Adipocyte-Specific Modulation of KLF14 Expression in Mice Leads to Sex-Dependent Impacts on Adiposity and Lipid Metabolism

Genome-wide association studies (GWAS) identified several genetic variants on chromosome 7 upstream of the KLF14 gene to be associated with a multitude of metabolic abnormalities, including insulin resistance, type 2 diabetes (T2D), and coronary artery disease (1–4). The associations were more pronounced in women than in men (5). GWAS of gene expression performed in the Multiple Tissue Human Expression Resource (MuTHER) and the Metabolic Syndrome in Men (METSIM) cohorts identified the maternally imprinted KLF14 gene as a master regulator of gene expression in subcutaneous adipose tissue only, although KLF14 is expressed at low levels across most human tissues (6–9). GWAS-associated single nucleotide polymorphisms (SNPs) were also associated with KLF14 expression levels in cis and nearly 400 genes in trans in adipose tissue (1,6,10). The locus is one of the largest trans-expression quantitative trait loci hotspots known in the human genome (7). The results suggested that disease-associated variants act in adipose tissue to increase disease risk. However, given that adipose tissue contains a multitude of cell types, it was not clear which cells were ultimately responsible for KLF14's adverse effects.

Several studies have implicated KLF14, a single-exon gene in the Krüppel-like factor family of transcription factors, in the development of metabolic diseases, including obesity, insulin resistance, and T2D. Small et al. (1) demonstrated that lowering KLF14 expression in preadipocytes from females prevented maturation to adipocytes, as measured by the expression of adipogenesis markers and lipid accumulation. Histological analysis of subcutaneous adipose tissue biopsy specimens showed that females who were homozygous for the T2D risk allele had larger adipocytes, suggesting that lower KLF14 expression is associated with adipocyte dysfunction (1). Collectively, these studies implied that KLF14 plays roles in adipogenesis and mature adipocyte size and function. T2D-associated variants also have strong associations with HDL and triglyceride (TG) levels in the serum (11,12). However, the mechanisms by which KLF14 regulates adipocyte differentiation and function are not known.

To further investigate the role of KLF14 in adipocytes and to establish a mouse model that is relevant to humans, we characterized mice with adipocyte-specific Klf14 deletion or overexpression. Our findings demonstrate that adipocyte KLF14 is a key regulator of lipid metabolism and consistent with findings that this regulation is female-specific in humans. Our findings may facilitate the development of novel therapeutics to treat obesity and related metabolic diseases specifically in females.

The generation of adipocyte-specific Klf14-knockout (KO) mice has been described in detail elsewhere (1). Briefly, in vitro transcribed Cas9 mRNA, two guide RNAs designed to target sites upstream and downstream of the Klf14 gene locus, two single-strand DNA oligonucleotides bearing loxP sequences, along with 80-nt homology arms matching the target sites, were injected into the cytoplasm of fertilized oocytes from C57BL/6J mice (Supplementary Fig. 1A). Genomic DNA samples from founders were screened for correct loxP sequences flanking the Klf14 gene by PCR and confirmed by Sanger sequencing. Mice with two loxP sites that segregated on the same chromosome were bred with Adipoq-Cre mice on the C57BL/6J background (B6; FVB-Tg [Adipoq-Cre]1Evdr/J, The Jackson Laboratory) to obtain litters with homozygous Klf14 loxP knock-in mice that were Cre⁺ (Klf14^fl/flAdipoq-Cre⁺ or KO) and wild-type (WT) mice that were positive for Klf14 loxP but negative for Adipoq-Cre (Klf14^fl/flAdipoq-Cre⁻ or WT) for experiments. The presence of the Adipoq-Cre allele was confirmed by PCR (Supplementary Table 1). Absence of KLF14 protein expression in adipocytes isolated from KO animals was confirmed by Western blotting (Supplementary Fig. 1B). To generate mice with adipocyte-specific overexpression of Klf14, Klf14 was PCR amplified from C57BL/6J cDNA and cloned downstream of the 5.43 kilobase (kb) adiponectin (Adipoq) promoter at an EcoRV site using Gibson Assembly into the pADNpcDNA3.1 KanR vector. The 8.2-kb expression cassette containing the Adipoq promoter, Klf14 cDNA, and 3′ untranslated region was gel purified after restriction digestion with NotI and NgoMIV. The transgene was injected into pronuclei of fertilized C57BL/6N eggs to produce random integrants at the University of California, Irvine Transgenic Mouse Facility. The presence of the transgene was confirmed by PCR (primers are listed in Supplementary Table 1). We labeled these mice Adipoq-Klf14-OE. Increased KLF14 protein expression in adipocytes isolated from transgenic animals was confirmed by Western blotting (Supplementary Fig. 1C).

All animal protocols were approved by the University of Virginia Animal Care and Use Committee. Mice were maintained in a 12-h light/12-h dark cycle with free access to water and standard chow (Envigo Teklad LM-485 irradiated mouse/rat sterilizable diet, Cat No. 7912). For high-fat diet (HFD) studies, the mice were fed a rodent diet with 40% kcal fat, 43% kcal carbohydrate, and 17% kcal protein (Research Diets, Cat No. D12079B). The mice were euthanized by cervical dislocation after induction of deep anesthesia using isoflurane, consistent with the recommendations of the Panel of the American Veterinary Medical Association.

Total body fat and lean mass were assessed in conscious mice using a noninvasive quantitative MRI system (Echo Medical Systems, Cat No. EchoMRI 3-in-1 v2.1), as previously reported (13).

Glucose and insulin tolerance tests were performed as previously described (14). For the glucose tolerance test, mice were fasted from 7 a.m. to 1 p.m. and were administered glucose at 1 mg/g body weight intraperitoneally. Blood glucose levels were measured with a OneTouch Ultra Glucometer (LifeScan) in blood drops obtained from nicked tail veins before and at 10, 20, 30, 60, 90, and 120 min after the injection. The insulin tolerance test was performed at 1–2 p.m. on mice fed ad libitum. Insulin (100 units/mL; Humulin, Eli Lilly) at 0.75 units/kg body weight was administered by intraperitoneal injection of 0.25 units/mL solution in 0.9% NaCl. Blood glucose levels were measured immediately before and at 15, 30, 45, 60, and 90 min after the injection.

