented for the BAT SUV_mean and BAT volume, in addition to total ¹⁸F-FDG uptake by BAT (SUV_mean multiplied by BAT volume) as previously described²⁰.

For skeletal muscle analysis, ROIs from the pectoralis major, psoas major, sternocleidomastoid, longus colli and trapezius muscles, and in abdominal subcutaneous adipose tissue, were drawn manually using the MR water image, ensuring ROIs were identically sized and positioned at both study visits. These ROIs were applied to the registered and filtered PET images as described above for analysis.

Thermal imaging

Thermal imaging was performed as described previously²⁰. In brief, thermal images of the volunteers’ upper body were obtained using a FLIR T650sc camera at defined intervals (generally 10–15 min apart during specific procedures such as adipose tissue biopsies). The camera was positioned 1 m from the participant and the emissivity set to 0.98 for human skin. Identically sized regions of interest were drawn around the left and right supraclavicular regions and the lower sternal region (as demonstrated in Fig. 3c) using FLIR Tools (FLIR). The mean and maximum supraclavicular (left and right) and lower sternal temperatures were recorded from each image. Data are presented as the mean of the mean supraclavicular and sternal temperatures recorded during warm and cold conditions.

Indirect calorimetry

EE was measured for 15 min on two occasions during both warm and mild cold exposure using a ventilated-hood indirect calorimeter (GEM Nutrition). The first 5 min of data were discarded and the mean value over the final 10 min was recorded. EE is presented as the mean of the two values obtained during warm and cold exposure. CIT was calculated by subtracting the mean EE obtained in the warm room from the EE measured in the cold room as previously described^20,35.

Abdominal adipose tissue biopsy

Anterior subcutaneous adipose tissue biopsies were performed by needle aspiration as described previously⁷¹. Following sterilization of the area and injection of 5 ml 2% lidocaine, a 14 G needle and 50 ml syringe was inserted ~5 cm lateral to the umbilicus. Following aspiration of ~200 mg adipose tissue, samples were washed with 0.1% diethylpyrocarbonate-treated water then immediately frozen on dry ice before storing at −80 °C. The biopsy obtained during cold exposure was aspirated from the contralateral abdominal depot to the one taken while in the warm room.

Biochemical assays

Plasma serotonin and noradrenaline (both LDN) and serum insulin (Mercodia) were measured by enzyme-linked immunosorbent assay (ELISA). Colorimetric assays were used to measure plasma glucose (Sigma-Aldrich) and NEFAs (Wako Diagnostics). For measurement of serotonin concentrations in platelet-poor and platelet fractions, EDTA plasma samples were subjected to centrifugation at either 2,000g (platelet-poor plasma) or 200g (platelet fraction) for 10 min at room temperature. The platelet fraction was further prepared by adding five volumes of saline to the supernatant which was then spun at 4,500g for 10 min at 4 °C; finally the pellet was resuspended in water. Difficulties in blood sampling in two participants on one phase of the SSRI study and during cold exposure in three participants in the case-control study led to platelet rupture meaning we were unable to measure platelet-poor or platelet serotonin levels in these volunteers, hence data are reported for the remaining 13 and 17 participants respectively. Screening blood tests were analysed by the Royal Infirmary of Edinburgh biochemical laboratory for serum kidney, liver and thyroid function and plasma glucose using an Abbott ARCHITECT c16000 analyser and for full blood count using a Sysmex XE-5000 analyser.

Analysis of circulating sertraline concentrations by liquid chromatography tandem mass spectrometry

Lithium heparin plasma samples were obtained after overnight fast during both phases following 7 days of sertraline and placebo tablets. Targeted analysis of sertraline and its major metabolite norsertraline was carried out by automated supported liquid extraction followed by liquid chromatography tandem mass spectrometry in multiple reaction mode. Calibration standards were prepared ahead of time by enriching blank human plasma with sertraline and norsertraline (N-049 and S-021 respectively; Cerilliant) across expected ranges (0.25–50 and 2.5–100 ng ml⁻¹, respectively). On the day of extraction, 200 μl of standard and samples were aliquoted into 96-well deep-well plates, enriched with 1 ng of d3-sertraline (S-026; Cerilliant) as an internal standard. These were diluted on an Extrahera liquid extraction robot (Biotage) with 0.5 M ammonium hydroxide (200 μl), on an SLE400 plate, followed by elution with ethyl acetate (1,800 μl), reduction to dryness, resuspension in water/methanol (100 μl; 90:10 v/v water/acetonitrile), and the plate was sealed with zone-free 96-well plate sealing film (Sigma-Aldrich) before liquid chromatography tandem mass spectrometry analysis.

