Introduction
Autism is characterized by impairments in the development of reciprocal social interaction, verbal/non-verbal communication and the presence of stereotyped behavior (APA, 2000). In addition to these deficits, known as core symptoms, there are several associated features, including under-reactivity and/or over-reactivity to sensory stimuli and self-stimulatory behaviors (APA, 2000). These associated symptoms, together with the core disturbances, have provided impetus to the development of a hypothesis proposing that autism is a disorder of sensory modulation best explained in terms of brainstem pathology (Ornitz, Reference Ornitz1983). However, the data from neuropathologic and neuroimaging studies do not suggest that the pathology of autism is localized to a single area of the brain. Nevertheless, the contribution of specific key regions, such as the brainstem, to the repertoire of autistic symptomatology should be investigated.
Over the past two decades, evidence for brainstem abnormalities in autism has emerged from a variety of investigations, applying widely different methodologies. With the advent of electrophysiological techniques for audiologic and neurologic assessment, several auditory response studies were reported in autism, suggesting brainstem involvement (Klin, Reference Klin1993). Structural magnetic resonance imaging (MRI) studies examining the brainstem reported morphometric abnormalities. Earlier imaging studies assessed its size using area measurements calculated from midsagittal images. Using this method, reductions in brainstem size were reported in some (Gaffney et al. Reference Gaffney, Kuperman, Tsai and Minchin1988; Hashimoto et al. Reference Hashimoto, Tayama, Murakawa, Yoshimoto, Miyazaki, Harada and Kuroda1995; Ciesielski et al. Reference Ciesielski, Harris, Hart and Pabst1997) but not all studies (Hsu et al. Reference Hsu, Yeung-Courchesne, Courchesne and Press1991; Garber & Ritvo, Reference Garber and Ritvo1992; Kleiman et al. Reference Kleiman, Neff and Rosman1992; Piven et al. Reference Piven, Nehme, Simon, Barta, Pearlson and Folstein1992; Elia et al. Reference Elia, Ferri, Musumeci, Panerai, Bottitta and Scuderi2000). More recently, two volumetric MRI investigations observed no significant differences in the brainstem volume between individuals with autism and healthy controls (Hardan et al. Reference Hardan, Minshew, Harenski and Keshavan2001; Herbert et al. Reference Herbert, Ziegler, Deutsch, O'Brien, Lange, Bakardjiev, Hodgson, Adrien, Steele, Makris, Kennedy, Harris and Caviness2003); however, neither of these studies examined gray- and white-matter differences.
In light of the mounting evidence implicating the brainstem in the pathophysiology of autism and taking into consideration the limitations of the previous morphometric studies, the primary aim of the current study was to examine gray- and white-matter volumes of the brainstem in a sample of children with autism. In line with prior studies, we hypothesized that brainstem volume would be decreased in the autism group. A secondary aim of this study was to assess the clinical relevance of imaging findings by identifying significant relationships between brainstem structures and sensory abnormalities as measured by the Sensory Profile Questionnaire (SPQ; Dunn, Reference Dunn1999).
Method
Participants
Quantitative volumetric analysis was performed on brain MRI scans obtained from 44 boys: 22 with autism and 22 healthy controls (age range 8–12 years). The study was confined to boys because the sample size was too small to allow for the structural variability associated with gender. Subjects with autism were referred to a research clinic from the community and met the following inclusion criteria: (1) diagnosis of autism through expert clinical evaluation and two structured research diagnostic instruments, including the Autism Diagnostic Interview – Revised (ADI-R; Lord et al. Reference Lord, Rutter and LeCouteur1994) and the Autism Diagnostic Observation Schedule (ADOS; Lord et al. Reference Lord, Rutter, Goode, Heemsbergen, Jordan, Mawhood and Schopler1989); and (2) absence of medical/neurological disorders. Those with autistic disorder met both ADI-R and ADOS criteria for autism. Subjects with pervasive developmental disorder, not otherwise specified (PDD NOS), had ADOS scores ranging from 7 to 10 while meeting ADI-R criteria for autism.
