Larger bodies are generally believed to require a larger number of sensory and motor neurons to operate them (Jerison, 1973; Fox and Wilczynski, 1986). In particular, the larger the muscle mass of an individual or species, the larger one expects the motor neuron pool that controls it to be (Krompecher and Lipák, 1966). However, what is the numerical relationship between numbers of neurons and muscle mass across species (and individuals), and how is it determined? In this article, we examine the role of the size of the target muscle group in determining the final number of motor neurons.
During embryonic development, a specific number of motor neurons are generated. After the axons grow out to make contact with muscles, a percentage of the motor neurons die (Hughes, 1961; Sohal, 1976; Lamb, 1977; Lance-Jones, 1982; Ashwell and Watson, 1983). Some authors have claimed that the earliest connections were of a random nature, and that the mature pattern of connection is not established until after the period of cell death. On this basis, these authors concluded that the role of cell death is to weed out inappropriately connected motor neurons (McGrath and Bennett, 1979; Pettigrew et al., 1979). However, a number of studies have shown that properly patterned connections are established before the period of cell death (Landmesser and Morris, 1975; Lamb, 1976; Hollyday et al., 1977; Ashwell and Watson, 1983).
An alternative explanation for the developmental death of motor neurons is that it is related simply to limitations in the amount of muscle available for innervation. If this was true, it would represent a very economical self-organizing mechanism to ensure that numbers of motor neurons are appropriate for the size of the animal. A number of experimental studies have examined this possibility, but found only a weak and nonlinear relationship between changes in motor neuron number in response to changes in target muscle mass during development (Beucker, 1945; Pollack, 1969; Hollyday and Hamburger, 1976; Hollyday and Mendel, 1976; Lamb, 1979). However, a chick-quail grafting study by Tanaka and Landmesser (1986) provided strong evidence for a direct relationship between final motor neuron number and the number of myotube clusters at the onset of motor neuron death. They found that the correlation between myotube cluster numbers and number of surviving motor neurons was 0.996. This very strong correlation indicates that either myotube number itself or some factor closely related to myotube number might determine the final motor neuron number. Here we address the question of quantitative matching by studying the relationship between motor neuron number and muscle mass (by proxy) in the adult facial nucleus of 22 different marsupial species, ranging in size from 10 to 50,000 g.
The consistent organization of the mammalian facial nucleus makes it a favorable site for the investigation of motor neuron development. The nucleus typically has six distinct subnuclei (Fig. 1) (Papez, 1927; Courville, 1966; Dom et al., 1973; Provis, 1977; Ashwell, 1982; Watson et al., 1982; Welt and Abbs, 1990; Sherwood, 2005). The musculotopic representation in the mammalian facial nucleus is very conserved; Sherwood (2005) summarizes the results of 22 retrograde tracing studies in marsupial and placental mammals by concluding that “a basic pattern of muscle representation exists in the VII that is common to all mammals.” Sherwood notes that the peri-oral muscles are represented in the lateral part of the nucleus, the posterior auricular and neck muscles are represented in the medial part, and the orbicularis oculi, frontalis, and anterior auricular muscles are represented in the intermediate part of the nucleus. The neurons supplying each muscle group form a longitudinal column that extends for almost the entire length of the nucleus, and this is responsible for the consistent pattern of representation of facial muscles in individual coronal sections (Welt and Abbs, 1990; Sherwood, 2005). The deep muscles of the second branchial arch, the stylohyoid and posterior belly of the digastric, are represented in an outlying group situated dorsal to the main nucleus (Matsuda et al., 1979; Watson et al., 1982). The study of this group is beyond the scope of the present study.
We hypothesize that muscle mass is the major determinant of adult motor neuron number. We have tested this in a series of marsupials with a very wide range of body mass. In the absence of specific data, we have used body mass as a proxy for muscle mass, but we believe that the substitution is justified. We also hypothesize that motor specialization (in this case whisking behavior) is not associated with local increases in motor neuron number.
