Modulators of Hepatic Lipoprotein Metabolism Identified in a Search for Small-Molecule Inducers of Tribbles Pseudokinase 1 Expression

Recent genome wide association studies have linked tribbles pseudokinase 1 (TRIB1) to the risk of coronary artery disease (CAD). Based on the observations that increased expression of TRIB1 reduces secretion of VLDL and is associated with lower plasma levels of LDL cholesterol and triglycerides, higher plasma levels of HDL cholesterol and reduced risk for myocardial infarction, we carried out a high throughput phenotypic screen based on quantitative RT-PCR assay to identify compounds that induce TRIB1 expression in human HepG2 hepatoma cells. In a screen of a collection of diversity-oriented synthesis (DOS)-derived compounds, we identified a series of benzofuran-based compounds that upregulate TRIB1 expression and phenocopy the effects of TRIB1 cDNA overexpression, as they inhibit triglyceride synthesis and apoB secretion in cells. In addition, the compounds downregulate expression of MTTP and APOC3, key components of the lipoprotein assembly pathway. However, CRISPR-Cas9 induced chromosomal disruption of the TRIB1 locus in HepG2 cells, while confirming its regulatory role in lipoprotein metabolism, demonstrated that the effects of benzofurans persist in TRIB1-null cells indicating that TRIB1 is sufficient but not necessary to transmit the effects of the drug. Remarkably, active benzofurans, as well as natural products capable of TRIB1 upregulation, also modulate hepatic cell cholesterol metabolism by elevating the expression of LDLR transcript and LDL receptor protein, while reducing the levels of PCSK9 transcript and secreted PCSK9 protein and stimulating LDL uptake. The effects of benzofurans are not masked by cholesterol depletion and are independent of the SREBP-2 regulatory circuit, indicating that these compounds represent a novel class of chemically tractable small-molecule modulators that shift cellular lipoprotein metabolism in HepG2 cells from lipogenesis to scavenging.


Introduction
Despite widespread use of cholesterol-lowering drugs, cardiovascular disease remains one of the leading causes of death worldwide and there is a need for novel approaches to improve therapies [1]. Epidemiological studies have repeatedly demonstrated that elevated levels of circulating LDL cholesterol (LDL-C) and triglyceride (TG)-rich remnant lipoproteins have strong associations with the development of coronary artery disease (CAD) and myocardial infarction (MI) [2][3][4]. Because 70% of LDL is removed from the circulation by LDL receptor-mediated uptake in the liver, therapeutic strategies that lead to elevated hepatic expression of the LDL receptor gene, LDLR, have proven to be efficacious in lowering LDL-C and provide protection from cardiovascular disease. Statins, through the inhibition of HMG CoA reductase, deplete cholesterol in the ER of hepatic cells, leading to the activation of the SREBP-2-dependent transcriptional program, which includes increased expression of LDLR. Paradoxically, clinical efficacy of statin therapy is limited by the fact that activation of SREBP-2 also leads to increased expression of proprotein convertase subtilisin/kexin type 9 (PCSK9), which acts as a negative regulator of LDL uptake by promoting degradation of LDL receptor. Recent results from clinical trials with anti-PCSK9 monoclonal antibodies suggest that PCSK9 blockade may indeed provide a more efficacious mechanism of elevating LDL receptor levels than traditional inhibition of HMG CoA reductase [5][6][7][8]. Alternative strategies to lower circulating LDL-C include treatments that lower hepatic secretion of very-low-density lipoprotein (VLDL) particles, the precursors of LDL particles, into the bloodstream. Recently introduced examples of such treatments include inhibitors of microsomal triglyceride transfer protein (MTP) and antisense DNA directed against apoB [9,10]. Reduction of VLDL secretion leads to reduction of plasma TG, which is an independent cardiovascular disease risk factor, and its reduction brings benefits beyond lowering LDL-C alone [3,4,11]. Limitations and the side effects of statins as well as the side effects-particularly hepatic fat accumulation and liver toxicity-associated with apoB-directed treatments underscore the critical need for development of new therapeutic strategies to lower LDL-C and TG to prevent MI.
Recent genome-wide association studies (GWAS) have uncovered novel genes associated with CAD and MI and suggest novel approaches to developing improved therapeutics. TRIB1 emerged in several GWAS as a novel cardiovascular locus, where the protective allele is strongly associated with decreased levels of circulating LDL-C and triglycerides (TG), increased levels of high-density lipoprotein (HDL) as well as with reduced incidence of CAD and MI [12,13]. Additional studies in mice confirmed the link between Trib1 and lipid levels and demonstrated that increased expression of Trib1 is protective against the disease [14]. Hepatic overexpression of Trib1 in mice reduced the secretion of VLDL particles from the liver into the bloodstream and, consistent with this observation, overexpression of TRIB1 in human hepatoma cells reduced apoB secretion. The precise molecular mechanism by which overexpression of Trib1 regulates the rate of VLDL particle formation and secretion is not known, although hepatic overexpression of Trib1 in mice correlates with decreased expression of TG biosynthetic genes (Fasn, Scd1, Dgat2) and decreased rate of TG formation, suggesting that reduced TG availability leads to insufficient apoB lipidation, targeting nascent apoB to ER-associated degradation [14]. TRIB1, as other members of tribbles family, is a pseudokinase and is thought to act as an adaptor protein in the MEK/ERK signaling pathways [15]. It regulates a wide array of cellular processes, ranging from cell migration in macrophages and in smooth muscle cells [16], through macrophage polarization [17] and myeloid leukemogenesis [18], to regulation of lipid metabolism in hepatocytes [14]. In addition to plasma lipid levels, the protective TRIB1 allele has been also linked to lower blood levels of liver enzymes, reduced risk of non-alcoholic fatty liver disease and to longer sleep [13,19,20]. SNPs leading to upregulation of TRIB1 likely have very selective effects that are restricted to one gene and it is unlikely that such selectivity could be achieved with drugs. Nonetheless identification of small-molecule upregulators of TRIB1 could potentially open up a path to identification of novel modulators of lipid metabolism and provide new tools for studying TRIB1 regulation. To this end we developed a qRT-PCR screen to identify compounds that can upregulate TRIB1 expression. We chose to screen the Broad Institute small-molecule library that includes 100,000 novel compounds derived from diversityoriented synthesis (DOS), a synthetic strategy to access complex and diverse compounds in an efficient manner [21][22][23][24]. The DOS compounds are enriched in sp 3 carbons and chiral centers leading to more 3-dimensionality compared to flat, achiral compounds often found in commercial libraries. In addition, the DOS compound collection is designed to elucidate initial stereochemical and appendage structure-activity relationships (SAR) from primary and secondary screens [25,26]. Herein, we describe the identification of BRD0418, a DOS molecule that regulates TRIB1 expression. Characterization of BRD0418 revealed a broader profile of gene-expression changes that lead to decreased rate of VLDL production and increased rate of LDL uptake in cells of hepatic origin. This data indicate that treatment with BRD0418 leads to reprogramming of hepatic lipoprotein metabolism from lipogenesis to scavenging.

