Scavenger receptors in neurobiology and neuropathology: Their role on microglia and other cells of the nervous system



Scavenger receptor class A (SR-A, CD204), scavenger receptor-BI (SR-BI), and CD36 are cell surface proteins that mediate cell adhesion to, and endocytosis of, various native and pathologically modified substances, and participate in intracellular signaling, lipid metabolism, and host defense against bacterial pathogens. Microglia, Mato cells, astrocytes, cerebral microvascular endothelial cells, cerebral arterial smooth muscle cells, and retinal pigment epithelial cells express one or more of these SR. Expression of SR-A and SR-BI by microglia is developmentally regulated. Neonatal microglia express SR-A and SR-BI, while microglia in normal mouse and human adult brain express neither. Astrocytes in adult brain express SR-BI. In Alzheimer's disease, microglial expression of SR-A is increased. Such findings, and evidence that SR-A and SR-BI mediate adhesion and endocytosis of fibrillar β-amyloid by microglia and astrocytes, respectively, and that SR-A, SR-BI, and CD36 participate in secretion of reactive oxygen species by microglia, suggest roles for these receptors in homeostasis and neuropathology. GLIA 40:195–205, 2002. © 2002 Wiley-Liss, Inc.


The term “scavenger receptor” (SR) was coined about 20 years ago to describe unidentified high-affinity binding site(s) on macrophages for acetylated low-density lipoproteins (acLDL) (Goldstein et al., 1979). Subsequently, it was shown that macrophages, and other cells in various species, express multiple structurally distinct receptors that mediate binding of acLDL, and other modified lipoproteins, such as oxidized LDL (oxLDL), entitling them to be included in the SR family, e.g., SR class A (SR-A, CD204), SR class B type I (SR-BI), Drosophila SR class C type I (dSR-CI), lectin-like oxidized low-density lipoprotein receptor-1 (LOX-1), macrophage receptor containing a collagenous domain (MARCO), and endothelial cell scavenger receptor (SR-EC), macrosialin/CD68 (Platt and Gordon, 1998). In addition, some of these receptors have been shown to participate in binding and uptake of a variety of structurally unrelated substances, including fibrillar amyloid-β (fAβ), thrombospondin-1 (TSP-1), anionic polysaccharides, polynucleotides, apoptotic eukaryotic cells, and various bacteria (Table 1). To describe the promiscuous affinity of SR for what on first examination appears to be a heterogeneous and unrelated group of ligands, Krieger and Stern (2001) termed SR-A, SR-BI, and CD36, the receptor for advanced glycation end products (RAGE) and the LDL-like receptor related protein (LRP) as “multiligand receptors,” while Franc et al. (1999) have called them “pattern recognition” receptors. Since the structural basis for ligand binding of these receptors remains uncertain, and many cell surface receptors (e.g., integrins) bind multiple structurally dissimilar ligands, we have retained the term “scavenger receptors.” This review focuses on SR-A, SR-BI, and CD36 expressed by various cells in the central nervous system (CNS). The roles of RAGE and LRP have been recently reviewed by Herz and Strickland (2001) and Schmidt et al. (2001). The role of glia in Alzheimer's disease (AD) also has been reviewed (Terry, 1999).

Table 1. Putative ligands for scavenger receptors
  1. LDL, low-density lipoprotein; acLDL, acetylated low-density lipoprotein; oxLDL, oxidized low-density lipoprotein; HDL, high-density lipoprotein; BSA, bovine serum albumin; mBSA, maleylated BSA; glyc CIV, glycated collagen type IV; TSP-1, thrombospondin-1; fAβ, fibrillar β-amyloid; AGE, advanced glycation endproducts; CS, chondroitin sulfate; poly I, polyinosinic acid; PS, phosphatidylserine; iRBC, Plasmodium falciparum-infected erythrocytes; LPS, lipopolysaccaride; LTA, lipoteichoic acid; −, not able to bind or compete; ?, not tested.

Long-chain fatty acids+
glyc CIV+??
poly I+?
Apoptotic cells+++
Bacteria, LPS, LTA+??


The principal structural features of SR-A, SR-BI, and CD36 have been reviewed elsewhere, and their putative ligands are listed in Table 1 (Krieger and Stern, 2001). We use the term “putative ligands” because most of these substances have not been shown to bind to purified SR-A, SR-BI, or CD36. They have been characterized as ligands on the basis of one or more of the following properties: (1) they are bound or endocytosed by cells that express SR, or by cells transfected with cDNA encoding one of these SR; (2) cells expressing one or more SR adhere in a divalent cation-independent fashion to matrices containing these SR ligands; (3) SR-specific antibodies block ligand uptake by SR-expressing cells, and/or adhesion of these cells to surfaces coated with putative SR ligands; and (4) putative ligands compete with other SR ligands for binding and/or uptake by cells, and/or block adhesion of cells expressing SR receptors to surfaces containing SR ligands. Although we have used several of these methods to identify SR ligands, we recognize that, with the exception of anti-receptor antibodies, these methods do not distinguish substances that mask SR ligands from substances that bind directly to SR (El Khoury et al., 1996; Maxeiner et al., 1998; Husemann et al., 2001a). For example, fucoidan (synonym: fucoidin), a sulfated polysaccharide, has been designated a ligand for SR-A on the basis of its ability to inhibit binding and uptake of acLDL by macrophages (Brown et al., 1980). However, the inhibitory activity of fucoidan may reflect its capacity to mask site(s) on ligands, which interact with one or more SR, rather than masking the receptor itself. Similarly, we suggest that an 18-amino acid peptide derived from the putative ligand-binding domain of SR-A blocks adhesion of macrophages and microglia to matrices containing fAβ 1–42 by masking the ligand, rather than SR-A (El Khoury et al., 1996).

