Elevation of blood triglycerides, primarily as triglyceride-rich lipoproteins (TGRL), has been linked to cerebrovascular inflammation, vascular dementia, and Alzheimer’s disease (AD). Brain microvascular endothelial cells and astrocytes, two cell components of the neurovascular unit, participate in controlling blood-brain barrier (BBB) permeability and regulating neurovascular unit homeostasis. Our studies showed that infusion of high physiological concentrations of TGRL lipolysis products (TGRL + lipoprotein lipase) activate and injure brain endothelial cells and transiently increase the BBB transfer coefficient (Ki = permeability × surface area/volume) in vivo. However, little is known about how blood lipids affect astrocyte lipid accumulation and inflammation. To address this, we first demonstrated TGRL lipolysis products increased lipid droplet formation in cultured normal human astrocytes. We then evaluated the transcriptional pathways activated in astrocytes by TGRL lipolysis products and found upregulated stress and inflammatory-related genes including activating transcription factor 3 (ATF3), macrophage inflammatory protein-3α (MIP-3α), growth differentiation factor-15 (GDF15), and prostaglandin-endoperoxide synthase 2 (COX2). TGRL lipolysis products also activated the JNK/cJUN/ATF3 pathway, induced endoplasmic reticulum stress protein C/EBP homologous protein (CHOP), and the NF-κB pathway, while increasing secretion of MIP-3α, GDF15, and IL-8. Thus our results demonstrate TGRL lipolysis products increase the BBB transfer coefficient (Ki), induce astrocyte lipid droplet formation, activate cell stress pathways, and induce secretion of inflammatory cytokines. Our observations are consistent with evidence for lipid-induced neurovascular injury and inflammation, and we, therefore, speculate that lipid-induced astrocyte injury could play a role in cognitive decline.
- blood-brain barrier
- cell stress
several studies have demonstrated a correlation between dietary lipid composition, particularly saturated fatty acids and cholesterol derived from animal products, and the risk of developing dementia, including vascular dementia and Alzheimer’s disease (AD) (22, 24, 45). Moreover, animal studies have shown that a high-fat/high-cholesterol diet exacerbates the accumulation of amyloid-β, increases neuroinflammation, alters blood-brain barrier (BBB) integrity, and induces cognitive decline (11, 12, 42, 51, 58). Furthermore, Ldlr−/− mice, which have elevated blood low-density lipoprotein cholesterol levels, show evidence of increased brain inflammation and cognitive decline (10, 44). Elevated triglyceride-rich lipoproteins (TGRL) and fatty acids released by the lipolysis of TGRL (63) may contribute to the pathogenesis of dementia by inducing the expression of proinflammatory genes and initiating inflammatory and stress-related signaling pathways in cells of the neurovascular unit.
The neurovascular unit, composed of endothelial cells, astrocytes, pericytes, neurons, and microglia, is essential for the maintenance and function of the central nervous system. BBB endothelia are fused together by tight junctional proteins, such as claudins and occludins, that connect with the cytoskeleton network (8). Endothelial cells and astrocytes work together to limit paracellular and transcellular transport from blood to brain (1, 29, 57). Previously, the BBB was believed to be a relatively static, low-permeability barrier for blood-brain transfer; however, recent work indicates that this barrier is much more dynamic than previously thought (47).
It is well established that in contrast to leaky vessels in peripheral organs (43), the BBB restricts entry of polar molecules into the brain. Nutrients such as vitamins, glucose, and amino acids cross the BBB using specific transporters (70). Peptides and proteins generally do not cross the BBB (72), but they can be transported into the brain via specific transport systems if they are expressed in brain endothelium (71). Furthermore, vascular injuries can initiate a cascade of events leading to cerebrovascular dysfunction, neurodegeneration, cognitive impairment, and dementia (46).
Our previous studies using human aortic endothelial cells showed decreased transendothelial electrical resistance in a time-dependent manner in response to TGRL lipolysis products (TGRL + LpL). Furthermore, immunofluorescent localization of zonula occludens-1, occludin, and vascular endothelial cadherin showed rearrangement along cell borders after treatment with lipolysis products (15). Thus TGRL lipolysis products may influence endothelial barrier permeability by altering cell junctional protein arrangement.
Our studies have shown that TGRL lipolysis products activate stress- and inflammatory-related pathways in aortic endothelial cells though a reactive oxygen species-mediated pathway (5, 64). Furthermore, when human brain microvascular endothelial cells were exposed to TGRL lipolysis products, a network of stress, inflammation, cell cycle, and apoptosis pathways was induced, indicating a lipotoxic effect (4, 6, 68). In this setting, as the BBB endothelia experience lipotoxicity, we explored the effects of blood lipids on the other cells of the neurovascular unit, specifically astrocytes. The influence of TGRL lipolysis products, such as fatty acids, on astrocyte physiology is poorly understood. Astrocytes are not only the most abundant cell in the brain comprising 80% of the cells of the brain, but their activation is associated with increased BBB permeability and the pathogenesis of neurodegenerative diseases such as AD (54, 69). For instance, studies suggest that palmitic acid increases caspase activity, induces Jun amino-terminal kinases (JNK) and NF-κB pathways, and promotes cell death in neurons (39), while activating inflammatory pathways and stimulating apoptosis in cultured astrocytes (23, 67). Therefore, we hypothesize that lipids released from TGRL lipolysis cause an increase in BBB transfer coefficient (Ki = permeability × surface area/volume), which could be exacerbated in hyperlipidemic models such as the Ldlr−/− mouse. Furthermore, TGRL lipolysis products activate astrocyte stress- and inflammatory-related pathways thereby resulting in increased neuroinflammation.
