Sex‐specific effect of AQP9 deficiency on hepatic triglyceride metabolism in mice with diet‐induced obesity

Studies in obese rats and human cell models of non‐alcoholic fatty liver disease have indicated that knockdown of the hepatic glycerol channel aquaporin 9 (AQP9) leads to decreased hepatic steatosis. However, a study in leptin receptor‐deficient mice did not find that knockout (KO) of AQP9 alleviated hepatic steatosis. The aim of this study was to investigate the effect of high‐fat diet (HFD) on hepatic glycerol and triglyceride metabolism in male and female AQP9 KO mice. Male and female AQP9 KO mice and wild‐type (WT) littermates were fed a HFD for 12 weeks. Weight, food intake and blood glucose were monitored throughout the study and tissue analysis included determination of hepatic triglyceride content and triglyceride secretion. The expression of key molecules for hepatic glycerol and triglyceride metabolism was evaluated using qPCR and western blotting. AQP9 KO and WT mice demonstrated a similar weight gain throughout the study period, and we found no evidence for AQP9 deficiency being associated with a reduced hepatic accumulation of triglyceride or a reduced blood glucose level. Instead, we show that the effect of AQP9 deficiency on hepatic lipid metabolism is sex‐specific, with only male AQP9 KO mice having a reduced hepatic secretion of triglycerides and an elevated expression of peroxisome proliferator‐activated receptor α. Male AQP9 KO mice had an elevated blood glucose level after 12 weeks of HFD when compared to baseline levels. Thus, we found no evidence for AQP9 inhibition being a target for alleviating the development of hepatic steatosis in mice with diet‐induced obesity.


Introduction
Non-alcoholic fatty liver disease (NAFLD) is closely associated with obesity and insulin resistance (Fabbrini et al., 2009;Korenblat et al., 2008).NAFLD is characterized by excessive hepatic accumulation of triglyceride and ranges from mild steatosis to non-alcoholic steatohepatitis (NASH).Development of NAFLD is promoted by increased fatty acid availability and increased de novo lipogenesis that is not paralleled by a sufficient increase in fatty acid β-oxidation or secretion of triglycerides in very low-density lipoprotein (VLDL) particles (Fabbrini et al., 2008;Kawano & Cohen, 2013).The development of NAFLD is influenced by sex, and men are more prone to develop NAFLD than women (Balakrishnan et al., 2021;Varlamov et al., 2015).The molecular mechanisms behind the sex-specific pathophysiology remain largely undetermined.
Triglycerides are predominantly synthesized by coupling fatty acids (after thioesterification into acyl CoA) to glycerol-3-phosphate (G3P) by acylation.Thus far, investigations of the mechanisms behind the increased hepatic triglyceride synthesis in NAFLD have focused on the role of fatty acid availability and the esterification process (Kawano & Cohen, 2013).However, little attention has been paid to the possible role of G3P availability (Xue et al., 2017).Glycerol is primarily stored as the backbone of triglycerides in adipose tissue.Lipolysis results in the release of glycerol into the bloodstream, and the glycerol channel aquaporin 9 (AQP9) facilitates its uptake into hepatocytes (Lebeck, 2014).Once inside the hepatocyte, glycerol is phosphorylated into G3P, which can be used for either gluconeogenesis or triglyceride synthesis.Studies in AQP9 knockout (KO) mice have demonstrated that AQP9 is an important provider of glycerol for hepatic gluconeogenesis in mice (Calamita et al., 2012;Jelen et al., 2011;Rojek et al., 2007).However, how AQP9 influences the development of NAFLD is still unclear.
On one hand, in vitro studies in human hepatocytes (HepG2) exposed to oleic acid to promote lipid accumulation (Gu et al., 2015;Wang et al., 2013) and studies in high-fat diet (HFD)-fed male rats (Cai et al., 2013;Zheng et al., 2019) support that HFD results in increased expression of AQP9 and that blocking AQP9 reduces the hepatic accumulation of triglycerides.Similarly, male mice fed a HFD for 16 weeks and obese male leptin-deficient (ob/ob) mice were both reported to have an increased hepatic Aqp9 mRNA expression when compared to lean mice (Hirako et al., 2016).On the other hand, reduced (Gena et al., 2013) or no significant difference (Rodriguez et al., 2015) in hepatic AQP9 protein expression has also been reported in male ob/ob mice when compared to lean controls.Moreover, in our previous study of male and female mice fed a HFD for 12 and 24 weeks, no significant difference in hepatic AQP9 protein abundance was observed when compared to lean controls (Iena et al., 2020).In addition, in contrast to the observations made in rats, obese leptin receptor-deficient (db/db) mice lacking AQP9 did not have reduced hepatic steatosis when compared to control mice (Spegel et al., 2015).Since leptin has been shown to modulate the hepatic expression of AQP9 (Rodriguez et al., 2011;Rodriguez et al., 2015), the interpretation of results obtained in leptin-or leptin receptor-deficient mice regarding the role of AQP9 in hepatic triglyceride accumulation must be made cautiously.With the conflicting results, more studies are warranted to determine the role of AQP9 in modulating the hepatic synthesis of triglycerides.
We hypothesized that AQP9 KO mice would respond to HFD with having reduced hepatic triglyceride accumulation and a reduced blood glucose level when compared to their WT littermates.Moreover, the hepatic expression of AQP9 is regulated in a sex-specific manner (Lebeck et al., 2012;Rodriguez, et al., 2015), and we therefore hypothesized that AQP9 deficiency would mainly alleviate hepatic steatosis in male mice.To test these hypotheses, we here analysed the effect of 12 weeks of HFD on hepatic triglyceride metabolism and blood glucose levels in both male and female WT and AQP9 KO mice.