At first glance, ligands for SR appear heterogeneous in composition and structure (Table 1). Closer examination reveals two properties shared by many, but not all, of them. First, many ligands are polyanions (e.g., polyinosinic acid, fucoidan, heparin sulfate, lipoteichoic acid), or proteins in which epsilon amino groups of lysine have been modified (e.g., acLDL, oxLDL, glycated collagen, maleylated albumin). Modification of lysine groups often renders the entire protein more anionic. Second, several ligands contain cross-pleated β-sheets, e.g., fibrillar forms of Aβ1–42 or Aβ 25-35 (Lorenzo and Yankner, 1994). Whether these motifs are the critical determinants recognized by one or multiple ligand-binding domains on SR-A, SR-BI and CD36 remains to be determined.


SR-A (CD204)

In peripheral tissues, SR-A is constitutively expressed by mononuclear phagocytes, e.g., macrophages, dendritic cells, and Kupffer cells (Yamada et al., 1998). In the nervous system, SR-A in human brain was first identified by Northern blot analysis (Matsumoto et al., 1990). In brain from normal adult humans, Christie et al. (1996) reported SR-A expression on microglia, but not on Mato cells (perivascular macrophages). In contrast, Honda et al. (1998) used a different antibody than that used by Christie et al. (1996) and reported little or no expression of SR-A by microglia in normal human brain, but observed staining on Mato cells. We have carried out similar studies of adult mouse brain using a monoclonal anti-SR-A antibody (clone 2F8) and observed staining on Mato cells, but not on microglia, consistent with the results of Honda et al. (1998) (J. Husemann, unpublished data). Other cells of the brain also may express SR-A. Lucarelli et al. (1997) found SR-A transcripts in bovine brain microvessel preparations and suggested that endothelial cells express SR-A. While immunochemical techniques revealed SR-A expression on cultured canine cerebrovascular smooth muscle cells, there are no reports of SR-A expression on vascular endothelium or smooth muscle cells in human brain (Prior et al., 2000).

Expression of SR-A appears to be downregulated postnatally in mouse brain. In the mouse neonate, SR-A is highly expressed on meningeal macrophages and detectable in low amounts on immature microglia in the supraventricular corpus callosum, cingulum, cavum septum, and the periaqueductal area (Bell et al., 1994). SR-A is expressed by cultured neonatal primary mouse microglia and N9 cells, a mouse neonatal microglial cell line (Righi et al., 1991; El Khoury et al., 1996; Husemann and Silverstein, 2001). In contrast, SR-A is not expressed by resident microglia in adult mouse brain but is expressed by stromal and epiplexus macrophages of the choroid plexus, meningeal macrophages and by Mato cells (Bell et al., 1994). Thus, SR-A expression by microglia in mouse brain appears to be developmentally regulated. Whether SR-A expression is also developmentally regulated in human brain remains to be explored.

In vitro, SR-A expression by mouse peritoneal macrophages is downregulated by tumor necrosis factor-α (TNF-α) or interferon-γ (IFN-γ), and upregulated by macrophage-colony-stimulating factor (M-CSF) (Yamada et al., 1998). In vivo, SR-A expression by microglia is upregulated by intracerebral administration of either lipopolysaccharide or kainic acid (Bell et al., 1994; Perry, 1994; Perry and Andersson, 1992). Grewal et al. (1997) reported that expression of both SR-A mRNA and protein by microglia in vivo and in vitro are upregulated by lipopolysaccharide, IFN-γ, and IL-1α, but not by TGF-β1.


Scavenger receptor class B type I (SR-BI), also known as CD36 LIMPII analogous-1 (CLA-1), is expressed on monocytes/macrophages, endothelium, in tissues that synthesize large amounts of cholesterol (i.e., liver), and in tissues that consume large amounts of cholesterol for steroid hormone synthesis (i.e., adrenal cortex, and ovary) (Murao et al., 1997; Buechler et al., 1999; Hirano et al., 1999; Krieger, 1999). SR-BI binds HDL (Arai et al., 1999) and several other substrates (Table 1). It participates in HDL-mediated transport of lipids and cholesterol to SR-BI-expressing cells in steroidogenic tissues or the liver (Krieger, 1999). The delivery of cholesterol from peripheral tissues via HDL and SR-BI to the liver for biliary excretion is called “reverse cholesterol transport”. It is believed to be one of the principal mechanisms by which HDL protects against atherosclerosis (Trigatti et al., 1999).

