5 January 2008. A transcriptome database comparing gene expression of cell types highly purified from fresh mouse brain is offering a trove of new information to define the distinct identities of the brain cells that are traditionally subsumed under the vague grab bag designation of “glia.” In the January 2 Journal of Neuroscience, researchers led by Ben Barres at Stanford University, Palo Alto, California, present a resource to the field that can help scientists finally determine what astrocytes are doing in the brain while synapses form during postnatal development. It might also provide clues to what astrocytes do while synapses degenerate later in life in neurologic diseases such as Alzheimer’s. The work has already produced new markers for CNS cell types, and begun to paint a picture of astrocytes as phagocytic cells.
Astrocytes make up half of all brain cells in humans, but years of research have not been able to pin a distinct functional profile on them. A single astrocyte can ensheath thousands of synapses in cortex, yet no one really knows what this elaborate morphological association does exactly. In previous years, Barres’s group found that cultured neurons can only form synapses when astrocytes are present, and they have begun isolating proteins astrocytes secrete to orchestrate this process (see Stevens et al., 2007; Allen and Barres, 2005; Christopherson et al., 2005; Ullian et al., 2004).
In the past, much of what is known about these cells has come from a classical astrocyte preparation, where the scientist cultures a suspension of neonatal brain in a medium containing serum (which native astrocytes in brain never encounter because of the blood-brain barrier), and assumes that the neurons die and astrocytes will survive and divide. Some do, but the cells that grow in these artificial conditions don’t look like astrocytes in vivo and are likely somewhat different from astrocytes in the mammalian brain. In a culture dish, proxy “astrocytes” look more like fibroblasts.
To get a comprehensive look at the real McCoys, first author John Cahoy and colleagues developed methods to separate mouse forebrain into neurons, astrocytes, oligodendrocytes, and other cell types such as endothelial cells of the blood-brain barrier. The scientists combined sequences of FACS scanning of fluorescent astrocytes from transgenic mice with rounds of depletion with antibodies. In this way, the scientists purified these cell types to greater than 99 percent purity. They did this with forebrain from newborn mice and maturing mice up to the age of one month, by which time synaptogenesis has largely completed. Then they extracted RNA from the cell types and put it on gene chips carrying 20,000 genes to catalog which cell type, at which developmental stage, expressed which complement of genes. (Validation was done by in-situ hybridization of selected genes.) That’s when the work of extracting biological information began. The large task of pathway analysis is not nearly completed, Barres said at a recent conference in Bar Harbor, Maine, but here are some initial findings about astrocytes that the data have yielded so far:
Dendrograms (i.e., diagrams that show “family trees” of cell types by their expression profiles) show that astrocytes are as different from oligodendrocytes as they are from neurons. “You cannot lump non-neuronal cells into one common class of glia. Glia are not related to each other,” Barres said.
“Heat Maps” of gene expression by cell type show that, besides some expected clusters of shared genes, there are some 2,000+ genes that are highly enriched in each of those cell types. The paper lists the top 40 cell type-specific genes expressed by astrocytes, oligodendrocytes, and neurons; the others exist in a huge Excel spreadsheet at this point, Barres said.
Of large numbers of genes that are highly and specifically expressed in astrocytes, many code for receptors, but nothing at all is known about their function. “It is humbling to look at these long lists and realize how much we still do not know about these cells,” Barres said.
Some of the specific genes serve as markers. For example, the gene aldehyde dehydrogenase 1 family, member L1 (Aldh1L1, aka FDH) makes for a better astrocyte marker than the commonly used GFAP, the authors argue. GFAP preferentially labels white matter astrocytes over grey matter astrocytes. Grey matter contains abundant astrocytes, as well, but because there was no good marker for them, some of the genes they express were mistakenly thought to be neuronal, Barres said. Aldh1L1 labels astrocyte populations and astrocyte processes more completely than does GFAP. In addition, the study generated markers for specific neural cell types, which the authors provided to Neuromab and other projects that make low-cost antibodies and mouse lines available to the community, such as NINDS GENSAT and NIH Neuromouse.
The Stanford scientists were surprised at the degree of metabolic specialization among the major cell types in the brain. Examples include synthesis or degradation of various amino acids including serine and glutamate, and synthesis of cholesterol primarily in astrocytes. By contrast, creatinine is made mostly in oligodendrocytes. The study confirmed previous knowledge that glycogen storage and ATP production via glycolysis and the Krebs cycle is cranked up in astrocytes compared to neurons. Secreted proteins such as ApoE and ApoJ/clusterin, Pdgf and Wnt7, among others, are also quite specific to astrocytes, as is the Notch signaling pathway that is known to this audience for getting in the way of therapeutic γ-secretase inhibition.
The astrocytes’ gene expression did not fit a readily discernable pattern. Generally, when scientists look at the 10 or so most expressed genes of a given cell type, they can infer the basic functions of the cell. For example, neurons express synaptic genes because they function in neurotransmission; oligodendrocytes make myelin basic protein because they myelinate axons. Astrocytes generate ApoE, ApoJ, Pla2g7, Sparc, among others—what do these do? “It highlights the continuing mystery about astrocyte function,” Barres said.
Of course, the scientists have made some preliminary conclusions. Perhaps the clearest is that astrocytes are professional phagocytes during development, and possibly also in neurodegeneration. Astrocytes eat in their own unique ways, though. They do not express the genes for the Fc receptor-related phagocytosis of antibody-coated debris, nor the Cd11/CD18b/integrin β2 pathway for complement-opsonized debris. Activated microglia employ those two pathways when they phagocytose. Instead, astrocytes express the ced-1/Draper—ced-7—ced-6 phagocytosis pathway that has been implicated in synapse pruning. The ApoE membrane receptor LRP1 is a homolog of ced-1 and enriched in astrocytes, raising the question of whether ApoE might function to coat debris for phagocytosis by astrocytes, Barres speculated. Astrocytes also express the ced-2—ced-5—ced-12 pathway of phagocytosis. They express phagocytosis receptors known to engulf opsonized debris, such as MertK and Axl, or the αvβ5 integrin pathway. The opsonin ligands for those receptors were also enriched in astrocytes.
Astrocytes plated onto brain slices have been reported to take up amyloid, as well (Wyss-Coray et al., 2003). A different speculation is that astrocytes might signal complement degradation of degenerating synapses (see ARF related news story). Microglia are thought to ingest amyloid after AD immunotherapy. The present paper does not include a transcriptome of microglia, partly because the lab’s interest is primarily in astrocytes and oligodendrocytes, and partly because microglia become activated during the isolation process and then express heat shock and other activation markers. Few specific markers for quiescent microglia are thus far available.—Gabrielle Strobel.
Cahoy JD, Emery B, Kaushal A, Foo LC, Zamanian JL, Christopherson KS, Xing Y, Lubischer JL, Krieg PA, Krupenko SA, Thompson WJ, Barres BA. A transcriptome database for astrocytes, neurons, and oligodendrocytes: a new resource for understanding brain development and function. J Neurosci. 2008 Jan 2;28(1):264-78. Abstract