β-KL is required for growth in postimplantation stages

To evaluate the impact of β-KL on growth, we examined the size of β-kl−/− mice through the developmental stages to the postnatal period. We first compared the size and cell numbers of blastocysts at E3.5 among the genotypes. Before implantation, no difference was observed in the size or number of cells in β-kl−/− blastocysts compared with β-kl+/+ and β-kl+/− blastocysts (Fig 1A and B).

Murine embryos absorb maternal nutrients through the yolk sac until E10 (Zohn & Sarkar, 2010). Hence, we next examined the size of embryos at E9.5 as a representative point in the yolk sac-dependent period. The crown-rump length (CRL) of β-kl−/− embryos was significantly shorter than that of β-kl+/− embryos at E9.5 (Fig 1C). The somite numbers were comparable among the three genotypes (Fig 1D). Because β-kl+/+ and β-kl+/− embryos were indistinguishable in terms of growth at E3.5 and E9.5, we used β-kl+/− embryos as a control to provide a sufficient number of embryos for experiments. We again confirmed that β-kl−/− embryos were significantly smaller than β-kl+/− embryos in different cohorts (Fig S1A). The body weight of β-kl−/− mice was lower in both female and male animals after birth (Ito et al, 2005). Likewise, no sex bias was observed in CRL at E9.5 (Fig S1B and C). On the other hand, the birth ratio of β-kl−/− mice was less than expected from Mendelian inheritance (Ito et al, 2005; Somm et al, 2017), though the embryonic genotypes were segregated to approximately Mendelian ratios at E9.5 (Table S1). The gross appearance of β-kl−/− embryos was normal at E9.5 (Fig S1D). We also evaluated the CRL at E12.5 and confirmed that β-kl−/− embryos remained smaller than β-kl+/− embryos during the placenta-dependent period (Fig 1E).

Consistent with our previous report (Kobayashi et al, 2016), the body weight of β-kl−/− mice remained significantly lower than that of their control littermates (Fig 1F). These findings demonstrate that β-KL is required for growth in the postimplantation embryos, but is not involved in morphogenesis or survival up to E9.5.

β-KL deletion leads to lipid abnormalities in the E9.5 embryo and yolk sac

To gain insight into the mechanisms underlying growth restriction in β-kl−/− embryos, we performed transcriptome analysis by RNA-sequencing. Among 16,314 genes significantly changed in β-kl−/− embryos compared with β-kl+/− embryos, the top 2,000 variable genes were divided into four groups using k-means clustering. The genes in cluster C, which is the cluster of genes related to cholesterol and sterol metabolism, were separated by genotypes (Fig 2A and Table S2). Another pathway analysis by the GAGE method also showed that sterol biosynthesis and metabolic processes were activated in β-kl−/− embryos (Fig 2B and Table S3). Quantitative real-time PCR (qRT-PCR) analyses confirmed that genes involved in lipid metabolism were up-regulated in β-kl−/− compared with β-kl+/− embryos (Fig 2C).

To judge the significance of changes in transcript levels, we compared the expression levels of genes related to lipid metabolism in fetal tissues from control β-kl+/− embryos with those in adult livers (Fig S2A). At E9.5, the level of 3-hydroxy-3-methylglutaryl-CoA reductase (Hmgcr) was higher in embryos than in adult liver, whereas 3-hydroxy-3-methylglutaryl-CoA synthase 1 (Hmgcs1) and the fatty acid synthase (Fasn) levels were comparable (Fig S2B). Low-density lipoprotein (LDL) receptor (Ldlr) expression in embryos was comparable with that in adult liver (Fig S2C). In contrast to Ldlr, high-density lipoprotein receptor (Scarb1) expression was markedly higher in yolk sacs than in embryos (Fig S2C).