SR-BI is present in mouse, rat and human brain homogenates, on cultured porcine brain capillary endothelial cells and canine leptomeningeal smooth muscle cells (Acton et al., 1994; Jurevics et al., 2000; Prior et al., 2000; Goti et al., 2001; Srivastava and Jain, 2002). We have used immunocytochemistry to identify SR-BI on astrocytes and cerebral arterial smooth muscle cells of the leptomeninges and parenchyma in adult mouse and human brain (Husemann and Silverstein, 2001). Capillaries in cortical and sub-cortical areas of normal adult human and mouse brain stain with SR-BI antibodies. However, it is difficult to determine whether the SR-BI antibodies react only with astrocyte processes associated with capillary astrocytic sheaths or with endothelial cells as well (Husemann and Silverstein, 2001). Thus, we are uncertain whether both astrocytes and capillary endothelial cells in these tissue sections express SR-BI. In addition, SR-BI and another splice product of the SR-B gene, SR-B type II (SR-BII), have been detected on retinal pigment epithelial cells in humans (Duncan et al., 2002). While no evidence is provided, these authors suggest that SR-BI and SR-BII participate in clearance of photoreceptor outer rod segments.

Like SR-A, SR-BI expression is developmentally regulated in microglia. SR-BI is expressed on microglia in newborn mouse brain and on cultured neonatal murine microglial cells but is absent on microglia in adult mouse brain (Husemann et al., 2001a; Husemann and Silverstein, 2001; Husemann, unpublished observation). Further work is required to determine whether SR-BI is expressed by neonatal human microglia, and whether cessation of SR-A and SR-BI expression by microglia occurs simultaneously or sequentially. Further work also will be required to identify the stage of brain maturation at which these changes in SR expression occur, to assess the mechanism(s) responsible for these changes, and to determine whether like SR-A, SR-BI expression is affected by CNS diseases.

Both TNF-α and IFN-γ downregulate SR-BI expression in blood monocytes and monocyte-derived macrophages (Buechler et al., 1999). Whether these cytokines have a similar effect on SR-BI expression by neonatal microglia, adult astrocytes or other CNS-derived cells remains to be determined.


In the periphery, CD36 is expressed on a variety of cell types, including monocytes/macrophages, microvascular endothelium, platelets, adipocytes, and cardiomyocytes (Febbraio et al., 2001). Furthermore, CD36 is highly expressed on retinal pigment epithelial cells in humans, where it has been suggested to play a role in the clearance of photoreceptor rod outer segments (Ryeom et al., 1996). In normal adult human brain, capillary endothelium expresses high levels of CD36, which are readily detected using conventional immunohistological methods (Barnwell et al., 1989). Coraci et al. (2002) report capillary and microglial staining for CD36 in normal adult human brain using a highly sensitive immunostaining amplification technique. Furthermore, they detected CD36 on primary neonatal human microglia and on N9 cells, a murine neonatal microglial cell line, using mouse antihuman CD36 antibodies (clones FA6-152 and SMΦ). However, other investigators reported that these and other antibodies of murine origin directed against human CD36 are not suitable for detection of murine CD36 (Daviet et al., 1995; Navazo et al., 1996; Puente Navazo et al., 1996). To address this issue, we used an antibody specific for murine CD36 (clone 63; murine mouse antimouse CD36 IgA) to examine expression of CD36 by mouse microglia both in vitro and in situ. We detected expression of CD36 on cultured neonatal microglia from wild-type mice, but not on microglia from CD36−/− mice (Fig. 1). These findings indicate that clone 63 is an appropriate antibody to detect mouse CD36 expressed by murine neonatal microglia in vitro. However, when we used clone 63 to detect CD36 in vivo, with acetone-fixed frozen brain sections from adult wild-type and CD36−/− mice, we observed staining of capillaries and unidentified cells in the gray and white matter in both wild-type and CD36−/− brain (not shown). Staining intensity in both wild-type and CD36−/− brain was identical and correlated positively with the concentration of clone 63 antibody used (not shown). Control experiments showed that a purified mouse IgA (Serotec) used as a negative control showed no staining, even at concentrations as high as 100 μg/ml (not shown). These findings show that clone 63 binds to unidentified sites, possibly microvasculature and microglia, in acetone-fixed wild-type and CD36−/− mouse brain and may not be suitable for immunohistochemistry of murine brain tissue. Furthermore, we advise caution in interpreting immunoreactivity in mouse tissue using clone 63 or other antibodies as “positive,” unless absence of staining in similar tissue from CD36−/− mice is observed. As noted above, the absence of staining with “control” IgA is unreliable as a measure of specificity.

Figure 1.

