Reducing Macrophage Proteoglycan Sulfation increases Atherosclerosis and Obesity through Enhanced Type I Interferon Signaling

Myeloid-specific Ndst1 gene disruption Increases atherosclerosis

To examine the impact of altering macrophage HS on atherosclerosis, we bred Ndst1^f/fLysMCre⁺ mice onto an Ldlr^−/− background. In preliminary studies, we found that feeding a high-fat, high-cholesterol (HFC) diet to the male mice led to greater weight gain than in female mice. Therefore, to avoid complications of differential weight gain, we fed female animals a HFC diet for up to 12 weeks. Total cholesterol and triglyceride levels did not differ between Ldlr^−/−Ndst1^f/fLysMCre⁺ and Ldlr^−/−Ndst1^f/fLysMCre⁻ mice prior to placing the animals on the diet (Fig. 2A–B). Plasma triglycerides and cholesterol progressively increased to the same extent in both strains on the HFC diet. FPLC analysis of plasma taken after 12-weeks demonstrated equivalent accumulation of chylomicron/VLDL triglycerides and chylomicron/VLDL and IDL/LDL cholesterol in both strains, while HDL cholesterol levels were unchanged (Fig. 2C–D). No difference in body weight was observed between Ldlr^−/−Ndst1^f/fLysMCre⁻ and Ldlr^−/−Ndst1^f/fLysMCre⁺ (24.2 ± 0.8 g [n=14] versus 25.0 ± 1.1 g [n=15]) mice at the end of the feeding period. Glucose homeostasis between the groups was similar based on their fasting glucose and insulin levels, although there was a trend towards higher glucose and lower insulin in the mutant (Fig. 2E–F). Glucose tolerance and insulin tolerance were comparable (Fig. 2G–H).

Open in a separate window

Figure 2

Macrophage NDST1 Inactivation Increases Atherogenesis and Lesion Macrophage Content

Plasma cholesterol and triglycerides levels from female Ldlr^−/−Ndst1^f/f and Ldlr^−/−Ndst1^f/fLysMCre⁺ mice at different times after commencing HFC diet feeding (A and B). FPLC analysis of plasma after 12-weeks HFC diet feeding (n = 6) (C and D). Fasting plasma glucose and insulin levels before and after an oral glucose gavage (n=5–7) (E and F). Glucose tolerance (G) and insulin tolerance (G) tests were performed after 11-weeks HFC diet feeding (n=5–7). En face analysis of atherosclerosis and quantification of Sudan IV positive area (E and F). Staining and quantification of atherosclerotic lesion size in the aortic sinus (n = 5–7) (G and H). Atherosclerotic lesions stained with Van Gieson for collagen or F4/80 for macrophages (I). Quantification of necrotic core size (J) and F4/80 positive area (K) in lesions after 12-weeks HFC diet feeding. Monocyte(L) and Neutrophil (L) counts in blood during HFC diet feeding. Shown are mean values ± SEM, *p<0.05 and **p<0.01.

En face analysis of atherosclerosis in the aorta showed plaque formation restricted to the aortic root and arch in both strains (Fig. 2I). However, inactivation of Ndst1 caused a two-fold increase in plaque surface area in Ldlr^−/−Ndst1^f/fLysMCre⁺ mice compared to Ldlr^−/−Ndst1^f/fLysMCre⁻ controls expressed either as a percentage of total surface area and by absolute plaque area (Fig. 2J). A similar significant increase was observed in male mice (Supp. Fig. 1B–C). Aortic root cross-sectional analysis confirmed that plaque volume in Ldlr^−/−Ndst1^f/fLysMCre⁺ mice increased as well (Fig. 2K–L). Visual inspection of van Gieson stained cross sections of lesions at the aortic root strongly suggested that plaques from Ldlr^−/−Ndst1^f/fLysMCre⁺ mice had larger necrotic cores (Fig 2M), which was confirmed by image quantitation (Fig. 2N), suggesting that the lesions were more advanced compared to those in control mice. Lesions from Ldlr^−/−Ndst1^f/fLysMCre⁺ mice also had a significantly higher macrophage content compared to lesions from control mice (40% of lesion area in the mutant vs. 25% in the wild-type; P<0.001) (Fig. 2M lower panel and Fig. 2O), but T-cell infiltration was not elevated (Supp. Fig. 1D–F). The proliferating and apoptotic cell content in the lesions was similar between the groups as well (Supp. Fig. 1G–J).

