To investigate long term Surf4 inactivation in hepatocytes in vivo, we generated mice with the Surf4 gene genetically inactivated specifically in the liver by combining a previously reported conditional Surf4 allele (in which Surf4 exon 2 is flanked by loxP sites, denoted Surf4^fl) (Wang et al., 2021b) with a Cre recombinase transgene under control of the Albumin promoter (Alb-Cre). Surf4^fl/fl Alb-Cre⁺ mice were observed at the expected Mendelian ratio (Table 1). Both male and female Surf4^fl/fl Alb-Cre⁺ mice are fertile and produce offspring of the predicted genotypes at expected Mendelian ratios (Table 1).

Excision of exon 2 is predicted to result in a frameshift mutation and the generation of a premature termination codon 8 base pairs downstream of the new exon1-3 junction (Figure 1A). Analysis of genomic DNA collected from mouse tails and livers of Surf4^fl/fl Alb-Cre⁺ mice demonstrated efficient Cre-mediated excision of Surf4 exon 2 only in the liver (Figure 1B), with the level of exon 2 containing Surf4 transcripts in Surf4^fl/fl Alb-Cre⁺ mice reduced to ~5% of controls (Figure 1C). This residual unexcised Surf4 mRNA is likely derived from nonhepatocyte cell types in the liver. Quantitative PCR of liver cDNA using primers outside of exon 2 demonstrated a 38% reduction in total Surf4 mRNA transcript in Surf4^fl/fl Alb-Cre⁺ mice compared to Surf4^fl/fl Alb-Cre^- littermates, likely due to nonsense mediated mRNA decay (Popp and Maquat, 2013). Analysis of Surf4 mRNA transcripts by RNA sequencing confirmed the expected reduction of reads spanning the exon 1–2 and exon 2–3 junctions in Surf4^fl/fl Alb-Cre⁺ mice compared to controls (150 and 148 vs 2,623 and 2490 reads, respectively, Figure 1D and Figure 1—figure supplement 1). Consistent with the qPCR data and incomplete nonsense mediated mRNA decay, we identified 928±51 reads mapping to the exon1-exon3 junction of the Surf4 mRNA in liver from Surf4^fl/fl Alb-Cre⁺ mice and zero in Surf4^fl/fl Alb-Cre^- samples (Figure 1—figure supplement 1). This residual exon 2 excised mRNA in Surf4^fl/fl Alb-Cre⁺ liver contains a premature stop codon near the start of the SURF4 coding sequence (codon 23 of 270), which is expected to be translated into a nonfunctional, truncated protein (Figure 1A).

We also detected 404±39 reads (none in controls) mapping to the exon1-4 junction of an alternatively spliced Surf4 mRNA in Surf4^fl/fl Alb-Cre⁺ mice (Figure 1D). Exclusion of exon 2 and 3 is predicted to restore the reading frame and result in the production of an internally deleted SURF4, missing ~1/3 of the full length sequence (Figure 1A). Though also likely to be nonfunctional, residual activity and/or a dominant-negative effect of this internally deleted SURF4 cannot be excluded.

We previously demonstrated a key role of SURF4 in the efficient trafficking of PCSK9 heterologously expressed in HEK293T cells (Emmer et al., 2018). To test the dependence of PCSK9 secretion on SURF4 in vivo, we first examined steady state serum PCSK9 levels in Surf4^fl/fl Alb-Cre⁺ and Surf4^fl/fl Alb-Cre^- control mice. Serum PCSK9 levels are reduced by ∼60% in Surf4^fl/fl Alb-Cre⁺ mice compared to Surf4^fl/fl Alb-Cre^- mice (from 46.0±19.0 ng/mL to 17.8±6.42 ng/mL, Figure 2A), although PCSK9 accumulation was not observed in liver lysates (Figure 2B–C). Quantitative RT-PCR revealed that Pcsk9 mRNA levels were also unchanged (Figure 2—figure supplement 2), consistent with a defect in PCSK9 protein secretion rather than gene expression as the cause for decreased plasma PCSK9 levels.