Expression of CD36 by murine neonatal microglia from wild-type mice (left) and CD36−/− mice (right). Wild-type or CD36−/− microglia adherent to glass slides were incubated with either commercially available murine antimouse CD36 IgA (clone 63; 1:100 dilution), purified mouse IgA (50 μg/ml; Serotec, Raleigh, NC) as a negative control, or in buffer without IgA for 1 h, followed by incubation with biotinylated goat antimouse IgA (Sigma; 1:250 dilution) and Alexa 488-conjugated streptavidin (Molecular Probes, Eugene, OR; 1:1,000) as described (Husemann et al., 2001a). Microglia from wild-type mice showed strong immunoreactivity with clone 63 (left), whereas CD36−/− microglia (right) showed no staining above control (mouse IgA, or buffer without IgA; not shown). Nuclei (blue) are stained with DAPI (Vector, Burlingame, CA).

Although there are no reports of developmental or cytokine-mediated regulation of CD36 in any cell of the brain, CD36 is expressed at low levels in freshly explanted human monocytes and highly expressed in monocyte-derived macrophages (Huh et al., 1995). CD36 expression by macrophages is upregulated by oxLDL or M-CSF and is downregulated by IFN-γ (Huh et al., 1996; Nakagawa et al., 1998).

In summary, cultured murine neonatal microglia express high levels of SR-A, SR-BI, and CD36, while cultured human neonatal microglia express high levels of CD36. In neonatal mouse brain, microglia express low levels of SR-A, whereas in adult mouse brain microglia do not express detectable levels of SR-A or SR-BI. In the adult human brain, microglia express low levels of CD36. Further work is required to determine expression levels for all SR in vitro and in situ for neonatal and adult microglia.

Astrocytes and vascular smooth muscle cells in the adult human and mouse brain express SR-BI but no data are available on whether expression of this receptor is developmentally regulated by these cells in vivo or in vitro.

Because neonatal microglia and astrocytes are easy to isolate and grow in culture, almost all studies heretofore have used glial cells from newborn mice. The findings reported above indicate that one must be cautious in extrapolating results obtained with neonatal microglia to the adult condition because they do not display the phenotype of adult microglia and possibly adult astrocytes, at least insofar as SR-A, SR-BI, and CD36 expression is concerned. Absent data on SR-BI expression by newborn astrocytes, a similar concern is relevant to experiments with these cells.


Conventional wisdom suggests that SR-A, SR-BI, and CD36 are involved in CNS homeostasis as mediators of brain lipid metabolism, phagocytosis of apoptotic cells, and endocytosis of native, denatured, and chemically modified proteins and lipoproteins, as well as activators of signal transduction pathways upon ligand binding. However, it remains unclear how essential these roles are since mice genetically deficient in SR-A, SR-BI, or CD36 exhibit no evident brain pathology (Suzuki et al., 1997; Febbraio et al., 1999; Trigatti et al., 1999). In addition, humans genetically deficient in CD36 show no more behavioral abnormalities than CD36 sufficient humans (Yamamoto et al., 1994). It is possible that the overlap in ligand specificity exhibited by these receptors reflects their functional redundancy and that double or triple knockout receptor mice will be more informative as to the function(s) of these receptors in the CNS.

Lipid Metabolism

In the blood, lipids and cholesterol are transported by lipoproteins. LDL contains mainly apolipoprotein B, while HDL contains apolipoproteins A-I, A-II, C, and E. However, little is known about lipid delivery and clearance within the CNS. Lipoproteins found in the cerebrospinal fluid (CSF) are HDL-like and are produced mainly by astrocytes (Ladu et al., 2000).

SR-BI mediates exchange of cholesterol and other lipids between cells and HDL. In the brain, SR-BI on astrocytes and cerebral artery smooth muscle cells may serve a similar function (Husemann and Silverstein, 2001). We have observed that cultured adult murine astrocytes endocytose DiI-labeled HDL (Fig. 2) and hypothesize that SR-BI on these cells mediates binding and uptake of HDL-like lipoproteins in the CNS. SR-BI is reported to promote uptake of HDL-associated vitamin E into cerebral endothelial cells (Goti et al., 2001). Assuming, SR-BI is expressed on capillary endothelium in the brain, it may play an important role in the transport of HDL, cholesterol, lipids, and fat-soluble vitamins, such as vitamin E, from plasma to CNS.

Figure 2.

Uptake of DiI-labeled HDL by adult murine astrocytes. Cells were incubated with 10 μg/ml DiI-labeled HDL (Intracel, Rockville, MD) in Krebs-Ringer buffer with 0.1% glucose and 0.1% BSA for 15 min (fluorescence, left; phase contrast, right).

Consistent with the role of SR-BI in lipid homeostasis, SR-BI−/− mice have total plasma cholesterol levels that are about two-fold above normal. Moreover, their circulating HDL particles are abnormally large, heterogeneous, and apoE-enriched (Trigatti et al., 1999). Effects of SR-BI deficiency on CNS lipid metabolism in these animals remain to be examined.