Ethical approval
The protocol for the animal experiment was approved by the Animal Experiments Inspectorate under the Ministry of Food, Agriculture and Fisheries and was in accordance with the guiding principles of the European Parliament Directive of 22 September 2010 (2010/63/EU) for animal experiments and the ARRIVE guidelines for reporting in vivo experiments (Kilkenny et al., 2010).

Animals
The generation of the AQP9 KO mice has been described earlier (Rojek et al., 2007).The mice were generated using 129S1/Sv embryonic stem cells and have been backcrossed into a C57BL/6J background for more than 10 generations.The study was performed in littermates and included male (M) and female (F) WT (n = 9 M/11 F), heterozygote (HET) (n = 15 M/13 F), and AQP9 KO (n = 16 M/11 F) mice (Fig. 1).Mice were housed in cages containing up to five mice and were kept on a 12:12 h artificial light-dark cycle; room temperature was kept at 21 ± 2°C with a humidity of 55 ± 2%.Animals had free access to tap water and the experimental diet throughout the experimental period except for the fasting periods specified below.

Experimental design
From age 8−9 weeks, all mice were fed a HFD containing 60 kcal% from fat (D12492, Research diets, New Brunswick, NJ, USA) for 12 weeks.Body weight (BW) was measured weekly and blood glucose levels were measured every other week after 4−5 h of fasting using a glucometer (Accu-chek Aviva, Roche, Copenhagen, DK).Food intake per cage was monitored weekly throughout the study period.After 12 weeks of HFD, animals were used for collecting blood and tissue samples, used for determining the hepatic triglyceride secretion rate, or simply euthanized (Fig. 1).
A large laparotomy was performed, a blood sample was collected from the abdominal aorta, and the liver was removed to measure liver weight (LW).Tissue samples were collected from liver, muscle and adipose tissue before euthanization by cervical dislocation.
Hepatic triglyceride secretion rate.Mice were fasted overnight (12-16 h) and received an intraperitoneal injection with buprenorphine (Temgesic) diluted in 0.9% NaCl at a concentration of 0.075 mg/kg.The analgesic was injected 30 min prior to the first blood drawing.
The tail was wiped with 70% ethanol, and approximately 30 μl of blood was collected from the tail using capillary tubes.After the first blood collection, mice were injected with poloxamer 407 (P-407) (16758, Merch Life Science, Søborg, DK) intraperitoneally at a concentration of 1000 mg/kg as previously described (Millar et al., 2005).Blood samples were then collected 2, 4 and 6 h after the injection of P-407.Immunoblotting.Protein separation was performed using Any kD gels (cat.no.5671125, Bio-Rad Laboratories, Hercules, CA, USA).Protein loading was adjusted using Gelcode Coomassie blue stain reagent (Thermo Fisher Scientific, Slangerup, DK).Separated proteins were electro-transferred to polyvinylidene difluoride membranes using the Trans-Blot Turbo transfer system (Bio-Rad), and membranes were blocked for 1 h in phosphate buffered saline with Tween 20 (PBS-T) with 5% milk.After washing, membranes were incubated overnight with primary antibody against AQP9 (RA2674-685; Elkjaer et al., 2000), glycerol kinase (GlyK; ab126599, Abcam, Cambridge, UK), glycerol-3-phosphate dehydrogenase 1 (GPD1; HPA044620, Merch Life Science, Søborg, DK), glycerol phosphate acyl transferase 1 (GPAT1; ab68925, Abcam, Cambridge, UK), perilipin 2 (PLIN2; cat.no.GP40, Progen, Heidelberg, Germany), fatty acid transport protein (FATP) 2 (Falcon et al., 2010), FATP5 (Doege et al., 2006) or actin (A2066, Merck Life Science, Søborg, DK) in PBS-T added 1% bovine serum albumin (BSA) and 2 mM NaN 3 .After washing, the membranes were incubated with appropriate horseradish peroxidase (HRP)-conjugated secondary antibodies.After , respectively) and male (n = 9 and 16 respectively) wild-type (WT) and AQP9 knockout (KO) mice (P-values for significant differences in body weight between males and females ranged from P < 0.0001 to P = 0.004).B, total body weight gain for each sex and genotype.C, total body weight gain as a percentage of the starting weight for each sex and genotype.No significant differences were observed; P = 0.263 for effect of genotype and P = 0.492 for effect of sex.D, cumulative food intake per body weight for both female (n = 4 cages) and male (n = 8 cages) mice, all genotypes included (P-values for significant differences in food intake per body weight between males and females ranged from P = 0.002 to P = 0.031 throughout the study).E, cumulative weight gain per food intake for both sexes, all genotypes included (P-values for significant differences in weight gain per food intake between males and females after week 7 ranged from P = 0.004 to P = 0.048).F, average blood glucose level for female (n = 11 and 11 respectively) and male (n = 9 and 16 respectively) WT and AQP9 KO mice every other week after 4−5 h of fasting (P-values for significant differences in blood glucose levels between males and females for WT mice: week 8: P = 0.043, week 10: P = 0.025 and for KO mice: week 0: P = 0.021, week 4: P = 0.036, week 8: P = 0.0009, week 10: P < 0.0001, week 12: P = 0.0002).The number in Fig. 2B is the exact P-value, which is indicated when P < 0.05.washing, ECL (Pierce ECL Western Blotting Substrate, Thermo Fisher Scientific, Slangerup, DK) was added, and chemiluminescence was detected by Image-Quant Las4000 (GE Healthcare, Freiburg, Germany).ImageJ software was used to subtract background and quantify band intensities.Band intensities were adjusted to actin band intensities from the same membrane.