We also examined the mRNA levels of enzymes in synthesizing bile acids because β-KL is indispensable for suppressing bile acid synthesis in adult mice (Ito et al, 2005; Tomiyama et al, 2010; Kobayashi et al, 2016). In the liver of adult β-kl−/− mice, the expression of Cyp7a1, a rate-limiting enzyme of hepatic bile acid synthesis, is strongly up-regulated, but Cyp7a1 expression was undetectable in E9.5 embryos and yolk sacs by qRT-PCR (data not shown). It is believed that the alternative pathway is the main contributor to production of bile acids in the fetal liver (Jia et al, 2021). However, to our surprise, the level of Cyp7b1, a gatekeeper to the alternative pathway, was much lower in E9.5 embryos than in the adult liver (Fig S2D). Furthermore, Cyp7b1 was expressed at similar levels in β-kl−/− and β-kl+/− embryos (Fig S2E), contrary to our expectation that β-KL would be required to suppress the alternative pathway in the embryo.

Transcription of Hmgcr and Ldlr is induced when cellular cholesterol levels are low (Silva Afonso et al, 2018). Therefore, we next evaluated lipid levels in E9.5 embryos and yolk sacs by TLC. The levels of cholesterol esters (CE) and triglycerides (TG) per μg of protein were significantly decreased in β-kl−/− embryos compared with their control littermates (Fig 2D). In contrast to the embryos, CE and TG were both increased in β-kl−/− yolk sacs (Fig 2E). The murine yolk sac is a membranous sac surrounding the embryo (Zohn & Sarkar, 2010), and its size as a tissue size is smaller than that of the whole embryo. Indeed, we confirmed this in the TLC samples by using total protein amount as a surrogate indicator (Fig S3A). Surprisingly, however, the CE content per tissue was higher in the yolk sacs than in the embryos, whereas TG levels were comparable in yolk sacs and embryos (Fig S3B). β-kl−/− yolk sacs were significantly smaller than β-kl+/− yolk sacs (Fig S3A), but no difference was observed in the lipid levels per yolk sacs (Fig S3B). Whether assessed per μg of protein or per tissue, β-kl−/− embryos showed significantly reduced lipid levels (Figs 2D and S3B). We also measured the lipid contents in the placenta, because β-kl mRNA was detected in the E12.5 placenta, even though at markedly lower levels than in the yolk sac or fetal liver (Fig S3C). We found that the lipid contents were similar in β-kl−/− and β-kl+/− placentas (Fig S3D). These results indicate that β-KL is required to regulate lipid metabolism in the embryo in the yolk sac-dependent period.

β-KL is required to regulate maternal cholesterol transport through the yolk sac

To better understand the role of β-KL in lipid metabolism during development, we focused on β-KL expression in the yolk sac. At E9.5, the level of β-kl mRNA in the yolk sac was ∼130-fold higher than that in the whole embryo (Fig 3A). Notably, β-kl expression in the yolk sac was also 1.8-fold higher than that in adult liver. We also confirmed the presence of β-KL protein in the E9.5 yolk sac (Fig 3B). The molecular weight of β-KL in the yolk sac was slightly smaller than that of β-KL expressed in the adult liver, which may suggest that there is some difference in form or modifications of the protein between adult and embryonic tissues. Hematoxylin and eosin staining revealed that the yolk sac from β-kl−/− embryos at E9.5 appeared grossly normal, but there were vacuoles in the basolateral surface (Fig 3C). Raabe et al reported that the vacuoles in yolk sacs lacking microsomal triglyceride transfer protein (Mttp) were cytosolic lipid droplets (Raabe et al, 1998). In accordance with this, we confirmed lipid accumulation in β-kl−/− yolk sacs by Oil Red O staining (Fig 3D).

