Site search   
  Home    Publications

Browse Publications
List of Contents

Quantitative Neurogenetics & QTL Mapping

Genetics of Myopia

Control of Neuron Number and Stereology

Growth Cones and Dying Axons

Retina Development and Visual System Mutants

Grant Application

U.S. Patent


Need Help?
Help with Publications
Help with Nervenet
Contact Us


Genetic Analysis of Variation in Neuron Number
Richelle Cutler Strom


Throughout mammalian evolution brain size has varied from a massive 6800 gm in whales down to a meager 0.10 gm in shrews (Count, 1947; Jerison, 1973) yet the regional organization has remained virtually the same (Butler and Hodos, 1996). Indeed, upon microscopic examination of the mouse and human brain Ramon y Cajal, a Nobel Laureate recognized for his detailed descriptions of the nervous system, stated: "In my opinion there are only quantitative differences, not qualitative differences between the brain of a mouse and that of a man." The genetic differences responsible for the vast differences in brain size are unknown.

Differences in brain size between species ultimately depend on the accumulation of micro-evolutionary changes among the individuals within a species (Butler and Hodos, 1996). Variation within a species can be substantial. For example, normal variation in brain weight among humans ranges two-fold, from 900 to 1,800 gm (Cobb, 1965). Even among closely related strains of inbred laboratory mice brain weight differs by as much as 13%, from 429 mg in C3H/HeJ to 490 mg in C3H/HeSnJ. Examining the genetic and developmental bases for variation in brain size within a species can reveal bases for the large diversity in brain size found across species.

Variation in brain weight results primarily from differences in glia and neuron number. This has been demonstrated by finding a high correlation between brain weight and total brain DNA (Zamenhof and van Marthens, 1978). Also, neocortical neuron number in humans is correlated with brain weight (r = 0.56) and with neocortical volume (r = 0.69) (Pakkenberg and Gundersen, 1997). Not surprisingly, estimates of neuron numbers for discrete structures within the brain are also highly variable. For example, the volume and cell number in the visual cortex of humans and other primates ranges almost three-fold (Gilissen and Zilles, 1996; Suner and Rakic, 1996). Large variation in neuron numbers have also been found among inbred mouse strains in numerous cell populations, e.g., midbrain dopaminergic neurons (Ross et al., 1976); forebrain cholinergic neurons (Albanese et al., 1985); and granule cell number in hippocampus area dentata (Wimer et al., 1978).

Variation in neuron number is the result of both environmental and genetic factors. A key question is how much of the variation is attributed to the environment and how much is attributed to genetic factors. A heritability estimate expresses the magnitude of the genetic contribution and is measured as the ratio of the genetic variance over the total phenotypic variance (VG/VP). The heritability of brain size in humans has been estimated to be as high as 94% (Bartley et al., 1997). Heritability estimates for brain weight and neuron number in mouse, are also high, 0.67 for brain weight (Roderick et al., 1973) and 0.80 for granule cell number (Wimer and Wimer, 1989). These high heritability estimates indicate that variation in brain size and neuron number is generated predominantly by genetic variation. Identifying the genetic variation that modulates neuron number can expose specific molecules and the developmental and regulatory mechanisms controlling neuron number.


Forward genetic approach

The first step to determine the genetic bases of variation is to map the genomic location of the genes responsible for the phenotypic variation. Using phenotypic variation to map the responsible genes is known as the forward genetic approach. One method for mapping a gene is linkage analysis. The ability to perform linkage analysis requires genetic variation between individuals. The genetic sites where DNA sequences differ between individuals are called polymorphic loci, with each variant form representing an allele type. Linkage analysis involves testing for an association between the inheritance of a particular allele and the inheritance of a particular phenotype. A strong association between an allele and a particular phenotype indicates that a gene that contributes to the phenotypic variation maps near the allele. The ability to use alleles as genetic markers to infer the presence of a nearby gene is dependent on the physical linkage of loci on chromosomes. The closer the physical proximity between two loci the less likely they are to form a chiasmata between them and recombine with their homologous chromosome during meiosis. The frequency that genetic loci recombine translates into units of genetic distance called centimorgans (cM). The frequency of recombination can be determined from the percentage of non-parental allele combinations in the offspring.

