Hypomorphic sialidase expression decreases serum cholesterol by downregulation of VLDL production in mice.

Lipoprotein metabolism is an important contributing factor in the development and progression of atherosclerosis. Plasma lipoproteins and their receptors are heavily glycosylated and sialylated, and levels of sialic acids modulate their biological functions. Sialylation is controlled by the activities of sialyltranferases and sialidases. To address the impact of sialidase (neu1) activity on lipoprotein metabolism, we have generated a mouse model with a hypomorphic neu1 allele (B6.SM) that displays reduced sialidase expression and sialidase activity. The objectives of this study are to determine the impact of sialidase on the rate of hepatic lipoprotein secretion and lipoprotein uptake. Our results indicate that hepatic levels of cholesterol and triglycerides are significantly higher in B6.SM mice compared with C57Bl/6 mice; however, VLDL-triglyceride production rate is lower. In addition, B6.SM mice show significantly lower levels of hepatic microsomal triglyceride transfer protein (MTP) and active sterol-regulatory element binding protein (SREBP)-2 but higher levels of diglyceride acyltransferase (DGAT)2; these are all indicative of increased hepatic lipid storage. Rescue of sialidase activity in hypomorphic sialidase mice using helper-dependent adenovirus resulted in increased VLDL production and an increase in MTP levels. Furthermore, hypomorphic sialidase expression results in stabilization of hepatic LDL receptor (LDLR) protein expression, which enhances LDL uptake. These findings provide novel evidence for a central role of sialidase in the cross talk between the uptake and production of lipoproteins.

A genomic fragment containing the mouse lysosomal sialidase gene and promoter was isolated from a BAC (NCBI accession number AF109906) containing 180 kb spanning the sialidase locus (Genome Systems) and was subcloned into a pBSK plasmid ( 60 ). pBSKS+msialR plasmid containing the mouse neu1 gene was digested with NotI and RsrII, yielding a 10.6 kb fragment. The helper-dependent plasmid pC4HSU was similarly digested with NotI and RsrII, producing a 19.3 kb fragment containing the adenoviral components necessary for viral encapsulation, the left and right inverted terminal repeats (ITR) and the packaging signal ( ⌿ ). The 10.6 kb mouse sialidase fragment and 19.3 kb adenoviral vector fragment were both purifi ed with the Geneclean II with Spin kit (Q-BIOgene). The 10.6 kb mouse sialidase fragment was then ligated by compatible cohesive ends to the 19.3 kb adenoviral vector fragment, producing a 29.9 kb helper-dependent vector containing the mouse sialidase gene, pC4HSUmsial, verifi ed by restriction digestion ( Fig. 4A ).

Propagation of HD-Ad containing mouse sialidase gene
The helper-dependent plasmid pC4HSUmsial was digested with PmeI to release the bacterial amplifi cation elements, ampicillin resistance, and the origin of replication, producing a 27.2 kb fragment and exposing the two ITRs on the ends of the fragment. Following heat inactivation, the digested plasmid was transfected into 293Cre4 cells by calcium phosphate transfection. After adsorption and addition of maintenance medium, complete cytopathic effect (CPE) was seen 48 h postinfection. The cells were then scraped and stored. This lysate was used to coinfect a 90% confl uent dish of 293Cre4 cells along with the helper virus at an MOI of 1 PFU/cell. After complete CPE was seen, the cell lysate was harvested and stored. Coinfections continued serially up until eight lysates had been harvested. After each lysate had been collected, pronase/SDS was added to the dishes to digest the remaining adenovirus to confi rm correct adenoviral propagation. Phenol extraction and ethanol precipitation purifi ed the adenoviral DNA, and restriction digestion confi rmed proper adenoviral sequence. Sequencing of the purifi ed adenovirus DNA for the mouse sialidase promoter also confi rmed the presence of the mouse sialidase gene within the viral genome. The helper-dependent vector AdC4HSULacZ was provided by that sialidase can affect LDL metabolism in vitro and that serum sialylation levels can be used as an indicator of cardiovascular disease risk (38)(39)(40)(41)(42). Additionally, low-density lipoprotein receptor (LDLR) and apolipoprotein (apo)B, CII, CIII, and E are all heavily sialylated, and several studies postulate that sialic acids can have functional significance on these proteins (43)(44)(45)(46)(47)(48)(49)(50)(51)(52). Despite their apparent roles in lipid metabolism and atherosclerotic disease progression, the functional effects of sialic acids on these glycoconjugates have not been signifi cantly addressed. To assess the effects of sialidase on lipoprotein and cholesterol metabolism in vivo, we sought to generate and analyze mice with hypomorphic sialidase expression. A partial defi ciency of sialidase was identifi ed in the SM/J mouse in the 1970s ( 53,54 ), and these animals have abnormal sialylation of glycoproteins and an impaired immune response ( 55 ). Campbell and colleagues have demonstrated that the SM/J liver expresses low sialidase mRNA ( 28 ), and we have recently identifi ed a point mutation (-519G>A) in the mouse lysosomal sialidase promoter (neu1), which results in reduced sialidase gene transcription in the SM/J mice. This promoter mutation creates a binding site for a transcriptional repressor, Nkx3.2, resulting in reduced gene expression ( 56 ). SM/J mice, however, harbor mutations in a number of other genes ( 57,58 ), complicating analysis of the physiological consequences of low sialidase. We have therefore isolated this mutation from the SM/J mice by backcrossing onto a C57Bl/6 genetic background, generating a hypomorphic sialidase mouse, named B6.SM, which has reduced sialidase protein levels and activity. In this study, we focus on the effect of hypomorphic sialidase (neu1) expression on the regulation of lipoprotein metabolism in the liver. We demonstrate that hypomorphic sialidase expression lowers serum cholesterol levels by modulating hepatic VLDL production as well as hepatic lipid metabolism. This study points to sialidase as an important player in lipoprotein metabolism and as a potential therapeutic target for metabolic syndrome diseases, such as hypercholesterolemia and atherosclerosis.