Immunohistochemistry
Sample preparation.Liver tissue was immersion-fixed in 4% formaldehyde in 0.01 mM PBS, pH 7.4.After dehydration using ethanol and xylene, the tissue was embedded in paraffin-wax, and 2 μm sections were cut using a rotary microtome.
Immunohistochemistry. Sections were dewaxed and rehydrated before the endogenous peroxidases were blocked by 1% H 2 O 2 in methanol.Target retrieval was performed by boiling sections in 1 mM Tris buffer, pH 9, with 0.5 mM EGTA in a microwave oven.Aldehydes were quenched by 50 mM NH 4 + before blocking in 1% BSA, 0.2% gelatine and 0.05% saponin in PBS.Sections were incubated overnight with primary antibody (PLIN2; cat.no.GP40, Progen, Heidelberg, Germany) in 0.1% BSA, and 0.3% triton X-100 in PBS.After rinsing, sections were incubated with HRP-conjugated secondary antibody for 1 h before detection with diaminobenzidine in PBS with 0.1% H 2 O 2 was performed.Sections were counterstained with Mayers haematoxylin before dehydration and covers were mounted with DPX.Microscopy was performed using a bright field microscope (Leica DM2500, Leica Microsystems, Brønshøj, DK) equipped with a Leica MC170 HD digital camera.

Oil Red O staining
Sample preparation.Liver tissue was immediately frozen in liquid nitrogen and kept at −80°C until embedding.The livers were embedded in Tissue-Tek O.C.T. compound (4583, Sakura, Torrance, CA, USA) and frozen using liquid CO 2 .The block was kept at −20°C until sectioning.Sections (10 μm) were cut on a cryostat and mounted on SuperFrost + glasses.Glasses were kept at room temperature for 10 min before being stored at −20°C.
Staining.A stock solution was made by adding 99% (v/v) 2-propanol alcohol to Oil Red O (ORO; cat.no.O0625, Merch Life Science, Søborg, DK), and on the day of the staining a working solution was prepared by diluting the stock solution in MilliQ H 2 O and filtering it through a 45 μm filter as described by Mehlem et al. (2013).The sections were placed at room temperature for 10 min before being immersed in ORO working solution for 5 min.Sections were immediately rinsed in running tap water for 45 min before being mounted with Glycergel mounting medium (C0563,Dako, Carpenteria, CA, USA).Bright-field images were obtained no later than 6 h after staining.Images were analysed using ImageJ software following the protocol of Mehlem et al. (2013).A threshold was set using a lean control mouse stained as described above.

Measurement of glycerol and triglyceride in serum
Glycerol and triglyceride measurements were performed using the serum triglyceride determination kit (TR0100, Merch Life Science, Søborg, DK).

Figure 3. Hepatic triglyceride (TG) content
A, liver weight in female (n = 6 and 5, respectively) and male (n = 5 and 7, respectively) wild-type (WT) and AQP9 knockout (KO) mice.B, correlation between body weight and liver weight, both genotypes combined for females and males.C, quantification of Oil Red O staining in livers from female (n = 5 and 4, respectively) and male (n = 4 and 6, respectively) WT and AQP9 KO mice.No significant differences were observed; P = 0.369 for effect of genotype and P = 0.531 for effect of sex.D, liver triglyceride content measured after Folch extraction in male (n = 4 and 7, respectively) and female (n = 6 and 4, respectively) WT and AQP9 KO mice.No significant differences were observed; P = 0.723 for effect of genotype and P = 0.333 for effect of sex.E, correlation between liver TG content and liver weight in female mice, both genotypes included.F, correlation between liver TG content and liver weight in male mice, both genotypes included.G, representative images of immunoperoxidase staining for PLIN2 in the liver is shown at ×25 or ×100 (insets) magnification.CV, central vein.Arrows indicate immunoperoxidase staining for PLIN2 at a small lipid droplet at both magnifications; arrowheads indicate staining for PLIN2 in larger lipid droplets.Scale bars indicate 100 μm.The number in Fig. 3A is the exact P-value, which is indicated when P < 0.05.
J Physiol 602.13 qPCR Sample preparation.Liver tissue was homogenized in cold TRIreagent (AM9738, Thermo Fisher Scientific).Samples were incubated for 5 min at room temperature and then centrifuged at 12 000 g for 15 min at 4°C.A lipid layer formed on top; this was discarded before further processing.
RNA extraction.Total RNA was extracted using the RiboPure RNA isolation kit (AM1924, Thermo Fisher Scientific).All steps were performed according to the manufacturer's instructions.
qPCR.Amplification was performed using the cDNA equivalent of 200 ng of RNA, 10 pmol of each primer, and SYBR Green I Master (cat.no.04887352001, Roche, Copenhagen, DK).The sequences of the primer pairs are shown in Table 1.In parallel with the samples, a dilution series with dilution factors 1, 3, 9, 27 and 81 was prepared and added to the plate along with primers and SYBR Green.The qPCR was run with a LightCycler 480 (Roche, Copenhagen, DK) with 40 cycles with denaturation at 95°C, annealing to primers at 60°C and elongation at 72°C.Fold change was calculated using the method of Pfaffl (2001).The relative expression of the mRNA of interest was normalized to 18S rRNA (18S) mRNA.

