Hyperlipidemia and hepatitis in liver-specific CREB3L3 knockout mice generated using a one-step CRISPR/Cas9 system

cAMP responsive element binding protein 3-like 3 (CREB3L3), a transcription factor expressed in the liver and small intestine, governs fasting-response energy homeostasis. Tissue-specific CREB3L3 knockout mice have not been generated till date. To our knowledge, this is the first study using the one-step CRISPR/Cas9 system to generate CREB3L3 floxed mice and subsequently obtain liver- and small intestine-specific Creb3l3 knockout (LKO and IKO, respectively) mice. While LKO mice as well as global KO mice developed hypertriglyceridemia, LKO mice exhibited hypercholesterolemia in contrast to hypocholesterolemia in global KO mice. LKO mice demonstrated up-regulation of hepatic Srebf2 and its corresponding target genes. No phenotypic differences were observed between IKO and floxed mice. Severe liver injury was observed in LKO mice fed a methionine-choline deficient diet, a model for non-alcoholic steatohepatitis. These results provide new evidence regarding the hepatic CREB3L3 role in plasma triglyceride metabolism and hepatic and intestinal CREB3L3 contributions to cholesterol metabolism.

genetically modified mice and rats has been greatly accelerated by recently developed approaches using direct injection of DNA or mRNA encoding site-specific nucleases into one-cell-stage embryos, thereby generating DNA double-strand breaks (DSBs) at specified sequences leading to targeted mutations. Co-injection of single-stranded or double-stranded DNA templates homologous to sequences flanking DSB can produce mutant alleles with precise point mutations or DNA inserts. Engineered endonucleases, including zinc-finger nucleases (ZNFs), transcription activator-like effector nucleases (TALENs), and the clustered regularly interspaced short palindromic repeat (CRISPR)/CRISPR associated protein 9 (Cas9) system have all demonstrated utility in the rapid generation of genetically modified animals. The CRISPR/Cas9 system is an RNA-mediated adaptive immune system found in bacteria and archaea that protects against the invasion of viruses and plasmids 7 . On the basis of locus organization and signature Cas gene composition, three major types of CRISPR systems (type I-III) have been identified 8 . The modified type II CRISPR/Cas9 system, derived from Streptococcus pyogenes, is widely used for gene editing 8,9 . The type II bacterial CRISPR/Cas9 system has been demonstrated as an efficient gene-targeting technology that facilitates multiplexed gene targeting. Because the binding of Cas9 is guided by simple base-pair complementarities between engineered single-guide RNA (sgRNA) and target genomic DNA sequences, it is possible to direct Cas9 to any genomic locus by providing engineered sgRNA. Previous reports have demonstrated that the generation of floxed mice in one step by the insertion of two loxP sites into the same allele of genes using the CRISPR/Cas9 system [10][11][12] . However, the generation of tissue-specific knockout (KO) mice by crossing floxed mice developed using the methods outlined above with tissue-specific Cre Tg mice has yet to reported.
Until now, CREB3L3 loss of function studies have relied on Creb3l3 −/− mice. However, the contribution of intestinal CREB3L3 to lipid metabolism has yet to be specifically examined. Accordingly, conditional Creb3l3 knockout mice are required to evaluate the tissue-specific functions of CREB3L3. In the present study, we generated Creb3l3 floxed mice using the one-step CRISPR/Cas9 system. For the first time, both liver-specific Creb3l3 knockout (LKO) and intestine-specific Creb3l3 knockout (IKO) mice were successfully generated by crossing Creb3l3 floxed mice with albumin-and villin-promoter Cre Tg mice, respectively. Phenotypic characteristics were then compared between tissue-specific CREB3L3 KO mice.

