Brothers in Arms: ABCA1- and ABCG1-Mediated Cholesterol Efflux as Promising Targets in Cardiovascular Disease Treatment

A. Initiation and Propagation of Reverse Cholesterol Transport

Each RCT cycle is initiated by the removal of cellular cholesterol from peripheral tissues, of which the determinants (i.e., HDL size and composition, gender, body mass index, and age) have been reviewed recently by Talbot et al. (2018). Active transport is involved in 70% of all cellular cholesterol efflux pathways (Ono, 2012; Phillips, 2014). Passive bidirectional cholesterol flux between HDL particles and the plasma membrane is mediated by aqueous diffusion and via interaction of HDL particles with scavenger receptor B1 (SR-B1). SR-B1 mediates cholesterol uptake in the cell without endocytic uptake and degradation of HDL particles by ABCG1 (Phillips, 2014). However, the contribution of this mechanism to peripheral cholesterol efflux is expected to be minimal, as SR-B1 is most abundantly expressed in the adrenal gland, and at lower levels in peripheral tissues. Aqueous diffusion includes passive transport through the intervening aqueous phase until collision occurs between cholesterol and an extra- or intracellular cholesterol acceptor (Yancey et al., 2003; Rosenson et al., 2012; Phillips, 2014). ATP-dependent cholesterol efflux is mainly facilitated by ABCA1 and ABCG1 (Fig. 2) (Oram, 2003; Yancey et al., 2003; Yvan-Charvet et al., 2010b; Phillips, 2014; Westerterp et al., 2014). ABCA1-mediated cholesterol and phospholipid efflux is initiated by the interaction of lipid- and cholesterol-free apoA-I particles with ABCA1 transporters on peripheral tissue (i.e., highest expression in macrophages and foam cells) and hepatocytes (Fig. 2). This results in the formation of lipid-poor pre-βHDL particles, which upon rapid esterification by lecithin:cholesterol acyltransferase are converted into lipid-rich HDL particles (Morgado et al., 2005). These particles mediate cellular cholesterol efflux via interaction with ABCG1, thereby further facilitating transport of cholesterol from peripheral tissues toward the liver, followed by uptake into hepatocytes via SR-B1 (Tall and Yvan-Charvet, 2015). As the final step of RCT, cholesterol is secreted into the bile via an ABCG5/8 heterodimer complex present at the canalicular membrane (Thomas et al., 2003; Roglans et al., 2004; Zanlungo et al., 2004; Valasek et al., 2007; Lee et al., 2016). These two ABCG transporters are only expressed in hepatocytes, gallbladder epithelium, and enterocytes (Tauscher and Kuver, 2003; Wang et al., 2015a; Patel et al., 2018). Formation of a heterodimer of ABCG5 and ABCG8 in the endoplasmic reticulum is required before the complex can be translocated to the apical membrane, where it facilitates sterol transport into the bile or gut lumen (Graf et al., 2003; Yu et al., 2014). The bilairy excretion rate of cholesterol is associated with the level of hepatic ABCG5 and ABCG8 expression (Yu et al., 2005). Consequently, the ABCG5/G8 complex is important in the correction of high plasma and hepatic cholesterol and sterol levels (Yu et al., 2002, 2014).

B. ATP-Binding Cassette A1 and ATP-Binding Cassette G1 as Master Effectors of Cholesterol Efflux

ABCA1 plays a crucial role in the cellular cholesterol efflux, and its importance is illustrated by Tangier disease, which is characterized by a severe deficiency in plasma HDL and cholesterol caused by autosomal recessive mutations in the ABCA1 gene (Brooks-Wilson et al., 1999; Langmann et al., 1999; Rust et al., 1999; Oram, 2000; Uehara et al., 2011; Phillips, 2018). The human ABCA1 gene has a total length of 149 kb, including a 1,453-bp promoter, 146,581 bp of introns and exons, and a 1-kb 3′ flanking region. Of the exons, 58 are comprised in the ABCA1 gene, and multiple bindings sites for transcription factors are detected in the promoter region (Gene, 1982–2019a; Santamarina-Fojo et al., 2000). Like many other transporters of the ABCA subfamily, ABCA1 is a full transporter, which contains 2,261 amino acids and is integrated into the membrane via two transmembrane domains (TMDs) that both comprise six transmembrane helices. In addition, ABCA1 has two nucleotide binding domains (NBDs), which contain the conserved Walker-A and Walker-B peptide motifs (Langmann et al., 1999; Santamarina-Fojo et al., 2000; Uehara et al., 2011). The tissue distribution of ABCA1 is ubiquitous, and expression is especially high in placenta, liver, small intestine, macrophages, adrenal glands, lung, and fetal tissues (Langmann et al., 1999; Uehara et al., 2011). ABCA1 initiates cellular cholesterol efflux and RCT to the lymph and bloodstream via a specific interaction with apoA-I, allowing cellular phospholipids and cholesterol to bind to these apolipoproteins, which almost exclusively mediates HDL biosynthesis and is therefore seen as its rate-limiting step (Lee et al., 2002; Murthy et al., 2002; Ohama et al., 2002; Mulligan et al., 2003; Singaraja et al., 2003). It remains, however, unknown whether phospholipid and cholesterol binding occurs at the plasma membrane surface or whether apoA-I is internalized and targeted to late endosomes after binding to ABCA1 at the plasma membrane, where it forms complexes with lipids that are subsequently released by exocytosis (Takahashi and Smith, 1999; Neufeld et al., 2001; Vaughan and Oram, 2003; Denis et al., 2008; Lorenzi et al., 2008; Azuma et al., 2009; Tang and Oram, 2009; von Eckardstein and Rohrer, 2009; Yvan-Charvet et al., 2010b; Westerterp et al., 2014; Du et al., 2015a) (Fig. 4). This is also supported by the localization of ABCA1 on early and late endosome and lysosome membranes, next to its plasma membrane localization, and in line with its reuptake to regulate cholesterol efflux rates, as described in the next section (Neufeld et al., 2001; Tang and Oram, 2009; von Eckardstein and Rohrer, 2009; Uehara and Saku, 2014; Westerterp et al., 2014; Du et al., 2015a). However, the majority of apoA-I lipidation is expected to occur at the plasma membrane (Denis et al., 2008; Faulkner et al., 2008). The exact modes of interaction between apoA-I and ABCA1 still need to be elucidated. At least six mechanisms have been proposed, as follows: 1) direct apoA-I binding to phosphatidylserine, which is translocated outward by ABCA1 floppase activity (Chambenoit et al., 2001; Alder-Baerens et al., 2005); 2) direct binding of apoA-I to extracellular ABCA1 loop domains (Wang et al., 2001; Fitzgerald et al., 2004); 3) apoA-I binding to protrusions as a result of ABCA1 floppase activity (Vedhachalam et al., 2007a); 4) outward translocation of phosphatidylinositol 4,5-bisphosphate mediated by ABCA1 floppase activity, which allows apoA-I to bind and unfold, followed by microsolubilization of the membrane (Gulshan et al., 2016); 5) low-affinity apoA-I binding to ABCA1 and high-affinity binding to cholesterol (Hassan et al., 2007; Vedhachalam et al., 2007b); 6) apoA-I binding to extracellular domains of ABCA1 dimers that have been formed from two ABCA1 monomers. These monomers have undergone a conformational change as a result of the translocation of phosphatidylcholine and cholesterol leading to dimer formation (Ishigami et al., 2018). Although ABCA1 mainly interacts with apoA-I, it can also associate with lipid-free apoE, which is most efficient when these apolipoproteins originate from small and dense HDL subfractions like HDL3b and HDL3c (Neufeld et al., 2001; Tang and Oram, 2009; von Eckardstein and Rohrer, 2009; Uehara and Saku, 2014; Westerterp et al., 2014; Du et al., 2015a).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Potential ABCA1- and ABCG1-driven cholesterol efflux modes. ABCA1-driven cholesterol efflux to lipid-poor apoA-I is hypothesized to occur either at the plasma membrane (upper left panel), or via endocytosis of the apoA-I–bound ABCA1 transporter, followed by intracellular lipid loading and exocytosis (upper right panel). ABCA1 is ubiquitously expressed, particularly at high levels in liver, small intestine, macrophages, adrenal glands, lungs, placenta, and fetal tissue. The transporter initiates cellular cholesterol efflux to the lymph and bloodstream via a specific interaction with apoA-I. Six different mechanisms have been proposed for the interaction between apoA-I and ABCA1 (small boxes), including the following: outward translocation of phosphatidylserine (PS) by ABCA1 floppase activity allowing apoA-I binding; direct binding of apoA-I to extracellular ABCA1 loop domains; ABCA1 floppase activity leading to the formation of protrusions facilitating apoA-I binding; ABCA1 floppase activity leading to the outward translocation of phosphatidylinositol 4,5-bisphosphate (PIP2), allowing apoA-I to bind and unfold, followed by microsolubilization of the membrane; low-affinity binding to ABCA1 and high-affinity binding to cholesterol; dimerization of ABCA1 transporter proteins is initiated by loading of the extracellular loop domains with cholesterol, followed by apoA-I binding to the dimerized cholesterol-loaded extracellular loop domains (ED). ABCG1-driven cholesterol efflux to HDL particles in the bloodstream or lymph is expected to be the result of a collision of these particles with cholesterol molecules that protrude from the plasma membrane (lower left panel), mediated either by a direct effect of the ABCG1 dimer on the membrane structure or by outward translocation via ABCG1 floppase activity (lower right panel). ABCG1 is expressed in many cell types, including macrophages, neurons, astrocytes, endothelial and epithelial cells, and many tissues (e.g., liver, intestine, kidney, spleen, lung, and brain), where it mediates basolateral cholesterol efflux.