Hepatic triglyceride content
Total liver lipids were extracted by the method of Folch et al. (1957).In brief, the frozen liver samples were weighed and homogenized in a 2:1 chloroform:methanol solution (20-fold the volume of the tissue).The tissues were homogenized and left to equilibrate for ∼30 min.Afterwards, 0.9% NaCl was added, and the samples were vortexed before being left for 2 h to separate the layers.The bottom chloroform phases were collected, and 100 μl Immunoblots and results of the densitometric analysis of the immunoblots for the indicated proteins in liver from female (n = 6 and 5, respectively) and male (n = 5 and 7, respectively) wild-type (+/+) and AQP9 knockout (−/−) mice.A, aquaporin 9 (AQP9) protein in liver from both sexes and genotypes.B, glycerol kinase (GlyK) protein in liver from both sexes and genotypes.No significant differences were observed; P = 0.147 for effect of genotype and P = 0.281 for effect of sex.C, cytosolic glycerol phosphate dehydrogenase (GPD1) protein in liver from both sexes and genotypes.D, relative mitochondrial glycerol phosphate dehydrogenase (Gpd2) mRNA abundance in liver from female (n = 5 and 5, respectively) and male (n = 5 and 7, respectively) wild-type and AQP9 knockout mice.No significant differences were observed; P = 0.682 for effect of genotype and P = 0.422 for effect of sex.The numbers in Fig. 4A and C are the exact P-values, which are indicated when P < 0.05.
aliquots were dried down and resuspended in 200 μl isopropanol and used for the measurement of triglycerides with the serum triglyceride determination kit (TR0100, Merch Life Science, Søbo).

Statistical analysis
Graph Pad Prism version 8.2 (GraphPad Software, San Diego, CA, USA) was used for statistical analysis.Data were analysed by two-way ANOVA or Student's unpaired t test and presented as means ± standard deviation unless otherwise stated in the figure legend.Correlation analysis was performed using Pearson's correlation coefficient data presented as R 2 and P-values.

Results
As outlined in Fig. 1, the study included female and male AQP9 WT, AQP9 heterozygotes (HET) and AQP9 KO animals.However, for the measured parameters, the HETs were comparable with the WT animals, and therefore we focused on results obtained in AQP9 WT and KO mice.

Body weight and food intake during the 12 weeks of HFD feeding
As expected, male mice had a significantly higher BW than females for both genotypes throughout the experiment (Fig. 2A and Table 2).At the termination of the experiment, male mice had an average BW of 41.0 ± 5.0 g, whereas, in females, the average BW was 31.2 ± 5.1 g (P = 0.004).Corresponding to that, males, on average, increased their weight by 79 ± 24.6% and females by 71 ± 24.6% (P = 0.283).No significant effect of genotype on BW was observed (Fig. 2A).When comparing the absolute weight gain after 12 weeks of HFD, male AQP9 KO mice had a significantly higher weight gain than the corresponding females (Fig. 2B).No significant differences between genotypes or sexes were observed in the relative weight gain (Fig. 2C).We have previously observed a higher weight gain in male than in female mice in response to 12 weeks of HFD (Iena et al., 2020), and therefore we monitored the food intake per cage to identify if a difference in weight gain between males and females would be due to a difference in food intake.When comparing the food intake per BW, the females had a significantly higher food intake than males throughout the study period (Fig. 2D).This was paralleled by males gaining more weight per food intake per BW after 7 weeks of HFD compared to the females (Fig. 2E).We conclude that males gain more weight per food intake per BW compared to females; however, AQP9 deficiency does not affect the BW in HFD fed mice.
AQP9 deficiency is not associated with reduced blood glucose levels after 12 weeks of HFD No significant effect of AQP9 deficiency was found on the blood glucose levels during the 12 weeks of HFD feeding (Fig. 2F).However, when comparing the starting and final blood glucose levels, both female (P = 0.027) and male AQP9 KO mice (P = 0.004) displayed a significant increase in their blood glucose levels in response to HFD.In females, the increase was due to a low starting blood glucose level.A similar difference was not observed in the WT mice (female mice: P = 0.389, male mice: P = 0.857).Moreover, especially in the KO animals, the male mice generally had a higher blood glucose level than their female counterparts (Fig. 2F).Overall, we found no evidence for AQP9 deficiency being associated with a reduced blood glucose level in mice with diet-induced obesity.

Serum glycerol and triglyceride levels
Female AQP9 KO mice had elevated serum glycerol levels when compared to their WT littermates; this did not reach statistical significance in male mice when using two-way ANOVA (Table 2).However, when using an unpaired t test, both male (P = 0.020) and female (P = 0.033) KO mice demonstrated an elevated serum glycerol level.The serum triglyceride (TG) values are generally determined by quantifying the glycerol concentration after hydrolysis of TGs into glycerol and fatty acids.Therefore, the free glycerol in the serum gives a false increase in the TG measurements (Lebeck & Brock, 2021).Thus, we here present TG values that have been corrected for the free glycerol in the sample.After correction, the TG levels did not differ between WT and KO mice (Table 2).