Mice
B6.SM mice were obtained by crossing SM/J mice with C57Bl/6 mice six times. The presence of the regulatory mutation, (-519G → A) within the neu1 promoter was confi rmed by PCR using DNA extracted from tail biopsies. The following primers were used for the PCR: 5 ′ ATC CCT GTC CAG GAA CTG GT 3 ′ and 5 ′ CTT AAG GGC ATT GGG GTC AT 3 ′ , synthesized by Mobix facility at McMaster University. PCR (40 cycles) was performed with denaturing temperature at 94°C for 2 min, annealing temperature at 60°C for 30 s, and elongation temperature at 72°C for 30 s. PCR products were digested with MspA1I (New England Bio-Lab), which serves as a genetic diagnostic as it only cleaves the PCR product carrying the B6.SM mutation. Mice were housed in microisolator cages in a room with a 12 h light and dark cycle and given unlimited access to food and water. All experimental protocols were approved by our Institutional Animal Research Ethics Committee. was as follows: 10 min 95°C, 40× 15 sec 95°C, 60 sec 60°C ). Primers (synthesized by MOBIX facility, McMaster University) for LDLR qRT-PCR were: 5 ′ TGACTCAGACGAACAAGGCTG 3 ′ and 5 ′ ATCTAGGCAATCTCGGTCTCC 3 ′ and for SREBP2 qRT-PCR were: 5 ′ GCAGCAACGGGACCATTCT 3 ′ and 5 ′ CCCCATGAC-TAAGTCCTTCAACT 3 ′ .

Immunoprecipitation and lectin pull-downs
Liver membrane-enriched lysates with equal amounts of protein were immunoprecipitated using LDLR antibody and Protein A prior to immunoblot analysis, as described before ( 62 ). We utilized SNA (Sambucus nigra agglutinin), which binds ␣ 2,6 linkages of sialic acid and MALII (Maackia amurensis leucoagglutinin), which binds ␣ 2,3 linkages of sialic acid ( 63 ). The biotin-labeled lectins (Vector Laboratories) were incubated with membrane-enriched lysates and pulled down with streptavidin beads. The enriched samples, containing glycoproteins pulled down by their sialic acids, were used for immunoblot analysis.

TrueBlot immunoprecipitation of PCSK9
Serum was incubated with anti mouse PCSK9 antibody (courtesy of Dr. Nabil Seidah, University of Montreal, Montreal, Canada ) and TrueBlot agarose anti-rabbit beads (eBioscience) overnight at 4°C. Samples were spun and washed with RIPA buffer with protease inhibitors (EDTA-free). The resulting pellet was boiled in Laemlli Sample Buffer, spun down, and then 30 l of the supernatant was subjected to SDS-PAGE and blotted with the same antibody. A special secondary antibody (TrueBlot anti-Rabbit IgG HRP, eBioscience), which only detects full-length immunoglobulin, was used on the Western blot to avoid nonspecifi c bands.

Lipid analyses
For hepatic lipid analyses, 150 mg of liver was homogenized in 1 ml of TNES [10 mM Tris (pH 7.5), 400 mM NaCl, 100 mM EDTA, 0.6% SDS]. Folch mixture (chloroform/methanol, 2:1; 3 ml) was added to 300 l of liver homogenates, and the tubes were mixed for 1 min. After that, 0.6 ml of distilled water was added to the tubes, and the tubes were mixed for 1 min. The extraction mixture was left at 4°C for 2 h. After 2 h, the tubes were centrifuged at low speed to facilitate phase separation. The lower phase (chloroform phase) was dried completely by sitting in a water bath at 37°C. The dry chloroform phase was resuspended in 60 l of isopropanol. Hepatic total cholesterol was analyzed with enzymatic Chang-Xin Shi (McMaster University, Hamilton, ON) and was serially passaged as above.