The results of this study demonstrate that TGRL lipolysis products increase murine BBB transfer coefficient transiently in vivo when data from our hyperlipidemic and normolipidemic models are combined. Furthermore, normal human astrocytes in culture develop numerous lipid droplets and upregulate the expression of various stress-, transcription-, and inflammatory-related genes when exposed to TGRL lipolysis products. Our study identifies increased transcription and translation of C/EBP homologous protein (CHOP), activating transcription factor 3 (ATF3), activation of JNK and the NF-κB stress pathway, and increased secretion of inflammatory cytokines in response to TGRL lipolysis products. Thus our work indicates that TGRL lipolysis products induce astrocyte inflammation and stress. We speculate that these pathophysiological effects may contribute to the cognitive impairment seen with consumption of a high-fat diet by inducing neurovascular unit lipotoxicity that is manifest as increased BBB permeability, lipid droplet accumulation, and astrocyte-mediated neurovascular unit inflammation.
MATERIALS AND METHODS
Postprandial blood samples were obtained from human subjects 3.5 h after consumption of a moderately high-fat meal. Triglyceride-rich lipoproteins (TGRL) were isolated from the plasma at a density of less than 1.0063 g/ml following an 18-h centrifugation at 40,000 rpm and 14°C in a SW41 Ti swinging bucket rotor (Beckman Coulter, Sunnyvale, CA) within a Beckman Optima L-70 ultracentrifuge as previously described (5). The top fraction (TGRL) was collected and dialyzed in Spectrapor membrane tubing (molecular weight cutoff 3,500; Spectrum Medical Industries, Los Angeles, CA) at 4°C overnight against a saline solution containing 0.01% EDTA. The TGRL routinely used in experiments was normalized based on triglyceride concentrations (determined calorimetrically by triglyceride determination reagents; cat. no. T2449, F6428, and G7793; Sigma-Aldrich, St. Louis, MO). Plasma from five to six donors was pooled to prepare the TGRL for the experimental replicates. Postprandial very low density lipoprotein, esterified fatty acid, and nonesterified fatty acid (NEFA) content, as well as the NEFA content after LpL incubation of samples from four healthy volunteers, has previously been published (63). Variation between subjects is relatively small and TGRL from subjects was pooled for these studies. All human subject studies were approved by the Human Subjects Review Committee at the University of California, Davis, and written informed consent was obtained from the subjects.
Reagents and antibodies.
Lipoprotein lipase (LpL) (L2254), stearic acid (85679), palmitic acid (P5585), oleic acid (O1008), linoleic acid (L1012), and oil red o (O-0625) were purchased from Sigma, bovine serum albumin (30-AB79) from Fitzgerald (Acton, MA), and PCR primers from Operon (Huntsville, AL). Antibodies were purchased from the following sources: monoclonal anti-β-actin antibody (A 5441) and monoclonal anti-GFAP were purchased Sigma; human perilipin-2 antibody [adipose differentiation-related protein (ADRP), MAB7634] from R&D Systems (Minneapolis, MN); ATF-3 (sc-188) and cJUN (sc-1694), IκBα (sc-203), NF-κB p65 (sc-7151) from Santa Cruz Biotechnology (Santa Cruz, CA); p- NF-κB p65 (3033), p-cJUN (9261), phosphorylated and total mitogen-activated protein kinase (MAPK) family antibodies (9910 and 9926), CHOP (2895), and HRP-conjugated secondary anti-rabbit (7074) from Cell Signaling; and HRP-conjugated secondary anti-mouse (NA931V) from GE-Biosciences (Pittsburgh, PA).
Four male C57BL/6J [wild type (WT)] mice and four B6.129S7-Ldlrtm1Her/J (Ldlr−/−) mice (Jackson Laboratory) were studied as approved by the Animal Use and Care Committee at the University of California, Davis. Starting at the age of 8 wk, mice were fed ad libitum a control diet (TD.08485; Harlan Laboratories, Indianapolis, IN) over the course of 12 wk before MRI analysis. Approximately 1 h before MRI, mice were anesthetized with ketamine and xylazine (90/10 mg/kg body weight ip) and the surgical plane anesthesia was maintained with a quarter dose (22.5/2.5 mg/kg body weight iv) every 35 to 45 min after the first hour. The left femoral vein was cannulated with PE-10 tubing for administration of the anesthesia, saline, gadopentetate dimeglumine (Gd-DTPA-Magnevist; Bayer Healthcare Pharmaceuticals, Wayne, NJ), and TGRL lipolysis products. Body temperature was maintained during surgery using an electric heating pad.
Dynamic contrast enhanced-magnetic resonance imaging analysis of BBB transport: experimental setup.