In contrast to the above findings, Wang et al reported no change in plasma PCSK9 levels in Surf4^fl/fl Alb-Cre⁺ mice (Wang et al., 2021a). To address this issue, and to exclude complex adaptation to hepatic SURF4 deletion induced in utero, we acutely inactivated hepatic Surf4 in adult mice using a previously reported Cas9 mouse system (Wang et al., 2021b). Analyses utilizing three different Surf4 targeting sgRNAs demonstrated a reproducible ~40% reduction in plasma PCSK9 levels for all mice receiving Surf4 targeting sgRNA compared to controls (Figure 2D), consistent with our findings in Surf4^fl/fl Alb-Cre⁺ mice.

Since PCSK9 is a negative regulator of LDLR, we next quantified LDLR levels in liver lysates collected from control and Surf4^fl/fl Alb-Cre⁺ mice. As shown in Figure 2E–F, Surf4^fl/fl Alb-Cre⁺ mice exhibit an ~1.5-fold increase in LDLR abundance in liver lysates compared to controls, consistent with the observed ~60% reduction in circulating PCSK9 level.

Humans with heterozygous loss of function mutations in PCSK9 exhibit an ~28–40% reduction in circulating cholesterol (Cohen et al., 2005; Cohen and Hobbs, 2013; Hooper et al., 2007) with a similar reduction observed in Pcsk9^+/- mice (Rashid et al., 2005). Though Surf4^fl/fl Alb-Cre⁺ mice exhibit similar reductions in PCSK9, total serum cholesterol is markedly reduced (from 54.3±15.1 mg/dL in Surf4^fl/fl Alb-Cre^- mice to 9.51±2.6 mg/dL in Surf4^fl/fl Alb-Cre⁺ mice) (Figure 2G). No change in cholesterol was observed in the limited numbers of mice haploinsufficient for Surf4 in the liver (Surf4^fl/+ Alb-Cre⁺). Analysis of fractionated pooled sera demonstrated marked reductions in cholesterol and triglyceride content in Surf4^fl/fl Alb-Cre⁺ mice among all three major classes of lipoproteins – very low density lipoprotein (VLDL), low density lipoprotein (LDL), and high density lipoprotein (HDL) (Figure 2H). This striking hypocholesterolemia phenotype is sustained through at least 1 year of age (Figure 2—figure supplement 2), with no difference in body mass between Surf4^fl/fl Alb-Cre^- and Surf4^fl/fl Alb-Cre⁺ mice (Figure 2—figure supplement 2A and Figure 2—figure supplement 2C).

Consistent with a role of SURF4 in the ER export of APOB-containing lipoproteins (Wang et al., 2021b), the mean serum APOB level in fasted Surf4^fl/fl Alb-Cre⁺ mice was 3.88±2.63 mg/mL, a>98% reduction compared to Surf4^fl/fl Alb-Cre^- mice (300±157 mg/mL, Figure 3A). Western blotting demonstrated a trend towards an accumulation of APOB in the livers of Surf4^fl/fl Alb-Cre⁺ mice compared to littermate controls, predominantly in an endoglycosidase H (endo H) sensitive form (Figure 3B–C), indicative of ER retention (Freeze and Kranz, 2010).

We next examined hepatic triglyceride secretion in Surf4^fl/fl Alb-Cre⁺ and control mice. For this experiment, mice were fasted to remove intestinal absorption of dietary fat and tissue lipid uptake was blocked by administration of a lipoprotein lipase inhibitor, with liver triglyceride output subsequently monitored by sampling of plasma triglycerides over 24 hours. Following fasting and inhibition of triglyceride hydrolysis, blood glucose levels fell equivalently between Surf4^fl/fl Alb-Cre⁺ and Surf4^fl/fl Alb-Cre^- littermates (Figure 3D). Although serum cholesterol and triglyceride levels steadily increased in both groups over time, both levels were consistently and significantly lower in Surf4^fl/fl Alb-Cre⁺ mice compared to littermate controls (Figure 3E–F). Following an initial decrease in the first hour after lipoprotein lipase inhibition, serum APOB levels steadily rose in control mice (Figure 3G). In contrast, serum APOB levels were markedly reduced at baseline in Surf4^fl/fl Alb-Cre⁺ mice, and showed minimal increase after lipoprotein lipase inhibition (Figure 3G).