Liquid chromatographic separation was achieved by injection (10 μl) on to an Acquity I-Class UPLC (Waters) using a Kinetex C18 column (150 × 2.1 mm; 2.6 μm; Phenomenex), protected by a Kinetex KrudKatcher (Phenomenex) at 40 °C. The mobile phase consisted of 0.1% formic acid in water (A) and 0.1% formic acid in acetonitrile (B) at a flow rate of 0.4 ml min⁻¹. Gradient elution was achieved with a total run time of 12 min from 20% to 90% B. The analytes were detected on a QTrap 6500+ mass spectrometer (AB Sciex) operated in positive electrospray mode (500 °C, 4.5 kV). The analytes and internal standard sertraline, d3-sertraline and norsertraline eluted from the column at 4.90, 4.89 and 4.75 min respectively. Multiple reaction mode transitions monitored for were sertraline m/z 306.1 → 159.0, 275.0 at 31 and 15 V, norsertraline m/z 275.0 → 159.0, 129.0 at 21 and 19 V and d3-sertraline m/z 309.1 → 159.0 at 16 V.

Linear regression analysis was applied to the ratio of the peak area of sertraline (1/x²) and norsertraline (1/x) to the internal standard d3-sertraline using Analyst software, and the amount of sertraline and norsertraline in the samples was calculated from the peak area ratio using the linear regression analysis equation.

Analysis of adipose tissue serotonin concentrations

Adipose tissue was homogenized in 0.2 N perchloric acid and subject to centrifugation at 10,000g for 5 min at 4 °C. The collected supernatant was mixed 1:1 in 1 M borate buffer (1 M boric acid, pH 9.25) and spun at 10,000g for 1 min. The supernatant was collected and stored at −80 °C until analysis. Serotonin concentrations were determined by ELISA (Beckman Coulter).

Analysis of retrospective ¹⁸F-FDG-PET–CT patient scans

Local Caldicott guardian approval was obtained. Patients were included who received scans between two separate periods, from December 2014 to January 2016 and from November 2017 to June 2019. Participants were assigned into three groups: (1) those currently prescribed an SSRI, (2) those currently prescribed any other class of antidepressant, and (3) those not currently prescribed antidepressants (Extended Data Table 2). Groups were matched for sex, age, body weight, fasting glucose, presence of diabetes, underlying cancer diagnosis and outdoor temperature in the month of scan. Patients received either a fixed a dose of 370 MBq or 4 MBq kg⁻¹ of ¹⁸F-FDG 1 h before being placed supine in a hybrid PET–CT scanner. Two models of hybrid PET–CT scanners were used, a Biograph mCT (Siemens Medical Systems) or a Discovery 710 (GE Healthcare). All scans were performed at room temperature and participants underwent a low-dose CT for attenuation correction (non-enhanced, 120 kV) with tube current modulation applied (50 mAs quality reference) followed by static PET imaging of the upper body using 3-min beds. On the Biograph, the data were reconstructed using Siemens UltraHD reconstruction (2 iterations and 21 subsets), on the Discovery 710 the QClear reconstruction was used with a beta value of 400. Images were analysed using PMOD v.3.7 (PMOD Technologies). ¹⁸F-FDG uptake by BAT was quantified by measuring the mean SUV from all pixels with an SUV of >1.5 g ml⁻¹ normalized to body mass, which corresponded to tissues with a radio density on the CT scan with Hounsfield units between −190 and −10⁹. Participants were classified into ‘BAT-positive’ or ‘BAT-negative’ based on the above analysis. The mean SUV and the volume of BAT with detectable ¹⁸F-FDG uptake were calculated for the BAT-positive patients.