Mice were either fed ad libitum, and samples taken between 7 a.m. and 8 a.m., or they were fasted for 6 h from 7 a.m. to 1 p.m. before samples were taken. Whether measurements were done in the fed or fasted state is indicated in the Results section. For insulin, fatty acids, and glycerol measurements, blood samples were collected through retro-orbital bleeding or cardiac puncture. For HDL, total cholesterol, and TG measurements, plasma was obtained using heparinized capillary tubes, followed by centrifugation at 7,800 relative centrifugal force in a microfuge for 10 min at 4°C. Insulin levels were measured with an ELISA (Alpco, Cat No. 80-INSMSU-E01). Total cholesterol, HDL, and TG levels were measured using colorimetric assays (FUJIFILM Wako Diagnostics, Cat Nos. 999-02601, 997-72591, 464-01601).

For determination of energy expenditure and respiratory exchange ratios (RERs), mice were placed in Oxymax metabolic chambers (Comprehensive Laboratory Animal Monitoring System [CLAMS] from Columbus Instruments), under a constant environmental temperature (22°C) and a 12-h light, 12-h dark cycle, as previously described (15). Vo₂, Vco₂, and ambulatory activity were determined for each mouse every 5 min over a 72-h period. Experimental mice were acclimated to the cages during the first 24 h. Measurements taken during the following 48 h were used for data analysis. RERs were calculated as Vco₂/Vo₂. Carbohydrate use was calculated as (20 kJ/L × Vo₂ uptake) × [(RER − 0.7)/0.3], and fat use was calculated as (20 kJ/L × Vo₂ uptake) × (1 − [(RER × 0.7)/0.3)] (16). Both body fat and lean mass were included as covariates in the models used to analyze data obtained from the CLAMS experiments. Mice in each chamber had free access to water and food. Locomotor activity was monitored by a multidimensional infrared light beam system surrounding each cage.

To determine lipolysis in vivo, two different experiments were performed. For lipolysis under fasting condition, blood samples were drawn from tail veins of 24-week-old mice using nonheparinized glass capillary tubes at 5 p.m. before food was withdrawn and at 9 a.m. after 16-h fasting to measure nonesterified fatty acids (NEFAs) and glycerol. For β-adrenergic receptor stimulation experiments, blood samples were drawn from tail veins of random-fed 25-week-old mice via nonheparinized glass capillary tubes before stimulation. After a short recovery period (30 min), mice were injected intraperitoneally with 10 mg/kg isoproterenol (prepared in saline; Sigma-Aldrich, Cat No. I6504). A second blood sample was collected 15 min postinjection. Blood from nonheparinized capillary tubes was allowed to clot on ice before centrifugation at 500g for 20 min at 4°C. Glycerol and NEFA concentrations in serum were determined colorimetrically using commercially available reagents (Sigma-Aldrich, Cat No. F6428; Wako Life Sciences, Cat No. NEFA-HR [2]).

Isolated mature adipocytes were diluted in DMEM/F12 with 2.5% BSA to a concentration of 0.5 million cells/mL, and 200 µL with 100,000 cells was used for each reaction. To induce lipolysis, 30 µmol/L isoproterenol was added to diluted adipocytes, followed by incubation at 37°C with shaking at 150 rpm. After incubation, infranatant was removed to a new tube, and NEFA measurements were performed (Wako Life Sciences, Cat No. NEFA-HR [2]).

Experimental mice were fasted for 4 h before the test (7–11 a.m.). Olive oil (6 µL/g body weight) was given by oral gavage, as described previously (17). Blood samples were drawn via heparinized (for TG)/nonheparinized (for NEFA) capillary tubes from tail veins before (time 0) and at 1, 2, 3, and 4 h after gavage. TG and NEFA measurements were performed as described above.

Subcutaneous white adipose tissue (sWAT) and parametrial-periovarian/epididymal (pWAT/eWAT) from four mice of each sex and genotype was excised and fixed in 10% neutral buffered formalin (18). Tissues were embedded in paraffin, and 5-µm-thick sections were prepared and stained with hematoxylin and eosin at the University of Virginia Research Histology Core. Slides were scanned, and images were acquired with an EVOS microscope camera. Quantification of adipocyte size was done using ImageJ 1.48 (19). Images were processed as described previously (20).

Mouse sera collected at the time of euthanasia were stored at −80°C. After thawing, they were allowed to equilibrate at room temperature and spun to remove aggregates. Then, 100 μL of 1:10 diluted sera per well were aliquoted for ELISA in 96-well plates. Free testosterone, progesterone, and estradiol levels were measured using the Testosterone Parameter assay kit (R&D Systems, Cat No. KGE010), Progesterone ELISA Kit (Cayman Chemical, Cat No. 582601), and Estradiol ELISA Kit (Cayman Chemical, Cat No. 501890), following the manufacturers’ instructions (21).

Adipocytes were isolated from sWAT and pWAT/eWAT, as previously described (22). Immediately after euthanization, fat pads were removed, placed in Krebs Ringer HEPES (KRH)–BSA buffer containing collagenase type I (1 mg/mL, 2 mg/g of tissue; Worthington Biochemical Corp.) and minced with scissors. Small tissue pieces were incubated in a 37°C shaking water bath (100 rpm) for 1 h. Fat cells were separated from nonfat cells and undigested debris by filtration through a 0.4-mm Nitex nylon mesh (Tetko) and four washes by flotation with KRH buffer without BSA. After centrifugation at 1,000g for 15 min at 4°C, the isolated fat cells floating on the surface were used for total RNA extraction and preparation of cell lysates for immunoblotting.