ApoE bound to HSPGs controls proliferation of hematopoietic stem and multipotential progenitor cells in the bone marrow, monocytosis and monocyte accumulation in atherosclerotic lesions in mice (Murphy et al., 2011). Ldlr^−/−Ndst1^f/fLysMCre⁺ and Ldlr^−/−Ndst1^f/fLysMCre⁻ mice showed no difference in monocyte (Fig. 2P) or neutrophil counts (Fig. 2Q). Also plasma lymphocyte counts (Supp. Fig. 1I) and total white blood cell counts (Supp. Fig. 1J) did not differ.

These findings suggest that altering macrophage proteoglycan sulfation exacerbates atherosclerosis, enhances the number of macrophages in the lesion and lesion composition, but it does not affect plasma lipoproteins, infiltration of neutrophils and T-cells, or plasma myeloid or lymphocyte counts.

Macrophage proteoglycans modulate foam cell formation

Macrophage syndecan-4, a type of HSPG, has been implicated in foam cell conversion in vitro (Boyanovsky et al., 2009). To test if inactivation of Ndst1, which affects sulfation of all HSPGs in targeted cells, also reduced foam cell conversion, we elicited peritoneal macrophages in Ldlr^−/−Ndst1^f/fLysMCre⁺ and Ldlr^−/−Ndst1^f/fLysMCre⁻ mice fed an HFC diet. In contrast to these earlier findings, macrophages isolated from Ldlr^−/−Ndst1^f/fLysMCre⁺mice stained more intensely with Oil red O and accumulated significantly more cholesterol esters than Ldlr^−/−Ndst1^f/fLysMCre⁻ macrophages (Fig. 3A–C). To examine foam cell conversion in vitro, bone marrow derived macrophages (BMDM) were isolated and challenged with normal LDL, aggregated LDL (agLDL) and oxidized LDL (oxLDL). Incubation with agLDL dramatically induced cholesterol ester accumulation compared to the other lipoprotein preparations and occurred to a greater extent in macrophages derived from Ndst1^f/fLysMCre⁺ mice than control mice (Fig. 3D–E). Treatment of the cells with lipopolysaccharide enhanced further the extent of cholesterol ester accumulation in wild-type macrophages, but did not accentuate the elevation caused by altering HS (Fig. 3F). No gross changes were observed in free cholesterol levels between Ndst1^f/fLysMCre⁺ and Ndst1^f/fLysMCre⁻ macrophages (Supp. Fig. 2A–D). Macrophages from both male and female mutant mice responded similarly.

Open in a separate window

Figure 3

Macrophage NDST1 Inactivation Increases Foam Cell Conversion

Oil Red O staining of peritoneal macrophages collected after 12-weeks HFC diet feeding (A). Oil Red O positive peritoneal macrophages (B) and cholesterol ester levels (C) in peritoneal macrophages collected after 12-weeks HFC diet feeding (n = 3/group). Cholesterol ester levels in BMDM after a 24h incubation with 0 or 50 µg/ml of LDL, agLDL and oxLDL (n = 3) (D). Cholesterol ester levels in BMDM after a 24h incubation with different concentrations of agLDL (n = 3) (E). Cholesterol ester levels in BMDM after a 24h incubation with 0 or 50 µg/ml of LDL and agLDL in the presence or absence of 100 ng/ml LPS (n = 3) (F). Binding at 4°C or uptake at 37°C for 1h of DiD-agLDL in BMDMs (n = 3) (G). Cholesterol efflux in BMDMs after 24h incubation with 10 µg/ml apoAI or 50 µg/ml HDL (n = 3) (H). Cholesterol ester content in BMDMs after cholesterol efflux with 10 µg/ml apoAI or 50 µg/ml HDL (n = 3) (I). Shown are mean ± SEM, *p<0.05 and **p<0.01.

