Role of ATP-binding cassette transporters in brain lipid transport and neurological disease

Authors

  • Woojin Scott Kim,

    1. Prince of Wales Medical Research Institute, Sydney, New South Wales, Australia
    Search for more papers by this author
  • Cynthia Shannon Weickert,

    1. Prince of Wales Medical Research Institute, Sydney, New South Wales, Australia
    2. School of Psychiatry, Faculty of Medicine, University of New South Wales, Sydney, New South Wales, Australia
    3. Schizophrenia Research Institute, Faculty of Medicine, University of New South Wales, Sydney, New South Wales, Australia
    Search for more papers by this author
  • Brett Garner

    1. Prince of Wales Medical Research Institute, Sydney, New South Wales, Australia
    2. School of Medical Sciences, Faculty of Medicine, University of New South Wales, Sydney, New South Wales, Australia
    Search for more papers by this author

Address correspondence and reprint requests to Dr Brett Garner, Prince of Wales Medical Research Institute, Sydney, Randwick, New South Wales 2031, Australia.
E-mail: brett.garner@unsw.edu.au

Abstract

The brain is lipid-rich compared to other organs and although previous studies have highlighted the importance of ATP-binding cassette (ABC) transporters in the regulation of lipid transport across membranes in peripheral tissues, very little is known regarding ABC transporter function in the CNS. In this study, we bring together recent literature focusing on potential roles for ABC transporters in brain lipid transport and, where appropriate, identify possible links between ABC transporters, lipid transport and neurological disease. Of the 48 transcriptionally active ABC transporters in the human genome, we have focused on 13 transporters (ABCA1, ABCA2, ABCA3, ABCA4, ABCA7 and ABCA8; ABCB1 and ABCB4; ABCD1 and ABCD2; ABCG1, ABCG2, and ABCG4) for which there is evidence suggesting they may contribute in some way to brain lipid transport or homeostasis. The transporters are discussed in terms of their location within brain regions and brain cell types and, where possible, in terms of their known functions and established or proposed association with human neurological diseases. Specific examples of novel treatment strategies for diseases, such as Alzheimer’s disease and X-linked adrenoleukodystrophy that are based on modulation of ABC transporter function are discussed and we also examine possible functions for specific ABC transporters in human brain development.

Abbreviations used
ALDRP

ALDP-related protein

APP

amyloid precursor protein

BBB

blood–brain barrier

HDL

high density lipoprotein

LCAT

lecithin cholesterol acyltransferase

LDLr

low density lipoprotein receptor

LXRs

liver X receptors

NBDs

nucleotide binding domains

PLTP

phospholipid transfer protein

PMP

peroxisomal membrane protein

RXRs

retinoid X receptors

SM

sphingomyelin

TMDs

transmembrane domains

VLCFA

very-long chain fatty acids

X-ALD

X-linked adrenoleukodystrophy

The brain is lipid-rich compared to other organs and although previous studies have highlighted the importance of ATP-binding cassette (ABC) transporters in the regulation of lipid transport across membranes in peripheral tissues, very little is known regarding ABC transporter function in the CNS. The purpose of the present review is to bring together recent literature focusing on potential roles for ABC transporters in brain lipid transport and, where appropriate, identify possible links between ABC transporters, lipids and neurological disease.

The human genome contains 48 transcriptionally active ABC transporter genes divided into seven subfamilies (designated A–G) based on sequence homology and the organization of their nucleotide binding domains (NBDs). Eukaryotic ABC transporters typically exist as either full-length transporters, consisting of two hydrophobic transmembrane domains (TMDs) each containing six α-helices and two NBDs (Fig. 1), or as half transporters, that contain a single TMD and NBD, and require homo- or heterodimer formation to function. There are exceptions in the domain structure of the full-length transporters; for example, several members of the ABCC subfamily contain an additional N-terminal TMD with five α-helices (Bakos et al. 1996; Stride et al. 1996). ABC transporters use energy derived from the hydrolysis of ATP to move a vast array of substrates across membranes against their concentration gradients (Hung et al. 1998; Klein et al. 1999; Higgins and Linton 2004). These substrates include xenobiotics, metals, inorganic ions, carbohydrates, vitamins, amino acids, peptides and lipids (Borst et al. 2000; Dean et al. 2001; Borst and Elferink 2002; Takahashi et al. 2005). ABC transporters are ubiquitously expressed in mammalian cellular membranes including the plasma membrane and intracellular membranes of endosomes, lysosomes, peroxisomes, multivesicular bodies, mitochondria, endoplasmic reticulum and Golgi. The rather specific substrate specificity described for a large number of mammalian ABC transporters underscores their tissue-specific expression in many cases.

Figure 1.

 Schematic representation of ABCA1 domain structure. Full-length ABC transporters are typically composed of two hydrophobic transmembrane domains each containing six α-helices (shown in blue) and two nucleotide binding domains (NBD-1 and NBD-2). Each NBD is characterized by Walker A and Walker B motifs (shown in red) and an ABC signature sequence (shown in yellow). ATP binds to each of the NBDs to provide the energy required to transport substrates across the membrane (indicated by vertical arrow). ABCA1 also contains two large extracellular loops with multiple sites for N-linked glycosylation (shown in green).

A detailed screen of tissue ABC transporter gene expression was published by Langmann et al. in 2003 (Langmann et al. 2003). In this report, the expression of 47 ABC transporter genes was measured in 20 human tissues by quantitative real-time PCR (qRT-PCR) and the data indicated reasonably distinct patterns of ABC transporter mRNA abundance for each of the tissues examined. Organs with particularly high ABC transporter expression included the adrenal gland, liver, lung, and reproductive organs (Langmann et al. 2003). Several ABC transporters are also highly expressed in the brain and recent investigations have aimed at determining both cell type-specific and brain region-specific ABC transporter expression in order to gain insights into their possible functions (Tachikawa et al. 2005; Kim et al. 2006).

Many ABC transporters, particularly in the A and G subfamilies, transport lipids, such as sterols, phospholipids and bile acids (Broccardo et al. 1999; Borst et al. 2000; Schmitz et al. 2000, 2001; Pohl et al. 2005; Takahashi et al. 2005; Kusuhara and Sugiyama 2007). There has been increasing interest in ABC transporters involved in lipid transport since it was recognized that several genetic diseases are caused by ABC transporter mutations. For example, ABCA1 plays a critical role in peripheral lipid transport by regulating cholesterol efflux from the plasma membrane to the lipid acceptor protein apolipoprotein-AI (apoA-I) and mutations in the ABCA1 gene cause Tangier disease which is associated with cellular cholesterol accumulation, premature atherosclerosis and peripheral neuropathy (Antoine et al. 1991; Serfaty-Lacrosniere et al. 1994). Mutations in another transporter, ABCA3, are associated with fatal neonatal pulmonary surfactant deficiency as ABCA3 expressed by lung alveolar type II cells mediates secretion of pulmonary surfactant lipids (Mulugeta et al. 2002; Shulenin et al. 2004). While mutations in ABCA4, which transports retinyldiene phospholipid complexes within the rod outer segment of the retina, can cause Stargardt disease and other related eye disorders (Allikmets et al. 1997). These examples highlight the fact that the loss of function of individual lipid transporters can have devastating organ-specific consequences.

The brain is the most lipid-rich organ in the body. Gray matter contains ∼ 40% lipid (as % of dry weight) whereas white matter contains ∼ 65% lipid; largely reflecting myelin enrichment which itself is ∼ 78% lipid (Johnson et al. 1948; O’Brien and Sampson 1965; Sastry 1985). Given that several ABC transporters predicted to regulate lipid transport are expressed in the brain, it is very likely they will be found to play key roles in brain lipid transport and homeostasis. Using cholesterol as an example, it is clear that tight regulation of brain cholesterol homeostasis is crucial for neurological function and that dysregulation of cholesterol homeostasis can contribute to neurodegeneration (Dietschy and Turley 2001; Puglielli et al. 2003). It follows that modulation of endogenous ABC transporter expression level or genetic mutations that alter transporter function could have a significant impact on intracellular and intercellular lipid transport, brain function and susceptibility to neurodegenerative diseases.

ABC transporters included in this review were selected based on their known expression in the brain (or isolated brain cells) and their known or predicted role in lipid transport. In certain cases, ABC transporters other than those in the more specialized ABCA/G lipid transporter subfamilies are mentioned only if there is evidence that they also have a lipid transport function. Therefore, of the 48 ABC transporters present in humans, the current review focuses mostly on 13 ABC transporters for which there is evidence that they may contribute in some way to brain lipid transport or homeostasis (Table 1).

Table 1.   Human brain ABC transporters and their role in lipid transport1
Gene name/Ensembl ID/SwissProt ID Chromosome/locationPredicted MW kDa/GlcNAc sites2Confirmed or potential disease association3Brain expression4Lipids transported
  1. 1This table provides information on ABC transporters that have established functions in brain lipid transport and on transporters that are known to transport lipids and are also expressed (at least at the mRNA level) in the brain. Therefore, not all transporters listed have been shown to be expressed at the protein level in the brain. Similarly, not all transporters have been shown to function in brain lipid transport. See text for further details.