Gene mapping before recombinant DNA technology required mutations that exhibited prominent visible or biochemical phenotypes in the offspring to carry out linkage analysis. For example, the first linkage demonstrated in the mouse was between the albino (c) mutation and the pink-eyed dilution (p) mutation (Haldane et al., 1915). In the last decade a dense map of genetic markers called microsatellites and their ease of typing by the polymerase chain reaction (PCR) has revolutionized genetic mapping. Microsatellites or simple sequence repeats (SSR) consist of di- or tri-nucleotide repeats, such as CACACA (Dietrich, 1992; Love, 1990) and are dispersed at a high frequency throughout mammalian genomes. Over 6,000 microsatellites have been genetically mapped in the mouse and recently 4,000 of these markers were physically mapped (Dietrich et al., 1994). Microsatellites have been typed in 12 inbred mouse strains and have been found to be highly polymorphic between strains.

Gene mapping has been greatly facilitated by the production of genetic variants produced by x-ray, chemical, and P–element induced mutations (Sentry et al., 1994). In Drosophila and zebrafish, the production of mutant phenotypes by mutagenesis and careful screening has lead to the identification of many genes involved in neurogenesis (Sentry et al., 1994). Large-scale mutagenesis screens are now underway in mouse and are expected to yield a plethora of mutant phenotypes for study (Kasarskis et al., 1998). However, one drawback of induced mutations that they have pleiotropic effects, and many mutations will affect the viability of the animal, making it difficult to study.


Exploiting natural genetic variation

Of the roughly 40,000 genes expressed in the human brain, approximately 10–20% are expected to have multiple alleles (Silver, 1995). Almost all of the heritable variation between individuals results from random segregation of different alleles that produce subtle quantitative effects (Falconer and Mackay, 1996). A quantitative trait, such as height, has a continuous range of phenotypes in a population and is controlled by multiple genetic and environmental factors. In contrast, qualitative traits have distinct phenotypes, such as coat color in mice, and are controlled by one or two genes with little environmental influence. Genes with allelic variants that produce quantitative variation are called quantitative trait loci (QTLs). Quantitative traits such as neuron number can be used in a forward genetic approach to identify the location of the QTLs. Mapping a QTL is different from mapping a gene with qualitative expression in that there is no one-to-one relation between a genotype and a phenotype. This is because the effects of other genes and environmental variation combine to produce the expressed phenotype. A QTL is mapped by comparing the phenotypic means from the progeny in each genotype class at every marker across the genome. Linkage is suggested when there is a significant difference between the means of the genotypic classes.

Mice are the ideal animal for genetic analysis, especially if the goal is to extrapolate to humans. Although, the evolutionary divergence of mice and humans occurred over 60 million years ago, there is still a close correspondence between the genomes of mice and humans. Large segments of the human and mouse genomes are conserved and the differences found today can be explained by breaking the mouse genome into roughly 130 segments and shuffling them into a new order (Nadeau and Taylor, 1984). The existence of a great variety of inbred mouse strains, which represent a wealth of quantitative variation, is crucial in a quantitative genetic analysis. Mice from a particular inbred strain are genetically identical or isogenic. Phenotyping a number of isogenic mice to obtain a strain average can minimize the variance due to environmental factors, measurement errors, or developmental factors, resulting in a more accurate mean phenotype. Mice are usually inbred by at least 20 generations of brother and sister mating. Inbreeding forces the alleles to become homozygous at all loci. Finally, in consideration of time and cost, mice reproduce quickly and are relatively inexpensive to house.


Dissertation research

In my dissertation research, I use the forward genetic approach to assess and dissect genetic sources of variation in neuron number. I have focused on variation in neuron number on a global scale by using the surrogate measure of whole brain weight. I have also focused on variation within a discrete neuron population, the retinal ganglion cells. I began my studies by determining the relative proportion of variance in brain weight and ganglion cell number that is due to environmental and genetic factors (Chapters 2 and 4). Subsequently, I used linkage analysis with composite interval mapping techniques to pursue QTLs that are responsible for genetic variation in brain weight and ganglion cell number (Chapters 3 and 5). To identify the developmental mechanisms generating variation in neuron number I have examined the role of cell production and cell death in the generation of ganglion cell variation among mouse strains (Chapter 6).