Synthesis of BRD0418
Detailed description of the synthesis of BRD0418 is provided in Supplemental Information (S1 Protocol. Synthesis of BRD0418).

Compound screening and generation of cDNA
HepG2 cells were plated in white 384 well plates (Corning) in 40 μL of growth media and incubated overnight. The next day 100 nanoliters of compound was transferred to the plates and incubated overnight. Each plate contained 32 wells treated with DMSO to statistically empower hit calling. After 18-22 hr of incubation RNA was harvested and cDNA was synthesized using the Ambion Cells-to-CT reagents (Life Technologies). Specifically, the cells were washed 2x with PBS and then lysed with 10 μL of Cells-to-CT Lysis with DNase reagent and incubated at room temperature for 10 minutes. After the incubation 1 μL of Stop solution was added. Two μL of the RNA lysis was added to 8 μL of Cells-to-CT cDNA reagent in a 384 well plate (Axygen) and incubated as follows: 37°C for 1 hr, 95°C for 5 min. then 4°C until frozen at-80°C.  [27]. For screening the hits were called for individual plates. Compounds that stimulated >2-fold upregulation of TRIB1 over GAPDH expression and had the ΔΔCt ((TRIB1 DMSO Ct-GAPDH DMSO Ct)-(TRIB Cpd Ct-GAPDH Cpd Ct)) Z-score <-2 and GAPDH Ct Z-score between 10 and-10 were considered hits. The Z-scores were calculated using standard deviations obtained for 32 DMSO treated wells present in each plate.

Luminex transcriptional profiling
The expression profiling of 1000 landmark genes selected based on potential to infer expression of remaining gene transcripts [28] was carried out using bead-based detection assay (Luminex, Austin, TX) essentially as described [29]. Briefly, the HepG2 cells were treated with compounds and the cell lysates were obtained as described in the Compound screening section above. The first strand of cDNA was synthesized from mRNA captured in oligo-dT coated turbocapture plates (Qiagen) using M-MLV reverse transcriptase and the selected cDNAs were amplified using gene specific probes by ligation dependent amplification (LDA) followed by detection using Luminex beads conjugated with gene specific oligonucleotide tags [29].

ELISA assays
HepG2 cells were plated in black clear bottom (Costar) 96-well plates at 40,000 cells/well in 100 μL of growth media. On Day 2 media was changed and cells were treated with compounds for 24 hr. After 24 hr medium was removed and used for the PCSK9 (R&D Systems) and apoB (MAb Tech) ELISAs. For the apoB and PCSK9 ELISAs 50 μL and 10 μL of medium were used, respectfully. For the LDLR ELISA (R&D Systems), cells were washed once with PBS (Invitrogen), lysed with 100 μL PBS/0.1% Igepal (Sigma) and frozen at-80°C until analyzed. Ten μL of lysate was used for the analysis.

Cell viability assay
HepG2 cells were plated in white 384 well plates (Corning) at 2000 cells/well in growth medium and incubated overnight at 37°C with 5% CO 2 . The next day cells were treated with 100 nL of compound for 24 hr. Cellular ATP levels were measured using CellTiter-Glo (Promega). For testing effects of compounds on mitochondria-dependent oxidative energy production cells were grown in media containing 10 mM galactose instead of 25 mM glucose as main carbon source [30].

Stable isotope labeling and analysis of TGs in HepG2 cells
The cells were plated at 40,000 cells/well in 96-well plate in growth media, incubated overnight and then treated for 24 hours with compounds as indicated. For stable isotope labeling of TG the cells were washed with PBS, incubated for at least 15 minutes with DMEM supplemented with 25 mM HEPES (pH 7.5) containing 0.1% FAF-BSA and then after media replacement incubated at 37°C with DMEM containing 25 mM HEPES (pH 7.5), 0.1% FAF-BSA and 20 μM 13 C 3 -D 5 -glycerol for indicated periods of time. The lipids were extracted after the media removal by incubating the cells with 150 μl isopropanol at room temperature for 15 min with shaking. The cell extracts (120 μl) were cleared by centrifugation and concentrated 5-fold through drying and resuspension in isopropanol. Liquid chromatography-tandem mass spectrometry (LC-MS) analyses of cell extracts were conducted using an Open Accela 1250 U-HPLC coupled to a Q Exactive hybrid quadrupole orbitrap mass spectrometer (Thermo Fisher Scientific; Waltham, MA). Extracts (10 μL) were injected directly onto a 150 x 3.0 mm Prosphere HP C4 column (Grace, Columbia, MD). The column was eluted isocratically at a flow rate of 350 μL/min with 80% mobile phase A (95:5:0.1 vol/vol/vol 10 mM ammonium acetate/methanol/acetic acid) for 2 minutes followed by a linear gradient to 80% mobile phase B (99.9:0.1 vol/vol methanol/acetic acid) over 1 minute, a linear gradient to 100% mobile phase B over 12 minutes, and then 10 minutes at 100% mobile-phase B. MS analyses were conducted using electrospray ionization in the positive ion mode using full scan analysis at 70,000 resolution and 3 Hz data acquisition rate. Monoisotopic and 13 C 3 -D 5 -TAG LC-MS peaks were integrated using TraceFinder 2.0 (Thermo Fisher Scientific; Waltham, MA).