Control subjects consisted of medically healthy individuals recruited from the community through advertisements in areas socio-economically comparable to those of the families of origin of the participants with autism. Control subjects had a full-scale IQ (FSIQ) ⩾70 and were screened by face-to-face interviews, questionnaires, telephone interviews and observation during psychometric tests. Individuals with a family history of any neuropsychiatric disorder (such as autism, learning disability, affective disorders and schizophrenia) were not included. Potential subjects with a history of birth asphyxia, head injury or a seizure disorder were also excluded. All control subjects had no present or lifetime history of psychiatric disorders and no learning disability as assessed by the Schedule for Affective Disorders and Schizophrenia for School-Age Children (Kaufman et al. Reference Kaufman, Birmaher, Brent, Rao, Flynn, Moreci, Williamson and Ryan1997) and the Wide Range Achievement Test-Revised (Jastak & Wilkinson, Reference Jastak and Wilkinson1985) respectively.
Evaluation of potential subjects also included obtaining a thorough history, physical examination, and laboratory testing when indicated. The Wechsler Intelligence Scale for Children was administered to measure cognitive functioning (Wechsler, Reference Wechsler1991). The Hollingshead method was used to assess socio-economic status (SES) of the family of origin of all participants (Hollingshead, Reference Hollingshead1975). The SPQ (Dunn, Reference Dunn1999) was obtained from parents for a better characterization of sensory processing dysfunction. In brief, the SPQ is a 125-item caregiver questionnaire reporting the frequency with which the child responds to various sensory experiences. The items are written such that low scores reflect greater symptom severity. For example, the frequency of a given behavior/symptom can be rated as never (5 points), seldom (4 points), occasionally (3 points), frequently (2 points), and always (1 point). Items of the SPQ generate nine summary factors: (1) sensory seeking, (2) emotionally reactive, (3) low endurance/tone, (4) oral sensory sensitivity, (5) inattention/distractibility, (6) poor registration, (7) sensory sensitivity, (8) sedentary, and (9) fine motor/perceptual.
After the procedures were fully explained, all subjects or their legal guardians provided written informed consent. Verbal assent was obtained from all subjects. The Institutional Review Board approved the methodology of the study, including MRI scanning of minors.
Measurements
MRI scans were acquired using a 1.5-T GE Signa MR Scanner (General Electric Medical Systems, Milwaukee, WI, USA). Final images for each subject were generated by obtaining T1-, T2- and PD-weighted images from all participants. The T1-weighted spoiled GRASS (SPGR) sequence was acquired using the following parameters: slice thickness=1.5 mm, slice number=124, echo time (TE)=5 ms, repetition time (TR)=24 ms, flip angle=40°, number of excitations (NEX)=2, field of view (FOV)=26 cm, matrix=256×192. Both PD- and T2-weighted images were obtained with the following parameters: slice thickness=5.0 mm, TE=96 ms for T2 and 36 ms for PD, TR=3000 ms, NEX=1, FOV=26 cm, matrix=256×192 with an echo train length=8. All images were obtained in the coronal plane. MRI data were identified by scan number alone to retain blindness of raters.
Image processing was performed on an SGI workstation (Silicon Graphics Inc., Mountain View, CA, USA) using the Brain Research: Analysis of Images, Networks, and Systems 2 (BRAINS2; University of Iowa, Iowa City, IA, USA) software package (Magnotta et al. Reference Magnotta, Harris, Andreasen, O'Leary, Yuh and Heckel2002). Six brain-limiting points (anterior, posterior, superior, inferior, left and right) were then identified to normalize the image data to the standard Talairach stereotactic three-dimensional space (Talairach & Tournoux, Reference Talairach and Tournoux1988) in which the anterior–posterior commisure line specifies the x axis, a vertical line rising from the x axis through the interhemispheric fissure specifies the y axis, and a transverse orthogonal line with respect to x and y coordinates specifies the z axis.
After fitting the image sequences to a standard three-dimensional space, the voxels representing gray matter, white matter and cerebrospinal fluid were identified using a segmentation algorithm applied to the T1-, T2- and PD-weighted image sequences as described elsewhere (White et al. Reference White, Andreasen, Nopoulos and Magnotta2003). Measurements were performed using the BRAINS2 masks as generated by a neural network and corrected by manual tracing [intra-class correlation (ICC) >0.9]. Total brain volume (TBV) was defined as the cerebrum, cerebellum and brainstem while excluding cerebrospinal fluid. The brainstem was defined as the infra-tentorial brain tissue volume superior to the foramen magnum and excluding the cerebellar volume (Magnotta et al. Reference Magnotta, Harris, Andreasen, O'Leary, Yuh and Heckel2002; Pfaendner et al. Reference Pfaendner, Reuner, Pietz, Jost, Rating, Magnotta, Mohr, Kress, Sartor and Hahnel2005).