To assess the potential role of hepatic Surf4 gene expression on dietary lipid absorption, mice were fed a ³H triolein-labelled lipid load following an overnight fast. No significant differences in blood glucose, serum triglycerides, non-esterified fatty acids, or total intestinal uptake of dietary lipids were observed between control and Surf4^fl/fl Alb-Cre⁺ mice (Figure 4A–D). Similarly, no significant differences were observed in tissue lipid uptake or fatty acid oxidation between Surf4^fl/fl Alb-Cre⁺ mice and littermate controls (Figure 4E–F).

Surf4^fl/fl Alb-Cre⁺ mice exhibited mildly enlarged livers (Figure 5A), with a small increase in hepatic fat content and a reduction in lean mass compared to littermate controls (Figure 5B). However, no differences were observed in fasting hepatic cholesterol, triglyceride, phospholipid, or nonesterified fatty acid content (Figure 5C). Hepatic lipid accumulation can lead to steatohepatitis and liver damage (Ipsen et al., 2018). However, at 8–12 week of age, serum albumin, bilirubin, and liver function markers were indistinguishable between Surf4^fl/fl Alb-Cre⁺ and control mice (Figure 5—figure supplement 1 and Figure 5D) and histologic analyses detected no evidence for steatohepatitis or fibrosis (Figure 5E). Finally, deep sequencing of liver mRNA identified only limited gene expression changes in response to Surf4 deletion (Figure 5F and Supplementary file 4). Significant downregulation was observed for several genes involved in fatty acid biosynthesis processes (Figure 5H). The most significantly upregulated gene in Surf4^fl/fl Alb-Cre⁺ mice is Derl3, a component of the ER-associated degradation (ERAD) pathway, which could be induced by protein accumulation in the ER of Surf4^fl/fl Alb-Cre⁺ mice. Genes involved in the unfolded protein response, such as Ire1, Aft6, and Perk (Hetz et al., 2020), are not upgregulated in Surf4^fl/fl Alb-Cre⁺ mice (Figure 5F).

To confirm the profound hypocholesterolemia with few adverse consequences observed in Surf4^fl/fl Alb-Cre⁺ mice, and to further explore SURF4 inhibition as a potential therapeutic approach, we next tested depletion of hepatic SURF4 using liver-targeted siRNA in adult mice. Mice were treated with control or Surf4 targeting siRNA at multiple doses between 0.5 and 4 mg/kg. Mice treated with Surf4 targeting siRNA demonstrated a dose-dependent reduction of liver Surf4 mRNA and protein levels (Figure 6A and Figure 6—figure supplement 1). As expected, plasma PCSK9, cholesterol, triglycerides, APOB, and APOA1 levels were inversely correlated with siRNA dosage, with the highest siRNA dose (4 mg/kg) resulting in cholesterol levels similar to those observed in Surf4^fl/fl Alb-Cre⁺ mice (Figure 6B–D and Figure 6—figure supplement 1). Finally, no differences in plasma ALT and AST levels were observed between control and siRNA treated mice, suggesting that siRNA treatment and Surf4 depletion does not lead to liver injury, even at the highest siRNA dose (Figure 6E–F).

All animal care and use complied with the Principles of Laboratory and Animal Care established by the National Society for Medical Research. Mice were housed in a controlled lighting (12 hr light/dark cycle) and temperature (22 °C) environment and had free access to food (5L0D, LabDiet, St. Louis, MO) and water. All animal protocols in this study have been approved by the Institutional Animal Care and Use Committee (IACUC) of the University of Michigan (protocol number PRO00009304) and the IACUC of Peking University. Both male and female mice were used in this study unless otherwise specified.