LDL uptake
HepG2 cells were plated at 3000 cells/well in a 384 well plate (Corning) in growth medium.
After an overnight incubation 10 μL of 5x compound was added and the plates were incubated overnight. The cells were then washed with phenol red free DMEM (Invitrogen) and incubated with 30 μl of BODIPY-FL-LDL (Invitrogen) diluted to final concentration 5 μg protein/mL in phenol red free DMEM for 0.1 to 5 hr. LDL uptake was stopped by fixing the cells with 3% formaldehyde. Nuclei were stained with Hoechst 33342 diluted in PBS for 10 minutes. The cells were then washed with PBS and imaged using an IXM microscope. Image analysis and quantification was performed with MetaXpress software (ver. 3.1.0.97, Molecular Devices) using the Multi Wavelength Cell Scoring application module.

Western blot analysis
HepG2 cells were seeded in 6-well plates at approximately 800,000 cells /well, incubated in growth media at 37°C for 24 hours, further incubated for 20 hours in serum-free DMEM and then treated with EGF (100 ng/mL), OSM (100 ng/mL) or BRD0418 (10 μM) for the indicated times. At the end of treatment, cells were washed with PBS and lysed in RIPA lysis buffer (50mM Tris-HCl, 1% IGEPAL, 0.5% Na-deoxycholate, 0.1% SDS, 150 nM NaCl, 2 mM EDTA, 50 mM NaF) with protease inhibitor cocktail (Roche). Lysates were incubated on ice for 1 hour with frequent agitation, centrifuged at 13,000 rpm for 15 minutes at 4°C and then equal amount of protein from each lysate (* 50 μg) was denatured, subjected to electrophoresis using NuPAGE Novex 4-12% Bis-Tris Gels and transferred to PVDF membranes with iBlot Gel Transfer Device (Invitrogen) according to manufacturer's instructions. The ERK1/2 phosphorylation was detected using rabbit polyclonal antibodies directed against p44/42 MAPK (Erk1/2) and Phospho-p44/42 MAPK (Erk1/2) (Cell Signaling Technology) and the goat antirabbit IgG (H+L) secondary antibodies conjugated with HRP (Thermo Scientific). After standard incubations and washing the membranes were exposed with SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Scientific) and band intensities were calculated using Image J software.

TRIB1 vector transfection
HepG2 cells (600,000 cells/well) were plated in a 6 well plate (Corning) in growth medium overnight at 37°C with 5% CO 2 . A TRIB1 expression vector (Origene) was transfected using Fugene 6 (Roche) in 1 mL of Opti-MEM (Invitrogen) for 6 hr. After 6 hr, 2 mL of growth medium was added to the wells and incubated overnight. After the overnight incubation the media was changed to growth media and supplemented with 500 μg/mL Geneticin (Invitrogen). After 10 days, Geneticin-resistant clones were transferred to a 96-well plate (Costar) for expansion. After expansion into a 48 well plate (Corning) clones were analyzed for TRIB1 mRNA expression by qRT-PCR.

CRISPR-Cas9 induced disruption of chromosomal TRIB1 locus
HepG2 cells were plated in a 6 well plate (400,000 cells/well) in growth medium and after overnight incubation they were transfected using X-tremeGENE HP DNA transfection reagent (Roche) with 1.5 μg CRISPR-Cas9 plasmid (pCMV-Cas9-GFP, Sigma) and 1.5 μg of the repair template DNA fragment. The CRISPR-Cas9 plasmid carried the target sequence 5'-GGATACACGCTTCGGCCTTCTC-3' in the guide RNA gene. The repair template consisted of three segments that were individually PCR amplified using HepG2 genomic DNA as a template for the TRIB1 target flanking segments and pCINeo plasmid as a template for Neo R cassette segment. The sequences of the PCR primers were designed to provide an overlap with the adjacent segments to allow for final PCR extension and amplification of the entire repair template. The sequences of PCR primers are listed in Table 2. All PCR amplification reactions were carried using Q5 High-Fidelity DNA Polymerase (New England Biolabs) and a touchdown amplification protocol. The repair template was cut with XhoI and SmaI and cloned in the pFN28A vector (Promega) cut with XhoI and PmeI. Gel purified XhoI/BbvI 2991 bp restriction fragment released from this plasmid was used for HepG2 transfection. Next day after transfection the media was replaced and supplemented with 500 μg/mL Geneticin (Invitrogen). Geneticin resistant clones were isolated and expanded after 3 weeks of selection. Clones that carry the designed chromosomal disruption were identified using diagnostic PCR reactions as described in the Results section. The sequences of the diagnostic primers T U , T D , N U and N D are listed in Table 2. Genomic DNA for PCR based screening was prepared using QuickExtract reagent (Epicentre) from cells grown in 96-well plate.