It is well established that CD36 participates in binding and transport of long-chain fatty acids (LCFA) in cells and tissues such as adipocytes or myocardium (Febbraio et al., 2001). CD36−/− mice have a significant increase in plasma free fatty acid and triacylglycerol levels and CD36-deficient humans show abnormal myocardial LCFA uptake (Hwang et al., 1998; Febbraio et al., 1999).

CD36 binds HDL with high affinity but has a much lower capacity to exchange HDL cholesterol than SR-BI (Febbraio et al., 2001). This suggests a minor role for CD36 in cholesterol metabolism. However, CD36-deficient humans and mice have significantly elevated plasma cholesterol levels (Febbraio et al., 1999; Yanai et al., 2000). Whether CD36 plays a role in brain lipid metabolism is unknown.


SR-A, SR-BI, and CD36 have been implicated in the clearance of apoptotic cells, cell debris, bacteria, myelin, and outer rod segments. However, most investigators have examined this issue utilizing non-CNS cell types. For example, SR-A on murine peritoneal macrophages has been implicated in the clearance of myelin debris and apoptotic thymocytes (Platt et al., 1996; da Costa et al., 1997). SR-BI-transfected Chinese hamster ovary (CHO) cells have been shown to endocytose apoptotic thymocytes and apoptotic MKM cells, a mouse bone marrow-derived mast cell line (Fukasawa et al., 1996; Murao et al., 1997). CD36 has been implicated in the clearance of apoptotic neutrophils by human monocyte-derived macrophages (Savill et al., 1992). COS cells (a monkey kidney cell line) and human Bowes melonoma cells are reported to ingest apoptotic neutrophils, lymphocytes, and fibroblasts when transfected with CD36, whereas nontransfected COS and Bowes melanoma cells do not (Ren and Savill, 1995; Puente Navazo et al., 1996).

In the CNS, SR on both microglia and astrocytes have been shown to participate in phagocytosis of apoptotic cells. For example, SR-A on neonatal microglia mediates the binding and phagocytosis of apoptotic cells that express phosphatidylserine (Fadok et al., 1992; Ashman et al., 1995). Chang et al. (2000) reported that astrocytes phagocytose apoptotic glioma cells, and we have shown that adult astrocytes express SR-BI. Thus, it seems likely that SR-BI participates in astrocyte clearance of apoptotic cells in vivo (Husemann and Silverstein, 2001).

Microglia in adult human brain express low levels of CD36 (Coraci et al., 2002). Whether CD36 plays a role in phagocytosis of apoptotic cells by microglia in the CNS, as it does in the uptake of apoptotic neutrophils, lymphocytes, and fibroblasts by monocyte-derived macrophages and dendritic cells, remains to be tested (Savill et al., 1992; Ren and Savill, 1995; Puente Navazo et al., 1996; Albert et al., 1998).

CD36 has been reported to be involved in the clearance of photoreceptor rod outer segments by retinal pigment epithelium (Ryeom et al., 1996). Thus far, there are no reported visual abnormalities associated with CD36 deficiency in humans, or in CD36−/− mice.

Lipoproteins, including CSF lipoproteins, are susceptible to oxidation (Bassett et al., 1999). Oxidatively modified lipoproteins are neurotoxic and stimulate macrophages, astrocytes and microglia to secrete neurotoxic substances, e.g., reactive oxygen species (ROS), in vitro (Kivatinitz et al., 1997; Maxeiner et al., 1998; Bassett et al., 1999; Keller et al., 1999, 2000). These reports suggest a need for rapid clearance of oxidized lipoproteins from the CNS to prevent cell damage and disease. In the periphery, oxidized lipoproteins are effectively removed from the circulation by SR-bearing cells in the liver (van Oosten et al., 1998, 2001). Similarly, SR may mediate this process in the brain. However, there are no reports of a role for SR-A, SR-BI, or CD36 in clearance of modified lipoproteins by CNS-derived cells. In fact, the mechanism(s) that mediate clearance of modified lipoproteins from the CNS have not been studied in detail.


Several studies suggest that CD36 mediates cell signaling events: Ockenhouse et al. (1989) report that antibody cross-linking of CD36 signals ROS production by human monocyte-derived macrophages. McGilvray et al. (2000) found that interaction of plasmodium-infected erythrocytes with macrophages stimulated clustering of CD36, phorphorylation of Erk and p38 MAPK, and phagocytosis of erythrocytes, and Huang et al. (1991) showed that CD36 is physically associated with Fyn, Lyn, and Yes protein-tyrosine kinases in platelets, and that ligation of CD36 activates these cells. Consistent with the concept that SR trigger signal transduction, Maxeiner et al. (1989) reported that antibodies against CD36 inhibit by ∼50% secretion of reactive oxygen species (ROS) by human monocyte-derived macrophages plated on matrices coated with oxLDL. It is unclear how to reconcile these results with Husemann et al. (2001b), who showed that peritoneal macrophages from wild-type and CD36−/− mice produce similar amounts of ROS when plated on these matrices. In addition, Janabi et al. (2000) reported that CD36-deficient human macrophages secrete significantly smaller amounts of TNF-α and interleukin-1β (IL-1β) and exhibit reduced nuclear factor-κB (NF-κB) activation in response to oxLDL than control macrophages, suggesting a role for CD36 and CD36 ligands in inflammation. Whether interaction of CD36 with specific ligands leads to activation of CD36-bearing cells in the CNS remains to be explored.

