Changes in Matrix Proteoglycans Induced by Insulin and Fatty Acids in Hepatic Cells May Contribute to Dyslipidemia of Insulin Resistance

Linoleic acid, bovine serum albumin (BSA; grade V), chondroitinase ABC, and HEPES were obtained from Sigma (St. Louis, MO). Heparitinase was obtained from ICN (Aurora, OH). NuSieve 3:1 agarose was obtained from FMC BioProducts (Rockland, ME). Disposable PD-10 columns (Sephadex G-25), HiTrap Q columns, and [³⁵S]sulfate were obtained from Amersham Pharmacia Biotech (Uppsala, Sweden). Primers for reverse transcriptase (RT)–polymerase chain reaction (PCR) were obtained from Life Technologies (Gaithersburg, MD). Reagents and fluorescently labeled probes for real-time PCR were purchased from Applied Biosystems (Stockholm). Trypsin and cell culture media were obtained from Biowhittaker (Verviers, Belgium), and fetal bovine serum (FBS) was obtained from Biochrom KG (Berlin, Germany). Human blood serum was obtained from young healthy donors. Scintillation fluid was obtained from Beckman (Fullerton, CA). Human recombinant apoE₃ was purchased from PanVera (Madison, WI). Darglitazone and bezafibrate were obtained from AstraZeneca (Mölndal, Sweden). Insulin was purchased from Novo Nordisk (Copenhagen). All other chemicals were of analytical grade and were obtained from Merck (Darmstadt, Germany).

LDLs (d = 1.019–1.063 g/ml) were isolated from fresh human plasma by using differential ultracentrifugation and were stored as previously described (9). β-VLDLs from rabbits fed 1% cholesterol for 1–5 weeks were isolated and combined with apoE₃, as previously described (5).

The experimental procedures were approved by the local ethics committee on animal experiments (Göteborg region) and were in accordance with Swedish laws on the use and treatment of experimental animals. Adult male obese Zucker fa/fa rats and their lean fa/+ littermates were maintained in a temperature-controlled (20–22°C) room with 12-h light/dark cycles, and they had free access to rodent diet (R3; Lactamin AB, Stockholm) and tap water.

HepG2 cells were bought from American Type Culture Collection (Manassas, VA) and subcultured by standard procedures. The cells were grown in Eagle’s minimum essential medium (EMEM) with 10% (vol/vol) FBS with the addition of glutamine, sodium pyruvate, and antibiotics. For incubations with NEFAs, media were prepared according to a previously described method (10). In brief, 100 μmol/l BSA was added to EMEM with 10% FBS. Free fatty acids were dissolved in a small volume of hexane, precipitated by dropwise addition of 6 mol/l NaOH, evaporated under nitrogen to a thin film, and dissolved in 1 ml warm water. The NEFAs suspended in water were added dropwise to the media while stirring. The pH was adjusted to 7.4 with NaOH, and the media were filtered through a 0.22-μm filter. The NEFA content in the media was determined with a colorimetric kit from Wako (Neuss, Germany). We previously have not been able to find evidence of NEFA free radical−mediated oxidation in the incubation conditions used (7). In the indicated studies, we added 10 or 50 nmol/l insulin to the culture medium. EMEM (Biowhittaker) contained 5.6 mmol/l glucose, which corresponded to a normal fasting glucose level. With dilution caused by the additions, the final glucose concentration was 5.4 mmol/l.

Biolabeling and purification of PGs from tissue culture were performed as previously described (7,11). HepG2 cells were trypsinized and seeded. The cells were then incubated overnight in EMEM with 10% FBS, glutamine (0.292 g/l), sodium pyruvate (0.11 g/l), and, as antibiotics, 100 units/ml penicillin, 0.1 mg/ml streptomycin, and 0.25 μg/ml amphotericin; afterward, fresh medium with the same ingredients was added, but with the addition of 100 μmol/l BSA (control), 100 μmol/l BSA plus 10 nmol/l insulin, or 100 μmol/l BSA plus 300 μmol/l linoleic acid. After a 24-h preincubation in these media, labeled GAGs were obtained by exposing cultured cells to corresponding fresh medium with the addition of 67 μCi [³⁵S]sulfate for 72 h before harvest. At harvest, the cultures were in a subconfluent state.