Purifi cation and concentration determination of adenovirus
After CPE was reached postinfection, cells were scraped into 10 mM Tris-HCl, pH 8.0. Sodium deoxycholate (5%) was added to lyse the cells, followed by the addition of 2M MgCl 2 and DNAase I to digest any unpackaged viral DNA as well as cellular DNA. The lysate was then spun, and the supernatant was collected and ultracentrifugated twice through CsCl density gradient ( 61 ). The lower viral bands were collected with an 18 gauge needle and syringe through the side of the tube. The collected adenovirus was then injected into a Slide-A-Lyzer dialysis cassette (Pierce) where the virus was dialyzed against three changes of 500 ml 10 mM Tris-HCl (pH 8.0) over 24 h. The adenovirus was collected from the dialysis cassette, and sterile glycerol was added to a fi nal concentration of 10%. The concentration of helperdependent adenovirus was determined through fl uorometric analysis using Hoechst dye (Boehringer Mannheim). CsClbanded adenovirus (20 l) was treated with 20 l of pronase/ SDS overnight at 37°C to degrade the viral capsid. The following day, 20 l of the pronase/SDS-treated virus was exposed to the Hoechst dye, and fl uorometric analysis was measured using the Hoefer Fluorometer (Hoefer). Adenoviral particle count was based on the fl uorometric result ( g/ml) of inserting this value into the following: Viral DNA Concentration ( g/ml) × 9.48 × 10 11 / Length of Viral DNA (Kb).

Adenoviral administration in vivo
Helper-dependent adenoviruses containing mouse sialidase gene or lacZ cDNA (100 l; 10 9 particles/mouse in sterile PBS) were injected into the tail vein of 5-month-old male B6.SM mice under isofl uorane anesthetic. Mice were monitored for the incubation period of 14 days until VLDL-production experiments.

Collection of blood and tissues
Mice were anesthetized with ketamine/xylazine and euthanized by exposure of their thoracic cavity. Blood was obtained by cardiac puncture. Serum was obtained by centrifugation of blood for 5 min at 15,000 rpm using serum collection tubes (Sarstedt). Mice were perfused with PBS through the left ventricle of the heart. The liver was harvested, frozen in liquid nitrogen, and stored at Ϫ 80°C for further use in protein and RNA studies.

Sialidase activity assay
Approximately 0.15 g of tissue was minced on ice and homogenized in 1.5 ml water. Tissue homogenate (50 l) was then incubated for 1 h at 37°C with 60 l of 0.4 mM 4-Mu-NANA in acetate buffer (pH 4.2) with 10% BSA. Assay was performed similarly for WG544 cell lysates infected with the HD-Ad sialidase. Activity is measured as the amount of fl uorescence generated from the liberation of umbelliferone (4-Mu) from the NANA substrate. The reaction was stopped by the addition of 2 ml of basic 0.1 M MAP buffer. Fluorscence was then measured using a plate fl uoremeter (PerkinElmer) and normalized to protein concentration.

RNA isolation and quantitative real-time PCR
Livers were homogenized in RNA lysis buffer, and then RNA was isolated using Norgen Total RNA Isolation Kit. Total RNA (1-5 g) was then reverse transcribed using oligoDT primers following the protocol of Invitrogen's SSIII RT reverse transcriptase. cDNA was then used for qRT-PCR using Applied Biosystems Power Sybr Green. Plates were loaded with a 20 l reaction per well and included appropriate blanks and standard. (PCR cycle software. Quantifi cation was performed by measuring the red stained area and dividing by nuclear area per cell and averaging each group.

Immunofl uorescence
Cells were grown in Optimem for 72 h then fi xed with 3.7% formaldehyde and permeabilized with 0.5% Triton-X 100 in PBS. Cells were blocked in 20% goat serum in PBS, incubated overnight with anti-LDLR antibody (R and D) at 1:400, washed, incubated with Texas Red anti-goat secondary at 1:400, stained with DAPI, and then mounted using ProLong Gold Antifade reagent from Invitrogen. Images were taken and analyzed with AxioVision software from Zeiss.

Fibroblast cholesterol assay
Fibroblasts were grown in Optimem for 72 h then incubated with 50 g/ml LDL for 24 h. Briefl y, cells were washed and lipids were extracted using a hexane/isopropanol mixture overlaid on the cells for 1 h. The mixture was allowed to evaporate and lipids were resuspended in 100 l isopropanol for enzymatic measurements of cholesterol as described above. After lipid extraction, cells were scraped in 0.1% SDS, and protein content was measured using a Lowry assay to normalize cholesterol readings.

Statistical analysis
Statistical analyses between multiple groups of data were analyzed by one-way ANOVA followed by Tukey comparison test and multiple comparison test using Prism 5 (version 5.04). Statistical analyses between two groups were performed using an unpaired Student's t -test. Error bars represent SEM unless otherwise noted. Data were considered statistically different if P < 0.05.

Sialidase protein expression and activity in B6.SM mice
Western blot analysis of hepatic neu1 sialidase reveals a signifi cant reduction (approximately 85%) in sialidase expression in B6.SM compared with C57Bl/6 mice ( Fig. 1A ). Sialidase activity is signifi cantly reduced in the brain, liver, spleen, and kidney of B6.SM males compared with C57Bl/6 controls ( Fig. 1B ). The sialidase enzymatic assay of organ assay (Infi nity Cholesterol Liquid Stable Reagent, Thermo Scientifi c). The enzymatic colorimetric assay product was measured at 500 nm. Free cholesterol was analyzed with Free Cholesterol E Reagent (Wako Diagnostics). The absorbance of the reaction product was measured at 600 nm. Cholesteryl ester concentration was calculated by subtracting free cholesterol measurements from total cholesterol concentration. Triglyceride was analyzed with enzymatic colorimetric assay (L-Type Triglyceride H, Wako Diagnostics). The absorbance of the reaction product was measured at 600 nm. Serum samples were measured directly as above. For lipoprotein cholesterol analyses, 300 l of serum was fractionated by gel fi ltration-FPLC using a Superose 6 column ( 64 ), and lipid levels were measured with enzymatic assay as above.

