The first years of life are uniquely important for brain development

Figure 1 is an example of the sparse neural connections in the frontal cortex of a newborn. The circuitry necessary for complex information processing and complex behavior does not yet exist. Instead, it will be created in the first postnatal years by neuronal differentiation and growth, dendritic and axonal growth, axonal myelination, and a tremendous increase in synapse numbers. Higher order social, emotional, cognitive, attention, speech, and language functions are all mediated by slowly maturing regions such as frontal cortex, while basic sensory and perceptual functions are mediated by relatively more rapidly and earlier maturing systems.

Golgi-stained sections showing the growth of pyramidal neuron soma and dendrites in the middle frontal gyrus. The normal newborn has sparse neural circuitry; then, with increasing age, there is a tremendous increase in the complexity of dendritic arborizations. In this frontal cortical area, the dendrite arbors for layer 3 pyramidal neurons, which are only 3% of mature size in the newborn, are still only about 50% by 2 years of age and do not reach 100% until the end of childhood (see text). From The Human Brain (3rd ed), by J. Nolte, 1993, St. Louis, MO: Mosby Year Book/Elsevier. Copyright 1993 by Mosby Year Book/Elsevier. Reprinted with permission from Elsevier.

In the frontal cortex at birth, the length of the pyramidal cell dendritic arbors in layer 3 is only 3% of full mature size, and by 2 years of age it is only 48% (see example in Figure 1); many additional years will pass before full size is reached. By comparison, in primary visual cortex pyramidal dendritic arbors are already 33% of full size at birth, and by 2 years they have reached mature size (table 1.3 in Huttenlocher, 2002). Myelination follows the same functional hierarchical pattern, with basic-level systems developing earlier than higher order association systems (Kinney, Brody, Kloman, & Gilles, 1988). Compared to more posterior cortices, frontal cortex undergoes synapse formation later, and for a longer period of time, and develops far larger pyramidal neurons with far more synapses (about 100,000 vs. about 20,000) and far larger dendritic (e.g., in layer 3, 6836 vs. 2900 μm) and axonal arbors and axonal projections (Huttenlocher, 2002).

It has long been recognized that this unique period of neural differentiation and circuit formation is also a time when the brain is particularly vulnerable to abnormal events that disrupt the formation of cortical connectivity, producing aberrant circuits and functions and behavioral deficits (Dobbing, 1981; Huttenlocher, 2002; Kinney et al., 1988). More slowly maturing brain regions such as frontal cortex have a longer window of vulnerability than more rapidly maturing ones, such as the occipital cortex.

Findings in young autistic children

Recent evidence indicates that autism may involve brain growth pathology during this very window of vulnerability; it is early, brief, and age delimited (for reviews, see Courchesne, 2004; Courchesne et al., 2004; Courchesne & Pierce, 2005a; Dementieva, Vance, Donnelly, Elston, Wolpert, Ravan, DeLong, Abramson, Wright, & Cuccaro, 2005). At birth, head circumference in autism, and therefore brain size (Bartholomeusz, Courchesne, & Karns, 2002), were found to be either equivalent to the normal average or slightly smaller than normal (Courchesne, Carper, & Akshoomoff, 2003; Courchesne et al., 2001; Dementieva et al., 2005; Gillberg & de Souza, 2002; Lainhart, Piven, Wzorek, Landa, Santangelo, Coon, & Folstein, 1997).¹

Both cerebral and cerebellar white matter volumes in autistic toddlers have been found to be abnormally enlarged (18 and 39%, respectively); cerebral gray matter is enlarged by 11% (Courchesne et al., 2001). Most importantly, there were striking regional differences in this overgrowth pathology: in 2- to 4-year-old autistic toddlers, frontal cortex and frontal white matter volumes were the most abnormally enlarged, but the occipital lobes did not differ significantly from normal (Carper, Moses, Tigue, & Courchesne, 2002). Within the frontal cortex, dorsolateral and mesial prefrontal cortices were most abnormal, but the precentral gyrus, like the occipital lobe, was not significantly different from normal (Carper & Courchesne, 2005). Temporal gray matter and parietal white matter were also enlarged but not to the magnitude of frontal gray and white volumes (Carper et al., 2002); in the limbic system, the amygdala was also enlarged by 4 years of age (Sparks et al., 2002). A PET study of 3- to 4-year-old autistic children reported hypoperfusion in frontal cortex in autism which was interpreted as evidence of delayed frontal maturation (Zilbovicius, Garreau, Samson, Remy, Barthelemy, Syrota, & Lelord, 1995).