  2. 2CAD, coronary artery disease; FHA, familial hypoalphalipoproteinaemia; RDS, respiratory distress syndrome; ARMD, age-related macular degeneration; PFIC, progressive familial intrahepatic cholestasis; X-ALD, X-linked adrenoleukodystrophy.

  3. 3Dot legend for expression level: Very high 5, High 4, Medium 3, Low 2, Very low 1. Expression is based primarily on mRNA data derived from (Annilo et al. 2001; Bhongsatiern et al. 2005; Chen et al. 1996; Choudhuri et al. 2003; Cooray et al. 2002; Cordon-Cardo et al. 1989; Croop et al. 1997; Engel et al. 2001; Fitzgerald et al. 2007; Fouquet et al. 1997; Fukumoto et al. 2002; Ikeda et al. 2003; Kemp and Wanders 2007; Kielar et al. 2001; Kim et al. 2005, 2006; Koldamova et al. 2003; Langmann et al. 1999, 2003; Matsumoto et al. 2003; Nagase et al. 1998; Nakamura et al. 2004; Oldfield et al. 2002; Savary et al. 1996; Seetharaman et al. 1998; Stahlman et al. 2007; Su et al. 2002; Tachikawa et al. 2005; Tatsuta et al. 1992; Tishler et al. 1995; Troffer-Charlier et al. 1998; Tsuruoka et al. 2002; Vulevic et al. 2001; Wang et al. 2003; Zhang et al. 2003; Zhao et al. 2000; Zhou et al. 2001). Data are provided only as an approximation of whole brain gene expression levels.

  4. 4Ch, cholesterol; PL, phospholipids; SM, sphingomyelin; GSL, glycosphingolipids; PG, phosphatidylglycerol; Ret-PE, retinyldiene phosphatidylethanolamine; PC, phosphatidylcholine; PS, phosphatidylserine; LTC4, leukotriene C4; VLCFA, very long-chain fatty acids.

ABCA1/ENSG00000165029/O954779/106 583 104–106 730 339254/21CAD, FHA, Tangier, Alzheimer••Ch, PL
ABCA2/ENSG00000107331/Q9BZC79/139 021 507–139 043 151270/25Alzheimer•••••SM, myelin lipids
ABCA3/ENSG00000167972/Q9975816/2265,878–2330 595191/0RDS••PC, PG
ABCA4/ENSG00000198691/P783631/94 230 981–94 359 279256/15Stargardt ARMDRet-PE
ABCA7/ENSG00000064687/Q8IZY219/992 361–1016 424234/1••Ch, PL
ABCA8/ENSG00000141338/O9491117/64 375 028–64 463 128179/5LTC4
ABCB1 (MDR1)/ENSG00000085563/P081837/86 970 884–87 180 500142/3Parkinson••PL, SM, GSL
ABCB4 (MDR3)/ENSG00000005471/P214397/86,869,302–86,942,991142/2PFICPC
ABCD1 (ALDP)/ENSG00000101986/P33897X/152 643 517–152 663 410 83/1X-ALD••VLCFA
ABCD2 (ALDRP)/ENSG00000173208/Q9UBJ212/38,232,814–38,300,237 83/2••VLCFA
ABCG1/ENSG00000160179/P4584421/42 509 335–42 590 421 76/0Alzheimer•••Ch
ABCG2 (BCRP1)/ENSG00000118777/Q9UNQ04/89 230 441–89 299 035 72/0Breast cancer••PC, PS
ABCG4/ENSG00000172350/Q9H17211/118 524 960–118 538 582 72/1•••••Ch

ABCA1

The A subfamily of ABC transporters consists of 12 members all of which are full-length transporters with two homologous halves, each containing a single hydrophobic TMD and NBD (Fig. 1) (Kaminski et al. 2006; Wenzel et al. 2007). ABCA1 is widely expressed, with particularly high expression in the adrenal gland and uterus and moderate expression in the liver (Luciani et al. 1994; Langmann et al. 2003). Hepatic ABCA1 plays an essential role in regulating high density lipoprotein (HDL) metabolism (Singaraja et al. 2006) and as mentioned earlier, defects in ABCA1 cause Tangier disease (Bodzioch et al. 1999; Brooks-Wilson et al. 1999; Lawn et al. 1999). Tangier disease is characterized by an absence of plasma HDL and a severely impaired capacity for cholesterol to be removed from peripheral tissues. Cholesterol therefore accumulates in the tonsils, liver, spleen, and artery wall and this is associated with premature coronary artery disease, hepatosplenomegaly, peripheral neuropathy and progressive muscle wasting and weakness (Bodzioch et al. 1999; Brooks-Wilson et al. 1999; Lawn et al. 1999). The link between Tangier disease and mutations in the ABCA1 gene fuelled intense interest in the role of ABC transporters in lipid transport and ABCA1 remains the most studied member of this subfamily.

Early investigations identified ABCA1/Abca1 in human and mouse brain (Luciani et al. 1994; Langmann et al. 1999). Semi-quantitative northern blot analysis indicated the highest ABCA1 mRNA expression in the putamen, occipital lobe, amygdala, caudate nucleus, hippocampus and substantia nigra of the human brain (Langmann et al. 1999). Although this initial data indicated that ABCA1 was expressed in the brain at levels comparable to the liver, subsequent qRT-PCR analysis indicated that total brain ABCA1/Abca1 expression is ∼ 15 to 40% of liver expression levels in humans (Kielar et al. 2001; Langmann et al. 2003) and mice (Su et al. 2002), respectively. In situ hybridization studies of Abca1 mRNA in the rat (Fukumoto et al. 2002) and mouse brain (Tachikawa et al. 2005) indicated the highest Abca1 mRNA expression in the olfactory bulb, hippocampus, cerebellar cortex and choroid plexus as well as in germinal regions of embryonic and early postnatal brains. This in situ hybridization data indicated Abca1 expression in laminar or nuclear patterns which suggested neuronal expression (Tachikawa et al. 2005). Consistent with this, isolated mouse primary neurons were found to express Abca1 protein by western blot (Fukumoto et al. 2002).

We used qRT-PCR to quantify the expression of ABCA1, 2, 3, 7, and 8 in isolated human fetal neurons, astrocytes, microglia, and oligodendrocytes (purified and grown in vitro) and detected ABCA1 expression in all of these cell types, with the highest expression in neurons and microglia (Kim et al. 2006). We reported that astrocyte ABCA1 expression was only ∼ 25% of neuronal expression level, however, our data revealed that of the five ABCA transporters analyzed, ABCA1 was still detected at the highest level relative to the other four transporters in astrocytes (Kim et al. 2006). Studies of isolated rat neurons, astrocytes, and microglia indicated Abca1 mRNA and protein expression by northern and western blotting, respectively (Koldamova et al. 2003). Studies from other groups indicated that ABCA1/Abca1 is expressed in mouse and rat astrocytes and the human CCF-STTG1 astrocytoma cell line (Hirsch-Reinshagen et al. 2004; Abildayeva et al. 2006), mouse microglia (Hirsch-Reinshagen et al. 2004) and porcine and rat brain capillary endothelial cells (Panzenboeck et al. 2002; Ohtsuki et al. 2004).

Given the known function of ABCA1 as a lipid transporter and the expression of ABCA1 in multiple brain cell types (albeit with significant cell type and brain region expression differences) numerous studies have focused on the potential function of ABCA1/Abca1 as a brain lipid transporter (Panzenboeck et al. 2002; Koldamova et al. 2003, 2005b; Quan et al. 2003; Hirsch-Reinshagen et al. 2004; Liang et al. 2004; Wahrle et al. 2004; Abildayeva et al. 2006; Burns et al. 2006; Karten et al. 2006; Fujiyoshi et al. 2007; Kim et al. 2007). Because membrane cholesterol regulates processing of the amyloid precursor protein (APP) to generate neurotoxic amyloid-β peptide (Aβ) (Simons et al. 1998; Puglielli et al. 2003), ABCA1/Abca1 has also been investigated as a modulator of Aβ production and amyloid deposition (Fukumoto et al. 2002; Koldamova et al. 2003, 2005a,b; Sun et al. 2003; Wollmer et al. 2003; Katzov et al. 2004; Rebeck 2004; Hirsch-Reinshagen et al. 2005, 2007; Wahrle et al. 2005, 2007; Burns et al. 2006; Kim et al. 2007; Koldamova and Lefterov 2007; Riddell et al. 2007); an area that we expand upon below.