Figure 6. Relative quantification of proteins involved in fatty acid transport and triglyceride synthesis/storage after 12 weeks of HFD
Immunoblots and results of the densitometric analysis of the immunoblots for the indicated proteins in liver from female (n = 6 and 5, respectively) and male (n = 5 and 7, respectively) wild-type (+/+) and AQP9 knockout (−/−) mice.A, fatty acid transport protein (FATP) 2 in liver from both sexes and genotypes.B, FATP5 in liver from both sexes and genotypes.The two-way ANOVA revealed a significant interaction (P = 0.001) between the effects of sex and genotype on the expression of FATP5, and therefore the effect of genotype was evaluated using Student's t test instead.C, glycerol phosphate acyl transferase 1 (GPAT1) in liver from both sexes and genotypes.No significant differences were observed; P = 0.849 for the effect of genotype and P = 0.263 for effect of sex.D, perilipin 2 (PLIN2) in liver from both sexes and genotypes.No significant difference was observed for the effect of sex (P = 0.261).The numbers in Fig. 6A and B are the exact P-values, which are indicated when P < 0.05.

No effect of AQP9 deficiency on liver weight after 12 weeks of HFD
We next wanted to evaluate whether AQP9 deficiency would reduce the hepatic accumulation of TG.No significant effect of genotype on liver weight (LW) was observed.Male AQP9 KO mice had a significantly higher LW than female AQP9 KO mice.In the WT mice, no significant difference was observed (P = 0.071) (Fig. 3A).When LW was correlated to BW, a significant positive correlation was observed in male mice (R 2 = 0.59, P = 0.003) but not in female mice (R 2 = 0.31, P = 0.075) (Fig. 3B).To further evaluate the influence of sex and genotype on hepatic TG accumulation, Oil Red O staining of liver sections was performed and quantified (Fig. 3C), and the hepatic TG content was measured after Folch extraction (Fig. 3D).Both measurements found no significant effect of sex or genotype on the hepatic TG content.A positive correlation between the TG content and LW was found in male (R 2 =0.74,P = 0.003) but not in female (R 2 =0.013,P = 0.769) mice (Fig. 3E and  F).Furthermore, immunohistochemical staining of liver tissue from obese WT and AQP9 KO mice for perilipin 2 (PLIN2), which coats intracellular lipid droplets, is consistent with no clear effect of AQP9 deficiency on the development of hepatic steatosis (Fig. 3G).In summary, we found no evidence for AQP9 deficiency resulting in a reduced hepatic TG accumulation in mice with diet-induced obesity.

Effect of genotype and sex on enzymes involved in glycerol metabolism
We speculated that AQP9 deficiency would result in compensatory changes in the enzymes involved in the synthesis of G3P.First, we confirmed the lack of AQP9 protein expression in AQP9 KO mice (Fig. 4A).Of note, HETs had a reduced protein expression of AQP9 compared to WT mice.In female HETs, an 83% decrease in hepatic AQP9 protein expression was observed (1.17 ± 0.28 vs. 0.34 ± 0.06, P < 0.0001), and in males, a 56% decrease was found (1.0 ± 0.26 vs. 0.43 ± 0.18, P < 0.011) (data not shown).Furthermore, male WT mice had a higher hepatic AQP9 expression than female WT mice after 12 weeks of HFD (Fig. 4A).Glycerol kinase (GlyK) phosphorylates glycerol into G3P.The hepatic expression of GlyK protein was not affected by genotype or sex (Fig. 4B).Glycerol-3-phosphate dehydrogenase 1 (GPD1) catalyses the reversible formation of G3P from dihydroxyacetone phosphate (DHAP), which is the final step in the synthesis of G3P from other precursors than glycerol such as glucose.The hepatic expression of GPD1 protein was not affected by AQP9 deficiency; however, in male AQP9 KO mice, the abundance of GPD1 was higher than in the female AQP9 KO mice (Fig. 4C).AQP9 expression in the liver has previously been shown to be heterogeneous, with a predominant expression in the perivenous hepatocytes (Nicchia et al., 2001).We therefore used the single-cell RNA-sequencing datasets from murine liver in the Liver Cell Atlas (https://www.livercellatlas.org/;Guilliams et al., 2022) to evaluate the cellular expression overlap between Aqp9 and the different glycerol metabolizing enzymes.The expression of Aqp9 and Glyk was mainly confined to the same group of murine hepatocytes (Fig. 5B and C), whereas the expression of Gpd1 mainly resided in other hepatocytes (Fig. 5D).Instead, glycerol-3-phosphate dehydrogenase 2 (Gpd2), which catalyses the conversion of G3P to DHAP, was predominantly expressed in the same group of hepatocytes as Aqp9 (Fig. 5E).Therefore, the relative expression of Gpd2 mRNA was investigated, and no affect of genotype or sex was observed (Fig. 4D).Thus, the lack of AQP9 expression was not associated with an increased protein expression of GlyK or GPD1/Gpd2.