Proteoglycans were prepared from cultured HepG2 cells and medium and from frozen rat livers. Complete mini protease inhibitors (Roche, Mannheim, Germany) were added to the culture medium; the medium was then dialyzed against binding buffer containing 8 mol/l urea, 2 mmol/l EDTA, 0.5% Triton X-100, and 20 mmol/l Tris-HCl (pH 7.5) during a 48-h period. Cells were dissolved to the same buffer (with protease inhibitors). A HiTrap Q (5 ml; Amersham Pharmacia) column was equilibrated with binding buffer, and the sample was applied at 5 ml/min. The column was washed with 0.25 mol/l NaCl to remove weakly bound glycoproteins and unincorporated [³⁵S]sulfate. The remaining bound material was eluted with a linear salt gradient composed of 0.25–2 mol/l NaCl in binding buffer. Fractions of 2.5 ml were collected. Total counts in each fraction were determined by liquid scintillation counting. The fractions containing PGs were pooled and dialyzed against H₂O at 4°C, lyophilized, and then dissolved in a small volume of H₂O and stored at −20°C until use.

Whole frozen livers of obese Zucker rats and lean littermates were minced, dissolved to cold binding buffer (8 mol/l urea, 2 mmol/l EDTA, 0.5% Triton X-100, and 20 mmol/l Tris-HCl [pH 7.5]) with protease inhibitors and centrifuged for 15 min at 20,000g. The supernatants were filtered and used to isolate total PGs with the same column (HiTrap Q; Amersham Pharmacia), and salt gradient that was used to purify PGs from tissue culture (above). [³⁵S]sulfate-labeled rat liver PGs were prepared with the same protocol after injecting two Zucker rats (one lean and one obese littermate) with 2 × 1 mCi [³⁵S]sulfate, as previously described (5).

Digestion and electrophoresis of GAGs were done as previously described (12). GAGs were fixed with 0.1% cetylpyridinium chloride and stained with 0.1% toluidine blue (13). A Blyscan GAG assay kit (Biocolor, Belfast, U.K.) was used for GAG determinations. Protein was determined with a Pierce (Rockford, IL) protein determination kit. Compositional disaccharide analysis of [³⁵S]-labeled HS was performed as described previously (14).

After incubations with NEFAs, cells were detached by a brief (5-min) incubation with trypsin, collected in cold phosphate-buffered saline with Ca²⁺ and Mg²⁺, and immediately put on ice. Cells were pelleted, and a kit was used to prepare total RNA (Invitrogen, Leek, the Netherlands). After converting mRNA to cDNA by incubation with RT and buffer containing 50 mmol/l Tris HCl, pH 8.3, 75 mmol/l KCl, 3 mmol/l MgCl₂, and 10 mmol/l dithiothreitol, samples were divided. The fluorogenic 5′ nuclease (TaqMan) assay (real-time PCR; Applied Biosystems, Foster City, CA) (15) and the ABI Prism 7700 sequence detection system (Applied Biosystems) were used to estimate mRNA content. Primers for human versican were selected in a region of versican that is common to all splice variants (16) and were sense GGTGCCTCTGCCTTCCAA (position 6754–6771) and antisense TGCCAGCCATAGTCACATGTC (position 6806–6826), defining a fragment of 73 nucleotides. A 6-carboxy-fluorescein (FAM)−labeled probe TTATGTTGGTGCACTTTGTGAGCAAGATACCG (position 6773–6804) was used. Primers for syndecan 1 were sense GAGGGCTGCTGAGGATGGA (position 772–790) and antisense ATTCTCCCCCGAGGTTTCAA (position 843–862), defining a fragment of 91 nucleotides; a FAM-labeled probe CCTCCAGTCAGCTCCCAGCAGCA (position 792–814) was used. A predeveloped primer/probe for control 18S ribosomal RNA (suitable for both human and rat) was purchased from Applied Biosystems.