In the next section I address the behavioral significance of variation in brain size and neuron number. Finally, I conclude this introduction with a summary of molecular neural development to acquaint the reader with the processes in which QTLs might be involved to produce variation in neuron number.


Please note that although I present the work in Chapters 4 and 5 using the first person singular pronoun, this work is actually the result of collaboration with Robert Williams, Dan Goldowitz, and Dennis Rice. For the most part, the methods and results in Chapters 4 and 5 match those in recent published papers. Differences include the addition of new strains and larger numbers of cases per strain and as a result I have also reanalyzed most of the data.


Importance of neuron number

Brain weight

Brain tissue is metabolically costly and an increase in brain size should be counterbalanced by an increased capacity for behavioral adaptations (Williams and Herrup, 1988). Increased neuron number is usually associated with more elaborate neural networks that have evolved for novel sensory or behavioral adaptations required for exploiting new ecological niches. The behavioral significance of variation in brain size is readily apparent across species, where more complex behaviors are generally associated with larger brains (Aboitiz, 1996). However, within species behavioral associations with variation in brain size are less evident. Controversial findings have been reported in humans, with positive correlation between cerebral volume measured by MRI and IQ in children (Reiss et al., 1996), but the absence of a correlation between forebrain volume, also measured by MRI, and IQ in young adults (Tramo et al., 1998). The disparity may have to do with the differences in the age of the subjects. Older subjects harbor the effects of years of environmental variance, as well as an extended period of brain growth, which could dissolve any initial correlation between brain size and IQ.

In mice, high and low brain weight lines, generated by selective breeding, initially showed behavioral differences in field activity, but these differences disappeared in later generations (Wimer et al., 1969). A recent study with rats reported a high correlation between general intelligence and brain weight (Anderson, 1993). In summary, an association with brain weight and behavior within a species is inconclusive and the outcome of individual studies may depend on the particular structures responsible for differences in brain size and the appropriateness of the behavioral tests.


Discrete populations

Behavioral associations are found with the size of discrete neuronal populations. The relative size of the hippocampus has been shown to be associated with spatial memory capacity in species of passerines, cowbirds (Healy and Krebs, 1996; Reboreda et al., 1996), and kangaroo rats (Jacobs and Spencer, 1994). In males of many seasonally breeding songbird species the volume of the song nuclei, hyperstriatum ventralis, pars caudalis (HVc) and archistriatum (RA), increases along with their singing ability during each breeding season (Smith et al., 1997). In male zebra finches, neuron number and the volume of HVc and RA in an individual correlates positively with the number of tutor syllables copied (Ward et al., 1998). Moreover, in humans, cerebellum volume significantly correlates with the individual’s verbal memory and fine motor dexterity (Paradiso et al., 1997). In conclusion, there is a strong association between the number of neurons in a discrete population and its behavioral output, particularly if the behavior is important for survival.

Numbers of retinal ganglion cells range from fifty thousand in nocturnal rodents to several million in diurnal birds and primates (Rager and Rager, 1978; Rakic and Riley, 1983). Variation is also marked within species: numbers range from 0.7 to 1.5 million in humans (Curcio and Allen, 1990), and from 40,000 to 80,000 in mice (Williams et al., 1996). It is not known whether visual acuity differs in animals with high and low ganglion cell number. However, visual acuity in glaucoma patients is compromised after a small decrease in ganglion cell number (Quigley et al., 1989). The difference in visual acuity will depend on what classes of ganglion cells are variable. There are three major morphological classes of ganglion cells in the mammalian retina (alpha, beta, and gamma)(Rodieck and Brening, 1983). The extent of variability within the represented classes of ganglion cells has not been studied. However, the numbers of beta and gamma cells are likely to be the most variable since they are the most numerous, with each representing roughly 45% to 50% of the ganglion cell population in several species (Wässle and Boycott, 1991; Williams et al., 1993). The high density and short dendritic fields of the gamma ganglion cell class are thought to set the limit for spatial resolution (Wässle and Boycott, 1991) and thus, even moderate variation in the density of this class should affect visual acuity. The two-fold variation in ganglion cell number among mice is significantly associated with retinal area (r = 67, t = 269, P <0.001) (Rice et al., 1995), and thus cells per degree of visual angle, the critical factor determining visual acuity, may not change with increased cell number.

