How many cells does the brain contain, and what is the proportion of each cell type? We know there are neurons, astrocytes, oligodendrocytes, microglia, endothelia, columnar epithelia, tanycytes, pericytes, cells of the choroid plexus, stem cells and a variety of more minor types. You would think that this was a question which had been settled decades ago with good solid data and that every neuroscientist would have a firm grasp of the basic findings. But if you thought that you would be wrong. There is still a widespread belief that glia vastly outnumber neurons in the mammalian brain. For example this is part of the introduction to a popular science book- “Back in the 1960s, it was discovered that glial cells are 90 percent of the brain. Neurons make up 10 percent” (Koob, 2009). And it is not just popular science books. To quote a highly influential text book “Glial cells far outnumber neurons- there are between 10 and 50 times more glia than neurons in the central nervous system of vertebrates” (Kandel et al. 2000). For a more recent example from the normally reasonably accurate journal Nature “The proportion of glia seems to be correlated with an animal’s size: the tiny nematode worm has only a few glia; some 25% of the fruitfly brain consists of glia; the mouse brain has roughly 65% of these cells; the human brain has about 90%; and the elephant brain consists of some 97% glia (Allen and Barres 2009). I contacted the authors of this article to try to find out where this to me outrageous claim originated but they could not find the original paper. It should be mentioned that not every text book has promulgated the extreme version of this idea. Some authors are more cautious “Glial are more numerous than neurons in the brain, outnumbering them by perhaps 3 to 1” (Purves et al, 2011). But still the claim in this case is that there is an awful lot more glia. I go to seminars and meetings at which people come up with various large numbers for the ratio of glia to neurons. For one recent example I was at the Society for Neuroscience meeting in Washington DC, (November 2014), and I went to an excellent presidential lecture by an extremely talented young scientist who also stated that glia far outnumber neurons in the mammalian brain. Clearly the idea that glia are enormously numerous has considerable traction. It is surely extraordinarily peculiar that the signaling cells of the brain are supposedly only a tiny minority of the whole. Anyone would be surprised to find, for example, that less than 5% of muscle cells were able to contract, or that less than 5% of liver cells were able to detoxify the blood. The purpose of the present article is to hopefully highlight the solid data which indicates what is the correct number of brain cells.
To start with the Koob claim that back in the 60s it was discovered that glia vastly outnumber neurons, where did that come from? It is noticeable that this book and in fact none of the other claims listed above cite any original literature. I assume that the original meme can only have come from histological studies, which were, in the 60s, in their infancy. At that time there was, believe it or not, no really reliable method of identifying a neuron in a section. What did exist was Golgi silver stain, Nissl stain and Bodian and other neurofibrillar stains. The Golgi stain allows you to identify cell types by their morphology but is by nature very non quantitative, for still unknown reasons staining only a subset of cells in a section. Both the Nissl and Bodian type stains will identify neurons but only the larger neurons. So you can see projection neurons in the cortex and spinal cord with these two stains, but these are unusually large and relatively non-abundant cells. The smaller cortical and especially cerebellar interneurons typically do not stain well or at all with these two stains and so could only be visualized in the 60s using nuclear stains. I believe that this is the reason why the idea that glia are so numerous first originated- the casual observer may have thought that all of the nuclei not associated with obvious Nissl substance or neurofibrils were glia. However, some researchers realized that the morphology of neuronal nuclei was distinct from that of other kinds of cells, the neuronal nuclei being larger and less dense. This did allow some reasonably accurate counting of neurons but only in very small well defined regions. These papers generally did not address the ratio of neurons to glia but certainly did not suggest that there was a huge over abundance of glia. However for whatever reason this work did not become part of the neuroscience canon.
As science progressed to the development of immunocytochemistry in 70s and onwards it became possible to specifically visualize different types of glia and neurons using antibodies to proteins such as glial fibrillary acidic protein or GFAP, neurofilaments, neuron specific enolase, Pgp9.5/UCHL1, Iba-1/Aif-1, CNPase and NeuN. Since these proteins are expressed only in certain types of brain cell it now became possible to identify the different types of cell and also of course a suitably motivated and competent person could now count each type of cell. However nobody seems to have systematically addressed this issue so again the meme that glia greatly outnumber neurons doctrine lived on.
Above: Rat hippocampus stained with antibody to NeuN (green), myelin basic protein (red) and DNA (blue). The neuronal cell bodies and nuclei in the layers of the hippocampus stain strongly with NeuN antibody, while non-neuronal cells are revealed by the DNA stain and are negative for NeuN. Antibodies and image courtesy of EnCor Biotechnology Inc.