Incubation of BMDM for 1h with DiD-agLDL at 4°C or 37°C did not show any significant difference in binding or cell association (e.g. the sum of binding and uptake) between the two genotypes (Fig. 3G). Reverse cholesterol transport, as measured by efflux of ³H-cholesterol to apolipoprotein AI or high density lipoproteins (HDL), was slightly decreased in Ndst1^f/fLysMCre⁺ macrophages compared to controls (Fig. 3H), but the difference did not reflect an alteration in HDL binding (Supp. Fig. 2E) or expression of cell surface ABCA1 transporters and apolipoprotein E (Supp. Fig. 2F–G). The small decrease in efflux of radioactive cholesterol did not correspond to any difference in total cholesterol ester loss measured chemically, suggesting that decreased efflux was not the cause for the accumulation of lipid in the mutant macrophages (Fig. 3I).

Ndst1-deficiency results in an inflammatory gene expression profile

To gain insight into the mechanism underlying the alterations in the mutant, we compared the transcriptome of BMDMs from Ndst1^f/fLysMCre⁻ and Ndst1^f/fLysMCre⁺macrophages in basal growth medium (Fig. 4A). Increased expression of a relatively small set of genes occurred in the mutant, treated or untreated with agLDL, mostly in genes with functional annotations linked to immunity, defense and inflammatory response (Fig. 4B and Table S1). A comparable number of genes were decreased in expression, but these were linked to transmembrane signaling or apoptosis (Fig. 4C). qPCR analysis confirmed the enhanced expression of CCL2, CCL5, CCL7 and CCL8 (Fig. 4D). ELISA confirmed increased secretion of CCL5 and CCL7 (Fig. 4E–G). Other inflammatory genes were elevated as well, including TNFα, Ly6a, and IL10 (Fig. 4D). LPS treatment of macrophages greatly amplified the expression of these genes and diminished the differences in gene expression between mutant and wild-type seen under basal conditions, with the exception of CCL5 and IL10 (Supp. Fig. 3A). While IL10 is an anti-inflammatory cytokine, it is one of the most highly induced genes in classically activated macrophages following exposure to Toll-like Receptor ligands (Re and Strominger, 2004). Expression of other anti-inflammatory markers in mutant macrophages was either reduced or unaltered compared to wildtype (Supp. Fig 3B).

Open in a separate window

Figure 4

Reducing Macrophage HSPG Sulfation Increases Expression of Inflammatory Genes

Microarray analysis on RNA isolated from BMDMs (n = 4/group). Gene expression was filtered for genes from untreated Ndst1^f/fLysMCre⁺ BMDMs at least 2-fold different from WT control (A). Venn diagram of > 1.5x upregulated genes (B) and < 1.5x downregulated genes (C) and their overlap by agLDL treatment in the transcriptome of wild-type and Ndst1^f/fLysMCre⁺ BMDMs, and p values for selected gene ontology (GO) terms. qPCR validation from WT and Ndst1^f/fLysMCre⁺ BMDMs (D–F). Secretion of CCL5, CCL2 and CCL7 in the media by BMDMs after 24h at 37°C (E–G). ACAT activity levels in cell lysates from WT and Ndst1^f/fLysMCre⁺ BMDMs (H). qPCR validation from atherosclerotic plaques isolated from Ldlr ^−/− Ndst1^f/fLysMCre ⁻ and Ldlr ^−/− Ndst1^f/fLysMCre ⁺ mice after 12-weeks HFC diet feeding.(n=3–4) (I). Micro-images and quantification of infiltrated fluorescent bead-labeled Ly6C high monocytes (green) in lesions of Ldlr ^−/− Ndst1^f/fLysMCre ⁺ mice. Atherosclerotic lesions were stained for CD68 positive macrophages (red) and with DAPI for nuclei (blue) (n=3/group) (J–K). Cholesterol ester levels in BMDM after a 24h incubation with 50 µg/ml agLDL in the presence or absence of 10 µg/ml Sandoz 58–035 (ACATi), 30µM chloroquine or a combination (n = 3) (L). Shown are mean values ± SEM, *p<0.05 and **p<0.01.