In vivo hepatic VLDL-lipid secretion
Hepatic production of VLDL-triglyceride, cholesterol, free cholesterol, and cholesteryl esters were measured in 3-month-old male C57Bl/6 and B6.SM after intravenous injection of Triton WR 1339 (Tyloxapol T0307-10G, Sigma BioXtra, Sigma-Aldrich) (15 g/dl in 0.9% NaCl). Mice were fasted overnight prior to the experiments, and 500 mg/kg mg of Triton WR 1339 was injected. Blood samples were taken from the cheek under light anesthesia before and at 1, 2, 3, and 4 h after Triton injection for triglyceride, cholesterol, free cholesterol, and cholesteryl ester measurements. VLDL-triglyceride, cholesterol, free cholesterol, and cholesteryl ester production rates were obtained by calculating the slope of the regression line of the graph with VLDL-triglyceride, cholesterol, free cholesterol, and cholesteryl ester concentration, respectively, versus time in hours.

Fibroblast Oil Red O staining
For staining of neutral lipids in cells, Oil Red O powder from Sigma-Aldrich (O0625) was prepared by dissolving 2.5 g in 500 ml of isopropanol. Prior to experimental use, this mixture was diluted 3:2 with isopropanol and fi ltered to remove any particulate matter. Cells were grown on uncoated glass coverslips in a 24-well plate and incubated with 50 g/ml LDL (BTI Inc.) for 24 h after growing in Optimem for 72 h. Cells were washed, fi xed with 3.7% formaldehyde, and then washed with 60% isopropanol prior to Oil Red O staining for 1 h. This was followed by another 60% isopropanol rinse and four PBS washes. Hematoxylin staining was then performed, and the coverslips were mounted on microscope slides using Aqua Mount from Fisher. Pictures were taken using brightfi eld and phase microscopy with Zeiss Axiovision Representative blots of n = 3 for each group. Liver lysates were subjected to SDS-PAGE (8%). Membranes were probed with anti-neu1 Sialidase antibody and anti-␤ -actin as a control. Intensities of bands were measured by ImageJ densitometry software. (B) B6.SM tissues have signifi cantly lower levels of sialidase activity, and this is especially prominent in the liver where levels are reduced to approximately 20% of C57Bl/6. Brain, liver, spleen, and kidney lysates were assessed for sialidase activity using fl uorescent 4-Mu-NANA. * P = 0.01, ** P = 0.001, *** P < 0.0001. mU = Mol/hr. ( Fig. 3E ), which is required for triglyceride loading onto apoB and subsequently VLDL synthesis (67)(68)(69). As a result, the decrease in VLDL-triglyceride appears to be a direct result of the decreased protein expression of MTP. These data show that B6.SM mice exhibit a drastically altered lipoprotein metabolism initiated by reductions in VLDL assembly and production.

Hepatic VLDL production and MTP expression after sialidase gene therapy in B6.SM mice
The mouse lysosomal sialidase gene (10.6 kb) was ligated via compatible cohesive ends into the 19.3 kb adenoviral vector pC4HSU that contained the two ITRs and the packaging signal ( ⌿ ). The resultant plasmid, pC4HSUmsial, was confi rmed by digesting with EcoRI , yielding the expected fragment sizes ( Fig. 4A ). After the helper-dependent mouse sialidase adenovirus was properly characterized and CsCl purifi ed (see Methods), we analyzed its effects on sialidase activity. Sialidosis fi broblast cells were infected with increasing doses of viral particles, which caused a steady, almost linear increase in sialidase activity with increasing virus concentration ( Fig. 4B ). These results demonstrate that this helper-dependent mouse sialidase (HD-AdSial) adenovirus is a functional vector capable of producing active mouse neu1 sialidase in vitro. Because B6.SM male mice have lower VLDL-TG production compared with C57Bl/6 controls ( Fig. 3 ), we sought to determine whether adenoviral sialidase gene therapy would rescue the phenotype in B6.SM animals. Thus, we infected B6.SM sialidase-defi cient male mice with helper-dependent mouse sialidase (HD-AdSial) or LacZ (HD-AdlacZ) adenovirus, and then measured VLDL-TG production. The HD-AdSial group had signifi cantly higher expression of neu1 sialidase protein in the liver compared with the HD-AdlacZ group, confi rming expression of the virus ( Fig. 4C ). B6.SM mice infected with sialidase virus had significantly higher VLDL-TG production compared with LacZ controls ( Fig. 4D ). These data directly demonstrate that extracts from B6.SM mice indicated that this strain recapitulates a tissue-specifi c hypomorphic sialidase mouse model.