MRI data were collected using a 7T Bruker Biospec MRS/MRI system (Bruker BioSpin MRI, Billerica, MA) with a 32-mm radiofrequency (RF) volume coil for transmission and detection interfaced with ParaVision 4 (PV4) software (Bruker BioSpin, Rheinstetten, Germany). Anesthetized, femoral vein cannulated mice were placed on a PVC animal stage in the prone position. Body temperature was maintained with circulating heated water within the animal bed (Gaymar, Orchard Park, NY) while the animal was in the magnet. The mice were placed inside the center of the coil and gradients such that the isocenter was 0.9 mm caudal to the bregma.
For each contrast agent Gd-DTPA (Magnevist) infusion, 10 consecutive T1-maps were acquired and analyzed using the Patlak plot to determine the BBB transfer coefficient Ki = permeability × surface area/volume (16). Each T1 map was obtained using a rapid-acquisition refocused-echo (RARE) sequence with variable TR = 200, 531.8, 958.6, 1557.2, 2568.7, and 7,500 ms; effective TE = 30.80 ms; RARE-factor = 16. A single slice was taken using 1-mm in-plane thickness, 32 × 32 mm2 field of view, 128 × 128 pixel resolution (thus 250 µm × 250 µm × 1 mm voxel size), with a 1.8-min acquisition time per T1 map (18 min for 10 consecutive maps required to create Patlak plots for each pixel). A 0.5 ml/kg bolus of the contrast agent, Gd-DTPA (0.5 mmol/ml) diluted 1:1 with saline, was injected intravenously followed by a 50-µl saline flush over ~10 s immediately after the acquisition of data for the first T1 map. TGRL lipolysis products (150 mg/dl TGRL + 2 U/ml LpL) were injected after the first series of 10 consecutive T1-maps (baseline) in a 50-µl volume followed by a 50-μl saline flush over ~10 s. Three more series of 10 consecutive T1-maps with contrast agent Gd-GTPA injection were taken post-TGRL lipolysis products injection to create the time course for each pixel over a total of four consecutive Patlak plots. T1 values were calculated using the T1 fit function in the PV4 software.
Postprocessing image analysis was done using an in-house MATLAB (Matlab 2014b; MathWorks, Natick, MA) script with Patlak linearized regression mathematical modeling for each pixel. The blood to brain transfer coefficient, Ki (min−1), was calculated from the slope of the Patlak plots that best fit an impermeable, unidirectional influx or bidirectional flux model selected using the F-test with P < 0.05 as previously described (16). Regions of interest (ROIs) were manually defined symmetrically about the midline, one from each hemisphere to exclude the median eminence and ventricles. Ki was calculated on a pixel-by-pixel basis and averaged to include both ROIs.
Cell culture and lipid treatments.
Normal human astrocytes (NHA; passage 3, Lonza, Walkersville MD) were cultured in astrocyte basal media (ABM, CC-3187) supplemented with AGM SingleQuot kit suppl. and growth factors (CC-4123) (Lonza, Walkersville MD) under and atmosphere of 5% CO2-95% air at 37°C. Cells were exposed for 3 h to the following conditions: media, TGRL (150 mg/dl), LpL (2 U/ml), TGRL lipolysis product (TGRL + LpL), ethanol control, or fatty acid treatments. The final concentrations of TGRL, LpL, and TGRL lipolysis products were diluted in media and preincubated for 30 min at 37°C before application. After incubation, supernatant was collected and cells washed with cold PBS and harvested as described below. For 24-h treatments, the supernatant was removed at 3 h and fresh media were applied. Both supernatant and cells were collected 24 h after initial treatment.
Preparation of fatty acids.
Fatty acid stock solutions of palmitic, stearic, oleic, and linoleic acid at 200 mM were prepared in 100% ethanol. Working water-soluble solutions of 5 mM fatty acids then were generated by incubating the fatty acids in 0.1 M Tris (pH 8.0) containing 10% endotoxin and fatty acid free bovine serum albumin (BSA) at 50°C for 15–30 min with occasional vortexing. This solution then was added to cells to obtain final fatty acid concentrations of 150 µM.
Lipid droplet quantification.
Astrocytes were grown on 12-mm round coverslips in a 24-well plate (BD Falcon) and were treated as described above (n = 3 coverslips/treatment group). Cells were then fixed with 4% paraformaldehyde then stained with oil red o. Bright-field photomicrographs were made from cells counterstained with hematoxylin. Lipid droplet counts were made with cells counterstained with 4',6-diamidino-2-phenylindole (DAPI) and coverslip mounted with Prolong Gold (Life Technologies, Carlsbad, CA). Fluorescent images from five regions per coverslip were taken with a ×63 objective on a Zeiss AxioObserver, LSM 700 microscope. The number of lipid droplets per cell was determined using ImageJ. Briefly, each red (oil red o) image was converted to an inverted black and white, threshold determined automatically, and the number of particles was measured if size was 0-infinity pixels with a circularity of 0–1. DAPI-stained nuclei were used to determine total number of cells per field. Average number of lipid droplet per cell was determined by dividing the total number of lipid droplet per field by number of nuclei per field.
Microarray analysis of gene transcription.