Developmental control of neuron number

Neuron number in the brain

The number of neurons in the adult central nervous system results from a balance between cell proliferation and developmental cell death. While there is considerable cell death in most vertebrate brains, the major process defining brain size in a species is neurogenesis. Neural progenitors originate from the embryonic ectoderm. The neural fate is the default pathway for embryonic ectoderm. The switch from the neural fate into the epidermal fate is initiated by an endogenous inhibitory signal within the ectoderm, possibly BMP-4 (Hemmati-Brivanlou and Melton, 1997). Release from the neural inhibitory signal may be mediated by follistatin, secreted from the mesoderm.

Once cells are committed to the neural fate they coalesce to form the neural plate. The neural plate then invaginates along the neuroaxis, folds inward, and fuses to form the neural tube. In the neural tube, neuroepithelial cells divide in the ventricular zone, leave the cell cycle, and migrate outward to reside in the appropriate cell layers. In neurogenesis, a complex molecular cascade controls the progressive hierarchy of cell fate competence that ultimately leads to a specific cell type in the appropriate place and numbers.

In mice, neocortical neurons are generated from a founding population in the pseudostratified ventricular epithelium (PVE) between embryonic day 11 (E11) and E17. During this interval of neurogenesis there are an average of eleven cell cycles, which increase the number in the founding population by 140 fold (Takahashi et al., 1995). At the onset of neurogenesis neural progenitors undergo symmetrical divisions in which both daughter cells return to the cell cycle (Rakic, 1988). During this time, cells increase in number exponentially. As development proceeds the neural progenitors begin to divide asymmetrically, with one daughter cell leaving the cell cycle and differentiating while the other daughter cell continues to divide. Gradually, as the neurogenesis proceeds, both daughter cells will cease to divide and differentiate. The portion of the population that remains proliferative represents the P fraction, and is equal to 1 when all cells are dividing, while the portion of cells leaving the cell cycle represents the Q fraction, for quiescent, and is equal to 1 when all cells have ceased to divide (Takahashi et al., 1997).

Developmental mechanisms for variability. Variability in neuron numbers could result from differences in the (1) number of founding neural progenitors, (2) parameters of the cell cycle, and (3) duration of cell proliferation. The variation of Purkinje cell number among chimeric mice produced between wildtype mice and lurcher, a mouse mutant whose Purkinje cell population is completely absent, appears to occur in quantum units equal to 10,200 cells. The integral units suggest that variation in Purkinje cell number results from differences in the initial proliferative fraction of wildtype progenitors (Wetts and Herrup, 1983). Quantum differences in the Purkinje cell number in chimeras made between the mouse strains C57BL/6 and AKR/J has also demonstrated that variation can result from differences in the initial clone size of progenitors (Herrup, 1986). These studies suggest that differences in cell number can originate from differences in the number of early progenitors, probably arising during regionalization of the neural plate or neural tube.

Variability in parameters of the cell cycle. Variation in neuron number could result from differences in the parameters of the cell cycle. One such parameter is the rate at which proliferative fraction Q approaches 1, which could be a major factor in controlling the number of neurons generated (Takahashi et al., 1997). In the mouse neocortex the Q fraction equals 0.5 at cell cycle number 8. However, if the cell cycle at which Q equals 0.5 were delayed from 8 to 10 the population would expand 5–fold. The number of neurons generated is also determined by the progressive increase in the length of the cell cycle. During the period of neocortical neurogenesis in mouse, the length of the cell cycle increases from ~8.1 hours to ~18.4 hours. This increased cycle time is a result of a 4-fold increase in the length of G1 (Takahashi et al., 1995). A broad range in the length of G1 has been found among different cell populations (Jacobson, 1991).

In primates the striate visual cortex has a higher density of neurons compared to other cortical areas. The differences in cell number between the striate and extrastriate cortex was shown to result from a higher proliferative fraction, but also from a higher portion of cells cycling, as shown by 3H-thymidine pulse labeling (Dehay et al., 1993). The proliferative portion, known as the labeling index, was 50% higher in the striate cortex than in the extrastriate. The authors concluded that differences in the mitotic rate must have produced the difference in the labeling index, however an equally plausible explanation is a difference in the progression of Q. Nevertheless, it is not known whether variability in the rate of Q or variability in the length of the cell cycle contributes to variation in cell number.