An equally striking and important second phase of growth pathology in autism follows on the heels of the early overgrowth: abnormally slow or arrested growth. A recent meta-analysis of 12 MRI brain volume studies showed an autism brain size difference from normal of about 11% between 1 and 3 years of age, which declined through childhood (when the normal, but not autistic, brain continues to grow); by adolescence the autism brain size difference from normal is only about 1–2% (Redcay & Courchesne, 2005). Between 2–4 and 6–8 years of age, frontal and temporal cortical gray matter increase by 20 and 17% in normal children but change by only 1 and −1%, respectively, in autism (Carper et al., 2002). The dorsolateral subregion of frontal cortex increases by 27% from 2–5 years to 5–9 years of age in normals, but by only 7% in autistic children (Carper & Courchesne, 2005). White matter growth is likewise retarded; for example, between 2 and 4 years and 7 and 11 years of age, frontal white matter volume increases by 45% in normal children, but by only 13% in autistic children (Carper et al., 2002), and between 2 and 3 years of age and adolescence, cerebellar white matter volume increases by 50% in normal children, but by only 7% in autistic children (Courchesne et al., 2001). Thus, regions that show the greatest early overgrowth in autism, frontal lobe and cerebellar white matter, also show sharply reduced or arrested growth thereafter.

Last, in contrast to the cerebellar white matter changes just described, during early development in autism there appears to be a reduction in the size of one or another region of the cerebellar vermis, which is largely gray matter (review in Courchesne, 2004). In one MRI study with over 200 autistic and control subjects (making this the largest autism MRI study ever done), Hashimoto, Tayama, Murakawa, Yoshimoto, Miyazaki, Harada, and Kuroda (1995, p. 229) reported underdevelopment of the cerebellar vermis in autistic individuals ranging from infants to adolescents. In another MRI study of 3- to 9-year-old children, reduction in the size of cerebellar vermis lobules VI–VII was found to be specific to autistic children compared with normal, fragile X, fragile X with autism, Down syndrome, and Down syndrome with autism children (Kaufmann, Cooper, Mostofsky, Capone, Kates, Newschaffer, Bukelis, Stump, Jann, & Lanham, 2003). In a third recent MRI study of autistic children, the posterior portion of the cerebellar vermis was found to be significantly reduced in size (Levitt, Blanton, Capetillo–Cunliffe, Guthrie, Toga, & McCracken, 1999). These findings are consistent with some previous developmental MRI studies reporting hypoplasia of the one or another region of the cerebellar vermis (e.g., Ciesielski, Harris, Hart, & Pabst, 1997; Courchesne, Yeung–Courchesne, Press, Hesselink, & Jernigan, 1988).

Findings in older autistic children and adults

Recent structural imaging results on older autistic children are compatible with this evidence of gray and white matter abnormality in autistic toddlers. For example, Herbert et al. (2003) reported greater white matter volume in 7- to 11-year-old autistic children and in a second study found that this deviation from normal was greatest in frontal white matter underlying cortex (so-called “radiate white matter”) and least in occipital lobe (Herbert, Ziegler, Makris, Filipek, Kemper, Normandin, Sanders, Kennedy, & Caviness, 2004). Hardan, Jou, Keshavan, Varma, and Minshew (2004) found abnormally increased gyrification in frontal cortex in an autistic sample ranging in age from 8 to 50 years. Levitt, Blanton, Smalley, Thompson, Guthrie, McCracken, Sadoun, Heinichen, and Toga (2003) reported that some frontal and temporal sulci are abnormally shifted superiorly and posteriorly in autistic children. Altered serotonin synthesis in the cerebello–thalamo–frontal pathway has been described in autistic children (Chugani, Muzik, Rothermel, Behen, Chakraborty, Mangner, da Silva, & Chugani, 1997) and a theory has been proposed linking serotonin abnormality during prenatal development to minicolumn and other neural abnormalities in autism (Chugani, 2000).