Total RNA was extracted from cells from three wells of a six-well culture plate per treatment group (media, LpL, TGRL, and TGRL lipolysis products) using an RNeasy Mini Kit (Qiagen, Valencia, CA) according to manufacturer’s protocol. An aliquot of RNA (5 μg) from each well in a defined treatment group was pooled and were processed for GeneChip analysis as previously described (5). A 200 ng aliquot of total RNA from each pooled sample was reverse transcribed, followed by aRNA (cRNA is also known as amplified RNA or aRNA) amplification, reverse transcription to synthesize first strand cDNA, second-strand cDNA synthesis, in vitro transcription to synthesize labeled aRNA, purification, and fragmentation of aRNA as described in the Affymetrix 3′ IVT Express Kit protocol (Affymetrix, Santa Clara, CA). The fragmented, biotin-labeled cRNA samples were hybridized to Human Genome U133A 2.0 Array (Affymetrix). The hybridizations, washings, labeling, and scanning of the GeneChips were performed as described in the Affymetrix protocols by Microarray Core Facility in the UC Davis Genome and Biomedical Sciences Facility. The microarray data set was deposited to Gene Expression Omnibus (GEO) with accession ID GSE76696.
Analysis and interpretation of GeneChip data.
The data were analyzed by GeneChip Operating System (GCOS) 1.4 (Affymetrix). The upper limit of P value for statistically reliable detection of an mRNA was 0.05 (except for “batch analysis,” see below), independent of its signal intensity (usually >10).
The lists of TGRL lipolysis-sensitive genes were obtained by using the “batch analysis” function in GCOS 1.4. The data for all of the 22,626 probe sets from the NHA cells treated with media were used as baseline and compared with those from the NHA treated with LpL, TGRL or TGRL lipolysis products. The lists of TGRL lipolysis product-sensitive genes were then sorted to satisfy three requirements: the mRNA must be detectable with P ≤ 0.05 in at least one of the two samples being compared, the fold change was (±) ≥2, and the gene products encoded by the mRNA must have an annotation that suggests either a known or a predicted function. The lower limit of twofold change was selected because our previous studies suggest that this selection criterion decreases the probability of obtaining “false positives” for selecting genes for further consideration (5).
The DAVID web server was used to decipher the Gene Ontology (GO) of up- and downregulated (±2-fold) expressed genes that were identified using subdata sets from LpL, TGRL, and TGRL lipolysis products treated, respectively. To perform functional enrichment analysis, Venn analysis was used to depict the overlapping ratios of the notably enriched biological process (also known as GO) terms (one-tailed Fishers’ extract test and Benjamini correction; adjusted P ≤ 0.05) between two groups (33). The enriched GO terms (adjusted P ≤ 0.05) for TGRL lipolysis product treated were observed.
Validation of changes in mRNA expression by quantitative RT-PCR.
Quantitative (q)RT-PCR was performed to validate and obtain statistical data verifying the significance of changes suggested by GeneChip assay of pooled RNA samples. These analyses were performed on individual archived aliquots of total RNA samples from each of the three wells from each treatment group that were pooled for GeneChip analysis.
Up to 5 μg of total RNA from each sample were reverse-transcribed to obtain cDNA in a final volume of 20 μl solution consisting of buffer, random hexamers, DTT, dNTPs, and Superscript-III reverse transcriptase (Invitrogen, Carlsbad, CA). qRT-PCR with SYBR Green as fluorescent reporter was used to quantify the expression of selected genes identified by GeneChip analysis. All the gene specific primers (Table 1) were designed with Primer Express 1.0 software (Applied Biosystems) using the gene sequences obtained from Affymetrix Probeset IDs. The reaction was carried out in 384-well optical plates containing 25 ng RNA in each well. The transcript levels were measured by quantitative RT-PCR using the ViiA 7 Real-Time PCR System (PE Applied Biosystems, Foster City, CA). The PCR amplification parameters were initial denaturation step at 95°C for 10 min followed by 40 cycles, each at 95°C for 15 s (melting) and 60°C for 1 min (annealing and extension). Relative changes in gene expression were determined from real-time quantitative PCR experiments by a ΔΔCT method (41) and were normalized with glyceraldehyde-3-phosphate dehydrogenase (GAPDH). The threshold cycle, Ct, which correlates inversely with the target mRNA levels, was measured as the cycle number at which the SYBR Green emission increases above a preset threshold level. The specific mRNA transcripts were expressed as fold difference from media control in the expression of the specific mRNAs.
Western blot analysis.