In contrast to ABCA1, ABCG1 has a larger ambiguity regarding its lipid acceptors, which include HDL, LDL, and phospholipid vesicles (Wang et al., 2004; Vaughan and Oram, 2005; Kobayashi et al., 2006; Sankaranarayanan et al., 2009; Yvan-Charvet et al., 2010b; Phillips, 2014; Westerterp et al., 2014). Unlike ABCA1, ABCG1 is a half transporter, which only contains a single TMD comprising of six transmembrane helices and a single NBD at the C terminus of the TMD that is responsible for ATP binding and hydrolysis (Kerr et al., 2011; Uehara et al., 2011). Therefore, the ABCG1 protein needs to either homo- or heterodimerize with other ABCG proteins to become functional. The human ABCG1 gene comprises 23 exons spanning 98 kb (Gene, 1982–2019b; Kennedy et al., 2001). Eight different isoforms of ABCG1 are produced by alternative splicing, with a length varying between 644 and 785 amino acids. The ABCG1 protein is expressed in many cell types, including macrophages, neurons, astrocytes, endothelial, and epithelial cells, and in many tissues, such as the liver, intestine, kidney, spleen, lung, and brain (Nakamura et al., 2004; Kennedy et al., 2005; Wang et al., 2008; Bojanic et al., 2010). Although the precise cellular localization of ABCG1 needs to be elucidated, the transporter was detected in the plasma membrane and membranes of the Golgi apparatus and endosomes of ABCG1-overexpressing HEK293 cells and macrophages (Kobayashi et al., 2006; Wang et al., 2006; Tarling and Edwards, 2011; Neufeld et al., 2014; Phillips, 2014; Westerterp et al., 2014). It has been demonstrated that cholesterol efflux facilitated by ABCG1 does not increase lipoprotein binding to the cell surface (Wang et al., 2004; Kobayashi et al., 2006; Sankaranarayanan et al., 2009; Yvan-Charvet et al., 2010b; Phillips, 2014), which makes a mechanism similar to ABCA1 unlikely. Several pathways have been suggested to explain the ABCG1-mediated cellular cholesterol efflux mechanism (Yvan-Charvet et al., 2010b; Neufeld et al., 2014; Phillips, 2014). One of the proposed mechanisms suggests that ABCG1 facilitates protrusion of cholesterol from the membrane pool into the hydrophilic water layer lining the plasma membrane (Fig. 4). Subsequently, cholesterol uptake by an acceptor occurs after transient collision (Small, 2003; Yvan-Charvet et al., 2010b; Neufeld et al., 2014; Phillips, 2014). A second model suggests that ABCG1 promotes changes in the organization of plasma membrane phospholipids functioning as a phospholipid floppase, which leads to a redistribution of sterols to the plasma membrane (Yvan-Charvet et al., 2010b; Phillips, 2014). This could result in an increased efflux of cholesterol out of the cell by aqueous diffusion because of increased cholesterol concentrations at the plasma membrane (Sankaranarayanan et al., 2009; Yvan-Charvet et al., 2010b; Phillips, 2014).

In summary, ABCA1- and ABCG1-mediated cholesterol efflux is indispensable in cholesterol and phospholipid loading of apolipoproteins and thereby the maturation of HDL particles, which makes both ABC transporters essential for the initiation of RCT (Tang and Oram, 2009; Phillips, 2014; Uehara and Saku, 2014; Westerterp et al., 2014). Consequently, stimulation of cellular cholesterol removal via ABCA1 and ABCG1 could provide a promising target to enhance RCT.

C. Regulation of ATP-Binding Cassette A1– and ATP-Binding Cassette G1–Mediated Cholesterol Efflux

Cholesterol efflux by ABCA1 and ABCG1 is regulated by a variety of different mechanisms. Serum HDL and its key apoA-I are key regulators of the cholesterol efflux rate, in which high apolipoprotein levels are associated with high cellular cholesterol efflux rates. Cellular mechanisms mainly regulate ABCA1- and ABCG1-mediated cholesterol efflux by controlling the plasma membrane expression of both transporters. This is accomplished by various mechanisms. First, the nuclear receptors retinoid X receptor (RXR), peroxisome proliferator-activated receptors (PPARs), and liver X receptor (LXR) regulate the transcription of the genes encoding ABCA1 and ABCG1 (Fig. 5). Second, the stability of ABCA1 and ABCG1 mRNA can be influenced by their protein expression. Third, the expression of the transporter on the plasma membrane is regulated via modulation of its internalization, degradation, and recycling. Finally, activity and localization are enhanced by cAMP-mediated transporter phosphorylation and the stimulatory role of extracellular ATP thereon (Lee et al., 2011). The different cellular mechanisms influencing ABCA1- and ABCG1-mediated cholesterol efflux are discussed in more detail below, along with an overview of small-molecule treatment strategies known to influence these mechanisms.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Cellular regulation of ABCA1 and ABCG1 transporter expression. Expression of ABCA1 and ABCG1 transporters is increased by stimulation of the nuclear PPAR, which upon dimerization with RXR (i.e., another nuclear factor) bind to the PPAR-responsive element (PPRE), leading to the transcription of the nuclear LXR and RAR, respectively (right upper panel). Next, dimerization of LXR or RAR with RXR enables binding to the LXRE, inducing the transcription of ABCA1 and ABCG1, respectively. Once transcribed, the stability of ABCA1 and ABCG1 mRNA can either be enhanced or decreased upon binding of miRNAs, of which an overview is provided in Table 4 (right middle panel). ABCA1 and ABCG1 plasma membrane expression is also mediated via modulation of its lysosomal (blue vesicle) or endosomal (orange vesicle) degradation. The latter is stimulated by phosphorylation of the PEST sequence by apelin-13 (AP-13) via PKCα (left lower panel). Upon phosphorylation, calpain (CALP) can bind and initiate proteolysis, a process that is stimulated by calpastatin (CPSTAT), and negatively affected by calmodulin (CM) or by the HO-1 axis. Finally, the ability of ABCA1 and ABCG1 to bind lipid poor apoA-I is stimulated upon transporter phosphorylation by PKA at Ser-1042 and Ser-2054, located in the nucleotide binding domain of ABCA1. PKA is positively regulated by cAMP levels that depend on the balance of cAMP breakdown by PDE, formation by adenosine receptor–stimulated AC, and efflux by multidrug resistance protein (MRP) 4 and 5 (right lower panel).

A. Nuclear Receptors Are Important Regulators of ATP-Binding Cassette A1 and ATP-Binding Cassette G1 Expression

Multiple nuclear receptors are involved in the regulation of ABCA1 and ABCG1 mRNA expression, of which the two LXR isoforms, LXRα and LXRβ, have demonstrated to play an important role in the cholesterol homeostasis, including RCT (Zhao and Dahlman-Wright, 2010; Hong and Tontonoz, 2014). LXRα is predominantly expressed in metabolically active tissue (e.g., liver, kidney, macrophages, adipose tissue, and small intestine), whereas LXRβ has a more ubiquitous distribution and is in particular highly expressed in the developing brain (Zhao and Dahlman-Wright, 2010; Li et al., 2016). LXRs act as cellular cholesterol sensors, as they are activated by accumulation of oxysterols, oxidized derivatives of cholesterol, which subsequently induce transcription of genes involved in the protection of cells against cholesterol overload (Zhao and Dahlman-Wright, 2010; Di et al., 2012). After activation by oxysterols, LXRα and LXRβ form heterodimers with isoforms of RXRs, including RXRα, RXRβ, or RXRγ (Di et al., 2012; Hong and Tontonoz, 2014) (Fig. 5). After heterodimerization, LXR/RXR initiate transcription of target genes (e.g., genes involved in lipid synthesis and metabolism, including, but not limited to, the following: ABCA1, ABCG1, ABCG5, ABCG8, SREBP-1C, and FAS) by binding to the LXR response element (LXRE), which consists of two direct repeats (i.e., a AGGTCA sequence) separated by four nucleotides (Edwards et al., 2002; Zhao and Dahlman-Wright, 2010; Jakobsson et al., 2012; Hong and Tontonoz, 2014; Li et al., 2016). These LXREs are found in the proximal promotors of genes involved in fatty acid, bile acid, cholesterol, and glucose regulation, but also of genes with high relevance to RCT, including ABCA1 and ABCG1 (Zhao and Dahlman-Wright, 2010; Di et al., 2012; Hong and Tontonoz, 2014). Combined with its high expression in macrophages, LXRα is an important regulator in the initiation of RCT via modulation of ABCA1 and ABCG1 expression.