Results
Generation of Creb3l3 floxed mice using a one-step CRISPR/Cas9 system. CREB3L3 is a key regulator of glucose and lipid metabolism. Although Creb3l3 −/− mice have previously been generated 13 , the generation of conditional Creb3l3 knockout mice has yet to be reported. Since CREB3L3 is specifically expressed in the liver and intestine, it is important to elucidate the differential functions of CREB3L3 in these organs. Accordingly, we aimed to generate Creb3l3 floxed (floxed) mice using the CRISPR/Cas9 system. Our plan was to insert flox site in intron 3 and 11 of the Creb3l3 gene, respectively, and then delete the about 7kbp between both sites by the Cre-loxP system. A px330 vector was used to express gRNA for the targeted region of the gene of interest (Creb3l3 in the present study) and Cas9 protein in mouse zygotes 14 . We identified appropriate CRISPR target sites in introns 3 and 11 of the Creb3l3 gene. Double-stranded DNAs (dsDNAs) of 20 bp were derived from regions of introns 3 and 11 of the Creb3l3 gene and inserted into the px330 vector, with the resultant plasmids designated as px330-Creb3l3 intron 3 and px330-Creb3l3 intron 11, respectively. We designed corresponding loxP site oligos with 250bp sequences homologous to each side surrounding each sgRNA-mediated DSB. To facilitate the detection of the intended insertions, DNA fragments targeting intron 3 were engineered to contain a BamHI restriction site and other oligos targeting intron 11 to contain an EcoRI site in addition to the loxP sequences, respectively ( Fig. 1a and Supplementary Fig. 1).
The EGxxFP system 15 involves the formation of a fragment of pCX-EGxxFP-Creb3l3 cleaved from the px330-Creb3l3 vector in order to produce full-length functional EGFP by HDR. The cleavage activity of the px330-Creb3l3 intron 3 and px330-Creb3l3 intron 11 vectors were confirmed using the EGxxFP system. Fragments of intron 3 and intron 11 of the Creb3l3 gene were inserted into the multi-cloning site of pCX-EgxxFP, with the resultant vectors designated pCX-EGxxFP-Creb3l3 intron 3 and pCX-EGxxFP-Creb3l3 intron 11, respectively. To confirm the above mechanism, the target sequence of Creb3l3 intron 3 or 11 for pCX-EGxxFP-Creb3l3 and px330-Creb3l3 vectors were co-transfected into HEK293T cells, respectively, and EGFP signals were observed in co-transfected cells. EGFP signals were clearly detected following co-transfection with both pX330 Creb3l3 and pCX-EGxxFP vectors (Fig. 1b). We further confirmed the ability of px330-Creb3l3 vectors to cleave both intron 3 and intron 11 of the Creb3l3 genome, respectively.
Next, the px330-Creb3l3 intron 3 and px330-Creb3l3 intron 11 vectors, and two dsDNA fragments intron 3 and intron 11 containing a loxP sequence, were injected into mouse 277 zygotes and 239 embryos were transferred into the oviducts of pseudopregnant recipient ICR mice with the aim of generating 20 pups (Fig. 1c). Genomic DNA from pups born after microinjection were amplified by PCR and subjected to digestion with BamHI and EcoRI separately. Genomic DNA from pups 4 and 8 (lanes 4 and 8 in Fig. 1d,e) were not detected by genotyping PCR for intron 3 and intron 11. This result demonstrates an important issue with the one step CRISPR/Cas9 system. Introns 3 and 11 of genomic Creb3l3 were predicted to be completely deleted in these mice. Fortunately, we identified one mouse (No. 6) which carried loxP sites at both the intron 3 and intron 11 sites of the same allele (Fig. 1d,e). It was confirmed that the loxP site within intron 3 was heterozygously inserted; however, the intron 11 site was homozygously inserted in this mouse (Fig. 1c-e).