Effect on serum cholesterol levels and hepatic lipid levels
Measurement of fasted serum levels of total cholesterol and triglycerides revealed no signifi cant difference between C57Bl/6 and B6.SM mice ( Table 1 ). However, when serum lipoproteins from unfasted male mice were fractionated by size exclusion using fast-protein liquid chromatography (FPLC) with a superose 6 gel fi ltration column ( Fig. 2 ), the cholesterol profi les of B6.SM mice show signifi cantly lower cholesterol levels in LDL fractions (fractions 20-30) than those of the corresponding C57Bl/6 controls. Also, the HDL peak shows a slight shift indicative of smaller-sized HDL particles in B6.SM mice ( 64-66 ) ( Fig. 2 ). Additionally, the VLDL peak appears to be slightly smaller in B6.SM mice ( Fig. 2 ) and to shift to the right, indicative of smaller-sized VLDL particles ( Fig. 2 , inset). These results indicate that hypomorphic sialidase expression appears to lower cholesterol levels of LDL-sized particles. To determine whether the altered cholesterol profi le is associated with altered hepatic lipid metabolism, we measured the hepatic total cholesterol, triglyceride, free cholesterol, and cholesteryl ester content in both B6.SM and C57Bl/6 mice. We observed a signifi cant increase in hepatic total cholesterol, cholesteryl esters, and triglyceride in B6.SM mice compared with C57Bl/6 mice ( Table 2 ). A trend of increase in hepatic free cholesterol of B6.SM mice compared with C57Bl/6 mice is also noted. These results point to sialidase activity as having a role in modulation of lipid metabolism and homeostasis in the liver.

In vivo effect on hepatic VLDL-lipid production and MTP expression
To determine whether the changes in serum and hepatic cholesterol levels are caused by a decrease in VLDLlipid production, we have measured the lipid concentrations at several intervals post lipoprotein lipase inhibition. Administration of Triton WR1339 prevents the hydrolysis of triglyceride and the uptake of VLDL and, as a result, allows for the assessment of hepatic production rates of VLDL.
Our results indicate that B6.SM mice had signifi cantly lower production rates of VLDL-triglyceride (VLDL-TG) over a 4 h period ( Fig. 3A and Table 3 ). These results show that hypomorphic sialidase expression decreases VLDL-TG production, which coincided with lower serum cholesterol levels in B6.SM mice. This decrease in VLDL-TG production is accompanied by lower microsomal triglyceride transfer protein (MTP) protein expression in the liver  Serum lipoproteins were fractionated by size using FPLC using a superose 6 gel fi ltration column. The FPLC cholesterol profi les of the unfasted male B6.SM animals (n = 3) have signifi cantly less total cholesterol in LDL-sized particles than those of the corresponding C57Bl/6 controls (n = 3). In addition, there is a tendency of the HDL and VLDL (inset) peaks to be shifted slightly to the right. Each circle represents mean ± SE (* P < 0.05).
(SREBP)-2 expression in B6.SM mice ( Fig. 5A ). It has been shown that the MTP promoter contains SREBP-2 response elements ( 70,71 ), suggesting that reduced SREBP-2 may contribute to the reduced MTP levels. Nevertheless, higher levels of cholesterol in the livers of B6.SM mice reduce active SREBP-2 and affect downstream gene expression. Additionally, protein levels of cleaved SREBP-1a/c remain unchanged between the two strains ( Fig. 5B ). It appears that the phenotype observed in B6.SM livers is primarily due to SREBP-2 and independent of SREBP-1. Furthermore, we assessed the expression of hepatic acylCoA:cholesterol acyltransferase (ACAT-2), which mediates esterifi cation of hepatic cholesterol ( 72 ), and we found a trend of higher ACAT-2 expression, implying that there is suffi cient excess of cholesterol to be esterifi ed ( Fig. 5C ). Due to the increases in triglyceride levels observed in the livers of B6.SM mice, it was important to analyze the diglyceride acyltransferase rescuing sialidase defi ciency via adenovirus increases VLDL-TG production, similar to what was observed in C57Bl/6 mice. The HD-AdSial group also had an increase in hepatic MTP protein ( Fig. 4E ) compared with the LacZ controls, indicating that sialidase can affect MTP levels, although the mechanism is yet to be determined. These fi ndings enable us to conclude that low levels of neu1 sialidase in B6.SM mice are directly driving reduced VLDL-TG production.

Hypomorphic sialidase expression decreases hepatic SREBP-2 and increases hepatic DGAT2 expression
To investigate the mechanisms behind the changes in hepatic lipid levels and lipoprotein metabolism, we analyzed protein expression of several important enzymes and transcription factors. We found a signifi cant decrease in cleaved hepatic sterol-regulatory element binding protein Lipids from livers of C57Bl/6 (n = 3) and B6.SM mice (n = 3) were isolated using Folch extraction. There is a signifi cant increase in hepatic total cholesterol, cholesteryl esters, and triglyceride levels in B6.SM male mice compared with C57Bl/6 ( P = 0.01, P = 0.02, and P = .048, respectively). In addition, there is a trend of increase in hepatic free cholesterol levels in B6.SM mice compared with C57Bl/6 ( P = 0.19). Mean ± SE are shown. a P < 0.05. by reduced sialidase expression, with potential functional consequences. Thus, hypomorphic sialidase expression leads to hypersialylation of LDLR, which in turn could affect the traffi cking, degradation, or turnover of the receptor.