Another class of nuclear receptors involved in the regulation of ABCA1 and ABCG1 expression is PPAR, which comprises the following three isoforms: PPARα, PPARβ/δ, and PPARγ (Berger et al., 2005; Ogata et al., 2009). PPARα is mainly expressed in liver, kidney, heart, muscle, and other metabolically active tissues that rely predominantly on fatty acid β-oxidation. PPARβ/δ is ubiquitously expressed (e.g., cardiovascular, urinary, respiratory, digestive, endocrine, nervous, and hemato organ system), whereas PPARγ is mainly found in adipose tissue, skeletal and cardiac muscles, and human monocytes (Escher et al., 2001; Berger et al., 2005; Higashiyama et al., 2007; Ogata et al., 2009). Like LXR, PPARs form obligate heterodimers with RXR upon activation by their ligands (e.g., fatty acid metabolites), which bind to isotype-specific peroxisome proliferator response elements in target genes (Berger et al., 2005; Grygiel-Górniak, 2014) (Fig. 5). Targets of PPARα and PPARγ include genes involved in lipid metabolism (e.g., SLC25A20, APOA1, LXRα, and SCD-1) and glucose metabolism (e.g., G6PC, PCK, and PDK4) (Muoio et al., 2002; Li and Glass, 2004; Rakhshandehroo et al., 2010). PPARs have been associated with an increased ABCA1 expression and consequently enhanced HDL biogenesis in vitro and in vivo (Chinetti et al., 2001; Ogata et al., 2009). All three PPAR isoforms mediate these effects via LXRα. However, only PPARγ is known to directly affect LXRα expression via interaction with a peroxisome proliferator response element proximal to the LXRα promotor (Chawla et al., 2001), whereas the exact mechanism remains unknown for the other isotypes and is expected to be indirect (e.g., via effects on its endogenous ligands) (Ogata et al., 2009).

In addition to RXR, retinoid-activated receptors (RARs; i.e., another group of retinoid nuclear receptors) play a role in cholesterol homeostasis. RARs are expressed in a wide variety of tissues and, like other retinoid nuclear receptors, also need to form a heterodimer with RXR (Matsumoto et al., 2007; Jung et al., 2010; Kuntz et al., 2015; Manna et al., 2015; Zhou et al., 2015). For all retinoid nuclear receptors, this heterodimerization allows them to bind to a response element in the promoter region of their target genes, including ABCA1, ABCG1, APOA1, GCK, UCP1 and UCP3, and FGF21 (Puigserver et al., 1996; Solanes et al., 2000; Balmer and Blomhoff, 2002; Cadoudal et al., 2008; Nishimaki-Mogami et al., 2008; Cui et al., 2011; Ayaori et al., 2012; Li et al., 2013; Zhang et al., 2013a, 2015b; Kuntz et al., 2015). PPAR/RXR, LXR/RXR, and RAR/RXR can be activated by agonists for either RXR and any other nuclear receptor (Nishimaki-Mogami et al., 2008; Cui et al., 2011; Kuntz et al., 2015) (Fig. 5). Consequently, RXR exerts pleiotropic effects due to crosstalk of RXR with other nuclear receptors, resulting in the simultaneous activation of multiple converging signaling pathways (Matsumoto et al., 2007; Nishimaki-Mogami et al., 2008). Both retinoic nuclear receptors are activated by their endogenous ligand retinoic acid (RA; i.e., all-trans-RA and 9-cis-RA) (Costet et al., 2003; Koldamova et al., 2003; Nishimaki-Mogami et al., 2008; Zhou et al., 2015). Recently, the relevance of this pathway for RCT was demonstrated in a mouse model of atherosclerosis (apoE^−/− mice on a high-fat diet) by the LXRE-dependent stimulatory effects of 9-cis-RA on macrophage ABCA1 and ABCG1 expression, cholesterol efflux, and HDL-C levels (Zhou et al., 2015). Consequently, RXR as well as the other nuclear factors are important effectors of the cellular cholesterol homeostasis, which makes them valuable targets to induce ABCA1 and ABCG1 expression and to eventually enhance cellular cholesterol efflux and RCT.

B. Liver X Receptor Activation to Induce ATP-Binding Cassette A1 and ATP-Binding Cassette G1 Expression

The increased understanding about the pivotal role of LXR target genes in cholesterol metabolism, including ABCA1 and ABCG1, has enhanced the interest to pharmacologically modulate LXR signaling. One of the first discovered synthetic steroidal LXR agonists, T0901317, a full and potent agonist of both LXRα and LXRβ, appeared to protect against atherosclerotic development in vitro and in animal studies (Terasaka et al., 2003; Thomas et al., 2003; Beyer et al., 2004; Quinet et al., 2004; Wang et al., 2006; Dai et al., 2008; Sato et al., 2008; Larrede et al., 2009; Verschuren et al., 2009; Yan et al., 2010; Honzumi et al., 2011; Chen et al., 2012; Kirchgessner et al., 2015; Manna et al., 2015; Jiang and Li, 2017) (Table 1). T0901317 enhanced the in vitro and in vivo expression of LXR target genes, including ABCA1 and ABCG1, leading to increased cholesterol efflux to apoA-I and HDL and decreased foam cell and atherosclerotic plaque formation (Fukumoto et al., 2002; Murthy et al., 2002; Terasaka et al., 2003; Thomas et al., 2003; Beyer et al., 2004; Miao et al., 2004; Quinet et al., 2004, 2006; Wu et al., 2004; Panzenboeck et al., 2006; Wang et al., 2006; Delvecchio et al., 2007, 2008; Fujiyoshi et al., 2007; Dai et al., 2008; DiBlasio-Smith et al., 2008; Sato et al., 2008; Zanotti et al., 2008; Larrede et al., 2009; Verschuren et al., 2009; Mogilenko et al., 2010; Morrow et al., 2010; Yan et al., 2010; Honzumi et al., 2011; Maejima et al., 2011; Chen et al., 2012; Di et al., 2012; Elali and Hermann, 2012; Jiang et al., 2012; Ma et al., 2014; Kaneko et al., 2015; Kirchgessner et al., 2015; Manna et al., 2015; Tamehiro et al., 2015; Carter et al., 2017; Jiang and Li, 2017; Marinozzi et al., 2017; Monzel et al., 2017; Kaseda et al., 2018). Unfortunately, T0901317 was associated with enhanced lipogenesis, resulting in elevated serum triglyceride levels and hepatic steatosis, which is most likely explained by LXRα-induced activation of the sterol regulatory-binding element protein 1c (SREBP-1c) pathway (Terasaka et al., 2003; Thomas et al., 2003; Beyer et al., 2004; Miao et al., 2004; Quinet et al., 2004, 2006; Delvecchio et al., 2008; DiBlasio-Smith et al., 2008; Sato et al., 2008; Verschuren et al., 2009; Yan et al., 2010; Honzumi et al., 2011; Chen et al., 2012; Kaneko et al., 2015; Kirchgessner et al., 2015; Manna et al., 2015; Carter et al., 2017; Marinozzi et al., 2017). A nonsteroidal LXR agonist, GW3965, also protected against atherosclerosis development (Ruan et al., 2003), but less effectively as T091317 (Sparrow et al., 2002; Bennett et al., 2006; Brunham et al., 2006; Naik et al., 2006; Quinet et al., 2006; Delvecchio et al., 2007; DiBlasio-Smith et al., 2008; Di et al., 2012; Kannisto et al., 2014). Moreover, it amplified SREBP-1c expression leading to increased hepatic and plasma triglyceride levels (Sparrow et al., 2002; Quinet et al., 2004). Various other LXR agonists, including acetyl-podocarpic dimer, LXR-623 (i.e., also known as WAY-252623), ritonavir (i.e., an antiretroviral drug), side-chain modified sterol, ergosterol derivatives, and C24-hydoxylated stigmastane derivatives, also effectively increased ABCA1 expression (Sparrow et al., 2002; DiBlasio-Smith et al., 2008; Pou et al., 2008; Quinet et al., 2009; Marinozzi et al., 2017; Castro Navas et al., 2018). However, they all produced unwanted effects on plasma triglyceride levels by stimulation of SREBP-1c gene transcription and LXR-623 enhanced plasma triglyceride levels, yet all to a lower extent than T0901317 (Sparrow et al., 2002; DiBlasio-Smith et al., 2008; Pou et al., 2008; Quinet et al., 2009; Marinozzi et al., 2017; Castro Navas et al., 2018).

Although the adverse effects limit the clinical use of these LXR agonists, they raised great interest in LXR stimulation as a potential target to stimulate RCT. Novel LXR agonists have been developed, like N,N-dimethyl-3β-hydroxycholenamide (DMHCA), a synthetic oxysterol that has the potential to enhance cholesterol transport in an LXR-dependent manner without increasing plasma triglyceride levels. Moreover, hepatic SREBP-1c mRNA expression was only slightly increased by DMHCA in rat aortic endothelial cells, whereas in macrophages SREBP-1c was reduced (Quinet et al., 2004; Kratzer et al., 2009; Hammer et al., 2017). The potential of this type of LXR agonist was emphasized by methyl-3β-hydroxy-5α,6α-epoxy-cholanate, a compound with similar structure and effects as DMHCA (Yan et al., 2010), and a novel analog of N,N-disubstituted 2,8-diazaspiro[4.5]decane, also known as IMB-151 (Li et al., 2014). This compound also upregulated ABCA1 and ABCG1 expression in RAW264.7 macrophages in an LXRα-dependent manner, whereas SREBP-1c protein expression levels in HepG2 cells were only slightly increased (Li et al., 2014). Other LXR-agonists, ibrolipim (also known as NO-1886) and BMS-779788, increased ABCA1 expression and cellular cholesterol efflux and reduced SREBP-1c expression in vitro and markedly lowered plasma triglyceride levels in vivo (Zhang et al., 2006; Ma et al., 2009; Chen et al., 2010; Tsou et al., 2014; Kick et al., 2015; Kirchgessner et al., 2015). The LXRβ-specific agonist E17110 (i.e., a benzofuran-2-carboxylate analog) and LXRα-specific synthetic chalcone derivative, (E)-1-(3,4-di-isopropoxyphenyl)-3-(4-isopropoxy-3-methoxyphenyl)prop-2-en-1-one, both induced ABCA1 expression in macrophages, whereas an enhanced cholesterol efflux was only described after treatment with E17110 (Li et al., 2016; Teng et al., 2018). Moreover, commonly used proton pump inhibitors (i.e., lansoprazole, omeprazole, pantoprazole) demonstrated to act as LXR agonists, of which lansoprazole was most potent (Cronican et al., 2010). By inducing LXR, proton pump inhibitors enhanced ABCA1 expression in human and mouse cells (Cronican et al., 2010).

