A summary of the animal protocol is illustrated in Supplementary Figure S1 and was undertaken under the United Kingdom Animals (Scientific Procedures) Act, 1986 with approval from the Local Ethics Committee of the University of Nottingham (Nottingham, United Kingdom).

Eighty-eight female and 16 male Sprague-Dawley rats were obtained at 4 weeks of age from Charles River Laboratories. Half of the male and female rats were immediately randomized to be housed either at a cool temperature (19–21°C) or at thermoneutrality (26–28°C). They were also randomized to receive either a low fat, low sucrose diet (L; 18% of energy from fat (soybean oil); 24% from protein; 58% from carbohydrates, of which 7% are mono- and disaccharides; 3.1 kcal/g; Harlan Teklad Global 18% Protein Rodent Diet; Teklad Diets, Madison, WI, United States) or a high fat, high sucrose diet (H; 39% of energy from fat (lard); 23% from protein; 39% from carbohydrate, of which 44% are mono- and disaccharides; 4.6 kcal/g; custom diet supplied by Abbott Nutrition, Granada, Spain; Supplementary Table S1), fed ad libitum. Food intake was measured over the course of three days before mating occurred. Starting numbers of males and females in the H groups were higher than in the L groups as we anticipated a lower fertility although this did not occur. Twelve females (3 at 20L, 2 at 20H, 2 at 27L, 5 at 27H) had reabsorbed all fetuses or failed to mate and were therefore not analyzed. Final groups were, therefore, 20L (18 females, 3 males), 20H (21 females, 5 males), 27L (19 females, 3 males), 27H (18 females, 5 males). All animals were kept in 12:12 h light:dark cycle at 40–50% humidity, and in groups of 3 females and 2 or 3 males to reduce social stress, but had no observable impact on their behavior.

Body weight was recorded three times a week. At 10 weeks of age, 6 randomly selected females of each group were euthanased and tissues were collected. The remaining females were mated with males of the same diet and temperature group. Vaginal plugs found in the morning indicated day 1 of pregnancy. Females were returned to their home cages and re-housed with the same cage mates, but because of the 4 day reproductive cycle of the rat, each mother did not become pregnant on the same day. This meant it was not possible to accurately measure food intake for each group of mothers through pregnancy. Females were euthanased and tissues were collected from 6 to 7 dams per group at 10 days gestation (d/g), coincident with mid-gestation, and from 6 to 8 dams per group near term i.e., 19 d/g.

Females were not fasted before tissue collection to avoid the fasting effects on brown adipocytes. They were euthanased between 10.00 and 12.00 h by CO₂ asphyxiation and subsequent cervical dislocation. A blood sample was collected into EDTA tubes after cardiac puncture, and interscapular, inguinal, omental (mesenteric), perirenal and gonadal adipose depots were collected and weighed. Blood was centrifuged at 4°C (15 min at 2000g) and plasma aliquoted and stored at −80°C until analysis. The interscapular adipose depot was halved, with one half snap frozen in liquid nitrogen and stored at −80°C until analysis, and the other half fixed in 4% formaldehyde in 0.9% saline before dehydration and blocking in paraffin. An aliquot of the remaining adipose tissues was snap frozen in liquid nitrogen and stored at −80°C until analysis.

Plasma was thawed gently on ice. Plasma concentrations of glucose (GAGO-20, Sigma-Aldrich, Gillingham, United Kingdom), triglycerides (GPO DAOS method, Wako, Neuss, Germany), non-esterified fatty acids [NEFA-HR(2), Wako, Neuss, Germany], insulin (80-INSRT-E01, Alpco, Salem, NH, United States), corticosterone (K014-H1, Arbor Assays, Ann Arbor, MI, United States) and leptin (EZRL-83K, Merck, Darmstadt, Germany) were measured with commercial assays, respectively, following manufacturer’s instructions.