Next, we analyzed gene expression in the yolk sac by qRT-PCR. Apolipoprotein B (Apob) and Mttp are essential for lipoprotein synthesis and their deletion causes lipid accumulation in the yolk sac (Farese et al, 1996; Raabe et al, 1998). However, the mRNA levels of Apob and Mttp were not altered in β-kl−/− yolk sacs (Fig 3E). Among the lipogenic genes, Hmgcs1, which converts acetyl-CoA to HMG-CoA in the mevalonate pathway, was up-regulated in β-kl−/− yolk sacs, but the expression of Hmgcr, the rate-limiting enzyme of de novo cholesterol synthesis, showed no difference between β-kl+/− and β-kl−/− yolk sacs (Fig 3E). Fasn mRNA levels were slightly but significantly increased in β-kl−/− yolk sacs (Fig 3E), whereas the levels of lipoprotein receptors (Ldlr, Scarb1) were not altered (Fig 3E). We previously demonstrated that the livers in β-kl−/− mice showed increased lipoprotein lipase (Lpl) mRNA levels, and the uptake of very low-density lipoprotein into the liver was increased (Kobayashi et al, 2016). Similarly, the expression levels of lipases (Lpl, Lipa, Lipc), digestive enzymes that are required for lipid uptake, were significantly increased in β-kl−/− yolk sacs (Fig 3E). Although the physiological significance of these enzymes in the yolk sac is unclear, we confirmed that these enzymes are detectable at the mRNA level in the yolk sac (Fig S4).

Because the murine embryo absorbs lipids derived from maternal blood through the yolk sac before placental maturation (Zohn & Sarkar, 2010), we speculated that β-KL expressed in the yolk sac might regulate the cholesterol supply from the dam to the embryo. To address this hypothesis, we injected ¹³C-labeled free cholesterol (FC) into pregnant mice at E8.5 and determined the ¹³C-cholesterol contents in the whole embryos after 24 h by GC-MS. As previously reported (Yoshida & Wada, 2005), ¹³C-FC was detected in the embryo (Fig 3F), confirming that maternal lipoproteins were transported to the embryo through the yolk sac. ¹³C-FC in β-kl−/− embryos was significantly decreased compared with β-kl+/− embryos (Fig 3F), indicating that β-KL is involved in feto-maternal cholesterol transport. Consistent with the TLC analyses, we found that ¹²C-FC (the sum of exogenously supplied and endogenously synthesized cholesterol) tended to be decreased in embryos lacking β-kl (Fig 3F).

To clarify the contribution of β-KL in the yolk sac to growth and lipid regulation, we analyzed β-kl−/−/Tg mice, which were generated by crossing β-kl−/− mice with hepatocyte-specific β-kl transgenic (Tg) mice (Kobayashi et al, 2016). As expected, the β-kl mRNA level in the β-kl−/−/Tg embryos was comparable with that in β-kl+/− embryos (Fig S5A). β-kl expression in β-kl−/−/Tg yolk sacs was very low (Fig S5B), confirming the loss of β-KL expression in yolk sacs of β-kl−/−/Tg mice. The CRL of the β-kl−/−/Tg embryos tended to be shorter than that of β-kl−/− embryos (Fig S5C). As in β-kl−/− embryos, Hmgcr and Ldlr mRNA levels were elevated in the β-kl−/−/Tg embryos (Fig S5D). The expression of Lpl in the β-kl−/−/Tg yolk sacs was increased, though not significantly (Fig S5E). We evaluated the LDL uptake in the yolk sac in an ex vivo experiment (Fig S5F), because cholesterol-rich lipoproteins, such as LDL, are absorbed via receptor-mediated endocytosis. At E9.5, the fluorescence of BODIPY-labeled LDL was observed in the β-kl+/− yolk sac after 30 min of incubation (Fig S5G). BODIPY signals were also detected in β-kl−/− and β-kl−/−/Tg yolk sacs, showing that LDL uptake occurred in the absence of β-KL. The dye emission in β-kl−/− yolk sac shifted from green to red, which indicates the accumulation of BODIPY molecules. In tissue sections prepared from the yolk sacs incorporating BODIY-LDL, BODIPY-labeled vesicles were spread throughout cells in the β-kl+/−/Tg yolk sac. The vesicles were mainly localized close to the nucleus and seemed larger in the β-kl−/−/Tg yolk sac than in the β-kl+/−/Tg yolk sac (Fig S5H). Taken together, these results suggest that β-KL expressed in yolk sac is required for lipid transport through the yolk sac.