Type II DNA topoisomerase inhibitors, which are used as chemotherapeutic medication, also enhanced ABCA1 expression in an LXRα-dependent manner (Zhang et al., 2013a; Tsou et al., 2014; Shen et al., 2015). Inhibition of type II DNA topoisomerase by etoposide and teniposide induced in vitro ABCA1 expression, free cholesterol efflux, and in vivo RCT. Teniposide exerted its effect by inhibiting expression of receptor-interacting protein 140, a corepressor of LXR, and consequently mediated its favorable effects at lower concentrations compared with etoposide (Zhang et al., 2013a). Like other LXR agonists, etoposide and teniposide both enhanced FAS expression, indicating that these inhibitors also activate genes involved in lipogenesis, which could limit their use in the treatment of atherosclerosis (Zhang et al., 2013a) in addition to their unfavorable cytostatic effects. In contrast, 3-(5′-hydroxymethyl-2′furyl)-1-benzyl indazole (YC-1), which in addition to LXR also activates soluble guanylyl cyclase, upregulated ABCA1 expression, enhanced cholesterol efflux, and decreased ox-LDL particle accumulation in macrophages in an LXRα-dependent manner, whereas ABCG1, SR-B1, SR-1, and CD36 expression levels were not altered (Tsou et al., 2014).

Indirect activation of LXR by a synthetic sphingosine analog (i.e., 2-amino-2-[2-(4-n-octylphenyl)ethyl]-1,3-propanediol hydrochloride, FTY720-P) and a synthetic sulfonylurea compound (i.e., G004) also enhanced in vitro expression of ABCA1 and ABCG1 and cholesterol efflux (Blom et al., 2010). FTY720-P upregulated ABCA1 expression in human macrophages via increased production of the endogenous LXR ligand 27-hydroxycholesterol,which subsequently enhanced LXR activity independent of sphingosine 1-phosphate receptor activation (Blom et al., 2010; Qian et al., 2017). In contrast, G004 increased ABCA1 and ABCG1 expression by targeting sirtuin 1 and thereby acting as an atheroprotective agent in vivo (Qian et al., 2017). Ibandronate, a bisphosphonate, also stimulated LXR indirectly, which was reversed upon addition of geranylgeranyl pyrophosphate, suggesting that low levels of this mevalonate pathway intermediate increased LXR expression via a negative feedback mechanism (Strobach and Lorenz, 2003). Anthracyclines, which are used as chemotherapeutic agents, also demonstrated to indirectly activate LXR-ABCA1/ABCG1 pathway by enhancing several oxysterol and cholesterol precursor levels, although a direct binding between anthracyclines and LXR cannot be excluded. Their clinical use is generally strongly limited due to cardiotoxic effects (Monzel et al., 2017), and their cytostatic effect omits the use of these drugs to stimulate RCT clinically. An indirect stimulatory effect on LXR expression by a yet unknown mechanism has also been observed with disodium ascorbyl phytostanol phosphate (FM-VP4), a potential cholesterol-lowering drug (Méndez-González et al., 2010).

The cardiac glycosides, digoxin and ouabain, stimulated cholesterol efflux via LXR-mediated upregulation of ABCA1 expression in cardiomyocytes, but only slightly increased ABCG1 expression (Campia et al., 2012). They also increased cholesterol and ubiquinone synthesis via upregulation of HMG-CoA reductase expression without affecting the intracellular cholesterol concentration. This could be explained by the stimulation of cholesterol efflux in cardiomyocytes, which might also contribute to the beneficial effect of cardiac glycosides in CVD (i.e., next to their antiarrhythmic effects). However, further research is warranted to demonstrate their antiatherosclerotic potential. Finally, the anti–tumor necrosis factor (TNF)-α monoclonal antibody infliximab was linked to a reduced foam cell formation in THP-1 macrophages induced by TNF-α, which was dependent on the reversal of TNF-α–induced inhibition of cholesterol efflux and LXR-α, ABCA1, and ABCG1 mRNA and protein expression (Voloshyna et al., 2014). This could also provide an explanation for the improved vascular function, as recently observed in patients with rheumatoid arthritis with TNF inhibitor therapy (Rongen et al., 2018). Although Voloshyna et al. (2014) demonstrated a lowering effect of TNF-α on LXR-α and ABCA1 levels in THP-1 monocytes, no differences in the mRNA levels of the LXR-target gene ABCA1 and apoA-I–dependent cholesterol efflux were reported in TNF-α–treated C57BL/6 mice peritoneal macrophages (Castrillo et al., 2003). The uncertainty about the exact role of TNF-α on LXR-α and ABCA1 expression is further emphasized by different findings in several cell types. TNF-α reduced LXR-α expression and LXRE activity level in HK-2 proximal tubular cells and Hep3B liver cells, whereas in Caco-2 cells ABCA1 was decreased without attenuation of LXR-α upon stimulation with TNF-α (Wang et al., 2005; Kim et al., 2007; Field et al., 2010). A stimulatory effect on cellular cholesterol efflux was found after lowering TNF-α production in human macrophages by GABA and the GABA agonist topiramate (Yang et al., 2014). These effects could not be observed with baclofen, another GABA agonist (Yang et al., 2014), which questions the direct involvement of GABA in the previously observed effects with GABA agonists. Interestingly, a reversal of TNF-α–induced cholesterol efflux inhibition could also be observed with the immunosuppressant sirolimus, which increased cholesterol efflux in human vascular smooth muscle cells accompanied by an increased ABCA1 expression (Ma et al., 2007). Although the structurally related drug tacrolimus did also increase ABCA1 expression, its effect seemed to be mediated via PPARγ and not via a TNF-α–dependent mechanism (Jin et al., 2004). Because both immunosuppressants have been also associated with elevated plasma cholesterol and triglyceride levels (Wlodarczyk et al., 2005; Kido et al., 2018), their clinical applicability to reduce atherosclerotic risk is limited.

Of all LXR agonists, only two (i.e., BMS-779788, also known as EXEL-04286652 and LXR-623) were studied in a clinical trial (Hong and Tontonoz, 2014). Unfortunately, no published report is available on the completed study of BMS-779788 (Hong and Tontonoz, 2014). LXR-623 stimulated ABCA1 and ABCG1 expression in peripheral blood cells of healthy individuals, which is expected to result in an enhanced RCT. However, the phase I clinical trial was terminated due to unfortunate neurologic adverse effects at the two highest concentrations tested. It remains unknown whether this effect was LXR-mediated or via other unknown off-target mechanisms of LXR-623 (Katz et al., 2009).

C. Peroxisome Proliferator-Activated Receptor Activation to Enhance ATP-Binding Cassette A1 and ATP-Binding Cassette G1 Expression

Fibrates have been used for decades as cholesterol-lowering drugs, but currently they are mainly administered as comedication with statins to treat severe hypertriglyceridemia (>15 mmol × l⁻¹). Although fibrates are mainly agonists of PPARα, they stimulate all three PPAR isoforms with different potencies, except for bezafibrate, which is an equipotent agonist of all isoforms. Gemfibrozil, fenofibrate, bezafibrate, and clofibrate all increased ABCA1 mRNA and protein expression upon PPAR activation in vitro (Oliver et al., 2001; Forcheron et al., 2002; Lin and Bornfeldt, 2002; Guan et al., 2003; Ruan et al., 2003; Thomas et al., 2003; Arakawa et al., 2005; Kooistra et al., 2006; Hossain et al., 2008; Tanabe et al., 2008; Ogata et al., 2009; Kobayashi et al., 2011; Jiang and Li, 2017), which was associated with enhanced apoA-I–mediated cholesterol efflux (Oliver et al., 2001; Ruan et al., 2003; Arakawa et al., 2005; Kooistra et al., 2006; Ogata et al., 2009) (Table 2). Next to an effect on peripheral ABCA1 expression, fenofibrate and gemfibrozil also induced hepatic ABCA1, ABCG1, and ABCG5/8 mRNA expression in a PPARα-dependent manner (Hossain et al., 2008; Rotllan et al., 2011). This positive effect on ABCA1 expression could also in part explain the antiatherosclerotic effects observed with fibrates in clinical trials (Steiner, 2005). The coronary risk reduction found in these trials could also be related to their LDL-C–lowering capacity. However, the therapeutic value of fibrates remains equivocal due to absence of a reduction in total cardiovascular mortality in several trials. Two large-scale intervention studies with fenofibrate (FIELD trial and ACCORD-LIPID trial) in patients with type 2 diabetes mellitus found no differences in coronary events, nonfatal myocardial infarction, and stroke (Keech et al., 2005; Ginsberg et al., 2010). Similar results were reported in two intervention trials with bezafibrate, in which no significant effect was found on both fatal and nonfatal myocardial infarction in the Bezafibrate Infarction Prevention (BIP) study, and no reduction in the incidence of coronary heart disease and of strokes in the LEADER trial (Bezafibrate Infarction Prevention (BIP) Study, 2000; Meade et al., 2002). In contrast, two trials using gemfibrozil [Helsinki Heart Study (HHS) trial and Veterans Affairs–HDL Intervention Trial (VA-HIT) study] showed a significant reduction in the incidence of coronary heart disease and myocardial infarction or cardiovascular death, demonstrating an overall benefit of gemfibrozil treatment (Frick et al., 1987; Rubins et al., 1999).

