Partial Agenesis and Hypoplasia of the Corpus Callosum in Idiopathic Autism

Abstract

To test the hypothesis that developmental anomalies of the corpus callosum (CC), contribute to the pathogenesis of autism, we characterized the type, topography, and severity of CC pathology corresponding to reduced CC areas that are detected by magnetic resonance imaging in the brains of 11 individuals with autism and 11 controls. In the brains of 3 autistic subjects, partial CC agenesis resulted in complete or partial lack of interhemispheric axonal connections in CC segments III–V. In these cases, a combination of focal agenesis and uniform axonal deficit caused reduction of CC areas by 37%, of axon numbers by 62%, and of the numerical density of axons by 39%. In the CC of 8 autistic subjects without agenesis, there was an 18% deficit of the midsagittal CC area, 48.4% deficit of axon numbers, and 37% reduction of the numerical density of axons. The significantly thinner CC, reduced CC area, and uniform axonal deficit in all autistic subjects were classified as CC hypoplasia. Thus, the byproduct of partial CC agenesis and hypoplasia is reduction of axonal connections between cortical areas known to be involved in behavioral alterations observed in people with autism.

INTRODUCTION

The autism spectrum disorder is a developmental disability characterized by deficits in social communication and social interactions and restricted, repetitive patterns of behavior, interests, and activities (1). In autism spectrum disorder, a lack of coherence (2), deficits of complex information processing (3), and theory of mind (4) coexist with intense and narrowly focused interests and the tendency to concentrate on systems that operate deterministically and repetitively, such as computers, games, or machines (5, 6). These functional anomalies are considered to reflect defects in connectivity (7–9), specifically with long-range underconnectivity and short-range overconnectivity (6, 10–13). The hypothesis of short-range connectivity anomalies is supported by postmortem morphological and morphometric studies of the brains of autistic subjects (13), but morphological characterization of the type, topography, and severity of structural changes that contribute to long-range underconnectivity in autism is still missing. This postmortem study of the corpus callosum (CC) in the brains of individuals diagnosed with autism was designed to characterize the type, topography, and severity of CC pathology corresponding to reduced CC midsagittal areas detected by magnetic resonance imaging (MRI) in autistic patients and to test the hypothesis that autism is associated with reduced number of interhemispheric axonal connections between cortical regions known to be involved in functional alterations observed in autism.

The CC is the largest white matter tract in the human brain (14, 15). Brain tractography based on diffusion tensor imaging reveals long-range axonal interhemispheric connections that transfer cognitive, sensory, and motor information and shows the distribution of cortical neurons that form the CC and the site where their axons cross the CC midline (16). One of the most consistent findings in MRI studies of individuals with autism is reduction or thinning of the CC (7, 8, 17–24). Chung et al reported lower white matter density, which is considered to be an index of altered neuronal connectivity in the CC genu, rostrum, and splenium; they concluded that the reduction may correspond to impaired interhemispheric connectivity of the frontal, temporal, and occipital cortices in autistic subjects (25). Preservation of a larger midsagittal area of the CC accompanies faster signal processing, higher intelligence, and reduced severity of autism behaviors (26).

The Association Between CC Agenesis and Behavioral Alterations Typical for Autism

CC agenesis or hypogenesis has been reported in more than 50 human congenital syndromes that are associated with other brain malformations (27). In the United States, callosal developmental anomalies have been reported in 0.7%–5.3% of individuals. A study of an unselected population estimated that the prevalence of CC agenesis is 1:1000 (28). Complete or partial CC agenesis is frequently associated with intellectual deficits and a wide range of cognitive, behavioral, and neurological consequences (9, 29). Many individuals with CC agenesis carry the diagnosis of cognitive impairments, attention deficit, and/or autism spectrum disorder (19, 30–32). Some subjects born with CC agenesis are considered asymptomatic, but detailed psychometric testing reveals social deficits, intellectual difficulties, problems with visual processing and higher-order language functions (19, 33).

In approximately 40% of children with CC agenesis, unusual social interactions were reported (30), but the most common social abnormalities are emotional immaturity, lack of introspection, impaired social competence, deficits in social judgment and planning, and poor communication of emotions (19, 34). Some differences have been reported between autistic groups with and without CC agenesis in the social domain, including onset of social deficits at the age of 2–3 years in autistic subjects and at the age of 6 years in children with CC agenesis (35). Moreover, repetitive and restricted behaviors are less common in children with CC agenesis than in children with autism (30).