Analysis
Statistical analyses were performed using the Statistical Package for the Social Sciences, version 16.0 (SPSS Inc., Chicago, IL, USA). Age, FSIQ, SES and all volumetric measures were compared using Student's t test with a two-tailed statistical significance level set at p<0.05.
Pearson's correlations were applied to examine potential relationships between brainstem volumes (total and also gray and white matter) with TBV and FSIQ. Analysis of covariance was applied to examine volumetric differences between the two groups while controlling for any confounding factors. SPQ summary factors were compared using Student's t test with a corrected significance level of p<0.005 (two-tailed). Partial correlations were used to examine the relationships between brainstem volumes and sensory abnormalities while controlling for subject group and FSIQ. Spearman's rank order correlation was used to examine the existence of any relationship between clinical measures as assessed by the ADI-R and brainstem volumes. This was used because the continuous nature of the scores on the ADI-R items has not been established (e.g. ADI-R domains/items are not intended for use as scales).
Results
The autism group included 18 individuals with autistic disorder and four with PDD NOS. Group-wise comparisons between the autism and control groups revealed no significant differences in age, SES and TBV; however, FSIQ was significantly higher in controls (Table 1).
Table 1. Subject characteristics
SES, Socio-economic status; TBV, total brain volume.
Correlations were observed between total brainstem volume and TBV in both groups (autism: R=0.516, p=0.014; controls: R=0.613, p=0.002) but not between total brainstem volume and FSIQ (autism: R=−0.007, p=0.977; controls: R=−0.326, p=0.139) (Table 2). Reductions in brainstem volumes were observed in the autism group when compared to controls (Table 3). Decreased gray-matter volume was observed before and after controlling for TBV. Differences in total brainstem volumes reached statistical significance only after controlling for TBV.
Table 2. Correlations between total brainstem volume and Full-scale IQ as well as total brain volume (TBV) in the autism and control groups
* Correlation is significant at the 0.05 level (two-tailed).
** Correlation is significant at the 0.01 level (two-tailed).
Table 3. Gray- and white-matter brainstem volumes
TBV, Total brain volume; df, degrees of freedom; s.d., standard deviation.
The SPQ findings are summarized in Table 4. Data from four participants were not available. All summary factors were significantly lower in subjects with autism indicating more abnormalities in this group. The relationships between brainstem volumes and sensory measures were examined. When controlling for subject groups, an association was observed between brainstem gray matter and oral sensory sensitivity factor (R=0.458, df=37, p=0.003) (Fig. 1). This relationship remained unchanged when controlling for subject group and FSIQ (R=0.443, df=36, p=0.005). No correlations were observed between brainstem white-matter volume and any sensory factors. Finally, no associations were found between any of the three main domains of the ADI-R (qualitative abnormalities in reciprocal social interaction; qualitative abnormalities in communication; and restricted, repetitive and stereotyped patterns of behavior) and brainstem volumes.
Fig. 1. Relationship between brainstem gray-matter volume and oral sensory sensitivity in the autism group as measured by the Sensory Profile Questionnaire (SPQ) (r=0.377).
Table 4. Sensory Profile Questionnaire (SPQ) findings in children with autism and controls
df, Degrees of freedom; s.d., standard deviation.
Lower scores reflect greater symptom severity.
Discussion
This study provides evidence of brainstem gray-matter volumetric reductions in a group of male children with autism and points to the existence of a relationship between this structure and sensory abnormalities. These observations are consistent with several previous neuroimaging studies that examined this structure in autism, but used only the midsagittal slice to measure the size of the brainstem and its three subdivisions: midbrain, pons and medulla. Hashimoto et al. (Reference Hashimoto, Tayama, Murakawa, Yoshimoto, Miyazaki, Harada and Kuroda1995) reported significant hypoplasia of brainstem structures using a large sample of patients with autism (102 cases). An earlier MRI study reported that the entire brainstem and pons were significantly smaller in a group of autistic youth when compared to controls (Gaffney et al. Reference Gaffney, Kuperman, Tsai and Minchin1988). A more recent investigation observed a reduction in the size of the pons but did not report on the total size of the brainstem (Ciesielski et al. Reference Ciesielski, Harris, Hart and Pabst1997). By contrast, several studies have failed to report findings similar to that of the current investigation; however, these investigations limited their measurements to subdivisions of the brainstem (i.e. pons only or pons and midbrain), which could account for the differences in findings (Hsu et al. Reference Hsu, Yeung-Courchesne, Courchesne and Press1991; Garber & Ritvo, Reference Garber and Ritvo1992; Kleiman et al. Reference Kleiman, Neff and Rosman1992; Piven et al. Reference Piven, Nehme, Simon, Barta, Pearlson and Folstein1992; Elia et al. Reference Elia, Ferri, Musumeci, Panerai, Bottitta and Scuderi2000).