MATERIALS AND METHODS
We counted the number of facial motor neurons in stained histological sections of single brains in each of 22 different marsupials. In all but one case (Monodelphis domestica; see below), the sections belong to the definitive collection of marsupial brains assembled by Dr John Johnson of Michigan State University, and now in the care of the National Museum of Health and Medicine at the Armed Forces Institute of Pathology, Washington, DC. The details of the preparation of the histological sections are to be found on a web site devoted to the presentation of the Michigan State and University of Wisconsin collections of mammalian brains (<brainmuseum.org>). The animals included in this data set include representatives from nine of the 20 extant families of marsupials (Table 1).
Table 1. The classification of marsupial species represented in this study
Family Didelphidae: Didelphis virginiana, Marmosa murina, Monodelphis domestica
Family Caluromydae: Caluromys lanatus
Family Caenolestidae: Lestoros inca
Family Dasyuridae: Dasyurus quoll, Sarcophilus harrisii, Smithopsis murina, Antechinus flavipes
Family Peramelidae: Perameles nasuta, Isoodon obesulus
Family Vombatidae: Vombatus mitchelli
Family Phalangeridae: Trichosurus vulpecula
Family Petauridae: Petaurus breviceps, Petaurus norfolcensis, Schoinobates volans
Family Pseudocheiridae: Psuedocheirus peregrinus
Family Potoridae: Potorous tridactylis
Family Macropodidae: Macropus rufus, Macropus rufogrisius Macropus fuliginosis, Thylogale billardieri
The brains we examined from the Michigan State collection were embedded in celloidin, cut at either 25 or 30 μm thickness in the coronal plane, and stained with Nissl and myelin stains. The histological quality of the stained sections is very high. In the case of Monodelphis domestica, we examined high-resolution images of Nissl stained sections presented on the website <brainmaps.org>. These sections were cut at 25 μm in the coronal plane, and stained with Nissl stain.
We determined the total number of motor neurons in the main body of the facial nucleus of one side of the hindbrain in a sample of Nissl stained sections from each brain. We counted motor neurons in either 1 in 4 or 1 in 8 sections through the nucleus. The counts were based on facial neuron profiles that contained a distinct nucleolus. We did not include the neurons in the outlying stylohyoid subnucleus in our counts.
To estimate the size of the muscle field available to the facial motor neurons, we were forced to use an indirect method of estimation. As we did not have data on the mass of the facial muscles, we took the weight of the animal as proxy for the total muscle mass in that animal. In lean mammals total muscle mass accounts for about 40% of body weight (Alexander, 1964; Hopwood et al., 1976; Shepherd, 1991; Lee et al., 2000 ). In addition, bone mass, which is about 15% of body mass, is highly correlated with muscle mass (Doyle et al., 1970), so the sum of muscle mass and bone mass makes a dominant contribution to body weight in lean animals.
Power laws describing how the number of motor neurons in the facial nucleus relate to body or brain mass were calculated using least squares regression to log-transformed data, and the scaling exponents are reported where P < 0.01. Reported r2 values refer to the adjusted r2. Spearman correlations were also calculated using the same software, and the adjusted r values are reported. Residuals calculated from the regression of number of facial motor neurons onto brain mass were compared between whisking and nonwhisking marsupials using a Mann-Whitney test after normalizing them to brain mass. All statistical analyses were performed using JMP 9.0 (SAS, NC). Controlling for phylogenetic relatedness in the dataset was not necessary because of the very high r2 values observed in the power law fits (see below). In the face of uncertainties in the phylogenetic relationships among the species studied, we, therefore, prefer to report the uncorrected scaling exponents, which apply directly to our dataset.