Effect of cold exposure on circulating serotonin levels in BAT-positive participants

Stored blood samples obtained from 11 male study participants with detectable ¹⁸F-FDG uptake by BAT during cold exposure^20,35 were analysed to determine whether acute cold exposure altered circulating serotonin concentrations. Plasma samples were analysed from these participants following <2 h of warm exposure at ~24 °C and following approximately 60 min of exposure to mild cold at ~17 °C. Serotonin concentrations were measured by ELISA analysis (LDN) as described above.

In vitro studies on primary adipocytes

Tissue collections

Male and female euthyroid participants were recruited (all participant data is given in Supplementary Table 1) who were due to undergo elective thyroid or parathyroid surgery in the Royal Infirmary of Edinburgh. Local ethical approval was obtained (ethics numbers 15/ES/0094 and 20/ES/0061) as was consent from each participant. Adipose tissue was obtained intra-operatively from the central compartment of the neck, superior to the clavicle and deep to the lateral thyroid lobe either adjacent to the longus colli muscle or to the oesophagus (BAT), and more superficially from the subcutaneous neck tissue (WAT). Tissue collections were all performed by the same surgeon. Tissue samples were either immediately frozen at −80 °C for qPCR, fixed for immunohistochemistry, or the stromal vascular fraction was isolated and cultured for the respective experiments detailed below.

For murine tissue collections, male 129/Ola mice were housed at 20–21 °C and a humidity of 50–55%, using a 12 h/12 h light/dark cycle and with food and water available ad libitum. Animals were killed at 8–9 weeks of age and their interscapular BAT, inguinal and epididymal WAT were immediately collected and either frozen at −80 °C for qPCR or their stromal vascular fraction was isolated and cultured.

Cell culture

Cells were cultured as described previously²⁰. Following collection, adipose tissue samples were incubated in Krebs–Henseleit buffer containing 0.2% collagenase type 1 at 37 °C for 45 min. Samples were subject to centrifugation at 800g for 10 min, the pellet resuspended and passed through a 100-μm filter, subject to centrifugation at 200g for 5 min and the pellet resuspended in DMEM containing 10% FBS and cultured in six-well plates. Cells were passaged when 80% confluent and differentiated in DMEM containing 10% FBS with the addition of 1 nM tri-iodothyronine, 20 nM insulin, 500 μM IBMX, 500 nM dexamethasone and 125 μM indomethacin for 7 days for human cells and 5 days for murine cells. Cells were then cultured for a further 7 days in DMEM medium containing 10% FBS, 1 nM tri-iodothyronine and 20 nM insulin before experiments.

Quantification of serotonin uptake by primary human adipocytes

For all the experiments involving addition of serotonin to medium, medium and samples were protected from light to prevent rapid degradation of serotonin. Cellular serotonin uptake was measured in duplicate by adapting a previously published protocol⁷². Paired white and brown adipocytes were pre-incubated for 1 h in DMEM medium containing either vehicle or ascending doses of sertraline (1 nM to 10 µM). Wells were then washed twice with uptake buffer (120 mM NaCl, 5 mM KCl, 1.2 mM CaCl₂, 1.2 mM MgSO₄, 1 mM ascorbic acid, 25 mM HEPES, 5 mM glucose; pH 7.4) before incubation with 20 nM [³H]-5-hydroxytryptamine (Perkin Elmer) and either vehicle or sertraline (1 nM to 10 µM) for 1 h. Serotonin uptake was terminated by aspiration of medium followed by three washes with uptake buffer. Cells were lysed by the addition of 1% SDS followed by a 30-min incubation at 37 °C. Wells were scraped and lysates were added to the scintillation vials containing Opti-Fluor (Perkin Elmer) for quantification by liquid scintillation counting.

Serotonin regulation of UCP1 mRNA levels in human adipocytes

The serotonin concentration in DMEM medium containing 10% FBS was quantified by ELISA and measured ~410 nM. Therefore, for experiments where serotonin was added to culture medium, serotonin was stripped by incubating FBS with dextran-coated charcoal⁷³, which successfully reduced the serotonin concentration in culture medium to ~3 nM. Paired white and brown adipocytes were incubated in DMEM medium containing stripped serum and insulin/tri-iodothyronine (T3) for 24 h. Thereafter, wells were incubated with medium containing either vehicle or 10 nM to 100 µM serotonin for 24 h before cell lysis for qPCR. In a separate experiment, following incubation with stripped serum medium for 24 h, white and brown adipocytes were incubated with medium containing vehicle, sertraline (1 μM), or sertraline in the presence of serotonin (10 nM to 10 μM) for 24 h.