Paraffin-blocked interscapular adipose tissue were sectioned to 6 μm and applied to Superfrost Plus slides (Thermo Fisher Scientific, Wilmingont, DE, United States). Slides were processed in a Bond Max IHC stainer (Leica Biosystems, Milton Keynes, United Kingdom) with rabbit anti-UCP1 antibody (ab10983, Abcam, Cambridge, United Kingdom; 1:3000), and compared to negative control slides without primary antibody applied. Slides were viewed and photographed at 10× magnification (Nikon Eclipse 90i, Nikon, Tokyo, Japan; with Hamamatsu ORCA-ER camera, Hamamatsu, Hamamatsu City, Japan). A minimum of five images per sample were analyzed with Fiji-native methods, (Schindelin et al., 2012) using Weka segmentation (Arganda-Carreras et al., 2017).

Adipose tissues were homogenized using a Dispomix (Wilten, Etten-Leur, Netherlands), and RNA was extracted from 50 to 100 mg of tissue using TRI reagent (Sigma-Aldrich, Gillingham, United Kingdom) and chloroform (Thermo Fisher Scientific, Wilmingont, DE, United States) within the Qiagen RNeasy kit (Hilden, Germany) with gDNA eliminator columns. Concentration and purity of eluted RNA was measured on a Nanodrop spectrometer (Thermo Fisher Scientific, Wilmington, DE, United States). One μg of RNA was reverse transcribed to cDNA with the High Capacity RNA-t-cDNA kit (Applied Biosystems, Thermo Fisher Scientific, Wilmington, DE, United States).

Gene expression was measured using fast SYBR Green (Applied Biosystems, Thermo Fisher Scientific, Wilmington, DE, United States) in StepOnePlus Real-Time PCR system (Applied Biosystems, Thermo Fisher Scientific, Wilmington, DE, United States). Specificity of primers (Sigma-Aldrich, Gillingham, United Kingdom; working concentration 250 nM) was confirmed by sequencing PCR product (primers are listed in Supplementary Table S2). Gene expression was assessed for the following pathways:

Gene expression was normalized to the geometric mean of housekeeping genes TATA-box binding protein (TBP) and tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein (YWHAZ) throughout, and was calculated using the 2-ΔΔCt method (Livak and Schmittgen, 2001).

All statistical analyses were carried out in SPSS (Version 22, IBM, Portsmouth, United Kingdom) and Graphpad (Version 6, GraphPad Software, La Jolla, CA, United States). Data was tested for normality using a Kolmogorov-Smirnov test. If normally distributed, data was analyzed (i) with two-way ANOVAs with LSD post hoc comparisons (20L vs. 20H, 27L vs. 27H, 20L vs. 27L, and 20H vs. 27H) within each of the three time points (0, 10, and 19 d/g) to test for changes due to the interventions, and (ii) with one-way ANOVAs with a priori comparisons (0 vs. 10 or 19, and 10 vs. 19 d/g) within each interventional group to test for changes with pregnancy. If data was not normally distributed, Kruskal-Wallis tests with equivalent Mann-Whitney tests were carried out.

Before mating, body weight was only higher due to the H diet when animals were kept at thermoneutrality, whereas fat mass was higher due to the H diet only at the low temperature (Figure 1). Body weight and fat mass increased with pregnancy in all interventional groups. By 10 d/g, body weight was higher with the H diet regardless of temperature, whereas fat mass was only raised in those dams maintained at thermoneutrality. Body weight and fat mass were similar between groups by 19 d/g. Fetal weight was not different between any of the groups, at either sampling gestations and increased with gestation e.g., 20H: 88 ± 2 mg at 10 d/g; 135 ± 3 mg at 19 d/g.