Generation of liver-and intestine-specific Creb3l3 knockout mice. As Creb3l3 is predominantly expressed in the liver and small intestine 2 , we aimed to specifically delete Creb3l3 in these tissues. The lox sites were homozygously inserted in both sites of genomic Creb3l3 in the floxed mice, which were confirmed by sequencing ( Supplementary Fig. 2). To analysis off-target cleavages, we searched candidates that matches 12 bases at the 3′ end and the NGG PAM sequence 15 with CRISPR direct (https://crispr.dbcls.jp/). Here we examined 6 potential off-target sites, resulting in no mutations ( Supplementary Fig. 3). To determine whether the px330-Creb3l3 intron 3 and px330-Creb3l3 intron 11 vectors were integrated into the chromosomes, we (a) Method used to generate Creb3l3 floxed mice using the one-step CRISPR/Cas9 system. (b) Creb3l3 CRISPR activity was evaluated using the EGxxFP system. Robust EGFP signals were observed following the co-transfection of HEK293T cells with the pCX-EGxxFP-Creb3l3 intron vector and either the pX330 empty vector or px330-Creb3l3 intron vector. Scale bars represent 100 μ m. (c) Generation of mutant mice via two vectors and two DNA fragments injection. (d,e) Genotyping to confirm insertion of loxP sites into Creb3l3 introns 3 (d) and 11 (e). Mouse tail genomic DNA was isolated and subjected to genotyping PCR. PCR products were cut using BamHI or EcoRI restriction enzymes. performed PCR assay with mouse genomic DNA and a primer pair for Cas9 detection. Cas9 PCR products were not observed in the floxed mice ( Supplementary Fig. 4). Generated floxed mice were then crossed with mice expressing tissue-specific Cre Tg mice (Fig. 2a,b). We used the albumin-promoter Cre Tg mice for the generation of LKO mice and the villin-promoter Cre Tg mice for IKO mice. To confirm the generation of LKO and IKO mice, we assessed gene and protein expression of CREB3L3 in the liver and intestine using qPCR and western blotting. CREB3L3 mRNA and protein were specifically deleted in the liver of LKO mice and intestine of IKO mice, respectively (Fig. 2c,d). Levels of CREB3L3 mRNA and protein in the intestine of LKO mice and liver of IKO mice were similar to levels measured in CREB3L3 floxed mice. As a result, we obtained tissue-specific CREB3L3 knockout mice using a one-step injection method in markedly shorter time than reported for previous techniques. To our knowledge, the present study represents the first report of the use of the one-step CRISPR/Cas9 system in generating tissue-specific knockout mice. Although previous reports have demonstrated the generation of floxed mice, none have reported the application of this approach to the generation of knockout mice 10 .
Liver-specific Creb3l3 knockout mice and not intestine-specific knockout mice demonstrate hyperlipidemia. No significant differences in body weight or lean weight were observed between floxed, LKO, and IKO mice in both fasted and fed states, except an increased weight proportion attributable to fat in IKO mice in the fed state compared to floxed mice as estimated by DEXA scanning (Supplementary Table 1). Plasma TG and glucose levels have been reported to be higher, with cholesterol and FGF21 levels reported lower, in Creb3l3 −/− mice compared to WT mice 5,16 . Plasma TG, cholesterol, and free fatty acid levels were significantly increased in LKO mice compared to floxed mice in both fasted and fed conditions (Fig. 3A,B). High-performance liquid chromatography (HPLC) demonstrated CM, VLDL, and LDL fractions of TG were markedly increased in LKO mice in both fasted and fed states (Fig. 3A). Plasma FGF21 levels were markedly decreased in LKO mice in the fasted state, but were unchanged in the fed state (Fig. 3B). No difference in plasma glucose levels were observed between LKO and floxed mice (Fig. 3C). Plasma insulin levels were increased in LKO mice in   Table 2). Histological analyses of liver sections following H&E staining, demonstrated no apparent morphological differences between LKO and floxed mice (Fig. 3D). No difference in hepatic lipid content, including TG and cholesterol, were observed between floxed and LKO mice in both fasted and fed states, with  Liver-specific Creb3l3 knockout mice have dysfunctional TG clearance but not VLDL secretion.