Total RNA was isolated from parametrial-periovarian/epididymal and subcutaneous adipocytes using a combination of TRIzol Reagent (Thermo Fisher Scientific, Waltham, MA) and miRNeasy kit (Qiagen, Hilden, Germany) according to the manufacturers' instructions (23). Adipocytes isolated from 200 mg of tissue in the previous step were homogenized in 2 mL of TRIzol Reagent using the Tissue-Tearor homogenizer (Model 985370, BioSpec Products). After homogenization, samples were incubated at room temperature for 5 min and centrifuged at 12,000g for 10 min at 4°C. The resulting fat monolayer on top was carefully avoided when pipetting the rest of the sample into a clean tube. Then, 400 μL of chloroform was added, and samples were mixed by vortexing. Samples were kept at room temperature for 3 min before centrifugation at 12,000g for 30 min at 4°C. After centrifugation, samples separated into three phases with the RNA in the upper phase. The RNA phase was transferred to a new tube without disturbing the interphase. Sample volumes were measured and 1.5× sample volumes of 100% ethanol added. Samples were mixed thoroughly by inverting tubes several times. Samples were then loaded on miRNeasy spin columns (Qiagen), and the manufacturer’s protocol was followed for subsequent steps. Total RNA concentrations were quantified using a Qubit Fluorometer and a Qubit RNA BR assay kit (Thermo Fisher Scientific). Then, 1 µg of total RNA was reverse transcribed using SuperScript IV reverse transcriptase (Invitrogen). Real-time PCR was performed in the QuantStudio 5 Real-Time PCR System (Thermo Fisher Scientific) using SYBR Green Master Mix (Roche) and gene-specific primers (Supplementary Table 2).

HEK293 cell line was from ATCC (Manassas, VA) and cultured in DMEM with 10% FBS (Hyclone), 1% l-glutamine, 100 units/mL penicillin, and 100 mg/mL streptomycin at 37°C and 5% CO₂. The pCMV-KLF14-3×FLAG and pCMV-ESR1-HA vectors were purchased from GeneCopoeia (Cat No. EX-I2219-M12 and EX-A0322-M06). These two constructs were transfected into HEK293 cells using Lipofectamine 3000 (Invitrogen) according to the manufacturer’s instructions. Stable cell lines were selected with G418 at 500 µg/mL.

Total protein was extracted from cells or tissues using cell lysis buffer (Cell Signaling, Cat No. 9803), containing 20 mmol/L Tris-HCl (pH 7.5), 150 mmol/L NaCl, 1 mmol/L Na₂EDTA, 1 mmol/L EGTA, 1% Triton, 2.5 mmol/L sodium pyrophosphate, 1 mmol/L β-glycerophosphate, 1 mmol/L Na₃VO₄, and 1 µg/mL leupeptin. Protease inhibitor cocktail (Thermo Fisher Scientific) and phosphatase inhibitor (Thermo Fisher Scientific) were added to the lysis buffer before extraction. Protein concentrations were determined using a Pierce BCA Protein Assay Reagent (Pierce Biotechnology). Proteins (10 μg protein/lane) were separated by SDS-PAGE. Electrophoresis was conducted using a XCell II Mini-Cell with 4–12% NuPAGE Tris-acetate gradient SDS-polyacrylamide gels (Invitrogen). Protein samples were transferred onto polyvinylidene difluoride membranes using an XCell II Blot Module (Invitrogen). The membranes were blocked in Intercept (TBS) Blocking Buffer (LI-COR) for 1 h at room temperature, and probed with primary antibody in 0.1% Tween LI-COR blocking buffer overnight at 4°C. The KLF14 antibody was a gift from Dr. Roger D. Cox (MRC Harwell Institute). It was generated at Eurogentec by immunizing rabbits against the C-DMIEYRGRRRTPRIDP-N peptide. Full-length purified KLF14 purchased from Creative BioMart (Cat No. KLF14-481H) served as a positive control. The β-actin antibody was from Cell Signaling Technology (Cat No. 3700s). For measuring hormone-sensitive lipase (HSL) phosphorylation, the antibodies used were phosphorylated (p)HSL (Ser660) antibody and HSL antibody (Cell Signaling Technology, Cat No. 4126S and 4107S). The membranes were washed three times for 10 min each in Tris-buffered saline–0.1% Tween buffer solution and then probed with IRDye-labeled secondary antibodies (926-32211 and 926-68072) in 0.1% Tween LI-COR blocking buffer for 1 h at room temperature. Several washes in Tris-buffered saline–0.1% Tween buffer solution were repeated after labeling with secondary antibodies. The blots were scanned using the LI-COR laser-based image detection method. For immunoprecipitation, cell lysate was incubated with Anti-FLAG M2 Magnetic Beads (MilliporeSigma Cat No. M8823) following the manufacturer’s instruction. The associated proteins were analyzed by immunoblotting using estrogen receptor-α antibody (Santa Cruz Biotechnology, Cat No. sc-8002), HA antibody (Abcam, Cat No. ab9110), FLAG M2 antibody (MilliporeSigma, Cat No. F1804), and KLF14 antibody described above.

Trial experiments were used to determine sample size with adequate statistical power. Since the scatter of the data follows normal distribution, all analyses were conducted with the Student t test with a two-tail distribution. Comparisons with P values <0.05 were considered significant. Effect size (β) was calculated by subtracting the mean difference between two groups, and dividing the result by the pooled SD: β = (M₂ − M₁)SD_pooled, SD_pooled = √([SD₁² + SD₂²]2), where M is the mean.

All data generated or analyzed during this study are included in the published article (and its online Supplementary Material).