Until recently, there has been no direct evidence that SR-BI transduces signals. Yuhanna et al. (2001) provide data that support SR-BI signaling by showing that interaction of HDL-associated apo A-I with SR-BI activates eNOS in arterial endothelial cells.


Alzheimer's Disease

AD is the most common cause of dementia in the elderly and is characterized by parenchymal deposits of fibrillar amyloid-β (fAβ) surrounded by microglia and astrocytes. In culture, these and other SR-bearing cells from the CNS interact with fAβ which results in activation of nuclear factor-κB (NF-κB) and its binding to DNA, increases in protein-tyrosine phosphorylation, and secretion of pro-inflammatory substances, e.g., reactive oxygen species (ROS), TNF-α, and complement components by these cells (McDonald et al., 1998; Combs et al., 1999, 2001; Terry, 1999; Nakai et al., 2001). SR-A, SR-BI, and CD36 have all been suggested to participate in the pathogenesis of AD.

SR-A is upregulated in a subset of reactive microglia in lesions of ischemia and in microglia associated with amyloid deposits in brains of patients with AD (Christie et al., 1996; Honda et al., 1998). While these investigators differ with respect to expression of SR-A by Mato cells and by microglia in normal human brain, they agree that SR-A is strongly expressed by microglia in AD brain. Similar observations were made in brains from APP23 mice, a transgenic mouse model for AD, where SR-A is localized on microglia associated with parenchymal and vascular amyloid deposits (Bornemann et al., 2001). These observations suggest induction of SR-A on microglia in pathological conditions. Whether upregulation of SR-A in AD brain results from direct microglial interaction with fAβ, and/or activation of these cells by cytokines/chemokines present in AD lesions are important areas to be investigated.

SR-A was the first receptor shown to participate in binding and internalization of fAβ 1–42 by cultured murine microglia and other mononuclear phagocytes (El Khoury et al., 1996; Paresce et al., 1996). Subsequently, we and others reported that neonatal microglia from SR-A−/− mice bound ∼50% less fluorescently tagged fAβ 1–42 than wild-type microglia (Chung et al., 2001; Husemann et al., 2001a). These findings, and the observation that microglia associated with Aβ deposits in AD brain express SR-A, suggested a role for this receptor in the pathogenesis of AD. However, Huang et al. (1999) found that transgenic mice lacking SR-A (SR-A−/−) and overexpressing human amyloid precursor protein (APP) in their brains (SR-A−/−/PDAPP mice) exhibited AD-like brain pathology similar to that observed in SR-A+/+/PDAPP mice. These investigators concluded that interaction between SR-A and fAβ is not required for the development of AD-like pathology in mice. However, their results do not rule out a role for microglia in AD. We have found that neonatal microglia from SR-A−/− mice adhere to matrix-associated fAβ, and secrete ROS in response to fAβ as efficiently as wild-type microglia. These findings suggest that cell receptors other than SR-A mediate Aβ-dependent ROS secretion by these cells in adult mice (Husemann et al., 2001a). Whether SR-A-fAβ interaction leads to the production of pro-inflammatory substances other than ROS by microglia remains to be examined.

Coraci et al. (2002) have implicated CD36 in the pathogenesis of AD. These investigators detected low levels of CD36 on microglia in normal adult brain and in brains of patients with AD, multiple sclerosis and Parkinson's disease but could not discriminate whether there were differences in CD36 expression in normal versus diseased brain. Furthermore, these authors identified CD36 as a receptor for fAβ 1–42 that participates in signaling ROS production by human monocyte-derived macrophages, human neonatal microglia, and N9 cells upon interaction with fAβ 1–42 in vitro. They found that murine antibodies directed against human CD36 (clones SMΦ and NL07) reduced by ∼50% secretion of ROS when these cells were plated on surfaces coated with fAβ 1–42 but only minimally affected cell adhesion to these surfaces. Although these authors observed a significant inhibitory effect of mouse antihuman CD36 antibodies on secretion of ROS upon interaction of N9 cells with fAβ 1–42, others reported that these antibodies do not bind to murine CD36 (Daviet et al., 1995; Navazo et al., 1996; Puente Navazo et al., 1996). Further studies are needed to determine the mechanism(s) by which these antibodies inhibit ROS secretion by mouse and human mononuclear phagocytes (Maxeiner et al, 1998; Coraci et al., 2002).