ABCA1 expression is regulated by liver X receptors (LXRs) and retinoid X receptors (RXRs) which, when activated by ligands, such as 22(R)-hydroxycholesterol and 9-cis-retinoic acid, bind to specific promoter regions of target genes to promote transcription (Venkateswaran et al. 2000). Initial studies correlated LXR/RXR-induced expression of Abca1 in isolated mouse neurons, astrocytes, and microglia with enhanced cholesterol efflux to lipid free apoA-I and apoE3 (Koldamova et al. 2003). Under both basal and LXR/RXR-induced conditions, apoE-induced cholesterol efflux was ∼ 50% as effective compared to apoA-I when these cholesterol acceptors were used at the same concentration (Koldamova et al. 2003). Previous work also indicated that lipid-free apoE stimulated cholesterol efflux from neurons and astrocytes, however, the mechanism was not identified (Michikawa et al. 2000). Based on observations that apoE is not present in the CNS in a lipid-free state but rather as a lipidated discoidal or spherical lipoprotein particle (Pitas et al. 1987; LaDu et al. 1998; Danik et al. 1999; Fagan et al. 1999; Ito et al. 1999; Demeester et al. 2000; Koch et al. 2001), we used reconstituted lipidated apoE discs to study neuronal cholesterol efflux and demonstrated potent cholesterol efflux from primary human neurons and neuronal cell lines (Kim et al. 2007). Furthermore, we showed that cholesterol efflux to apoE discs was enhanced by ABCA1 over-expression in HEK293 cells (Kim et al. 2007). Importantly, < 25% of neuronal cholesterol efflux in the presence of apoE discs was due to conversion of cholesterol to more polar products, such as 24-hydroxycholesterol which can diffuse from cell membranes independently of ABC transporter activity (Kim et al. 2007). Therefore, while it was established that apoE lipoproteins in the CNS can deliver cholesterol to neurons via LDL receptor family members (Pitas et al. 1987; Herz and Beffert 2000; Pfrieger 2003), it is now clear that at least in the presence of apoE discs, ABCA1 can also stimulate removal of cholesterol from neurons. This pathway is also likely to contribute to cholesterol efflux from astrocytes (Abildayeva et al. 2006) and indeed any brain cell type that expresses ABCA1.

Since membrane cholesterol distribution is thought to modulate APP processing (Sparks et al. 1994; Bodovitz and Klein 1996; Simons et al. 1998; Dietschy and Turley 2001; Riddell et al. 2001; Runz et al. 2002; Cordy et al. 2003), several studies have focused on the potential regulation of APP proteolysis and brain amyloid deposition by ABCA1. An initial report suggested that induction of ABCA1 in the Neuro2a neuroblastoma cell line using the LXR agonists TO-901317 or 22(R)-hydroxycholesterol (both in the presence of the RXR agonist 9-cis-retinoic acid) increased Aβ secretion (Fukumoto et al. 2002). However, subsequent work indicated that LXR/RXR-induced ABCA1/Abca1 expression inhibited Aβ secretion from Neuro2a cells, rat primary cortical neurons, and from a CHO cell line stably expressing human APP695 with the Swedish K670M671-->N670L671 mutations (Koldamova et al. 2003; Sun et al. 2003; Brown et al. 2004). In agreement with these latter studies, we found that transient transfection of ABCA1 into a CHO cell line stably expressing human wild-type APP695 inhibited Aβ secretion by 50% (Kim et al. 2007). The reason for the discrepancies in the earlier work with the latter studies, may be related to different cell types used and methods of Aβ detection or possibly due to differences in the extent that membrane cholesterol transport was modified under the different in vitro settings (Burns et al. 2006; Koldamova and Lefterov 2007).

In vivo studies in mice have added further insights into the potential function of Abca1 in the brain. ABCA1/Abca1 clearly plays an important role in the lipidation and steady-state concentration of apoE (Hirsch-Reinshagen et al. 2004; Wahrle et al. 2004; Hirsch-Reinshagen and Wellington 2007). In Abca1−/− mice, total apoE levels were reduced by 80% in the cortex and 98% in the CSF (Wahrle et al. 2004). CSF from Abca1−/− mice also had reduced cholesterol concentration and contained smaller, poorly lipidated apoE-containing lipoproteins as compared to CSF from wild-type mice (Wahrle et al. 2004). Regarding the potential for Abca1 to modulate Aβ production and/or deposition in vivo, amyloidogenic transgenic mice have been used to investigate the impact of LXR-induced Abca1 expression (Koldamova et al. 2005b; Burns et al. 2006; Riddell et al. 2007) and the direct impact of Abca1 gene knockout (Hirsch-Reinshagen et al. 2005; Koldamova et al. 2005a; Wahrle et al. 2005) or over-expression (Hirsch-Reinshagen et al. 2007) as described below.

The LXR agonist TO-901317 significantly inhibited Aβ deposition in the brains of amyloidogenic APP23 and Tg2576 transgenic mice (Koldamova et al. 2005b; Riddell et al. 2007). Evidence for Abca1-mediated inhibition of Aβ production (which is not necessarily correlated with Aβ deposition) was provided only in the first of these studies (Koldamova et al. 2005b). Interestingly, in an in vitro system using APP C-terminal fragment (C99) as a substrate, TO-901317 was shown to directly inhibit γ-secretase activity (as assessed by Aβ1-40 production) with an IC50 of 5 μmol/L (Czech et al. 2007). While this raises some concerns about the interpretation of early studies using TO-901317 (as at least part of this compound’s action may be independent of LXR activation), the capacity for brain LXRs to regulate Aβ deposition has now been confirmed using another approach. Coming from the opposite direction, genetic loss of either Lxrα or Lxrβ in amyloidogenic transgenic APP/PS1 mice was recently shown to significantly increase cerebral Aβ deposition (Zelcer et al. 2007). Furthermore, the increase in Aβ deposition was correlated with a ∼68% decrease in brain Abca1 mRNA and ∼ 40% decrease in brain Abca1 protein in the Lxrα−/− and Lxrβ−/− APP/PS1 transgenic models (Zelcer et al. 2007). In these studies, there was no evidence for altered APP processing in the LXRαβ−/− setting. Therefore, in vivo studies consistently show that modulation of LXR significantly impacts on Aβ deposition in mice and that this is correlated with ABCA1 expression; however, it is not yet clear if modulation of APP processing, apoE mediated Aβ clearance, or other factors (e.g., related to inflammation) are primarily responsible for the LXR-mediated modulation of Aβ deposition and brain pathology.

Another series of studies from three laboratories investigated the impact of genetic loss of Abca1 on brain Aβ deposition in four different amyloidogenic transgenic mouse models (Hirsch-Reinshagen et al. 2005; Koldamova et al. 2005a; Wahrle et al. 2005). The overall conclusion from these studies was that in the absence of Abca1, Aβ load was significantly increased in the brains of three of the four transgenic animal models, significant reductions of brain total apoE concentration or altered apoE solubility was observed in all four animal models, and in the two studies that presented data on APP processing (Koldamova et al. 2005a; Wahrle et al. 2005), no consistent evidence for a modified rate of Aβ production in the absence of Abca1 was found. Regardless of the molecular mechanisms responsible for the protective function of LXR activation and/or enhanced ABCA1 expression on brain amyloid deposition, this remains a critical target in the prevention of neuropathology in AD and the preclinical data clearly indicates therapeutic LXR activation is worthy of further investigation.

Regarding genetic variation in the ABCA1 locus, eleven studies have been published so far investigating an association between ABCA1 gene polymorphisms and AD in a variety of populations (Wollmer et al. 2003; Katzov et al. 2004; Li et al. 2004; Kolsch et al. 2006; Shibata et al. 2006; Chu et al. 2007; Rodriguez-Rodriguez et al. 2007; Sundar et al. 2007; Wahrle et al. 2007; Wang and Jia 2007; Wavrant-De Vrieze et al. 2007). Seven of these concluded that genetic variation in the ABCA1 locus is significantly associated with AD risk (Wollmer et al. 2003; Katzov et al. 2004; Chu et al. 2007; Rodriguez-Rodriguez et al. 2007; Sundar et al. 2007; Wang and Jia 2007; Wavrant-De Vrieze et al. 2007). The underlying biological mechanisms for these associations remain to be determined.

ABCA2

In humans and rodents ABCA2/Abca2 is expressed predominantly in the brain as compared to other organs; with particularly high expression in oligodendrocytes (Luciani et al. 1994; Zhao et al. 2000; Vulevic et al. 2001; Zhou et al. 2001; Su et al. 2002; Langmann et al. 2003; Kim et al. 2006). It has been suggested that since Abca2 is selectively expressed in rat oligodendrocytes it may be a useful oligodendrocyte marker (Zhou et al. 2001); however, human fetal brain cells (including neurons) also express ABCA2 mRNA (Kim et al. 2006), and it is clear that ABCA2/Abca2 is detectable in neuronal subsets of the adult rodent brain and in rat and human brain endothelial cells (Ohtsuki et al. 2004; Broccardo et al. 2006). Very little ABCA2 is expressed on the cell surface, rather there is evidence for a predominantly lysosomal localization at least in oligodendrocytes (Zhou et al. 2001; Broccardo et al. 2006). This led to the proposition that ABCA2 may play a role in intracellular lipid trafficking rather than trans-plasma membrane transport as is the case for ABCA1 (Schmitz and Kaminski 2002; Davis et al. 2004; Mack et al. 2006, 2007b). Consistent with this, ABCA2 over-expression in HEK293 cells did not result in enhanced cholesterol efflux to acceptors including apoA-I, apoE, or apoE discs (Kim et al. 2007).

