Human FABP1 T94A variant impacts fatty acid metabolism and PPAR-α activation in cultured human female hepatocytes

Abstract

Although human liver fatty acid-binding protein (FABP1) T94A variant has been associated with nonalcoholic fatty liver disease and reduced ability of fenofibrate to lower serum triglycerides (TG) to target levels, molecular events leading to this phenotype are poorly understood. Cultured primary hepatocytes from female human subjects expressing the FABP1 T94A variant exhibited increased neutral lipid (TG, cholesteryl ester) accumulation associated with 1) upregulation of total FABP1, a key protein stimulating mitochondrial glycerol-3-phosphate acyltransferase (GPAM), the rate-limiting enzyme in lipogenesis; 2) increased mRNA expression of key enzymes in lipogenesis (GPAM, LPIN2) in heterozygotes; 3) decreased mRNA expression of microsomal triglyceride transfer protein; 4) increased secretion of ApoB100 but not TG; 5) decreased long-chain fatty acid (LCFA) β-oxidation. TG accumulation was not due to any increase in LCFA uptake, de novo lipogenesis, or the alternate monoacylglycerol O-acyltransferase pathway in lipogenesis. Despite increased expression of total FABP1 mRNA and protein, fenofibrate-mediated FABP1 redistribution to nuclei and ligand-induced peroxisome proliferator-activated receptor (PPAR-α) transcription of LCFA β-oxidative enzymes (carnitine palmitoyltransferase 1A, carnitine palmitoyltransferase 2, and acyl-coenzyme A oxidase 1, palmitoyl) were attenuated in FABP1 T94A hepatocytes. Although the phenotype of FABP1 T94A variant human hepatocytes exhibits some similarities to that of FABP1-null or PPAR-α-null hepatocytes and mice, expression of FABP1 T94A variant did not abolish or reduce ligand binding. Thus the FABP1 T94A variant represents an altered/reduced function mutation resulting in TG accumulation.

early studies of liver fatty acid-binding protein (FABP1), discovered in rat liver over 40 years ago, primarily characterized the murine protein in vitro and in transfected cells, suggesting short-term roles in long-chain fatty acid (LCFA) uptake, cytosolic transport, esterification, and oxidation (1, 37). More recent findings with purified recombinant proteins and murine hepatocytes from Fabp1-null mice suggested that FABP1 protein also functions in longer-term regulation of LCFA metabolism by facilitating LCFA nuclear targeting to induce transcriptional activity of peroxisome proliferator-activated receptor-α (PPAR-α) (16, 39, 46, 51). Like ablation of other FABP family members, Fabp1 gene ablation or Fabp1 antisense RNA treatment are not lethal in mice (1, 55). Instead, dietary challenge studies in mice revealed more subtle effects of FABP1 on lipid metabolism and whole body phenotype dependent on sex, age, background strain, and other factors (1, 11, 38).

Despite a high degree of amino acid sequence identity and overall structural similarities, the recombinant human FABP1 T94T differed significantly from rat and other species' FABP1s in having a much larger ligand-binding site structure, suggesting altered ligand affinity/specificity (6). The functional data available for human FABP1 T94T were obtained with only transformed tumor cell lines. Variation in the level of human FABP1 in HepG2 cells, a human liver-derived cell line, correlated directly with fatty acid uptake (65). Compared with normal liver or primary hepatocytes, hepatoma lines express no or less FABP1 (21, 65) and are less responsive to fibrates (49). Interestingly, a previous study has concluded that human hepatocytes are a better in vitro model than HepG2 cells relating to drug metabolism and toxicology (63).

Despite the paucity of studies addressing the physiological function of human FABP1, human FABP1 has also proven clinically useful as a marker for tissue injury. Appearance of human FABP1 in serum or urine reflects damage to liver (8), intestine (8), and kidney, especially in type 2 diabetes (62). Because humans but not rodents express FABP1 in the kidney, transgenic mice overexpressing the human FABP1 have proven especially useful in the latter regard (23).

Genomic studies demonstrated the existence of a highly conserved T->C single nucleotide polymorphism (SNP) in exon 3 of the human FABP1 gene (rs2241883) (53). This SNP gave rise to humans that are homozygous in either the wild-type, where both alleles are T (TT), variant, where both alleles are C (CC), or heterozygous with both T and C alleles (TC). This substitution resulting in a codon change in the mRNA from ACT to GCT with a subsequent threonine-to-alanine substitution (T94A) in the coded protein variant has spurred renewed interest in the functions and potential pathological roles of human FABP1 (5). The threonine (T94) residue is highly conserved in mammals (human, rat, mouse, pig, cattle) (5, 53). Unlike rare genetic polymorphisms in other FABP family members, the human FABP1 SNP has a high minor allele frequency of 26–38% with 8.3 ± 1.9% homozygous variant [minor allele frequency (MAF) = 0.264 for 1,000 genomes in NCBI dbSNP database; ALFRED database]. Whereas some studies suggested that FABP1 T94A did not alter serum TG or low-density lipoprotein (LDL) cholesterol (LDL-C), interpretation is complicated by combining data from female plus male subjects, heterozygous plus homozygous FABP1 T94A variant subjects, and/or a very small subject number. In contrast, other studies showed that female and not male homozygous human subjects (CC alleles) exhibit increased plasma TG and LDL-C, which are independent risk factors for cardiovascular disease risk, particularly in uncontrolled diabetes (5, 9, 44). Likewise, female FABP1 gene-ablated mice exhibit a more pronounced phenotype than their male counterparts (26–30). Higher prevalence of atherothrombotic cerebral infarction in metabolic syndrome occurs with FABP1 T94A subjects (67).

The FABP1 SNP may contribute to nonalcoholic fatty liver disease (NAFLD) (44), and fenofibrate is less effective in lowering plasma TG to target levels in subjects with this SNP (5), possibly contributing to inconclusive results of fibrate therapy for NAFLD (58). Overexpression of human wild-type FABP1 T94T, but not human FABP1 T94A, in human Chang liver cells increased LCFA uptake and TG mass compared with either the vector-transfected or nontransfected cells (12). However, the relevance of these findings with transformed cells to human liver hepatocytes is unclear in light of the observation that Chang liver cells are not of hepatic origin but instead are derived from human cervical cancer cells (36).

This underscored the need to resolve molecular details of FABP1 T94A effects on LCFA uptake and lipid accumulation directly in human hepatocytes. Because human and murine FABP1 bind their respective PPAR-α (16, 18, 51, 54, 60) to facilitate ligand-mediated PPAR-α transcription of LCFA β-oxidative enzymes (20, 45, 46), the human FABP1 SNP may also impact the FABP1/PPAR-α signaling pathway. Therefore, cultured primary human hepatocytes from female subjects possessing the FABP1 SNP TT, TC, and CC alleles were used to determine its impact on LCFA uptake/metabolism, fibrate-induced FABP1 targeting to nuclei, and ligand-mediated PPAR-α transcription of β-oxidative enzymes.

MATERIALS AND METHODS

Materials.