We have used CD36−/− microglia to further explore the role(s) of CD36 in fAβ 1–42-stimulated ROS production. Since adhesion of microglia to plastic or glass stimulates them to produce ROS, we used surfaces coated with collagen IV, a matrix protein to which microglia adhere weakly, if at all, to measure basal ROS release by these cells (Husemann et al., 2001a). We compared H2O2 production by wild-type and CD36−/− murine neonatal microglia plated on surfaces coated with collagen IV alone with surfaces coated with collagen IV and fAβ, or plated on collagen IV and stimulated with phorbol myristate acetate (PMA), a protein kinase C-activator known to signal maximal H2O2 production by mononuclear phagocytes (Table 2). The results of these experiments were informative and surprising. First, CD36−/− microglia released 68% more H2O2 than wild-type microglia when plated on collagen IV alone. Second, there was no significant difference in H2O2 production by CD36−/− microglia plated on surfaces coated with collagen IV alone than on surfaces coated with collagen IV and fAβ. As expected, microglia from wild-type mice produced 23% more H2O2 when plated on surfaces coated with collagen IV and fAβ than on surfaces coated with collagen IV alone (Coraci et al., 2002; Husemann et al., 2001a). Third, PMA stimulated a 57% increase in H2O2 production by wild-type microglia but no significant increase in H2O2 production by CD36−/− microglia. These experiments indicate that, at least with respect to H2O2 production, neonatal microglia from CD36−/− mice are constitutively “activated.” They suggest that in wild-type microglia CD36 exerts a negative regulatory effect on protein kinase C and/or other enzymes that regulate H2O2 production. Therefore, it seems likely that CD36−/− mice will be less helpful than hypothesized in dissecting the role(s) of CD36 in signaling ROS production stimulated by fAβ 1–42 in wild-type microglia. Finally, these experiments emphasize the need for caution in ascribing changes in phenotype and/or physiological functions to specific receptors and biochemical pathways in knockout mice. In the case of CD36 deficient microglia, absence of CD36 appears to correlate with a constitutive gain of function.

Table 2. H2O2 production (in pmoles/150,000 cells) by neonatal microglia from wild-type or CD36−/− mice under unstimulated conditions and in response to fAβ 1–42 or PMA
 Without stimulus+ fAβ 1–42+PMA
  • PMA, phorbol myristate acetate; BSA, bovine serum albumin.

  • Neonatal microglia from wild-type and CD36−/− mice (Febbraio et al., 1999) were generated as described (Bard et al., 2000). To measure release of H2O2 by wild-type or CD36−/− microglia, 150,000 cells suspended in 100 μl Krebs-Ringer buffer containing 0.1% glucose (Sigma, St. Louis, MO) and 0.1% BSA (Sigma) and Amplex Red/horseradish peroxidase (Molecular Probes, Eugene, OR) without or with 100 ng/ml PMA (Sigma), were added to wells of 96-well plates coated with either collagen IV alone or collagen IV and fAβ 1–42. incubated for 1 h at 37°C in a 5% CO2/95% air atmosphere, assayed as described (Coraci et al., 2002; Husemann et al., 2001a). Data are presented as mean ± SEM of four independent experiments. Δ, mean of pmoles H2O2 released by cells in response to fAβ 1–42 or PMA-pmoles H2O2 released by cells with no stimulus ± SEM. Statistical significance was tested with Student's t-test.

  • *

    Compared with wild-type microglia without stimulus.

  • **

    Compared with CD36−/− microglia without stimulus; n.s.: not significant).

Wild-type microglia422 ± 20519 ± 19 (Δ: 97 ± 25; P ≤ 0.05*)663 ± 63 (Δ: 241 ± 67; P ≤ 0.05*)
CD36−/− microglia711 ± 53; P ≤ 0.005*)731 ± 48 (Δ: 21 ± 12; n.s.**)801 ± 83  (Δ: 91 ± 42; n.s.**)

SR-BI participates in binding and endocytosis of fAβ 1–42 by adult astrocytes and cerebrovascular smooth muscle cells in vitro (Prior et al., 2000; Husemann, Wyss-Coray et al., submitted). The role of SR-BI in Aβ clearance by these cells in vivo in normal and AD brain remains to be determined. Whether interaction of SR-BI with fAβ 1–42 activates signal transduction pathways in astrocytes and cerebrovascular smooth muscle cells, as it does with HDL in endothelium (Yuhanna et al., 2001), is unknown.

Increased production of human Aβ in the brains of adult mice expressing a transgene encoding this protein resulted in AD-like pathology, synaptic transmission deficits, and behavioral impairments. Immunization of these mice with human fAβ resulted in marked reductions in the Aβ burdens in the brains of these mice and protected them from the age-related behavioral impairments seen in un-immunized littermates (Schenk et al., 1999; Morgan et al., 2000). Several mechanisms have been suggested as mediating these effects (Bard et al., 2000; DeMattos et al., 2001). To these should be added the possibility that antibodies versus fAβ increase the accessibility of this protein to SR-bearing CNS cells, facilitating the removal of this protein.