Livers from lean and obese Zucker rats were excised and immediately frozen in liquid nitrogen. The frozen livers were kept at −80°C until use. The same kit that was used for isolation of RNA from HepG2 cells (Invitrogen) was used to prepare RNA from frozen rat livers, following the manufacturer’s instructions. Primers for rat versican were sense TGCCTCTGCCTTCCGAGTT (position 6435–6453) and antisense CAGCCATAGTCGCATGTCTCA (position 6482–6502), defining a fragment of 68 nucleotides; a FAM-labeled probe TGTCGGTGCACTCTGCGAACAAGAC (position 6455–6479) was used. Primers for rat syndecan 1 were sense GTTGTGGAGGATGAAACTACCAATC (position 803–827) and antisense AGCTGTGTTCTCCCCAGATGTT (position 874–895), defining a fragment of 93 nucleotides; a FAM-labeled probe CAAAGGTGAAGTCTTGTTCTCCAGAGCCCT (position 843–872) was used.

Gel mobility shift studies to evaluate PG-lipoprotein interaction were performed as previously reported (17). The same amounts of biolabeled HepG2 PGs as determined by [³⁵S]sulfate content or rat liver PGs (as determined by GAG content) were incubated with increasing amounts of lipoproteins for 1 h at room temperature and electrophoresed. Evaluation of the gels was done by autoradiography (HepG2 cells) or by staining the gels for GAG (liver PGs).

All densitometric evaluations were done with a Bio-Rad molecular imager system and evaluated with Quantity One software (Bio-Rad, Hercules, CA). Graph Pad Prism software (San Diego, CA) was used to fit one-site hyperbola binding curves to the data. Results are given as means ± SD. Differences between two groups were identified with a Student’s t test (Fig. 5). For multiple groups, a one-way analysis of variance (ANOVA) and Newman-Keuls post hoc test were used to identify differences (Figs. 2 and 3). P > 0.05 was considered nonsignificant.

To study PG expression in hepatic cells, we used an in vitro model (HepG2 cells) and an ex vivo model (livers from obese Zucker rats and their lean littermates). Isolated PGs from each model were subsequently used to estimate binding of β-VLDL and model remnant particles (β-VLDL + apoE) with gel mobility shift (17).

HepG2 cells were exposed to low and high levels of albumin-bound linoleate or insulin that were within physiological ranges. In all experiments performed, HepG2 cells incubated with 10 nmol/l insulin secreted ∼15% more extracellular matrix PGs, measured as incorporated [³⁵S]sulfate per milligram of cell protein, whereas cells incubated with 300 μmol/l albumin-bound linoleate appeared to decrease their PG secretion by 10%. To determine the GAG distribution in the PGs produced under the different growth conditions, the PG preparations were digested with chondroitinase AC that hydrolyzes CS-4 and -6, chondroitinase ABC that degrades CS and DS, and heparitinase that hydrolyzes HS. The largest effect appeared to be on CS secretion: insulin increased CS secretion, whereas linoleic acid markedly decreased CS secretion (Fig. 1). In cells incubated with both insulin and linoleic acid, the NEFAs appeared to almost neutralize the insulin action. Cell surface PGs appeared to contain mostly HS, with very little CS and almost no DS. No changes in the cell surface PGs were observed with insulin and linoleic acid (not shown). The degree of 6-O and 2-O sulfation (18) of the cell-associated and secreted HS was similar between the two incubation conditions (not shown). However, other structural changes of HS could not be excluded.

RT-PCR showed that in HepG2 cells, insulin induced an increase in the mRNA for the core protein of versican, a secreted CS PG (Fig. 2A). High levels of NEFAs (300 μmol/l linoleic acid), on the other hand, markedly decreased mRNA for versican. Combining linoleic acid and insulin partially neutralized the effect of linoleic acid (Fig. 2A). Syndecan 1, a membrane-bound PG with predominantly HS GAG, appeared to be similarly affected by insulin and linoleic acid (Fig. 2B). However, linoleic acid had a less intense effect on syndecan 1 mRNA.