Fenofibrate, stearic acid (C18:0), eicosapentaenoic acid (EPA, C20:5n-3), docosahexaenoic acid (DHA, C22:6n-3), fatty acid-free albumin fraction V (10% solution for tissue culture), insulin from bovine pancreas, and dexamethasone were from Sigma-Aldrich (St. Louis, MO). Collagenase B was from Roche, (Life Technologies, Carlsbad, CA). RNEasy kit and RNase-Free DNase set were from Qiagen Sciences (Valencia, CA) and Qiagen (Hilden, Germany), respectively. TaqMan, One-Step RT-PCR Master Mix reagents, TaqMan Gene Expression Assays for mRNAs encoding the following human proteins were purchased from Life Technologies (Carlsbad, CA): FABP1 (FABP1, Hs00155026_m1), fatty acid transport protein-5 (FATP5) (SLC27A5, Hs0020073_m1), fatty acid transport protein-2 (FATP2) (SLC27A2, Hs00186324), carnitine palmitoyltransferase 1A (liver) (CPT1A) (CPT1A, Hs00912671_m1), carnitine palmitoyltransferase 2 (CPT2) (CPT2, Hs00988962_m1), acyl-coenzyme A oxidase 1, palmitoyl (ACOX1) (ACOX1, Hs01074241_m1), PPAR-γ coactivator 1α (PGC-1α) (PPARGC1A, Hs01016719_m1), CCAAT/enhancer-binding protein α (C/EBP-α) (CEBPA, Hs00269972_s1), forkhead box protein A1 (FOXA1) (FOXA1, Hs00270129_m1), hepatocyte nuclear factor 1 homeobox A (HNF1-α) (HNF1A, Hs00167041_m1), hepatocyte nuclear factor 4α (HNF-4α) (HNF4A, Hs00230853_m1), PPAR-α (PPARA, Hs00947539), human acetyl-CoA carboxylase-α (ACACA, Hs01046047_m1), acetyl-CoA carboxylase-β (ACACB, Hs00153715_m1), fatty acid synthase (FASN, Hs01005622_m1), monoacylglycerol O-acyltransferase 2 (MOGAT2, Hs00228268_m1), monoacylglycerol O-acyltransferase 3 (MOGAT3, Hs00228268_m1), mitochondrial glycerol-3-phosphate acyltransferase (GPAM, Hs01573680_m1), patatin-like phospholipase domain containing 3 (PNPLA3, Hs00228747_m1), 1-acylglycerol-3-phosphate O-acyltransferase 2 (AGPAT2, Hs00944961_m1), lipin 2 (LPIN2, Hs00206237_m1), diacylglycerol O-acyltransferase 2 (DGAT2, Hs01045913_m1), microsomal triglyceride transfer protein (MTTP, Hs00165177_m1), and apolipoprotein B (APOB, Hs01071209_m1). [9,10-³H]-stearic acid (1 mCi/ml in EtOH) was from Moravek Biochemicals (Brea, CA). For Western blotting, rabbit polyclonal antibody against human FABP1, CPT1A, CPT2, FATP2, FATP4, and FATP5 was from Santa Cruz Biotechnology (Dallas, TX). Nunc LabTek chambered coverglass was from VWR (Radnor, PA). 7-Nitrobenz-2-oxa-1,3-diazol aminostearate (NBD)-stearic acid was custom synthesized by Avanti Polar Lipids (Alabaster, AL). 1-Anilinonaphthalene-8-sulfonic acid, TO-PRO-3 monomeric cyanine nucleic acid state, and SlowFade reagent were from Life Technologies (Grand Island, NY). Goat anti-rabbit IgG conjugated to FITC was from Sigma-Aldrich (St. Louis, MO). LR White resin and donkey anti-rabbit IgG conjugated to 15-nm gold were from Electron Microscopy Sciences (Hatfield, PA). The following culture medium products were purchased from Life Technologies: Williams Medium E (1×, no phenol red), Cryopreserved Hepatocytes Recovery Medium (CHRM), Hepatocyte Plating Supplement Pack (serum-containing), and Hepatocyte Maintenance Supplement Pack (serum-free). BD Gentest High Viability CryoHepatocyte Recovery Medium and BD Gentest CryoHepatocyte Plating Medium were from BD Biosciences (San Jose, CA).

Recombinant human FABP1 T94T and FABP1 T94A proteins and binding.

The cDNA encoding the human FABP1 obtained from OriGene was sequenced (Research Technology Support Facility, Michigan State University, East Lansing, MI) and shown to code for the human FABP1 T94A mutant. Recombinant human FABP1 T94A variant was prepared from this cDNA coding for human FABP1 T94A variant (OriGene, Rockville, MD) with standard procedures in our laboratories (10, 26), whereas the recombinant human FABP1 T94T was prepared from the cDNA coding for human FABP1 T94A variant (OriGene, Rockville, MD) using standard mutagenesis procedures in our laboratories (10, 26). Identity and purity were confirmed by sequencing (Gene Technology Laboratory, Texas A & M University) and MS/proteomics (LBMS Laboratory, Texas A & M University) as we have shown (14, 32, 41, 47). Binding of fenofibrate and fatty acids to recombinant human T94T and T94A was determined by aminonaptholsulfonic acid displacement as described earlier (60).

Human hepatocyte culture and genotyping.

Multiple vials of cryopreserved plateable human primary hepatocytes from 30 different lots were purchased from Life Technologies and BD Biosciences and shipped using cryo shippers containing adsorbed liquid nitrogen. The vials, labeled by lot number, containing the hepatocytes were stored in a −135°C freezer. These lots represent cryopreserved plateable hepatocytes isolated from livers of female (49.5 ± 2.7 yr) Caucasian subjects by the respective companies. Personnel working with the purchased human cryopreserved hepatocytes received BL1 and BL2 training and certification for work with human hepatocytes in our approved laboratory (Institutional Biosafety Committee, Office of Research Compliance and Biosafety, Division of Research, Texas A & M University).

Hepatocytes were thawed and cultured overnight according to the manufacturer's instructions using the manufacturer's specified medium for each step: 1) vials containing cryopreserved human hepatocytes were taken out of the −135°C freezer and immediately put in a 37°C water bath for less than 2 min to thaw; 2) the outside of each vial was wiped with 70% alcohol, and then the contents of each vial were transferred either into 1 tube (50 ml) of CHRM Medium (Life Technologies) or BD Gentest High Viability CryoHepatocyte Recovery Medium (BD Biosciences) per manufacturer's instructions; 3) tubes were centrifuged at room temperature 100 g for 10 min, and the supernatant was carefully removed and discarded; 4) plating medium (∼1 ml) per 1 × 10⁶ total cells was added to the cells (cells were counted using a hemocytometer), and the viability was determined in PBS buffer containing Trypan blue with a final concentration 0.1%. After overnight culturing, the medium was removed, and the hepatocytes were incubated for 1 h, using glucose-free Williams medium E. Further incubation was for 24 h with serum-free, Williams medium E (6 mM glucose) with 100 nM insulin, 10 nM dexamethasone, and 40 μM fatty acid-free albumin (Alb) or 40 μM Alb complexed with 40 μM fenofibrate, 200 μM EPA (C20:5n-3) or 200 μM DHA (C22:6n-3) as described earlier (45). Hepatocytes were washed and prepared for RNA isolation or Western blotting. Based on stability of marker proteins determined by Western blotting (see below), all experiments were performed with human primary hepatocytes maintained in culture ≤2 days.

DNA isolated from a small sample of a vial of the cryopreserved plateable human hepatocytes from each of the 30 different lots were genotyped using the genotyping assay ID C_25473098_10 (Life Technologies) for the FABP1 T/C rs2241883 SNP. 13 lots were associated with FABP1 TT, 14 lots with TC, and 3 lots with CC SNP alleles, thus determining whether expression is of wild-type FABP1 T94T, both wild-type FABP1 T94T and FABP1 T94A variant (heterozygous), and FABP1 T94A variant as described (9).

RNA isolation and gene expression analysis by quantitative real-time PCR.

Total mRNA was collected from cultured primary human hepatocytes using the RNeasy mini kit from Qiagen Sciences according to the manufacturer's instructions. mRNA concentration was determined spectrophotometrically and used to measure the relative level of human mRNA expression for translation. Quantitative real-time PCR (qRT-PCR) was performed with an ABI PRISM 7000 Sequence Detection System (SDS) from Applied Biosystems (Foster City, CA) with thermal protocol: 48°C for 30 min, 95°C for 10 min before the first cycle, 95°C for 15 s, and 60°C for 60 s, repeated 40 times. Each was performed in triplicate followed by analysis with ABI PRISM 7000 SDS software (Applied Biosystems) to determine ΔCt relative to 18S as a housekeeping gene positive control. The relative abundance of human mRNAs relating to the expression of the proteins CPT1A, CPT2, and ACOX1 was calculated for each albumin/lipidic ligand (fenofibrate, EPA, DHA) treatment compared with treatment with albumin only. The comparative 2^−ΔΔCt calculation method was used as described in User Bulletin 2, ABI Prism 7000 SDS (Applied Biosystems) and earlier (20, 45, 46).

Western blotting.

Primary human hepatocytes, cultured overnight as above, were homogenized, and protein was determined using a Bio-Rad Protein Assay (Bio-Rad Laboratories, Hercules, CA) as described previously (56). Each was then subjected to SDS-PAGE gel electrophoresis and Western blotting with rabbit anti-human FATP5, FATP4, FATP2, sterol carrier protein-2 (SCP-2), and SCP-x, as well as anti-pig glutamate oxaloacetate transaminase/aspartate aminotransferase (GOT) and anti-rat wild-type FABP1 with antibodies against either of the housekeeping proteins GAPDH and β-actin as described previously (56). Human hepatocytes maintained stable expression of multiple markers for up to 2 days in culture as determined by Western blotting of FATP5, FATP2, FATP4, mitochondrial GOT, CPT1A, CPT2, and SCP-2 (not shown). Analogous protein marker studies showed that primary mouse hepatocytes maintained marker expression constant for even longer time in culture, 3–4 days (56). Therefore, all studies were performed with primary human hepatocytes maintained in culture ≤2 days.