Effect on proteins involved in triglyceride metabolism
Next, we wanted to investigate how proteins involved in TG formation and storage were affected by 12 weeks of HFD feeding in male and female AQP9 KO mice.Fatty acid transport protein 2 (FATP2) and 5 (FATP5) are important facilitators of FFA uptake in hepatocytes.The protein expression of FATP2 was not significantly affected by AQP9 deficiency in either sex, as shown in Fig. 6A.However, male WT mice had a higher expression of FATP2 compared to female WT mice, whereas, in the KO animals, no significant difference was observed (P = 0.072).The effect of genotype on the hepatic protein expression of FATP5 was sex-specific, with female AQP9 KO mice having a reduced abundance of FATP5, whereas, in males, an increased expression was observed when compared to WT mice (Fig. 6B).The Liver Cell Atlas shows strong expression of Fatp2 in most of the hepatocytes, including the group of hepatocytes expressing Aqp9, whereas Fatp5 expression was predominantly confined to other groups of hepatocytes (Fig. 7B and C).Glycerol-3-phosphate acyltransferase 1 (GPAT1) facilitates the first step in TG synthesis, the formation of lysophosphatidic acid from G3P and activated fatty acids.Gpat1 expression was mainly found in the same group of hepatocytes as Aqp9 (Fig. 7D).The hepatic expression of GPAT1 protein was not significantly affected by genotype or sex (Fig. 6C).The statistical analysis of the hepatic expression of PLIN2 showed a significant effect of genotype (P = 0.015); however, this did not reach statistical significance in the direct comparison of WT and KO (female: P = 0.360 and male: P = 0.189) (Fig. 6D).Interestingly, Plin2 demonstrated a rather broad expression profile, with the lowest expression found in hepatocytes expressing Aqp9 (Fig. 7E).In summary, apart for the reduced hepatic expression of FATP5 in female AQP9 KO mice, the analysed parameters do no support that AQP9 deficiency is associated with reduced hepatic synthesis of TG after 12 weeks of HFD.

Effect of genotype on hepatic peroxisome proliferator-activated receptor α mRNA expression in male mice
To evaluate if AQP9 deficiency would be associated with an increased channelling of FAs towards β-oxidation, we analysed the mRNA expression of genes central to the regulation of β-oxidation.We measured the mRNA levels of peroxisome proliferator-activated receptor α (Pparα), carnitine palmitoyltransferase 1a (Cpt1a), acyl-CoA dehydrogenase, medium chain (Acadm) and acyl-CoA synthetase long-chain family member 1 (Acsl1) to get a

Figure 10. Effect of genotype on hepatic triglyceride (TG) secretion rate after 12 weeks HFD
A, corrected TG concentration in serum after subtraction of the free glycerol values in female wild-type (WT) (n = 5) and AQP9 knockout (KO) (n = 6) mice 0, 2, 4 and 6 h after injection of poloxamer 407 (P-407).B, corrected TG concentration in serum after subtraction of the free glycerol values in male WT (n = 4) and AQP9 KO (n = 6) mice 0, 2, 4 and 6 h after injection of P-407.The numbers in Fig. 10A and B are the exact P-values.
first insight.Male AQP9 KO mice demonstrated a 61% increase in the Pparα mRNA levels compared to WT males (Fig. 8A).However, the effect of genotype on Cpt1a (P = 0.087), Acadm (P = 0.083) and Acsl1 (P = 0.191) did not reach statistical significance in male mice.By contrast, in female mice, the mRNA expression of genes involved in β-oxidation was not affected by genotype (Fig. 8B).Interestingly, the expression of Pparα was confined to the same group of hepatocytes as Aqp9 and a similar overlap was seen for Cpt1a and Acsl1, even though Acsl1 expression was found in a broader range of hepatocytes (Fig. 9 B-D).By contrast, Acadm expression was mainly found in other groups of hepatocytes (Fig. 9E).In summary, AQP9 deficiency in male mice is associated with an increased expression of Pparα, which likely promotes an increased β-oxidation of FAs.

Effect of genotype on hepatic triglyceride secretion in male mice
We also wanted to investigate how the lack of AQP9 would influence hepatic triglyceride secretion.After injection with poloxamer-407, which inhibits the lipoprotein lipase, female AQP9 KO and WT mice show a similar increase in their serum TG levels (Fig. 10A).Instead, male AQP9 KO mice had significantly lower levels of serum TG at 2, 4 and 6 h after the injection compared to WT mice (Fig. 10B).
To further evaluate the influence of AQP9 deficiency on hepatic triglyceride secretion, we analysed the mRNA expression of genes central to the hepatic synthesis of VLDL.However, none of the investigated genes were significantly influenced by genotype in either male or female mice (Fig. 11A and B).Apolipoprotein B (Apob), microsomal triglyceride transfer protein (Mttp), diglyceride acyltransferase 2 (Dgat2) and glycerol-3-phosphate acyltransferase 4 (Gpat4) were all expressed in the same group of hepatocytes as AQP9, whereas cell death-inducing DFFA-like effector b (Cideb) was expressed in other groups of hepatocytes (Fig. 12  B-E).Overall, the results show that only male AQP9 KO mice have reduced hepatic secretion of triglycerides.To get a general overview of which of the measured parameters were affected by the HFD in a parallel manner, a correlation analysis was performed, which clearly demonstrated that more parameters are correlated in male mice than in female mice (Fig. 13A and B).In addition, this demonstrates that, in male mice, the hepatic expression of AQP9 was negatively correlated with weight gain and the expression of PLIN2 and PPARα.A similar pattern was not observed in female mice, which instead had a positive correlation between AQP9 and GPAT1.