Neuron numbers: Excess in cerebrum, reduction in cerebellum

In collaborative and as yet unpublished research projects, neuron numbers in the cerebrum, cerebellum, and subcortical structures in a set of autistic and control postmortem cases aged 4–67 years were counted (Schmitz, 2004; Wegiel, 2004). These were among the first studies to utilize modern stereological procedures in a study of the autistic brain. There appeared to be a tendency toward an excess number of cerebral cortical neurons was found in autistic cases as compared to controls (Schmitz, 2004). Interestingly, the magnitude of this excess showed an age-related decline across child to adult autism cases. It is not yet known whether there are regional differences in the excess neuron numbers or in the rate and magnitude of the age-related decline in numbers.

In contrast, decreased neuron numbers were reported for the cerebellum and basal ganglia (Wegiel, 2004). Cerebellar Purkinje neuron numbers were decreased by about 30% and the volume of the cerebellum was reduced by about 19%. This loss of Purkinje neurons was seen at all ages in the autism sample from 4- to 67-year-olds. Large decreases in neuron numbers were also reported for the nucleus accumbens and basal ganglia, but neuron numbers in the hippocampus were remarkably similar in autistic and control cases. This quantitative and stereological study adds to a large number of observational reports that cerebellar Purkinje neuron numbers are abnormally reduced in autism (see reviews in Courchesne, 1997, 2004). Cerebellar Purkinje neuron loss may be the single most commonly reported neuronal abnormality in the autism literature to date.

Implications of increased cortical neurons and decreased Purkinje neurons

Many fundamental questions remain before the implications of an increase in cortical neurons and axons can be fully interpreted. Are there regional increases in neurons and axons that parallel regional increases in MRI gray and white matter volume? The only existent stereological count of cerebral neurons was global, not regional (Schmitz, 2004). In our pilot study, we reported what appeared to be an increased number of neurons in layer 3 throughout frontal cortical regions but not in primary visual cortex. Although this fits the reports of regional MRI volumetric differences, it was not a stereologically conducted count. It will be valuable for future studies to perform regional stereological neuron counts. Because cerebral neurons in different layers are generated at different but somewhat overlapping developmental times, it will also be important to determine whether or not the increases are layer specific. There have been hypotheses that autism involves deficient cortical inhibitory control, and so it will be interesting to know if there is an imbalance in the numbers of excitatory pyramidal neurons and inhibitory interneurons such as chandelier cells. Because roughly 75–80% of cortical neurons are pyramidal cells (Jones, 1984), an aberrant increase in their numbers could have a significant overall effect on both total neuron counts and gray matter volume even if the number of inhibitory neurons remained constant or even decreased slightly.

An excess number of cerebral neurons could be because of any one of several possibilities: a failure to correctly regulate the number of neurons produced during the neurogenesis stage of prenatal development in autism, a deficit or delay in apoptosis so that too many survive into postnatal life, or a compensatory neural genesis during perinatal or postnatal life that is triggered by adverse events such as those that ignite the neuroinflammatory reaction reported by Vargas, Nascimbene, Krishnan, Zimmerman, and Pardo (2005).

There is no experimental data as yet that could speak to the abnormal neurogenesis possibility. However, if the excess resulted from an early prenatal failure to regulate genesis, then it might be expected that this early excess would cause the brain to appear to be enlarged at birth as well as after birth, but in fact, in the majority of cases, head circumference at birth in autism is normal to slightly smaller than normal. On the other hand, if the excess in neuron numbers is regionally limited to only late maturing frontal and other higher order association cortices, then perhaps at birth its presence might not be apparent and only becomes so across the first months and years of postnatal life.

The second possibility involving abnormal deficits or delays in apoptosis is compatible with observations of abnormally increased levels of molecules that should promote apoptosis molecules in frontal and cerebellar cortices in adult autistic postmortem cases (Araghi–Niknam & Fatemi, 2003) and of an age-related decline in the magnitude of the excess numbers from childhood to adult years in autism (Schmitz, 2004).

