Equal numbers of neuronal and nonneuronal cells make the human brain an isometrically scaled-up primate brain

It is repeatedly stated in the literature and in neuroscience textbooks that the human species is an unusually encephalized primate species, whose brain, five to seven times larger than expected for a mammal of its body size (Jerison,1973; Marino,1998), contains 100 billion neurons and about ten times more glial cells (Kandel et al.,2000; Ullian et al.,2001; Doetsch,2003; Nishiyama et al.,2005; Noctor et al.,2007). The supposedly unusual scaling of the human brain, however, derives from comparisons across orders (Jerison,1973) and, even when restricted to primates, regards only the brain–body relationship (Marino,1998) rather than addressing how its cellular composition compares with that expected from other primates.

Moreover, to our knowledge, the widespread numbers on the cellular composition of the human brain have never been supported by experimental studies. The high anisotropy and large size of the human brain hinder stereological determination of cell numbers and their distribution in the brain as a whole. Estimates of the cellular composition of the human brain are available only for some structures, such as the cerebral cortex (von Economo and Koskinas,1925; Shariff,1953; Pakkenberg,1966; Pakkenberg and Gundersen,1997; Pelvig et al.,2008), cerebellum (Lange,1975; Andersen et al.,1992), and some subcortical nuclei (Pakkenberg and Gundersen,1988). Such studies have estimated the number of cells in the human cerebral cortex as 3, 7, 14, 19–23, or 21–26 billion neurons and, very recently, 28–39 billion glial cells (Pelvig et al.,2008), and the number of cells in the human cerebellum has been estimated as 70 or 101 billion neurons (Lange,1975; Andersen et al.,1992) and fewer than 4 billion glial cells (Andersen et al.,1992). From such studies, the total number of neurons in the human brain might be inferred to fall anywhere between about 75 and 125 billion plus an undetermined number of neurons in the brainstem, diencephalon, and basal ganglia that may or may not be comparatively small.

Additionally, no evidence is found to support the common quote of ten times more glial cells than neurons in the human brain. The glia:neuron ratio in subcortical nuclei can be as high as 17:1 in the thalamus (Pakkenberg and Gundersen,1988), but, given the relatively small combined number of glial cells reported for the cerebral (Pelvig et al.,2008) and cerebellar (Andersen et al.,1992) cortices, the only possible explanation for the quote of ten times more glial than neuronal cells in the entire human brain would be the presence of nearly one trillion glial cells in the remaining structures.

We have recently determined the cellular scaling rules that apply to the brain of a number of rodent and primate species and found that, whereas the rodent brain increases in mass faster than it gains neurons (defined as NeuN-positive cells) across species, suggesting that the average neuronal cell size increases in larger rodent brains (Herculano-Houzel et al.,2006), the primate brain increases in mass linearly with increases in its number of neurons across species, suggesting that the average neuronal cell size does not increase significantly with brain size (Herculano-Houzel et al.,2007). The power laws relating body mass, brain mass, and number of neurons for rodent and primate species allowed us to predict that, if built according to the cellular scaling rules that apply to rodents, a brain of 100 billion neurons should weigh over 45 kg and belong to a body of 109 tons. In contrast, if built according to the scaling rules that apply to primates, this brain of 100 billion neurons should weigh 1.45 kg and belong to a body of 73 kg, values that approach those observed in humans, suggesting that the human brain is indeed constructed according to the same rules that apply to other primates.

We thus set out to determine the total cellular composition of the human brain with the aid of the same method (the isotropic fractionator; Herculano-Houzel and Lent,2005) and relying on the same criterion of NeuN labeling to identify “neurons” and “nonneuronal cells” in order to evaluate how its composition compares with the expected composition of a primate brain of its size. While supporting several independent stereological estimates, our results challenge the values so often cited in the literature and suggest that, with regard to brain cellular composition, humans are just scaled-up, large primates.