Previous reports have described lower plasma LPL activity in Creb3l3 −/− mice leading to the lower TG clearance compared to WT mice, with no effect on VLDL secretion 1 . LKO mice were subjected to evaluation of VLDL secretion and TG clearance. In VLDL secretion tests, no difference in plasma TG levels was observed between LKO and floxed mice, indicating normal VLDL secretion in LKO mice (Fig. 3F). There was a trend toward increased plasma cholesterol levels in LKO mice compared to floxed mice from 1 h after injection (Fig. 3F). In TG clearance testing, plasma TG levels were significantly higher in LKO mice compared to floxed mice (Fig. 3G). These results indicate that the dysregulation of TG clearance in LKO mice can be explained by hepatic Creb3l3 deficiency.
Hepatic deletion of Creb3l3 dysregulates the expression of genes related to lipid metabolism.
Next we evaluated the expression of genes related to lipid metabolism including TG, cholesterol, and fatty acids in the livers of LKO mice in fasted and fed states (Fig. 4). No significant changes in gene expression levels were observed in the fed state. CREB3L3 is predominantly active in the fasted state, rather than the fed state. Thus, this result was expected. In the fasted state, expression levels of Ppara, another key regulator of fatty acid oxidation, were reduced in LKO mice. Consistently, hepatic expression levels of the PPARα target gene, carnitine palmitoyltransferase 1a (Cpt1a), were also reduced. The expression of another target gene, acyl-Coenzyme A oxidase 1, palmitoyl (Acox1), were significantly increased. No differences in the expression levels of regulators of fatty acid synthesis, such as sterol regulatory element binding transcription factor 1a (Srebf1a) and Srebf1c, and their corresponding target genes, stearoyl-Coenzyme A desaturase 1 (Scd1), and ELOVL family member 6, elongation of long chain fatty acids (yeast) (Elovl6), were observed between floxed and LKO mice in either nutritional state, but fatty acid synthase (Fasn) was significantly increased. On the other hand, the cholesterol synthesis regulator, Srebbf2, was significantly increased in the fasted state, leading to an increase in its corresponding target genes including 3-hydroxy-3-methylglutaryl-Coenzyme A synthase 1 (Hmgcs1), 3-hydroxy-3 -methylglutaryl-Coenzyme A reductase (Hmgcr), and low density lipoprotein receptor (Ldlr). Consistent with a previous report in Creb3l3 −/− mice 16 , fatty acid elongases, such as Elovl2 and Elovl5, were significantly decreased in LKO mice in the fasted state indicating a changed composition of fatty acids in the livers of LKO mice. Of the examined lipoprotein lipase (LPL) activity regulators, expression levels of activators such as Apoa4, Apoa5, and Apoc2 were significantly reduced, while expression levels of a suppressor, Apoc3, were conversely increased, not significant, in livers of LKO mice compared to floxed mice in the fasted state. These results provide evidence of the dysregulation of TG clearance in LKO mice, consistent with previously reported phenotypes of Creb3l3 −/− mice. No significant differences in expression levels of bile acid metabolic genes, nuclear receptor subfamily 1, group H, member 4 (Fxr encoded by Nr1h4), small heterodimer partner (Shp), or cytochrome P450, family 7, subfamily a, polypeptide 1 (Cyp7a1) were observed between LKO and floxed mice. In consistent with previous reports about Creb3l3 −/− mice 1,4,5,14 , changes in the hepatic expression of genes related to fatty acid oxidation, fatty acid elongation, and apolipoprotein in LKO mice were same as Creb3l3 −/− mice. LKO mice especially exhibited the increased hepatic expression of genes related to cholesterol synthesis. Given these changes in gene expression, we suggest the increased hepatic cholesterol synthesis and reductions in plasma LPL activity observed in LKO mice contribute to hyperlipidemia.

Deficiency of Creb3l3 in the liver accelerates non-alcoholic steatohepatitis.