Sialidase-null mutation increases LDL uptake in human fi broblasts
To address the direct functional effect of sialidase on LDLR, we utilized human fi broblasts that have null neu1 sialidase activity. The aim of this experiment was to complement the in vivo data with human cells in vitro. We observed no differences in LDLR protein expression between normal and sialidase-null cells, but sialidase-null cells had slightly higher molecular weight LDLR protein, potentially due to hypersialylation ( Fig. 7A ). Furthermore, LDLR immunofl uorescence has shown that the receptors appear to cluster next to the nucleus to a greater extent in the sialidosis cell line ( Fig. 7B , arrows), although there were no gross changes in expression. To measure the function of the LDLR, we analyzed LDL uptake and lipid droplet formation via Oil Red O staining. Both cell types showed minimal staining during serum starvation but exhibited lipid droplet formation after 24 h of LDL treatment, indicating signifi cant LDL uptake. Neutral lipid accumulation appeared to be slightly higher (although not signifi cant) in sialidase-null versus normal control cells ( Fig. 7C ), as measured by Oil Red O quantifi cation. To assess this uptake more quantitatively, we utilized lipid extraction followed by enzymatic cholesterol level measurements in these cells. Both cell types had signifi cant increases in cholesterol levels when treated with LDL, indicating internalization of the lipoprotein. However, sialidase-null fi broblasts treated with LDL had signifi cantly higher total cholesterol levels than wild-type cells treated with LDL ( Fig. 7D ). These data indicate that the absence of neu1 sialidase activity results in higher cholesterol levels due to increased LDL uptake in sialidase-null fi broblasts, consistent with our fi ndings in livers of B6.SM mice.

DISCUSSION
In this study, we examined lipoprotein metabolism in a unique mouse model expressing hypomorphic levels of sialidase (neu1). Although sialylation of lipoproteins and lipoprotein receptors has been invoked previously as an important determinant in cholesterol metabolism, little has been reported toward dissecting the impact of sialidase (neu1) on lipoprotein production or clearance in vivo.
(DGAT) enzyme, which facilitates the fi nal step in endogenous triglyceride synthesis ( 73 ). We observed an increase in DGAT2 (the major enzyme) in B6.SM mice ( Fig. 5D ), which is indicative of increased triglyceride synthesis and changes in hepatic lipid homeostasis ( 73 ). Thus, increased hepatic retention of lipids, along with downregulated SREBP-2 and upregulated DGAT2 protein levels, appears to alter lipid homeostasis in B6.SM mice.