The subnucleation in the marsupial facial nucleus is consistently present in marsupials, but is more distinct in some species than others. Figure 1 compares a coronal section from an animal with distinct subnuclei (the brush-tailed possum Trichosurus vulpecula of Australia) with sections from two marsupials with less distinct subnuclei (the mouse opossum Monodelphis domestica of South America and the Australian sugar glider Petaurus breviceps). In the brush-tailed possum facial nucleus, the six major subnuclei (lateral, dorsolateral, dorsal intermediate, ventral intermediate, dorsal medial, and ventral medial) can be clearly seen (Fig. 1). In Monodelphis and the sugar glider Petaurus some subnuclear boundaries are less distinct.
In marsupials and placentals, the two medial subnuclei supply the posterior auricular muscles and the two lateral subnuclei supply the nasolabialis muscle, which moves the mystacial vibrissae (Provis, 1977; Ashwell, 1982; Watson et al., 1982). A seventh subnucleus, which supplies the stylohyoid and the posterior belly of the digastric (Matsuda et al., 1979; Watson et al., 1982), lies at varying distances dorsal to the main nucleus. We did not include this outlying subnucleus in our neuronal counts.
The number of facial motor neurons in each marsupial species is presented in Table 2. Across the 22 species, ranging in body mass from 10 to 50,000 g (a 5,000-fold variation), brain mass varies 140-fold from 0.44 to 61.4 g, scaling as a power function of body mass with exponent 0.590 (P < 0.0001, r2 = 0.937; 95% confidence interval 0.521–0.660; Fig. 2a). In contrast, the number of motor neurons in the facial nucleus varies only 4.8-fold, from 2,530 to 12,160, also in a very strong correlation with body mass (Spearman, 0.919; P < 0.0001). This indicates that most of the variation in motor neuron number accompanies differences in body weight.
Table 2. The body weight, brain weight, and number of facial motor neurons in one side of the brain in 22 species of Australian and American marsupials
Body wt (g)
Brain wt (g)
Neurons in 7N
The sex and maturity of each specimen in indicated. The data are derived from the Michigan State University Collection (MSU) or from the BrainMaps website (<brainmaps.org>).
Brazilian short-tailed opossum
? not mature
Western gray kangaroo
In contrast to brain mass, variations in the number of motor neurons in the facial nucleus can be described as a power function of body mass with a much smaller exponent of 0.184 (P < 0.0001; r2 = 0.878; 95% confidence interval, 0.153–0.216; Fig. 2b). With such an exponent, doubling in body mass from one hypothetical species to another is accompanied by a modest increase of only 14% in motor neuron number, while halving the body mass results in a decrease of only 12% in motor neuron number.
The number of neurons in the facial nucleus also scales as a power function of brain mass with a small exponent of 0.307 (P < 0.0001; r2 = 0.904; 95% confidence interval, 0.263–0.351; Fig. 2c). Although the relationship between brain mass and number of brain neurons could not be determined at this time, the small exponent suggests that brain size increases across marsupial species occur with the addition of smaller numbers of neurons to the facial nucleus than to the rest of the brain.
To examine whether whisking behavior is associated with an increased number of motor neurons in the facial nucleus, we calculated the normalized residuals of numbers of facial motor neurons regressed onto brain mass for species known to exhibit whisking behavior (Monodelphis domestica, Trichosurus vulpecula) or not (Perameles nasuta, Petaurus breviceps, Petaurus norfolcensis, Potorous tridactylus Macropus rufus, and Macropus rufogriseus). We find no significant difference in the normalized residuals between whisking (−0.230 ± 0.423) and nonwhisking marsupials (−0.022 ± 0.155; P = 0.5050, Mann-Whitney test). Because these values were normalized to brain mass, differences due to brain mass across species can be discarded. The lack of difference in normalized residuals between whisking and nonwhisking marsupials thus suggests that behavioral specialization occurs in the absence of relative increases in the numbers of facial motor neurons.
In this article, we have provided evidence for a nonlinear scaling of numbers of motor neurons as a function of body mass—used here as a proxy for muscle mass. In addition, our results argue against the contention that increased motor sophistication is based on a relative increase in the number of motor neurons for a given muscle mass.