To determine the effects of hepatic CREB3L3 on NASH pathology, LKO mice were fed the methionine-choline deficient (MCD) diet for 4 weeks leading to the induction of non-alcoholic fatty liver disease (NAFLD). Plasma alanine aminotransferase (ALT) and aspartate aminotransferase (AST) levels were acutely and significantly increased in LKO mice fed an MCD diet, indicating severe and acute liver injury (Fig. 5a). Histological analysis of sections stained with H&E demonstrated disrupted hepatic architecture in LKO mice, with Masson's trichrome (MT) staining demonstrating greater induction of fibrosis in LKO mice compared to floxed mice (Fig. 5b). Surprisingly, no differences in hepatic TG and cholesterol contents were observed between floxed and LKO mice (Fig. 5c). Consistent with the results of the MT staining analysis, hepatic gene expression studies demonstrated significantly increased hepatic expression of fibrosis-related genes such as transforming growth factor, beta 1 (Tgfb1), actin, alpha 2, smooth muscle, aorta (Acta2), and collagen, type I, alpha 1 (Col1a1) in LKO mice (Fig. 5c). There was a trend toward decreased hepatic expression of the CREB3L3 target gene, In FGF21 KO mice fed the MCD diet, non-alcoholic steatohepatitis (NASH) phenotypes were induced 17 . Fgf21 was decreased in LKO mice, thereby in part contributing to the pathology of NASH. There was a trend toward increased expression levels of inflammatory mediators including Il6, Il1b, and Ccl2; however, these differences were not statistically significant. Interestingly, no significant differences were observed in the expression levels of fatty acid and cholesterol synthesis regulator genes, such as Srebf1c and Srebf2, between LKO and floxed mice (Fig. 5c). These results indicate hepatic Creb3l3 deficiency acutely exacerbates MCD diet-induced liver injury and fibrosis, without abnormal hepatic lipid accumulation.

Discussion
There are three previous studies reporting the generation of conditional mutant mice by introduction of two flox sites using the CRISPR/Cas9 system [10][11][12] . However, no reported studies have used the one-step CRISPR/Cas9 system to generate floxed mice and subsequently tissue-specific knockout mice. The present study is the first to report the generation of tissue-specific knockout mice and to analyze the function of the gene of interest in mice. We generated Creb3l3 floxed mice by direct microinjection of 2 vectors and 2 DNA fragments into C57BL/6J zygotes using the CRISPR/Cas9 system. Creb3l3 is predominantly expressed in the liver and small intestine. We intended to generate liver-specific and intestine-specific Creb3l3 knockout mice by crossing Creb3l3 floxed mice with tissue-specific promoter-driven Cre Tg mice, respectively. In contrast to Creb3l3 −/− mice reported to have high plasma TG levels only 1 , LKO mice demonstrated higher plasma TG and cholesterol levels but not hepatic steatosis in the fasting condition. These changes were considered to be attributable to intestinal expression of Creb3l3 in addition to hepatic expression. Collectively, we propose that CREB3L3 functions as a mediator of lipid metabolism via regulation of the enterohepatic circulation. Previous reports have demonstrated the generation of floxed mice using the one-step CRISPR/Cas9 system; however, no studies have demonstrated the generation of tissue specific knockout mice 10 . The major limitation of the one step system is that in situations where 2 sites at the same allele are ligated, the genomic portion between 2 sites could be deleted leading to failed insertion of the loxP sites. Furthermore, the frequency of the insertion of 2 loxP sites into the same allele may be extremely low. We obtained only one line of Creb3l3 floxed mice with 2 loxP sites within the same allele.