Impact of fatty acid load.

Primary human hepatocytes were cultured overnight in the maintenance medium. Medium was then removed, replaced with serum-free Williams medium E supplemented with 100 nM insulin, 10 nM dexamethasone, and either 40 μM fatty acid-free BSA or 40 μM fatty acid-free BSA plus 200 μM oleic acid/palmitic acid in 2:1 molar ratio, followed by incubation for 24 h. The choice of 200 μM oleic acid/palmitic acid bound to albumin was based on the following: 1) serum concentrations of unesterified LCFAs range from 200–600 μM under normal conditions, nearly 7,000-fold higher than LCFA monomeric solubility (37); 2) serum albumin concentrations, 624–789 μM, are in the same range as normal unesterified LCFA levels, such that >99% of serum unesterified LCFA are albumin bound (37); 3) to avoid potential toxicity of micellar unesterified LCFA, the LCFAs are commonly solubilized for delivery to cells at concentration comparable to that in serum, i.e., 200 μM (45, 64); 4) oleic acid and palmitic acid are the two most common fatty acids found in serum (3). Although cellular uptake/efflux of unesterified fatty acid is bidirectional, LCFA-free albumin was chosen as a control because 1) bidirectional LCFA translocation across the plasma membrane is protein facilitated (FATPs, GOT) but not energy driven (25); 2) L-FABP is present at high concentrations (200–400 μM) in hepatocyte cytosol, where it facilitates LCFA desorption from the plasma membrane, accounting for 90% of liver cytosolic LCFA binding capacity (31, 37); 3) once taken up, unesterified LCFAs are rapidly (half-time of a few minutes) esterified or oxidized, processes facilitated by l-FABP, thereby driving net uptake (1, 37); 4) LCFA-free BSA was used as a control in previous studies of BSA/LCFA complex supplementation (45, 64); 5) results in Table 1 indicate that incubating TT, TC, and CC variant hepatocytes for increasing time in medium containing fatty acid-free BSA did not significantly alter TG mass.

Table 1.

Basal levels of TG in cryopreserved plateable human hepatocytes

Genotype Group	6 h	12 h	24 h
TT	197 ± 32	183 ± 5	203 ± 13
TC	564 ± 60	600 ± 60	556 ± 28
CC	301 ± 45	290 ± 24	264 ± 16

Values are means ± SE in nmol/mg protein; replicates = 8. Hepatocytes were cultured in lipid-free BSA only (without addition of fatty acids), and the triglycerides (TG) were measured at 6, 12, and 24 h using a fluorescent assay kit and presented as a function of genotype (see materials and methods).

Lipid composition.

For TG mass determinations, hepatocytes were cultured in 96-well plates with BSA and BSA/LCFA complexes as above. The highly sensitive fluorescence detection adipogenesis assay kit (BioVision, Milpitas, CA) was used to determine TG mass in the hepatocytes after PBS wash. For determination of secreted ApoB and other lipids in hepatocytes, cells were cultured in 24-well plates with BSA and BSA/LCFA complexes; medium was removed, and hepatocytes were washed with PBS. Secreted ApoB in medium was determined by human ApoB ELISA PRO (horseradish peroxidase) kit (Mabtech, Cincinnati, OH). Lipids in the medium (for secreted TG analysis) and in the hepatocyte [for cholesterol, cholesteryl ester (CE), free fatty acids (FFA), and phospholipid (PL) analysis] were then extracted with chloroform/methanol (2:1), the organic phase separated and dried with N₂, and dried lipids were solubilized with detergent buffer (PBS with 0.55% IGEPAL). Lipid components were then determined in aliquots of solubilized lipids using highly sensitive fluorescence detection kits (using PBS with 0.55% IGEPAL detergent buffer as blank) as follows: 1) PL was determined using the EnzyChrom PL assay kit (BioAssay Systems, Hayward, CA); 2) cholesterol and CE were determined using the Amplex Red cholesterol assay kit (Life Technologies); 3) the amount of unesterified FFA was determined using EnzyChrom FFA assay kit (BioAssay Systems); 4) quantitation of TG in the medium was determined with the aforementioned kit.

Laser scanning confocal microscopy: NBD-stearic acid uptake.

Human hepatocytes were plated at a density 1 × 10⁵ cells per well on collagen-coated four-well Lab-Tek chambered coverglass (VWR). After overnight culturing in the maintenance medium, the hepatocytes were washed three times with warm (37°C) PBS buffer and transferred to a 37°C heated microscope stage on the on a MRC-1024MP imaging system (Carl Zeiss MicroImaging, Thornwood, NY). After addition of 200 nM of NBD-stearic acid, the cells were imaged at 1-min intervals as described earlier (56). Excitation was by the 488-nm line of the Ar/Kr+ laser and the emission imaged using a HQ530/40 band pass filter (Chroma Technology, Bellow Falls, VT) and a ×40 oil Zeiss Apochromat objective (Carl Zeiss MicroImaging). Images were analyzed in MetaMorph Image Analysis Software (Molecular Devices, Sunnyvale, CA) to obtain mean grayscale intensity/cell. Cells were averaged and the means ± SE plotted as a function of time using SigmaPlot (Systat Software, San Jose, CA). Nonlinear regression was used to fit the curves in SigmaPlot (Systat Software, San Jose, CA) to the Wiebull equation of the form t = A [1 − e^−(t/b)c], where f(t) is the intensity as a function of time, A is the maximal intensity, 1/b is an apparent rate constant, t is time (min), and c is a shape parameter (c > 1 for sigmoidal curves) as described previously (56).

Laser scanning confocal microscopy: FABP1 T94T nuclear distribution.

TT genotyped human hepatocytes were plated in LabTek chamber slides (VWR), incubated with glucose and BSA ± fenofibrate as above for 0.5–24 h, washed, fixed, immunolabeled with rabbit anti-rat FABP1 polyclonal serum for 1 h, and then FITC-conjugated in goat anti-rabbit IgG for 1 h. After fixation, counterstaining was for 30 min with 1 μM of TO-PRO-3 nucleic acid stain (Life Technologies). An MRC-1024 laser scanning confocal microscope (LSCM) system (Carl Zeiss MicroImaging) simultaneously detected FITC-FABP1 and TO-PRO-3 using 540/30 and 680/32 emission filters with excitation at 488 nm and 647 nm, respectively. Analysis yielded nuclear:cytoplasmic ratios as a function of incubation time with BSA or BSA/fenofibrate.

Fluorescence microplate detection: NBD-stearic acid uptake.

Human hepatocytes cultured in 96-well plates (4,000 cells per well) were washed twice with PBS. Measurement of the fluorescence intensity was started immediately after addition of 500 nM NBD-stearate in PBS buffer. The wells with hepatocytes plus PBS buffer without NBD-stearate were used to control for scatter while wells without hepatocytes plus 500 nM NBD-stearate were used to control for photobleaching. A Synergy 2 microplate reader (BioTek Instruments, Winooski, VT) was used to record NBD-stearic acid fluorescence intensity every 2 min under the following conditions: excitation filter, 460/40; emission filter, 540/35; optics position, bottom; light source, tungsten; read type, endpoint; read speed, normal; temperature, 37°C.

[9,10-³H]-stearic acid uptake and β-oxidation.

Human hepatocytes were diluted to 0.5 × 10⁶ live cells per well on six-well dishes with plating medium and incubated at 37°C for 4–6 h. Plates were shaken gently to loosen debris, and the plating medium was removed. Maintenance medium was added, and cells were cultured overnight in the 37°C incubator. Fatty-acid free BSA (40 μM) or 200 μM stearic acid/40 μM BSA and 0.01 nmol/well [9,10-³H]-stearate (1 nmol/63.5 μCi stearic acid [9,10-³H], Moravek Biochemicals, Brea, CA) were used for an additional 18 h at 37°C. [9,10-³H]-stearic acid uptake and oxidation were then quantitated as previously described for [9,10-³H]-stearic acid and other radiolabeled fatty acids with mouse hepatocytes (2, 45, 56). Briefly, medium was removed, the cells were rinsed twice with PBS buffer to remove excess probe, and the removed medium was pooled with PBS washes and saved. The multiwell culture plates were floated on liquid nitrogen, and cells were scraped from individual wells in hexane:isopropanol (3:2, vol/vol). Lipids were extracted from each medium/PBS wash pool and from cell samples with hexane:isopropanol (3:2, vol/vol) (3 × 1,500 rpm for 10 min), organic phase lipid fractions were pooled, aqueous fractions were pooled, and radioactivity in each was quantified by scintillation counting. Uptake and oxidation values were calculated as the combined ³H disintegrations per minute of the cellular lipid fraction and medium/wash aqueous fraction. The specific activity of the [9,10-³H]-stearic acid probe and the cold-labeled ratio were used to calculate the total nanomoles of stearic acid taken up and oxidized, which was then expressed on the basis of sample total cell protein quantified by the Bradford assay.

