Search for Natural Compounds That Increase Apolipoprotein A‐I Transcription in HepG2 Cells: Specific Attention for BRD4 Inhibitors

Abstract Although increasing apolipoprotein A‐I (apoA‐I) might lower the cardiovascular disease risk, knowledge on natural compounds that elevate apoA‐I transcription is limited. Therefore, the aim of this study was to discover natural compounds that increase apoA‐I transcription in HepG2 cells. Since BRD4 inhibition is known to elevate apoA‐I transcription, we focused on natural BRD4 inhibitors. For this, the literature was screened for compounds that might increase apoA‐I and or inhibit BRD4. This resulted in list A, (apoA‐I increasers with unknown BRD4 inhibitor capacity), list B (known BRD4 inhibitors that increase apoA‐I), and list C (BRD4 inhibitors with unknown effect on apoA‐I). These compounds were compared with the compounds in two natural compound databases. This resulted in (1) a common substructure (ethyl‐benzene) in 60% of selected BRD4‐inhibitors, and (2) four compounds that increased ApoA‐I: hesperetin, equilenin, 9(S)‐HOTrE, and cymarin. Whether these increases are regulated via BRD4 inhibition and the ethyl‐benzene structure inhibits BRD4 requires further study.

Supporting information Additional supporting information may be found online in the Supporting Information section at the end of the article.

Introduction
Cholesterol efflux capacity is defined as the amount of cholesterol taken up from cholesterol-loaded macrophages by high-density lipoprotein (HDL) particles. It is inversely associated with the incidence of cardiovascular events (Rohatgi et al., 2014). As an elevated in vitro cholesterol efflux capacity may reflect increased reverse cholesterol transport in vivo; the efflux capacity may be a useful parameter for the development of cardiovascular disease (CVD)-lowering strategies (Rohatgi et al., 2014). Apolipoprotein A-I (apoA-I) is the principal component of HDL, which can take up cholesterol by binding to the ATPbinding cassette A1 (ABCA1), the trans membrane cholesterol transporter on macrophages (Phillips, 2014). The plasma concentration of apoA-I is associated with increased cholesterol efflux capacity (Saleheen et al., 2015). Therefore, a promising strategy to increase cholesterol efflux capacity is to increase the amount of nascent HDL particles by increasing de novo apoA-I production (Dullens et al., 2007;Smits et al., 2014). The effectiveness of increasing apoA-I concentrations in the combat against CVD is supported by several in vivo animal (Rubin et al., 1991;Schultz et al., 1993) and human studies (Nissen et al., 2003;Tricoci et al., 2015). For example, intravenous infusion of recombinant apoA-I particles decreased atherosclerosis progression, as it reduced the atheroma volume in patients with acute coronary syndromes (Nissen et al., 2003). Moreover, the use of apoA-I mimetics like CSL112 (Tricoci et al., 2015) clearly enhanced cholesterol efflux capacity. Besides the involvement of apoA-I in enhancing cholesterol efflux capacity, apoA-I may also provide other cardioprotective effects. ApoA-I is antiinflammatory (Umemoto et al., 2013), antithrombotic (Epand et al., 1994), and has glucose-lowering properties (Dalla-Riva et al., 2013;Drew et al., 2009). Altogether, this illustrates the crucial role for elevating apoA-I production in CVD risk management.
In addition, studies have indicated a positive role for the family of bromodomain and extra-terminal (BET) protein inhibitors to increase apoA-I production. For example, in in vitro as well as in in vivo studies, the BET inhibitor RVX208 (or apabetalone) increased apoA-I transcription and protein production (Gilham et al., 2016). Additionally, there are many other compounds with BET-inhibiting function and the capacity to increase apoA-I synthesis, at least in vitro, such as JQ1(+) (Kempen et al., 2013), Ro11-1464 (Zanotti et al., 2011), GW841819X (Chung et al., 2011), GSK1210151A or I-BET151 (Seal et al., 2012), alaprazolam (Filippakopoulos et al., 2012), GSK1324762A or I-BET762 , and thieno-or benzotriazolodiazepines (Kempen et al., 2013) such as U-34599 and U-51477 (Princen JMG May 28;Princen and Kooistra, 2003;Kempen et al., 2013). In humans, four types of BET proteins have been identified, namely, bromodomaincontaining protein (BRD) 2, BRD3, BRD4, and testesspecific BRDT. Although many BET inhibitors are multi-BET active, in vitro experiments have shown that specifically the silencing of BRD4 is involved in increasing apoA-I production (Chung et al., 2011). For example, JQ1 (+) and RVX208 inhibit BRD4, which may explain their effects on increasing apoA-I production. Currently, BETinhibition is considered a promising route to increase apoA-I transcription and most BET inhibitors under development are of synthetic origin. Possibly, natural compounds can-assuming they pass safety assessment, affordable sourcing, and have favorable ADME properties-be used as a functional food ingredient. Therefore, the aim of this study was to identify natural compounds that increase apoA-I transcription, by an in silico and in vitro approach based on a literature review. Specific attention was paid to the role of BRD4 inhibition.