Adipocyte-specific Klf14-KO (Klf14^fl/flAdipoq-Cre⁺ or KO) mice were born at expected Mendelian ratios and were indistinguishable from wild-type (Klf14^fl/flAdipoq-Cre⁻ or WT) littermates at birth. At 8 weeks of age, mice were placed on a 45% HFD (Research Diets D12079B). We measured body weights over a 12-week period (Fig. 1A). Female Klf14^fl/flAdipoq-Cre⁺ mice showed similar body weight increases as WT littermates. However, male Klf14-deficient mice had lower body weight starting at 8 weeks on the HFD (with 8% difference after 12 weeks on HFD, P = 0.018) (Fig. 1B, left). In humans, lower KLF14 expression is associated with higher visceral fat mass (1). To determine whether loss of KLF14 in adipocytes affected adiposity on the HFD, we measured body composition using EchoMRI (Fig. 1A). We found higher fat mass relative to lean mass in female Klf14^fl/flAdipoq-Cre⁺ mice, while in male mice, we observed the opposite (Fig. 1B, middle and right). When determining individual organ weights normalized to tibia length, we found no differences between WT and Klf14-deficient mice for most organs, including brain, heart, kidney, muscle, spleen, and pancreas (Supplementary Fig. 2). However, we observed that parametrial-periovarian fat depots in female Klf14-deficient mice were 70% larger (β_perigonadal = 1.105, P_perigonadal = 0.02), whereas subcutaneous and epididymal fat depots of male Klf14-deficient mice were 33% and 27% smaller compared with WT littermates (β_subq = −1.073, P_subq = 0.01; β_epididymal = −1.006, P_epididymal = 0.01) (Fig. 1C). Notably, adipocyte Klf14 deficiency resulted in decreased liver weight in males and a trend toward a decrease in females (β_male = −0.626, P_male = 0.05; β_female = −0.639, P_female = 0.18) (Fig. 1C).

In humans, lower KLF14 expression in adipose tissue is associated with higher serum insulin and lower HDL levels, with a stronger effect in females (1). Consistent with this, we observed 78% higher insulin and 25% lower HDL levels in Klf14-deficient female but not in male mice after a 6-h fast (β_{female, insulin} = 0.947, P_{female, insulin} = 0.036; β_{female, HDL} = 1.352, P_{female, HDL} = 0.004) (Fig. 1D and E). Total cholesterol and NEFA were lower by 19% and 26%, while TG levels were 10% higher in female mice with adipocyte Klf14 deficiency (β_{female, cholesterol} = 1.176, P_{female, cholesterol} = 0.011; β_{female, NEFA} = 1.173, P_{female, NEFA} = 0.031, β_{female, TG} = 1.250, P_{female, TG} = 0.044) (Fig. 1F–H). Mutant male mice also had lower NEFA levels (β_{male, NEFA} = 1.022, P_{male, NEFA} = 0.01) (Fig. 1G) but normal serum cholesterol and TG levels (Fig. 1F and H).

The parametrial-periovarian fat depot in female mice and the epididymal fat depot in male mice are considered visceral fat. Larger visceral fat depots with bigger adipocytes are a significant risk factor for T2D, cardiovascular disease, and hypertension (24,25). Adipocyte size is increased in obese conditions (26,27). To test whether adipocyte size was altered as a result of Klf14 deficiency, we performed histological analysis using hematoxylin and eosin staining of subcutaneous and perigonadal fat depots (Fig. 2A). We observed that adipocyte size in female mice with adipocyte deletion of Klf14 was 25% smaller in the subcutaneous depot (β_subq = −0.422, P_subq = 4.02 × 10⁻⁸⁵) and 70% larger in the parametrial-periovarian depot relative to WT mice (β_periovarian = 0.706, P_periovarian = 2.25 × 10⁻¹⁷⁵) (Fig. 2B). In contrast, adipocyte size in male mice with adipocyte deletion of Klf14 was 6% larger in the subcutaneous depot (β_subq = 0.09, P_subq = 1.66 × 10⁻³) and 14% smaller in the epididymal depot relative to WT mice (β_epididymal = −0.233, P_epididymal = 2.66 × 10⁻¹⁵) (Fig. 2B). These findings suggest that KLF14 regulates adipocyte size and adipose tissue mass in a depot-specific and sex-specific manner.

Adipose tissues regulate systemic glucose metabolism and insulin sensitivity (28,29). Genetic studies showed that T2D risk alleles associated with lower KLF14 expression in adipose tissue led to higher fasting insulin in humans, indicating decreased insulin sensitivity (1,5). As described above, we also observed increased insulin levels in female Klf14-deficient mice (Fig. 1D). To further explore whether the differences in adipose tissue mass as a result of Klf14 deletion in adipocytes affected systemic glucose homeostasis and insulin sensitivity, we performed glucose and insulin tolerance tests with intraperitoneal injections of glucose or insulin after 7 and 9 weeks of the HFD, respectively. Mice with adipocyte-specific Klf14 deletion had a similar response to an intraperitoneal glucose bolus as WT littermates (Fig. 3A). However, in the insulin tolerance test, female mice with adipocyte-specific Klf14 deletion displayed increased glucose levels compared with WT littermates, with 19% higher calculated area under the curve (AUC) (β_AUC = 1.233, P_AUC = 0.005) (Fig. 3B) indicating increased insulin resistance. Insulin sensitivity for male mutant mice was similar to WT (β_AUC = 0.027, P_AUC = 0.943) (Fig. 3B).