The third possibility involving an aberrant late compensatory genesis could help explain why the brain at birth in autism is not abnormally enlarged in the great majority of cases. It is now appreciated that such a “compensatory” late genesis of neural and glial cells can occur in response to some adverse conditions (Vaccarino & Ment, 2004).

Abnormally slowed or arrested growth that follows the brief period of overgrowth in the autistic toddler could be because of a late onset of apoptosis as just discussed above, because of the conclusion of the period of late compensatory neural and glial genesis also just discussed, or because the excess of neurons and connections produces dysfunctional neural activity that causes elimination or retarded growth of ineffective neurons, axons, and dendritic and axonal arbors. Certainly there are a number of clear microstructural signs of arrest of growth, loss, and insufficient development in the older child or adult with autism, each of which could underlie the arrested or slowed growth of gray and white matter that has been observed in MRI studies of older autistic children and adults. Already mentioned were abnormal age-related declines in cerebral neuron numbers, the presence of abnormal levels of molecules that might promote apoptosis, and abnormally narrow minicolumns. Small cell size has been noted in the cingulate cortex (Kemper & Bauman, 1998) and in other frontal regions (Buxhoeveden, Semendeferi, Schenkar, Switzer, & Courchesne, 2005) in older autistic postmortem cases.

Increased numbers of cerebral neurons could underlie the volume increases in cerebral gray matter that have been reported in MRI studies. In addition, an excess of neurons likely means an excess of axons, and this could underlie volume increases in white matter that have also been reported in MRI studies of autism. These excesses mean cortical connectivity must be abnormal. Some have speculated that autism involves cerebral anatomical overconnectivity (Casanova, 2004), while others have argued in favor of cerebral functional underconnectivity (Belmonte & Yurgelun–Todd, 2003; Horwitz et al., 1988; Just et al., 2004). It has also been suggested that the key is an abnormally increased ratio of excitation to inhibition (Rubenstein & Merzenich, 2003). Here it is proposed that each of these defects underlies autism: there is an abnormal ratio of excitation to inhibition, there is anatomical overconnectivity but primarily within frontal cortical regions, and there is functional underconnectivity but principally in long-distance frontoposterior reciprocal pathways. Moreover, it is proposed that connectivity patterns and defects change with development.

Minicolumns are abnormally narrow in frontal and temporal cortex but not occipital cortex

The minicolumn is a fundamental unit of information processing. Figure 2 shows schematics of a minicolumn. The minicolumn is a roughly columnar vertical assembly of pyramidal neurons and interneurons, their interconnections, and input and output axons that extend from layer 6 up to the cortical surface. Each is 30–60 μm in diameter with perhaps 80–100 pyramidal neurons per minicolumn. In humans, minicolumns in frontal association cortex are nearly twice the diameter and several times the volume of those in primary sensory cortices such as the primary visual cortex (Buxhoeveden & Casanova, 2002). These 80–100 pyramidal neurons are thought to have originated from a common precursor cell during neurogenesis in the second and third trimesters. Thus, the number of minicolumns in cerebral cortex reflects the number of precursor cells that gave rise to each minicolumn. Expansion of cerebral cortex and increase in its processing power across evolution up to humans is due to an increase in the number of prenatally generated minicolumns. The vertical array of pyramidal cells is surrounded by neuropil space; in humans, this neuropil space that contains synapses and dendrites has expanded greatly, allowing for more complex microcircuitry and refined and powerful neural computation within each minicolumn (Buxhoeveden & Casanova, 2002). These vertical assemblies of interconnected pyramidal neurons have large inhibitory interneurons (e.g., chandelier cells) that modulate the synchronous output of clusters of minicolumns. Synchronous processing and signaling locally in minicolumns provide more powerful coherence of output to other cortical regions and cortico–cortical coupling. The function of minicolumns is to provide fine tuning of information processing and learning within a cortical column.