A Non-Linear Relationship Between Motor Neuron Number and Muscle Mass
Many previous studies of the relationship between motor neuron number and muscle mass were apparently based on an expectation that the relationship would be linear. In general, these studies manipulated the available muscle mass in order to either double or halve it before the period of formation of synapses between motor neuron and muscle cells. While these studies found that doubling the target muscle field in amphibians and birds resulted in an increase in the number of motor neurons that survive, many of the authors expressed surprise that the increase in motor neuron survival was so small—only 15–20% (Beucker, 1945; Pollack, 1969; Hollyday and Hamburger, 1976; Hollyday and Mendel, 1976; Lamb, 1979). Similarly, an effective reduction of the muscle field by 50% resulted in only an 8% decrease in motor neuron survival (calculated from the data of Lamb, 1980, 1981). None of these authors discussed the possibility that the relationship between motor neuron number and muscle mass might be nonlinear, despite the well-known fact that the relationship between the growth of different body parts is most commonly nonlinear (Huxley, 1932).
The present study indicates that the number of motor neurons in the facial nucleus of marsupials correlates strongly with body mass, growing with body mass raised to a small exponent of 0.184. With such a small exponent, doubling the body mass is accompanied by an increase of only 14% in motor neuron number, while halving the body mass would result in a decrease of only 12%. The figure of 14% is remarkably similar to the 15–20% increase in the number of spinal cord motor neurons that results from natural or experimental doubling of the target muscle field of birds and amphibians (Beucker, 1945; Pollack, 1969; Hollyday and Hamburger, 1976; Hollyday and Mendel, 1976; Lamb, 1979). The predicted 12% decrease following a halving in mass is broadly consistent with the 8% decrease that followed an experimental halving of target muscle mass (Lamb, 1980, 1981). Overall, these data strongly support the quantitative matching theory of motor neuron development that was advocated by the findings of Tanaka and Landmesser (1986).
Facial motor neuron number has been measured in many different placental mammals. The most comprehensive data set is that presented by Sherwood (2005), based on 18 species of primates. He found that the number of motor neurons varied from 3,517 in Galago to over 10,000 in large primates (Papio, Pan, Gorilla, Homo). Sherwood did not present data on body weight, but he measured the allometric correlation between motor neuron number and the volume of the medulla oblongata, which yielded a moderate correlation coefficient of 0.56. In an attempt to compare the primate data series with that of marsupials, we have taken Sherwood's data on motor neuron count and correlated them with estimates of body weight in each of the species. The body weight estimates were based on the median of body weight ranges quoted in the web site of the National Primate Research Center of the University of Wisconsin and Madison (http://pin.primate.wisc.edu). The correlation can be seen in Fig. 3, with a correlation coefficient of 0.732 (P = 0.0013). The slope of the graph shows that the number of facial motor neurons in primates varies with body mass raised to an exponent of 0.127 (P = 0.0013; 95% confidence interval 0.061–0.193), which is very close to the figure we obtained from our analysis of the marsupial data (0.184). Remarkably, the data points for marsupial and primate species are well aligned, suggesting that the relationship between numbers of motor neurons and body mass (or muscle mass) may be shared across the two mammalian orders.