The correlation between the area of the CC and the number of axons detected in postmortem studies in the control group (15) suggests that the reduced CC area detected by MRI in autistic subjects may correspond to an axonal deficit in the midline CC. The aims were to (1) establish the type, prevalence, and severity of CC developmental anomalies and (2) determine the pattern of developmental alterations in 5 CC segments with known cortical connections and function.

MATERIALS AND METHODS

Patients

The CC was examined in 11 brain hemispheres of 4- to 60-year-old subjects diagnosed with idiopathic autism, including 8 males and 3 females, and 11 age-matched (4–52 years old) control subjects, including 7 males and 4 females (Table).

TABLE.

Preserved Tissue and Clinical and Postmortem Records

Case ID	Sex	Age (y)	Intellectual deficit	Seizures age of onset	Cause of death	PMI (h)	BW (g)	H	Weight loss (%)
A1	M	4	—	—	Drowning	30	1280	R	49
A2	F	5	—	—	Traumatic multiple injuries	13	1390	R	52
A3	M	8	—	ab EEG	Rhabdomysarcoma	22	1570	R	45
A4	F	11	Mild	4.5 m	Seizure-related drowning	13	1460	L	52
A5	M	13	Severe	2 y	Seizure-related	8	1470	L	39
A6	F	21	Moderate	5 y	Seizure-related respiratory failure	50	1108	R	43
A7	M	22	Moderate	15y	Seizure-related	25	1375	R	38
A8	M	23	Severe	23 y	Seizure-related r respiratory failure	14	1610	R	60
A9	M	36	Severe	—	Cardiac arrest	24	1480	R	44
A10	M	52	—	—	Heart failure	11	1324	L	53
A11	M	60	Moderate	3 y	Pancreatic cancer	26	1210	R	38
Mean		23				21	1388		48
C1	F	4	—	—	Acute bronchopneumonia	17	1530	R	49
C2	F	4	—	—	Lymphocytic myocarditis	21	1222	R	43
C3	M	14	—	—	Electrocution	20	1464	R	44
C4	F	15	—	—	Traumatic multiple injuries	9	1250	R	49
C5	F	20	—	—	Traumatic multiple injuries	9	1340	R	52
C6	M	23	—	—	Ruptured spleen	6	1520	R	41
C7	M	29	—	—	Traumatic multiple injuries	13	1514	R	49
C8	M	32	—	—	Asphyxia	24	1364	R	42
C9	M	48	—	—	Heart atherosclerosis	24	1412	L	39
C10	M	51	—	—	Myocardial infarct	18	1450	L	43
C11	M	52	—	—	Heart atherosclerosis	13	1430	R	48
Mean		26				16	1408		45

PMI, postmortem interval; h, hours; BW, brain weight; H, hemisphere; R, right; L, left; weight loss, decrease (%) of hemispheric brain sample weight during dehydration in ethyl alcohol; ab EEG, abnormal EEG; y, years; m, months; M, male; F, female.

The diagnosis of autism was confirmed with the Autism Diagnostic Interview-Revised (36). Intellectual disability ranging from mild to severe was diagnosed with the Wechsler Intelligence Scale for Children III and the Woodcock–Johnson Tests of Achievement-Revised in 8 subjects. Six of the 11 autistic subjects were diagnosed with seizures and 1 with abnormal EEG activity requiring Depakote treatment. Death was seizure-related in 5 cases.

Hemispheric brain samples were obtained from the Harvard Brain Tissue Resource Center, Belmont, MA (supported in part by PHS grant number R24-MH 068855), the Brain and Tissue Bank for Developmental Disorders of the National Institute of Child Health and Human Development at the University at Maryland, and the New York State Brain and Tissue Bank at the Institute for Basic Research in Developmental Disabilities (IBRDD), Staten Island, NY. Between 2003 and 2013, tissue and clinical record acquisition was coordinated by the Autism Tissue Program (Director: Jane Pickett, PhD) of Autism Speaks. Tissue was processed, cut, and stained at IBRDD and was shared with partnering institutions consistently with Autism Speaks/Autism Tissue Program recommendations.