Several of these chemokines act as chemoattractants for monocytes, thus providing a plausible explanation for the enhanced content of macrophages in the atherosclerotic lesions (Fig. 2M and 2O). To test if similar changes in gene expression occurred in vivo we dissected plaques from the aortic arch and extracted mRNA for qPCR analysis, correcting transcript expression for macrophage content in the tissue (F4/80 expression). The results confirmed significant elevated expression of inflammatory genes (CCL2, CCL7, CCL8, TNF-α, and IL6) in the atherosclerotic plaques with the exception of CCL5 and IL-10 (Fig. 4I). These findings suggest that the accumulating lesion macrophages express inflammatory genes and are thus proinflammatory M1-like (Ly6c-hi) macrophages. To evaluate if monocyte recruitment was altered, we tagged Ly6c-hi monocytes with fluorescent beads (Tacke et al., 2007) and showed enhanced accumulation into lesions of Ldlr^−/−Ndst1^f/fLysMCre⁺ mice compared to control mice (Fig. 2J–K).

ACAT2, an enzyme that converts cholesterol into cholesterol ester using long chain fatty acyl-coenzyme A, also was enhanced in Ndst1^f/fLysMCre⁺ macrophages (Fig. 4D). Western blot analysis of ACAT1 and ACAT2 expression revealed that both enzymes were significantly increased in Ndst1^f/fLysMCre⁺ macrophages (Supp. Fig. 3C). Interestingly, the increase in ACAT1 protein expression did not correlate with increased ACAT1 mRNA levels, except under conditions in which Ndst1^f/fLysMCre⁺ BMDMs were stimulated with LPS (Supp. Fig. 3D). Incubation of macrophages with agLDL and the ACAT inhibitor (ACATi) Sandoz 58–035 blocked accumulation of cholesterol esters, normalizing the levels in the mutant to that seen in the wildtype (Fig. 4L). Treatment of BMDMs with chloroquine, which inhibits lysosomal degradation of LDL particles, also prevented cholesterol ester accumulation (Fig. 4L). Taken together, these findings suggest that elevated expression of both ACAT1 and ACAT2 in Ndst1^f/fLysMCre⁺enhanced foam cell conversion.

Myeloid-specific Ndst1 Gene disruption promotes diet-induced obesity

Given the activated state of macrophages in Ndst1^f/fLysMCre⁺ mice, we wondered if the animals also might be susceptible to diet-induced obesity. Aged-matched male mice on a chow diet did not show any differences in body weight gain between the genotypes (Fig. 5A). However, when fed a high fat diet, Ndst1^f/fLysMCre⁺ mice were consistently heavier, with a significant increase in the rate of weight gain after 6 weeks (Fig. 5B). The increased body weight gain of the mutant mice correlated with increased mass of white adipose tissue (WAT), brown adipose tissue (BAT) and the liver, whereas other organs such as muscle were unaffected (Fig. 5C). The diameter of the adipocytes in white adipose tissue increased as well, consistent with accumulation of lipid droplets (Fig. 5D). Histological analysis revealed that Ndst1^f/fLysMCre⁺ mice exhibited increased lipid accumulation in liver, WAT, BAT and skeletal muscle (Fig. 5D–E).

Open in a separate window

Figure 5

Reducing Macrophage HSPG Sulfation Promotes Diet-Induced Obesity

Body weights of 8- and 18-weeks old male Ndstf^/fLysMCre- and Ndst1^f/fLysMCre⁺ mice on a chow diet (n = 4–8) (A). Male Ndstf^/fLysMCre- and Ndst1^f/fLysMCre⁺ mice on a HFC diet were weighed weekly (n = 8–10) (B). Wet weights of the liver, eWAT, sWAT, BAT and biceps femoris muscle after 18-weeks on the HFC diet (C). Adipocyte diameter distribution in eWAT (D) and H&E staining of Liver, eWAT, sWAT, BAT and biceps femoris muscle (E) of Ndst1^f/fLysMCre and Ndst1^f/fLysMCre⁺ mice 18-weeks on a HFC diet (n = 4/group). Feeding behavior of 26-week old Ndst1^f/fLysMCre ⁻ and Ndst1^f/fLysMCre ⁺ on a chow or HFC diet for 18 weeks (n = 4–6) (F). Average VO₂ (ml/kg/hr) (G) and average VCO₂ (H) in the light and dark cycles. Energy Expenditure (I) and Respiratory Exchange Ratio (RER) (J) in Ndst1^f/fLysMCre ⁻and Ndst1^f/fLysMCre ⁺mice littermates over a 24-hour period. Average activity in the dark and light cycles (K). Fasting plasma glucose levels (L), fasting NEFA levels (M) and insulin levels before and 15 min after an oral glucose gavage (N) in Ndst1^f/fLysMCre ⁻ and Ndst1^f/fLysMCre ⁺ mice on a HFC diet (n = 4–6). (O) Glucose tolerance and insulin tolerance (P) tests were performed after 16-weeks on a HFC diet (n = 9). Shown are mean ± SEM, *p<0.05 and **p<0.01.