High-throughput chemical screen
Building on the observations that linked the increased expression of the TRIB1 gene to improved lipid profile and reduced incidence of CAD and MI, we developed a quantitative RT-PCR-based phenotypic assay to identify compounds that upregulate TRIB1 expression in HepG2 human hepatocellular carcinoma cells. We performed a high-throughput screen using the Broad small-molecule library, which includes 100,000 DOS compounds, testing each compound in duplicate at 12 μM concentration and looking for upregulation of TRIB1 expression in HepG2 cells. We focused on compounds inducing TRIB1 >2-fold relative to GAPDH and compared to DMSO controls, selecting only compounds with no significant effect on the reference gene expression. Initial hits were retested in a dose-response format in two hepatoma cell lines (HepG2 and HuH7) using a different reference gene (B2M) and screened for effects on viability as measured by cellular ATP levels. Confirmed hits with no cell toxicity were further tested for effects on apoB secretion in HepG2 cells. After completing the triaging process and reviewing existing SAR generated from the DOS collection, we identified a benzofuran scaffold [24] (Fig. 1A), which upregulated TRIB1 in both cell lines and was able to reduce apoB secretion in HepG2 cells >30%, thus mimicking the effects of TRIB1 cDNA overexpression [14]. Initial SAR from the benzofuran library ( Fig. 1A) gave us confidence that the observed effect was due to the small molecule and not to an artifact or impurity. Moreover, the enantiomer of BRD0418 was largely inactive suggesting that biological effects of BRD0418 result from selective and specific interactions with a cellular target (Fig. 1B, C). Changes at the aniline site (R 1 ) were mostly tolerated, with both the tertiary amine and sulfonamide showing activity and therefore indicating that none of the particular properties of neither of those substituents is important for producing the phenotype. The amide building block (R 2 ) showed less tolerability to change, as bulky substituents off of the benzyl amine were required for increased potency. Based on the observed potency, we selected BRD0418 (Fig. 1B) for more detailed characterization. BRD0418 phenocopies the effects of TRIB1 overexpression Overexpression of TRIB1 cDNA in HepG2 cells reduced the rate of triglyceride synthesis and downregulated apoB secretion, presumably due to insufficient lipidation of apoB, which leads to targeting of apoB for ER-associated degradation (ERAD) [14]. Treatment of HepG2 cells with BRD0418 for 6 hours resulted on average in 3.5-fold up-regulation of TRIB1 transcript levels, with half-maximal effect achieved at 2.3 μM concentration of the compound (Fig. 1C). BRD0418 inhibited secretion of apoB from HepG2 cells with similar potency, reducing by about 50% the level of apoB100 present in the media over the cells treated with the compound for 24 hours (Fig. 1D). The secretion of transferrin was not affected by BRD0418 treatment, indicating that the compound did not have a general effect on secretion (Fig. 1E). In the same range of doses, BRD0418 had no effect on cell viability as measured by cellular ATP levels (Fig. 1F). Oligomycin A, an inhibitor of OXPHOS at the mitochondrial ATP synthase level, was recently shown to induce TRIB1 mRNA in HepG2 cells through a MAPK dependent mechanism [31]. Since immortalized and highly proliferative cells lines, like HepG2, tend to synthesize majority of its ATP from glycolysis the cellular ATP levels in cells grown in standard glucose-containing media do not change as result of drug induced mitochondrial defects [30]. Therefore in order to determine if benzofurans might be acting through a similar mechanism we also measured the effects of BRD0418 on cellular ATP levels in cells grown in galactosecontaining media, conditions that force cellular ATP production through mitochondrial oxidative phosphorylation. The ATP levels for cells grown in the presence of oligomycin on glucose were inhibited no more than 21%, whereas for cells grown on galactose they were inhibited more than 95%. Treatment with BRD0418 did not affect ATP levels in cells grown under either condition (Fig. 1F) suggesting that it does not act as mitochondrial stressor. Next, we examined the effect of BRD0418 on the rate of triglyceride synthesis by measuring the incorporation of stable isotope-labeled glycerol into triglycerides over 4 hours in cells preincubated for 24 hours with 5 μM compound. BRD0418 treatment significantly reduced formation of several species of triglycerides that were quantified in lipid extracts by mass spectrometry (Fig. 1H, and data not shown). These results indicate that the BRD0418-induced upregulation of TRIB1 expression mimics the effects of TRIB1 cDNA overexpression.