Like gemfibrozil and fenofibrate, the specific PPARα agonist, Wy14643, enhanced ABCA1 protein expression and apoA-I–mediated cholesterol release in vitro, primarily mediated via LXRα activation (Chinetti et al., 2001; Ruan et al., 2003; Thomas et al., 2003; Beyer et al., 2004; Arakawa et al., 2005; Lee et al., 2008; Maejima et al., 2011). Although beneficial effects on ABCA1 expression and cholesterol efflux were observed after treatment with gemfibrozil, fenofibrate, and Wy14643, all three compounds stimulated PPARα with a relatively low affinity in the micromolar range (Ferri et al., 2017). This initiated the exploration of high-affinity PPARα agonists like GW7647, which effectively increased ABCA1 mRNA. However, no effect of GW7647 on apoA-I–driven cholesterol efflux was observed in THP-1 macrophages (Oliver et al., 2001; Li et al., 2004; Wang et al., 2010; Nakaya et al., 2011). Another high-affinity PPARα agonist, LY518674, did increase both ABCA1 expression and apoA-I–mediated cholesterol efflux via LXRα activation, resulting in increased HDL biogenesis, whereas Wy14563 only increased cholesterol efflux via PPARα stimulation (Chawla et al., 2001; Hossain et al., 2008; Ogata et al., 2009; Ferri et al., 2017). Surprisingly, high aspirin concentrations (≥250 μM) slightly enhanced ABCA1 expression and cholesterol efflux in macrophages in a PPARα-dependent manner (Viñals et al., 2005; Wang et al., 2010).

Statins most likely also affect ABCA1-dependent cholesterol efflux beneficially via indirect stimulation of PPARα (Zanotti et al., 2004, 2006; Argmann et al., 2005; Kobayashi et al., 2011; Maejima et al., 2011; Song et al., 2011; Shimizu et al., 2014; Nicholls et al., 2015). However, the results of a variety of studies are equivocal and report large differences with different statins in vitro and in vivo (e.g., in various cell types, animal models, and patients). In hepatocytes, only pitavastatin seemed to enhance ABCA1 mRNA expression at low micromolar concentrations in all studies (Zanotti et al., 2004, 2006; Kobayashi et al., 2011; Maejima et al., 2011; Song et al., 2011). Mechanistically, PPARα activation was seen in all studies, but LXR was activated as well as suppressed in hepatocytes after statin treatment. Moreover, this effect seems to depend on downstream products of the cholesterol synthesis pathway (e.g., mevalonate, geranylgeranyl pyrophosphate), with a major role for Ras homolog gene family member A, which fully reversed statin-mediated ABCA1 upregulation and ABCA1-mediated cholesterol efflux (Zanotti et al., 2004; Argmann et al., 2005). Not all statins did increase macrophage ABCA1 expression in vitro (Zanotti et al., 2006), but rosuvastatin exposure in vivo increased total and ABCA1-dependent cholesterol efflux (Shimizu et al., 2014). Although cotreatment with atorvastatin and evacetrapib revealed an enhanced cholesterol efflux, whereas single treatment with atorvastatin resulted in a decreased cholesterol efflux, the clinical relevance is limited, because evacetrapib also increased overall atherogenic risk (Nicholls et al., 2017). Therefore, the absence of an effect or even negative effect of statins on ABCA1-mediated cholesterol efflux in some clinical studies (Nicholls et al., 2015, 2017), as well as the high incidence of dose-dependent muscle complaints associated with these drugs, limits their ability to increase RCT.

Next to compounds that beneficially affect PPARα, a vast number of drugs can stimulate PPARγ. Thiazolidinediones are probably the most well-known and widely used PPARγ agonists, due to their antidiabetic properties. Thiazolidinediones, including pioglitazone, ciglitazone, and rosiglitazone, enhanced LXRα expression, which subsequently induced ABCA1 protein expression and cholesterol efflux toward apoA-I in vitro (Chawla et al., 2001; Chinetti et al., 2001; Claudel et al., 2001; Cabrero et al., 2003; Li et al., 2004, 2015; Llaverias et al., 2006; Panzenboeck et al., 2006; Nakaya et al., 2007; Lee et al., 2008; Tanabe et al., 2008; Ogata et al., 2009; Cocks et al., 2010; Ozasa et al., 2011; Wang et al., 2014b, 2015b; Jiang and Li, 2017) (Table 2). Contradictory to these results with pioglitazone in THP-1 macrophages and J774 macrophages, an attenuated or unaffected ABCA1 expression was found in peritoneal macrophages isolated from 15PGJ2-, troglitazone-, and pioglitazone-treated C57BL/6 mice, and in the liver of pioglitazone-treated LDLR^−/− C57BL/6 mice (Ruan et al., 2003; Ogata et al., 2009; Ozasa et al., 2011; Zhao et al., 2015; Jiang and Li, 2017; Silva et al., 2018). Clinically, the cardioprotective antiatherosclerotic effects of thiazolidinediones have been heavily debated (Liebson, 2010). Although increased HDL-C levels were observed after pioglitazone and rosiglitazone treatment, increased cardiovascular morbidity and heart failure have been associated with rosiglitazone (Gerstein et al., 2006; Home et al., 2009; Liebson, 2010; Chandra et al., 2017). Although in various clinical trials cardioprotective effects have been observed with pioglitazone (Chandra et al., 2017), its chronic use is limited by the observed association with an increased risk of developing bladder cancer, bone fractures, and congestive heart failure (Tang et al., 2018). Another PPARγ agonist, GW7845, which is an analog of the PPARα agonist GW7647, did not affect apoA-I–mediated cholesterol efflux, whereas ABCA1 mRNA expression was enhanced (Chawla et al., 2001; Oliver et al., 2001). The PPARγ agonist, GW1929, reduced ABCA1 protein expression in HepG2 cells, whereas ABCG1 gene expression was increased probably due to an interplay between PPARγ and LXRβ, resulting in dissociation of LXRβ from the ABCA1/LXRβ complex (Mogilenko et al., 2010). In contrast, mycophenolic acid, 4010B-30, telmisartan, propofol, and lysophosphatidylcholine exposure did result in an increased in vitro ABCA1 expression and apoA-I–mediated cholesterol efflux, mediated via the PPARγ–LXRα–ABCA1 axis (Hou et al., 2007; Nakaya et al., 2007; Xu et al., 2011; Du et al., 2015b; Ma et al., 2015; Hsu et al., 2018). Lysophoshatidylcholine treatment enhanced apoE secretion from peritoneal macrophages (Hou et al., 2007) and 4010B-30 enhanced apoA-I production, whereas telmisartan and propofol also enhanced ABCG1 expression in macrophages (Nakaya et al., 2007; Du et al., 2015b; Ma et al., 2015). The clinical antiatherosclerotic applicability of propofol seems limited, because of its sedative effect and risk of hypertriglyceridemia induced by the lipid emulsion formulation (Eddleston and Shelly, 1991). Additionally, E3317 enhanced in vitro ABCA1 expression and apoA-I–dependent cholesterol efflux via PPARγ, and a novel thiazolidine, GQ-11, which is a partial PPARγ and PPARα agonist, increased hepatic ABCA1, APOA-I, and HDL mRNA expression in LDLR^−/− C57BL/6 mice (Silva et al., 2018; Wang et al., 2018). Thus, PPARγ agonists show controversial results in the stimulation of ABCA1 expression, which could be due to their relatively low potency for PPARα and high PPARγ potency.

Although explored to a lesser extent, activation of PPARβ/δ may be promising, as activation with GW501515 promoted RCT by enhancing in vitro apoA-I and HDL levels, ABCA1 expression, and apoA-I–mediated cholesterol efflux (Oliver et al., 2001; Sprecher et al., 2007; Ogata et al., 2009) (Table 2). Another PPARβ/δ agonist, GW0742, did not affect ABCA1 or ABCG1 mRNA expression and only stimulated cholesterol efflux in bone marrow–derived macrophages, but not in peritoneal macrophages from LDLR^−/− mice (Li et al., 2004; Briand et al., 2009). Carbaprostacyclin, a PPARβ/δ agonist, only increased cholesterol efflux in macrophages (Chawla et al., 2001).