The minicolumn is a fundamental unit of information processing. (a) It is a roughly columnar vertical assembly of pyramidal neurons and interneurons, their interconnections, and input and output axons that extend from layer 6 up to the cortical surface. Each is 30–60 μm in diameter with perhaps 80–100 pyramidal neurons per minicolumn. (b) A schematic closeup of cortical layer 3 within a minicolumn. The shaded area represents the neuropil space that surrounds the vertical assembly of neurons. This area contains mostly synapses, dendrites, and unmyelinated axons. The unshaded section is where the majority of the cell soma reside and it contains myelinated axons. From “Morphological Differences Between Minicolumns in Human and Nonhuman Primate Cortex,” by D. P. Buxhoeveden, A. E. Switala, E. Roy, M. Litaker, and M. F. Casanova, 2001, American Journal of Physical Anthropology, 115. Copyright 2001 by American Journal of Physical Anthropology. Adapted with permission from Wiley–Liss, a subsidiary of John Wiley & Sons, Inc.

The first quantitative study of minicolumn size in autism examined layer 3 and found minicolumns to be abnormally narrow in one frontal and two temporal cortical areas (BA 9, 21, and 22; Casanova, Buxhoeveden, Switala, & Roy, 2002). In a pilot study, we aimed to determine which frontal regions have this minicolumn abnormality, and whether this abnormality is present at the time of clinical onset of autism. We measured minicolumn size in layer 3 throughout dorsal, mesial, and orbital frontal cortex in a postmortem 3-year-old autistic case; minicolumn size in primary visual cortex was measured as a contrast site (Buxhoeveden et al., 2005). To test whether there is arrest of minicolumn growth in frontal cortex in autism, we compared results from this 3-year-old to those from a postmortem 41-year-old adult with autism. We measured between 1,600 and 2,000 individual minicolumns per case, and we also measured controls. Minicolumn and neuropil sizes were significantly reduced throughout dorsal, mesial, and orbital frontal cortices in the 41-year-old adult with autism, being almost half the volume of normal adults. In addition, minicolumns and their surrounding neuropil space in the 3-year-old autistic case were nearly the same size as this 41-year-old autistic adult, which suggests the hypothesis that in autism minicolumn growth is arrested sometime during early childhood. In contrast to findings in frontal cortices, in both the 3-year-old and 41-year-old autistic cases, minicolumn size was normal in primary visual cortex. In addition, in the frontal cortex in these autistic cases, there appeared to be an excess number of neurons in each minicolumn, but this was not stereologically confirmed. The reduction in frontal minicolumn size in our small pilot study was greater than that reported for minicolumns in temporal cortex in the Casanova et al. (2002) study. Together, the Casanova study and our pilot study reveal regional differences in minicolumn size in autism, with maximum maldevelopment in the dorsal and orbital frontal cortex, somewhat less in the temporal cortex, and none detectable in the primary visual cortex.

Implications of underdeveloped minicolumns

An interesting question is whether the increase in cerebral neuron numbers indicates too many minicolumns or too many neurons per minicolumn. The number of minicolumns in cortex in autism has not been quantified. Certainly, a modest increase in the number of otherwise normal minicolumns cannot explain the emergence of autistic behavior in the first 2 years of life. On the contrary, one can imagine the opposite view: a modest increase in the number of otherwise normal minicolumns could produce a modest improvement in information processing power.

In contrast, an excess number of neurons per minicolumn would likely disrupt the normal intrinsic microcircuitry and physiological functioning within each minicolumn, especially if there is an imbalance between the number of excitatory pyramidal cells per column and the number of inhibitory cells. Because normally more neurons are generated than are later retained, an excess of neurons per minicolumn would be indicative of a failure of the normal neuron selection and pruning process of apoptosis. The retention of an abnormal number of neurons and their connections would alter minicolumn circuit design and function. Because minicolumn cell assemblies are thought to be the most basic unit of refined information processing in cortex, their defect could be part of the explanation of information processing deficiencies in autism.

A more profound disturbance of minicolumn circuit design and function would result from migration abnormality. Migration defects have been observed in several postmortem autism cases (Bailey et al., 1998), and Kennedy, Semendeferi, and Courchesne (unpublished) have also seen rafts of cells in aberrant locations and disoriented pyramidal neurons in the frontal cortex. Migration abnormality could fracture or interrupt the fundamental vertical integrative feature of the minicolumn organizational design, cause an excess of neurons in some minicolumns but reduction in others, and/or produce an excess of neurons settling in certain layers within minicolumns and a deficit in other layers. Such migration-based defects would lead to fractionated and incompletely or aberrantly formed minicolumn vertical circuitry, as well as an imbalance between excitation and inhibition within and between minicolumns. This could explain why minicolumns are “narrow” or “underdeveloped” and an excess of neuron numbers could explain why cortical gray matter volumes are larger than normal.