Before mating, plasma glucose concentrations were notably higher in animals maintained at thermoneutrality but were similar between groups through pregnancy (Table 1). Plasma triglyceride and leptin increased with gestation in all groups, whereas NEFA concentrations were only increased at 19 d/g in the 20H group. Insulin was raised with the H diet before mating, but this only reached statistical significance for animals raised at 20°C. Although insulin concentrations were similar between groups through pregnancy there was high variability. Corticosterone was similar between groups, although mean values were highest in the 27L group, this difference was not significant. It was low in the 27L group at 10 d/g to increase by 19 d/g. Food intake, measured just before mating, was higher in the H diet group only when animals were kept at thermoneutrality when consumption of the L diet was reduced [20L: 57 ± 3; 20H 67 ± 3; 27L: 45 ± 3; 27H: 66 ± 5 kcal/d (P < 0.05)].

Interscapular adipose mass was similar between unmated animals (Figure 1C,D), but during pregnancy it was lower in animals at thermoneutrality, in which there was a smaller rise with gestation compared with those kept at a 20°C. Additionally, at 19 d/g, 20H animals had a greater interscapular fat mass than 20L, whereas diet had no effect at thermoneutrality.

Uncoupling protein 1 protein was more abundant in unmated animals fed the H diet at 20°C than at thermoneutrality (Figure 2). In contrast, UCP1 gene expression was similar between groups and declined with gestation, and was accompanied with similar changes in PGC1α (Figure 3). There was also a higher gene expression of PGC1α in 20H compared with 27H groups prior to mating. Gene expression of leptin, a marker of white fat, was higher before, and in mid-pregnancy, in the 27H, compared with the 20H group. When comparing the ratio of UCP1 and leptin gene expression, i.e., the main genes characteristic of brown and white adipocytes, respectively, the ratio of gene expression was markedly higher in unmated animals at 20°C, and then declined to similar mean values for all groups through pregnancy. PPARγ gene expression also declined with pregnancy in the animals kept at 20°C, irrespective of diet.

Other thermogenic genes examined i.e., B3AR and SERCA2B were unaffected by pregnancy, diet, or temperature, whilst TRPV1 gene expression was higher at 10 than 19 d/g (Table 2). Gene expression of BHSD11 was consistently higher in 27H than 20H groups and higher in 27L than 20L at 19 d/g, as it rose substantially by the end of pregnancy in those dams maintained at thermoneutrality. Energy sensing genes i.e., TCFL2 and mTOR were consistently upregulated through pregnancy but were not affected by diet or temperature. IRS1 gene expression was lower at thermoneutrality at 19 d/g regardless of diet, and was downregulated with pregnancy in the 27H group, whilst IRS2 was reduced with the H diet prior to pregnancy. Subsequently, there was no difference between groups as mRNA abundance increased up to 10 d/g and was similar between all groups by term. Of the genes involved in fat transport, CD36 showed an increase in late pregnancy, but only at thermoneutrality. ATGL expression was higher in 27L than 27H, a difference that was not present at 20°C, as it was downregulated with pregnancy in the 20L group. At both environmental housing temperatures, LPL peaked at 10 d/g in those dams fed the control diet. FATP4 gene expression increased with pregnancy regardless of any intervention.

UCP1 mRNA was undetectable in the inguinal (or omental) depot. PPARG and IRS2 gene expression declined through gestation in all groups (Table 3), whilst B3AR was undetectable at the end of pregnancy. Leptin gene expression declined by mid gestation in the 20L group only, whilst FABP4 only decreased with gestation in the 27L group.

There were few differences in gene expression between groups in omental fat (Table 4). In contrast to the inguinal depot, B3AR was expressed throughout pregnancy, though declined by 19 d/g. RYR2 gene expression was also higher in unmated animals and declined with pregnancy in all groups except 20L. Gene expression of IRS2 was high before mating in the animals kept at thermoneutrality and declined during pregnancy. FATP4 gene expression increased at 10 d/g but then decreased by 19 d/g, a pattern that is also seen for FABP4 in 27L but no other groups. Expression of genes of the immune response pathways were not affected, with the exception of a reduction in MCP1 through pregnancy.