β-KL–FGF15 axis regulates growth and lipid metabolism in the embryonic stages

Hepatic β-KL regulates lipid and bile acid synthesis by mediating FGF15 signaling from the ileum in adult mice (Tomiyama et al, 2010; Kobayashi et al, 2016). Because Fgf15 mRNA is expressed in the developing central nervous system in mice (McWhirter et al, 1997; Gimeno et al, 2002), we examined whether β-KL interacts with FGF15 in the embryonic stages.

In contrast to β-kl, the Fgf15 mRNA level in E9.5 embryos was ∼170-fold higher than that in the yolk sac (Fig 4A). Notably, the Fgf15 level in E9.5 embryos was 1.6-fold higher than that in the adult ileum (Fig S6A). Our previous studies showed that Fgf15 mRNA expression in the ileum was elevated in adult β-kl−/− mice (Tomiyama et al, 2010; Kobayashi et al, 2016). Similarly, the level of Fgf15 mRNA in β-kl−/− embryos was significantly higher than that in their β-kl+/− littermates (Fig 4A).

Activation of FGF signaling induces ERK phosphorylation, which results in modulation of the expression levels of the downstream genes (Tomiyama et al, 2010). Dual-specific phosphatases (DUSPs) are a subclass of protein tyrosine phosphatases that promote a negative feedback loop to specifically dephosphorylate ERKs (Jeffrey et al, 2007). The genes encoding DUSP are targets of ERK, and therefore, their mRNA levels reflect ERK activity (Ekerot et al, 2008; Lan et al, 2017). To examine whether β-KL is required to transduce FGF signaling in embryonic tissues, we measured mRNA levels of DUSPs. At E9.5, Dusp4 and Dusp6 mRNA levels were decreased in β-kl−/− yolk sacs, but not in β-kl−/− embryos (Fig 4B). These data suggest that FGF15 is mainly produced in the embryo and activates ERK1/2 signaling through β-KL expressed in the yolk sac.

Next, we analyzed Fgf15−/− embryos and yolk sacs at E9.5 to verify the contribution of FGF15 to growth and lipid regulation. The CRL of Fgf15−/− embryos was significantly smaller than that of their WT or heterozygous littermates (Fig 5A), confirming that FGF15 is required for fetal growth. Fgf15−/− embryos displayed increased expression levels of genes involved in lipid synthesis and lipoprotein uptake (Fig 5B), in accordance with the pattern of expression in β-kl−/− embryos. The expression profile of genes involved in lipid metabolism in Fgf15−/− yolk sacs was consistent with that in β-kl−/− yolk sacs (Fig 5C). Oil Red O staining showed lipid accumulation in Fgf15−/− yolk sacs (Fig 5D). The levels of Dusp4 and Dusp6 mRNAs were down-regulated in Fgf15−/− yolk sacs (Fig 5E). The expression of Dusp4 and Dusp6 mRNA was significantly up-regulated in Fgf15−/− embryos, unlike β-kl−/− embryos. These data supported the idea that ERK1/2 signaling is activated in the yolk sac via the β-KL–FGF15 interaction.

In this study, we mainly used β-kl+/− embryos as a control. To examine whether one allele of β-kl was enough to suppress Fgf15 expression, we measured the Fgf15 levels in β-kl+/+ embryos and yolk sacs. β-kl mRNA was expressed dose-dependently in β-kl+/+ and β-kl+/− tissues (Fig S6B and C), whereas the Fgf15 mRNA levels were similar in β-kl+/+ and β-kl+/− embryos (Fig S6D). We also compared β-kl expression in Fgf15 mutant lines. Fgf15 mRNA was expressed dose-dependently in Fgf15+/+ and Fgf15+/− embryos (Fig S6E), but there was no difference in β-kl expression levels in the yolk sac among the genotypes (Fig S6F). These data show that one allele of β-kl is sufficient to regulate Fgf15 expression.