MATERIALS AND METHODS

Isotropic fractionator

The isotropic fractionator method has been described elsewhere (Herculano-Houzel and Lent,2005). Briefly, it consists of a chemomechanical dissociation of fixed biological tissue in a saline detergent solution (1% Triton X-100, 40 mM sodium citrate) using 40–200-ml glass tissue grinders, followed by intense agitation of the suspension containing all nuclei in the original structure, in order to achieve isotropy. After adding the fluorescent DNA marker 4′-6-diamino-2-phenylindole dihydrochloride (DAPI) to the suspension, the density of nuclei is quantified by use of a hemocytometer under a fluorescence microscope (Fig. 1A,C). The total number of nuclei is calculated by multiplying the density of nuclei by the total suspension volume and heretofore is referred to as “total number of cells” in each structure.

RESULTS

We find that the male human brain, aged ∼50 years (n = 3) or 70 years (n = 1) and weighing 1,508.91 ± 299.14 g, contains on average 170.68 ± 13.86 billion cells. Among these, 85.08 ± 6.92 billion cells are located in the cerebellum, 77.18 ± 7.72 billion cells are in the cerebral cortex (including both gray and white matter), and 8.42 ± 1.50 billion cells are found in the remaining regions (RoB; Fig. 2). Because no significant differences were found in mass or in neuronal, nonneuronal, and total cell numbers between right and left hemispheres (t-test, P values typically well above 0.1), all numbers given refer to the combined hemispheres.

Overall, the nonneuronal/neuronal ratio in the whole human brain is close to 1 (Fig. 2), insofar as half of the cells in the human brain, or 86.06 ± 8.12 billion, are neurons (range 78.82–95.40 billion neurons). The fractional distribution of neurons in the human brain does not correspond to the fractional distribution of mass among brain structures (Figs. 2, 3). Although 82% of brain mass consists of cerebral cortex (including subcortical white matter) and 42% consists of cerebral cortical gray matter alone, the 16.34 ± 2.17 billion neurons found in this structure represent only 19% of all brain neurons. In contrast, the cerebellum, which represents only 10% of total brain mass, contains 69.03 ± 6.65 billion neurons, or 80% of all neurons in the human brain. Fewer than 1% of all brain neurons are located in the RoB, comprising basal ganglia, diencephalon, and brainstem (Figs. 2, 3), although this percentage is necessarily underestimated as a result of the known lack of NeuN staining in at least some structures in the brainstem (see Materials and Methods).

The other half of all human brain cells, or 84.61 ± 9.83 billion, are nonneuronal cells, yielding a nonneuronal/neuronal ratio of 0.99 for the human brain as a whole (Fig. 2). Nonneuronal cells outnumber neuronal cells in the RoB, with a nonneuronal/neuronal ratio of 11.35 (Fig. 2). In the gray matter of the cerebral cortex, the nonneuronal/neuronal ratio is 1.48. In contrast, in the cerebellum, this ratio is only 0.23 (Fig. 2).

In contrast to the distribution of neurons, the fractional distribution of nonneuronal cells in the brain resembles more closely the fractional distribution of mass in the structures (Fig. 3). The cerebral cortex, including the subcortical white matter, holds 72% (or 60.84 ± 7.02 billion) of all nonneuronal cells in the brain, whereas the cerebellum has 19% (or 16.04 ± 2.17 billion) and the RoB holds 9% (or 7.73 ± 1.45 billion) of all the nonneuronal cells in the brain.

The separate analysis of white matter (WM) and gray matter (GM) of the cerebral cortex shows that the former contains 69.6%, or 19.88 ± 2.83 billion, of the nonneuronal cells of the cerebral cortex of one hemisphere (Fig. 4). The ratio between nonneuronal and neuronal cells is 1.48 ± 0.42 for the GM (Fig. 4) and 3.76 ± 0.55 for the combined GM and WM (Fig. 2).

The numbers of neuronal and nonneuronal cells found in the human brain fall very close to the values expected for a primate brain of human dimensions built according to the linear, isometric cellular scaling rules found to apply to primate brains (Herculano-Houzel et al.,2007; Fig. 5a–c). According to these rules, a primate brain weighing 1,508 g would be expected to have 94 billion neurons, with 22 billion neurons in the cerebral cortex, 78 billion neurons in the cerebellum, and 0.6 billion neurons in the RoB (Table 1). The observed numbers of neurons are actually slightly smaller than expected in the human cortex and cerebellum and larger in the RoB (Table 1). The percentage deviations in the observed values from the numbers of neuronal and nonneuronal cells expected in each structure of the human brain given its mass fall within the same range as the values found for each of the species from which the cellular scaling rules were originally obtained (Herculano-Houzel et al.,2007; Fig. 6). The neuronal and nonneuronal cell densities found in the human brain also fall in the range observed in nonhuman primates in that study (Herculano-Houzel et al.,2007; Fig. 5d,e).