Immunogold electron microscopy.

Cultured primary human hepatocytes expressing wild-type FABP1 or FABP1 T94A variant (genotyped as described above) were plated in triplicate on slides for electron microscopy as described earlier (16, 18). After incubation with BSA or BSA/fenofibrate complex as above, hepatocytes were washed three times with HBSS, fixed with 4% formaldehyde, 0.1% glutamine in 0.1 M sodium phosphate buffer (pH 7.4) for 20 h at 4°C, then washed with 0.1 M sodium phosphate, dehydrated in an ethanol series, and embedded in LR White resin at 48°C for 2 days. Ultrathin sections (60 to 80 nm) were placed on Formvar-coated nickel grids, labeled with primary antibody (i.e., rabbit anti-rat FABP1 polyclonal serum) for 1 h, followed by secondary donkey anti-rabbit IgG conjugated to 15-nm gold (18). Primary human hepatocytes without primary antibodies were used as controls. Sections were poststained with aqueous 2% uranyl acetate and Reynold's lead citrate and imaged with a Zeiss 10c TEM (Carl Zeiss Microimaging). Anti-FABP1 labeling particle density was determined in nucleoplasm and cytoplasm and statistically analyzed as described earlier (18). Gold particles in cytoplasm and nucleoplasm were identified and manually counted using MetaMorph Image Analysis Software (Molecular Devices).

Statistics.

Statistical analysis was performed using GraphPad Prism (La Jolla, CA) by one-way ANOVA with Newman-Keuls posttest. The threshold chosen for statistical significance was P < 0.05.

RESULTS

Genotyping of cryopreserved plateable human primary hepatocytes.

Genotyping of the hepatocytes from 30 different lots with each lot representing a different female Caucasian human subject (see materials and methods) revealed a MAF of 0.317 for the FABP1 rs2241883 SNP with the homozygous mutant at 10% of the total. The MAF, C, was within the range of that reported in literature for Caucasian populations (8.3 ± 1.9% homozygous variant; MAF = 0.264 for 1,000 genomes in NCBI dbSNP database; ALFRED database) (5, 9, 44).

Impact on neutral lipid content.

Genotyped primary hepatocytes were plated and cultured with low or high LCFA for determination of the two key neutral storage lipids (TG; CE) as in materials and methods. In basal medium (lipid-free BSA without LCFA), the expression of the FABP1 T94A variant (C allele) significantly increased the cellular mass of TG (Fig. 1A) at all incubation times examined (Table 1). As expected, human hepatocytes had much lower level of CE (Fig. 1C) than TG (Fig. 1A). However, FABP1 TC and CC variant expression increased CE mass (Fig. 1C). TG levels were not significantly altered over time within each genotype group while culturing the hepatocytes in the basal medium containing lipid-free BSA only (Table 1). Accumulation of TG and CE in FABP1 TC and CC variant hepatocytes was concomitant with 56 ± 4% and 41 ± 3% decreased mRNA for MTTP, respectively (Fig. 1E), suggesting diminishing capacity to load ApoB with TG, especially at high LCFA load. Culturing with a mixture of fatty acids [200 μM oleic acid/palmitic acid (2:1 molar ratio) complexed with BSA] increased cellular TG mass in FABP1 TC variant even more than in CC variants (Fig. 1A), whereas high LCFA load did not increase CE mass in hepatocytes of either genotype (Fig. 1C).

Fig. 1.

Liver fatty acid-binding protein (FABP1) T94A variant alters cellular lipids. Cultured cryopreserved primary hepatocytes (Life Technologies) from 30 subjects were genotyped to determine which lots were from subjects with the FABP1 T>C single nucleotide polymorphism (SNP) alleles; TT (n = 14), TC (n = 13), or CC (n = 3) as described in materials and methods. The hepatocytes were plated and cultured overnight per manufacturer's protocol, and the hepatocytes were incubated for 24 h with serum-free Williams medium E supplemented with 100 nM insulin, 10 nM dexamethasone, and either 40 μM fatty acid-free BSA (solid bars) or 40 μM fatty acid-free BSA plus 200 μM oleic acid/palmitic acid in 2:1 molar ratio (open bars). Triglycerides (TG) (A), secreted TG (B); cholesteryl ester (CE) (C), secreted apolipoprotein B (ApoB) protein (D), relative abundance of microsomal triglyceride transfer protein (MTTP) (E) and APOB (F) mRNA (normalized to internal control) determined by qRT-PCR and fold-change relative to TT hepatocytes. Phospholipids (PL) (G), free fatty acids (FFA) (H), free cholesterol (Chol) (I), and total Chol (J) were determined as described in materials and methods. mRNA measurements were performed in triplicate with the rest in sextuplicate. Means ± SE, *P < 0.05 vs. TT BSA, #P < 0.05 vs. BSA in each group by 1-way ANOVA and Neuman-Keuls posttest.

Impact on neutral lipid secretion.

Cultured primary human hepatocytes were preloaded with [9,10-³H]-stearic acid, and esterified [9,10-³H]-stearic acid appearing in TG and CE in the culture medium was determined. After 24 h, the culture medium of TT hepatocytes contained 0.14 ± 0.01 nmol/mg cell protein of [9,10-³H]-stearic acid esterified to TGs. Radiolabeled TG appearing in the culture medium was significantly (P < 0.05) increased in CC (0.30 ± 0.09 nmol/mg), but not TC genotypes. After 24-h incubation, the culture medium of TT hepatocytes also contained 0.016 ± 0.01 nmol/mg cell protein of [9,10-³H]-stearic acid esterified to CEs, which was significantly (P < 0.05) increased in the CC variant (0.220 ± 0.030 nmol/mg) but not the TC variant hepatocytes.

TG mass appearing in the culture medium was increased slightly in CC (Fig. 1B) but not TC (Fig. 1B) genotypes. Appearance of ApoB protein [a key apolipoprotein of secreted nascent very-low-density lipoprotein (VLDL)] in the culture medium was increased in both TC and CC genotypes (Fig. 1D), consistent with 2.0 ± 0.1- and 1.6 ± 0.2-fold increased mRNA encoding ApoB in the TC and CC variants, respectively (Fig. 1F). High LCFA load increased secreted TG mass in TT (Fig. 1B) and TC (Fig. 1B) but not in homozygous CC variants (Fig. 1B), whereas ApoB secretion was unchanged (Fig. 1D).

Impact on polar lipid content.

Both the TC and CC genotyped hepatocytes showed significantly increased cellular mass of PL (Fig. 1G). Under high LCFA load (200 μM oleic acid/palmitic acid in a 2:1 molar ratio complexed with BSA), FABP1 T94A expression also increased the cellular mass of PL, more so in the CC genotypes (Fig. 1G). Nevertheless, it is important to note that FABP1 T94A expression increased the mass of PL much less than of TG mass in TC and CC variant hepatocytes. Preferential LCFA targeting toward neutral lipids rather than phospholipids was confirmed with radiotracer distribution. Under high (200 μM) stearic acid load, the incorporation of [9,10-³H]-stearic acid into PL was decreased from 1.90 ± 0.07 nmol/mg protein in the wild-type TT genotyped hepatocytes to 1.02 ± 0.04 and 0.92 ± 0.03 nmol/mg protein in the TC and CC variant hepatocytes, respectively, whereas [9,10-³H]-stearic acid incorporation into neutral lipids was concomitantly increased.