BRD0418 broadly affects lipid metabolic genes
Early in the triage process, we performed an abbreviated transcriptional profiling of 1000 landmark genes (L1000) using Luminex detection [29] on HepG2 cells treated for 6 or 24 hours with selected screening hits. Evaluation of L1000 data not only confirmed by an alternative quantification methodology (Luminex LDA vs. qRT-PCR and quantile normalization vs. single reference gene transcript) the TRIB1 upregulating activity of BRD0418 and other benzofuran hits, but also indicated that TRIB1 upregulators modify gene expression more broadly affecting gene sets involved in signal transduction, in responses to external stimuli and in metabolism, including lipid metabolism (S1 Table. GO terms enrichment analysis). For example, expression of cholesterol biosynthetic genes ACAT2, HMGCR, HMGCS1, FDFT and NSDHL was inhibited more than 3-fold in benzofuran treated cells (Fig. 2A). The expression of TRIB1 was more strongly elevated at 6 hours than at 24 hours, whereas the downregulation of cholesterol biosynthetic genes appeared to be secondary and was apparent only after the longer treatment.
To confirm and to extend the L1000 observations, we measured changes in several key lipid metabolic genes by qRT-PCR. A 24-hour treatment with 25 μM compound led to significant downregulation of cholesterol biosynthetic genes (HMGCR, HMGCS), fatty acid biosynthetic genes (FASN, SCD1) and genes involved in lipoprotein assembly (MTTP, APOC3) (Fig. 2B). The finding that BRD0418 reduces expression of APOC3 is consistent with TRIB1-driven inhibition of lipoprotein assembly and apoB secretion, and was further confirmed in the dose-response experiments that showed potent downregulation of APOC3 transcript (data not shown) and apoCIII protein secretion in HepG2 cells treated with BRD0418 (Fig. 1F). Notably, BRD0418 treatment led to decoupled changes in the expression of two SREBP-2 co-regulated genes directly involved in LDL uptake, namely to upregulation of LDLR and downregulation of PCSK9 levels (Fig. 2B). The dose-response experiment demonstrated that the changes in expression of LDLR and PCSK9, measured 6 hours and 24 hours post-treatment, respectively, are produced by the similar dose of the compound as upregulation of TRIB1 (Fig. 2C). Moreover, changes in gene expression were mirrored closely in changes in protein levels, leading to 4-fold increase of the LDL receptor in cells and to 2-fold decrease of secreted PCSK9 in the media (Fig. 2D).
Interestingly, we found in the L1000 data analysis that the changes in TRIB1 and cholesterol metabolism genes in HepG2 cells treated with BRD0418 were strongly inversely correlated with the responses to MEK inhibitors, U0126 and PD98059, which were included in the analysis since they were identified in the primary screen as inhibitors of TRIB1 expression ( Fig. 2A).

Small-molecule TRIB1 inducers
The inverse correlation of transcriptional responses to benzofuran compounds and to MEK inhibitors was further highlighted by marker selection analysis, which ranks genes according to difference in response between the two groups (Fig. 3). The L1000 gene expression data showed that the response to benzofurans was rather broad and included a number of early response genes (e.g. EGR1, JUN, IER3, ZFP36, SERPINE1) suggesting activation of MAPK signaling pathways (Fig. 3).

BRD0418 stimulates LDL uptake
Since treatment of HepG2 cells with BRD0418 increases the levels of LDL receptor in cell lysates and decreases the levels of extracellular PCSK9 protein, we evaluated the effect of BRD0418 on cellular LDL uptake. We quantified the uptake of BODIPY-FL-conjugated human LDL at various time intervals in cells pretreated for 20 hours with BRD0418. Treatment with BRD0418 resulted in about 5-fold increase in the rate of labeled LDL uptake (Fig. 4A). In comparison, cholesterol depletion also resulted in about 5-fold increase in the rate of labeled LDL uptake, although the stimulation of uptake by BRD0418 in 10% serum was about 5-fold higher than stimulation achieved through cholesterol starvation in 0.5% lipid-deprived serum (Fig. 4B). BRD0418 produced the half-maximal effect on LDL uptake at 3.3 μM (Fig. 4C), which corresponded to its effects on LDLR and PCSK9 transcript expression.

TRIB1 is upregulated by oncostatin M
Since, in HepG2 cells treated with BRD0418, the change in LDLR expression appears to be decoupled from changes in expression of PCSK9 and cholesterol metabolic genes, we sought to more closely characterize the gene-expression changes in a time-course experiment. In contrast to atorvastatin, which induced a nearly 2-fold increase of LDLR and PCSK9 mRNA levels and did not significantly affect expression of TRIB1 (Fig. 5A), BRD0418 robustly upregulated TRIB1 and LDLR expression but strongly reduced PCSK9, MTTP and APOC3 expression (Fig. 5B). Maximal increase in LDLR and TRIB1 levels occurred after 6-8 hours, whereas the decrease in PCSK9, MTTP and APOC3 levels was a late response and occurred after 24 hours of treatment. We also tested the effects of oncostatin M (OSM), as it was previously demonstrated to upregulate LDLR and downregulate PCSK9 expression in HepG2 cells [32]. Interestingly, OSM strongly induced the expression of TRIB1, and resulted in a pattern of changes resembling the responses to BRD0418 (Fig. 5C), except that the peak responses occurred after one hour rather than six hours.

Effects of BRD0418 treatment and cholesterol depletion are additive
The decoupling of LDLR from PCSK9 responses to BRD0418 treatment suggested an SREBP-2-independent mechanism of regulation. To verify this supposition, we tested the effects of BRD0418 on HepG2 cells under conditions of cholesterol depletion. Expectedly, after a 24-hour incubation in lipoprotein-depleted media, the levels of LDLR and PCSK9 transcripts were increased 3-and 9-fold, respectively, compared to cholesterol-supplemented media (Fig. 6A). TRIB1 levels were not regulated by cholesterol, and remained the same under lowand high-cholesterol growth conditions. Treatment of cholesterol starved cells with 5 μM BRD0418 elevated LDLR an additional 4-fold to reach a combined 12-fold increase, but decreased PCSK9 transcript level by 40%. This result demonstrates that the effects of BRD0418 are additive with the effects of cholesterol depletion. We then sought to determine whether the effects of BRD0418 were additive with the effects of pharmacologically induced cholesterol depletion. Combination of overnight treatment with a statin with a subsequent treatment with BRD0418 had an additive effect on the induction of LDLR expression (Fig. 6B) and led to Small-molecule TRIB1 inducers significantly reduced expression of PCSK9 (Fig. 6B). These results further support the conclusion that BRD0418 acts through a SREBP-2-independent mechanism.
Limited effect of BRD0418 on non-hepatic cells TRIB1 is ubiquitously expressed in most tissues with high levels of expression found in lungs, lymphocytes, pancreas, liver and placenta [14]. To begin address tissue selectivity of TRIB1 induction by BRD0418 we tested the levels of TRIB1 and LDLR transcripts in twelve cell lines originating from various tissues (Table 3). Although we observed induction of TRIB1 in a muscle and lung cell lines the strongest and the most consistent effects on both genes were found in cells of hepatic origin.