Increased fat mass in female mice with Kl14 deficiency suggested a defect in energy balance. To investigate the role of KLF14 in energy homeostasis, we performed indirect calorimetry studies using metabolic cages. Food consumption was comparable between Klf14-deficient and WT mice (Supplementary Fig. 4), and mice of both genotypes maintained body weights within 5–10% of their initial weight during the 3-day period in the metabolic cages. Despite these similarities, female Klf14^fl/flAdipoq-Cre⁺ mice showed a 38% decrease in locomotor activity (β_{locomotor activity} = 2.212, P_{locomotor activity} = 5.2 × 10⁻⁴) compared with WT mice during the dark cycle (Fig. 4A). Vo₂ was lower in both the light and dark cycles by 26% and 19%, respectively (β_{O2, dark} = 2.320, P_{O2, dark} = 0.003; β_{O2, light} = 1.569, P_{O2, light} = 0.0095), whereas CO₂ release was 14% lower in the light cycle and trended toward a decrease in the dark cycle (β_{CO2, dark} = 0.900, P_{CO2, dark} = 0.093; β_{CO2, light} = 1.262, P_{CO2, light} = 0.026) (Fig. 4B and C). The calculated RER (RER = V̇co₂/V̇o₂), an indicator for the metabolic energy source (carbohydrate or fat), was higher (close to 1) for female Klf14^fl/flAdipoq-Cre⁺ mice than for WT littermates (β_{RER, dark} = 4.258, P_{RER, dark} = 1.0 × 10⁻⁶) in the dark cycle (Fig. 4D), suggesting predominant use of carbohydrates as an energy source, despite being fed the HFD. The indirect calorimetry measurements allowed us to estimate consumed substrates using the gas exchange ratios as a surrogate for substrate use (16). Female Klf14^fl/flAdipoq-Cre⁺ mice had 82% higher carbohydrate use (β_{carbohydrate, dark} = 2.015, P_{carbohydrate, dark} = 0.01) and 79% decreased fat use (β_{fat, dark} = −5.400, P_{fat, dark} = 1.1 × 10⁻⁶) in the dark cycle compared with WT mice. We did not observe differences for any of the measured parameters for male mice in dark or light cycles (Fig. 4).

We reasoned that defects in lipid breakdown (lipolysis) or lipid uptake and synthesis (lipogenesis) in adipocytes could explain the shift in energy source for female mice. Compatible with these possibilities, circulating NEFA and TG levels were decreased and increased under fasting conditions, respectively, in female mutant mice (Figs. 1G and H). To follow-up on these changes, we first measured lipolytic products, NEFAs, and glycerol after 16 h of fasting. NEFA levels increased as a result of fasting, as expected (Fig. 5A). However, female Klf14^fl/flAdipoq-Cre⁺ mice had 17% and 38% lower NEFA levels compared with WT littermates in fed and fasted states (β_fed = −1.416, P_fed = 0.013; β_fasted = −3.231, P_fasted = 1 × 10⁻⁴) (Fig. 5A), with NEFA levels in Klf14-deficient female mice increasing to a lesser extent compared with WT littermates in response to fasting (β = −1.292, P = 0.021) (Fig. 5B). Glycerol in female mice was only decreased in the fed state (β_fed = 2.157, P_fed = 7 × 10⁻⁴) (Fig. 5C and D). No differences in NEFA and glycerol were observed in male mice, except that NEFAs under fed conditions were slightly increased (P = 0.036) (Fig. 5).

Adipocyte lipolysis is regulated by hormonal and neuronal stimulation (30), with β-adrenergic receptor signaling potently increasing lipolysis (31). To further corroborate that female mice with Klf14-deficient adipocytes indeed had a defect in lipolysis, we quantified serum levels of NEFA and glycerol in response to isoproterenol that stimulates β-1 and β-2 adrenergic receptors (32). Female Klf14^fl/flAdipoq-Cre⁺ mice had 21% and 22% lower NEFA levels compared with WT littermates under the fed condition in nonstimulated and stimulated states, respectively (β_{nonstimulation} = −2.692, P_{nonstimulation} = 9 × 10⁻⁵; β_stimulated = −2.699, P_stimulated = 8 × 10⁻⁵) (Fig. 5E), with NEFA levels in Klf14-deficient female mice increasing to a lesser extent than in WT littermates in response to stimulation (β = −1.022, P = 0.06) (Fig. 5F). No differences were observed in male mice (Fig. 5E and F). Similar to fasting experiments, we observed a difference in the serum glycerol level in female mice before stimulation (β_female = −0.307, P_female = 0.021) but not after (β_female = −0.301, P_female = 0.559) (Fig. 5G). Changes in serum glycerol levels upon stimulation did not differ between genotypes in both female and male mice (β_female = 0.815, P_female = 0.166; β_male = −0.031, P_male = 0.951) (Fig. 5H). The difference in NEFA and glycerol responses to induced lipolysis is most likely due to different metabolic fates of NEFA and glycerol after release into the circulation.

Given that the adipocyte size changed in opposite directions in subcutaneous and visceral depots, we tested whether differences in cell size are caused by differences in lipolysis. We assessed lipolysis with and without isoproterenol stimulation ex vivo using adipocytes isolated from subcutaneous and visceral depots. We observed that female Klf14-deficient compared with WT adipocytes showed decreased lipolysis irrespective of whether they were derived from subcutaneous or visceral depots (Fig. 5I and J). However, nonstimulated lipolysis was more drastically reduced in visceral than in subcutaneous Klf14-deficient adipocytes. There was no significant difference between Klf14-deficient and WT adipocytes from male mice (Fig. 5I and J). This suggests that the difference between subcutaneous and visceral adipocyte size is caused by other processes, such as differential changes in lipid uptake and synthesis (see below and Fig. 6F and G).

TG hydrolysis to NEFA and glycerol is achieved in a three-step enzymatic pathway (Fig. 5I). We measured the expression of adipocyte TG lipase (ATGL), the first enzyme in the TG breakdown pathway, and found that mRNA levels of Pnpla2 in adipocytes isolated from subcutaneous and parametrial-periovarian/epididymal fat were decreased by 24% and 29%, respectively, as a result of Klf14 deficiency in female mice (β_sWAT = −9.320, P_sWAT = 5 × 10⁻³; β_pWAT = −13.139, P_pWAT = 1 × 10⁻⁴) (Fig. 5J), but were normal in male mice. We also measured HSL, the second enzyme in the TG breakdown pathway, but did not observe differences in Lipe mRNA expression in adipocytes of male or female mice (Fig. 5K). HSL activity is predominantly regulated by phosphorylation (33). We thus measured pHSL levels in adipocytes isolated from subcutaneous and visceral depots following isoproterenol stimulation. We observed that pHSL was reduced in Klf14-deficient adipocytes isolated from visceral and subcutaneous depots of both female and male mice (Fig. 5N and O). Lower Pnpla2 expression together with decreased pHSL is consistent with the above-described decreased lipolysis and increased fat mass and adipocyte size of visceral adipocytes in female mutant mice. Normal Pnpla2 expression in male Klf14-deficient adipocytes is associated with normal levels of lipolysis despite decreased pHSL. Visceral adipocytes are normally more lipolytically active compared with subcutaneous adipocytes (34).