An important and I think somewhat overlooked series of papers have been very influential to me at least, and deserve to be much better know. These are from the laboratories of Roberto Lent and Suzana Herculano-Houzel in Rio de Janiero. The first of these is entitled “Isotropic fractionator: a simple, rapid method for the quantification of total cell and neuron numbers in the brain” (Herculano-Houzel and Lent 2005). This very short paper describes how to make a nuclear preparation from a fixed brain sample and count the nuclei of brain cells after staining with DAPI. The cunning part of this paper is to counterstain the isolated nuclei with antibody to NeuN, a protein found in almost but not quite all neuronal nuclei. You can therefore count two classes of nuclei, those which are DAPI positive and NeuN positive, which must have come from neurons, and those which are just DAPI positive, most of which must come from one of the several classes of non neuronal cells. The results were surprising if you believed the supernumerous glia theory, there were actually more NeuN positive nuclei in a rat brain than NeuN negative nuclei. This means that even the combination of the astrocytes, oligodendrocytes, microglia, endothelia, pericytes, lymphocytes, macrophages, ependyma, tanycytes, pia cells and so on was still the minority and consequently that no one class of non-neuronal cell is likely to represent more than about 10% of the total brain cell count. The actual data is that the typical adult rat brain contains about 330 million nuclei, and so 330 million nucleated cells, of which about 60% are NeuN positive. The ratio of NeuN positive to NeuN negative was variable in different brain regions, with about 40% in cortex, 53% in olfactory bulb and a perhaps surprisingly high 83% in cerebellum. The reason for the very high neuronal content in the cerebellum is comprehensible on reflection as being due to the very large number of cerebellar granule cells. These small and densely packed interneurons are the major cell type of the cerebellar granular layer, and this cell type is very likely the single most abundant kind of brain cell in the rat. The basic fractionator method has been tried on several other species with rather comparable results. The same researchers went on to show that the average human brain contains about 86 billion NeuN positive cells and about 85 billion NeuN negative cells, so humans also have a preponderance of neuronal cells, albeit by a smaller margin than in rodents (Azevedo et al., 2009). The elephant brain, far from having 97% glial cells according to the one claim mentioned above, also has a preponderance of neurons although the number of cortical neurons compared to glia is lower than in humans. However the number of neurons in the cerebellum compensates for this, so that about 97% of the neurons in the elephant brain are in this region, again mostly granule cells. If you want to see a TED talk by Hercolano-Houzel outlining some of this interesting data, along with many other links, press here.
If you think about the construction of the nervous system these kind of findings become quite comprehensible. For example cortical astrocytes and other glia show “tiling” and “self avoidance”. These interesting properties, likely mediated by specific cell adhesion molecules, result from two related phenomena. Self avoidance means that the processes of each individual cell avoids contact with oneanother so they naturally and efficiently fill a three dimensional region of brain tissue. Tiling results from the processes of each astrocytes being inhibited from extending on contact with the processes of neighboring astrocytes. As a result of these two mechanisms each astrocyte occupies a small three dimensional space and only one astrocyte has processes in any one brain region. In cortical regions this block of tissue may contain numerous neurons, I have heard claims of up to 150 per astrocyte. Clearly there are also regions where there are astrocytes and no neurons, but a lot of brain is cortex and astrocytes are very large cells, so the absolute number of astrocytes is necessarily not enormous. Similar arguments can be made about microglia which are likely to be more numerous than astrocytes as they are relatively small but also “tiled” cell and “self avoiding” cells, limiting their absolute number. Oligodendrocytes are found only in white matter regions, and each oligodendrocyte makes myelin around numerous axons, so again there is no particular reason to think that these would be enormously numerous. The endothelia, in my opinion, are likely to be more numerous than any of the glial cells. After all the brain is very heavily vascularized in all regions and the vasculature is made up of densely packed and relatively small nucleated endothelial cells. What is needed to answer these kinds of quantitative questions definitively are robust markers similar to NeuN which are present only in subclasses of non-neuronal cell nuclei. Unfortunately such markers have not been described to date, though there is some possibility that such may eventually be found. One possible so far unexplored avenue might be to use tissues from transgenic mice in which Cre recombinase is expressed in specific subclasses of glia. The form of Cre used in these transgenic mice is engineered to be nuclear targeted. As a result the Isotropic fractionator method should allow the identification and counting of Cre positive nuclei using a suitable Cre antibody. As far as I know nobody has done this. Similarly GFP expression under cell type specific promoters might be used in the same way, especially if the GFP includes a nuclear localization sequence.
So why has the dogma proved to be so durable? One reason is likely the practice of text books to repeat material found in previous editions of the same book or from other peoples earlier text books without bothering to look for the original reference. The other is I think that the old glia doctrine is one of those odd thoughts that seems counterintuitive at first sight, a fun filled fact like that New York sewers are full of alligators. Since neurons do the really important work in the brain should they not be the most abundant? So suggesting that some other cell type is far more abundant becomes one of those little known fun filled facts that people like to spread about. In this case counterintuitive is also counterfactual and the original intuitive thought, that there should be more neurons than other cell types, is actually correct.