The TC genotyped hepatocytes showed significantly increased cellular mass of free cholesterol (Fig. 1I), total cholesterol (Fig. 1J), and unesterified FFA (Fig. 1H). High LCFA did not further increase but instead decreased free cholesterol mass (Fig. 1I) and total cholesterol mass (Fig. 1J). In contrast, high LCFA load markedly increased unesterified FFA mass in all groups with the highest levels appearing in the CC phenotype hepatocytes (Fig. 1H). [9,10-³H]-stearic acid appearing in PL secreted into the culture medium was decreased from 0.104 ± 0.008 nmol/mg protein in the wild-type TT genotyped hepatocytes to 0.077 ± 0.007 and 0.046 ± 0.004 nmol/mg protein in the TC and CC variant hepatocytes, respectively.

Protein ligand binding affinity.

The altered accumulation and secretion of esterified lipids in T94A expressing human hepatocytes could be due to a total loss of ligand binding ability analogous to FABP1 gene ablation, as has been postulated (12, 44), or due to partially impaired ability to bind ligands. Because single amino acid modifications in or outside the murine and bovine FABP1 binding cavity can increase or decrease LCFA affinity and/or specificity (52, 59, 66), ligand binding assays were performed.

1-Anilinonaphthalene-8-sulfonic acid (ANS) binding detects the higher affinity of the two binding sites of human FABP1 (60). Displacement of ANS revealed that the human T94T FABP1 has high affinity for the saturated LCFA stearic acid, C18:0, a weak PPAR-α activator (46) and lipidic ligands that are potent PPAR-α activators, including very-long-chain polyunsaturated fatty acids (DHA > EPA) and fenofibrate (Table 2). Importantly, affinities for these ligands were not significantly altered (Table 2).

Table 2.

Impact of FABP1 T94A variant on ligand binding

	Human FABP1 Ligand-Binding Affinity, μM
Ligand	T94T	T94A
Stearic acid (C18:0)*	0.023 ± 0.001	0.027 ± 0.001
Eicosapentaenoic acid (EPA, C20:5n-3)*	0.12 ± 0.01	0.18 ± 0.01
Docosahexaenoic acid (DHA, C22:6n-3)*	0.06 ± 0.01	0.05 ± 0.01
Fenofibrate*	0.012 ± 0.001	0.014 ± 0.001

Values are means ± SE, replicates = 3. Ligand binding to recombinant human wild-type liver fatty acid-binding protein (FABP1) T94T protein and FABP1 T94A variant was determined by 1-anilinonaphthalene-8-sulfonic acid (ANS) displacement as described in materials and methods.

^*ANS displacement assay.

Impact on LCFA uptake: fluorescent and radiolabeled stearic acid.

NBD-stearic acid is a poorly metabolizable fluorescent analog of stearic acid (C18:0) in hepatocytes and similarly binds to FABP1 T94T and T94A (not shown). Therefore, the impact of the T94A variant on real-time LCFA uptake was measured by NBD-stearic acid and LSCM, which simultaneously measures the uptake in a small number (n = 20) of individual cells per image field. Expression of the FABP1 T94A significantly decreased NBD-stearic acid uptake (Fig. 2A) evidenced by 46 ± 2% reduction in maximal rate from 0.42 ± 0.01- to 0.78 ± 0.01- and 1.9 ± 0.05-fold longer half-time from 21.4 ± 0.2 to 40 ± 1 min, although maximal uptake was unaffected at longer times.

Fig. 2.

FABP1 T94A decreased 7-nitrobenz-2-oxa-1,3-diazol aminostearate (NBD)-stearic acid uptake and [9,10 ³H]-stearic acid β-oxidation. A–E: genotyped (C represents the FABP1 T94A SNP) human hepatocytes in Fig. 1 were cultured overnight in either chambered coverglass or 96-well plates per manufacturer's protocol described in materials and methods. NBD-stearic acid uptake as measured on chambered coverglass in triplicate by confocal imaging (A) and as measured in octuplicate using 96-well plates using a fluorescence plate reader (B). The data from B were fitted to a 2-component fit to obtain the initial rates (C), maximal intensities (D), and associated half-times (E). F: for radiolabeling, the hepatocytes were cultured in quadruplicate for 24 h with serum-free Williams medium E (6 mM glucose) supplemented with 100 nM insulin, 10 nM dexamethasone, 40 μM fatty acid-free BSA, and 200 μM stearic acid containing trace amount of [9,10-³H]-stearic acid and [9,10-³H]-stearic acid oxidation, and then measured. Stearic acid oxidation as a function of genotype was expressed as nanomoles [9,10-³H]-stearic acid oxidized/mg cell protein. Means ± SE, *#P < 0.05 relative to TT or TC hepatocytes, respectively, by 1-way ANOVA and Neuman-Keuls posttest.

The uptake experiment was repeated with a much larger number of cells using a fluorescence plate reader. FABP1 T94A expression decreased NBD-stearic acid uptake in both TC and CC genotyped hepatocytes (Fig. 2B). Initial rates were significantly decreased by 12% and 14% (P < 0.05), respectively (Fig. 2C). Uptake for each genotype was best fit to two components reflecting rapid (F¹_max, t¹_1/2) and slow (F²_max, t²_1/2) phases (Fig. 2, D and E). TT hepatocytes exhibited a rapid (t¹_1/2 = 2.2 min) component with F¹_max comprising 56% of total uptake as well as a 10-fold slower (t²_1/2 = 22 min) component with F²_max comprising 44% of total uptake (Fig. 2, D and E). In the TC hepatocytes, the fractional contribution of rapid component F¹_max (but not rapid t¹_1/2) while both t²_1/2 and associated F²_max were significantly decreased (Fig. 2, D and E). Finally, in the CC hepatocytes, t¹_1/2, F¹_max, and F²_max (but not t²_1/2) were all decreased by 14%, 27%, and 12%, respectively (Fig. 2, D and E).

The impact of the FABP1 T94A on uptake of a metabolizable LCFA, i.e., [9,10-³H]-stearic acid, was determined to resolve any contribution by intracellular metabolism on uptake. By 24 h of incubation, the uptake of [9,10-³H]-stearic acid in TT, TC, and CC hepatocytes was 10.5 ± 0.6, 7.8 ± 0.3, and 8.2 ± 0.4 nmol/mg protein, respectively. Thus hepatocytes from donors with the C allele showed significantly decreased uptake of the metabolizable [9,10-³H]-stearic acid as determined by one-way ANOVA and Neuman-Keuls posttest (P < 0.05).

Influence on cytosolic LCFA uptake and transport protein levels.

LCFA uptake is regulated by, not only intracellular metabolism, but also PPARα-regulated membrane-bound (FATP2, 5, 4, GOT) and soluble LCFA transport proteins such as FABP1 (1) or SCP2 (40, 57).

FABP1 mRNA levels were increased in both TC (Fig. 3A) and CC (Fig. 3A) compared with TT (Fig. 3A) genotype hepatocytes. Western blot analysis corroborated that the total level of FABP1 protein also increased in both TC (Fig. 3B) and CC (Fig. 3B) genotype hepatocytes. The cytosolic and peroxisomal matrix protein SCP-2 was unaltered in the TC (Fig. 3G) and decreased slightly in the CC (Fig. 3G) genotype hepatocytes. Levels of SCP-x, another SCP-2 gene product in peroxisomal LCFA β-oxidation, were unaltered (Fig. 3H). Transcription of membrane fatty acid transport proteins, which can affect LCFA uptake, such as FATP2, FATP5, FATP4, and GOT (Fig. 3, C–F), showed no significant altered basal levels in CC hepatocytes. Although TC hepatocytes had slightly higher levels of FATP2 (Fig. 3C), this was offset by unaltered FATP5 (Fig. 3D) and slightly decreased levels of FATP4 (Fig. 3E) and GOT (Fig. 3F).

Fig. 3.

Effect of human FABP1 T94A variant on expression of peroxisome proliferator-activated receptor (PPAR-α)-regulated genes involved in hepatic transmembrane long-chain fatty acid (LCFA) uptake and intracellular transport. Genotyped human primary hepatocytes from Fig. 1 were cultured in triplicate for 2 days as described in materials and methods. A: relative abundance of FABP1 mRNA (normalized to internal control) was determined by qRT-PCR, and the fold-change relative to TT hepatocytes was graphed. B: FABP1 protein levels (see materials and methods) were determined by Western blotting using rabbit anti-human FABP1 and anti-GAPDH and normalized to an internal control (GAPDH). Insets = representative blots. mRNA C and D: relative abundance of SLC27A2 (fatty acid transport protein-2, FATP2) and SLC27A5 (FATP5) mRNA (normalized to internal control) was determined by qRT-PCR, and the fold change relative to TT hepatocytes was graphed. E–H: Western blotting to determine relative protein levels of FATP4 (E), glutamate oxaloacetate transaminase (GOT) (F), sterol carrier protein (SCP)-2 (G), and SCP-x (H) together with the internal control (GAPDH) were obtained (see materials and methods). Relative protein was normalized to internal control, and TT was set to 100. Means ± SE, *#P < 0.05 relative to either TT or TC hepatocytes, respectively, by 1-way ANOVA and Neuman-Keuls posttest.