The increase in weight gain in mutant mice fed a HFC diet was not associated with differences in food intake (Fig. 5F). Indirect calorimetry measurements at the end of HFC diet showed a trend towards reduced oxygen consumption and CO₂ production in Ndst1^f/fLysMCre⁺ mice, which translated into a significant reduction in energy expenditure during the dark phase (Fig 5G–I). These changes were independent of alterations in respiratory exchange ratio (RER) or activity levels between both mouse groups (Fig. 5J–K).

The increase in weight gain on the HFC diet was associated with pronounced differences in glucose homeostasis between the groups. Ndst1^f/fLysMCre⁺ mice had elevated fasting plasma glucose and NEFA (non-esterified fatty acid) levels and exhibited a pronounced hyperinsulinemia before and after glucose challenge (Fig. 5L–N). Reduced glucose and insulin tolerance indicated that Ndst1^f/fLysMCre⁺ mice had developed severe type-2 diabetes compared to control mice on the high-fat diet (Fig. 5O–P).

The increase fat content in adipose tissue and liver also correlated with increased macrophage infiltration (F4/80 and CD11c+ cells, Fig. 6A–C). More F4/80 macrophages stained for CD11c, suggesting that the cells were in an activated state (Fig. 6D). Expression of inflammatory cytokines, CCL2, CCL5, CCL7, CCL8, TNF-α, and IL-6 was elevated as well (Fig. 6E–J), but changes in IL-10 expression were modest (Fig. 6K). Accumulation of pro-inflammatory macrophages in adipose tissue is a hallmark of obesity-associated inflammation, which exacerbates adipogenesis and lipid deposition in tissues (Jin et al., 2013). We found that the inflammation in BAT correlated with reduced expression of PRDM16, PGC-1 and Evolv3, all markers for BAT associated non-shivering thermogenesis (Fig. 6L). Thus, the macrophage status in the adipose tissue was similar to that observed in macrophages present in the atherosclerotic lesions and their accumulation in adipose tissues is consistent with enhanced diet-induced obesity in the Ndst1^f/fLysMCre⁺ mice.

Open in a separate window

Figure 6

Macrophage HSPGs regulate adipose tissue inflammation in a model of diet-induced obesity

Expression of F4/80 and CD11c in BAT, eWAT and liver after 18-weeks on a HFC diet (A–B). Immunostaining images from eWAT of HFD-fed Ndst1^f/fLysMCre ⁻ and Ndst1^f/fLysMCre ⁺ mice co-stained with anti-F4/80 (green) and anti-caveolin-1 (red) antibodies (C). FACS analysis of F4/80+/CD11b+/CD11c+ cells in stromal vascular cells (n4–5/group) (D). Expression of CCL2, CCL5, CCL7, CCL8, IL6, TNF-a and IL10 in BAT, eWAT and liver after 18-weeks on a HFC diet (E–K). Expression of UCP1, Prdm16, Pgc-1α and Evolv3 in BAT liver after 18-weeks on a HFC diet (n = 4– 5/group) (L). Shown are mean values ± SEM, *p<0.05 and **p<0.01.

Heparan sulfate purification and disaccharide analysis

BMDM were treated with a trypsin for 10 min at 37°C to release cell surface proteoglycans. Disaccharides were quantified by mass spectrometry as described (Lawrence et al., 2008).

Plasma lipids and lipoprotein distribution and blood cell analysis

Blood was obtained by cardiac puncture or via tail the vein. Plasma cholesterol and triglyceride levels (Sekesui, San Diego, CA, US) and NEFA levels (WAKO Diagnostics, VA, US) were measured by commercially available enzymatic kits.