Combinations of PPAR and LXR agonists were investigated to overcome disadvantages observed with either PPAR or LXR agonists (e.g., increased plasma triglyceride concentrations), which demonstrated to be a promising strategy. Fenofibrate could abolish the LXR-mediated induction of SREBP1 by T0901317 in THP-1 macrophages without affecting the endogenous ABCA1 expression (Thomas et al., 2003). Similar results were observed after coadministration of T0901317 and Wy14643, as shown by the attenuation of T0901317-induced increases in serum and plasma triglycerides, without posing an effect on ABCA1 mRNA levels in C57BL/6 mice (Beyer et al., 2004).

In summary, both PPARα and PPARγ agonists increase ABCA1 and ABCG1 expression in an LXRα-dependent manner, whereas PPARα agonists were more effective inducers of apoA-I levels and apoA-I–mediated cholesterol efflux. Furthermore, coadministration of LXR and PPARα agonists may overcome the adverse effects on plasma triglyceride levels of LXR agonists. However, more studies are needed to elucidate the mechanism and effectivity of such combinatorial strategies for treatment of atherosclerosis.

D. Enhancement of ATP-Binding Cassette A1 and ATP-Binding Cassette G1 Expression by Retinoid X Receptor Agonists

The notion that RXR activates and orchestrates the signaling of other nuclear receptors, including LXR and PPAR, led to the idea that RXR stimulation could be of special interest to stimulate RCT (Costet et al., 2003; Nishimaki-Mogami et al., 2008; Cui et al., 2011; Zhang et al., 2013a; Zhou et al., 2015). Such crosstalk with other nuclear receptors is illustrated by RXR/LXRα heterodimer formation after stimulation of RXR by endogenously synthesized 9-cis-RA (Manna et al., 2015; Zhou et al., 2015) and with synthetic RXR agonists, including PA024 and HX630 (Nishimaki-Mogami et al., 2008; Zhou et al., 2015), which all increased ABCA1 expression (Table 3). In RAW264 cells, PA024, unlike HX630, directly influenced LXRE and positively modified the promoter activity to enhance ABCA1 mRNA expression, resulting in generation of HDL particles and stimulation of cholesterol efflux (Nishimaki-Mogami et al., 2008). In contrast, HX630 is expected to stimulate ABCA1 expression by activating PPARγ/RXR heterodimer, leading to an enhanced LXR expression in RAW264 cells (Nishimaki-Mogami et al., 2008). Similar mechanisms were observed with two other RXR agonists, tributyltin chloride and LG268, which activated PPARγ/RXR and LXRα/RXR signaling. This treatment modulated cellular lipid homeostasis and cholesterol efflux in RAW264 and THP-1 macrophages via increased expression of ABCA1 and ABCG1 (Repa et al., 2000; Chawla et al., 2001; Cui et al., 2011; Sun et al., 2015). Additionally, LG101305 enhanced ABCA1 mRNA expression and cholesterol efflux in macrophages (Claudel et al., 2001). The RXR agonists, bexarotene and methoprene, enhanced ABCA1 expression and cholesterol efflux in astrocytes and increased the overall cerebral cholesterol efflux into the circulation (LaClair et al., 2013; Kuntz et al., 2015; Tachibana et al., 2016).

However, as RXR interacts with many other nuclear receptors, RXR agonists are associated with a wide spectrum of adverse events (Costet et al., 2003), including enhanced lipogenesis due to LXR/RXR heterodimerization (Manna et al., 2015), which severely limits their therapeutic potential to stimulate RCT.

Compounds that stimulate RAR do not suffer from these adverse mechanisms, as they are less ambiguous in the targets they activate. The RAR agonists, 4-[(E)-2-(5,6,7,8-Tetrahydro-5,5,8,8-tetramethyl-2-naphtalenyl)-1-propenyl]benzoic acid (TTNPB) and Am580, enhanced ABCG1 expression by modulating ABCG1 promoter activity, especially by interacting with LXRE-B in macrophages (Ayaori et al., 2012). Furthermore, TTNPB stimulated ABCA1 expression in mouse and human macrophages mediated via a stimulatory effect of RARγ/RXR on the ABCA1 promoter (Costet et al., 2003). Chen et al. (2011b) found opposite effects of TTNPB in astrocytes, indicating that the effects may be cell-type dependent. The beneficial effects in macrophages are, however, most relevant for a potential antiatherosclerotic effect of RAR agonists, which makes them an interesting class of compounds to target RCT.

A. mRNA Degradation as a Post-Translational Mechanism to Regulate ATP-Binding Cassette A1 and ATP-Binding Cassette G1 Expression

Another mechanism regulating plasma membrane abundance of ABCG1 and ABCA1 transporters is the rapid post-translational degradation of their mRNA transcripts. For ABCG1 this depends on the interaction of the 3′ untranslated region of its transcript with miRNAs, resulting in repression of translation or mRNA degradation (Li et al., 2010; Rayner et al., 2011; Rotllan and Fernandez-Hernando, 2012; Lv et al., 2014). Consequently, these small noncoding RNAs appear to be important modulators of gene expression (Rotllan and Fernandez-Hernando, 2012), which upon binding with other mRNA regions can lead to degradation as well as increased mRNA expression and translation (Rotllan and Fernandez-Hernando, 2012) (Fig. 5). Several miRNAs, which are summarized in Table 4, are known to affect ABCA1 mRNA expression. Although less is known about ABCG1 regulation, some miRNAs directly decreased ABCG1 mRNA expression levels (Table 4) Moore et al., 2010; 2011; Fernández-Hernando et al., 2011; Fernández-Hernando and Moore, 2011; et al., 2011, 2012; Hazen and Smith, 2012; Iatan et al., 2012; Rotllan and Fernandez-Hernando, 2012; Sun et al., 2012; Wang et al., 2012, 2014a; Adlakha et al., 2013; Dávalos and Fernandez-Hernando, 2013; de Aguiar Vallim et al., 2013; Kang et al., 2013; Ramárez et al., 2013; Canfrán-Duque et al., 2014; Goedeke et al., 2014; Mao et al., 2014; DiMarco and Fernandez, 2015; He et al., 2015; Mandolini et al., 2015; Yang et al., 2015; Feinberg and Moore, 2016; Ono, 2016; Rotllan et al., 2016; Aryal et al., 2017).

B. Targeting Post-Transcriptional Regulation of ATP-Binding Cassette A1 mRNA

The therapeutic potential of miRNA in the regulation of cholesterol metabolism will not be further addressed in this review, as this was extensively reviewed by others (Moore et al., 2010, 2011; Fernández-Hernando and Moore, 2011; Rotllan and Fernandez-Hernando, 2012; Dávalos and Fernandez-Hernando, 2013; Canfrán-Duque et al., 2014; Goedeke et al., 2014; DiMarco and Fernandez, 2015; Rotllan et al., 2016; Aryal et al., 2017). In this study, other drugable mechanisms involved in the post-transcriptional regulation of ABCA1 and ABCG1 mRNA expression will be discussed (Moore et al., 2011; Rotllan and Fernandez-Hernando, 2012; Dangwal and Thum, 2014; van Rooij and Kauppinen, 2014; Rotllan et al., 2016). One of these mechanisms is mediated by the cellular energy sensor AMP-activated protein kinase (AMPK), through its regulation of cholesterol metabolism. AMPK activation by 5-aminoimidazole-4-carboxyyamide ribonucleoside (AICAR; an AMP mimetic) has led to enhanced ABCG1 mRNA and protein expression, reduced ox-LDL uptake, and enhanced HDL-mediated cholesterol efflux (Table 5). These effects were found to be independent of LXRα, but mediated through ABCG1 mRNA 3′ untranslated region without affecting ABCA1, SR-A, CD36, and SR-B1 protein expression (Li et al., 2010). In contrast, Kemmerer et al. (2016) demonstrated predominant LXRα-mediated upregulation of ABCA1 mRNA expression in macrophages after exposure to AICAR and the allosteric AMPK activators A769662 and salicylate (Li et al., 2010). The increased ABCA1 expression by A769662 and AICAR was proposed to be mediated via inhibition of extracellular signal-regulated kinase (ERK) or the mammalian target of rapamycin pathways (Kemmerer et al., 2016). Indeed, ERK signaling is described as a negative regulator of ABCA1 protein stability (Mogilenko et al., 2010). Compounds that affect the ERK pathway will be discussed in Section VI.B (Pharmacological Inhibition of ATP-Binding Cassette A1 and ATP-Binding Cassette G1 Protein Degradation). Thus, AMPK agonists could increase cholesterol efflux via a dual mechanism, either by enhancing ABCG1 mRNA stability or via increased mRNA expression of ABCG1 and ABCA1.

A. The Role of ATP-Binding Cassette A1 and ATP-Binding Cassette G1 Reuptake and Degradation in the Regulation of Their Plasma Membrane Abundance

Like other plasma membrane transporters and receptors, ABCA1 and ABCG1 protein abundance is dependent on their cellular reuptake and degradation (Yokoyama et al., 2012; Wang et al., 2017). After internalization, ABCA1 is mainly degraded via lysosomal and ubiquitin- or calpain-mediated proteolysis (Yokoyama et al., 2012; Huang et al., 2015; Wang et al., 2017) (Fig. 5). Although the function of lysosomal and ubiquitin-mediated proteolytic degradation in ABCA1 turnover and activity needs to be elucidated, the importance of calpain-mediated ABCA1 breakdown has been established, especially in THP-1 macrophages (Yokoyama et al., 2012; Wang et al., 2017).