Discussion
In the present study, we investigated the effect of 12 weeks of HFD in male and female AQP9 KO and WT mice and report several novel findings.First, we found no evidence for AQP9 deficiency being associated with a reduced hepatic accumulation of TG or a reduced blood glucose level in mice with diet-induced obesity.Instead, we demonstrate that the effect of AQP9 deficiency on hepatic lipid metabolism is sex-specific, with only male AQP9 KO mice having a reduced hepatic secretion of triglycerides and an elevated expression of PPARα.
We and others have previously reported a sex-specific response to HFD in C57BL/6J mice (Hong et al., 2009;Iena et al., 2020;Yang et al., 2014) and the development of obesity-related complications (Iena et al., 2020;Varlamov et al., 2015).In our previous investigation of the effect of 12 and 24 weeks of HFD, we found that the sex-specific response to HFD was most pronounced after 12 weeks of HFD, whereas, after 24 weeks, the male and female mice had a more similar response to the diet (Iena et al., 2020).Therefore, the effect of 12 weeks of HFD was investigated in the present study.However, in this study, the overall sex difference was less pronounced after 12 weeks of HFD, with no significant difference in absolute or relative weight gain or LW between male and female WT mice.It is likely that an extension of the HFD feeding period would have resulted in more pronounced results.However, the higher number of correlations between the different analysed parameters in male mice likely, at least in part, reflects that male mice, in general, were more severely affected by the HFD than female mice.In addition, a clear sex difference was observed in weight gain per food intake per BW, with males gaining more weight per caloric intake than females.This is in accordance with previous reports (Hwang et al., 2010) and indicates that any sex differences observed are not due to a higher caloric intake in males.
AQP9 deficiency did not significantly influence the effect of HFD on BW or LW in either females or males and, in contrast to our hypothesis, we found no evidence for a decreased blood glucose level or a decreased hepatic accumulation of triglycerides.Overall, this is in line with the results obtained by Spegel et al. (2015), where AQP9 deficiency was investigated in obese db/db mice.In the initial study of AQP9 KO mice, a reduced blood glucose level was found in obese db/db mice (Rojek et al., 2007) and a recent investigation of a novel AQP9 inhibitor administered to obese db/db mice found a trend towards a reduced blood glucose level, which, however, did not reach statistical significance (Florio et al., 2022).However, similar to the study by Spegel and coworkers, we here report an elevated blood glucose level when comparing the baseline and 12 weeks blood glucose level in AQP9 KO mice.This was especially evident in male AQP9 KO mice, whereas in the female AQP9 KO mice, the increase was due to the low baseline value.Thus, our results suggest that male AQP9 KO mice are more susceptible to developing insulin resistance than their WT littermates, which contrasts with the assumption that lack of AQP9 would lower the hepatic glucose  production by diminishing the hepatic availability of glycerol (Mendez-Gimenez et al., 2014).Overall, as also outlined in the Introduction, the results obtained in mice for the role of AQP9 in obesity are ambiguous, which at least in part could be due to differences in the obesity model used, the age of the animals when investigated and the sex of the animals investigated likely influencing the results.
We speculated that the lack of a significant effect of AQP9 deficiency on the hepatic accumulation of triglyceride would be due to compensatory mechanisms involved in the hepatic synthesis of G3P.Our results show no compensatory increase in the protein expression of GlyK.Similarly, we found no difference in the protein expression of GPD1 that would promote the synthesis of G3P from other precursors than glycerol.Interestingly, the data from the Liver Cell Atlas clearly show that the expression of Aqp9 and Glyk is largely confined to the same group of hepatocytes, whereas Gpd1 is predominantly expressed in other hepatocytes.Therefore, it could be speculated that the synthesis of G3P via AQP9/GlyK is linked to different metabolic pathways than G3P synthesized via GPD1.The lack of an association between AQP9 deficiency and a reduced hepatic accumulation of triglyceride is in contrast to the results reported from studies using cells and rats, which report that silencing of AQP9 leads to decreased hepatic steatosis (Cai et al., 2013;Wang et al., 2013;Zheng et al., 2019).Further studies are needed to clarify this species difference and potentially which animal model mimics the role of AQP9 in human glycerol metabolism and development of NAFLD.One study investigating the association between NAFLD in humans and AQP9 found a reduced protein expression of AQP9 in individuals with NAFLD (Rodriguez et al., 2014).However, it remains to be determined whether a reduced expression of AQP9 is a cause or a consequence of NAFLD.Moreover, the hepatic expression of AQP9 has been shown to be reduced by insulin in rodent models (Carbrey et al., 2003;Kuriyama et al., 2002).Instead, an increased expression of AQP9 protein was reported in response to insulin stimulation in the human hepatocyte cell line; HepG2 (Rodriguez et al., 2011), which could indicate fundamental species differences in the regulation of AQP9.
In a healthy liver, TG accumulation is maintained at a low level by a balance between the uptake of FFA/de novo lipogenesis for the synthesis of TG on the one hand and FFA oxidation/TG secretion in VLDL particles on the other hand (Kawano & Cohen, 2013).Here, we show that male AQP9 KO mice have an elevated expression of PPARα, a major regulator of hepatic β-oxidation and ketogenesis (Pawlak et al., 2015), which likely here, at least in part, reflects an increased hepatic β-oxidation of fatty acids.The data from the Liver Cell Atlas clearly show that the expression of Ppara and Cpt1a is confined to the same group of hepatocytes as AQP9 and GlyK as well as GPAT1.Moreover, we here are the first to show that AQP9 deficiency in male mice is associated with a reduced hepatic secretion of TG.Further studies are needed to identify the mechanisms behind the decreased secretion of TG.The expression of ApoB and Mttp is largely found in the same group of hepatocytes as Aqp9 and Glyk.It could therefore be speculated that a reduced availability of G3P in these hepatocytes results in a reduced synthesis of TG, which, in turn, results in reduced secretion of TG in VLDL particles in male mice and a compensatory increase in the hepatic β-oxidation of fatty acids.No effect of AQP9 deficiency on the hepatic TG secretion or the expression PPARα was observed in female mice.
In conclusion, in contrast to previous assumptions, we here show that AQP9 deficiency in mice with diet-induced obesity does not alleviate the development of hepatic steatosis, and we find no support for AQP9 KO mice being less susceptible to developing insulin resistance.Instead, our results rather support that the effect of AQP9 deficiency is sex-specific, with male AQP9 KO mice having a reduced hepatic secretion of TG and an elevated hepatic expression of PPARα.Thus, we find no evidence that AQP9 inhibition is a target in alleviating the development of hepatic steatosis nor insulin resistance.Since these results are in contrast with results obtained in rats, this calls for further investigation of which animal model could mimic the human glycerol metabolism best in relation to its influence on hepatic steatosis and insulin resistance.