Because oligodendrocytes play a key role in myelination and because ABCA2 expression was developmentally up-regulated in concert with myelin sheath-associated proteins, a role for ABCA2/Abca2 in myelin lipid transport was suggested by several groups (Vulevic et al. 2001; Tachikawa et al. 2005; Wang et al. 2005). The recent generation of Abca2 gene knockout mice from two independent groups supports this prediction (Mack et al. 2007a; Sakai et al. 2007). In the first of these studies, deletion of Abca2 resulted in a distinct shaking phenotype and the mice were easily startled and therefore termed ‘skittish’ (Mack et al. 2007a). The Abca2−/− mice had greatly increased myelin sheath thickness in the spinal cord and a significant reduction in myelin membrane periodicity (compaction) was observed in both spinal cords and cerebra (Mack et al. 2007a). Total CNS tissue lipid composition (ceramide, sphingosine and sphingomyelin species) was apparently unaffected by Abca2 deficiency at least as assessed in four mice at 7 weeks of age (Mack et al. 2007a). The second report on Abca2−/− mice confirmed the shaking and startle phenotype and conducted extensive lipid analysis of both whole brain and purified myelin fractions over a range of ages from 11 days to 64 weeks and in doing so discovered several major alterations in brain lipid composition in the Abca2−/− mice (Sakai et al. 2007). While total brain cholesterol and phospholipids (glycerophospholipids) were not altered at any of the ages investigated, brain sphingomyelin (SM) levels were significantly lower from 32 weeks of age and brain SM, cerebrosides and sulfatides were all significantly lower at 64 weeks of age in the Abca2−/− mice (Sakai et al. 2007). In contrast, brain ganglioside concentrations were significantly increased in Abca2−/− mice from as early as 4 weeks of age (Sakai et al. 2007). Interestingly, analysis of myelin lipids revealed selective deficiencies in two glycerophospholipids (phosphatidylethanolamine and phosphatidylserine) and SM and a doubling of ganglioside GM1 concentration (Sakai et al. 2007). Therefore, it appears that ABCA2 does play an important role in the regulation of neural sphingolipid transport and (because this can modify subcellular sphingolipid localization) metabolism. It should be noted that in addition to the lysosomal location of ABCA2, there is evidence for ABCA2 localization in late endosomes and in the Golgi membrane (Zhou et al. 2001). It is clear that further research is needed to clarify how ABCA2 regulates brain sphingolipid transport.

It has been suggested that ABCA2 may contribute to AD by regulating myelination and/or oligodendrocyte cholesterol homeostasis (Bartzokis 2004). Interestingly, variations in the ABCA2 gene have been identified as risk factors for both early- and late-onset AD (Mace et al. 2005; Wollmer et al. 2006). ABCA2 expression appears to be associated with over-expression of several genes implicated in AD and there is evidence that cellular APP levels and Aβ production are increased by ABCA2 expression in vitro (Chen et al. 2004b); although in studies conducted by our group, ABCA2 transient expression in CHO cells expressing wild-type human APP695 had no impact on APP processing or Aβ secretion (Kim et al. 2007). Since ABCA2 over-expression in HEK293 cells was associated with increased expression of gene clusters related to oxidative stress and APP amyloidogenic processing, the impact of ABCA2 on AD may not be specifically related to lipid transport (Chen et al. 2004b; Mace et al. 2005). Further studies are therefore required to understand precisely how ABCA2 may modulate AD risk.

Other research indicates that ABCA2 expression is increased (along with other oligodendrocyte genes) in anterolateral temporal cortical samples taken from individuals with temporal lobe epilepsy with abnormal spiking electrical activity (Arion et al. 2006). In these studies, alterations in myelination were speculated to play a role in the abnormal spiking electrical activity although this was not demonstrated directly (Arion et al. 2006). ABCA2 has also been suggested as a possible molecular marker for oligodendrocytomas (Soichi et al. 2007). Such cancers are difficult to diagnose by routine histological criteria and the high expression of ABCA2 in oligodendrocytes may provide a reasonably specific marker (Soichi et al. 2007). Finally, a number of neurological disorders have been linked to the ABCA2 locus on chromosome 9q34 including juvenile amyotrophic lateral sclerosis 4, lethal congenital contraction syndrome (LCCS), and Joubert syndrome-related disorders associated with cerebellar vermis hypoplasia (Makela-Bengs et al. 1998; Chen et al. 2004a; Valente et al. 2005; Kaminski et al. 2006). While there is strong evidence that the DNA/RNA helicase senataxin is the gene responsible for the 9q34 association with amyotrophic lateral sclerosis 4 (Chen et al. 2004a), the genes responsible for LCCS and Joubert syndrome-related disorders remain to be defined and ABCA2 represents one candidate worth considering. Although several of the neurological conditions mentioned in relation to ABCA2 function are related to situations where alterations in myelination may be a factor, there are currently no reports linking ABCA2 mutations with well characterized human myelination disorders as far as we are aware.

ABCA3

The highest expression of ABCA3 mRNA in humans is in the lung, followed by the brain, thyroid, and testis (Langmann et al. 2003). In the mouse, western blotting for Abca3 indicates high expression in the lung and moderate expression in brain, kidney, spleen, and adipose tissue (Fitzgerald et al. 2007; Stahlman et al. 2007). During mouse brain development, Abca3 mRNA is widely detected in a pattern very similar to Abca2 expression (e.g., with higher expression in the mantle zone than the ventricular zone during embryonic stages and upregulation in white matter at postnatal day 14) (Tachikawa et al. 2005). In isolated human fetal brain cells, we found that ABCA3 mRNA expression was highest in oligodendrocytes but also detectable at low levels in neurons, astrocytes, and microglia (Kim et al. 2006). Which, given the high expression of ABCA2/Abca2 in oligodendrocytes discussed above, could explain some of the similarities observed for regional expression of Abca2 and Abca3 during mouse brain development.

A role for ABCA3 as a lipid transporter in the lung is firmly established (Yamano et al. 2001; Mulugeta et al. 2002). ABCA3 plays a key role in the formation of pulmonary lung surfactant which accumulates in lamellar bodies of lung alveolar epithelial type II cells (Cheong et al. 2006). Lamellar bodies are the intracellular storage organelles for surfactant and it is thought that ABCA3 expressed on the limiting membrane of the lamellar bodies regulates the transport of pulmonary lung surfactant lipids, such as phosphatidylcholine, SM and to a lesser degree cholesterol into lamellar bodies (Cheong et al. 2006). It is clear that ABCA3 mutations are associated with defective lamellar body formation in humans which leads to fatal respiratory distress syndrome in newborn infants and interstitial lung disease (Yamano et al. 2001; Mulugeta et al. 2002; Nagata et al. 2004). ABCA3 with mutations linked to respiratory distress syndrome was not correctly processed in vitro and remained in the endoplasmic reticulum whereas expression of recombinant wild-type human ABCA3 in HEK293 cells induced formation of lamellar body-like vesicles that contained lipids (Cheong et al. 2006). The recent generation of Abca3 knockout mice confirmed the essential role for Abca3 in the formation of pulmonary lung surfactant and indicated that lipids most affected by Abca3 deletion were phosphatidylglycerol and phosphatidylcholine while there was essentially no impact on lung SM or cholesterol levels (Fitzgerald et al. 2007).

Given the known substrate specificity for ABCA3 and its presence in oligodendrocytes (Kim et al. 2006) it seems possible that this transporter could contribute to phosphatidylcholine incorporation into myelin. In contrast, a role for ABCA3 as phosphatidylglycerol transporter in this context seems unlikely as there are currently no data indicating the presence of phosphatidylglycerol in myelin. Phosphatidylglycerol is a mitochondrial membrane lipid and precursor to cardiolipin. Total brain phosphatidylglycerol levels are reported to be in the order of 700 nmol/g wet weight in mouse brain and it is known to be synthesized from phosphatidic acid in the mitochondrial inner membrane (Pumphrey 1969; Zhang et al. 2002; Ellis et al. 2005; Barcelo-Coblijn et al. 2007). The exact role of ABCA3 in brain lipid transport is thus far unknown; however, future analysis of lipid composition of whole brain, myelin or oligodendrocytes derived from Abca3−/− mice may provide valuable insights (Fitzgerald et al. 2007). (see also ‘Dynamics of ABC transporter expression during human brain embryonic and post-natal development’ section below).

ABCA4

ABCA4/Abca4 is highly expressed in the retina with only very low expression in whole brain of humans, rats and mice (Sun and Nathans 1997; Langmann et al. 2003; Bhongsatiern et al. 2005; Tachikawa et al. 2005). ABCA4 mediates the transport of retinyldiene phospholipid complexes within the rod outer segment of the retina and mutations in this gene can cause Stargardt disease and other related eye disorders (Allikmets et al. 1997); as mentioned in the ‘Introductary part’. Interestingly, Abca4 appears to be expressed at the choroid plexus of the mouse and rat brain and rat choroid plexus epithelial cells (Bhongsatiern et al. 2005; Tachikawa et al. 2005). If the same cell-specific expression pattern also occurs in humans this may explain why there is low but detectable ABCA4 mRNA expression in whole human brain samples (Langmann et al. 2003). While speculative, it has been suggested that ABCA4 expression in choroid plexus epithelial cells could regulate the concentration of retinoids in the CSF (Bhongsatiern et al. 2005) and, because certain retinoids regulate signalling and development and play a role in motor neuron disease (Zetterstrom et al. 1999; Corcoran et al. 2002; Maden 2002), that ABCA4 could have an important impact on retinoid modulation of CNS function. However, it is not known if CNS retinoids are substrates for ABCA4 and there is so far no direct evidence that ABCA4 has an impact on CSF retinoid levels or on CNS development/function. An Abca4 deficient mouse line has been created and could be used to directly answer these important questions (Weng et al. 1999).