Tissue Preservation for Neuropathologic and Morphometric Studies

The difference between postmortem interval in the autistic (21 hours) and control (16 hours) group and the difference between the average weight of the brain in the autistic group (1388 g) and control group (1,408 g) was insignificant. The brain hemispheres were fixed in 10% buffered formalin for an average of 443 days in the control group (range of 95–1819 days) and 1043 days in the autistic group (range of 75–4560 days). Neither the average length of time of brain hemisphere dehydration in ethyl alcohol (36 days in the autistic group and 38 days in the control group) nor the reduction of brain hemisphere weight during dehydration (48% in the autistic group and 45% in the control group) was significantly different. Brain hemispheres were embedded in 8% celloidin (37) and cut into serial 200-μm-thick sections.

For neuropathological examination, approximately 120 equidistant coronal serial sections per case were stained with Cresyl violet and mounted in Acrytol. The study revealed focal developmental abnormalities in neurogenesis, migration, and dysplastic changes in the brains of 92% of the autistic subjects as a marker of fetal multiregional defects of brain development (38).

Tissue Sampling and Staining

From each twelfth serial hemispheric section, a 200-µm-thick, 2-mm-wide strip of CC was removed. These samples were impregnated with 2% osmium tetroxide for 75 minutes to visualize myelin sheaths. After 40 minutes of dehydration in ethyl alcohol with increasing concentrations from 70% to 100% and a 2 × 15–minute final dehydration with propylene oxide (RT20401; Electron Microscopy Services [EMS], Hartfield, PA), tissue samples were infiltrated and embedded using an EMBed-812 (EMS, 14120) kit. The sample was positioned in the mold to preserve section orientation and to cut axons perpendicularly in the CC midline. Blocks were polymerized in an oven at 60° C for 24 hours. One-µm-thick sections were cut using a Reichert Ultracut S ultramicrotome equipped with an 8-mm diamond knife (Diatome-U.S., Fort Washington, PA). Tissue strip length usually ranged from 3 mm to 12 mm and required a modified elongated platinum loop to pick up and transfer individual sections to histological slides. After drying at 50 °C, sections were stained for 30 minutes with a solution of 2% p-phenylenediamine ([PPD]; P6001, Sigma-Aldrich, Co., St. Louis, MO) in 50% ethyl alcohol (15). The PPD solution was stirred for 3 hours, stored for more than 5 days, and filtered with a 0.2-µm filter before use. Stained samples were washed several times to remove residues of PPD precipitates, dried in 60 °C for ∼30 minutes to increase adhesion to the histological slide, and mounted with Acrytol (Leica Biosystems, Richmond, IL).

Stereological and Statistical Analyses

The length and midsagittal area of the CC (mm²) were determined on the basis of measurements of serial sections. The total number (millions) and the numerical density (number per mm²) of myelinated axons were estimated in 5 CC segments and in the entire CC. Morphometric measurements were executed without knowledge of each subject’s age, gender, clinical diagnosis, or neuropathological status. Axonal morphometry was performed using a workstation consisting of an Axiophot II (Carl Zeiss, Gottingen, Germany) light microscope with Plan Apo objectives 2.5× (N.A. 0.075) for tracing the contour of each section and 100× (N.A. 1.3) for counting myelinated axons; a specimen stage with a 3-axis computer-controlled stepping motor system (Ludl Electronics; Hawthorne, NY); and a CCD color video camera (CX9000; MicroBrightField Bioscience, Inc., Williston, VT).

Partitioning based on diffusion tensor imaging and fiber tractography distinguishes 5 segments in the human CC. They correspond to the position of cortical neurons and their projections to the CC and homotopic and heterotopic areas in the other hemisphere. The length of segments I, 17%; II, 33%; III, 17%; IV, 8%; and V, 25% was calculated using Hofer and Frahm segmentation to ensure comparability of callosal sectors in control and autistic subjects (16) (Fig. 1). A systemic sampling scheme was achieved with Stereo Investigator (MicroBrightField Bioscience, Inc.). On average, 21 equidistant sections were examined in the CC (every twelfth 200-µm-thick section, with a 2.4-mm interval). The applied sampling resulted in a proportional representation for each of the 5 segments of the CC. The size of the sampling grid (130 × 130 µm) was selected after tests to reduce the Schaeffer coefficient of error to 0.05 or less. The total number of counting frames was on average ∼224 per case, and the number of counting frames per CC subregion was proportional to the midsagittal area of each of the 5 CC regions. Application of a 5 ×5–µm counting frame with a 25-µm² test area resulted in counting approximately 790 axons per case. This overrepresentation for the entire CC was necessary to maintain a sufficient representation for individual CC regions, including the smallest, region IV, which occupies only 8% of the linear rostrocaudal CC extent.