Our metabolic cage data indicated that female Klf14^fl/flAdipoq-Cre⁺ mice were not able to efficiently use dietary fatty acids, leading us to speculate that there was also a defect in lipid uptake. Therefore, we challenged mice with a bolus of olive oil by oral gavage and measured TG and NEFA levels to quantify lipid clearance from the bloodstream. We observed that blood TG and NEFA clearance were both decreased in female mutant mice compared with WT mice (β_{4 h-TG} = 4.610, P_{4 h-TG} = 7.2 × 10⁻⁹; β_{4 h-NEFA} = 3.528, P_{4 h-NEFA} = 8 × 10⁻⁶) (Fig. 6A and B), consistent with decreased use of lipids as an energy source in female Klf14-deficient mice. No differences were observed in male mice (Fig. 6C and D). These results are consistent with our metabolic cage data that female mice use less fat as their energy source.

Uptake of fatty acids into cells is achieved predominantly via a saturable protein-facilitated process (35) (Fig. 5I). Several proteins facilitate the uptake of fatty acids, among which the best characterized are CD36, FABPs, and FATPs (36–38). Fatty acids are also synthesized in cells through de novo lipogenesis that is catalyzed by ATP-citrate lyase (encoded by Acly) (39), acetyl-CoA carboxylase (encoded by Acaca and Acacb) (40), and fatty acid synthase (encoded by Fasn) (41). Fatty acids are converted into fatty acyl-CoAs by acyl-CoA synthetase (encoded by Acsl) and then esterified with glycerol to form TGs, with the last step catalyzed by diglyceride acyltransferase (encoded by Dgat) (42). We assayed several key enzymes in adipocyte fatty acid uptake and lipogenesis and observed sex- and depot-specific differences in Fabp4, Fatp4, and Dgat1 mRNA expression in isolated mature adipocytes. Fabp4 was 36% lower in the parametrial-periovarian depot of female Klf14-deficient mice (β_{female pWAT} = −10.750, P_{female pWAT} = 3 × 10⁻⁴) (Fig. 6E), while Fatp4 (Fig. 6F) and Dgat1 (Fig. 6G) were 20% and 54% higher in the parametrial-periovarian depot of the same mice compared with WT controls (βf_{emale pWAT} = 14.336, P_{female pWAT} = 8 × 10⁻⁵). We also observed differences in other pathway genes that were either sex- or depot-specific. mRNA levels for Fatp1 were decreased in both female sWAT and pWAT/visceral WAT, Dgat2 was decreased in male sWAT and eWAT/visceral WAT. We also observed that Fabp5 was decreased in both female and male sWAT and visceral WAT (Supplementary Fig. 4). Acaca, Acacb⁻, and Fasn showed no differences. (Supplementary Fig. 4).

Our results showed that KLF14 deficiency resulted in an adverse metabolic phenotype in female mice. We therefore hypothesized that overexpression of Klf14 may be metabolically beneficial. To address this, we generated a transgenic mouse with Klf14 overexpression under the regulation of the Adipoq promoter (Adipoq-Klf14-OE) (see Research Design and Methods section for a detailed description). This resulted in 1.7-fold induction of KLF14 in adipocytes (Supplementary Fig. 1C). We fed these mice the HFD for 18 weeks and performed body composition analysis using EchoMRI. Male and female mice from both genotypes gained similar amounts of body weight, and no differences in body weight between the Klf14-OE mice (Adipoq-Klf14-OE) and WT littermates were observed at the end of the 18-week diet challenge (Fig. 7A). However, starting at 12 weeks of the HFD, female Adipoq-Klf14-OE mice showed a trend toward decreased fat mass–to–lean mass ratios compared with WT littermates (P = 0.074) (Fig. 7B). At the end of the study, after 18 weeks of the HFD, female mice with Klf14 overexpression in adipocytes had 38% lower body fat compared with WT littermates (β_female = −1.446, P_female = 0.002) (Fig. 7B). No differences were observed in male mice with Klf14 overexpression (P = 0.99) (Fig. 7C).

Sex hormones play a significant role in sex-specific body fat distribution, as evidenced by the fact that android and gynoid fat distribution patterns in males and females appear as early as puberty (43). Therefore, we measured the circulating levels of testosterone, progesterone, and estradiol after 29 weeks of HFD feeding. There was no significant difference between female and male mice with Klf14-deficient adipocytes and WT mice (Supplementary Fig. 5A), suggesting that the sex differences we observed in mice are likely hormone independent. It is, however possible that differential expression of sex hormone receptors or sex-specific interactions of KLF14 with sex hormone receptors could play a role in transcriptional target expression since other KLF family transcription factors have been shown to co-bind DNA with the estrogen receptor (44). Therefore, we tested the hypothesis that KLF14 interacts with estrogen receptor-α (ESR1). We overexpressed FLAG-tagged KLF14 (KLF14-FLAG) and HA-tagged ESR1(ESR1-HA) in HEK293 cells and performed immunoprecipitations with anti-FLAG antibody. Untransfected cells or cells that were only transfected with one of the two vectors served as negative controls. We blotted the immunoprecipitated samples with anti-HA or anti-ESR1 antibodies, respectively. We detected the presence of ESR1 protein in the sample that overexpressed both KLF14 and ESR1 but not in negative control samples, suggesting that KLF14 can bind ESR1 (Supplementary Fig. 5B and C).