ABCA1 proteolysis is initiated by calpain binding to a calpain-specific cleavage sequence [Pro-Glu-Ser-Thr (PEST)] (Iwamoto et al., 2010; Huang et al., 2015; Wang et al., 2017), which is located within the large cytosolic loop of ABCA1 (Reiss and Cronstein, 2012) (Fig. 5). The relevance of this region was illustrated by deletion of the PEST sequence, which prevented the breakdown of ABCA1 (Yokoyama et al., 2012). This region is also important for ABCA1 half-life, which is 1 to 2 hours in the absence of helical apolipoproteins, such as apoA-I, apoA-II, and apoE (Iwamoto et al., 2010; Yokoyama et al., 2012). Phosphorylation of the PEST sequence is suggested to reduce ABCA1 half-life (Iwamoto et al., 2010; Yokoyama et al., 2012), whereas apoA-I binding or pre-exposure (i.e., apoA-I presence before internalization of prebiotinylated surface ABCA1 expressed on THP-1) prolongs the half-life by increasing the resistance of ABCA1 calpain-mediated proteolysis (Lu et al., 2008; Iwamoto et al., 2010; Yokoyama et al., 2012). This did not affect ABCA1 internalization in endosomes, but resulted in a larger portion of ABCA1 that is recycled from the endosomes to the plasma membrane. Besides apoA-I, calpain-mediated ABCA1 degradation is regulated by various endogenous ligands, including calmodulin, calpeptin, calpastatin, protein kinase C (PKC)α, and heme oxygenase-1 (HO-1). Calmodulin interacts with a region in the large cytoplasmic loop near the PEST sequence, which protects ABCA1 against calpain-mediated proteolysis (Iwamoto et al., 2010). In contrast to calmodulin, calpastatin and the nuclear factor-like (Nrf)2–HO-1 axis directly inhibit calpain activity and decrease ABCA1 degradation (Tsai et al., 2010; Wissel et al., 2015). Finally, phosphorylation of the PEST region by activation of apelin-13–mediated PKCα also protects ABCA1 against calpain-mediated proteolysis (Liu et al., 2013). In short, inhibition of calpain activity and ABCA1 internalization, or stimulation of its turnover rate, may result in an increased ABCA1-mediated cellular cholesterol efflux and increased HDL biosynthesis.

B. Pharmacological Inhibition of ATP-Binding Cassette A1 and ATP-Binding Cassette G1 Protein Degradation

As proteolytic breakdown of ABCA1 and ABCG1 plays a significant role in the regulation of their plasma membrane abundance, different pharmacological treatments have been investigated and aimed at increasing RCT via this mechanism (Table 6). One is the use of inhibitors of thiol proteases (e.g., calpain), like leupeptin and N-acetyl-leu-leu-norleucinal (ALLN), which both prevented ABCA1 degradation (Arakawa and Yokoyama, 2002; Yokoyama, 2004). In contrast, other thiol protease inhibitors (i.e., pepstatin A, aprotinin, and phosphoramidon) did not affect ABCA1 degradation (Arakawa and Yokoyama, 2002), questioning the specificity of the effect of leupeptin and ALLN. In addition, ABCA1 protein expression levels were not affected by nonspecific protease inhibitors like the proteasome inhibitor lactacystin (Arakawa and Yokoyama, 2002).

The stabilizing role of PKC on ABCA1 also seems to hold for ABCG1, as the ABCG1-dependent cholesterol efflux is decreased by the PKC inhibitors calphostin C and PKC19-36 (Gelissen et al., 2012). As mentioned above, apelin-13, a cleaved peptide from the adipocytokine apelin, phosphorylated ABCA1 in THP-1 macrophages, leading to ABCA1 stabilization and enhanced cholesterol efflux (Liu et al., 2013). Remarkably, several other PKC inhibitors (i.e., Gö6983, Gö6976, rottlerin, and doxazosin) enhanced ABCA1 mRNA and protein expression as well as cholesterol efflux in macrophages via an inhibitory effect on protein kinase D (PKD), with the strongest effect by Gö6983 (Iwamoto et al., 2007, 2008; Tsunemi et al., 2014). This effect is probably mediated by a decreased phosphorylation of activator protein 2α, which represses transcription of ABCA1 as phosphorylated activator protein 2α binds to the promoter region of ABCA1 (Iwamoto et al., 2007, 2008; Remaley, 2007). The potential of this promoter region as a target to increase cellular cholesterol efflux is emphasized by pyrrole–imidazole polyamides that bind to this region and enhanced ABCA1 mRNA and protein levels as well as cholesterol efflux in 3T3-L1 adipocytes and RCT in C57BL/6 mice (Tsunemi et al., 2014). In this study, ABCA1 levels were increased via PKD inhibition, which illustrates the importance of the balance between PKC and PKD activity.

A more specific calpain inhibitor, triacetyl-3 hydroxyphenyl-adenosine (IMM-H007), increased ABCA1 plasma membrane expression in THP-1 cells and enhanced ABCA1-mediated cholesterol efflux to apoA-I (Huang et al., 2015). In apoE^−/− mice, IMM-H007 delayed ABCA1 protein degradation, promoted ABCA1 cell surface localization, enhanced RCT, and suppressed atherosclerotic lesion development (Huang et al., 2015). Interestingly, IMM-H007 activated AMPK, which may have also contributed to these effects via ABCA1 stabilization, as described above. In summary, the adenosine analog IMM-H007 may be a promising candidate to upregulate ABCA1 expression with proven capacity to increase RCT in vivo.

More indirect strategies to inhibit calpain, including the induction of HO-1 production, have also been shown to suppress ABCA1 degradation. Tertiary-butyl-hydroquinone, a synthetic phenolic antioxidant, mediates these effects by activation of Nrf2 via its dissociation from kelch-like ECH-associated protein (Keap) 1, translocation of Nrf2 toward the nucleus, and activation of antioxidant responsive element (ARE)-dependent transcription of HO-1 gene. The increased production of HO-1, which is an endogenous inhibitor of calpain, resulted in reduced ABCA1 degradation (Lu et al., 2013).

In addition, two oxidative products of probucol, spiroquinone and diphenoquinone, suppressed ABCA1 and ABCG1 degradation both in vitro and in vivo, and thereby increased RCT and reduced lipid deposition in atherosclerotic plaques in vivo (Yokoyama, 2004; Arakawa et al., 2009; Lu et al., 2016; Yakushiji et al., 2016). These effects are expected to be due to disruption of the caveolin-1 interaction with ABCA1 by spiroquinone and diphenoquinone, resulting in the stabilization of ABCA1 protein against degradation (Arakawa et al., 2009). This mechanism is also expected to mediate the increase in ABCA1 abundance by ezetimibe, which lowers caveolin-1 expression through suppression of SREBP-1 expression (Lu et al., 2016). Another mechanism directly affecting ABCA1 stability is the reduction of its disulfide bonds (Bungert et al., 2001), as described with N-acetyl-cysteine (Machado et al., 2014). N-acetyl-cysteine treatment is, however, associated with an increased cellular cholesterol efflux, which was mediated by increased ABCG1 expression and HDL-dependent cholesterol efflux, whereas ABCA1 expression and apoA-I–dependent cholesterol efflux were decreased in J774 cells (Machado et al., 2014).

Finally, calpain-mediated ABCA1 proteolysis can also be inhibited by alteration of the phosphorylation status of the ABCA1 PEST sequence, as described above (Reiss et al., 2008; Chen et al., 2011a). This could explain the beneficial effects of the ERK1/2 inhibitors, PD98059 and U0126, on ABCA1 and ABCG1 mRNA and protein expression, cholesterol efflux in macrophages, and RCT in apoE^−/− mice (Mulay et al., 2013; Kemmerer et al., 2016; Xue et al., 2016; Zhang et al., 2016). Besides a direct effect, ERK1/2 inhibition could also increase ABCA1 and ABCG1 mRNA stability and expression via an LXRα-dependent mechanism, possibly, by enhancing the LXRα to LXRE-A binding, without affecting PPARγ and LXRα expression (Xue et al., 2016; Liu et al., 2019). In addition to an increased cholesterol efflux, lipid deposition and CD36 expression were suppressed upon U0126 treatment in ox-LDL–stimulated macrophages (Xue et al., 2016). The effect of ERK1/2 inhibition is cell-type dependent, as ABCA1 and ABCG1 protein stability was reduced in CHO and HuH7 cells after ERK1/2 inhibition (Gelissen et al., 2012; Mulay et al., 2013). In addition, the beneficial effect on ABCA1 and ABCG1 expression of zerumbone (i.e., a wild ginger-derived natural compound) and tanshinone IIA was abolished in THP-1 macrophages by ERK inhibitor PD98059 (Liu et al., 2014; Mostafa et al., 2015, 2016; Zhu and Liu, 2015; Yin et al., 2016). Moreover, LXRα, PPARα, ABCA1, and ABCG1 expression were not altered after inhibition of c-Jun N-terminal kinase and P38 mitogen-activated protein kinase (MAPK) phosphorylation by SP600125 and SB203580, respectively (Liu et al., 2019). The uncertainty about the precise role of ERK inhibition in cellular cholesterol efflux is emphasized by the observation that stimulation of MAPK/ERK increased LXR-mediated ABCA1 expression, as found for the glucagon-like peptide 1 receptor agonist exendin-4 and the dipeptidyl peptidase-4 inhibitor vildagliptin (Mostafa et al., 2015, 2016; Yin et al., 2016). Similar effects upon MAPK/ERK stimulation were observed using the melanocortin 1 receptor (MC1-R) agonist MSG606, which upregulates cholesterol efflux and ABCA1 and ABCG1 protein levels in bone marrow–derived macrophages, probably via ERK stimulation without increasing cAMP levels (Rinne et al., 2017). The MC1-R agonist, LD211, which stimulated ERK1/2 and p38 MAPK phosphorylation along with a strong increase in cAMP levels, also positively affected cholesterol efflux (Rinne et al., 2017). However, the exact mechanism behind the effect of MC1-R agonists on ERK1/2 is yet unidentified.