We next examined the contribution of FGF21, another partner of β-KL (Owen et al, 2015). Unlike Fgf15, the normalized read counts of Fgf21 were extremely low in E9.5 embryos, and no difference was observed between β-kl+/− and β-kl−/− embryos (Fig S6G).

Finally, we examined the levels of FGFRs in the yolk sac and embryo (Fig S6H). In adult mice, β-KL serves as a coreceptor of FGFR4 in the liver (Tomiyama et al, 2010). Among the four subtypes, Fgfr1 and Fgfr2 mRNAs showed significantly higher levels in embryos than in yolk sacs. Fgfr3 mRNA was expressed at a slightly, but significantly higher level in yolk sacs than in embryos. The Fgfr4 mRNA level in yolk sacs was 15-fold higher than that in embryos, suggesting that FGFR4 in the yolk sac is required to receive the FGF15 signal from the embryo.

Animals

β-kl−/− mice, Fgf15−/− mice, and hepatocyte-specific β-kl transgenic mice were generated as previously described (Wright et al, 2004; Ito et al, 2005; Kobayashi et al, 2016). β-kl−/− mice were maintained on a mixed background. Hepatocyte-specific β-kl transgenic mice were identified by using forward primer 5′-CAT ATA AAT TCT GGC TGG CGT G-3′ and reverse primer 5′-GTT GAC ACC TCT CAG GTG TGA GT-3′. The sex of the embryos was determined by using primers for Zfy (forward: 5′-CCT ATT GCA TGG ACT GCA GCT TAT G-3′, reverse: 5′-GAC TAG ACA TGT CTT AAC ATC TGT CC-3′) (Bradford et al, 2009). C57BL/6J mice were purchased from CLEA Japan. To obtain embryos, one male mouse was allowed to cohabit with one or two female mice overnight. Noon on the day of detection of the vaginal plug was considered as E0.5. The genotype of the mated mice is indicated in the figure legends. Mice were maintained on a 12-h dark–light cycle and given free access to food and water. All animals were euthanized by cervical dislocation or isoflurane inhalation. All experiments were approved by RIKEN Center for Biosystems Dynamics Research, Foundation for Biomedical Research and Innovation at Kobe, and Kyoto University.

Sample preparation

Based on the milestones of intrauterine development in mice (Cross et al, 1994; Rossant & Cross, 2001), tissues were collected at E3.5, E9.5, and E12.5, corresponding to preimplantation, yolk sac-dependent, and placenta-dependent stages, respectively. Blastocysts were obtained at E3.5 as previously described (Nagy et al, 2002). After microscopic observation, the blastocysts were lysed and used as a template for genotyping PCR (Honda et al, 2019). Whole embryos and yolk sacs were collected at E9.5 under microscopic observation. Fetal livers, yolk sacs, and placentas were dissected at E12.5. To avoid contamination with maternal tissues or blood, the specimens were washed with PBS. A portion of the yolk sac, allantois or embryo was used for genotyping. Adult livers and small intestines were collected from male C57BL/6J mice (10 wk of age) in a fed state. The small intestine was divided into three parts (0%, 33%, 66%, and 100%; relative distance from the stomach to distal ileum) and a piece of tissue collected from the middle of the 66–100% part was used as the ileum. Each sample was prepared from a different animal, unless mentioned in the figure legends.

RNA-sequencing analysis

Total RNA was extracted from E9.5 embryos using an RNeasy Mini kit (QIAGEN) and treated with DNase I (QIAGEN) on-column according to the manufacturer’s instructions. The quality of RNA was confirmed by using an Experion Automated Electrophoresis System (Bio-Rad). Sequencing was performed at 100 bp paired-ends on an Illumina HiSeq 2500 with all eight libraries multiplexed in one lane. Base-calling was processed with RTA 1.18.64 (Real-Time Analysis, HiSeq Control Software). Fastq files were generated with bcl2fastq 1.8.4 (Illumina). The quality of the RNA-Seq reads was evaluated using trimmomatic (ver. 0.32). Having established the high quality of the data, sequence reads for each library were mapped using BWA (ver. 0.7.15). Next, for each library, we estimated the number of sequence reads overlapping at any given nucleotide position in the reference genome at 100-bp resolution. The expression data were quantified using samtools version 0.1.19. Differentially expressed genes were analyzed using integrated Differential Expression and Pathway analysis (iDEP ver. 0.93) (Ge et al, 2018; Ge, 2021).