We generated both LKO and IKO mice through breeding with mice generated using the one step CRISPR/ Cas9 system. IKO mice had no apparent differences in metabolic parameters compared to floxed mice, while LKO mice demonstrated hyperlipidemia. These findings indicate hepatic CREB3L3 contributes to cholesterol and TG homeostasis. Consistent with previous reports, hepatic expression of LPL activator genes was reduced and suppressor genes was increased in LKO mice, underlying the observed increases in plasma TG levels in LKO mice. Hepatic de novo cholesterol synthesis was apparently activated leading to increased plasma cholesterol levels in LKO mice. Accordingly, hepatic expression levels of genes related to cholesterol synthesis were higher in LKO mice compared to floxed mice. Despite the higher plasma cholesterol levels in LKO mice, no differences in hepatic cholesterol content were observed between floxed and LKO mice. Expression levels of Srebf2 are known to be regulated by a feedback loop between gene expression and cholesterol levels. Therefore, increased expression Scientific RepoRts | 6:27857 | DOI: 10.1038/srep27857 of Srebf2 and its corresponding target genes led to up-regulation of de novo cholesterol synthesis in LKO mice, resulting in increased plasma cholesterol levels. SREBP-2, a regulator of the expression of genes involved in cholesterol synthesis, is similar to CREB3L3 as both molecules have a transmembrane domain cleaved by site-1 and site-2 proteases. We hypothesized that CREB3L3 and SREBP-2 compete as targets of these cleavage enzymes, the detailed mechanism of which is described in a manuscript currently under submission (Nakagawa Y, paper in preparation). In addition, FGF21 suppresses Srebf2 expression in the liver of Apoe −/− mice 18 . Plasma FGF21 levels were lower in LKO mice, but not IKO mice than floxed mice. The results of the present study indicate liver-specific deletion of CREB3L3 leads to activation of Srebf2 expression and the SREBP cleavage process resulting in the stimulation of SREBP-2 protein activity. These changes were shown in only LKO mice, not Creb3l3 −/− mice, thereby supporting that the function of CREB3L3 in enterohepatic circulation contributes to the regulation of hepatic SREBP2 activation.
We hypothesized that lipid metabolism dysregulation in LKO mice influences the pathogenesis of NAFLD in response to the MCD diet. Surprisingly, LKO mice demonstrated increased liver injury and fibrosis in the absence of the excess hepatic lipid accumulation compared to floxed mice, indicating liver-specific deletion of Creb3l3 accelerates the development of non-alcoholic steatohepatitis. MCD diet-fed Fgf21 −/− mice also exhibit NASH phenotypes 17 , thereby supporting that a decrease of Fgf21 in LKO contributes to same phenotypes. In addition, we are aware of data demonstrating increased inflammation and infiltration of macrophages in Creb3l3 −/− mice fed the MCD diet along with liver injury and fibrosis (Nakagawa Y, unpublished data). NAFLD patients, and in particular patients with NASH, are more likely to exhibit increased intestinal permeability compared with healthy controls 19 . Likewise, MCD-fed mice develop intestinal permeability changes after an initial phase of liver injury 19 . NASH is associated with increased intestinal permeability suggesting the early phase of liver injury and inflammation contributes to injury to the intestinal barrier, thereby indicating the contribution of gut-liver crosstalk in NASH pathogenesis 20 . Taken together, these results indicate enhanced disease progression in Creb3l3 −/− mice is due to aberrant inflammation in mice with intestine-specific deletion of CREB3L3 in addition to mice with liver-specific deletion of CREB3L3. According, CREB3L3 may contribute to the pathogenesis of metabolic syndrome through its role in regulating enterohepatic circulation. However, further studies are required to verify the intestinal function of CREB3L3 in models of metabolic disease such as those fed the MCD diet. The present study demonstrated the utility of the one-step CRISPR/Cas9 system as a simple and less time-consuming method for generating genetically modified mouse models. The findings of the present study further demonstrate a crucial role for CREB3L3 in the pathogenesis of NAFLD through its function in regulating enterohepatic circulation. Future studies using intestine-specific Creb3l3 knockout mice may demonstrate CREB3L3 as a therapeutic target for improving hyperlipidemia.