Stimulation of the ERK1/2–PPARγ–LXLα axis has been associated with the favorable effects of the G protein–coupled receptor (GPR)109 agonist, niacin, on triglyceride and total, LDL, and HDL cholesterol levels in plasma. Moreover, it has been shown that this first clinically available cholesterol-lowering drug may upregulate ABCA1 expression, stabilize newly produced HDL, and promote cholesterol efflux possibly via cAMP/protein kinase A (PKA) and PPARγ–LXRα pathways (Rubic et al., 2004; Siripurkpong and Na-Bangchang, 2009; Wu and Zhao, 2009; Yvan-Charvet et al., 2010a; Zhang et al., 2012; Connolly et al., 2013; Gaidarov et al., 2013). The GPR109A agonists, acifran and acipimox, but not isoniacin, also activated ERK1/2 and Ca²⁺ flux (Gaidarov et al., 2013). Although niacin exerted beneficial effects, it is also associated with vasodilation and flushing side effects, possibly mediated via the secretion of prostaglandins as a result of activation of GPR109A (Rubic et al., 2004; Gaidarov et al., 2013). Besides niacin, the full GPR109A agonist, MK-1903, gave similar effects on ABCA1- and apoA-I–mediated cholesterol efflux, whereas the effect on ABCG1 expression was lower compared with niacin (Gaidarov et al., 2013). Another GPR109A agonist, MK-0354, induced neither flushing nor activation of GPR109A signaling in macrophages and HDL modulation (Gaidarov et al., 2013). This suggests that GPR109A agonists that do not cause flushing are probably also unable to exert an antiatherogenic effect.

ERK1/2 inhibition by the commonly used calcium channel blocker nifedipine leads to in vitro anthiatherogenic effects at clinically relevant low nanomolar concentrations via inhibition of monocyte chemoattractant protein-1 and stimulation of ABCA1 expression (Ishii et al., 2010). Two other calcium channel blockers, verapamil and nicardipine, also increased ABCA1- but not ABCG1-mediated cholesterol and phospholipid efflux at suprapharmacological concentrations in the low micromolar range (Suzuki et al., 2004). Verapamil was able to enhance ABCA1 promotor activity in an LXR-independent manner (Suzuki et al., 2004). In summary, ERK1/2 has the potential to beneficially affect cellular cholesterol efflux, but some of the controversies of this relation will need to be clarified to assess its true therapeutic relevance.

To conclude, inhibition of ABCA1 and ABCG1 degradation has demonstrated to be a promising strategy to enhance their plasma membrane expression, in combination with beneficial in vivo effects on RCT and the progression of atherosclerotic plaque formation. However, many compounds are rather unspecific, which may limit their therapeutic applicability due to adverse effects as a consequence of off-target activities.

A. cAMP Is a Potent Regulator of ATP-Binding Cassette A1 Function and Expression

ABCA1 expression and function can also be enhanced by increasing cellular cAMP levels (Sakr et al., 1999). The transporter is phosphorylated upon cAMP-mediated activation of PKA, which leads to an increased capacity to interact with apoA-I and subsequent apoA-I–dependent cellular cholesterol efflux (Haidar et al., 2002, 2004) (Fig. 5). PKA has two major phosphorylation sites, Ser-1042 and Ser-2054, which are located at the NBDs of ABCA1 (See et al., 2002). These effects are most likely specific to macrophages (Bortnick et al., 2000; Suzuki et al., 2004). Interestingly, phosphorylation can also be autoregulatory, as apoA-I binding to ABCA1 has been demonstrated to stimulate adenylate cyclase (AC) activity by an unknown mechanism and boost intracellular cAMP levels (Haidar et al., 2002). Besides influencing cholesterol efflux, cAMP is also involved in the regulation of glucose, lipid, and cholesterol metabolism (Haidar et al., 2002, 2004; Tresguerres et al., 2011), and its cellular concentrations are a result of a delicate balance between its production from ATP by AC and its hydrolysis into AMP by phosphodiesterases (PDEs) and cellular efflux by the ABC transporters multidrug resistance protein 4/ABCC4 and multidrug resistance protein 5/ABCC5 (Lin and Bornfeldt, 2002; Wielinga et al., 2003; Copsel et al., 2011; Tresguerres et al., 2011) (Fig. 5). The activity of AC is regulated by different adenosine receptor subtypes, of which A₁ and A₃ receptors have an inhibitory and A_2a and A_2b a stimulatory effect (Reiss and Cronstein, 2012). Consequently, A_2a receptor activation enhanced cAMP levels and promoted RCT via ABCA1-mediated cholesterol efflux (Bingham et al., 2010).

B. Stimulation of cAMP Levels Enhances ATP-Binding Cassette A1–Mediated Cholesterol Efflux

Treatment with cAMP analogs, 8-bromo-cAMP, 8-(4-chlorophenylthio)adenosine-cAMP, (Bu)2cAMP, and dibutyryl cAMP, stimulated apoA-I–dependent cholesterol efflux in different cell types (Sakr et al., 1999; Abe-Dohmae et al., 2000; Haidar et al., 2002; Lin and Bornfeldt, 2002; Kellner-Weibel et al., 2003; Bingham et al., 2010; Gaidarov et al., 2013; Manna et al., 2015) (Table 7). The 8-bromo-cAMP enhanced cholesterol efflux toward apoA-I, most likely via increased PKA-mediated ABCA1 phosphorylation (Iwamoto et al., 2008; Katz et al., 2009), as PKA inhibition by H89 reversed the beneficial effects of 8-bromo-cAMP on D4-F–mediated cholesterol efflux in RAW264.7 macrophages (Bingham et al., 2010; Xie et al., 2010). Similar results were observed with 8-(4-chlorophenylthio)adenosine-cAMP, which elevated ABCA1 protein expression (Sakr et al., 1999; Haidar et al., 2002; Lin and Bornfeldt, 2002; Kellner-Weibel et al., 2003). A direct PKA agonist, 6-Benz-cAMP, also stimulated ABCA1 protein expression and cellular cholesterol expression, which emphasizes the role of PKA-dependent phosphorylation in enhancing ABCA1-mediated cholesterol efflux capacity (Bingham et al., 2010). Surprisingly, PKA seemed to have contradictory effects on ABCG1 as inhibition of its activity by H89 or KT5720 in CHO-K1 cells enhanced ABCG1-mediated cholesterol efflux and stabilized ABCG1. However, this effect was only observed in cells overexpressing a mutant ABCG1 containing a 12-amino-acid internal segment (Gelissen et al., 2012). Interestingly, treatment with the platelet inhibitor dipyradimole, which increases intracellular cAMP levels, demonstrated antiatherosclerotic effects in the ESPRIT trial (Halkes et al., 2006)).

In line with a central role for cAMP, modulation of AC and PDEs has demonstrated to be another useful strategy to enhance RCT. The AC activator forskolin enhanced RCT via elevated ABCA1 protein levels and phosphorylation (Haidar et al., 2002; Lin and Bornfeldt, 2002). AC stimulation via A_2a receptor activation with CGS-21680 and ATL313 enhanced cholesterol efflux and ABCA1 protein expression (Bingham et al., 2010; Voloshyna et al., 2013). Moreover, stimulation with 8-pcPT-2′-O-Me-cAMP of Epac, a signaling molecule downstream of A_2a receptor activation, showed similar results. A_2a receptor activation decreased foam cell formation in THP-1–derived macrophages by regulation of cholesterol influx and efflux (Bingham et al., 2010). Activation of this adenosine receptor is also associated with an increased ABCA1 expression in peripheral blood mononuclear cells by methotrexate, which is used to treat cancer and autoimmune diseases (Reiss et al., 2008; Chen et al., 2011a).

Increased cAMP levels as a result of reduced breakdown with PDE4 inhibitors (3-isobutyl-1-methylxanthine (IBMX), rolipram, and cilomast) also resulted in elevated ABCA1 expression levels and apoA-I–dependent cholesterol efflux in macrophages (Haidar et al., 2002; Lin and Bornfeldt, 2002). Similar effects were observed with the PDE3 inhibitor cilostazol through increased ABCA1 and ABCG1 expression and cholesterol efflux in vitro and in vivo (Nakaya et al., 2010).

In summary, cAMP analogs, A_2a receptor agonists, and other compounds that increase intracellular cAMP levels have the potential to induce ABCA1 expression, which could increase apoA-I–mediated cholesterol efflux. The involvement of cAMP in many different metabolic pathways (i.e., glucose, lipid, and cholesterol metabolism) may though render these strategies vulnerable to adverse effects.