Quantification of aortic atherosclerosis and monocyte labeling in vivo

Mice euthanized by isoflurane intoxication were perfused with 10 ml PBS. The heart and ascending aorta down to the iliac bifurcation were removed and stained for neutral lipids using Sudan IV. In addition, aortic root cross-sectional atherosclerosis lesion size was determined using ImageJ software. Classical Ly-6Clo monocytes were labeled in vivo and bead infiltration was analyzed as described (Tacke et al., 2007).

Immunohistochemistry

Mice euthanized by isoflurane intoxication were perfused with PBS and 10% formalin (Fisher, SF93-4). The upper part of the hearts were placed onto a tissue mold, covered in OCT (Tissue-Tek), and frozen. Serial 10-µm cryosections from similar parts of the aortic sinus were used for staining. All other tissues were embedded in paraffin. The extent of staining was measured with ImageJ software.

RNA analysis

Total RNA was isolated from homogenized tissue and cells in Trizol and purified using RNeasy columns (QIAGEN). qPCR (SYBR Green) analysis was performed on an Applied Biosystems 7300 Real-time PCR system (Invitrogen). Global gene expression was measured by microarray analysis using the Affymetrix HT_MG430A array. Data are uploaded to the Gene Expression Omnibus and can be accessed via the following link: http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?token=sxkbymogflovlsn&acc={"type":"entrez-geo","attrs":{"text":"GSE51009","term_id":"51009"}}GSE51009.

Lipoprotein binding and uptake

Human LDL (density δ 1.019–1.063 g/mL) and HDL (density δ 1.063–1.21 g/mL) was isolated from plasma by buoyant density ultracentrifugation and quantified by BCA protein assay (Pierce).

Foam cell assay and reverse cholesterol transport

Macrophage cholesterol ester content was quantitated by the Amplex Red Cholesterol Assay Kit (Life Science Technologies, San Diego). Cholesterol ester content was calculated by subtracting free cholesterol from total cholesterol for each sample. Cholesterol efflux from BMDMs was performed as described (Nakaya et al., 2011).

Indirect Calorimetry

Mice were placed into Comprehensive Lab Animal Monitoring System (CLAMS; Columbus Instruments). Rates of O2 consumption (VO2; ml/kg/h) and CO2 production (VCO2) were measured for each chamber every 17 minutes throughout the study.

Glucose tolerance (GTT) and insulin tolerance test (ITT)

Mice were fasted for 5 hours. Blood samples were collected via tail vein bleeding immediately before and at 15, 30, 60, 90 and 120 min after oral glucose (2 mg/g body weight) gavage for GTT, after intraperitoneal injection of insulin (0.5U/g body weight) for ITT. Blood was immediately used for determination of glucose levels via the Accu-Chek® Aviva glucose monitoring system (Roche Applied Science GmbH, Mannheim, Germany). Insulin levels were determined via the ultra sensitive rat insulin ELISA kit (Chrystal Chem, Downers Grove, IL).

Nitrocellulose filter binding assay

Recombinant IFNAR1 (Life Science Technologies), IFNAR2, IFN-α4 and IFN-β (400 or 800 ng; PBL Interferon) were incubated with 10,000 counts of ³⁵S-labeled BMDM HS for 1 hr at 22 °C. The mixtures were applied to a prewashed nitrocellulose membrane to determine HS binding.

Stromal Vascular Fraction (SVF) isolation

Stromal vascular cells were prepared from collagenase-digested adipose tissue and FACS analysis of stromal vascular cells for macrophage content and subtypes was performed. The antibodies for surface staining were F4/80 (BM8), Ly6C (AL-21), CD11b (M1/70), and CD11c (N418) (eBioscience, San Diego, CA).

We would like to thank Jennifer Pattison and Karen Bowden for their assistance with FPLC, lipid analysis and aortic root sectioning. We would also like to thank UCSD Histology core facility for assistance with histology and the UCSD hematology core facility for hematology analysis and Tiffany Lee for assistance with sample processing for qPCR analysis. This work was supported by NIH grant GM33063 and P01 HL107150 (to J.D. Esko), HL 088093 (to J.L. Witztum), HL30568 and HL28481 (to A.J. Lusis) and the European Community FP7 Award PIOF-GA-2010-273994 (to P.L.S.M. Gordts).