Linoleic acid or its metabolites acid may stimulate proliferator activated receptor-α in hepatic cells (19). We were unable to find regions with high homology to the proposed peroxisomal proliferator response element in the promoter region of versican (20). We tested to see whether darglitazone and bezafibrate, both proliferator-activated receptor-α (PPAR-α) agonists, could mimic the effect of linoleic acid (Fig. 3). Bezafibrate decreased mRNA for versican to the same extent as did linoleic acid, whereas darglitazone had no observable effect. The concentrations of darglitazone and bezafibrate used usually show effects in vitro (7,19). The decrease in versican mRNA that linoleic acid induced, although observed in all studies, varied somewhat. The difference in magnitude of the effects seen in individual experiments is illustrated by the differences between Figs. 2A and 3.

As an ex vivo model of the possible chronic effects of high NEFA levels and insulin on liver PGs, we used obese Zucker fa/fa rats and their lean fa/+ littermates. The obese animals had insulin levels of 4.0 ± 1.0 nmol/l, plasma NEFA levels of 550–1,200 μmol/l, and triglyceride levels of 7–11 mmol/l, mostly associated with remnant particles. The lean littermates had insulin levels of 0.4 ± 0.2 nmol/l, NEFA levels of 50–140 μmol/l, and triglyceride levels of 0.7–1.2 mmol/l, mostly VLDLs. To estimate the GAG composition in rat liver PGs, the PGs were labeled in vivo with [³⁵S]sulfate and purified as described in research design and methods. The GAG distribution of the isolated PGs was estimated by enzymatic digestion, as were those of HepG2 cells; the percent distribution of [³⁵S]sulfate as HS, DS, and CS was calculated. These data indicated a shift toward increased CS content at the expense of HS in the liver of the obese rat (Fig. 4). It should be stressed that the PGs extracted from whole liver were a mixture of cell surface and secreted (extracellular matrix) PGs. When mRNA for versican and syndecan 1 core proteins from liver of obese and lean Zucker rats were compared, the obese animals show increased expression of these proteins (Fig. 5). This resembled the pattern induced by insulin on HepG2 cells (Fig. 2). Preliminary data indicated that when both obese and lean littermate Zucker rats were older, the differences remained, but were less pronounced (not shown).

ApoE-enriched rabbit β-VLDLs have been used as a model for the retention of remnant particles mediated by liver PGs (6). Using these particles, we estimated lipoprotein-binding affinities to PGs isolated from HepG2 cells and Zucker rat livers. The altered PG composition after treatment with insulin appeared to decrease the binding of rabbit apoE-enriched β-VLDLs to secreted HepG2 cell PGs (Fig. 6), whereas the binding of LDLs was unaffected (not shown). Binding of β-VLDLs without apoE to PGs secreted from HepG2 cells was low and was unaffected by insulin or NEFAs (not shown). Linoleic acid partially reversed the effect of insulin on the binding isotherm. Thus, high insulin combined with high NEFA levels appeared to decrease the affinity of the remnant lipoproteins for the PGs. β-VLDL (without additional apoE) bound to PGs isolated from lean rat livers in a saturable manner and with an apparent K_d of 25–50 ng/ml, whereas the rate at which β-VLDLs bound to PGs isolated from obese rat livers was lower than the detection limit of the assay (not shown). The maximal binding of apoE-enriched β-VLDLs to PGs isolated from obese rat livers was ∼40% of that bound to PGs isolated from lean rat livers (Fig. 7). The electrophoretic mobility of PGs isolated from all rat livers showed a higher negative charge than that of PGs isolated from obese rat liver. This probably reflected the higher relative content of HS-containing PGs in lean liver. A low negative-charge density of PGs from the obese animals could explain their reduced binding of apoE-enriched remnant VLDLs.

We found that the addition of insulin and albumin-bound linoleic acid to standard culture medium increased and decreased, respectively, the PGs synthesized by HepG2 cells and induced qualitative changes in their distribution (Fig. 1). Alterations were also observed in PGs from livers of obese Zucker rats—liver cells that were chronically exposed to high insulin and NEFA levels (Fig. 4). Interestingly, PGs isolated from both insulin-treated HepG2 cells and obese Zucker rats showed decreased binding of apoE-enriched VLDL remnants, a step that may be critical for lipoprotein internalization by liver cells (6). The most important effects of insulin and NEFAs on HepG2 cells were seen on secreted PGs. To what extent secreted or membrane-bound PGs contribute to binding and internalization of lipoproteins in intact liver remains to be explored in detail. Mahley and Ji (6) suggested that secreted PGs may contribute to this process by presenting or transporting lipoproteins to other receptors, such as the LDL receptor−related protein.