qRT-PCR analysis

Total RNA was extracted from whole embryo (E9.5), yolk sac (E9.5 or E12.5), liver (E12.5) or placenta (E12.5) using an RNeasy Mini Kit (QIAGEN) or AllPrep DNA/RNA/Protein Mini Kit (QIAGEN). Total RNA treated with DNase I (QIAGEN) on-column was transcribed into first-strand cDNA using a SuperScript VILO Master Mix (Thermo Fisher Scientific) according to the manufacturer’s instructions. qRT-PCR analysis was performed by using TaqMan Expression Assay Primers (Applied Biosystems, Table S4). Expression levels were calculated by the relative standard curve method with 18S rRNA as an internal control and the data were shown as relative expression compared with the control group. In the comparison between fetal tissues and adult tissues, data were shown as fold increase over the average expression levels in the tissues of male C57BL/6J mice.

Western blotting

Tissue homogenates of yolk sac (E9.5) and adult liver (male β-kl+/− mice, 12 wk of age) were prepared in buffer (20 mM 4-[2- hydroxyethyl]-1-piperazineethanesulfonic acid, 150 mM NaCl, and 0.5% Nonidet P-40 [pH 7.4]) containing protease inhibitor cocktail (cOmplete, Mini, EDTA-free; Roche), and phosphatase inhibitor cocktail (Nacalai tesque). The supernatant was electrophoresed on 4–20% gradient SDS-polyacrylamide gel and transferred to PVDF membranes. After blocking with 5% skim milk in TBS-T for 1 h at room temperature, the membrane was incubated overnight with goat anti-mouse KLB polyclonal antibody (AF2619, 1:1,000; R&D Systems) at 4°C. Then, the membrane was incubated with anti-goat HRP (705-035-147, 1:100,000; Jackson) for 45 min at room temperature. The signals were developed by using the ECL Prime Western Blotting Detection Reagents (Cytiva) and detected with a FUSION SYSTEM (Vilber-Lourmat). After stripping, the membrane was reprobed with rabbit anti-β-actin monoclonal antibody (4970, 1:1,000; Cell Signaling) and anti-rabbit HRP (A16110, 1:200,000; Thermo Fisher Scientific).

TLC

Whole embryo and yolk sac without allantois were collected at E9.5. Allantoises were used for genotyping. Placentas were dissected at E12.5. Tissues were homogenized in 880 μl of hexane/isopropanol (3:2, vol/vol) by using zirconia beads (Sarstadt) in a Micro Smash (TOMY) at 4°C. The homogenate was centrifuged at 10,000g for 5 min at 4°C. Lipids were extracted from the supernatants as previously described (Hara & Radin, 1978; Matsuoka et al, 2001). The pellets were solubilized with 2% SDS solution, and protein concentrations were determined by the BCA method. Extracted lipids were evaporated and resolubilized in chloroform, then applied to HPTLC plates (Merck). TLC nonpolar Lipid Mixture B (Matreya) was loaded as a loading control. Free cholesterol solubilized in ethanol (20 μg/plate) was used as an internal standard. The plates were developed in hexane/diethyl ether/acetic acid (100:15:1, v:v:v) at room temperature. Lipids were stained with Coomassie brilliant blue (Nakamura & Honda, 1984). Spot images were obtained by scanning with a FUSION SYSTEM (Vilber-Lourmat). The signal intensity corresponding to cholesteryl ester and triglyceride in each sample was quantified using ImageJ’s Fiji software (Schindelin et al, 2012), and normalized by the signal intensity of the internal standard. Signals were divided by protein concentration and expressed as normalized relative intensity per μg of protein.