Diabetic Peripheral Neuropathy

Glucose-modified proteins, called advanced glycation end products (AGEs) accumulate on long-lived proteins such as myelin and extracellular matrix proteins in patients with diabetes (Brownlee, 1992). These AGE-induced changes can affect cell adhesion, growth, and matrix accumulation. SR-A, SR-BI, and CD36 all serve as receptors for AGEs (El Khoury et al., 1994; Ohgami et al., 2001a, b). AGE-modified proteins also alter vascular function by interacting with specific receptors on macrophages and endothelial cells, inducing changes that are associated with matrix overproduction, focal thrombosis, and vasoconstriction (Baynes, 2001). Thus, interactions of SR with AGEs may contribute to the development of diabetic peripheral neuropathy.

Demyelinating Diseases

While diabetic peripheral neuropathy may reflect excessive interactions of SR with glycated myelin and matrix proteins, deficits in SR activity may lead to accumulation of potentially neurotoxic debris in the nervous system. SR-A mediates phagocytosis of myelin by mouse macrophages (da Costa et al., 1997). Consistent with the hypothesis that SR-A participates in removal of damaged myelin, SR-A−/− mice develop significantly more “onion bulbs” after compression injury of peripheral nerve than wild-type mice (Naba et al., 2000). Onion bulbs are observed in acquired neuropathies associated with chronic inflammation, and in demyelining polyradiculopathy and diabetic neuropathy. Thus, these observations imply role(s) for SR in these diseases.


Historically, the capacity of a plasma membrane receptor to recognize modified lipoproteins was used to classify it as a “scavenger receptor” (SR). Subsequently it was discovered that various SR bind multiple, structurally dissimilar ligands (Table 1). Recently, other “multi-ligand receptors”, e.g. RAGE and LRP, have been added to the expanding family of SR. Today, over 20 structurally distinct receptors exhibiting overlapping ligand specificities and functions have been termed “scavenger receptors”.

In mice, SR-A, SR-BI and CD36 are highly expressed on cultured neonatal microglia and SR-A is expressed by microglia in neonatal brain. In contrast, microglia in adult mouse brain express no SR-A or SR-BI. These findings show that expression of SR-A, and possibly other SR, in the mouse brain is developmentally regulated. Whether microglia express SR-BI and CD36 in newborn mouse brain, and whether CD36 is expressed by microglia in the adult mouse brain, and whether cultured adult microglia have different expression levels of SR compared to their neonatal counterparts remains to be determined.

In humans, CD36 is highly expressed by fetal microglia in culture and low levels of CD36 and SR-A, but not SR-BI, have been detected on microglia in adult brain. More research is required to examine expression levels of SR in neonatal/adult human microglia in culture and in situ.

Futhermore, in adult mouse and human brain SR-A is expressed on Mato cells, SR-BI is expressed on astrocytes and vascular smooth muscle cells, and CD36 is expressed on microvasular endothelial cells. Whether expression of SR is developmentally regulated on these cells, or whether SR levels change in culture is unknown.

The functions of these receptors in the CNS most likely parallel their roles in the periphery. They are summarized in figure 3 and include: 1) phagocytosis of apoptotic cells; 2) removal of cellular debris such as membrane fragments and myelin; 3) activation of intracellular signaling events; 4) internalization and degradation of a variety of substances such as oxidized lipoproteins Aβ, bacteria and bacterial products; and 5) cell adhesion to matrixes that contain ligands for SR.

Figure 3.

Putative functions of SR-A, SR-BI, and CD36 (see text).

Neonatal microglia are relatively easy to grow in culture while adult microglia are not. Hence most studies of microglial interactions with apoptotic cells or with matrixes containing membrane fragments, denatured myelin proteins, Aβ, oxidized lipoproteins, bacteria, or bacterial products have been performed with neonatal microglia. Phenotypic differences in expression of SR-A, SR-BI, CD36 and other (SR) receptors by neonatal vs. adult microglia may lead to erroneous conclusions with respect to the functions of these cells in homeostasis and diseases (e.g., Alzheimer's disease) of adult brain. In pathological states, such as in Alzheimer's disease, only SR-A has been shown to be upregulated, and only on microglia.

The availability of mice genetically deficient in each of these scavenger receptors has facilitated exploration of some of their functions. However, as this review indicates, we know very little about their roles in the CNS since mice genetically deficient in each of these scavenger receptors exhibit no obvious deficits in neural development, homeostasis, or pathology. Perhaps mice with genetic deficits in two or three of these receptors will prove more revealing.


The authors thank Dr. Ira Goldberg and Ms. Sadna Budhu for critical reading of this manuscript and insightful suggestions. The studies from our laboratory were initiated with the support of grants AI20516 from the NIH (to S.C.S.) and RG96-067 from the Alzheimer's Association (to S.C.S.) and are currently supported by NIH grant AG19772 (to S.C.S.), “Pilot Grant Award” from Columbia University's Alzheimer's Disease Research Center (to J.H.) and IIRG-02-3510 from the Alzheimer's Association (to J.H.).