ABCA7

In the studies reporting the initial cloning of human ABCA7 a lack of expression for ABCA7 mRNA in the brain was assumed based on a dot blot method (Kaminski et al. 2000). In contrast, subsequent analysis revealed strong expression for both Abca7 mRNA and protein in the mouse brain (Ikeda et al. 2003; Wang et al. 2003; Kim et al. 2005; Tachikawa et al. 2005). In situ hybridization studies in the adult mouse revealed that Abca7 is expressed throughout the brain with particularly high expression in CA1 hippocampal neurons (Kim et al. 2005). Additional northern blot and qRT-PCR analysis confirmed Abca7 is expressed in human brain (Ikeda et al. 2003; Langmann et al. 2003) and analysis of isolated fetal human brain cells indicated that microglia express the highest level of ABCA7 mRNA overall; at a level approximately 10-fold higher than neurons (Kim et al. 2006). While ABCA7 is the closest homologue of ABCA1 (54% homology based on amino acid identity), the role of ABCA7 in lipid transport is not entirely clear. There are two isoforms of ABCA7, arising as a result of alternative splicing, both of which are detected in the human brain (Ikeda et al. 2003). While ABCA7 has been detected on the cell surface and in intracellular compartments, the so-called Type II ABCA7 splice variant appears to remain in the endoplasmic reticulum (Ikeda et al. 2003). Thus, it is possible that these variant proteins have different biological functions.

Given the homology of ABCA7 with ABCA1 it was predicted that ABCA7 would potently stimulate cellular cholesterol efflux to lipid free acceptors, such as apoA-I and apoE. Most studies have confirmed that ABCA7 expression promotes efflux of phospholipids, including phosphatidylcholine and SM, however, ABCA7/Abca7 is not as effective at promoting cholesterol efflux (Wang et al. 2003; Abe-Dohmae et al. 2004, 2006). The initial identification of ABCA7 in hemopoietic cells pointed towards a role for regulating macrophage lipid efflux in particular (Kaminski et al. 2000). In contrast to this expectation, macrophages isolated from Abca7−/− mice exhibited normal cholesterol and phosphatidylcholine efflux thereby indicating that ABCA7/Abca7 may have another distinct biological function (Kim et al. 2005). More recent work identified a role for ABCA7 in phagocytosis (Iwamoto et al. 2006; Jehle et al. 2006). The available evidence suggests that macrophage ABCA7 plays a key role in phagocytosis of apoptotic debris (Jehle et al. 2006). Because microglia actively phagocytose apoptotic debris in the CNS (Stolzing and Grune 2004), and data from our own laboratory indicates that ABCA7 has the highest expression in human microglia (Kim et al. 2006), a function for microglial ABCA7 in phagocytosis of apoptotic debris (e.g., during development or neurodegenerative disease) is also plausible.

ABCA8

ABCA8 was originally cloned as a gene with unknown function from human brain (Nagase et al. 1998). ABCA8 mRNA was subsequently detected at low levels in human brain by northern blot and qRT-PCR analysis and in fetal brain neurons and astrocytes by qRT-PCR (Tsuruoka et al. 2002; Langmann et al. 2003; Kim et al. 2006). In contrast to the relatively low expression in human brain or isolated brain cells, Abca8 mRNA was expressed at a high level in the choroid plexus of adult and postnatal P5 mouse brain as assessed by northern blot and in situ hybridization (Matsumoto et al. 2003). Very little is known regarding the possible function of ABCA8/Abca8 other than it has the capacity to transport lipophilic substrates including the bioactive lipid, leukotriene C4, across Xenopus laevis oocyte membranes (Tsuruoka et al. 2002). Whether ABCA8 has a specific function in the choroid plexus remains to be determined.

ABCB1

Nomenclature of the P-glycoproteins ABCB1(MDR1)/Abcb1a(Mdr1a)/Abcb1b(Mdr1b) and ABCB4(MDR3)/Abcb4(Mdr2) was initially confusing. As noted previously (Oude Elferink and Paulusma 2007), this was due to the assignment of different names to the different orthologues in different species before the function of the proteins was established and was not helped by the fact that Mdr1a has also been called Mdr3. Use of the ABC nomenclature should help resolve this issue. ABCB1 in humans and the two orthologues Abcb1a and Abcb1b in rodents are established as important multidrug resistance proteins. ABCB1 is expressed in the apical plasma membranes of many types of epithelial and endothelial cells and is widely studied as a drug pump that effectively effluxes a variety of amphipathic drugs from the cytosol to the extracellular compartment (Higgins 1992; Ambudkar et al. 1999). It is also recognized that ABCB1 is a lipid transporter with broad specificity. For example, studies in LCC-PK1 pig kidney epithelial cells transfected with ABCB1 demonstrated accelerated efflux of phosphatidylcholine, phosphatidylethanolamine, glucosylceramide and SM (van Helvoort et al. 1996). Although the expression of ABCB1 in the brain is low (Langmann et al. 2003), there is evidence suggesting that ABCB1/Abcb1b expressed on the luminal surface of brain capillary endothelial cells is a functional component of the blood–brain barrier (BBB) (Cordon-Cardo et al. 1989; Tatsuta et al. 1992; Tishler et al. 1995; Seetharaman et al. 1998). Rat Abcb1a is expressed at very low levels in the choroid plexus (Choudhuri et al. 2003). Another study in rats has shown that Abcb1a mRNA is clearly expressed in the frontal cortex, dorsal cortex, hippocampus, midbrain, pons/medulla and cerebellum (Kwan et al. 2003); although cell type-specific expression remains unknown. Interestingly, Abcb1b mRNA and protein has been reported to be expressed in isolated rat astrocytes (Decleves et al. 2000) and, at least at the mRNA level, in rat hippocampus (Kwan et al. 2003).

Studies in Abcb1a−/− mice suggest an important function in the regulation of drug efflux from the CNS through the BBB (Schinkel et al. 1996); however, the possible role that ABCB1/Abcb1 may play in regulating transport of phospholipids and glycosphinglolipids from the brain to the peripheral circulation has not been assessed. Based on the known functions of ABCB1 and its localization in brain capillary endothelial cells, a contribution to the transport of these lipids across the BBB seems plausible.

Dysfunction of ABCB1 has been implicated in Parkinson’s disease, using uptake of [11C]-verapamil in the midbrain as a measure of BBB function (Kortekaas et al. 2005). However, it is not clear if altered lipid transport plays a role in this context. In this study linking ABCB1 dysfunction with Parkinson’s disease, the causal link was thought to be related to increased midbrain concentrations of environmental toxins (Kortekaas et al. 2005). In a different setting, increased ABCB1 expression has been linked to intractable epilepsy and this is thought to be due to accelerated efflux of intraparenchymal antiepileptic drugs (Tishler et al. 1995).

ABCB1, in general, has a very broad substrate specificity and recent studies have shown that it is also a key regulator of Aβ transport across the BBB into the bloodstream (Lam et al. 2001; Cirrito et al. 2005; Kuhnke et al. 2007). Given that Aβ forms soluble assemblies with the glycosphingolipid GM1 (Hayashi et al. 2004), investigating whether ABCB1 transport of Aβ through the BBB is dependent on or modulated by GM1 represents a worthwhile study for the future. Although there is circumstantial evidence suggesting that ABCB1 could regulate the flux of specific lipids/lipid complexes across the BBB, direct evidence for such pathways remain to be provided.

ABCB4

Human ABCB4 is also known as MDR3 and the rodent orthologue Abcb4 is also known as Mdr2. ABCB4/Abcb4 mRNA expression is low but detectable in human brain and rat choroid plexus (Choudhuri et al. 2003; Langmann et al. 2003). ABCB4 selectively transports phosphatidylcholine across membranes and exhibits very little transport activity for SM and GM1 (van Helvoort et al. 1996). Studies in Abcb4−/− mice indicated a crucial role for ABCB4/Abcb4 in hepatic phosphatidylcholine transport to form bile (Smit et al. 1993). ABCB4 plays a major role in the translocation of phosphatidylcholine across the canalicular membrane of the hepatocyte and defects in ABCB4 are the cause of progressive familial intrahepatic cholestasis type III (PFIC3) (Deleuze et al. 1996; Oude Elferink and Paulusma 2007). While ABCB4 is an extremely effective phosphatidylcholine transporter, such a function in the brain has not been investigated thus far.

ABCD1 and ABCD2

In contrast to the ABC transporters discussed above, ABCD1 is a half transporter that requires either homo- or hetrodimerization with one of at least three other closely related transporters (ABCD2, ABCD3, and ABCD4) in order to function. All four transporters are expressed in peroxisome membranes and defects in the ABCD1 gene cause X-linked adrenoleukodystrophy (X-ALD) which is characterized by demyelination of the CNS, peripheral adrenal insufficiency (Addison’s disease), mental deterioration, corticospinal tract dysfunction, and cortical blindness (Kamijo et al. 1992; Mosser et al. 1993; Lombard-Platet et al. 1996; Holzinger et al. 1997; Shani et al. 1997). ABCD1–4 are also commonly known as adrenoleukodystrophy protein (ALDP), ALDP-related protein (ALDRP), 70 kDa peroxisomal membrane protein (PMP70) and 69 kDa PMP (PMP69), respectively.