General Approach
To identify new, natural compounds, or functional (sub-) structures that increase apoA-I transcription, three lists (Lists A, B, and C) were compiled based on a literature review, and via a database of bioactivities (Gaulton [22]). The compounds in these lists were compared with those from two databases containing natural compounds: a company-owned database and a commercially available one. Next, most similar compounds were tested in vitro for their ability to increase apoA-I transcription. For a schematic representation of the study design, see Fig. 1.

Literature Review
To identify known natural compounds that increase apoA-I production or HDL, the literature was scrutinized using PubMed for articles published until August 2015. As search term (((BRD*) OR bromodomain)) AND (((apo*) OR High-Density Lipoproteins, Pre-beta) OR apolipoprotein A-I) was used. In addition, a well-curated database of bioactivities (Gaulton et al., 2017) was used to identify natural compounds and synthetic compounds, which inhibit BRD4. Additionally, we recently identified cymarin and 9 (S)-HOTrE as apoA-I transcriptional elevating compounds (van der Krieken et al., 2018). Therefore, these two compounds were included in our literature review results as well.
This resulted in the formation of three lists ( Fig. 1): List A contained compounds that increase apoA-I production with unknown BRD4 inhibiting capacity. List B contained compounds that were BRD4 inhibitors and increased apoA-I production. List C was composed of known BRD4 inhibitors, with unknown effects on apoA-I production. Only BRD4 inhibitors with IC 50 values <500 nM were selected.
In Silico Screening for Natural Ingredients with Potential to Increase apoA-I Two databases were searched for natural compounds that may increase apoA-I transcription: a database provided by DSM (DSM, Delft, The Netherlands) containing 2000 natural compounds) and a commercially available one, the Dictionary of Natural Products (version 18.1; Francis & Taylor) containing about 260.000 natural compounds.
Structural similarities between compounds identified from the literature and molecules present in those databases were determined using the Tanimoto algorithm based on circular fingerprints of four atoms (FCFP4). This was performed for molecules from each list.
The method used to screen for molecules depended on the list of molecules: since lists A and B contained relatively few compounds, a structural similarity search was performed for each compound from these lists in the aforementioned natural product databases via Tanimoto-based similarities. In addition, the structural similarity of compounds from lists A and B was shown in a structural similarity matrix. By performing a substructure search in list C, we determined if the BRD4 inhibitors in this list contain a common substructure.
In addition, retrieved compounds were filtered on having a structural similarity of at least 0.5, and being commercially available.
Finally, BioVIA's PipelinePilot version 9.2 was used for all cheminformatic calculations (determination of structural similarity and identification of a common substructure).