The presence of normal minicolumns in primary sensory cortex may signal that the autistic brain retains the capacity for detailed and refined processing of lower level visual information, and in fact, normal visual activation in autism has been found by recent fMRI studies (Hadjikhani et al., 2004). Thus, cortical areas of impaired and spared minicolumns parallel areas of impaired and spared function in autism. To our knowledge, this is the first evidence of such a regionally specific cerebral cytoarchitectonic–cerebral function parallel in autism. Major strides in understanding the fundamental neural bases of autistic behavior may ensue from combined investigations of regional differences in minicolumns, MRI volumetric growth patterns, and neurofunctional activity.

The double defect of aberrant migration and an excess of neurons due to failure of normal apoptosis would be devastating to minicolumn circuit formation and function. Although a simple increase in minicolumn numbers does not by itself explain abnormal function and behavior in the first 2 years of life in autism, dysfunctional minicolumn assembly could explain numerous findings of dysfunctional frontal activity. The underdevelopment of minicolumns in frontal cortex likely signals the failure of the normal emergence of a diversity of highly specialized vertical functional units that are necessary for the refined processing of and learning about information critical to higher order functions. Indeed, abnormalities in a variety of higher order social (Castelli et al., 2002; Pierce et al., 2004), emotional (Kennedy et al., 2004), and cognitive functions (Belmonte & Yurgelun–Todd, 2003; Just et al., 2004; Luna, Minshew, Garver, Lazar, Thulborn, Eddy, & Sweeney, 2002) have been demonstrated by neuroimaging in many studies of older children and adults with autism.

Abnormal glia activation and neuroinflammation are present in frontal lobes and cerebellum

Vargas et al. (2005) found evidence of astroglial and microglial activation and neuroinflammation in both white and gray matter samples taken from posterior cerebellar hemispheres, the middle frontal gyrus and anterior cinguate gyrus in 5- to 44-year-old autistic postmortem cases. Macrophage chemoattractant protein (MCP-1) and tumor growth factor-beta-1 (TGF-β-1) derived from glia were the most prevalent cytokines. CSF taken from living autistic children also showed a marked increase in MCP-1. In all three regions, there was enlargement of astroglial cell bodies and their processes. Microglial activation was present in the cerebellum and cerebral cortex and its underlying white matter; the cerebellum had the most pronounced microglial activation. In the cerebellum, glial activation was associated with degenerating Purkinje neurons, granule cells and axons, and in nearly all autistic brain examined, there was Purkinje and granule cell loss. Degenerating Purkinje cells were strongly immunoreactive for TGF-β-1. In the anterior cingulate and middle frontal gyrus, microglial activation was prominent at the junction of cortex and underlying white matter. There was no evidence of an adaptive immune reaction (e.g., T-cell infiltration or deposition of immunoglobulin) in the autistic brains, but evidence of deposition of complement membrane attack complexes was observed in the cerebellum, apparently associated with Purkinje cells. Interestingly, compared to the cerebellum and middle frontal gyrus, the anterior cingulate gyrus had a more complex array of increased proinflammatory and modulatory cytokines.

Implications of neuroinflammation and glial activation

Glial activation in cerebellar and frontal white matter reported by Vargas et al. (2005) might contribute to the significant increases in cerebellar and frontal lobe white matter volumes documented by in vivo MRI in 2- to 4-year-old autistic children (Carper et al., 2002; Courchesne et al., 2001), as well as to the retarded growth in white matter after this early overgrowth period. The reported glial activation involved enlargement of glial cell bodies in the autism postmortem cases (Vargas et al., 2005); although in that study the number of glial cells was not counted, it is possible that in autism there might also be an increase in glial cell numbers as part of the neuroinflammation reaction. Glial activation in white matter that increases glial cell size and number could be part of the explanation for the volume increases seen in structural MRI of young autistic children. It might also be speculated that when glial activation reaches a plateau or steady state, it would no longer drive volume increases; if so, this might explain the second phase of growth pathology in autism: abnormally slow or arrested growth.