DISCUSSION

It is often stated in the literature that glial cells outnumber neurons in the human nervous system by a factor of 10. Our finding that the human brain has an approximately 1:1 ratio of nonneuronal:neuronal cells implies a necessarily smaller glia/neuron ratio, if endothelial and other mesenchymal cells were removed from the nonneuronal pool. Although this ratio of approximately 1:1 for the whole brain is important counterevidence to the common overestimation of numbers of glial cells in the brain, it conceals the fact that specific structures of the human brain can have maximal glia/neuron ratios (if all nonneuronal cells were glial cells) as small as 0.23, such as the cerebellum, and as large as 11.35, such as the RoB. The maximal glia/neuron ratio of 1.48 that we observed in the total GM of the cerebral cortex is close to the ratio of 1.65 observed by Sherwood et al. (2006) in layer II/III of human prefrontal area 9L and similar to the ratios between 1.2 and 1.6 that Pelvig et al. (2008) encountered in the whole human neocortical GM. These similarities corroborate our findings.

According to the common view in the literature, the glia:neuron ratio increases with brain size (Reichenbach,1989), leading to a predominance of glial cells in large brains (Nedergaard et al.,2003) that would be compatible with a 10:1 ratio in humans. We have shown that, although the average nonneuronal cell size is relatively invariant across brain structures and species, an increasing predominance of glial cells with brain size is indeed found in rodents (Herculano-Houzel et al.,2006), in which average neuronal size increases together with neuronal number, but not in the primates examined so far (Herculano-Houzel et al.,2007), in which average neuronal size, as with average nonneuronal cell size, is estimated to remain relatively stable as numbers of neurons increase. All primates we have analyzed until now (Herculano-Houzel et al.,2007) exhibit ratios of nonneuronal/neuronal cells that are similar to the approximately 1:1 ratio found in the human brain as a whole.

The glia/neuron ratio has been considered to be of great relevance because of the multiple functional relationships recently demonstrated between these cell types (Nedergaard et al.,2003; Shaham,2005) and has been hypothesized to reflect neuronal activity (Reichenbach,1989). Alternatively, the glia:neuron ratio might simply follow the ratio between average neuronal and average glial cell mass. We have proposed (Herculano-Houzel et al.,2006) that this occurs as the nearly all-neuronal parenchyma is invaded in early postnatal development by glial progenitors that divide until the newly formed glial cells, of a relatively constant average size, reach confluence (Zhang and Miller,1996). In this scenario, a neuronal parenchyma of a given volume built of a large number of small neurons is invaded by progenitors that will give rise to the same number of glial cells as a parenchyma of the same volume built of a smaller number of larger neurons, but the latter, with larger neurons, will have a much larger glia:neuron ratio than the former. It is noteworthy that the low glia:neuron ratio in the cerebellum is due not to a conspicuous lack of glial cells but rather to a very large number of very small neurons, because its nonneuronal cell density is even somewhat larger than that in other structures. The relative constancy of glial cell densities observed across rodents and primates, humans included, may therefore be more physiologically meaningful in terms of their role in the metabolic maintenance and functional support of brain tissue than their numeric ratio to neurons.

Our estimate of an average total of 86 billion neurons in the human brain is compatible with previous stereological determinations for individual structures such as the cerebral cortex (von Economo and Koskinas,1925; Pakkenberg and Gundersen,1997; Pelvig et al.,2008) and the cerebellum (Lange,1975; Andersen et al.,1992). Exact numbers are probably highly variable among humans, particularly given the variation of over 50% in the number of cortical neurons among individuals of the same sex described recently in the literature (Pelvig et al.,2008). Although the numbers of neurons that we report for the RoB (which includes NeuN-negative neurons in the inferior olive and possibly other structures as well) are necessarily underestimated, the number of NeuN-negative neurons included in the “nonneuronal” population of the RoB is likely to be negligible compared with the 85 billion neurons found in the ensemble of cerebral cortex and cerebellum (because the RoB is found to contain a total of only about 8 of the 170 billion cells in the human brain), or to the 690 million neurons in the RoB. Moreover, even in the improbable scenario in which all cells in the RoB, composed of massive fiber tracts, were NeuN-negative neurons, they would still amount to only 5% of all brain cells.