In Vitro apoA-I Transcription
Human hepatocellular liver carcinoma (HepG2) cells (kindly provided by S. Braesch-Andersen, Mabtech, Nacka Strand, Sweden) were cultured at 37 C in a humidified atmosphere and 5% CO 2 . For cell culturing, minimum essential medium (MEM) was used supplemented with 10% fetal calf serum (v/v, South-American, Greiner Bioone, Frickenhausen, Germany), L-glutamine (Invitrogen Life Technologies, Carlsbad, CA, USA), 1% penicillin/ streptomycin (v/v), 1% nonessential amino acids (NEAA, v/v), and 1% sodium pyruvate (v/v, all from Invitrogen Life Technologies). To gain insight into the effect of the selected natural compounds on apoA-I transcription, HepG2 cells were exposed for 48 h to different doses of each compound. Stocks were prepared in recommended carrier solutions and were diluted in culture medium. For carrier controls, maximally 0.5% DMSO or ethanol was used. After incubation, cells were microscopically inspected and pictures of each condition were made to confirm cell vitality (data not shown). Different doses of the BRD4 inhibitor RVX208 (Bioconnect, Huissen, The Netherlands) were used as positive controls for their ability to increase apoA-I transcription. Furthermore, in each Fig. 1 Study design. In order to find natural compounds with the ability to increase apoA-I transcription in HepG2 cells, literature was reviewed. This resulted in lists A, B, and C. Molecules from those lists were compared with those of two databases of natural compounds. One of the lists (C) contained many hits, allowing a common substructure search. Finally, four hits from the screening, and for each compound a structurally comparable compound, were tested for their effect on apoA-I transcription in HepG2 cells experiment 3 μM JQ1(+) (Tocris Bioscience, Abingdon, UK), another known BRD4 inhibitor was used as a control for increased apoA-I transcription in HepG2 cells.

ApoA-I qPCR Measurement
Total RNA was isolated according to the Qiagen Trizol protocol. Next, cDNA was produced using Taqman reagents. ApoA-I (Hs00163641_m1) and internal control cyclophilin A (Hs99999904_m1) mRNA expression was determined using the 7300 Real-Time PCR System. Both TaqMan Gene Expression Assays and reagents were obtained from Applied Biosystems (Warrington, UK).

Literature-Derived ApoA-I Increasing Compounds and/or BRD4 Inhibitors
Eight compounds were identified that increased HDL-C or apoA-I protein and/or mRNA expression according to literature. It is unknown whether these compounds are also BRD4 inhibitors (Table 1; list A). Among the compounds of List A, hesperetin, equilenin, 9(S)-HOTrE, and cymarin (Fig. 2, upper panel) were confirmed to be of natural origin as they are listed in DNP and DSM natural databases. There were six synthetic compounds that increased apoA-I production (either measured as transcription, protein secretion, or luciferase reporter activity) and were known BRD4 inhibitors (Table 1; list B). Two of the synthetic compounds (U-34599 and U-51477) from list A and three of the synthetic compounds of list B (JQ1(+), Ro11-1464 and RVX208) also increased in vitro apoA-I protein secretion, besides increasing apoA-I transcription.
In addition, the literature search also resulted in the identification of 48 synthetic compounds that were described to bind to BRD4 (IC 50 < 500 nM) (Table S1; list C), suggesting that they may affect apoA-I. However, their effects on apoA-I transcription and/or apoA-I production have not been reported in the literature.
In Silico Screening for Natural Ingredients with Potential to Increase apoA-I By comparison of the molecular structures within lists A and B, a structural similarity of 0.7 was found between U-   (Table S2). Searches for structural similarity of the compounds found in the literature (list A and B) via in silico screening are available upon request. Among them, four natural compounds were selected based on their commercial availability (Fig. 2, lower panel). For compound hesperetin, eriodictyol was selected (0.79 similarity); for compound equilenin hordenine was selected (0.53 similarity); for compound 9(S)-HOTrE we selected 5 (S),15(S)-DiHETE (0.91 similarity), and for the compound cymarin we selected the structurally comparable compound emicymarin (0.81 similarity). These eight compounds were further tested for their effects on apoA-I transcription in HepG2 cells (Fig. 2).
After determining structural similarities of these synthetic BRD4 inhibitors within list C, a common substructure (ethylbenzene moiety) was discovered in 60% of all compounds (Fig. 3). Furthermore, the DNP and DSM databases were searched for natural compounds that contained this ethylbenzene moiety. This search resulted in the identification of approximately 179 compounds with a similarity of >0.5. This data is available upon request.