Variability in the number of cycles can have a large impact on cell number. For example, if the number of cell cycles doubled while the relative progression rate of P to Q was the same, the population would increase 60-fold. Of course, an increase in the cell cycle number, and consequently the duration of neurogenesis, means that the over all progression of Q is slower. It has been known for some time that the large differences in neuron number found between species is best explained by differences in the period of neurogenesis (Passingham, 1975). A recent study confirmed this finding by finding a high correlation between the peak day of neurogenesis and the size of 51 brain structures among 7 different mammals (Finlay and Darlington, 1995).

Extrinsic signals in cell cycle control. Although much is known about cell cycle mechanics and the molecular components, little is known about the regulatory mechanisms governing variation in the exit from the cell cycle (Ross, 1996). In vertebrates the decision of a neural precursor to leave the cell cycle during development involves the Notch signaling pathway (Beatus and Lendahl, 1998). Notch, a transmembrane protein, and its ligands, Delta and Serrate, are the mediators of cell-cell interactions that control the competence of neighboring cells. Through a process called lateral inhibition, a prospective neuron expressing Delta or Serrate on its cell surface binds the Notch receptor on a neighboring cell thereby activating the cell’s Notch pathway, which inhibits cell differentiation and holds them in a proliferative state. Growth factors may regulate the expression of genes in the Notch pathway. For example, the epidermal growth factor (EGF), which can induce widespread proliferation of neuronal precursors in the brain (Kuhn et al., 1997), was recently shown in the developing retina to down regulate Mash-1, a proneural factor within the Notch signaling pathway (Ahmad et al., 1998). A connection between EGF and Notch suggests that one way growth factors may act is by down regulating the Notch pathway, thereby holding cells in a proliferative state. Other extrinsic factors such as retinoic acid and thyroid hormone influence the extent of cell proliferation in the brain by inducing the withdrawal from the cell cycle and subsequent differentiation (Jacobson, 1991; Kelley et al., 1994). How the expression patterns and concentration levels of specific growth factors and hormones are regulated to set proliferative limits to defined populations of neurons is not known.

Finally, extrinsic factors ultimately regulate the progression of the cell cycle by inhibiting or promoting the build up of cyclins and the activity of their dependent kinases (Hutchinson and Glover, 1995). The complete pathway between the proliferative effects of growth factors and hormones and their effectors within the cell cycle during neurogenesis remain to be mapped out. In summary, the developmental control of neuron number involves a complex collaboration of molecular pathways that converge to determine the progression of the cell cycle through specific checkpoints.


Retinal ganglion cell development

In mouse the neural retina is part of the brain, comprising ~0.4% of the brain’s weight. The eyes begin to form in the mouse around E9 with the bilateral out-pocketing of the forebrain into a structure called the optic vesicle (Mann, 1964; Pei and Rhodin, 1970). The optic vesicle grows outward toward the surface ectoderm and then invaginates to form the optic cup. The optic cup then separates into an inner portion, the neural retina and an outer portion, the pigment epithelium.

Proliferating retinal cells migrate in a to-and-fro pattern, moving to the inner retina to synthesize DNA and then back to the outer retina to divide. During retinal development the percentage of proliferating cells (P) decreases as cells from asymmetric divisions leave the cell cycle, migrate toward the vitreal surface and differentiate. The neural retina develops into three distinct cell layers, a ganglion cell layer, a bipolar layer and an inner photoreceptor layer. The onset of differentiation for retinal cells occurs in a temporally-ordered manner, with retinal ganglion cells, amacrine cells, and horizontal cells beginning to differentiate early, while photoreceptors, bipolar cells and Müller glia cells begin to differentiate later (Sidman, 1961). Although the generation of cell types begin and peak at different times the cell types are being generated simultaneously throughout most of the developmental period. In mouse, ganglion cells start to differentiate on E11 and continue until just before birth (Dräger, 1985). Developmental cell death of ganglion cells begins at, or just before, birth, peaks between postnatal days 4–6, and is essentially complete by P12 (Linden and Pinto, 1985; Young, 1984). Variation in ganglion cell number could result from differences in the extent of ganglion cell production or ganglion cell death.