FIGURE 1.

Corpus callosum (CC) partition and sampling. (A) Partition of the CC into 5 segments was based on diffusion tensor imaging and fiber tractography identifying cortical projection areas (16). (B–D) To estimate the total number of myelinated axons, a 2-mm-wide strip of the CC was cut off from every twelfth 200-µm-thick coronal hemispheric section embedded in Epon (B), and cut in the midline plane into 1-µm-thick sections (C), and stained with PPD (D). The sampling grid (130 ×130 µm) and the counting frame (5 × 5 µm) were applied (D). (E) Axonal profiles in the square and on green lines of the frame were included in counts and on red lines were excluded.

Comparisons of the number of axons, the cross-sectional area, and the density of neurons of the CC and its subdivisions were performed using one-way ANOVA. Post hoc comparisons were performed using Scheffé's method.

RESULTS

Partial CC Agenesis and Other Developmental Anomalies in Autistic Subjects

In 3 autistic subjects, sectioning of the brain in the midline revealed defects of interhemispheric connectivity with a several-mm-wide gap in the CC midline and some inter-individual differences in the topography and morphology of these developmental defects (Fig. 2, left panels). A medial view of the hemisphere shows a 2- to 4-mm-deep gap in the isthmus and anterior and posterior CC part adjacent to the isthmus in the first case (5 years old); a much deeper gap (5–6 mm) in the isthmus area as well as the genu in the second case (8 years old); a very narrow gap in the isthmus and anterior splenium area; and a similar narrow gap just behind the genu CC in the third case (11 years old).

FIGURE 2.

Corpus callosum (CC) partial agenesis combined with hypoplasia. (A–C) In the brains of 5-, 8-, and 11-year-old children diagnosed with autism, focal lack of interhemispheric connectivity in the brain midline was detected macroscopically in the isthmus and in the adjacent anterior and posterior part of the CC (arrows). Additional fetal developmental defects were found, including fragmentation of the genu (arrow) in the 8-year-old subject, and another narrow gap just behind the genu (arrow) in the 11-year-old subject. Morphometry (right panels) confirmed a consistent focal lack of axons in segments III, IV, and V in the 3 subjects, and additional gaps in segments I and II in the 11-year-old subject. Moreover, the average density of axons/mm² in the preserved parts of the CC (∼116 000) was significantly less than in control subjects (∼197 ,000).

Morphometric estimates of the numerical density and numbers of axons confirmed topographic selectivity in the focal lack of axons and revealed a significant deficit of axons in areas free of partial agenesis in all 3 cases with agenesis (Fig. 2, right panels). In the 5-year-old, there was total lack of midline connectivity in the entire segment IV, the caudal part of segment III, and the anterior part of segment V, resulting in a 23-mm-long gap. In the 8-year-old, there was lack of axons similarly affected segments III–V, resulting in a 21-mm-long gap in midline connectivity. In the 11-year-old, a lack of axons was found not only in segments III–V but also in a portion of segments I and II (21-mm-long gap in interhemispheric connectivity). The average numerical density of axons in preserved parts of the CC was only 116 682/mm² (41% deficit when compared to control group), whereas the estimated total number of axons was 34.0 × 10⁶ (62% deficit).

Examination of serial Cresyl violet-stained coronal sections revealed that a common feature in areas of agenesis was a several mm gap between truncated arms of the left and right part of the CC (Fig. 3). Serial sections from both hemispheres of a 5-year-old subject illustrate asymmetry of the truncated CC with a rather thin shortened left part pointing downward and a short thick right part pointing upward. Disruption also included the hippocampal commissure resulting in only partial contribution to the splenial CC region. In the 8-year-old, the truncated part of the CC was only 2- to 3-mm thick, and thinning and distortion of the shape was also detected in the hippocampal commissure. Morphology of the CC in 11-year-old subject suggested that axons approached the midline but that their expansion was arrested a few mm from the midline.

FIGURE 3.