Effects of BRD0418 are MAPK-dependent
The primary screening results and the subsequent L1000 analysis indicated that the responses of HepG2 cells to BRD0418 are in many aspects opposite to the responses to MEK1/2 inhibitors, suggesting possible involvement of the MAP kinase signaling pathway in mediating BRD0418 activity. To begin mapping the interaction of BRD0148 with signal transduction pathways we used two highly specific kinase inhibitors, CP-69550, a JAK inhibitor, which blocks receptor proximal signaling and U0126, a MEK inhibitor, which blocks MAPK Erk signaling cascade. First, we evaluated effects of CP-690550 and U0126 on the stimulating activities of BRD0418 and OSM after short duration of treatment. As reported previously, the LDLR upregulating activity of OSM was both JAK-and MEK-dependent, whereas activity of BRD0418 was inhibited by the MEK inhibitor, but not by the JAK inhibitor (Fig. 7A left panel). Upregulation of TRIB1 expression showed similar sensitivity to U0126 (Fig. 7A right panel). In addition, the basal level of TRIB1 in HepG2 cells was reduced by U0126 treatment indicating tight regulation of TRIB1 expression by MAPK Erk signaling. We also examined the effects of MAPK Erk blockade on late responses after 24h treatment with BRD0418. Similar as in case of stimulation of TRIB1 and LDLR the downregulation of PCSK9, MTTP, APOC3 and HMGCR transcripts was also reversed in the presence of U0126 pretreatment (Fig. 7B). Notably, most of the genes tested-with the exception of SCD1 and FASN, which were not strongly affected by U0126-showed opposite responses to treatments with individual drugs confirming the results of L1000 profiling.
Next we found that BRD0418, similar to OSM, stimulates signaling through MAPK Erk , as treatment of HepG2 cells with those agents led to ERK1/2 phosphorylation, which temporally preceded upregulation of TRIB1 expression (Fig. 7C). The response to OSM was rapid and short-lived, as it was strong at 15 minutes and reverted to basal level 1 hour after the treatment, whereas the response to BRD0418 occurred 1 hour after treatment and persisted longer, remaining strong at 6 hours after the treatment. Interestingly, constitutively elevated expression of TRIB1 in HepG2 cells stably transfected with the expression plasmid also resulted in elevated phosphorylation of ERK (Fig. 7C). These data indicate that BRD0418 affects elements of the network that operate downstream of JAK and upstream of the ERK signaling pathway, leading to broad MEK1/2-dependent transcriptional changes. It is unlikely that BRD0418 acts directly through nuclear hormone receptors known to modulate lipid metabolism, as PPARγ, PPARα, FXR and RAR agonists had no effect on TRIB1 transcript levels in HepG2 cells (data not shown).

Effects of genetic perturbation of TRIB1 expression on lipid metabolic genes and on BRD0418 responses in HepG2 cells
Changes in the level of TRIB1 expression in animal models have been reported to regulate expression of TG but not cholesterol biosynthetic genes [13,14]. To evaluate the role of TRIB1 in lipid metabolism in HepG2 cells and in mechanism of BRD0418 action we constructed HepG2 clones stably overexpressing or lacking the expression of TRIB1. First, to help determine whether changes in LDLR and PCSK9 expression entailed by BRD0418 are related to upregulation of TRIB1 expression or are independently induced by BRD0418 treatment, we measured the level of all three transcripts in a collection of clones of HepG2 cells selected after Table 3. Effect of BRD0418 on TRIB1 and LDLR levels in cell lines of various origins. Twelve cell lines were subjected to treatment with BRD0418 in concentrations ranging from 0.2 to 25 μM and tested for TRIB1 and LDLR up-regulation relative to GAPDH expression after 6 hour and 24 hour exposure to the compound. The values listed in the table represent the lowest concentration of the compound at which 2-fold up-regulation (or maximal up-regulation shown in brackets) of the measured transcript was observed. IA, inactive.  transfection with TRIB1 expression vector. We found no changes in the levels of LDLR or PCSK9 transcripts in clones in which TRIB1 overexpression ranged between 3-fold and 30-fold (Fig. 8). In contrast, the expression levels of SCD1 and APOC3 were significantly decreased in all clones (Fig. 8). These results indicate that the regulation of LDL uptake genes by BRD0418, at least under conditions of chronic overexpression of TRIB1, is not affected by TRIB1 upregulation. Second, given the established effects of TRIB1 overexpression on lipogenic genes we examined the impact of TRIB1 knockout in HepG2 cells on modulation of lipid gene expression by BRD0418. Since the TRIB1 mRNA has a very high turnover rate [33] and our attempts to downregulate TRIB1 expression with 14 different shRNA constructs did not yield more than a 50% reduction in the transcript level, we constructed TRIB1 knockout HepG2 cell line using a CRISPR-Cas9 gene editing approach [34]. HepG2 cells harbor an inactivating mutation in the PRKDC gene that is essential for non-homologous end joining (NHEJ) mechanism of DNA repair [35]. This information, retrieved from the CCLE database (http://www.broadinstitute.org/ ccle/home; 1 Dec 2014), provided plausible explanation for the resistance of this cell line to NHEJ based gene editing procedures that we encountered in early attempts. However, induction of the homologous recombination (HR)-based DNA repair by co-transfecting HepG2 cells with CRISPR-Cas9 plasmid together with the HR repair template DNA (Fig. 9A) yielded from 171 Geneticin resistant clones, 17 heterozygous clones carrying one copy of ΔTRIB1::Neo R allele and one clone with homozygous deletion of the wild-type TRIB1 locus that carried ΔTRIB1::Neo R allele in both copies of chromosome 8. PCR amplification of genomic DNA from this cell line produced the expected size DNA fragments for T U /N U , N D /T D and T U /T D primer pairs (1167 bp, 1228 bp and 3395 bp, respectively (Fig. 9A)) but failed to produce the 2020 bp intact TRIB1 fragment present in reactions carried out for the parental cell line or for heterozygous clones. Moreover, qRT-PCR reactions carried out for the HepG2 ΔTRIB1::Neo R cell line using the Taqman primers flanking the target sequence (Table 1) failed to amplify any products or resulted in ΔCt values at least 10 cycles higher than for the HepG2 cells consistent with complete abrogation of TRIB1 expression in those cells.
Next we compared levels of expression of key lipid metabolic genes in HepG2 cells and HepG2 ΔTRIB1::Neo R cells and compared responses of both cell lines to treatment with 10 μM BRD0418. Abrogation of TRIB1 expression in HepG2 cells resulted in a greater than 5-fold reduction in MTTP expression, a 2-fold upregulation of LDLR expression and a smaller but consistent elevation of PCSK9 and SCD1 transcript levels (Fig. 9B) further confirming the modulatory function of TRIB1 in hepatic lipid metabolism. Asymmetrically to the TRIB1 overexpression results, the TRIB1 knockout had no significant effect on BRD0418 driven downregulation of SCD1 and APOC3, genes involved in TG synthesis and lipoprotein assembly (Fig. 9C), indicating that BRD0418 effects are TRIB1 independent or that in the absence of TRIB1 expression, BRD0418 treatment can evoke a compensatory response. Therefore, TRIB1 overexpression is sufficient but not necessary to induce downregulation of SCD1 and APOC3 expression in HepG2 cells. As expected from TRIB1 overexpression results, the TRIB1 knockout had no large effect on responses of cholesterol metabolic genes to BRD0418 treatment (Fig. 9C).