Key transcription factors of FABP1 upregulation.

Human FABP1 gene transcription is regulated by PPAR-α as well as several nuclear transcription factors, most of which increase FABP1 transcription (15). Furthermore, in a positive feedback mechanism, FABP1 directly interacts with PPAR-α to induce transcription of PPAR-α via a DR1 response element in the PPARA promoter (1).

Expression of the T94A increased the levels of PPARA more than twofold (Fig. 4A), did not alter HNF1A (Fig. 4C), and decreased that of FOXA1 (Fig. 4D) with slight decreases in HNF4A (Fig. 4B) and PPARGC1A (Fig. 4E). Conversely, expression of the T94A significantly reduced the mRNA levels of the potential negative regulator CEBPA (Fig. 4F). Although PGC-1α is known to coactivate HNF-4α, PPAR-α, and FOXO1 in other genes, it does not do so in the human FABP1 (35).

Fig. 4.

FABP1 T94A regulatory transcription. Genotyped human hepatocytes from Fig. 1 were cultured overnight, washed, and incubated 24 h with 40 μM LCFA-free BSA in triplicate. Total mRNA was isolated for quantitative (real-time) reverse transcriptase PCR (qRT-PCR) to determine the levels of nuclear receptor mRNAs, which were normalized to an internal control. Values represent the relative fold change compared with the TT group for mRNA levels of PPAR-α (PPARA) (A), hepatocyte nuclear factor 4 homeobox A (HNF4-α) (HNF4A) (B), HNF1A (C), forkhead box protein A1 (FOXA1) (D), PPAR-γ coactivator 1α (PGC-1α) (PPARGC1A) (E), and CCAAT/enhancer-binding protein α (C/EBP-α) (CEBPA) (F). Means ± SE. *#P < 0.05 relative to either TT or TC genotyped hepatocytes, respectively, by 1-way ANOVA and Neuman-Keuls posttest.

Impact on lipid (triacylglyceride) biosynthesis.

The possibility that increased TG accumulation in TC and CC genotype hepatocytes was associated with upregulation of hepatic enzymes in TG biosynthesis was considered. However, the mRNA levels of ACACB and FASN, key hepatic enzymes in de novo LCFA synthesis, were significantly decreased in the CC compared with the TT hepatocytes (Fig. 5A). Similarly, the MGATs (MOGAT2, MOGAT3), key enzymes in TG biosynthesis via the monoacylglycerol pathway, had significant differences (Fig. 5B). mRNA levels of specific enzymes in the predominant sn-glycerol-3-phospahte triacylglycerol biosynthesis pathway of TG synthesis (GPAM, LPIN2, and DGAT2) were significantly decreased in the CC hepatocytes with no significant change in the mRNA levels of AGPAT2. In the TC hepatocytes, mRNA levels of GPAM and LPIN2 were increased. mRNA levels of PNPLA3, which acts primarily as lysophosphatidic acid hydrolase (may also have lipogenic acyltransferase activity in liver), was also decreased significantly in the CC hepatocytes compared with the TT and TC hepatocytes. TG accumulation in TC and CC genotypes was not associated with upregulation of the de novo LCFA synthesis pathway but rather upregulation of two key enzymes in the predominant GPAT lipogenesis pathway (GPAM, LPIN2).

Fig. 5.

Relative mRNA abundance of key enzymes involved in lipid (triacylglyceride) biosynthesis. Genotyped human hepatocytes from Fig. 1 were cultured overnight, washed, and incubated for 24 h with 40 μM LCFA-free BSA in triplicate. After isolation of total mRNA, qRT-PCR was used to determine relative mRNA abundance (normalized to internal control) of key enzymes in fatty acid synthesis and lipogenesis. Values represent the relative fold change compared with the TT group for mRNA levels of key enzymes in fatty acid synthesis, including acetyl-CoA carboxylase-α (ACACA), ACACB, fatty acid synthase (FASN) (A), key enzymes in the triacylglycerol biosynthesis of the monoacylglycerol pathway, including monoacylglycerol O-acyltransferase 2 (MOGAT2) and MOGAT3 (B), and key enzymes in the predominant sn-glycerol-3-phosphate triacylglycerol biosynthesis pathway, including mitochondrial glycerol-3-phosphate acyltransferase (GPAM), patatin-like phospholipase domain containing 3 (PNPLA3), 1-acylglycerol-3-phosphate O-acyltransferase 2 (AGPAT2), lipin 2 (LPIN2), and diacylglycerol O-acyltransferase 2 (DGAT2) (C). *,#P < 0.05 relative to either TT or TC genotyped hepatocytes, respectively, by 1-way ANOVA and Newman-Keuls posttest.

Impact on stearic acid β-oxidation.

[9,10-³H]-stearic acid β-oxidation was found to be reduced in both TC and CC (Fig. 2F) genotype hepatocytes but was not due to decreased transcription of LCFA β-oxidative enzymes. Instead, the mRNA levels of CPT1A (Fig. 6A), the key rate-limiting enzyme of mitochondrial LCFA β-oxidation, as well as CPT2 (Fig. 6C), were significantly increased in both TC and CC genotype hepatocytes. Likewise, expression of ACOX1 (Fig. 6E), the rate-limiting enzyme in peroxisomal LCFA β-oxidation, was also increased in both TC and CC genotype hepatocytes.

Fig. 6.

FABP1 T94A basal and ligand-induced PPAR-α transcription of carnitine palmitoyltransferase 1A (liver) (CPT1), CPT2, and acyl-coenzyme A oxidase 1, palmitoyl (ACOX1). Genotyped human hepatocytes from Fig. 1 were cultured overnight, washed, and incubated for 24 h with 40 μM LCFA-free BSA (Alb) without or with 40 μM fenofibrate (FF), 200 μM eicosapentaenoic acid (EPA) (C20:5n-3), or 200 μM docosahexaenoic acid (DHA) (C22:6n-3) in triplicate as described in materials and methods. qRT-PCR was used to determine mRNA levels (normalized to internal control) for human CPT1 (A and B), CPT2 (C and D), and ACOX1 (E and F). Values represent the fold change in basal mRNA levels (LCFA-free BSA only) relative to TT hepatocytes (A, C, and E) or the fold change induced by ligand (FF, EPA, or DHA)/BSA complex relative to albumin only (B, D, and F). Means ± SE, *#P < 0.05 relative to either TT or TC hepatocytes, respectively, by 1-way ANOVA and Neuman-Keuls posttest.

FABP1 trafficking to hepatocyte nuclei.

The possibility that the increased transcription of PPAR-α-regulated LCFA β-oxidative enzymes in TC and CC genotypes was associated with increased distribution of FABP1 T94A protein to the nucleus was considered. FABP1 distribution to the nucleus of human hepatocytes was examined by immunolabeling LSCM (200 nm resolution) and electron microscopy (subnanometer resolution). Cultured TT or CC genotype hepatocytes were fluorescently immunolabeled to detect FABP1 (green) followed by counterstaining with a nuclear dye (red), which were detected simultaneously through separate photomultipliers. Under basal conditions (albumin), wild-type FABP1 T94T protein was qualitatively localized in both cytoplasm (Fig. 7B, green) and nucleoplasm (Fig. 7B, red). Codistribution of FABP1 T94T in nuclei was shown by superposition of the images and display of only colocalized pixels (Fig. 7B, yellow). Quantitation revealed a nuclear/cytoplasmic distribution of FABP1 T94T in cultured primary human hepatocytes (BSA only) of ∼0.7, regardless of time in culture (Fig. 7C). Higher resolution immunogold EM confirmed this distribution using TT hepatocytes. Quantitative analysis of multiple electron micrographs demonstrated that the nucleus/cytoplasm ratio of FABP1 T94T in TT genotype hepatocytes was ∼0.8 (Fig. 7D) and in the same range as shown earlier for murine FABP1 T94T in cultured primary mouse hepatocytes (18, 45). The ratio of nuclear/cytoplasmic immunogold labeling increased to ∼1.4 in homozygous FABP1 T94A expressing CC genotype hepatocytes (Fig. 7D).