Variability in ganglion cell production. Variation in ganglion cell production could result from differences in: (1) numbers of retinal progenitor cells, (2) the kinetics and duration of progenitor cell proliferation, or (3) bias in cell fate determination in progenitors. The variance of retinal cell numbers in embryonic chick to peaked during the exponential phase of cell division and were lowest after cell differentiation (Morris and Cowan, 1984). A regression analysis of the variance in cell number at the start of incubation and growth rate indicated that the variation in the exponential phase was due mostly to the variance in the number of retinal progenitors at the start of incubation. However, variation in retinal cell number at the start of incubation could result from subtle differences in the correspondence of the developmental stage and not necessarily to the variability in the number of retinal progenitors

Variance in numbers of multipotent retinal progenitors should have consistent effects on the number of many retinal cell types. However, a comparison of horizontal cell and ganglion cell numbers for six strains has demonstrated that ratios in these early-generated cell types are not always matched (Williams et al., 1998). In fact, preliminary data indicates a weak negative correlation exists between the number of ganglion cells and photoreceptors within inbred strains (Williams et al., 1998). If true, that would suggest that a reciprocal relationship exists between early- and late-generated retinal cell types and that variation in ganglion cell number is produced by differentiation bias. An example of the reciprocal relationship between retinal cell types is found when the inhibition of Notch signaling results in an increase of early-generated cell types at the expense of later-generated cell types (Dorsky et al., 1997). It is possible that a genetic bias in the Notch signaling pathway may play a role in executing the differentiation ratios of retinal cell types.

Variation in the ganglion cell number could result from differences in the proliferation kinetics of progenitors that give rise to ganglion cells. The rates of mitosis may be influenced through inhibitory molecules and pathways. One example is dopa—a tyrosine metabolite that normally has inhibitory effects on cell genesis in the retina. The absence of dopa in albino rats leads to an anomalous up-regulation of ganglion cell production followed by an increase in the severity of cell death (Ilia and Jeffery, 1999). Finally, as with brain weight, large differences between species within a defined population is best explained by differences in the duration of neurogenesis. For example, a two-fold difference in the duration of retinal neurogenesis can explain the near two-fold difference in retinal ganglion cell number between gerbil and hamster (Wikler et al., 1989).

Variability in ganglion cell death. Differences in cell death can produce large differences in cell number across species. Estimates of retinal ganglion cell death range from none at all in fish to 40% in chicken, 50–60% in rat (Lam et al., 1982), and up to 80% in cat (Williams et al., 1986). The ganglion cell population in the domestic cat is 30% lower than that of wildcat —the species from which the domestic cat has evolved. Retinal ganglion cell number in a Spanish wildcat fetus and domestic cat fetus obtained at E38, were remarkably close, suggesting that cell death rather than cell production generates the marked species differences (Williams et al., 1993).

Variation in the severity of cell death may result from differences in titers of neurotrophic factors. The neurotrophins—BDNF and neurotrophin-3/4—have been found to increase survival of retinal ganglion cells in chicken and rat (Ma et al., 1998; Rosa et al., 1993). Neuregulin, found on the cell surface and as a secreted protein, can also increase survival of neonatal rat retinal ganglion cells (Bermingham-McDonogh et al., 1996). Differences in the concentration or expression time of these neurotrophic factors, their receptors, or components within their signaling pathways could produce variation in the severity of naturally occurring ganglion cell death.

In summary, the generation of variation in cell number could involve any of the molecular processes described above. Determining the genetic bases of variation in neuron number could identify key regulatory genes in neurogenesis, which may lead to a greater understanding of cell-cycle regulation. In addition, examining the molecular variants in the QTLs that produce natural variation in neuron number will lead to a better understanding of the genetic bases of quantitative variation and provide insight into the molecular nature of brain evolution.

Return to Table of Contents


Neurogenetics at University of Tennessee Health Science Center

Text Only Top of Page

Home Page  |  Genome DBs  |  Phenome DBs  |  Publications  |  People & Associates
Mouse Brain Library  |  Related Sites  |  |   MBL.ORG

Robert W. Williams | Alex Williams © 2002,