Morphological inter-individual differences of corpus callosum (CC) focal agenesis. Examination of Cresyl violet-stained serial sections from both hemispheres of the 5-year-old autistic subject revealed asymmetry with a very thin left arm of the CC and short and thick Probst-like fibers in the right CC, with ingrowth of the gyrus cinguli in a distorted CC (shown in section number 343), and truncated hippocampal commissure that also does not reach the midline. In the 8-year-old subject, the right CC is thin and ends several mm from the midline. In the 11-year-old child, serial sections reveal that the central portion of the CC ends 1–2 mm from the midline, whereas the dorsal and ventral parts of the CC form thin bridges between hemispheres detectable in some sections. Serial sections from the brain of a 7-year-old control subject show normal morphology of the CC and hippocampal commissure. L, left hemisphere; R, right hemisphere. The numbers of the serial sections are shown at the bottom of each image.

Developmental abnormalities in the CC coexisted with other types of developmental alterations including dysplastic changes and defects of neuronal migration (Fig. 4). In the 5-year-old girl, 2 types of fetal developmental defects were closely related to CC focal agenesis. Ingrowth of the dysplastic gyrus cinguli in the dorsal region of the CC resulted in deformation of the CC and trajectories of truncated CC axons, whereas subcortical ectopia distorted the gyrus cinguli white matter structure a few millimeters from the CC. In the 8-year-old boy, there were 3 other types of developmental abnormalities: temporal cortex dysplasia with focal loss of vertical and horizontal cortex organization and large acellular areas, dysplasia in the hippocampus with disorganization of laminar organization of the CA1 sector, and cortical dysplasia in the cerebellar vermis. In the 11-year-old girl, there were 2 types of developmental defects found: dentate gyrus dysplasia and a large heterotopia of cortical cells in the cerebellar white matter.

FIGURE 4.

Multiregional fetal origin anomalies of brain development. In all 3 subjects with autism, corpus callosum (CC) partial agenesis coexisted with other fetal developmental defects detected in sections stained with Cresyl violet. In the 5-year-old child, ingrowth of the cingulate gyrus (GC) into the CC caused a distortion of the trajectory of axons in the truncated CC (A), whereas ectopia [E] caused focal distortion of subcortical axonal connections in this gyrus (B). In the 8-year-old boy, there was dysplasia [D] in the CA1 sector (C), temporal cortex (D), and cerebellar vermis (E). The 11-year-old child was affected with dysplasia in the granule cell layer in the dentate gyrus (DG) (F) and heterotopia [H] in the cerebellar white matter (G).

CC Hypoplasia in Autistic Subjects

A thin CC, reduced midsagittal area, and axonal deficit indicate CC hypoplasia (19, 39). The average CC thickness, estimated by averaging measurements of 21 equidistant serial CC sections, was 7.18 mm (SE  =  0.27) in the control group and 6.11 mm (SE  =  0.25) in autistic subjects; CC thickness was 15% less in autistic subjects (p  <  0.0072). Postmortem estimates of the area of the CC cut in the midline revealed that in control subjects, the average area was 437  ±  21.2 mm², whereas in autistic subjects with partial agenesis, the area was less by 37% (276  ±  27.0 mm²; p  <  0.004), and in autistic subjects without partial agenesis, the CC area was 16% less (368 ±  20.6 mm²; p  <  0.095) (Fig. 5).

FIGURE 5.

Corpus callosum (CC) midsagittal area, total number and numerical density of axons. Significant reduction of the CC midsagittal area, and significant deficit in the total number of axons (millions) and numerical density of axons (number/mm²) in 3 autistic subjects with partial CC agenesis (A-Ag) and 8 autistic subjects without partial agenesis (A) compared to 11 control subjects (C) (*p  <  0.05; **p  <  0.01; ***p  <  0.001; + 0.095).

The average total number of axons in the CC of 11 control subjects was estimated to be 89.7 (± 3.41) × 10⁶. The average number of axons in 8 autistic subjects without CC agenesis was 46% less (48.4  ±  4.69 × 10⁶; p < 0.001), whereas in autistic individuals with partial CC agenesis, the number of axons was less by 62% (34.0  ±  2.38 × 10⁶; p  <  0.001). The difference among autistic subjects in the total number of axons ranged from 27.5 to 68.2 × 10⁶; the number was always below the range of inter-individual differences in the control group (76.4 to 109.9 × 10⁶). In the control group, the total number of axons (N/mm²) in the CC correlated with the CC area (mm²) (p  <  0.0224; n  =  11 cases). A similar correlation between the CC area and the number of axons was found in the autistic group without agenesis (p  <  0.0014; n  =  8 cases).