Cell were lysed in radioimmunoprecipitation assay buffer assay (RIPA) buffer containing 50 mM Tris (pH 7.4), 150 mM NaCl, 1% NP40, 0.25% sodiumdeoxycholate, 0.1% SDS, protease inhibitor cocktail set 1 (539131; Calbiochem, La Jolla, CA), 1 mM NaF, and 1 mM Na3VO4. Protein concentration was determined with the bicinchoninic acid assay (Pierce), and equal amounts of proteins (15 μg) were separated on a 4–15% Mini-PROTEAN TGX SDS-PAGE gel on a Biorad Mini (Bio-Rad, Hercules, CA) at 100 V for an average of 1.5 h. Proteins were then transferred onto 0.22 μm PVDF membranes (Bio-Rad) that were subsequently blocked with 5% nonfat milk for 1 h and then probed with either ATF3 (1:200), CHOP (1:500), cJUN (1:200), p-cJUN, p- and total MAP kinases (1:1,000), ADRP (0.2 µg/ml), and APP (1:5,000) at 4°C overnight or blotting control mouse monoclonal anti-β-actin (1:10,000) at room temperature for 1 h. Blots were subsequently washed 3 times with TBS with 0.05% tween (TBST). Membranes were then incubated with HRP-conjugated secondary anti-mouse or anti-rabbit antibody (1:5,000) for 1 h, followed by three washes with TBST. Blots were developed with the ECL Prime according to manufacturer's instructions (GE Healthcare, Pittsburgh, PA) followed by exposure to X-ray film. Protein expression levels were determined using ImageJ.
Astrocytes were grown to confluence on 12-mm round coverslips placed in 24-well plates (BD Falcon) and were treated (n = 5 coverslips per treatment group) as previously described for 3 h. After treatment, cells were fixed with 4% paraformaldehyde in PBS for 30 min at room temperature and then washed five times with PBS. Cells then were permeabilized and blocked with block/perm (0.05% saponin in 1% BSA/PBS) for 30 m and incubated with mouse anti-CHOP (1:5,000 dilution) overnight at 4°C. They were then washed with block/perm four times and then treated with goat anti-mouse antibody conjugated to Alexa 555 for 1 h. Unbound material was removed as above. The antibodies were then fixed with 4% paraformaldehyde for 10 min and washed two times with warm PBS. The nucleus was counter stained for 5 min with DAPI (1 μg/ml). After cells were mounted with Prolong gold, they were imaged as described for lipid droplet quantification. CHOP-related nuclear fluorescent intensity and percentage of CHOP positive cells were determined using ImageJ. Briefly; area of nuclei was defined using the DAPI stain. The fluorescent intensity of CHOP within the nuclei region was measured and CHOP-positive cells were defined as nuclei with an intensity >35 arbitrary units (background).
Supernatants from 3- and 24-h-treated NHA were assayed for cytokine secretion using ELISA kits for human IL-6, IL-8, IL-1β, TNF-α (BD Bioscience, San Diego, CA), GDF-15, macrophage inflammatory protein (MIP)-3α, MIP-1α, and MCP-1 (R&D Systems, Minneapolis, MN) according to manufacturer’s protocol. Colorimetric analysis was completed using a Bio-Rad xMark plate reader and concentrations calculated following the manufacturers’ instructions.
All statistical calculations were done using Sigma Stat software (Systat Software, San Jose, CA) or GraphPad Prism 6 software (GraphPad Software, La Jolla, CA). For the BBB transfer coefficient (Ki) assay, Grubb’s test was used to determine outliers. Two-way repeated measures were used to test for differences between genotypes across time and where affects were found the Student-Newman-Keuls (SNK) post hoc test for multiple comparisons was used to test for differences over time. Data from the lipid droplet abundance, gene expression, ELISA, and Western blot analysis are presented as means ± SE. Nonnormal samples were first transformed using natural log before analyzing. Thereafter, a one-way ANOVA with multiple comparisons (Tukey) was completed. Values were considered statistical significant when P ≤ 0.05.
Infusion of TGRL lipolysis products transiently increased the BBB transfer coefficient.
We first extended our understanding of lipotoxicity in neurovascular unit by evaluating the acute effect of an infusion of TGRL lipolysis products on BBB transport in vivo. C57BL/6J (WT) and Ldlr−/− mice fed a control diet were infused via a cannula in the femoral vein with TGRL lipolysis products (150 mg/dl TGRL + 2 U/ml LpL) and the BBB transfer coefficient, Ki = permeability × surface area/volume (min−1), was assessed by dynamic contrast enhanced-magnetic resonance imaging (DCE-MRI). By two-way repeated measures ANOVA, there was not a statistically significant difference between the genotypes (P = 0.897) or interaction between the genotype and time points (P = 0.355). However, we observed a significant effect for the time variable on Ki (Fig. 1 data combined; P = 0.024, n = 8). Post hoc SNK tests showed a significant difference between the 0–20 min and the baseline, 20–40 min, and 40–60 min postinjection time points (P = 0.035, 0.022, and 0.029, respectively) when C57BL/6 and Ldlr−/− mice were pooled. There were no significant differences among other time comparisons.
TGRL lipolysis products induced lipid droplets in human astrocytes.
To evaluate the influence of blood lipids on lipid sequestration, we asked whether lipolysis products of TGRL induce lipid droplet formation in normal human astrocytes, as has previously been seen in human macrophages (14). Astrocytes treated with media, LpL, TGRL or TGRL lipolysis products (TGRL + LpL) for 3 h were subjected to oil red o staining. Lipid droplets per cell were significantly increased (~13-fold) when treated with TGRL + LpL compared with media (Fig. 2, A and B). However, there was no significant change in number of lipid droplets in astrocytes exposed to LpL or TGRL alone. We then determined whether increased lipid droplet accumulation was associated with a change in expression of ADRP, a protein commonly associated with lipid droplets (9). Cell lysates from astrocytes exposed to media, LpL, TGRL, and TGRL + LpL showed no significant alteration in ADRP expression (data not shown).