Tracer experiment

One milligram of [2,3,4–¹³C]-cholesterol (Cambridge Isotope Laboratories) dissolved in 40 μl of ethanol was mixed with 10% Intralipos (Otsuka Pharmaceutical) (1:1, vol/vol) just before administration, and the lipid solution was intraperitoneally injected into pregnant β-kl+/− mice at E8.5. Then, the whole embryos were collected at E9.5 under microscopic observation. The yolk sacs and allantoises were removed. The collected embryos were homogenized in 1.1 ml of hexane and isopropanol (3:2, v:v), and the homogenate was centrifuged at 15,000g. The resulting supernatant (1.0 ml) was collected and stored at −80°C until use. Free cholesterol contents in each whole embryo with/without [2,3,4–¹³C] labeling were determined by GC-MS (GCMS-QP2010Ultra; Shimadzu) as previously described (Yoshida & Wada, 2005). Cholesterol (D7) (Avanti Polar Lipids) was added to the sample tube as an internal standard. The content of free cholesterol in the tissue was calculated using standard curves generated by using unlabeled cholesterol or [2,3,4–¹³C]-cholesterol (Cambridge Isotope Laboratories).

Histologic analysis

Yolk sacs (E9.5) were fixed in 10% formalin (Wako) and embedded in paraffin. The paraffin sections were stained with hematoxylin and eosin. For Oil Red O staining, yolk sacs (E9.5) were fixed in 4% PFA for 1 h at 4°C, left on a 30% sucrose solution for 1 h at 4°C, and frozen in Tissue-Tek optimum cutting temperature compound (Sakura) for cryo-sectioning. Sections (14 μm) were mounted on glass slides (MATSUNAMI). Oil Red O (0.3%) in isopropanol solution was freshly prepared, mixed with H₂O (0.3% Oil Red O isopropanol: H₂O 3:2), and filtered through filter paper (Whatman). Sections were incubated in 60% isopropanol for 1 min and then in Oil Red O solution for 15 min and 60% isopropanol for 1 min. After rinsing in water, the sections were mounted in mounting media (Fluoromount-G; Thermo Fisher Scientific).

LDL uptake assay

E9.5 embryo with yolk sac was dissected in PBS and incubated in 75% lipoprotein-deficient fetal calf serum (Sigma-Aldrich) and 25% DMEM for 30 min at 37°C in 5% CO₂. BODIPY FL-labeled LDL (BODIPY-LDL) (Thermo Fisher Scientific) was added at 40 μg/ml for 30 min at 37°C. After uptake, the embryo with yolk sac was washed in ice-cold PBS (+) three times to remove surface-bound BODIPY-LDL. The yolk sac was separated from the embryo and fixed in 4% PFA for 1 h at 4°C. The embryo was used for genotyping PCR. Fixed yolk sac was embedded in an optimum cutting temperature compound and sectioned by a cryostat. Fourteen μm cryostat sections were rehydrated in PBS and permeabilized in 0.1% Triton X-100 in PBS for 20 min at room temperature. Tissues were washed in PBS containing 0.1% Tween 20 (PBST) 3 times, and incubated with 4′,6-diamidino-2-phenylindole (DAPI, 1:1,000) (Dojindo) for 15 min. Sections were washed in PBS three times and mounted using Fluoromount-G (Thermo Fisher Scientific). Fluorescence images were obtained by Nikon A1R (Nikon). The color, brightness, and contrast of the images were adjusted using ImageJ’s Fiji software.

Statistics

All statistical analyses were performed with EZR (version 1.52) (Kanda, 2013). Statistical analyses were performed using a two-tailed t test or Mann–Whitney’s U-test (two groups) as indicated in figure legends. Bonferroni correction was applied for multiple-group analysis (three or more groups). P < 0.05 was considered significant. The number of replicates is given in each figure legend.