Experimental procedures
Animals. This project was approved and performed in accordance with the guidelines of the Animal Care Committee of the University of Tsukuba. Creb3l3 tm1.1Sad /J (Creb3l3 −/− ) 13 , B6.Cg-Tg (Alb-cre) 21Mgn/J (Albumin Cre Tg) 21 , and B6.Cg-Tg (Vil1-cre) 1000Gum/J (Villin Cre Tg) mice were purchased from Jackson Laboratory. For the NAFLD study, eight-week-old male mice were fed the MCD diet (Oriental Yeast, Tokyo, Japan) for 4 weeks. Fat and lean mass analyses were performed using dual-energy X-ray absorptiometry (DEXA) with a PIXImus mouse densitometer (GE Medical Systems Lunar). All animal husbandry procedures and animal experiments were consistent with the University of Tsukuba's Regulation of Animal Experiment and were approved by the Animal Experiment Committee, University of Tsukuba.
Vector construction. The px330 vector (Addgene plasmid 42230) was a gift from Dr. Feng Zhang 14 .
Microinjection. Female C57BL/6J mice were injected with pregnant mare serum gonadotropin (PMSG) and human chorionic gonadotropin (hCG) at 48-h intervals and mated with male C57BL/6J mice. Fertilized one-cell embryos were collected from oviducts. Then, 5 ng/μ l of px330 vector (circular) and 10 ng/μ l dsDNA donors were injected into the pronuclei of one-cell-stage embryos according to standard protocols (Gordon and Ruddle 1981). Injected one-cell embryos were then transferred into pseudopregnant ICR mice.
Off-target analysis. Potential off-target sites were found using CRISPR target with rules outlined previously 15 . Thirteen bases proceeding the PAM sequence with NGG were aligned with mouse genome. The genome fragments containing the off-target were sequenced.
Determination of metabolic parameters. Plasma glucose, TG, cholesterol, NEFA, AST, and ALT concentrations were measured using Wako enzymatic kits. Plasma insulin was measured using mouse insulin ELISA Kits (Sibayagi). Plasma FGF21 concentrations were measured using a mouse/rat FGF21 Quantikine ELISA assay (R&D Systems). Hepatic TG and cholesterol levels were measured using previously described protocols 22 .
Scientific RepoRts | 6:27857 | DOI: 10.1038/srep27857 High-performance Liquid Chromatography Analysis. For lipoprotein distribution analyses, pooled plasma samples from 4 to 5 mice per group were analyzed using an upgraded high-performance liquid chromatography (HPLC) technique as previously described (Skylight Biotech) 23 .
Immunoblotting. Total cell lysates from mouse livers were prepared as previously described 22 , separated by SDS-PAGE, and subjected to Western blot analysis using anti-CREB3L3 5 , α -Tubulin (05-829, Millipore), and β -Actin (sc-47778 Santa Cruz) antibodies. TG Production assay. Mice were fasted for 24 h prior to injection with Triton WR-1339 (0.5 mg/g body weight, Sigma-Aldrich) via the tail vein to block the clearance of nascent ApoB-containing lipoproteins. Blood samples were collected at 0, 30, 60, and 120 min after Triton injection 23 .
Postprandial TG Response assay. Mice were fasted for 16 h prior to the oral administration of 200 μ l olive oil 1 . Blood samples were collected at 0, 3, 6, and 9 h after olive oil administration.
Analysis of gene expression. Total RNA from cells and tissues was prepared using Trizol reagent (Invitrogen). Templates for real-time PCR analysis were prepared by cDNA synthesis (Invitrogen) using total RNA. Real-time PCR was performed using the ABI Prism 7300 System (ABI) and SYBR Green Master Mix (Roche) 24 . Primer sequences are listed in Supplementary Table 3.

Statistical analyses.
Statistical significance was determined using unpaired Student's t-tests. Differences with P < 0.05 were considered significant. Data are expressed as the mean ± SEM.