We evaluated versican and syndecan 1 expression in this study because syndecan 1 is abundant in liver cells, and our preliminary experiments indicated the presence of versican in hepatocytes (21). We failed to find evidence of the presence of decorin in HepG2 cells. Changes in the expression of versican and syndecan 1 may have reflected the quantitative changes in PG production and should have been the cause of qualitative differences in the most common GAGs, which are CS, DS, and HS (Figs. 1 and 4). Syndecan 1 appeared to mediate a large part of the binding of chylomicrons and VLDL remnant particles to hepatic cells, and inhibition of its synthesis markedly decreased the binding of remnant lipoproteins to HepG2 cells (21). Our results showed that exposure to insulin slightly increased PG synthesis in HepG2 cells and shifted the GAG composition in secreted PGs toward PGs with a lower affinity for VLDL remnants. Linoleic acid appeared to affect PG synthesis in the opposite direction (Figs. 1 and 2), as did bezafibrate (Fig. 3). We speculate that this may have been a manifestation of NEFA-induced insulin resistance at the level of extracellular matrix metabolism. At the concentrations used, downregulation of versican by linoleic acid appeared to be more intense than its upregulation by insulin in HepG2 cells in vitro (Fig. 2). We found a higher level of versican in the liver of obese Zucker rats chronically exposed to high circulating insulin and NEFA levels than in the liver of lean rats (Fig. 5). It should be pointed out, however, that the in vivo labeling was performed in only two animals (Fig. 4). We speculate that in this in vivo model, the insulin effect overrode the NEFA effect. Although this preliminary observation may indicate that the PG alterations of HepG2 cells and liver cells of Zucker rat may have had similar origin, one should be careful to not overinterpret the resemblance, as these two models have many differences. Specifically, the effects of insulin and NEFAs studied in HepG2 cells were from relatively short exposure experiments, whereas in the hyperinsulinemic dyslipidemic Zucker rat, these agents acted chronically.

Studies on smooth muscle cells (7) and endothelial cells (8) have indicated that albumin-bound oleic and linoleic acids have similar effects on PG modulation. However, Hennig and colleagues (10,22,23,24,25) found that linolenic acid and saturated fatty acids were less potent modulators than linoleic acid on endothelial barrier function and PG synthesis. Although the effects on smooth muscle and endothelial cells are not necessarily transferable to hepatic cells, the above considerations led us to choose linoleic acid as a first model to study NEFAs’ effects on hepatic cells. The level of 300 μmol/l of BSA-bound linoleic acid was selected as a value that showed clear effects, but that was well within the physiological concentrations that could exist in extracellular fluid (26). An alternative approach would have been to use a mixture of fatty acids that mimic a more physiological situation. The use of mixtures with several physiologically relevant compositions is an important question that should be addressed in depth in a separate set of studies.

There were several nonexclusive mechanisms that could explain the observed effects of insulin and NEFAs on metabolism of extracellular matrix PGs. It has been proposed that the hexosamine biosynthetic pathway is a nutrient-sensing pathway that can be primed by several stimuli, including infusion of NEFAs (27). The end products of this pathway, UDP-Glc-NAc and UDP-Gal-Nac, are the building blocks for the synthesis of the GAG chains of PGs. The concentration of some of the end products of the hexosamine pathway can be altered by NEFAs (27). On the other hand, activation of protein kinase C (PKC) isoforms appears to be involved in several diabetic complications, including basement membrane thickening by excess extracellular matrix production (28). PKC also modulates some of the effects of NEFAs on smooth muscle cells, probably through the action of diacylglycerol, which interferes with insulin signaling (29,30), on PKC isoforms. In preliminary experiments, we found that some of the effects of NEFAs on PG secretion in HepG2 cells could be partially reversed by PKC inhibition (not shown). The observed opposite effects of linoleic acid and insulin on PG biosynthesis and GAG composition observed in HepG2 cells (Figs. 1 and 2) are compatible with such interference of NEFAs with insulin action.