Induction of transcription factors, inflammatory pathways, and stress-related genes by TGRL lipolysis products.
To gain a better understanding of the various pathways activated by blood lipids in the brain, the transcriptional response of astrocytes to media, LpL, TGRL, and TGRL + LpL was determined with high-density oligonucleotide arrays containing over 22,000 gene probe sets (Human Genome U133A 2.0 array, Affymetrix). Over 13,000 genes were transcribed reliably (P ≤ 0.05) among all four treatment groups. Table 2 gives as summary of the total number of genes affected (± ≥2 fold) by each of the treatments (LpL, TGRL, and TGRL +LpL) with the highest number in the TGRL + LpL group. DAVID analysis of microarray results was competed for functional enrichment analysis. We determine TGRL lipolysis products significantly enriched over 100 biological process (GO terms, Table 3) with no significant groups found with LpL or TGRL alone. The genes regulated by TGRL + LpL are segregated into three major groups: transcription factors, endoplasmic reticulum (ER)- and stress-related, and those involved with inflammatory pathways, which have been further divided into upregulated (Table 4) and downregulated (Table 5) genes. While there were many genes regulated by LpL or TGRL alone, we focused on a TGRL+ LpL treatment as it has the most physiologically relevance in regards to astrocytes.
Since pooled samples were utilized for microarray analysis, we then verified the upregulation of gene transcripts found by microarray using qRT-PCR. Inflammatory-related genes confirmed to be significantly upregulated by qRT-PCR (Fig. 3A) included MIP-2α, MIP-2β, MIP-3α, CXCL1, and prostaglandin-endoperoxide synthase 2 (PTGS2 also known as COX-2) (27.1-, 12.2-, 4.8-, 9.9-, and 6.2-fold of media, respectively). We also confirmed the transcription factor nuclear receptor-related 1 protein (NR4A2, 14.2-fold), Kruppel-like factor 4 (KLF4, 10.4-fold), and ATF3 (13.8-fold). All were significantly upregulated. Furthermore, stress-related pathway DNA damage-inducible transcripts 3 and 4 (DDIT3 and DDIT4) were significantly upregulated 15.9- and 6.8-fold, respectively. In addition to these highly expressed genes, we verified significant increases in other inflammatory and lipid-related genes including GDF-15 and ADRP (2.6 and 2.3). IL-8, while upregulated (2.4-fold) in the microarray, was not significantly increased by qRT-PCR validation (P = 0.083, Fig. 3B). We further evaluated the major TGRL lipolysis downregulated genes and validated that SMAD7 and SOX9 significantly downregulated (Fig. 3C). However, the same was not true for ITGB, SIX1, and PGF.
TGRL lipolysis products activate stress-related pathways.
Based on upregulation of stress regulated transcripts, we asked whether canonical stress signaling pathways are activated by TGRL lipolysis products. Western blot analysis demonstrated a significant increase (1.6-fold) in phosphorylation of stress-activated protein kinases (SAPK)/JNK in human astrocytes when exposed to TGRL lipolysis products when compared with media, while demonstrating no significant change by TGRL alone (Fig. 4A). Other MAPK pathways including p38 and extracellular signal-regulated kinase (ERK) phosphorylation also were evaluated but no significant change was observed (data not shown). We then evaluated specific downstream pathways in the JNK/cJUN/ATF3 signaling cascade for activation by lipolysis products. When compared with media, TGRL lipolysis products not only significantly increased phosphorylation of cJUN (3.4-fold) but also ATF3 (3.3-fold) (Fig. 4, B and C).
Another stress-related protein, C/EBP homologous protein (CHOP, also known as DDIT3), showed increased nuclear accumulation and the percentage of CHOP-positive cells increased from 7.5% in media to 76.7% after exposure to TGRL lipolysis products (Fig. 5, A–C). Furthermore, CHOP cellular abundance increased by 4.7-fold upon treatment with lipolysis products as determined by Western blot (Fig. 5, D and E). We also evaluated proteins in the NF-κB inflammatory pathway and found significant increase in phosphorylation of p65 subunit of NF-κB (Fig. 6A) and reduction of IκBα (inhibitory subunit of NF-κB, Fig. 6B).
Fatty acids released from TGRL lipolysis induce secretion of MIP-3α, GDF-15, and IL-8.
To evaluate if TGRL lipolysis products induced secretion of inflammatory-related cytokines, we treated astrocytes with media, LpL, TGRL, and TGRL + LpL for 3 h and changed to fresh media for 21 h (24 h total). Supernatant from the 24-h time point showed a significant increase in MIP-3α (2.6-fold), GDF-15 (2.5-fold), and IL-8 (2.4-fold) in TGRL + LpL when compared with control, with no change with LpL or TGRL alone (Fig. 7). Other cytokines evaluated, including TNF-α, IL-1β, MCP-1, MIP-1α, and IL-6 (despite increases seen in gene expression by microarray), were either undetectable or no significant change was detected at 24 h. The treatment effect of TGRL lipolysis products also was evaluated at 3 h with no significant changes in GDF-15 and MIP-3α, but a significant decrease in IL-8 (data not shown).