Although many previously published studies have used the midsagittal slice to measure brainstem size, two recent reports have in fact examined the volume of this structure (Hardan et al. Reference Hardan, Minshew, Harenski and Keshavan2001; Herbert et al. Reference Herbert, Ziegler, Deutsch, O'Brien, Lange, Bakardjiev, Hodgson, Adrien, Steele, Makris, Kennedy, Harris and Caviness2003). These investigations did not detect any volumetric alterations of the brainstem in autism, but differences in sample characteristics and study methodology may explain the inconsistent findings. Hardan et al. (Reference Hardan, Minshew, Harenski and Keshavan2001) reported on a sample of adolescents and adults with autism who had a mean age of 22 years, versus 10.5 years in the current study. In addition, volumetric measurements were performed manually and no gray-/white-matter segmentation was conducted. In the more recent study by Herbert et al. (Reference Herbert, Ziegler, Deutsch, O'Brien, Lange, Bakardjiev, Hodgson, Adrien, Steele, Makris, Kennedy, Harris and Caviness2003), neuroanatomic segmentation was applied on scans obtained from a sample of 17 boys with autism (age range 7–11 years) and 15 age- and gender-matched controls. No volumetric alterations were detected between the two groups and gray/white segmentation could not be performed reliably (Herbert et al. Reference Herbert, Ziegler, Deutsch, O'Brien, Lange, Bakardjiev, Hodgson, Adrien, Steele, Makris, Kennedy, Harris and Caviness2003). In light of these inconsistencies, additional research is warranted to determine whether brainstem volumetric alterations exist in autism.
The role of the brainstem in the pathophysiology of autism has long been suspected. Specific clinical features such as under/over-reactivity to sensory stimuli, self-stimulatory behaviors (APA, 2000) and even deficits in social interactions and communication have supported a hypothesis linking autism to dysfunctional brainstem sensory modulation (Ornitz, Reference Ornitz1983). The diversity of deficits possibly resulting from brainstem abnormalities is consistent with its complex neuroanatomy; it is the source of various neurotransmitter systems thought to be implicated in autism, including serotonergic, cholinergic and gamma-aminobutyric acid (GABA)ergic systems (Pickett & London, Reference Pickett and London2005). Moreover, the brainstem houses the nuclei of the vagus nerve, which has extensive enervation to many parts of the body such as the heart, gastrointestinal system and larynx/mouth (Nolte, Reference Nolte2002). This nerve controls varied tasks such as heart rate, gastrointestinal peristalsis and several muscle movements in the mouth, including speech and eating. Impairments in all these domains have been reported in autism (Volkmar et al. Reference Volkmar, Cohen, Paul and Klin2005). However, the most direct evidence of brainstem abnormalities comes from post-mortem investigations that have reported abnormal inferior olives (Bauman & Kemper, Reference Bauman and Kemper1985; Bailey et al. Reference Bailey, Luthert, Dean, Harding, Janota, Montgomery, Rutter and Lantos1998), reduction of neurons in the facial nucleus, shortened brainstem and absent superior olive (Rodier et al. Reference Rodier, Ingram, Tisdale, Nelson and Romano1996). Moreover, Bauman & Kemper (Reference Bauman and Kemper1985) reported that neurons appeared larger in youth with autism and smaller in adults with the disorder. Additional post-mortem investigations reported on a variety of histological abnormalities in the brainstem, including enlarged arcuate nuclei in the medulla (Bailey et al. Reference Bailey, Luthert, Dean, Harding, Janota, Montgomery, Rutter and Lantos1998) and a major reduction of neuron numbers in the facial nuclei (Rodier et al. Reference Rodier, Ingram, Tisdale, Nelson and Romano1996).