The numerical density of axons in control subjects was 207 631  ±  7513. There were lower numerical densities in subjects with partial CC agenesis (126 401  ±  18 381; p  <  0.001) and without agenesis (129 834  ±  6524; p  <  0.001), reflecting a similar 39% and 37% axonal deficit, respectively.

An average of 21 examined equidistant serial sections per case represented the entire rostrocaudal extent of the CC and 5 CC segments. The average numerical density of axons in 11 control subjects and 8 autistic subjects without agenesis revealed a uniform deficit of axons in the autistic group when compared to control subjects (Fig. 6); however, in the control group, the regression line showed a decrease in the numerical density of axons from the highest in the genu of the CC to lower density in the middle and caudal portion of the CC. The opposite trend was found in the autism group, with the lowest density in the anterior part, and the highest in the splenium. The difference between regression lines in autistic and control subjects was significant (t  =  2.68; p  <  0.011).

FIGURE 6.

Difference between numerical density of axons in rostrocaudal extension of the CC in autism and control groups. The graph demonstrates a reduced average numerical density of axons in 21 equidistant serial sections representing the rostrocaudal extension of the CC in 8 autistic subjects without partial CC agenesis (A) compared to 11 control (C) subjects. Regression lines reveal a significant difference in the decrease of the axonal density in the control group and in the increase in the autistic group (t  =  2.68; p  <  0.011).

Agenesis was very topographically selective. To test whether the axonal deficit was also topographically selective, the total number of axons and numerical density were compared in all 5 segments between autistic subjects without agenesis and control subjects. The deficit between control and autistic subjects was significant in all segments. A comparable deficit in the total number of axons in segments I–V (by 51%, 43%, 58%, 47%, and 47%, respectively) and in the numerical density of axons (by 41%, 34%, 36%, 36%, and 33%, respectively) (Fig. 7) indicates that in autistic subjects, the axonal deficit affects the entire rostrocaudal extension of the CC and is similar in all 5 CC anatomical subdivisions connecting the prefrontal (I), premotor and supplementary motor (II), motor (III), sensory (IV), and parietal, occipital, and temporal cortex (V).

FIGURE 7.

Distribution of axonal deficits in 5 corpus callosum (CC) segments. In comparison to control subjects (C), the total number of axons (millions) and the numerical density of axons (n/mm²) are significantly less in all 5 CC segments in autistic subjects with partial agenesis (A-Ag) as well as in autistic subjects without partial agenesis (A). *p  <  0.05; **p  <  0.01; ***p  <  0.001.

DISCUSSION

Types of CC Developmental Anomalies in Autistic Subjects

Among 11 autistic subjects, 3 matched the diagnostic criteria for congenital partial agenesis of the CC. In 8 autistic subjects, a fully formed but thinner CC matched the criteria for diagnosis of CC hypoplasia (19). The prenatal or postnatal origin of CC developmental anomalies is of critical importance for determination of the etiology and potential prevention of harmful factors. In a study by Glass et al, classification of callosal anomalies, including CC hypoplasia in 63% of 630 subjects as congenital pathology, was supported by diagnosis within 10 days after birth or in the first year of life (39). However, in 4- to 60-year-old autistic subjects examined postmortem, the distinction between fetal CC hypoplasia and postnatal CC thinning mimicking congenital hypoplasia is limited. The development of the CC is regulated by a number of factors and is vulnerable to the environment and insult. The size of the CC decreases in male primates that were isolated during early development (40). The possibility that harmful factors may alter the size of the human CC in postnatal life is supported by reports showing reduction of CC area in neglected children diagnosed with posttraumatic stress disorder (PTSD). The rostrocaudal CC myelination sequence suggests varying regional vulnerabilities to insult (41). The difference between the topographically selective 15% to 18% reduction in CC area in regions 1, 4, 5, and 7 in PTSD subjects (42), but rather uniform deficits of the CC area and of the total number and numerical density of axons in all CC segments detected in this study of autistic subjects may reflect a different pathogenesis and potentially fetal origin of CC hypoplasia in autistic individuals.

The mean area of the CC in autistic subjects with partial agenesis was 37% less than that in controls (p  <  0.004); in subjects without agenesis it was 16% less (368  ±  20.6 mm²; p  <  0.095). Results of this postmortem study are consistent with clinical MRI estimates demonstrating reduced CC area in autistic patients (7, 8, 17–24, 43), including high-functioning individuals (44).