To elucidate which fatty acids released during lipolysis of TGRL (63) contribute to inflammatory response, we treated astrocytes with stearic, palmitic, lauric, or oleic acid and measured cytokine secretion at 24 h. IL-8, IL-6, and MIP-3α (4.6, 2.0, and 2.0-fold) were significantly increased in stearic acid-treated cells when compared with vehicle control but not with any of the other fatty acid treatments (Fig. 8).
Our earlier studies and those of others have demonstrated that the BBB transfer coefficient (Ki) in rats can be altered by diet, obesity, and/or TGRL lipolysis infusion (12, 47). Our studies are the first to model the postprandial state after a single meal in mice and show increased BBB Ki in as little as 20 min following TGRL lipolysis infusion and that this coefficient subsequently falls significantly to a level not significantly different from baseline within 40 min. This transient increase in BBB leak or transfer has been implicated in the development and/or progression of a variety of pathological conditions (52). We also show TGRL lipolysis products induce astrocyte lipid droplet formation, activate cell stress pathways, and induce secretion of inflammatory cytokines. Taken together, the data suggest elevated levels of circulating lipids, and likely saturated fatty acids, after a single meal may cause acute lipotoxic injury and can lead to activation of the neurovascular unit.
Because of their size, minimal transfer of gadolinium (Gd)-containing MR contrast agents across an intact BBB is expected. Gd-contrast agent extravasation is a validated and highly precise technique for measuring increases in brain transfer coefficient (Ki = permeability × surface area/volume) in human and rodent studies (17, 30, 36, 48, 50, 66). Here we used a similar MRI methodology (47, 61) to measure alterations in BBB Ki by measuring extravasation of gadolinium contrast agent and found a transient increase with TGRL lipolysis treatment. The transient nature of the change in Ki (i.e., the fall in Ki following the initial increase after TGRL lipolysis product infusion) remains to be investigated. However, these experiments are consistent with a repetitive injury model of lipotoxicity. Possible explanations for the increased BBB Ki are fatty acid integration into endothelial cell membranes to increase BBB permeability and/or direct activation of endothelial cells and astrocytes to increase transcytosis and/or paracellular transport. Although the data presented here address only an acute injury, we speculate that a chronic high-fat diet or genetic hyperlipidemia may lead to a chronic lipotoxicity and increase BBB compromise.
The observed increase in Ki after TGRL lipolysis injection suggests postprandial increases in Ki may facilitate migration of blood cells (e.g., monocytes) in response to inflammation and allow blood products including lipid-containing molecules to more easily move from blood into brain, where they can interact with pericytes, neurons, and astrocytes. Astrocytic feet are part of the BBB and once astrocytes are activated, their inflammatory response can provide a feedback mechanism that can affect BBB integrity at a later time point (1).
Previous studies have shown increased abundance of lipid droplets in cells of the frontal subventricular zone in triple transgenic mice expressing mutant APP, tau, and presenilin 1 (26) and in neurons from the hippocampus and cortex of AD patients that are positive for amyloid β (20). Here, we see a significant increase in the number of lipid droplets in astrocytes after exposure to TGRL lipolysis products. These lipid droplets could be acting to sequester excess lipid and lipid byproducts one might expect to cross the more permeable postprandial BBB. While there was an increase in ADRP gene expression, there was no change in protein abundance. This same phenomenon is seen in other models such as lipid accumulation (adipocyte maturation) where there is a change in ADRP mRNA but not protein as well (56). Perhaps TGRL + LpL alters the abundance of another perilipin or fatty acid binding proteins (i.e., Plin1 or FABP4), or the existing perilipins are redistributed into and across the lipid droplets. Results of previous studies and the current study suggest a link among TGRL lipolysis products, lipid accumulation, and AD. However, it is still unclear whether lipid droplet accumulation is a cause or consequence of cellular stress related to cognitive decline.
To determine the scope and function of transcriptional alterations in astrocytes due to TGRL lipolysis product exposure, we examined astrocytes by microarray analysis. Over 250 genes were upregulated by astrocytes in response to lipolysis products with the largest changes in transcripts related to gene transcription (29 genes), inflammation (22 genes), apoptosis/cell cycle (19 genes), and metabolism (11 genes) accounting for 31% of the changes observed. Genes involved in stress pathways (NR4A2, ATF3, and DDIT3) and cytokines (CXCL2, CXCL3, CCL20, and IL-8) were also found to be upregulated. These results indicate lipolysis products from TGRL induce a proinflammatory and stress-related gene response similar to that implicated with neuroinflammation, a response associated with and perhaps a mediator of AD (2, 3). While there were many genes regulated by both LpL and TGRL alone, we focused on a TGRL+ LpL treatment as it has the most physiologically relevance in regards to astrocytes. Lipoprotein lipase (LpL) anchored to brain microvascular endothelium binds to and hydrolyzes TGRL into lipolysis products, primarily free fatty acids, contributing to the influx of free fatty acids into the brain (7, 21, 53, 55, 62).