Modulation of hepatic expression of LDLR
To determine whether the decrease in serum cholesterol levels and the increase in hepatic cholesterol levels in B6.SM mice are caused by altered expression of lipoprotein receptors, we have evaluated the expression of LDLR and LRP-1. Although there is no signifi cant difference in hepatic protein levels of LDLR or LRP-1 in B6.SM male mice compared with C57Bl/6 controls ( Fig. 6A , B ), there is a signifi cant reduction in LDLR mRNA levels as measured by qRT-PCR ( Fig. 6C ). The decrease in SREBP-2 protein expression observed in B6.SM livers is consistent with a decrease in LDLR transcript, as the LDLR promoter contains SREBP-2 response elements. Thus, the maintenance at the protein level caused by hypomorphic sialidase expression could be due to a posttranslational mechanism, such as slower receptor traffi cking/ recycling or decreased degradation. In view of these results, we have measured the serum protein expression of proprotein convertase subtilisin/kexin 9 (PCSK9), a SREBP-2 target gene that affects the turnover of the LDLR ( 74-77 ). Our results indicate that B6.SM mice show a decrease in the serum levels of PCSK9 compared with C57Bl/6 mice ( Fig. 6D ). PCSK9 gets secreted and can bind to LDL receptors at the cell surface and target them for degradation instead of recycling ( 78 ). Therefore, the lower level of PCSK9 expression may result in an increase in the rate of receptor recycling and retention of LDLR protein, despite lower mRNA levels. To determine whether hypomorphic sialidase expression affects the sialylation of LDLR directly, we performed lectin pull-downs followed by Western blotting. Membrane-enriched liver lysates were pulled down with streptavidin beads using biotin-labeled SNA or MAL II, which bind specifi c ␣ -2,6 and ␣ -2,3 linkages of sialic acid, respectively. This was followed by LDLR blotting to assess how much LDLR was pulled down via the sialic acids. We included control samples immunoprecipitated with LDLR and blotted for total LDLR to ensure equal starting amounts. We observed higher levels of LDLR-associated sialic acids in livers of B6.SM mice compared with C57Bl/6 ( Fig. 6E ). This indicates that sialic acid molecules on LDLR are directly affected Male C57Bl/6 (n = 3) and male B6.SM mice (n = 3) were fasted overnight. Serum was collected just before the injection of the lipoprotein lipase inhibitor Triton WR1339 (time 0 h) and at 1, 2, 3, and 4 h post administration. There is a signifi cant decrease in hepatic VLDL-triglyceride, VLDL-cholesterol, VLDL-free cholesterol, and VLDLcholesteryl esters production rate in B6.SM mice compared with C57Bl/6 mice. Mean ± SE are shown. a P < 0.05.
phenotype is a direct result of the hypomorphic neu1 gene, we transduced neu1 sialidase expression using helper-dependent adenovirus in B6.SM livers and were able to show a signifi cant increase in both VLDL-TG production and hepatic MTP protein levels compared with HD-AdlacZ controls.
To dissect the mechanisms of increased hepatic lipid storage and decreased VLDL-production in B6.SM mice, hepatic cholesterol and triglyceride metabolism in the liver were investigated. The SREBP transcription factors are master regulators of hepatic lipid homeostasis ( 79 ). Our observations of higher hepatic cholesterol levels resulting in lower active SREBP-2 and lower VLDL production The B6.SM mouse shows a drastic reduction in sialidase protein and activity, and our study revealed a signifi cant decrease in cholesterol levels in IDL/LDL-sized particles in B6.SM mice with a shift in the size of HDL and VLDL particles as revealed by FPLC cholesterol profi les. We demonstrated that hypomorphic sialidase mice have lower serum cholesterol levels as a result of lower hepatic VLDL production and potentially higher hepatic LDL uptake. This is coupled with an increase in hepatic lipid storage and a decrease in hepatic MTP protein, all of which are suggestive of an atheroprotective phenotype, which appears to be driven by reduced VLDL secretion. To confi rm that the lower VLDL-TG production helper-dependent adenovirus displaying fragment sizes corresponding to an Eco RI digestion. Restriction analysis of pC4HSUmsial using Eco RI reveals the appropriate fragments following digestion (lane 1, pC4HSUmsial; lane 2, 1 kb ladder). This virus was combined into a helper-dependent vector, yielding HD-AdSial (see Methods). (B) Sialidase activity of sialidosis fi broblast cells (WG544) that were infected with HD-AdSial at various particles/cell. Viral infection increases sialidase activity at a particle-dependent rate (n = 3, error bars represent SE). B6.SM male mice (n = 4 for each group) were infected with helper-dependent adenovirus containing mouse sialidase (HD-AdSial) or LacZ (HD-AdlacZ) and monitored for 14 days. (C) Neu1 sialidase protein is upregulated in liver lysates of the HD-AdSial group as measured by Western blotting, indicating expression of the viral vector. VLDL-TG production was analyzed by hourly serum triglyceride measurement following injection of LPL inhibitor Triton WR1339, as described above. (D) B6.SM mice infected with mouse sialidase helper-dependent adenovirus have signifi cantly higher VLDL-TG production compared with B6.SM mice infected with HD-AdlacZ control adenovirus. (E) MTP protein expression is increased in B6.SM mice that were infected with HD-AdSial compared with HD-AdlacZ as measured by Western blotting. * P < 0.05. by sialidase levels. It appears that the phenotype observed in B6.SM livers is primarily due to SREBP-2 and is independent of SREBP-1a/c. In addition, higher levels of hepatic esterifi ed cholesterol as a result of high levels of ACAT-2 expression were observed in hypomorphic sialidase mice. ACAT-2 is an ER-bound enzyme that forms cholesterol esters from cholesterol ( 85 ). ACAT-2 activity decreases the solubility of cholesterol and prevents its incorporation into lipid membranes. As ACAT-2 expression is limited to hepatocytes and enterocytes, ACAT2-derived cholesteryl esters can be packaged directly into VLDL via MTP or stored as neutral lipid droplets in the cytosol. The latter option is more likely in our model as B6.SM mice show lower production levels of VLDL and higher hepatic cholesterol content. We also observed an increase in DGAT2 protein in the livers of B6.SM mice compared with C57Bl/6, which is indicative of increased triglyceride synthesis ( 73 ). These results, along with higher levels of hepatic triglyceride, favor the idea that these animals have increased lipid droplet formation and triglyceride storage, as DGAT does not affect the VLDL production rate ( 86 ). Interestingly, DGAT2 is insensitive to SREBP regulation ( 79 ). Clearly there are other factors that infl uence levels of DGAT2; nevertheless, changes in DGAT2 and SREBP-2 protein levels contribute to alterations in hepatic lipid homeostasis in hypomorphic sialidase mice. Overall, the decrease in VLDL-lipid production rate observed in B6.SM mice is caused by decreased MTP expression, which appears to be caused by increased hepatic retention of lipids and downregulated SREBP-2 levels.