A third mechanism involving the transcription factor PPAR-α is supported by the observed similitude of the action of the PPAR-α agonist bezafibrate and linoleic acid on gene expression of the versican core protein (Fig. 3). In hepatocytes, polyunsaturated fatty acids and their oxidation products may be ligands for nuclear transcription factors that modulate expression of genes controlling lipid and carbohydrate metabolism in liver cells (20). Such mechanisms may also have been involved in the effects of insulin and NEFAs on PG synthesis reported here, which in this case was modulated by a PPAR-α but not a PPAR-γ agonist (Fig. 3). Such results contrast with those from studies of human arterial smooth muscle cells (7). In these cells, the PPAR-γ agonist darglitazone opposed the stimulatory action of linoleate on decorin gene expression. These differences should be subject to detailed investigation, but we speculate that they may be related to the differential effects of NEFAs on gene expression for the core protein of different PGs. The lack of effect of darglitazone on HepG2 PGs may also have been related to the more prominent expression of PPAR-α in liver cells than in smooth muscle cells.

One of the important functions of hepatic extracellular PGs is their participation in lipoprotein metabolism as high-capacity low-affinity receptors of chylomicron and VLDL remnants (6). Our binding studies showed a decreased binding of apoE-enriched β-VLDLs to secreted PGs from HepG2 cells when the cells were cultured in medium enriched in insulin (Fig. 6). The curves in Fig. 6 indicated no change of apoE-enriched β-VLDL binding to PGs produced by HepG2 cells incubated in medium supplemented with 300 μmol/l linoleic acid when compared with PGs produced by control cells. Addition of insulin alone to the culture medium drastically changed the binding characteristics, and we were unable to fit a one-site binding curve to this data. Interestingly, PGs secreted by HepG2 cells that were cultured in medium supplemented with insulin and linoleic acid exhibited about half the total PG binding of control cells or cells incubated only with linoleate (Fig. 6). Lipoproteins and apolipoproteins show specificity in their binding to different GAGs (2). Thus NEFA- or insulin-induced extracellular GAG changes of hepatic cells could modify the binding of specific lipoproteins, especially changes involving HS PGs that in hepatic cells appear critical for binding and internalization of apoE-enriched β-VLDLs (Fig. 6) (6). The importance of liver HS PGs for postprandial clearance of remnants, and the fact that their postprandial clearance rate is reduced in type 2 diabetes, have been previously described (31). However, to our knowledge, whether liver PGs and GAGs are affected in type 2 diabetes has not been investigated. In our studies, obese rat liver appeared to contain relatively more CS, at the expense of DS and HS, than lean rat liver (Fig. 4). This observation is in line with studies showing a marked increase in the core protein genes for both syndecan 1 and versican, which are CS PGs, in liver from obese rats compared with liver from lean rats (Fig. 5). Our studies indicated a decreased binding of β-VLDLs and apoE-enriched β-VLDLs to PGs isolated from livers of obese rats (Fig. 7). If a similar situation existed in vivo, it could contribute to increased circulating remnant particles in obese Zucker rats. In a type 1 diabetes model, Ebara et al. (32) found delayed catabolism of apoB-48 lipoproteins because of decreased HS PG production in diabetic mice. They suggested that the effect was attributable to the effects of hyperglycemia on matrix production. Our findings suggest that in insulin resistance, the chronically elevated levels of insulin and circulating NEFAs affect PGs of the liver extracellular matrix and decrease the clearance rate for triglyceride-rich remnant lipoproteins. This is a condition that could contribute to the dyslipidemia of insulin resistance and type 2 diabetes.

This work was supported by grants from the Swedish Medical Research Council, the Swedish Heart-Lung Foundation, Strategic Funds National Network for Cardiovascular Research, the foundations of the National Board of Health and Welfare, the Medical Faculty at Göteborg University, Magnus Bergvalls Foundation, and AstraZeneca.

We thank Eva Gottfridsson for her technical help.