Inhibition of 5-HT_2A/B receptor on UCP1 mRNA levels in human brown adipocytes

Brown adipocytes were incubated in DMEM medium containing stripped serum and insulin/T3 for 24 h. Thereafter, wells were incubated with the above medium with the addition of either vehicle or 100 µM serotonin for 24 h; wells receiving serotonin were further treated with either vehicle, the 5-HT_2A receptor inverse agonist pimavanserin (10 µM)⁷⁴, the 5-HT_2B receptor antagonist SB-204741 (10 µM)⁷⁵ or both compounds. Thereafter, cells were lysed for qPCR analysis.

Knockdown of HTR2A and HTR2B in human brown adipocytes

siRNA-mediated knockdown of the 5-HT_2A/2B receptors was undertaken based on a previously published protocol⁷⁶. Following differentiation, human primary brown adipocytes were transfected with 20 nM On-TARGETplus Smartpool siRNA (Dharmacon) either targeting HTR2A (L-005638-00-0005), HTR2B (L-005639-02-0005) or a non-targeting pool (D-001810-10-05), using Lipofectamine RNAimax reagent (Thermo Fisher Scientific) according to the manufacturer’s instructions. After 48 h, cells were washed with PBS buffer and replaced with medium containing stripped serum for 24 h. Thereafter, wells were incubated with this medium containing either vehicle or 100 µM serotonin for a further 24 h. Cells were then lysed in QIAzol (Qiagen) for RNA.

RNA extraction and quantitative real-time PCR measurements

Whole adipose tissue was homogenized in QIAzol reagent using a TissueLyser (Qiagen). mRNA was extracted from both tissue and cells using the RNeasy Lipid Kit (Qiagen) and complementary DNA was generated using the Qiagen QuantiTect reverse transcription kit. qPCR was performed in triplicate using a Roche Lightcycler 480, using gene-specific primers (Invitrogen) and fluorescent probes from the Roche Universal Probe Library, or with Taqman assays as detailed in Extended Data Table 4. Transcript levels are presented as the ratio of the abundance of the gene of interest: mean of abundance of control genes (PPIA and RNA18S5 in humans, Tbp and Rn18s in mice).

RNA-seq of human primary adipocytes

Paired brown and white human primary adipocytes were incubated in DMEM containing 10% stripped FBS, 1 nM T3 and 20 nM insulin for 48 h. Cells were lysed and mRNA extracted as above. Before transcriptomics, UCP1 mRNA levels were confirmed to be substantially higher in the brown adipocytes (see Source data for Fig. 1; P < 0.01 versus white adipocytes). RNA-seq was performed by the ECRF Genetics Core. Total-RNA quality and integrity were assessed on an Agilent Bioanalyser using the RNA 6000 Nano kit. Samples were DNase-treated using the TURBO DNA-free kit (Thermo Fisher Scientific), then quantified using the Qubit 2.0 fluorometer and the Qubit RNA BR assay, and assessed for residual DNA contamination using the Qubit dsDNA HS assay.

Nucleic acid library construction protocol

TURBO DNase-treated total-RNA samples were used to generate the libraries using the TruSeq Stranded Total-RNA with Ribo-Zero kit (catalogue no. RS-122-2201). Total-RNA (100 ng) was depleted of ribosomal RNA before purification, fragmentation and were primed using random hexamers. The RNA fragments were reverse transcribed using reverse transcriptase and random primers. Double-stranded cDNA was synthesized following removal of RNA templates using a replacement strand incorporating dUTP in place of dTTP. Blunt-ended cDNA was generated using AMPure XP beads (Beckman Coulter) to separate the double-stranded cDNA from the 2nd strand reaction mix, before the addition of an ‘A’ nucleotide to 3′ ends of the blunt fragments (and a ‘T’ on the dapter 3′ end). Thereafter, multiple indexing adaptors were ligated to the double-stranded cDNA for hybridization onto a flow cell, before enrichment and amplification of adaptor-containing DNA fragments using 15 cycles of PCR to ensure the library was suitable for sequencing. Libraries were quantified using the Qubit dsDNA HS kit and by PCR using the Kapa Universal Illumina Library Quantification kit. An Agilent Bioanalyser was used for quality control.