More important than the exact number of neurons in the human brain, however, are the implications of how this number compares with that expected for a primate brain of human proportions. We have shown before that a brain with about 100 billion neurons built according to the cellular rules that apply to scaling rodent brains would weigh 45 kg (Herculano-Houzel et al.,2007), well above the largest known whale brain. Humans of 70 kg of body mass built according to these rules would be expected to have a brain of only 145 g, instead of 1,500 g. That is, humans do indeed have a brain that is about ten times larger and holds seven times more neurons than predicted for a nonprimate mammal of its body size.

Remarkably, however, here we find for the first time that the human brain conforms to the scaling rules observed for a given group of mammals: six other primate species (Herculano-Houzel et al.,2007), the only ones so far whose total brain cellular composition is known. We show that humans, despite the large relative size of the cerebral cortex, hold only 19% of all brain neurons in this structure, as do other primates and rodents of different brain sizes, and demonstrate that the mass and cellular composition of the human brain deviate from the values expected for a primate of 75 kg by only 10%. Given that the cellular scaling rules that apply to primate brains are linear, the conformity of the human brain to these rules strongly indicates that the human brain is a linearly scaled-up primate brain in its cellular composition. The isometric scaling of the human brain cellular composition relative to other primates is in line with other observations that have established that the cerebellum (Frahm et al.,1982) and frontal cortex (Semendeferi et al.,2002) of the human brain have the same relative size as in great apes, even though the distribution of mass within the WM and individual cortical areas in humans may differ from that in other primates (Semendeferi et al.,2001; Rilling and Seligman,2002; Schoenemann et al.,2005).

Our notion that the human brain is a linearly scaled-up primate brain in its cellular composition is in clear opposition to the traditional view that the human brain is 7.0 times larger than expected for a mammal and 3.4 times larger than expected for an anthropoid primate of its body mass (Marino,1998). However, such large encephalization is found only when body-brain allometric rules that apply to nonprimates are used, as stated above, or when great apes are included in the calculation of expected brain size for a primate of a given body size. Our finding thus suggests that the rules that apply to scaling brains are more conserved than those that apply to scaling the body and raises the intriguing possibility that, rather than humans having a larger brain than expected, it is the great apes such as orangutans and, more notably, gorillas that have bodies that are much larger than expected for primates of their brain size. Indeed, the inclusion of great apes (Marino,1998) in the primate species (Herculano-Houzel et al.,2007) that we compare to humans would increase the body size expected of our species, with a brain of 1,509 g, from 77 kg to 216 kg, and decrease the expected brain size for a body of 70 kg from 1,247 g to 557 g. One piece of evidence in support of the possibility that gorillas and orangutans, rather that humans, are outlier species in terms of body size is that, whereas in most primate species, humans included, the brain represents about 2% of total body mass (Marino,1998), the brains of gorillas and orangutans, at about 500 g (Semendeferi and Damasio,2000), represent at most 1% of a body of 50 kg, and only 0.5% or less in typical male gorillas of 100 kg or more. The adaptive value of an enlarged body size in these species can be appreciated from the status of social dominance that comes with the large investment of time and energy necessary to develop large bodies in alpha-male gorillas and orangutans (Leigh,1995).

We are currently investigating whether the brains of gorillas and orangutans also conform to the cellular scaling rules found to apply to other primates, including humans. Such conformity would substantiate the intriguing possibility that, rather than humans having too large a brain for their bodies, gorillas have too large a body for their brains, although both species have brains built according to the same rules that apply to other primates. Body size (Jerison,1973), after all, may not be a relevant parameter when it comes to cognition (Roth and Dicke,2005). In light of the recent finding that absolute brain size is the parameter that best correlates with cognitive abilities (Deaner et al.,2007), our cognitive advantage over other primates might be simply a consequence of having the largest brain, built with an isometrically enlarged number of neurons compared with other smaller-brained primates, regardless of body size.