Discussion
The TRIB1 locus has been linked through GWAS to changes in blood lipid levels and to risk of CAD and MI. Genetic studies in mice revealed that increased expression of Trib1 in liver is associated with decreased apoB-100 secretion and VLDL production and decreased plasma TG and cholesterol levels. In the present study, we identified, through a high-throughput qRT-PCR screen, small-molecule upregulators of TRIB1 expression and characterized the effects of the lead compound, BRD0418, on lipid metabolism in HepG2 hepatoma cells. BRD0418 increased expression of TRIB1 4-fold in HepG2 cells in a dose-dependent manner (EC 50 = 2.5 μM). Accordingly, BRD0418 also decreased the rate of TG synthesis and inhibited secretion of apoB-100. Unexpectedly, however, treatment with BRD0418 also led to a 4-fold upregulation of LDLR expression, a 3-fold increase of the LDL receptor protein and a proportional increase in the rate of LDL uptake. Furthermore, BRD0418 was equally potent in inhibiting PCSK9 mRNA expression and protein secretion, inhibiting PCSK9 protein secretion by 70% after 24-hour treatment.
The effects of BRD0418 on LDLR and PCSK9 expression appeared to be independent of TRIB1 upregulation, as overexpression of TRIB1 cDNA had no impact on LDLR and PCSK9 transcript levels. As expected, based on previous studies in mice [14], plasmid-driven TRIB1 overexpression led to downregulation of the expression levels of SCD1, a rate-limiting gene in hepatic TG synthesis. Interestingly, we found equally strong correlation between overexpression of TRIB1 cDNA in HepG2 cells and downregulation of APOC3 expression. ApoC-III is a major component of lipoproteins linked to the rate of VLDL secretion and a negative regulator of lipoprotein lipase and blood triglycerides [36,37]. APOC3 is an independent CHD and MI risk factor, and hypomorphic apoC-III variants are cardioprotective [38][39][40]. Treatment of HepG2 cells with BRD0418 also downregulated, as a part of late response, APOC3 expression. We detected significant changes in steady state levels of genes involved in VLDL production (MTTP) and LDL uptake (LDLR and PCSK9) in HepG2 ΔTRIB1::Neo R cells that lack TRIB1 expression due to CRISPR-Cas9 induced inactivation of the chromosomal TRIB1 locus. This observation further highlights the role of TRIB1 in regulation of hepatic lipoprotein metabolism albeit the observed changes differ from those reported for mice, where Trib1 knockout was associated with elevated expression of Fasn and Scd1 and no changes in Mttp expression in liver [14]. HepG2 ΔTRIB1::Neo R cells responded to BRD0418 treatment with similar magnitude and pattern of responses as parental HepG2 cells (Fig. 9C) indicating that changes in SCD1 and APOC3 expression produced by BR0418 are also independent of TRIB1 upregulation. Alternatively, it is likely that TRIB1 plays a role in driving some of the responses to the drug but its absence in the knockout strain allows for engagement of a compensatory mechanism (e.g. through activity of TRIB2 or TRIB3, which is also upregulated in response to BRD0418 treatment (Fig. 3)) that masks effects of the TRIB1 knockout.
L1000 transcriptional profiling indicated that BRD0418 had a broad effect on lipid metabolic genes, and comparison of L1000 profiles revealed that cells treated with MEK inhibitors U0126 and PD-98059 produced opposite responses compared to cells treated with the BRD0418 and its analogs. Upon further characterization, we found that the responses to BRD0418 are MEK-dependent, as they are blocked by treatment with U0126, and that BRD0418 stimulates ERK1/2 phosphorylation. MAPK Erk signaling has been previously shown to regulate apoB secretion in HepG2 cells, as ERK inhibition stimulates VLDL assembly and secretion [41,42], whereas active ERK signaling is required for downregulation of apoB secretion by nobiletin [43]. ERK signaling has also been shown to be required for downregulation of PCSK9 expression by OSM [32] and for stimulation of LDLR expression by berberine [44]. Activation of ERK1/2 signaling by BRD0418 was reflected in the L1000 transcriptional profile through induction of early-response genes including EGR1 (Fig. 3). EGR1 has been previously reported to drive upregulation of LDLR expression in response to OSM [45], which is known to stimulate LDL uptake both in vitro and in vivo [46,47]. Other treatments that stimulate ERK phosphorylation in hepatoma cells that had been previously shown to upregulate LDL uptake include HGF [48], FGF21 [49], hGH [50], PMA [51] and AICAR [52]. The AICAR effect was shown to stem from stimulation of ERK signaling, rather than from activation of AMPK [52].