Nucleic acid sequencing protocol and analysis of RNA-seq data

Sequencing was performed using the NextSeq 500/550 High-Output v2 (150 cycle) Kit on the NextSeq 550 platform. The Illumina RNA-Seq Alignment Workflow was used to generate count files (genome assembly was performed (hg19; Illumina)). Gene expression level normalization was performed by DESeq2 (v.1.26.0, Bioconductor v.3.10)⁷⁷ which was used for downstream analysis. Pathway analysis of gene identifiers extracted from RNA-seq was performed using Ingenuity Pathway Analysis (Ingenuity Systems, www.ingenuity.com, Qiagen). Volcano plots and heat maps to visualize significantly differentially expressed transcripts (adjusted P value < 0.05, log₂(fold change) > 2) were generated using Prism v.9 (GraphPad). RNA-seq data generated this research study have been uploaded to ArrayExpress (E-MTAB-101123; details in Source data for Fig. 1).

Respirometry

Pre-adipocytes from human WAT and BAT were plated in Seahorse XF24 V7 PS cell culture plates (Agilent Technologies) and differentiated as above. White and brown adipocytes were cultured in serum-stripped medium for 24 h, and then cultured in serum-stripped medium containing either vehicle or serotonin (10 nM to 100 μM) for 24 h. In a separate experiment, following culture in serum-stripped medium for 24 h, white and brown adipocytes were cultured in serum-stripped medium containing either vehicle, sertraline (1 μM) or sertraline in the presence of serotonin (10 nM to 10 μM) for 24 h. Cells were analysed on a Seahorse XFe24 analyser as described previously²⁰. Measurements were performed during basal respiration and following sequential addition of noradrenaline (2 μM), oligomycin (2.5 μM), carbonyl cyanide-p-trifluoromethoxyphenylhydrazone (FCCP, 2 μM) and rotenone (0.2 μM)/antimycin A (2.5 μM). The oxygen consumption rate (OCR) was calculated following subtraction of non-mitochondrial respiration. Basal respiration was calculated by taking the mean of three cycles, whereas stimulated respiration was measured using the results of the sixth cycle following the addition of noradrenaline. Uncoupled respiration was calculated by taking the mean of three cycles following addition of oligomycin, whereas maximal respiration was measured after the addition of FCCP for one cycle.

Immunohistochemistry

Human WAT and BAT were fixed for 24 h in 10% formalin before embedding in paraffin. Tissue sections were incubated with mouse anti-human SERT antibody (1:1,500 dilution; Millipore) or rabbit anti-human UCP1 antibody (1:8,000 dilution; Sigma-Aldrich, catalogue no. U6382) as the primary antibodies, and goat anti-mouse or goat anti-rabbit (both 1:200 dilution; Vector Laboratories) as the secondary antibodies following the avidin–biotin complex method as described previously²⁰. Antigen retrieval was conducted in Tris-EDTA buffer (pH 9) for 5 min in a pressure cooker. Images were taken on a NIKON Eclipse Ci-L mounted with a NIKON DS-Fi3 camera and DS-L4 controller set.

Statistical analyses

Data are presented as mean ± s.e.m. All analyses (except RNA-seq, detailed above) were performed either using Prism software v.9 (GraphPad) or SPSS v.25 (IBM). A P value <0.05 was considered statistically significant and all tests performed were two-sided. The group numbers in the figure captions indicate the number of independent biological replicates. Data were analysed for typical distribution within each experimental group. Comparisons between two related groups were examined using the paired t-test, whereas comparisons between two unrelated groups were performed by unpaired t-test. Comparisons involving three or more related groups were analysed using two-way repeated measures analysis of variance (ANOVA) with post-hoc testing. Where data was not normally distributed, for comparisons between two related groups the Wilcoxon signed-rank test was used. Associations were tested using Pearson’s correlation coefficient. The specific statistical tests used are detailed in the respective figure captions along with the number of biologically independent samples used for each comparison. All data generated or analysed during this research study are included in this published article (including supplementary information, extended information figures, tables and source information).

Reporting summary

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