The generation of mice carrying a conditional Surf4 allele in which exon 2 of the gene is flanked by 2 loxP sites (Surf4^fl) has been previously described (Wang et al., 2021b). Surf4^fl/+ mice were crossed with mice carrying an Alb-Cre transgene (Adams et al., 2014) to obtain Surf4^fl/+ Alb-Cre⁺ mice. These mice were then crossed to Surf4^fl/fl mice to generate Surf4^fl/fl Alb-Cre⁺ mice. The Surf4^fl and Alb-Cre alleles were maintained on the C57BL/6 J background by continuous backcrosses to C57BL/6 J mice (0006640, Jackson Laboratory, Bar Harbor ME).

Tail clips were obtained from 2 weeks old mice for genomic DNA isolation and genotyping. PCR was performed using Go-Taq Green Master Mix (Promega, Madison, WI) and resulting products were resolved by 3% agarose gel electrophoresis. All primers used for genotyping are listed in (Supplementary file 1). Those used for genotyping the Surf4 locus are also depicted in Figure 1A. For the Alb-Cre transgene, parental mice were genotyped using promoter-specific Cre primers and offspring were genotyped with primers that detect the Cre transgene (Supplementary file 1).

Mice were fasted overnight for up to 16 hr prior to sample collection. For non-terminal experiments, blood was collected from the superficial temporal vein using a 4 mm sterile lancet. For terminal experiments, mice were first euthanized by isoflurane inhalation and blood was drawn from the inferior vena cava using a 23 G needle and a 1 ml syringe. Blood was collected into a serum separator tube (365967, BD, Franklin Lakes NJ), allowed to clot at room temperature for at least 10 min, and centrifuged at 15,000 g for 10 min to separate serum. Sera were aliquoted and stored at –80 °C. Liver tissue were collected as previously described (Emmer et al., 2020).

Sera were analyzed by a colorimetric assay for total cholesterol (SB-1010–225, Fisher Scientific, Hampton NH) and by ELISAs for PCSK9 (MPC900, R&D Systems, Minneapolis MN) and APOB (ab230932, abcam, Cambridge UK). Serum lipoprotein fractionation assays were performed at the University of Cincinnati Mouse Metabolic Phenotyping Center. Sera were pooled from 5 mice for each genotype and fractionated by fast liquid protein chromatography (FPLC) into 50 fractions. Cholesterol (NC9343696, Fisher, Hampton NH) and triglyceride (TR213, Randox Laboratories, Crumlin UK) content in each fraction were determined using a microliter plate enzyme-based assay. Liver function tests were performed at the University of Michigan In-Vivo Animal Core (IVAC) with sera collected from individual mice using a Liasys analyzer (AMS Alliance).

Hepatic triglyceride secretion assays were performed at the University of Michigan Mouse Metabolic Phenotyping Center as previously described (Millar et al., 2005). Blood levels of glucose were measured using a glucometer (Acucheck, Roche, Basel Switzerland) and plasma levels of cholesterol (SB-1010–225, Fisher Scientific, Hampton NH) and triglycerides (10010303, Cayman Chemical, Ann Arbor MI) were determined using colorimetric assay kits. Plasma APOB levels were determined by ELISA (ab230932, abcam, Cambridge UK).

Oral fat tolerance test and lipid flux assay were performed at the University of Michigan Mouse Metabolic Phenotyping Center. Mice were fasted overnight and ³H triolein-labeled olive oil (0.026 µCi/µl) was given via oral gavage at 5 µl/g of body mass. Blood samples were collected at time 0, 30, 60, 120, and 240 min after the gavage via tail vein bleeding. Plasma levels of triglyceride (10010303, Cayman Chemical, Ann Arbor MI) and non-esterified fatty acid (NEFA-HR (H2), Wako Pure Chemical Industries, Ltd, Richmond VA) were determined using the colorimetric assays. Plasma radioactivity, reported as ³H disintegration per minute (dpm), were determined from 2 µl of serum at each time point.

Tissues samples (liver, heart, gastrocnemius muscle, and perigonadal fat) were collected at the 240 min time point, flash frozen in liquid nitrogen and stored at –80 °C. For the liver, tissue composition was measured using an NMR-based analyzer (EchoMRI) immediately upon harvest and prior to freezing. ³H-triolein flux was estimated as previously described (Kusminski et al., 2012; Ye et al., 2014).