Fig. 7.

FABP1 intracellular nucleus/cytoplasmic distribution. A: impact of the T94A substitution on anti-human FABP1 antibody specificity against recombinant human FABP1 T94A and FABP1 T94T proteins purified as described in materials and methods. Specificity of rabbit anti-mouse (A, left) and rabbit anti-human (A, right) FABP1 was determined by immunoblotting as in materials and methods. No significant differences were noted between the antibodies to either the T94T or the T94A variant FABP1 protein. B and C: primary human hepatocytes genotyped as TT and expressing the wild-type T94T FABP1 from Fig. 1 were cultured overnight followed by incubation in 40 μM LCFA-free BSA (Alb) without or with 40 μM fenofibrate, 200 μM EPA (C20:5n-3), or 200 μM DHA (C22:6n-3) and imaged using LSCM as described in materials and methods. Hepatocytes were fixed, labeled with FITC-anti FABP1 and TO-PRO (nuclear dye), and analyzed in triplicate. B: representative confocal fluorescence images of FABP1 (green, left), nuclei (red, middle), and colocalized pixels (yellow, right) after 0.5 h incubation. C: images of hepatocytes incubated in 40 μM LCFA-free BSA (Alb) without or with 40 μM fenofibrate, 200 μM EPA (C20:5n-3), or 200 μM DHA (C22:6n-3) for 0.5, 1, 4, or 24 h were analyzed with Image J software. FABP1 fluorescence intensity per surface area inside and outside the nucleus was used to calculate the nucleus/cytoplasm ratio. Means ± SE, n = 40; #P < 0.05 compared with Alb at each time point. D: homozygous TT (T94T) and CC (T94A) hepatocytes were cultured overnight followed by incubation for 24 h in medium with 40 μM LCFA-free BSA (Alb) without or with 40 μM fenofibrate in triplicate. Hepatocytes were fixed, embedded, sectioned, anti-FABP1 immunogold (15 nm) labeled, and subsequently imaged using transmission electron microscopy. The ratio of the density of FABP1 labeling particle density (nucleoplasm/cytoplasm) was plotted. Means ± SE, *P < 0.05 vs. TT BSA, #P < 0.05 vs. TT fenofibrate by 1-way ANOVA and Neuman-Keuls posttest.

Effect on ligand-induced PPAR-α transcription of LCFA β-oxidative enzymes.

Fenofibrate is less effective in lowering elevated plasma TG to normal target levels in the T94A expressing subjects (5, 9). Therefore, the impact of the T94A expression on ligand (fenofibrate, VLCn-3PUFA)-induced PPAR-α transcription of key proteins in LCFA β-oxidation (CPT1, CPT2, ACOX1) was examined.

In TT genotype hepatocytes, fenofibrate enhanced CPT1A, CPT2, and ACOX1 mRNAs by 1.7-fold (Fig. 6, A vs. B), 2-fold (Fig. 6, C vs. D), and 2.2-fold (Fig. 6, E vs. F), respectively. In contrast, fenofibrate-mediated transcription of these proteins was significantly reduced in the TC (Fig. 6, B, D, and F) and CC (Fig. 6, B, D, and F) genotypes. Similar effects were observed in human hepatocytes treated with the VLCn-3PUFA inducers of PPAR-α. EPA enhanced transcription of CPT1A and CPT2 (but not ACOX1) mRNAs by 3-fold and 1.2-fold, respectively, TT genotype hepatocytes (Fig. 6, B, D, and F). DHA appeared slightly more potent in enhancing transcription of, not only CPT1A and CPT2, but also ACOX1 mRNAs by 3-fold, 1.7-fold, and 1.3-fold, respectively, in the TT genotype (Fig. 6, B, D, and F). In contrast, EPA and DHA were significantly less effective in stimulating transcription of these enzymes in TC (Fig. 6, B, D, and F) and CC genotype hepatocytes (Fig. 6, B, D, and F).

Ligand-induced FABP1 distribution to hepatocyte nuclei.

Because fibrate induces wild-type FABP1 T94T redistribution in mouse hepatocytes (46), it was important to determine whether the reported resistance of FABP1 T94A subjects to fenofibrate lowering (5, 9) is associated with reduced ability of fenofibrate to redistribute FABP1 T94A variant to hepatocyte nuclei. The specificity of anti-rat and anti-human FABP1 was found to be unaffected by T94A amino acid substitution (Fig. 7A).

Under basal conditions (albumin) human T94T FABP1 was qualitatively localized in both cytoplasm (Fig. 7B, green) and nucleoplasm (Fig. 7B, red). Codistribution of FABP1 in nuclei was shown by superposition and display of only colocalized pixels (Fig. 7B, yellow). Fenofibrate increased the distribution of T94T human FABP1 to the nucleus (Fig. 7, C and D) as observed by immunogold EM and also by LSCM, qualitatively in comparison of colocalized yellow pixels (Fig. 7B) and quantitatively by analysis of multiple images (Fig. 7C). Basically, similar results were obtained with the VLCn-3PUFA ligand activators of PPAR-α, i.e., EPA (Fig. 7C; C20:5n-3) and DHA (Fig. 7C; C22:6n-3). Despite similar fibrate binding affinities between T94T and T94A (Table 2), the fenofibrate did not induce FABP1 T94A redistribution into the nucleus because the relative nuclear/cytoplasmic ratio was similar to albumin only (Fig. 7D).

DISCUSSION

The human FABP1 T94A variant has been associated with a number of clinical disorders including NAFLD, elevated serum TG and LDL cholesterol, and atherothrombotic cerebral infarction (5, 9, 44, 67). Because our knowledge about cellular and molecular defects of liver steatosis is based mainly on rodent models (22), the role of genetic background in human liver diseases such as NAFLD is unclear (13). The work presented herein with cultured human primary hepatocytes from female subjects demonstrated for the first time that expression of the FABP1 T94A variant was associated with 1) upregulation of total FABP1, a key protein stimulating GPAM, the rate-limiting enzyme in lipogenesis; 2) increased mRNA expression of key enzymes in lipogenesis (GPAM, LPIN2) in heterozygotes; 3) decreased mRNA expression of MTTP; 4) increased secretion of ApoB100 but not TG; and 5) decreased LCFA β-oxidation.

First, expression of the FABP1 T94A variant induced TG accumulation, exacerbated by high LCFA load-in cultured primary human hepatocytes. A clinical study involving high hepatic lipid accumulation resulted in the diagnosis of higher NAFLD prevalence in human subjects expressing the human FABP1 T94A variant but was based solely on ultrasound determination (44). Such techniques visualize lipid droplets within liver cytoplasm but do not represent a chemical analysis of the lipids therein (13). Thus our findings of TG accumulation provide a molecular basis for the increased NAFLD detected by ultrasound in human FABP1 T94A variant-expressing subjects. Although it is not clear how the T94A variant substitution increases hepatic TG accumulation, several factors contribute including the following: 1) increased total FABP1 in TC and CC genotype hepatocytes may stimulate enzymes in the TG synthesis pathway. This possibility is based on the ability of murine FABP1 to directly stimulate GPAM, the rate-limiting enzyme in de novo phosphatidic acid and TG synthesis (4, 50). In addition, transcription of GPAM and LPIN (another key enzyme in the TG synthesis pathway) were increased in the heterozygous TC genotype hepatocytes. Elevation of total FABP1 together with GPAM and LPIN2 in the heterozygous TC genotype may contribute to the higher level of TG accumulation therein compared with the CC genotype hepatocytes wherein only total FABP1 was elevated.

Second, expression of the FABP1 T94A variant elicited accumulation of CE in cultured primary human hepatocytes. CE accumulation was not exacerbated by high LCFA load because this requires, not only exogenous LCFA, but also exogenous cholesterol (7). The finding was consistent with increased CE in NAFLD (48). In vitro studies have shown that murine T94T FABP1 stimulates LCFA-CoA utilization by cholesteryl acyltransferase (ACAT), the key enzyme in CE synthesis (7, 42). The elevated level of total FABP1 in the FABP1 T94A variant-expressing hepatocytes would be expected to increase CE synthesis catalyzed by ACAT.