Since many proinflammatory responses are mediated by transcriptional activation of cytokine genes, we examined the upstream signaling of these key transcription factor complexes, phosphorylation of MAPKs including ERK, p38, and JNK. Activation of JNK, as indicated by an increase in phosphorylation, occurred after 3 h of treatment with TGRL lipolysis products, but both p38 and ERK yielded no change in phosphorylation status at this time point. This JNK pathway has been shown to be activated and induce transcription and activation of transcription factors including ATF3 in endothelial cells (5, 6). Previous studies have shown ATF3 expression is inhibited by ERK, while transcriptionally activated by JNK (25, 34). JNK phosphorylates cJUN, which binds to and activates ATF3. This complex then translocates to the nucleus, binding to AP-1 response elements initiating transcription (65, 68). Our findings here show both phosphorylation of cJUN and increased expression of ATF3 protein after three hours’ treatment with lipolysis products, indicating activation of the JNK/cJUN/ATF3 response pathway. Additionally, we see activation of the NF-κB pathway as indicated by phosphorylation of p65 and decreased protein expression of IκBα. Also, we found a significant increase in the cellular abundance and accumulation in the nucleus of CHOP, the major proapoptotic transcription factor induced by ER stress (6). Interestingly, others have found not only a correlation but also a cascade interaction of JNK-ATF3-CHOP that precedes cell death (28, 37, 60). While activation of apoptotic pathways including the caspases were not evaluated here, future studies are needed to determine how the inhibition of JNK, ATF3, CHOP, and/or NF-κB affects cell survival in the presence of TGRL lipolysis products.
Studies with rats on a high-fat diet showed increases in TNF-α, IL-1β, and IL-6 in the hypothalamus (13), while in vitro studies have shown that palmitic acid induces mitochondrial dysfunction and inflammatory signaling in neuronal cells (39) and inflammatory and apoptotic signaling in astrocytes (23, 67). Also, while predominantly associated with cardiovascular disease, GDF-15 plasma levels also have been shown to correlate with neurological function (19). Here, we show that TGRL lipolysis products induced astrocyte secretion of cytokines including MIP-3α, GDF-15, and IL-8, while stearic acid increased IL-8, IL6, and MIP-3α. TGRL lipolysis products consist of a pool of fatty acids released from the TGRL particle that includes saturated, mono-, and polyunsaturated fatty acids. Individual fatty acids may activate a unique pathway while TGRL lipolysis products may activate multiple pathways. However, all of these cytokines are linked to an inflammatory response: chemotaxis of neutrophils or lymphocytes, angiogenesis, apoptosis, and disruption of the BBB (18, 27, 31, 32, 38, 40, 49). Studies have shown that activation of the JNK/cJUN/ATF3 and NF-κB pathways have been linked to proliferation, prosurvival (65), and regulation of MIP-3α, GDF-15, and IL-8 (31, 40, 49). Furthermore, CHOP has been shown to enhance gene transcription by interacting with cJUN/FOS (59) and ATF3 (35). TGRL lipolysis products may induce expression of these cytokines through activation of JNK/cJUN/ATF3-, CHOP-, and NF-κB-related pathways. However, whether secretion of these cytokines contributes to astrocyte toxicity or exacerbates the overall neuroinflammatory response remains to be determined.
In conclusion, we have found that lipids from the lipolysis of TGRL increase the BBB transport coefficient transiently after intravenous infusion in mice, increase astrocyte lipid droplet accumulation, induce a strong proinflammatory and stress-related response expression in transcription-, apoptotic-, and inflammation-related genes, activate cytokine stress-kinase pathways, and stimulate the secretion of IL-8, GDF-15, and MIP-3α cytokines. Our study thus demonstrates TGRL lipolysis product-induced stress-mediated inflammation is dependent on the activation of multiple signaling pathways. Given these and our previous findings, we speculate TGRL lipolysis products may contribute to the cognitive impairment seen with consumption of a high-fat diet by inducing lipotoxicity that increases BBB transport, astrocyte-mediated neuroinflammation, and early AD pathology.
This work was supported by the National Institutes of Health (NIH) Grants AG-039094 and AG-045541 and the Richard A. and Nora Eccles Harrison Endowed Chair in Diabetes Research Fund and NIH Grants U24-DK-092993-05S1 (to J. C. Rutledge) and U24-DK-092993 (to K. Lloyd).
No conflicts of interest, financial or otherwise, are declared by the author(s).
L.L.L., J.C.R., and J.M.R. conceived and designed research; L.L.L., H.H.A., and J.M.R. performed experiments; L.L.L., H.H.A., and J.M.R. analyzed data; L.L.L., H.H.A., D.W.W., S.E.A., and J.M.R. interpreted results of experiments; L.L.L., H.H.A., and J.M.R. prepared figures; L.L.L., H.H.A., and J.M.R. drafted manuscript; L.L.L., H.H.A., D.W.W., S.E.A., J.C.R., and J.M.R. edited and revised manuscript; L.L.L., H.H.A., D.W.W., S.E.A., J.C.R., and J.M.R. approved final version of manuscript.
We thank Dr. Jeffrey Walton from the University of California, Davis Nuclear Magnetic Resonance Facility for assistance and expertise with MRI. We also acknowledge Dr. Kent Lloyd and the University of California, Davis Mouse Metabolic Phenotyping Center for support.
- Copyright © 2017 the American Physiological Society