Effects of the Selected Compounds on apoA-I mRNA Expression in HepG2 Cells
Addition of different doses of the known BET inhibitor RVX208 (Fig. S1) clearly increased dose dependently the production of apoA-I mRNA. Likewise, the positive control JQ1(+) increased apoA-I transcription in all experiments (Fig. 4), indicating that it could serve as a positive control for apoA-I production in vitro. Hesperetin, did not increase apoA-I transcription in our in vitro testing system. If anything, apoA-I transcription was reduced (Fig. 4a). However, eriodictyol, the selected compound with a high similarity to hesperetin, did induce apoA-I transcription by 35% at a dose of 50 μM (Fig. 4b). Eriodictyol in doses ranging from 100 to 250 μM decreased apoA-I and had no effect at lower doses. Equilenin raised apoA-I transcription in HepG2 cells, although no clear dose-response pattern was evident (Fig. 4c). Hordenine, the compound with a structure comparable to equilenin, did not affect apoA-I transcription (Fig. 4d). 9(S)-HOTrE increased apoA-I  Table S1 with unknown effects on apoA-I Fig. 4 Relative apoA-I mRNA expression in HepG2 cells treated with different doses of (a) hesperetin, (b) eriodictyol, (c) equilenin, (d) hordenine, (e) 9(S)-HOTrE, (f) 5(S),15(S)-DiHETE, (g) cymarin, and (h) emicymarin. Known BET inhibitor and apoA-I increaser JQ1(+) (3 μM) was used as a positive control, whereas thapsigargin (0.01 μM) was used as a control to confirm decreased apoA-I expression. *The data presented in panels (e) and (g) are adapted from van der Krieken et al. [23]. Compounds were tested in duplo; error bars indicate the SD transcription by 35% at a dose of 170 nM (Fig. 4e), while the structural variant of this compound 5(S),15(S)-DiHETE, reduced apoA-I transcription (Fig. 4f). Cymarin slightly decreased apoA-I mRNA expression at low concentrations, but increased its expression by 37% at doses ranging from 18 to 45 μM (Fig. 4f). The structurally comparable compound emicymarin increased apoA-I up to 77% (Fig. 4h). All compound doses were tested in duplo in HepG2 cells and measurements were performed in duplo.