Figure 2 .
Figure 2. Effect of high-fat diet (HFD) on body weight, weight gain, food intake and blood glucose levelsA, body weight of female (n = 11 and 11, respectively) and male (n = 9 and 16 respectively) wild-type (WT) and AQP9 knockout (KO) mice (P-values for significant differences in body weight between males and females ranged from P < 0.0001 to P = 0.004).B, total body weight gain for each sex and genotype.C, total body weight gain as a percentage of the starting weight for each sex and genotype.No significant differences were observed; P = 0.263 for effect of genotype and P = 0.492 for effect of sex.D, cumulative food intake per body weight for both female (n = 4 cages) and male (n = 8 cages) mice, all genotypes included (P-values for significant differences in food intake per body weight between males and females ranged from P = 0.002 to P = 0.031 throughout the study).E, cumulative weight gain per food intake for both sexes, all genotypes included (P-values for significant differences in weight gain per food intake between males and females after week 7 ranged from P = 0.004 to P = 0.048).F, average blood glucose level for female (n = 11 and 11 respectively) and male (n = 9 and 16 respectively) WT and AQP9 KO mice every other week after 4−5 h of fasting (P-values for significant differences in blood glucose levels between males and females for WT mice: week 8: P = 0.043, week 10: P = 0.025 and for KO mice: week 0: P = 0.021, week 4: P = 0.036, week 8: P = 0.0009, week 10: P < 0.0001, week 12: P = 0.0002).The number in Fig.2Bis the exact P-value, which is indicated when P < 0.05.

Figure 4 .
Figure 4. Relative quantification of AQP9 and enzymes involved in glycerol metabolism after 12 weeks of HFDImmunoblots and results of the densitometric analysis of the immunoblots for the indicated proteins in liver from female (n = 6 and 5, respectively) and male (n = 5 and 7, respectively) wild-type (+/+) and AQP9 knockout (−/−) mice.A, aquaporin 9 (AQP9) protein in liver from both sexes and genotypes.B, glycerol kinase (GlyK) protein in liver from both sexes and genotypes.No significant differences were observed; P = 0.147 for effect of genotype and P = 0.281 for effect of sex.C, cytosolic glycerol phosphate dehydrogenase (GPD1) protein in liver from both sexes and genotypes.D, relative mitochondrial glycerol phosphate dehydrogenase (Gpd2) mRNA abundance in liver from female (n = 5 and 5, respectively) and male (n = 5 and 7, respectively) wild-type and AQP9 knockout mice.No significant differences were observed; P = 0.682 for effect of genotype and P = 0.422 for effect of sex.The numbers in Fig.4Aand C are the exact P-values, which are indicated when P < 0.05.

Figure 9 .
Figure 9. UMAP plot of the mRNA expression of different genes in murine liver cells from the Liver Cell Atlas (https://www.livercellatlas.org/)A, overview of the cell clusters identified in the murine liver.B, UMAP plot for peroxisome proliferator-activated receptor α (Ppara) expression.C, UMAP plot for carnitine palmitoyltransferase 1A (Cpt1a) expression.D, UMAP plot for acyl-CoA dehydrogenase medium chain (Acadm) expression.E, UMAP plot for acyl-CoA synthetase long-chain family member 1 (Acsl1) expression.

Figure 12 .
Figure 12.UMAP plot of the mRNA expression of different genes in murine liver cells from the Liver Cell Atlas (https://www.livercellatlas.org/)A, overview of the cell clusters identified in the murine liver.B, UMAP plot for apolipoprotein B (Apob) expression.C, UMAP plot for microsomal triglyceride transfer protein (Mttp) expression.D, UMAP plot for diglyceride acyltransferase 2 (Dgat2) expression.E, UMAP plot for glycerol phosphate acyltransferase 4 (Gpat4) expression.F, UMAP plot for cell death-inducing DFFA-like effector b (Cideb) expression.

Figure 13 .
Figure13.Heatmap with correlations between different parameters and proteins/genes, predominately located in the same cells as AQP9 as evaluated using the Liver Cell Atlas (https://www.livercellatlas.org/)A, correlations in female mice after 12 weeks of HFD (wild-type and knockout mice combined).B, correlations in male mice after 12 weeks of HFD (wild-type and knockout mice combined).Indicated values are Spearman's R, when P-values are <0.05.Results from western blots (WB) and results from qPCR analysis (mRNA).

Table 2 . Biological parameters in female and male AQP9 WT and KO mice
Values are expressed as mean and standard deviation (SD); n indicates the numbers of measurements.Included measurements are body weight (BW), start and final blood glucose (BG), plasma glycerol, triglyceride (TG) and TG corrected.TG corrected values are calculated by subtracting the absorbance of free glycerol from the absorbance detected for triglyceride in the same sample.P indicates P-values for comparison of genotypes within the same sex and P comparison of sex with the same genotype.Values in bold indicate statistical significance.KO, knockout; WT, wild-type.