In addition to VLDL production, hepatic cholesterol content is primarily dependent on LDL endocytosis via LDLR or on chylomicron uptake via LRP-1. The stabilized levels of hepatic LDLR protein in the hypomorphic sialidase mice appear to be due to posttranslational events, as the LDLR mRNA level is in fact lower in B6.SM mouse livers.
is consistent with previous studies ( 80 ). SREBP-2 is typically endoplasmic reticulum (ER) -resident and bound to SREBP cleavage activating protein (SCAP), and it can be escorted to the Golgi when sterol levels are low, where site-1 and site-2 proteases can convert SREBP to its active form, which induces the transcription of genes involved in cholesterol synthesis ( 81 ). As intracellular cholesterol levels increase in hepatocytes of hypomorphic sialidase mice, the insulin-induced gene 1 (Insig) can bind to SCAP, and the Insig and SCAP complex will retain SREBP in the ER ( 82 ), limiting the transcriptional activation of cholesterol synthesis genes as well as MTP and LDLR. Thus, the decreased hepatic expression of MTP and lower VLDL production in B6.SM mice could be due to a downstream effect of decreased levels of active SREBP-2. Interestingly, B6.SM mice have no signifi cant difference in hepatic SREBP-2 mRNA compared with C57Bl/6 mice as determined by quantitative RT-PCR. Therefore, our model indicates that the decreased SREBP-2 protein expression is a direct result of the altered cholesterol levels in the liver and not a downstream negative feedback on SREBP-2 gene transcription. In fact, Horton and colleagues have reported that transgenic mice overexpressing SREBP-2, but not SREBP-1, exhibit increased MTP gene transcription ( 71,83 ). This implies that hepatic cholesterol levels could be the driving force for our observed phenotype, primarily through a decrease in active SREBP-2. In contrast, Sato and colleagues have shown that SREBP negatively regulates MTP gene transcription, although their work was performed in human HepG2 cells ( 84 ). Further work is required to consolidate these results with our observations. Furthermore, we have shown that protein levels of cleaved SREBP-1a/c, which is a master transcriptional regulator of fatty acid synthesis ( 80 ), are unchanged between C57Bl/6 and B6.SM male mice, indicating that SREBP-driven regulation of hepatic fatty acid synthesis is unaffected more toward differences in ligand-receptor interactions between sialidase-null and normal fi broblasts, as earlier reports had suggested dependency on sialylation for LDL uptake in vitro ( 38 ). This also provides verifi cation that sialidase can affect LDLR function directly and complements the mechanism delineated in vivo. These human fi broblast data provide strong evidence for the role of sialidase in LDLR function and suggest a role for sialidase in human lipoprotein metabolism.
Overall, we demonstrate that the changes in lipoprotein metabolism observed in hypomorphic sialidase mice are mediated by a decrease in the production of VLDL-TG in the liver, which is driven by reduced MTP expression and also by hepatic retention of cholesterol and triglycerides. Despite having similar levels of LDLR protein in the liver, hypersialylated LDL receptors in the B6.SM mice are more effective in internalizing LDL, and as a result, hypomorphic sialidase mice show lower serum LDL cholesterol. Although the exact mechanism is unknown, sialylation of LDLR appears to affect The lower LDLR mRNA levels are as expected due to lower SREBP-2 protein levels, but the maintenance of the protein is a unique fi nding in the B6.SM mice. The increased sialic acid content of the LDLR protein appears to result in its retention, potentially by increasing its half-life or recycling or decreasing its degradation. We postulate that signifi cant alterations in the sialylation of LDLR (as observed in B6.SM mice) could have functional implications by altering its interaction with LDL, PCSK9, or the recycling machinery. Interestingly, the lower levels of PCSK9 in hypomorphic sialidase mice are likely responsible for maintaining steady-state levels of LDL receptors in spite of lower LDLR transcripts. As a result, the reduced level of LDLR degradation enables its recycling and its higher effi ciency for LDL internalization. The dependency on PCSK9 for LDLR stabilization does not seem to be a requirement for LDL uptake in human fi broblasts, where signifi cantly higher levels of LDL were internalized by sialidase-null compared with normal cells. The latter points (E) LDLR from B6.SM mice has higher levels of terminal ␣ -2,6 linked and ␣ -2,3 linked sialic acids compared with C57Bl/6 mice, due to lower sialidase levels. Sialylated conjugates from membrane-enriched liver lysates (equal protein) were pulled down using biotinylated SNA and MALII lectins followed by streptavidin sepharose. Pulled down glycoproteins were immunoblotted for LDLR. Control LDLR blotting of LDLR IP samples show equal quantities of LDLR in starting lysates.
its recycling or internalization, potentially through PCSK9. Taken together, these events are expected to lead to an atheroprotective effect through the lowering of LDL cholesterol in the serum. These fi ndings provide evidence of a central role for sialidase in cholesterol metabolism and set the stage for examining polymorphisms in the human neu1 gene and potential links to cardiovascular disease. Fig. 7. LDLR expression and function in human fi broblasts with a Neu1 sialidase-null mutation. Cells were serum starved in Optimem media to upregulate LDLR and analyzed by Western blotting, immunofl uorescence, Oil Red O staining, and enzymatic cholesterol assay. (A) Sialidase-null fi broblasts do not show signifi cantly different LDLR protein expression, although their receptors appear to be of a higher molecular weight (representative of four blots). (B) LDLR immunofl uorescence shows localization and clustering of receptors in fi broblasts (arrows), but no gross expression changes between the two cell types (bar = 20 m). (C) Oil Red O staining after LDL incubation shows lipoprotein uptake and lipid droplet formation (arrows) in the fi broblasts, although there was no signifi cant difference in neutral lipid accumulation (bar = 20 m). (D) Sialidase-null fi broblasts show signifi cantly higher total cholesterol levels as measured by enzymatic cholesterol assay after LDL incubation compared with wild-type cells, indicating more LDL uptake ( P = 0.048, n = 3 in each group).