In summary (Fig. 8), Klf14 deficiency in adipocytes caused major abnormalities in HFD-fed female but not male mice. Female mice with Klf14-deficient adipocytes had increased total body fat with larger visceral fat mass and bigger visceral adipocytes, but they had smaller subcutaneous adipocytes. Concomitant with the changes in body fat distribution and adipocyte size, the mice demonstrate metabolic abnormalities, including insulin resistance and altered serum lipid levels. Despite HFD feeding, Klf14-deficient female mice used carbohydrates as the predominant source of energy and simultaneously had lower energy expenditure and activity. Further evaluation of lipid metabolism uncovered defects in fasting- and β-adrenergic–induced lipolysis, and impaired TG and NEFA clearance after lipid gavage. These defects were coupled with decreased mRNA expression of Pnpla2 (a key lipolytic enzyme) in both perigonadal/visceral and subcutaneous adipocytes and decreased pHSL. Furthermore, mRNA levels for proteins involved in cellular fatty acid uptake, including Fabp4, Fabp5, and Fatp1, were decreased in both sWAT and pWAT/visceral WAT, but mRNA levels of the fatty acid transport protein Fatp4 and Dgat1, an enzyme that catalyzes the conversion of DAG into TG, were increased only in perigonadal/visceral adipocytes. In contrast, male mice with Klf14-deficient adipocytes fed the HFD gained less body weight and had decreased total body fat with smaller epididymal/visceral and subcutaneous fat depots and adipocytes, but had, relative to WT, normal metabolic parameters and insulin sensitivity. The male mice did not show defects in stimulated lipolysis (despite decreased pHSL in pWAT) and lipid clearance, and mRNA levels of enzymes involved in cellular lipid metabolism were not different from WT littermates, except that Fatp1 was increased in sWAT, and Fabp5 and Dgat2 decreased in sWAT and eWAT/visceral WAT. In conclusion, our findings suggest that KLF14 regulates adipose tissue mass and adipocyte size in a depot-specific and sex-specific manner, most likely by differentially modifying adipocyte lipid metabolism.

Visceral adipose tissue expansion independent of overall adiposity is associated with an increased risk for developing T2D and cardiovascular disease (45). How deposition of excess calories into the different adipose depots is regulated is currently unknown. However, sex-dependent differences in adipose tissue distribution are well documented, with higher adiposity in women than men (46). Men usually display “android” or “apple shape” distribution, with more visceral fat in the abdominal region, while women have “gynoid” or “pear shape” distribution, with more subcutaneous fat and less visceral fat in the lower body (47). A shift in fat deposition from subcutaneous depots around the hip to visceral depots in the waist region in men confers increased risk for metabolic disorders (48). However, women can harbor increased upper-body fat without enhanced metabolic disease risk as excess fat is mainly stored in subcutaneous depots. Our results suggest that the transcription factor KLF14 plays a role in modifying adipocyte function and thereby adipose tissue distribution in a sex- and depot-specific manner. Deletion of the Klf14 gene in mature adipocytes resulted in an increase of visceral fat in female mice but a decrease in male mice. The shift in lipid storage from subcutaneous to parametrial-periovarian (visceral) depots in female mice with adipocyte Klf14 deletion was characterized by changes in adipocyte size, smaller adipocytes in subcutaneous and larger adipocytes in perigonadal/visceral adipose tissues, and major abnormalities in lipid metabolism.

Identifying the molecular mechanisms of how KLF14 affects adipocyte function in a sex-specific manner remains a challenge. Although we did not observe sex hormone differences at the end of our experiments, we cannot rule out the possibility that earlier differences during development exist. Further, we showed that when we overexpressed KLF14 and ESR1 in HEK293 cells, they can co-bind. However, this result needs to be verified for endogenous proteins in adipocytes, especially since HEK293 cells do not express ESR1.

The absence of Klf14 in adipocytes causes significant abnormalities in lipid metabolism, both lipolysis and fatty acid uptake in female mice. Whether these are primary consequences of KLF14 deficiency or secondary metabolic abnormalities due to changes in adipose tissue mass and distribution and adipocyte size is not entirely clear. Future identification of the primary consequences of KLF14 deficiency on metabolism and primary KLF14 targets will require the careful analysis of metabolic phenotypes at various time points before and after the transition to a HFD. Abnormalities in circulating lipids in female mice with Klf14 deficiency in adipocytes can partly be explained by the increased insulin resistance in female mice with Klf14-deficient adipocytes. This includes the lower circulating HDL-cholesterol (HDL-C) and increased plasma TG levels under fasting condition, and impaired TG and NEFA clearance after lipid gavage. TG clearance requires the action of lipoprotein lipase in adipose tissues, and insulin-stimulated lipoprotein lipase activity is impaired in insulin resistance. Insulin-stimulated NEFA uptake under fed conditions is also defective in insulin resistance. Furthermore, expression of fatty acid transporters is decreased in the presence of impaired insulin sensitivity (49). However, the lower-than-normal NEFA levels in female mice under fed and fasting conditions and in response to β-adrenergic stimulation cannot be attributed to impaired insulin sensitivity. On the contrary, NEFA levels should be increased due to defective suppression of lipolysis by insulin. Currently, we do not know the mechanisms by which Klf14 deficiency in adipocytes impairs lipolysis. But impaired lipolysis in female mice is accompanied by a decrease in Pnpla2 mRNA and decreased isoproterenol-induced pHSL in adipocytes of both sWAT and pWAT. Decreased pHSL in male Klf14-deficient adipocytes alone was not sufficient to decrease lipolysis when Pnpla2 mRNA was normal. Future studies will establish whether Pnpla2 is a primary target of KLF14 and whether transcriptional Pnpla2 regulation is sex specific.