Support for brainstem abnormalities as contributors to the pathophysiology of autism also comes from the fields of teratology and embryology. Prenatal exposures to medications have also suggested the existence of a relationship between brainstem abnormalities and autism. There is an early period in embryonic development (<6 weeks) when the forebrain is absent and brainstem and cranial nerve nuclei are developing; this results in a vulnerable period when chemical contact could be neurotoxic (Rodier et al. Reference Rodier, Ingram, Tisdale, Nelson and Romano1996). For instance, exposure to thalidomide in utero (20–24 days of gestation) has been associated with the development of autistic features (Stromland et al. Reference Stromland, Nordin, Miller, Akerstrom and Gillberg1994). A report of misoprotol exposure during the sixth gestational week has also reportedly led to an increased risk of autism in children with Möbius sequence (Miller & Ventura, Reference Miller and Ventura2001; Bandim et al. Reference Bandim, Ventura, Miller, Almeida and Costa2003). This link between medication exposure early in the pregnancy and autism has led to speculation that brainstem injury sustained in utero plays a role in the development of autism (Rodier, Reference Rodier2002).
Decreased gray-matter volume in the brainstem observed in the current study provides indirect support for the potential contribution of altered cortico-cerebellar and brainstem–cerebellum networks to the pathophysiology of autism. Higher cognitive functions rely on communications between the cerebral cortex, brainstem and cerebellum; and it has been hypothesized that some core symptoms of autism are related to the disconnection between these brain regions (Skoyles, Reference Skoyles2002). Additionally, specific behavioral abnormalities reported in autism could be related to brainstem anomalies. Studies comparing eye-blink conditioning (with individuals with autism developing the conditioned response more rapidly than controls) are suggestive for alterations of the brainstem–cerebellum loops (Sears et al. Reference Sears, Finn and Steinmetz1994). Although other behaviors such as visual orientation and facial expression have been linked to brainstem alterations, no associations were observed in the current study between these clinical features and structural findings. These observations could be related to the limitation of the current morphometric software, including the inability to generate midbrain, pons and medulla measurements, or to the contribution of other brain regions to gaze and facial abnormalities.
Findings reported in this study must be interpreted in the context of several methodological limitations. The two groups were not matched on FSIQ, with the autism group having lower scores than controls. The relationship of cognitive functioning and brain structures in autism is complex and it remains unclear whether matching for IQ is necessary (Jarrold & Brock, Reference Jarrold and Brock2004). Limited evidence is available to support a relationship between cognitive functioning and brain structures in autism, as suggested by the existence of increased brain size in individuals with low IQ (Cody et al. Reference Cody, Pelphrey and Piven2002). Additionally, the difficulty with IQ-matching is that the cognitive profiles in individuals with autism are markedly uneven (Joseph et al. Reference Joseph, Tager-Flusberg and Lord2002). Therefore, matching groups on FSIQ may ensure that only the group averages are not significantly different, but matching on all subsets is clearly impractical (Hobson, Reference Hobson1991; Happe, Reference Happe1994; Joseph et al. Reference Joseph, Tager-Flusberg and Lord2002; Jarrold & Brock, Reference Jarrold and Brock2004).
In addition to the above limitations, the autism sample was relatively heterogeneous, consisting of a majority of children with autistic disorder and a minority with PDD NOS. Including these individuals in one autism group assumes that these disorders have a common neurobiological underpinning, an assertion that currently lacks strong scientific support as PDD NOS continues to be considered a separate disorder according to DSM-IV (APA, 2000). In addition, the accuracy and validity of gray- and white-matter segmentation of the brainstem have not been confirmed by post-mortem studies and the image analysis procedure used to measure these structures was semi-automated, and was not confirmed by traditional manual tracing methods. Finally, this study did not separate the brainstem into individual components (i.e. midbrain, pons and medulla), which limits more specific localization of gray-matter reductions and does not allow direct comparison with previous studies that have examined brainstem subdivisions.
Although there are several studies examining brainstem volume in autism, results have been inconsistent and no studies have measured gray- and white-matter volumes separately. The findings from this study are suggestive of brainstem abnormalities in autism involving gray-matter structures and are supported by prior neuropathologic investigations. This reduction may possibly point to disconnectivity between the brainstem and cerebrum and cerebellum. However, in light of the limitations of this investigation, additional studies are needed before any conclusions can be drawn about gray-matter volumes of the brainstem in autism. Cross-sectional and longitudinal studies are warranted in large samples of individuals with autism and a wide range of intellectual abilities to examine the three subregions of the brainstem using well-validated segmentation techniques.
Acknowledgments
This work was supported by NIMH grant MH 64027 (to A.Y.H.). We gratefully acknowledge the effort and commitment of the participants and their families in this study.
Declaration of Interest
None.