Liver lipid extraction and quantification were performed at the University of Cincinnati Mouse Metabolic Phenotyping Center. Lipids were extracted using the Folch’s extraction method as previously described (Folch et al., 1957). Levels of cholesterol, triglycerides, free fatty acids, and phospholipids in each sample were quantified using specific colorimetric assays.

Lysates were prepared from snap frozen liver tissues and resolved on a 4–20% Tris-glycine gel as previously described (Emmer et al., 2020). Immunoblots were probed with antibodies against APOB (70 R-15771, 1:1000, Fitzgerald Industries International, Acton MA), PCSK9 (ab31762, 1:1000, abcam, Cambridge UK), LDLR (ab52818, 1:1000, abcam, Cambridge UK), HSP90 (4877, 1:1000, Cell Signaling Technology, Danvers MA), SURF4 (Wang et al., 2021b), APOA1 (70 R-15769, 1:1000, Fitzgerald Industries International, Acton MA), Albumin (66051, 1:1000, Proteintech, Rosemont, IL), Tubulin (10094–1-AP, 1:1000, Proteintech, Rosemont, IL), and GAPDH (ab181602, 1:5000, abcam, Cambridge UK). For endoglycosidase H assays, 30 µg of lysate was analyzed as previously described using the above antibodies (Emmer et al., 2018).

Tissue processing, embedding, sectioning, hematoxylin and eosin (H&E), and picrosirius red staining were performed at the University of Michigan In-Vivo Animal Core (IVAC). Slides were reviewed by an investigator blinded to the genotype.

Liver RNA was isolated from tissue using an RNeasy Plus Mini Kit according to the manufacturer’s instructions (74134, Qiagen, Hilden Germany) and reverse transcription was performed using oligo(dT)_12-18 primers (18418012, Invitrogen, Waltham MA). Quantitative PCR reactions were performed using Power SYBR Green PCR Master Mix (4367659, Invitrogen, Waltham MA) and primers listed in Supplementary file 1. Total Surf4 mRNA abundance was calculated using data from primers that bind to exon 5 and 6 on the Surf4 transcript. Abundance of Surf4 mRNA that contains exon 2 was obtained using primers specific to exon 2 and exon 3 of the transcript. Normalized transcript abundance was calculated by the 2^-ΔΔCt method using Gapdh and Rpl37 as housekeeping controls.

Library preparation and next generation sequencing was performed at the University of Michigan Advanced Genomics Core. Demultiplexed fastq files were aligned against the mouse reference genome (GRCm38.92) using STAR (Dobin et al., 2013) and quantified with RSEM (Li and Dewey, 2011). Differential expression analysis was performed by DESeq2 (Love et al., 2014; Supplementary file 4). Sequencing coverage for the Surf4 transcript was analyzed using the ggsashimi package (Garrido-Martín et al., 2018). Raw and processed sequencing data have been deposited to GEO (accession number GSE214393).

Hepatic Surf4 was selectively inactivated in adult mice by injection of adeno-associated virus (AAV) delivering a hepatocyte-specific Cre and a guide RNA targeting Surf4 or LacZ (control) (Supplementary file 2) into a Cre-dependent spCas9 knockin mice (Platt et al., 2014) as previously described (Wang et al., 2021b). Blood samples were collected as previously described (Wang et al., 2021b) and plasma PCSK9 concentrations were measured by the commercial kit (CY-8078 of MBL) according to the manufacturer’s protocol.

To deplete Surf4 mRNA, N-acetylgalactosamine (GalNAc) conjugated siRNA oligos are synthesized to ensure liver targeting. SiRNA targeting murine Surf4 or GalNAc conjugated siRNA with scrambled sequence (siCTL) were injected into 6 weeks old male mice subcutaneously with concentrations indicated in the figures. Seven days after injection, blood samples were collected by tail vein, and then centrifuged at 6000 rpm, 4℃ for 10 min to harvest plasma. The plasma PCSK9 concentrations were measured by the commercial kit (CY-8078 of MBL) according to the manufacturer’s protocol. SiRNA sequences are listed in Supplementary file 3.