Since peroxisomes are present in essentially all tissues, it is not surprising that ABCD1 is widely expressed in human tissues (Mosser et al. 1993). It is clear that peroxisomes are particularly abundant in the liver and kidney where they have an average diameter of 0.2–1 μm whereas in the brain they are less abundant and have a diameter of 0.05–0.2 μm (Mosser et al. 1993). Although there is in vitro evidence from yeast two-hybrid and mouse 3T3 fibroblast co-transfection experiments that ABCD1 and ABCD2 can form heterodimers (Liu et al. 1999), there are striking tissue-specific differences in their expression indicating that they may function as homodimers with similar functions in different tissues (Troffer-Charlier et al. 1998; Pujol et al. 2004; Kemp and Wanders 2007). ABCD1/Abcd1 protein expression in the human and mouse cortex is very low but is clearly detected in the astrocyte processes at the junction of the cortex and white matter and in cultured astrocytes (Fouquet et al. 1997). ABCD1/Abcd1 protein is also expressed in a subpopulation of oligodendrocytes localized in the corpus callosum, internal capsules and anterior commissure and in most microglial cells in both humans and mice (Fouquet et al. 1997). In contrast, ABCD2, which is expressed at much higher levels in the brain than ABCD1 (Langmann et al. 2003; Kemp and Wanders 2007), was detected at the mRNA and protein levels at high concentration in the cerebral cortex, in the pyramidal and granular cell layers of the dentate gyrus of the hippocampus and in cerebellum in Purkinje cells and the inner granular layer (Troffer-Charlier et al. 1998).

X-linked adrenoleukodystrophy is a severe neurological disorder caused by mutations in the ABCD1 gene and characterized by progressive demyelination of the CNS, adrenal insufficiency (Addison’s disease) and the accumulation of very-long chain fatty acids (VLCFA) in brain white matter, adrenal cortex, plasma and fibroblasts (Schaumburg et al. 1975; Mosser et al. 1993; Kutsche et al. 2002; Dean 2005). The accumulation of VLCFA appears to be due to defective β-oxidation of these fatty acids in peroxisomes and based on studies in yeast, where ABCD1/ABCD2 homologues PXA1P/PXA2P form heterodimers in peroxisome membranes, the most plausible explanation is that ABCD1/ABCD2 normally imports VLCFA (that have been activated with CoA in the cytosol) into the peroxisome (Shani et al. 1995; Hettema et al. 1996; Verleur et al. 1997; Hettema and Tabak 2000). Two ABCD1 mutations associated with X-ALD (Ser606Leu and Gly512Ser) result in decreased ATP binding and reduced ATPase activity, respectively, thereby indicating that an energy dependent transport activity is crucial to the function of ABCD1 (Roerig et al. 2001).

ABCG1

There are five ABC subfamily G transporters in the human genome ABCG1, ABCG2, ABCG4, ABCG5 and ABCG8; all of which are half transporters containing a single hydrophobic TMD and NBD (Schmitz et al. 2001; Baldan et al. 2006; Kusuhara and Sugiyama 2007). ABCG1 is the founding member of the ABCG subfamily and was identified as the human homologue of the Drosophila white gene (Chen et al. 1996; Savary et al. 1996; Croop et al. 1997). ABCG1/Abcg1 is expressed widely in tissues of humans and mice and in macrophages (Langmann et al. 2003; Baldan et al. 2006; Kusuhara and Sugiyama 2007). Several studies confirm significant ABCG1/Abcg1 mRNA expression in the brain (Chen et al. 1996; Savary et al. 1996; Croop et al. 1997; Su et al. 2002; Langmann et al. 2003; Nakamura et al. 2004; Tachikawa et al. 2005). In situ hybridization studies in mouse brain indicate Abcg1 mRNA is widely expressed in both the ventricular and mantle zones of embryonic brains and in both gray and white matter of postnatal brains (Tachikawa et al. 2005). Homologous recombination of an IRES-LacZ-Neo-pA cassette into the murine Abcg1 locus demonstrated that Abcg1 is highly expressed in neurons, with particularly abundant expression in hippocampus, where it is found in CA1, CA2, and CA3 neurons as well as in the dentate gyrus (Tansley et al. 2007). In these same studies, Abcg1 was also found to be expressed in all cortical layers as well as in the striatum and thalamus (Tansley et al. 2007). Similar studies recently confirmed that Abcg1 is expressed in many different classes of neurons in the mouse CNS (Tarr and Edwards 2007). qRT-PCR analysis and western blot analysis of human fetal brain cells indicated the highest ABCG1 expression in microglia followed by oligodendrocytes, neurons and astrocytes (Kim et al. 2007). Abcg1 has also been detected in mouse primary cerebellar astroglia preparations and at low levels in rat primary astrocytes and the CCF-STTG1 astrocytoma cell line (Abildayeva et al. 2006; Karten et al. 2006). Similar to ABCA1, LXR agonists 24(S)-hydroxycholesterol, GW683065A and TO-901317 all potently stimulate astroglial ABCG1/Abcg1 expression (Abildayeva et al. 2006; Karten et al. 2006). Abcg1 mRNA and protein is also expressed in rat choroid plexus and choroid plexus epithelial cells where it is also potently induced by 24(S)-hydroxycholesterol (Fujiyoshi et al. 2007).

Given that ABCG1 is a key transporter of cholesterol across the macrophage plasma membrane (Jessup et al. 2006), it is perhaps not surprising that a similar role has also been proposed for ABCG1 in neurons, astroglia and choroid plexus epithelial cells (Abildayeva et al. 2006; Karten et al. 2006; Fujiyoshi et al. 2007; Kim et al. 2007). One of the features that differentiates ABCA1- and ABCG1-mediated cholesterol efflux is the nature of the extracellular cholesterol acceptor. Several studies have shown that ABCA1 preferentially stimulates cholesterol efflux to lipid free acceptors, such as apoA-I and lipid-free apoE whereas ABCG1 is more selective for lipidated lipoprotein complexes, such as HDL (relevant for peripheral reverse cholesterol transport) and lipidated apoE discs (relevant to the CNS) (Nakamura et al. 2004; Wang et al. 2004; Kennedy et al. 2005; Kim et al. 2007). In both peripheral tissues and the CNS a synergistic process relying on initial ABCA1-dependent efflux of phospholipids to lipid free acceptors (which may occur concomitantly with cellular apolipoprotein secretion) is apparently followed by ABCG1-dependent enrichment of the lipoprotein complexes with cholesterol (Gelissen et al. 2006; Karten et al. 2006; Vaughan and Oram 2006). Thus, apoE discs that initially arise from astrocytes and microglia may interact with multiple brain cell types, including neurons, to promote cholesterol efflux through ABCG1 and to induce signaling pathways through low density lipoprotein receptor (LDLr) family members (Herz and Beffert 2000; Herz and Bock 2002; Li et al. 2003; Kim et al. 2007).

An understanding of the interactions of different brain apolipoprotein species with ABCG1 (and ABCA1) is important as this represents crucial early steps in CNS lipoprotein biogenesis. Subsequent to acquisition of additional cholesterol from cells that express ABCG1, apoE discs may be converted to spherical lipoprotein particles via lecithin cholesterol acyltransferase (LCAT) in a process that is analogous to plasma HDL maturation (Zannis et al. 2006). Through the action of LCAT, apoE discs (that are predicted to carry in the order of 50–200 molecules of free cholesterol and PL) (LaDu et al. 1998; Xu et al. 2000) would accumulate a core of cholesteryl esters and the resulting spherical particle would represent a rich source of cholesterol for cells expressing LDLr family members responsible for cholesterol uptake through receptor-mediated endocytosis (Pitas et al. 1987). In contrast to the factors regulating maturation and remodeling of plasma lipoproteins, very little is known regarding CSF lipoprotein remodeling. It is clear though that LCAT is present in CSF, and that the key LCAT activator, apoA-I, is transported across the BBB (Demeester et al. 2000). Phospholipid transfer protein (PLTP) is also present in the brain where it may play a role in remodeling of the apoE lipoprotein subpopulations, although this has not been studied directly (Vuletic et al. 2003). It is very likely that the physical state of apoE (i.e., lipidation state and disc versus sphere) will have a major impact on CNS apoE function as both a lipid transporter and inducer of cell signalling and, in the AD pathophysiological context, as an Aβ transporter (Bell et al. 2007).