The correlation between CC area and number of axons detected in the control group (p  <  0.0224; n  =  11 cases) is similar to the findings of Riise and Pakkenberg in their control group studies (15). Detection of a significant correlation between the reduced CC area and the reduced number of axons in the autistic group (p  <  0.0014; n  =  8) examined postmortem suggests that clinical MRI-based deficits of the CC area in autistic subjects are predictors of axonal deficit whose magnitude may correspond to CC hypoplasia.

Clinical MRI studies revealed not only a reduction of CC area on medial sagittal slices but also a reduced CC volume (21, 23). This observation may match postmortem images of serial coronal sections showing thin CC not only in the midline but also 10–15 mm from the midline.

A meta-analysis of data from 10 studies that characterized 253 patients with autism revealed a CC reduction in size, with the largest reduction in the rostral portion containing axons of pre-/supplementary motor neurons and a less prominent reduction in the caudal direction (43). These clinical data correspond to a significant deficit in the total number of axons and numerical density detected in the anterior part of the CC in the present study. This pattern suggests more severe deficits in interhemispheric connections of the frontal, prefrontal, premotor, and motor cortex in people with autism.

Number of Axons in Typically Developing Subjects

Postmortem studies provide information beyond the range of MRI resolution, including estimates of the number of axons and their numerical density in the CC. The estimated numbers of axons reported in control groups vary across a broad range. Our results are close to those published by Highley et al, who examined Palmgren silver-stained nerve fibers in the CC of 15 males with an average age of 67 years (14). In our control group, the average estimated total number of axons was 89.7 × 10⁶, whereas according to the estimates of Highley et al, the human CC comprises 84 × 10⁶ axons. The numerical density of axons was also similar, with 210 000 in the genu (segment I in our study) and 199 000 in the Highley et al report. Differences between published estimates of the number of axons are likely due to the effects of application of different methods of tissue staining and morphometric evaluation and inter-individual differences. Counting of myelinated and unmyelinated axons detected with Holmes (neurofibrils) and Loyez (myelin) stains revealed approximately 200 × 10⁶ axons in the human CC (45). Using PPD staining, Riise and Pakkenberg estimated that 39- to 60-year-old men have on average 138 × 10⁶ myelinated fibers, with inter-individual differences ranging from 104 to 195 × 10⁶ (15). In our control group of 4- to 52-year-old subjects, inter-individual differences ranged from 76 to 110 × 10⁶. However, the average fiber density was relatively similar in the Riise and Pakkenberg study (216 000/mm²) and Highley et al study (163 000–209 000/mm²), and in our control group, where the density ranged from 189 000/mm² to 209 000/mm² in different CC segments.

Pathogenesis of Reduced Size of the CC and Reduced Total Number and Numerical Density of Axons in the CC of Subjects Diagnosed With Autism

CC anomalies have been reported in some autistic individuals (17, 19, 32, 46–8), but the origin of these anomalies is not clear. This study reveals that the anomalies are a combination of partial agenesis, with a strong preference to segment IV, the posterior part of segment III, and the anterior part of segment V, and axonal deficits detected as reduced number and numerical density of axons in all 5 segments. In the control group, the midsagittal CC area correlates with the number of fibers (15, 45). One may assume that the smaller size of the CC and axonal deficit in autistic subjects might be a result of (1) midline axonal misguidance, (2) reduced number of cortical neurons and proportional reduction in the number of axons sent to the other hemisphere, (3) reduced ability of small neurons to produce long-range interhemispheric axons, and (4) modification of the number of callosal axons by noxious factors during the extremely dynamic process of prenatal increase and postnatal reduction of axon number, or a combination of these mechanisms.

Partial agenesis and uniform CC axonal deficit are most pronounced in the midline or close to the midline suggesting that during CC formation, axons are not able to cross the midline and therefore terminate several mm before the midline. The 10- to 15-mm-long and only 2- to 3-mm-thick arms of a truncated CC suggest that access of the majority of axons to the midline zone is blocked not only 3–4 mm but also ∼10–15 mm from the midline. CC agenesis is usually associated with alterations of its midline guidepost cells, including the glial wedge, indusium griseum, midline zipper glia, and subcallosal sling cells (19). Both the indusium griseum and glial wedge cells express Slit2, a potent chemorepellant that restricts callosal axon penetration through the midline (49–51), whereas subcallosal sling cells and midline zipper glia provide a substratum over which CC pioneer axons extend (50, 52). The arrest of axons close to the midline in areas of partial agenesis and in the hypoplastic CC detected in autism might be an effect of downregulation of midline guidepost promoters of axonal migration and/or activation of cells producing chemorepellants. In the fetal brain, pioneer gyrus cinguli axons expressing the guidance receptor neuropilin 1 cross the midline first and provide a path for axons arriving from other neocortical regions (53). The ingrowth of the gyrus cinguli into the CC detected in a 5-year-old autistic subject and the focal distortion of the trajectory of CC axons appears to be a remnant of a congenital anomaly contributing to misguidance of interhemispheric axonal connections and partial CC agenesis.