In the anterior cingulate gyrus and middle frontal gyrus in the Vargas et al. (2005) study, microglial activation was prominent at the junction of cortex and underlying white matter. In vivo diffusion tensor imaging also reports white matter abnormality in these same two frontal zones in older autistic children (Barnea–Goraly, Kwon, Menon, Eliez, Lotspeich, & Reiss, 2004) and an in vivo MRI study reports greater volume in autism in the white matter underlying cortex, the zone of so-called radiate white matter tracts (Herbert et al., 2004). Because brain size is normal to smaller than normal at birth in autism and the abnormal acceleration of brain size begins in the first postnatal months, if glial activation does play a role in the overgrowth then it must begin to do so sometime between late in the third trimester and the first postnatal months of life.

It is uncertain how the neuroinflammation described by Vargas et al. (2005) might be related to the recent report of an increase in the number of cerebral neurons (Schmitz, 2004), but there are several possibilities. One is that the neuroinflammatory activity signals compensatory reactions that limit the process of apoptosis in cerebral cortex thus sparing too many neurons. Another is that the neuroinflammatory activity signals a late onset, compensatory proliferation of glia and cererbral cortical neurons (see review in Vaccarino & Ment, 2004). Still another is that genetic factors lead to overproduction of both neural and glial cells as in the example of the p27 gene knock-out model mentioned below.

In either event, the overproduction of neurons, and probably glial cells, may itself trigger a still later compensatory response; in the Schmitz (2004) study there was an age-related decline in the amount of excess cerebral neuron numbers; another study (Araghi–Niknam & Fatemi, 2003) found evidence in postmortem autism tissue of reduced anti-apoptotic and increased pro-apoptotic molecules in the cerebral cortex. The age-related decline in excess cerebral neuron numbers is compatible with the in vivo MRI studies of the young autistic brain reviewed earlier. Namely, by 2–4 years of age, frontal and temporal gray matter volumes are abnormally enlarged, but thereafter they fail to increase in volume, whereas normal cortex grows by about 20% in size across childhood (Carper et al., 2002; Courchesne et al., 2001).

Glial cells play key roles in brain organization during development as well as in neuroinflammatory reactions. They are involved in neural migration, minicolumn structural formation, and minicolumn function, as well as in apoptosis, including that of Purkinje cells (Marin–Teva, Dusart, Colin, Gervais, van Rooijen, & Mallat, 2004). For example, they normally provide crucial signaling to migrating neurons and growing axons, and disruption of glial development and activity would disrupt neural development and axon connectivity patterns. They are hypothesized to play an essential role in minicolumn structural and functional development (Colombo & Reisin, 2004). Disruption of this role may prevent the emergence of functionally discrete minicolumns. Glial developmental abnormalities, which apparently occur in autism, may play a part in the genesis of microstructural abnormalities in autism including migration defects, minicolumn abnormality, connectivity defects and abnormal neuron numbers. Further research is needed to determine whether the glial and molecular abnormalities described by Vargas et al. (2005) are fundamentally neuroinflammatory reactions that begin prenatal or early postnatal or reflect aberrations in genetic mechanisms that regulate the normal role of glia during neural development and organization.

Excess glial production and/or activation have the potential to produce any or all of the previously discussed microstructural findings, including frontal minicolumn abnormalities and increased neuron counts. Abnormal glial activation might also help explain some macrostructural findings in autism such as cerebellar and frontal white matter overgrowth and gray matter reduction in the cerebellum. In addition to its potential explanatory capacity for micro- and macrostructural findings, the increase in glial activity also could underlie theories of autism based on functional imaging studies, such as local overconnectivity and long-distance underconnectivity, based on the capacity of glia to alter synaptic strength (Ullian, Christopherson, & Barres, 2004). Excess glial activation in the cerebellum is also in accord with previously discussed findings of abnormal cerebellar function (Allen & Courchesne, 2003; Allen et al., 2004).