Third, expression of the FABP1 T94A variant elicited less increase in PL mass (compared with larger increase in TG mass) and decreased targeting of [9,10-³H]-stearic acid into PL of TC and CC genotyped human hepatocytes. One potential adverse consequence of such a limitation may be a diminution of newly synthesized PL needed for assembly of TG-rich VLDL. In support of this possibility, TC and CC hepatocytes exhibited 1) decreased [9,10-³H]-stearic acid appearing in PL secreted into the culture medium and 2) secretion of TG-poor VLDL (Fig. 1, D and F). The smaller increase (vs. TG) in basal PL level in TC and CC hepatocytes may be associated with increased FATP2 and 4 expression therein; increased PL was not observed in a liver lipidomic study of NAFLD in the general human population (48). The basis for the discrepancy may be due to differences between liver and cultured hepatocytes. Nevertheless, the finding of increased cholesterol was consistent (48).

Fourth, the T94A substitution did not abolish ligand binding. The human FABP1 T94A variant bound LCFA (stearic acid, EPA, DHA) and fenofibrate with overall similar affinities as the T94T FABP1 (19, 34). In contrast, FABP1 gene ablation in mice results in the complete absence of FABP1 and reduces hepatic cytosol LCFA binding capacity by >80% in mice (31), thereby decreasing TG in mouse livers and primary hepatocytes (2, 33, 43). Thus the FABP1 T94A SNP results in an altered function and not a complete loss-of-function mutation as in the FABP1-null mouse.

Fifth, FABP1 T94A variant decreased hepatic expression of MTTP but not ApoB, which increased. MTTP is the key microsomal protein responsible for loading ApoB with TG. FABP1 T94A expression increased secretion of relatively TG-poor ApoB-rich VLDL, leading to TG accumulation in hepatocytes, especially on high LCFA load, and potentially causing fatty liver disease (24). However, increased secretion of relatively TG-poor VLDL by FABP1 T94A-expressing hepatocytes does not account for the increased total TG observed in blood of human subjects with the FABP1 T94A variant (5, 9, 44), suggesting that in vivo other factors such as VLDL clearance contribute.

Sixth, the FABP1 T94A variant significantly reduced LCFA uptake in heterozygous or homozygous FABP1 T94A variant-expressing human hepatocytes. Reduced LCFA uptake was not due to altered intracellular LCFA metabolism because the uptake of the nonmetabolizable fluorescent NBD-stearic acid was also reduced, similar to the observation in Chang liver cells (12). Hepatic LCFA uptake has been observed to directly correlate with FABP1 protein level in other model systems (31, 37, 56, 65). However, despite increased levels of the FABP1 expression, LCFA uptake was decreased in the FABP1 T94A variant-expressing human hepatocytes and not associated with reduced overall expression of membrane and other intracellular membrane LCFA transport proteins. Thus despite phenotypic similarities in impact on LCFA uptake, neither the FABP1-null mice nor transformed murine or human cell lines (regardless of hepatic or nonhepatic origin) recapitulate the mechanism whereby the FABP1 T94A variant expression reduces LCFA uptake. Because murine FABP1 directly interacts with membrane LCFA transporter (56), the altered structure of the FABP1 T94A protein and conformational response to ligand binding (19, 34) may impair its interaction with membrane LCFA transporters to facilitate LCFA uptake.

Seventh, expression of the FABP1 T94A variant decreased LCFA β-oxidation in cultured human primary hepatocytes, likely contributing to TG and CE accumulation. The decreased LCFA β-oxidation in the FABP1 T94A variant was observed despite increased total FABP1, distribution of FABP1 to nucleus vs. cytoplasm, and expression of LCFA β-oxidative enzymes. Several considerations may account for this conundrum: 1) the FABP1 T94A variant exhibits altered structure and conformational response to ligand binding (19, 34), which may thereby diminish its ability to interact with mitochondrial CPT1A to deliver bound LCFA-CoA for LCFA β-oxidation (17) by altering the secondary structure, conformational stability, and conformational response to ligand binding of this protein (19, 34). Such structural alterations may adversely impact its ability to likewise function with human hepatocyte mitochondrial CPT1A because single amino acid substitutions that alter CPT1A conformation and diminish FABP1 binding (but not ligand binding) to CPT1A result in diminished mitochondrial LCFA β-oxidation (17). 2) Differences in allelic expression are such that the T94T protein level was more highly expressed than the T94A variant protein in the heterozygous FABP1 TC genotype. Detecting quantitative differences by RT-PCR is not possible (L. Cameron, ABTechnicalsupport@lifetech.com) by Western blotting because the available antibodies reacted similarly to both proteins (Fig. 7A). While isolation of the two forms of FABP1 followed by mass spectrometry may quantitatively resolve differential expression, acquisition of larger amounts of human hepatocytes would be required than presently feasible. 3) There is an adaptation of the human primary hepatocytes in culture 2 days after plating. However, CPT1A, CPT2, and other protein levels remained constant in primary human hepatocytes maintained in culture for at least 2 days, similar to that of primary mouse hepatocytes (56).

Eighth, the T94A amino acid substitution diminished the ability of fenofibrate to induce FABP1 T94A variant redistribution from cytoplasm to nuclei for inducing transcription of PPAR-α-regulated genes involved in LCFA β-oxidation (e.g., CPT1A, CPT2, ACOX1). In contrast, PPAR-α ligands (fibrates, VLCn-3PUFA) induced redistribution of human FABP1 T94T into human hepatocyte nuclei similar to murine FABP1 into murine hepatocyte nuclei (45, 46). Both human and murine WT FABP1 are known to directly bind their respective PPAR-α to transfer bound ligand (16, 18, 51, 60) and induce ligand-mediated PPAR-α transcription of genes involved in LCFA uptake (FABP1), intracellular transport (FABP1, DBI), β-oxidation (CPT1A, CPT2, ACOX1), as well as PPARA itself (20, 45, 46). Despite similar binding affinities, circular dichroism has shown differences in altered conformational changes resulting from fibrate binding, thereby providing a potential structural basis to affect PPARA gene regulation (19, 34). These findings could help explain why homozygous FABP1 T94A variant expressing human subjects (especially females) exhibited increased plasma TG (5, 9) and reduced effectiveness of fenofibrate in lowering plasma TG to target levels in individuals with this variant (5). Differences observed in heterozygous individuals, wherein both forms are produced, could result from synergistic protein-FABP1 interactions resulting from these conformational changes. The result could be upregulation of specific metabolic enzymes beyond the singular action of either the T94T or the T94A in homozygous individuals. These effects have typically been attributed to carriers of multiple SNPs (61). Unfortunately, multiple SNPs combined with low sample number makes interpretation of the complex gene interaction difficult.

In summary, despite the genetic complexity and low number of individuals sampled for the homozygous T94A, the cultured primary human hepatocyte model, independent of tissue-specific cross talk or endocrine variables, demonstrated for the first time that the human FABP1 T94A variant induced triglyceride accumulation in hepatocytes consistent with the observed increased incidence of NAFLD in human subjects expressing the FABP1 T94A variant (44). In addition, the FABP1 CC genotype human hepatocytes were much less responsive than wild-type FABP1 TT genotype hepatocytes to fenofibrate-induced FABP1 T94A protein redistribution to the nucleus and ability to activate PPAR-α-regulated genes, consistent with the reduced ability of fenofibrate to lower serum TG to target levels in human subjects expressing this variant (5). These findings suggest that cultured primary human hepatocytes reflect TG accumulation in FABP1 T94A variant human subjects better than other models, such as Chang liver cells or hepatocytes from l-FABP-null mice. Finally, studies of distinct population subgroups with one or more significant SNP variations create opportunities to integrate genomic information with clinical phenotype and function.

GRANTS

This work was supported in part by the USPHS National Institutes of Health DK41402 (F. Schroeder, A. Kier) and DK70965 (B. Atshaves).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

Author contributions: A.L.M., H.H., S.G., B.P.A., A.B.K., and F.S. conception and design of research; A.L.M., H.H., S.M.S., K.K.L., D.L., A.D.P., and S.G. performed experiments; A.L.M., H.H., S.M.S., K.K.L., D.L., A.D.P., S.G., and B.P.A. analyzed data; A.L.M., H.H., S.M.S., B.P.A., A.B.K., and F.S. interpreted results of experiments; A.L.M., H.H., S.M.S., D.L., and A.D.P. prepared figures; A.L.M. drafted manuscript; A.L.M. and F.S. edited and revised manuscript; A.L.M., H.H., S.M.S., K.K.L., D.L., A.D.P., S.G., B.P.A., A.B.K., and F.S. approved final version of manuscript.