The hypothesis that reduced volume of the CC is a direct result of interhemispheric commissural pyramidal neuron deficit or loss is consistent with local neuronal deficits in the superior frontal cortex (54), Brodmann areas 44 and 45 (55), fronto-insular cortex (56), visual cortex, and the entire cerebral cortex (57), but is in conflict with a regional 67% increase in the number of neurons in the prefrontal cortex reported in 2- to 16-year-old individuals with autism (58).

The possibility that in autism, smaller neurons may produce fewer long-range connections, including interhemispheric callosal axons, might be explained by a significantly smaller neuronal body in 4- to 8-year-old autistic children in 13 of the examined subcortical structures, archicortex, cerebellum, and brainstem (59), superior and middle frontal gyrus (60), inferior frontal cortex (55), anterior mid-cingulate (61), and cingulate cortex (62), and in other brain structures (63).

The risk of fetal and postnatal distortion of CC development appears to be very high because of the extremely dynamic process of CC growth during fetal life, with an increase in the number of axons by 1.8 × 10⁶ per day in the last 100 days of rhesus fetal life, and CC reorganization, with the loss of 4.4 ×10⁶ axons per day during the first 3 weeks of postnatal life (64). One may speculate that even a short exposure of the human fetus or infant to noxious factors may permanently distort CC structure and function and contribute to the autism phenotype. However, it is still not known whether defects in axonal pruning affect CC size and contribute to callosal hypoplasia in humans (19).

Study Limitations

The results of clinical studies (which are often based on brain MRI) of hundreds of individuals with autism and typically developing control subjects, may provide partially different results than postmortem studies, which are usually limited to much fewer brains of subjects diagnosed with autism and a similar number of age-matched control cases. One of the causes of these differences might be a higher mortality rate in the population diagnosed with autism, epilepsy, and intellectual disabilities than in the general population (65), resulting in higher prevalence and more severe developmental anomalies in postmortem than in clinical studies. The presence of partial agenesis in 3 of 10 examined subjects, as well as ectopias and dysplastic changes in other brain regions, appears to be a reflection of this higher prevalence of multiregional and diverse developmental anomalies detected in the postmortem-examined group (38, 66, 67). The concentration of focal agenesis in the youngest subjects (5, 8, and 11 years of age), but its absence in examined postmortem adults might be an effect of a small group or may indicate that CC partial agenesis is a sign of a more complex and severe developmental pathology that increases the risk of death, especially related to epilepsy. Another factor contributing to the lower prevalence of CC developmental anomalies in clinical studies might be the limited resolution of routine clinical imaging in comparison to macroscopic and microscopic examination of hundreds of serial 200-µm-thick sections in this postmortem study.

Closing Remarks

Our studies and those of others of developmental anomalies of the brains of autistic subjects and this first postmortem study of the CC suggest that the detected developmental pathology of the largest brain commissure is part of multiregional developmental defects. The markers of CC maldevelopment are partial agenesis and hypoplasia resulting in an approximately 40% axonal deficit. These data support and expand clinical MRI studies demonstrating reduction of CC area on medial sagittal slices and reduced CC volume (21, 23). A significant deficit of CC axons in all 11 examined postmortem cases matches the conclusion from MRI studies that reduction of the CC is the most consistent finding in autism. The association of a larger CC with higher intelligence and reduced severity of autism clinical manifestations (26) suggests a link between postmortem detected deficits of interhemispheric axonal connections and functional deficits in autistic subjects.

Financial support: This study was supported in part by funds from the New York State Office for People With Developmental Disabilities, a grant from the Department of Defense Autism Spectrum Disorders Research Program (AS073234, Jerzy Wegiel, Thomas Wisniewski) and a grant from Autism Speaks (Princeton, NJ; Jerzy Wegiel).

Conflict of interest. The authors have no duality or conflicts of interest to declare.