Behavioral Specialization and Motor Neuron Number
The idea that the size of motor neuron pools correlates with the sophistication of movements executed by a particular muscle group probably stems from the extensive evidence of increased size of sensory systems in animals with sensory specializations. Examples of increases in size of sensory parts of the brain include the barrel cortex in mice (Woolsey and Van der Loos, 1970), the visual cortex of primates (Van Essen, 1985), and the auditory cortex in bats (Jones and Teeling, 2006). On this basis, it seems at first reasonable to expect that specialized motor systems, such as the use of the fingers in primates and the whisking of vibrissae in rodents, might be associated with a generous increase in the number of motor neurons in the relevant motor nucleus. We will refer to this proposal as the behavioral specialization theory of motor neuron development. Sherwood (2005) summarized a wide range of data that claim to link facial muscle specialization with increased motor neuron number. In support of this link, he cites published data on the number of neurons innervating the muscles of facial expression in hominoids, the pinna of the ear in the echolocating bat (Friauf and Herbert, 1985), the perioral muscles of the pig (Marshall et al., 2005), and the nasolabial muscles of the mouse (Ashwell, 1982). However, Sherwood points out that some of these studies are observational rather than strictly quantitative, and he concludes that overall the evidence for an increase based on skill is weak at best. In the present sample of marsupials, we do not find the significant increase in the number of facial motor neurons in whisking over nonwhisking animals, such as would have been predicted by the behavioral specialization theory of motor neuron development. It appears, therefore, that behavioral specialization in marsupials can occur without a concurrent increase in the number of motor neurons involved in the particular behavior.
However, a fine-grain direct comparison between the innvervation of pinna muscles in rat and bat (Friauf and Herbert, 1985) does supports the behavioral specialization theory of motor neuron development. Friauf and Herbert (1985) compared the location and number of motor neurons supplying five individual auricular muscles in a rat (a roof rat-Rattus rattus) and a bat (an echolocating flying fox-Rousettus aegyptiacus), with the aim of determining whether the more sophisticated pinna movements in the bat were associated with special features in the facial nucleus. They found that there was almost no overlap between the representations of individual pinna muscles in the bat, whereas there was significant overlap in the rat. They found that the bat had more pinna motor neurons than the rat (1,646 versus 1,110). They concluded that the presence of discrete motor neuron populations and the increased motor neuron number both contribute to the existence of specialized ear movements in bats. As the body weight of the two animals is similar (110–170 g for the bats and 145–250 g for the rats), bats would have about 50% more motor neurons than expected compared to the rat, which would be associated with more sophisticated movements of the pinna.
In addition, Dobson and Sherwood (2011) have examined the relationship between facial nucleus volume on the one hand, and visual cortex size and social group size on the other in catarrhine and platyrrhine anthropoid primates. They found a correlation between facial nucleus volume and visual cortex size, and between facial nucleus volume and social group size in catarrhine but not in platyrrhine monkeys. This suggests that the size of motor nuclei may adapt to behavioral specialization in some taxonomic groups but not in others. We conclude that the relationship between motor specialization and behavior is still an open question, which might be clarified by the collection of more data from different taxonomic groups.
Does Facial Motor Neuron Number Depend on the Number of Cycles of Division During Development?
The maximum number of motor neurons in an individual cannot be greater than the number of embryonic motor neurons generated before these cells make contact with muscles. A study in the mouse showed that the embryonic peak number was over 6,000 (Ashwell and Watson, 1983). In marsupials, the peak number of facial motor neurons during development should therefore be at least as great as the largest number of motor neurons counted in an adult (12,160 in a wombat). The generation of the final number of motor neurons depends on the number of progenitor facial motor neurons and the number of doublings of this number during development. Using chimeric mice, Herrup et al. (1984) have determined that the number of progenitor cells in the murine facial nucleus is 12. If 12 progenitors were to pass through nine cycles of division, this would generate a maximum number of 6,144 motor neurons, which is very close to that observed in the mouse before the period of cell death (Ashwell and Watson, 1983). In this case, a single extra round of cell division would double the number of motor neurons to 12,288, which is similar to the maximum number we found in marsupials. This would indicate that the number of doublings in large marsupials should be at least 10.
There is some support for the possibility that the number of mitotic cycles is tightly controlled. Hamburger (1975) has pointed out that the number of mitotic cycles involved in the production of spinal motor neurons seems to be precisely programmed. He observed that the total number of spinal motor neurons produced before the onset of neuron death in a number of 5.5-day and 6-day chicken embryos was relatively constant (ranging from 18,900 to 21,600). On the other hand, Herrup et al. (1984) did not find a consistent quantal pattern in the development of the mouse facial nucleus, and they suggested that external factors might interfere with the pattern of cyclical division of each progenitor line.