HSL was once thought to be the key enzyme that regulates cellular TG breakdown. Its activity is modulated by phosphorylation (33) However, Hsl-deficient mice had a normal rate of free fatty acid production, indicating that other enzymes are playing a key role in TG breakdown (50). Haemmerle et al. (51) identified ATGL as the rate-limiting enzyme in the mobilization of cellular TG. Notably, they reported that mice with inactivated ATGL displayed deleterious metabolic phenotypes similar to those of our female mice with Klf14-deficient adipocytes. These include increased adipose tissue mass, decreased insulin sensitivity, impaired energy metabolism, and decreased availability of free fatty acids (51). The latter fostered the use of carbohydrates as the primary fuel despite the presence of increased amounts of fat in the adipose tissue. Reduced TG hydrolysis then led to reduced energy expenditure and total activity. This study supports the notion that defective lipolysis in mice with Klf14-deficient adipocytes leads to a switch to carbohydrates as the predominant source of energy, decreased energy expenditure and activity, and increased fat mass gain. The fact that glucose tolerance is not impaired in mice with Klf14-deficient adipocytes despite impaired insulin sensitivity may be explained by the energy substrate switch to glucose and decreased cellular NEFA uptake facilitating instead glucose uptake into tissues. The decreased locomotor activity in female Klf14-deficienct mice is possibly attributed to decreased lipolysis and decreased lipid uptake as well, which may mean that they do not have enough nutrients to use as an energy source.

The increased size of parametrial/visceral but normal size of subcutaneous adipocytes in females with adipocyte Klf14 deficiency could, at least in part, be explained by the more drastically reduced nonstimulated lipolysis in pWAT-derived adipocytes. In addition, other processes, such as increased expression of Dgat1 in parametrial/visceral adipocytes, may further contribute to increased cell size. Mice that overexpress Dgat1 in adipocytes have increased deposition of TG in WAT (52). Although fatty acid uptake is impaired in adipocytes, as decreased expression of key fatty acid binding/transporter proteins and impaired NEFA clearance suggest, fatty acids taken up could be efficiently esterified to TG due to the increase in Dgat1. Increased Dgat1 expression cannot be explained by increased insulin resistance. In contrast, Dgat1 expression positively correlates with insulin sensitivity in human subjects (53). Specific transcription factors regulating Dgat1 expression have not been identified, but peroxisome proliferator–activated receptor-γ is a candidate (54). Whether KLF14 directly or indirectly controls Dgat1 expression will need to be investigated.

Interestingly, male mice, despite decreased fat mass and smaller adipocytes, did not show improved metabolic parameters and greater insulin sensitivity compared with WT littermates. It is possible that simultaneous changes in fatty acid transporter/binding proteins in the smaller adipocytes are leading to unfavorable changes that are, however, offset by improved insulin sensitivity.

Previous mouse studies showed conflicting results for KLF14 on metabolic phenotypes and atherosclerosis. We recently reviewed these studies (5), several of which only used male mice to study the impact of KLF14 deletion or overexpression, a consequential omission given that in humans and in our study, KLF14 plays sexual-dimorphic roles. Some of the studies also used different diet compositions, which may be another source for conflicting results.

Klf14 adipocyte-specific KO mice have been partially characterized before by Small et al. (1). Similar to our findings, female Klf14-deficient mice had decreased HDL-C and impaired insulin sensitivity. However, Small et al. also observed impaired insulin sensitivity in male Klf14-deficient mice, contrasting with normal insulin sensitivity in our male mice. Furthermore, glucose tolerance was normal in our Klf14-deficient mice, while Small et al. showed defects in both male and female mice. The differences in results could be attributed to several factors. First, different diets were used. Small et al. used a regular chow diet while we fed the mice an HFD. Second, glucose tolerance and insulin tolerance tests were assessed differently. For the glucose tolerance test, Small et al. fasted mice overnight, while we fasted them for 6 h. For the insulin tolerance test, Small et al. used insulin at 1 unit/kg body weight following 4 h of fasting while we used it at 0.75 units/kg body weight without fasting. Our more thorough characterization, including metabolic cage studies and lipolysis assays, then uncovered the clear differences between male and female Klf14-deficient mice that we report in here.

Finally, we showed a metabolically favorable consequence of increased expression of adipocyte KLF14: female Klf14-OE mice accumulated less body fat in response to the HFD challenge. Although a comprehensive characterization will be required to further flesh out the phenotype of these mice, results from this initial assessment is of therapeutic interest. Previous studies showed that perhexiline, an approved therapeutic small molecule presently in clinical use to treat angina and heart failure, induced KLF14 expression and reduced atherosclerosis (55). Our results, together with the results from this previous pharmacological study, suggest that therapeutic targeting of KLF14, specifically in adipose tissue, could serve as a novel approach to improve metabolic abnormalities in obesity and related metabolic diseases specifically in females.

Acknowledgments. The authors thank the members of the Civelek laboratory for their feedback and discussion, Dr. Stephen B. Abbott (University of Virginia) for assistance in indirect calorimetry analysis, and Dr. Michael M. Scott (University of Virginia) for the assistance in EchoMRI.

Funding. This work was supported by National Institute of Diabetes and Digestive and Kidney Diseases (R01 DK118287 to M.C.) and by the American Diabetes Association (1-19-IBS-105 to M.C.).

Duality of Interest K.M. is an advisor to and holds equity in Verve Therapeutics and Variant Bio. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. Q.Y., S.R.K., and M.C. drafted the article. Q.Y. and M.C. conceived the study, J.H., J.N.R., R.A., and Z.X. participated in data collection and analysis. T.E.H., E.J.S., and S.R.K. contributed to the experimental design and data interpretation. K.M. generated Klf14^fl/flAdipoq-Cre⁺ mice. M.C. directed the study. All authors edited the final article. M.C. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the Don S. Fredrickson Lipid Research Conference, University of Kentucky, Louisville, KY, 9–11 September 2021.