Interestingly, TRIB1 itself has been shown to directly interact with MEK1 and to enhance ERK signaling [53,54]. A novel connection between mitochondrial function, ERK1/2 signaling and TRIB1 expression was recently established in a report demonstrating the MEK-dependent upregulation of TRIB1 expression by the oxidative phosphorylation inhibitor Oligomycin A [31]. Treatment of HepG2 cells with BRD0418 had no obvious effects on mitochondrial function, as measured by cellular ATP levels in cells grown on galactose, or on mitochondrial membrane depolarization, as measured by JC-1 staining (data not shown). However, we cannot preclude more nuanced effects of BRD0418 on mitochondria as we observed that berberine, which was reported to inhibit mitochondrial complex I [55], and AICAR, which simulates changes in cellular AMP/ATP ratio, also strongly stimulated upregulation of TRIB1 and LDR1 expression in HepG2 cells (S1 Fig. AICAR and berberine treatment). Effects of AICAR were observed at relatively high concentrations of the compound (TRIB1 EC 50 = 980 μM) consistent with reported doses required to stimulate LDLR expression in HepG2 cells [52] and with doses needed to produce physiological responses [56]. Our observation differs from a recent report indicating that AICAR has no effect on TRIB1 expression in HepG2 cells perhaps due to a difference in a tested drug dose [31].
The differential effect of BRD0418 on genes regulating LDL uptake (increased expression of LDLR and decreased expression of PCSK9) was preserved in cells grown under conditions of cholesterol depletion, and was additive with effects of statins, suggesting that BRD0418 is not acting along the SREBP2 regulatory pathway. The decoupled response of LDLR and PCSK9 expression in HepG2 cells treated with BRD0418 resembled the responses to OSM, a cytokine involved in liver development and regeneration [32], and to berberine, a plant alkaloid with pleiotropic effects and undefined mode of action [57]. OSM was efficacious in lowering triglyceride and cholesterol levels in hyperlipidemic syrian hamsters demonstrating that the OSM signaling pathway that drives changes in lipoprotein metabolism in hepatoma cells is also operational in vivo [47]. Berberine was also shown to have hypolipidemic activity, both in animal models and in patients [44], however the efforts to optimize its PK properties and to identify its molecular target(s) are hampered by the nature of its chemical scaffold.
The molecular target of BRD0418 is currently unknown, but our data indicate that its perturbation produces a broad response that involves activation of MAPK erk signaling followed by upregulation of a number of early response genes including TRIB1 and LDLR and leads to remarkable change in the lipoprotein metabolism, which results in a decrease in VLDL production and upregulation of LDL uptake in cells of hepatic origin. Identification of the molecular target of BRD0418 should help to unravel how those two arms of hepatic lipoprotein metabolism are coordinated. Overexpression of TRIB1 has been also linked to leukemia [18] and, while BRD0418 has limited effects on non-hepatic cells and the potency of TRIB1 as an oncogene has not yet been fully evaluated, the oncogenic potential of TRIB1 inducing agents will need to be closely monitored in future animal studies.
BRD0418 was identified from a library designed using the diversity oriented synthesis (DOS) strategy, which yields molecules with diverse stereochemistries and skeletons and provides modular synthetic pathways to facilitate downstream medicinal chemistry efforts. Initial SAR derived from the closer characterization of the screening hits from the benzofuran library indicated that the R 1 position of the benzofuran scaffold can tolerate a wider range of substituents than R 2 position, and suggests possible derivatization strategies for BRD0418 to generate affinity and cross-linking probes that can be employed for target-identification efforts.
Collectively, our data highlight the presence in hepatic cells of a naturally occurring, MAP-K Erk -dependent lipoprotein rheostat that through the two arms of lipoprotein metabolism can downregulate the VLDL output by decreasing TG synthesis and upregulate the LDL uptake by increasing the expression of LDL receptor on the cell surface. This rheostat can be naturally regulated by endogenous modulators like OSM or growth factors but can also be accessed by small-molecule natural products like berberine or by more chemically tractable synthetic compounds like BRD0418 described here. Identification and targeting of this new regulatory node will help to find new hypolipidemic agents and to evaluate their therapeutic utility.
Supporting Information S1 Dataset. L1000 data.  Table. GO enrichment analysis. GO terms (process) enriched in the list of L1000 genes affected by the treatment of HepG2 cells with 10 μM BRD0418 were identified using the web tool GOrilla (cbl-gorilla.cs.technion.ac.il/). Only terms with FDR < 1.0E-1 and P-value < 1.0E-4 are listed in the table. Since L1000 approach provides an abbreviated transcriptional profile comprising 961 genes only one term in this analysis reached a significance level of FDR < 0.05. (DOCX)