Discussion
We aimed to identify new natural compounds that increase apoA-I transcription via a literature search, in silico screening, and in vitro testing. Specifically, we searched for natural BRD4 inhibiting compounds, since a growing number of studies point toward a role for BRD4 inhibitors for increasing apoA-I transcription (Chung et al., 2011).
Eriodictyol, the compound that was structurally comparable to hesperetin, is a metabolite of hesperitin and both are present in oranges. We observed that, eriodictyol, but not hesperetin, induced apoA-I transcription by 35% at a dose of 50 μM. Possibly, hesperetin needs to be converted into eriodictyol to exert its effects on apoA-I transcription (Brand et al., 2008). If so, it would be of interest to study also effects of other metabolites of hesperetin on apoA-I transcription. The optimal dose of eriodictyol in our study was 50 μM, which has been reported as nontoxic in various cell lines, such as macrophage mouse cells, NIH3T3 (fibroblast) cells, and HaCaT (human skin) cells (Lee et al., 2013). In vivo studies in mice suggested that eriodictyol is antiinflammatory (Zhu et al., 2015), while in vitro studies in mouse spleen cells showed that this compound has antioxidant properties (Mokdad-Bzeouich et al., 2016). In HepG2 cells, eriodictyol increased insulinstimulated glucose uptake (Zhang et al., 2012) and inhibited of c-Jun N-terminal kinase (JNK) in macrophages (Lee et al., 2013). Since JNK is also activated during ER-stress (Lee et al., 2011), it is of interest to examine if eriodictyol is also able to prevent ER-stress-induced decreases in apoA-I transcription.
Equilenin is a naturally occurring estrogenic steroid that can be extracted from the urine of pregnant mares (Bhavnani and Stanczyk, 2014). Equilenin (10 μM) is a potent activator of the apoA-I promoter and increased apoA-I transcription with 140% in HepG2 cells (Zhang et al., 2001). At a dose of 10 μM, we observed that equilenin increased apoA-I transcription by 19% in HepG2. However, the optimal dose of equilenin in our study was only 0.6 μM, which increased apoA-I transcription by 27%. In women, estrogen replacement therapy in postmenopausal women increases plasma HDL-C and apoA-I concentrations (Lamon-Fava et al., 2003). Long-term estrogen treatment reduced the risk of mortality, myocardial infarction, or heart failure in early postmenopausal women, but not in women who start hormone therapy 5-20 years after menopause. Moreover, unwanted effects such as thromboembolic disease or breast cancer have been observed (Mosca et al., 2001). Hordenine, the structurally comparable compound of equilenin (0.53 similarity) did not influence apoA-I transcription. Unfortunately, no other compounds with a resemblance to equilenin above 0.53 were commercially available.
9(S)-HOTrE increased apoA-I transcription in HepG2 cells by 35% at a dose of 170 nM (Fig. 4). 9(S)-HOTrE is a monohydroxy polyunsaturated fatty acid found in the leaves of a plant called Glechoma hederacea (Kim et al., 2015). Studies in primary macrophages, suggested a possible protective role for a Glechoma hederacea extract on rheumatoid arthritis and osteoporosis, due to antiinflammatory actions (Hwang et al., 2014). In humans, 9(S)-HOTrE can be produced from the essential fatty acid alpha-linolenic acid by the action of 5-lipoxygenase (Yandava et al., 1999). Unfortunately, the compound with a comparable structure to 9(S)-HOTrE-5(S),15(S)-DiHETE did not increase apoA-I transcription.
Cymarin, a cardiac glycoside produced by the digitalis plant (Gozalpour et al., 2013), increased transcription of apoA-I by 37% at doses ranging from 18 to 45 μM. The structurally comparable compound of cymarin, emicymarin increased apoA-I transcription up to 77%. Emicymarin is, like cymarin, a cardenolide and cardiac glycoside. Although digoxins are used in the clinic, there is concern about its toxicity, as an in vivo plasma concentration above 3 μg/L could give rise to symptoms of toxicity, (Vivo et al., 2008).
The bioavailability of a compound, which was not tested in our studies, is of importance for its possible in vivo effects on apoA-I transcription. In future experiments, bioavailability can therefore be addressed by using a transwell system that combines intestinal and hepatic cells. Moreover, by using the Simulator of the Human Intestinal Microbial Ecosystem (SHIME), the in vivo situation could be better mimicked (Van Rymenant et al., 2018). Also, more structural derivatives of the main structures that were presented could be tested. In fact, all natural compounds that have a minimum similarity above 0.5 could be examined for their effect on apoA-I transcription. Likewise, one should investigate their effect on toxicity. Our findings further suggest that the lack of effect could be caused by the removal of a specific structural element that is essential for increasing apoA-I transcription. It would be interesting to discover which part of a molecular structure is responsible for the increased apoA-I transcription.
In summary, this study confirmed that the reported natural compounds equilenin, cymarin, and 9(S)-HOTrE increase apoA-I transcription in HepG2 cells. In addition, as predicted by in silico analysis, two new (structurally comparable) compounds eriodictyol and emicymarin, increased apoA-I in vitro. Additional experiments are needed to confirm whether these compounds increase apoA-I transcription by inhibiting BRD4. Additionally, it would be interesting to investigate the effects of all BRD4 inhibitors that were predicted by in silico modeling, on their effect on apoA-I production, in vitro and in vivo.