Our in vitro studies indicated that ABCG1 potently regulates neuronal cholesterol efflux (Kim et al. 2007). Based on this, and on in vitro and in vivo studies indicating that ABCA1/Abca1 could modulate either neuronal Aβ production or amyloid deposition (Koldamova et al. 2003, 2005a; Sun et al. 2003; Hirsch-Reinshagen et al. 2005; Wahrle et al. 2005), we examined the impact of ABCG1 expression on Aβ production in CHO cells stably expressing human wild-type APP695. Similar to in vitro observations regarding ABCA1 function (Sun et al. 2003), our data indicated a significant reduction in Aβ production with ABCG1 transient expression in vitro (Kim et al. 2007). In contrast, a subsequent study by Tansley et al. (Tansley et al. 2007) investigated the effect of ABCG1 transient expression in HEK cells stably expressing human APP with the Swedish FAD mutations (APPsw, K670M671-->N670L671 which leads to increased β-secretase cleavage, increased supply of γ-secretase substrate and increased Aβ production (Sinha et al. 1999)) and found that with ABCG1 transfection, Aβ production was increased.

While the reasons for these apparently conflicting results are not known at present, there are important differences in the experimental conditions and in several aspects of the resulting data that may contribute. In the studies of Tansley et al., transient expression of ABCG1 resulted in an increase in total HEK cell APP protein with no change in APP mRNA expression and this was associated with increases in both sAPPα and Aβ secretion (Tansley et al. 2007). This was accompanied by increases in cellular CTFα and CTFβ, so it was clear that, in these studies, the total level of APP was increased and both the amyloidogenic and the non-amyloidogenic APP processing pathways were stimulated (Tansley et al. 2007). In contrast, in the studies by our group no increase in total APP protein, APP mRNA or sAPPα were detected with ABCG1 transfection whereas the amyloidogenic pathway was significantly inhibited and resulted in a 64% reduction in Aβ secretion (Kim et al. 2007). While these experimental models are therefore not directly comparable, differences in cell type, APP wild-type versus APPsw mutant gene expression, and possibly the methods used to detect Aβ (ELISA (Tansley et al. 2007) and western blot (Kim et al. 2007)) may contribute to the final result. It is also clear that wild-type human APP and APPsw are differentially trafficked in HEK293 cells and that an intracellular acidic compartment plays a critical role in the generation and secretion of Aβ from APPsw but not from wild-type APP (Knops et al. 1995). It remains possible that ABCG1 only inhibits the production of wild-type APP which would be of particular significance in late onset AD (Kim et al. 2007). Analysis of ABCG1 function in an in vivo setting should help to clarify these in vitro findings.

At present the role of ABCG1/Abcg1 in APP processing has not been specifically studied in any of the amyloidogenic mouse models. It is also not yet clear if increases in Abcg1 could partially account for the protective effects of LXR agonists in transgenic amyloidogenic mice (i.e., LXR activation would be predicted to induce transcription of both ABCA1 and ABCG1) mentioned above (Koldamova et al. 2005b; Burns et al. 2006; Koldamova and Lefterov 2007; Riddell et al. 2007). Of potential relevance, a recent study has shown that amyloid plaque load is significantly increased when amyloidogenic APP/PS1 mice were crossed with Lxrα/β null animals (Zelcer et al. 2007). Under these conditions basal expression of both Abca1 and Abcg1 in the brain were significantly reduced (Zelcer et al. 2007). Although this data suggests a protective function for ABCG1 in AD, further in vivo studies utilizing Abcg1−/− mice crossed with amyloidogenic transgenic mice may help to establish this. Of relevance to the human situation, a recent study of cholesterol-related genes identified significant associations between ABCG1 single nucleotide polymorphisms and AD in Swiss and Polish study populations (Wollmer et al. 2007). Intriguingly, these associations were not confirmed in other European populations studied in Germany, Belgium, Sweden and Greece (Wollmer et al. 2007). Future research in additional larger populations should help to clarify the potential association of ABCG1 genetic variation with AD.

ABCG2

ABCG2 was originally identified in a multidrug-resistant breast cancer cell line and called breast cancer resistance protein (Doyle et al. 1998). Other than cancer cells, ABCG2 is expressed in the human brain, liver, kidney, small intestine, mammary gland and placenta (Langmann et al. 2003; Zhang et al. 2003; Sarkadi et al. 2004). In situ hybridization studies of developing mouse brain revealed that Abcg2 is expressed at low levels in a diffuse pattern throughout embryonic and postnatal development (Tachikawa et al. 2005). ABCG2 protein is also specifically localized apically in the epithelial cells of the BBB (Cooray et al. 2002; Zhang et al. 2003). Cellular localization and in vivo studies using Abcg2−/− mice suggested that this transporter plays a major role in restricting the penetration of drugs and dietary toxins into the brain (Jonker et al. 2002; Zhou et al. 2002). In addition, ABCG2 may also act as a flippase, increasing the outward transport of phosphatidylserine and NBD-labeled phosphatidylcholine, as well as an estradiol exporter (Janvilisri et al. 2003; Woehlecke et al. 2003). In contrast to other ABCG subfamily members, ABCG2 does not appear to transport cholesterol or sitosterols (Albrecht et al. 2002; Wang et al. 2004; Janvilisri et al. 2005). Interestingly, cellular cholesterol level has been shown to impact on ABCG2 ATPase activity and thereby modulate ABCG2 function (Pal et al. 2007). There is no data to suggest that ABCG2 is directly involved in neurological disease; however, its role as a xenobiotic transporter may prevent or limit the entry of therapeutic drugs to the brain which could impact on the treatment of diseases, such as epilepsy, depression and schizophrenia (Fava and Davidson 1996; Hellewell 1999; Kwan and Brodie 2000). (see also ‘Dynamics of ABC transporter expression during human brain embryonic and post-natal development’ section below).

ABCG4

Like ABCG1, ABCG4 is another human homologue of the Drosophila white gene. However, unlike ABCG1 which is expressed in many tissues, human ABCG4 mRNA is highly expressed only in the brain and the neural retina of the eye as assessed by northern blot analysis (Oldfield et al. 2002). In both humans and mice, ABCG4/Abcg4 mRNA is expressed throughout the brain (Annilo et al. 2001; Engel et al. 2001; Tarr and Edwards 2007) and has also been identified in the choroid plexus of the rat (Fujiyoshi et al. 2007). In situ hybridization studies of developing mouse brain have revealed that Abcg4 is strongly expressed in the mantle zone of embryonic brain and in the gray matter of postnatal brain (Tachikawa et al. 2005). ABCG4 is closely related to ABCG1 (74% amino acid identity) and it is therefore not surprising that ABCG4 also stimulates cholesterol efflux from cells to HDL but not to lipid-free apoA-I (Wang et al. 2004; Vaughan and Oram 2005, 2006). Apart from the overlapping functions of ABCG4 and ABCG1, there is evidence that ABCG4 and ABCG1 form functional heterodimers on the cell surface in order to promote cholesterol efflux (Cserepes et al. 2004). A direct comparison of Abcg1 and Abcg4 mRNA regional expression in both the developing and adult mouse brain indicates there are distinct differences implying that the ABCG1-ABCG4 heterodimer may not be critical for ABCG4 function (Tachikawa et al. 2005). Other data derived from primary mouse neurons and astrocytes transiently transfected with epitope-tagged ABCG1 and ABCG4 indicated that both transporters predominantly reside within intracellular vesicles (Tarr and Edwards 2007). These same studies also indicated that in contrast to Abcg1, Abcg4 is not regulated by Lxr in murine neurons, astrocytes, or microglia (using the LXR ligand GW3965) and furthermore, that transient expression of either transporter in primary mouse neurons and astrocytes increased the expression of Srebp-2 mRNA and induced several target genes involved in cholesterol biosynthesis (Tarr and Edwards 2007). Human ABCG4 brain cell-specific expression and functional analysis particularly regarding this transporter’s role in brain lipid homeostasis and neurological disease are potentially important areas that remain to be explored.

Therapeutic targeting of ABC transporter expression to treat neurological disease

If a clear relationship between the activity of specific ABC transporters and neurological disease can be identified then a novel therapeutic approach may arise. In the two examples discussed below much is known already regarding the regulation of gene expression for the ABC transporters in question and therefore novel pharmacologic interventions have already been assessed in vivo. It is anticipated that both genetic and pharmacologic regulation of ABC transporter expression in humans may be feasible in the near future.

In the first example, several groups have shown that the LXR agonist TO-901317 can significantly reduce amyloid deposition in mouse models of AD (Koldamova et al. 2005b; Burns et al. 2006; Cao et al. 2007; Koldamova and Lefterov 2007; Riddell et al. 2007). There is evidence that additional LXR agonists, such as GW3965 also potently modulate LXR-regulated genes in the brain (Zelcer et al. 2007); this compound may also similarly reduce amyloid deposition in vivo through an LXR dependent pathway, although this remains to be established. While the initial focus of the LXR-inducible genes responsible for this anti-amyloidogenic effect was on ABCA1, it is now clear that induction of ABCG1 as well as other non-ABC transporter genes including APOE, SREBP-1c and attenuation of genes regulating inflammation may play a role (Liang et al. 2004; Kim et al. 2007; Zelcer et al. 2007). It is therefore likely that LXR activation induces multiple pathways that together reduce amyloid production, deposition and clearance in vivo (Cao et al. 2007; Koldamova and Lefterov 2007).