This raises the important question of whether numbers of motor neurons are adjusted in animals endowed with larger muscle masses by decreasing cell death—until a step increase occurs in the peak number of motor neurons generated (by adding an extra round of cell division). This scenario raises the intriguing possibility that all marsupials (and perhaps all mammals) of similar body size might be endowed with the same maximum embryonic number of motor neurons in development, in a one-size-fits-all solution that can be subsequently tailored by cell death for individuals of different size within a same range of body mass. Further increases could only be achieved with a step increase (one more cycle of doubling) in numbers of motor neurons generated. In this scenario, a small marsupial such as the marsupial mouse would incur a drastic reduction of the number of facial motor neurons after the period of synaptogenesis, from the supposed 12,288 down to 2,308. In the case of very large marsupials such as the wombat and kangaroo, on the other hand, the loss of motor neurons by cell death would be very small. It is worth noting here that reports of massive developmental motor neuronal death in vertebrates have all been based on very small species, notably the chick, and mouse.
While we cannot assume that the numbers of progenitors and of cell cycles is the same in all marsupials, this hypothesis provides a useful basis for further comparative analyses. The hypothesis has the attraction of economy, by implying that across species, a simple, shared matching mechanism could adjust brain components to match body size. While extensive developmental cell death has been documented in a variety of brain regions, including motor nuclei, almost all such studies have focused on small animals (toads (Hughes, 1961), chicks (Hamburger, 1975), and mice (Ashwell and Watson, 1983); see the extensive review by Oppenheim, 1991). Comparative studies of motor neuron development comparing small and large related animals, such as turkey and quail, or capybara and mouse, would address this hypothesis.
On the Meaning of a Small Scaling Exponent for Numbers of Motor Neurons
Here we find a striking difference in the scaling of brain mass and of numbers of facial motor neurons with increasing body mass across marsupial species, with the former increasing at a faster pace than the latter. We do not yet know how brain mass scales with the total number of brain neurons in marsupials, and our previous work on glires (rodents/lagomorphs), primates and insectivores showed that brain mass can no longer be considered a similar proxy for numbers of neurons across mammalian orders (reviewed in Herculano-Houzel, 2011). Therefore, the possibility remains that the facial nucleus gains motor neurons at the same rate as the marsupial brain as a whole. However, the finding that the number of facial motor neurons scales with brain mass raised to a small exponent of 0.307 suggests that it lags behind the addition of neurons to other brain structures. This is consistent with current views on how larger brains add complexity to the processing of environmental and somatic information by gaining neurons faster than would be required to deal with information pertaining to the body.
The probable lag in the scaling of numbers of motor neurons in the facial nucleus relative to the whole brain is consistent with our findings on primate spinal cord scaling, in which we observed that the spinal cord gains neurons at a slower rate than the brain (Burish et al., 2010). In that study, we found that the primate spinal cord gains neurons with increasing body mass raised to an exponent of 0.346, which is higher than the exponent of 0.184 found here for facial motor neurons alone in marsupials. However, two significant observations arise in common from the two studies. First, a relatively small number of neurons seems to be required to operate the body, or parts of it. Just a few thousand neurons are required to control facial muscles in marsupials, and only several million neurons to provide the sensory and motor interface with the body, from the neck down (Burish et al., 2010). Second, the finding that numbers of facial motor neurons in marsupials and spinal cord neurons in primates are scaled to larger bodies at a rate that is far below linearity might be related to a decrease in average motor unit size in larger animals. However, we lack direct evidence to support this speculation. Further comparative studies relating the size of particular motor neuron pools, such as those that innervate the hands, and manual dexterity are required to address this issue.
The authors owe an enormous debt to Dr John I. Johnson of Michigan State University who collected the marsupial brain specimens and produced the beautiful histological sets on which this work was based. They thank Dr Hironobu Tokuno for providing high resolution images of the sugar glider brain to improve the quality of our illustration.