It is established that ABCA1/Abca1 and ABCG1/Abcg1 are regulated by LXR receptors and are major regulators of cellular cholesterol homeostasis (Zelcer and Tontonoz 2006). Both LXRα/Lxrα and LXRβ/Lxrβ subtypes are expressed in the brain and have been shown to regulate neuronal ABCA1/Abca1 resulting in increased cholesterol efflux and reduced Aβ generation (Wang et al. 2002; Whitney et al. 2002; Koldamova et al. 2003; Sun et al. 2003). Endogenous activation of LXRs in the brain is most likely via naturally occurring oxysterols, such as 24(S)-hydroxycholesterol and 24(S), 25-epoxycholesterol (Wong et al. 2007). However, at least in the case of 24(S)-hydroxycholesterol, previous research indicates that micromolar concentrations of oxysterols may be neurotoxic and pro-inflammatory (Kolsch et al. 1999; Alexandrov et al. 2005). Therefore, development of non-toxic synthetic LXR agonists may represent a better strategy than modulation of endogenous LXR ligands.

These studies indicate that LXR is an attractive pharmacologic target for regulating cerebral amyloid deposition and a prospective avenue for therapeutic treatment/prevention of AD. A common obstacle to this strategy is the development of LXR agonists that are permeable to the BBB. Importantly, in vivo data indicating that orally administered TO-901317 or GW3965 significantly activates brain LXR target genes including Abca1 is very promising (Koldamova et al. 2005b; Burns et al. 2006; Riddell et al. 2007; Zelcer et al. 2007). It remains to be seen if these experimental findings can be translated to the human clinical context.

In the second example, a novel treatment for the peroxisomal disorder X-ALD is considered. The approach is based on pharmacologic induction of redundant genes that can compensate for the loss of ABCD1 (ALDP). In vitro and in vivo studies using Abcd1 knockout mice have shown that over-expression of ABCD2 (ALDRP) and ABCD3 (PMP70) can overcome the biochemical defects in VLCFA β-oxidation and prevent the onset of disease symptoms (Pujol et al. 2004). Several pharmacologic agents, including peroxisome proliferators, fibrates and 4-phenylbutyrate (4-PBA), have been shown to up-regulate ABCD2 in vitro (Kemp and Wanders 2007). Unfortunately, an inability to pass through the BBB and issues with target cell specificity have precluded the use of some of these compounds in vivo; although 4-PBA does appear to cross the BBB (Kemp et al. 1998). Further studies indicated that 4-PBA-induced transcription of ABCD2 is dependent on PEX11α and current research is focused on increasing the potency of this approach by modifying 4-PBA structure in ways that prevent its rapid degradation via β-oxidation in the liver (McGuinness et al. 2000; Wei et al. 2000; Kemp and Wanders 2007). It therefore appears likely that a small molecule approach targeting ALDRP gene expression will offer a viable potential treatment for X-ALD patients.

With both of the therapeutic targeting examples mentioned above, pharmacologic effects on extracerebral tissues will not necessarily be problematic as firstly, LXR activators are also under investigation for the treatment of peripheral vascular disease which is increasingly recognized as contributing to the progression of AD (Casserly and Topol 2004; Martins et al. 2006), and secondly, in the case of X-ALD, it is clear that extracerebral tissues, particularly the adrenal gland, accumulate VLCFA (Kemp and Wanders 2007). Treatment of neurological disease related to ABC lipid transporters is, so far, only experimental and it remains to be seen if such approaches can be successfully transferred to the human clinical setting.

Dynamics of ABC transporter expression during human brain embryonic and post-natal development

As already noted, the human brain contains more lipid than any other organ (Johnson et al. 1948; O’Brien and Sampson 1965; Sastry 1985) and the human brain increases in mass ∼4-fold over the first two decades of postnatal life (Dekaban 1978; Beltaifa et al. 2005). Contributing to the dramatic developmental growth of the human brain is an increase in lipids and building of myelin sheaths in both the gray and white matter of the cerebral hemispheres. This process of myelinogenesis is detected as a protracted increase in myelin stain in sections of the developing human cortex (Benes 1989; Benes et al. 1994) and at the molecular level, as an over 30-fold up-regulation of myelin basic protein mRNA and protein in the frontal cortex from the neonatal period until young adulthood. Since ABC transporters may play a role in brain lipid transport necessary for efficient myelination, we predict that the expression of specific ABC transporters would change dramatically during brain development.

Essentially nothing is known regarding the expression of ABC transporters during human brain development and this knowledge may help elucidate the distinct functions of ABC transporters in the brain. High levels of specific ABC transporter expression early in human life would be consistent with a role in active myelination whereas an up-regulation at maturation could suggest a role in myelin maintenance or synapse stabilization. We have charted changes in human brain ABC transporter expression using microarray analysis and found dramatic alterations during postnatal development. For example, we found that ABCA3 was highly expressed in newborns and decreased with maturation, whereas ABCG2 had low expression in newborns, increased expression until school-age and then declined toward adult levels during the teenage years (Fig. 2). While we plan to confirm developmental changes in ABC transporters in these same samples by qRT-PCR and western blotting and to determine which anatomical and cellular sites of expression may contribute to these changes, this gene microarray data clearly indicates that profound changes in the expression of ABC transporters do occur during human brain development and this may eventually provide further insights as to their functions in brain lipid transport and neurological disease.

Figure 2.

 Gene microarray data for human ABCA3 and ABCG2 expression during development (postnatal). Forty-five cases ranging in age from 6 weeks to 49 years were obtained from the University of Maryland Brain and Tissue Bank for Developmental Disorders (NICHHD contract # NO1-HD8-3283; 29 males and 16 females). Total RNA was extracted from gray matter of the middle frontal gyrus (Brodman’s area 46) and was purified through a Qiagen RNA miniKit column (Qiagen Inc, Valencia CA USA). Samples of high quality total RNA were subjected to microarray analysis (hybridized to HG-U133 version 2.0+ GeneChips, Affymetrix, Santa Clara, CA, USA). Rigorous quality control including analysis of 5′ 3′ ratios (included range 0.40–0.79), percent present (included range 37–47%), average pair-wise correlation analysis and principle component analysis (PCA) were preformed, resulting in the exclusion of three individuals. ABCA3 (a) and ABCG2 (b) mRNA expression (log scale) is plotted against postnatal age (log scale). The black circles represent males and the white circles represent females.

Additional ABC transporters may have lipid-dependent and lipid-independent functions in the brain

Our focus has been on the ABC transporters that are expressed in the brain and are thought to transport lipids in a broad sense. Additional ABC transporters are also expressed in the human brain, for example ABCA5, 6, 9 and 10 (Langmann et al. 2003; Ohtsuki et al. 2007); however, nothing is known regarding their possible function as lipid transporters so they are not discussed further here. Similarly ABCA12 mRNA is detectable at a low level in the brain and appears to play a specific role in glucosylceramide transport during skin lipid barrier formation (Langmann et al. 2003; Akiyama et al. 2005). Conversely, ABCG5/Abcg5 and ABCG8/Abcg8 are potent sterol transporters but are not expressed in the brain to a significant degree (Choudhuri et al. 2003; Langmann et al. 2003; Tachikawa et al. 2005). It should also be noted that ABC transporter functions in peripheral tissues may have an impact on the homeostasis of specific CNS lipids. As a case in point, Abcg5/8−/− mouse brain accumulates campesterol, sitosterol and stigmasterol (but not cholesterol) and this is probably due to a selective increase in the concentration of these lipids in the peripheral circulation and their subsequent transport across the BBB (Yu et al. 2004). It is also clear that there are exciting developments highlighting roles for brain ABC transporters in functions not related to lipid transport. For example in regulation of drug efflux at the BBB (Loscher and Potschka 2005; Soontornmalai et al. 2006) and in neural stem cell biology (Broccardo et al. 2006; Lin et al. 2006). However, a discussion of these areas is beyond the scope of the present review.

Conclusion

It is clear that many of the ABC transporters expressed in the brain have already been shown to play a significant role in lipid transport. Mounting evidence indicates that ABC transporter functions may be very specialized in terms of their location at the level of brain region (e.g., ABCA7 in hippocampus), cell type (e.g., ABCA2 in oligodendrocytes), and organelle (e.g., ABCD1 in peroxisomes). Another tier of specialization is added when ABC transporter substrate specificity and selectivity for lipid transport to acceptor molecules is considered. With some notable exceptions (discussed above), much of the information regarding the potential function of ABC transporters in the brain is based on biochemical or in vitro research and it will be important now to assess speculated roles for ABC transporters in vivo where transgenic and gene knockout mouse models can provide important insights both in terms of normal physiological function and, when mouse models of neurological disease are also available, of their role in various pathological settings. A detailed analysis of gene expression profiles during development of the human brain should also provide novel insights regarding ABC transporter functions. We envisage that these approaches together will provide crucial information that will accelerate the development of therapeutic strategies that target brain lipid transport/homeostasis in order to treat specific neurological diseases.

Acknowledgments

We are grateful to Dr Maree J. Webster (Stanley Laboratory of Brain Research, Department of Psychiatry, USUHS, Bethesda, USA) for granting access to the gene microarray data used in Fig. 2 and to Dr Jenny Wong for proof reading the manuscript. Dr Garner is supported by an Australian National Health and Medical Research Council R. D. Wright Fellowship.

Ancillary