|Home | About | Journals | Submit | Contact Us | Français|
Individuals with autism spectrum disorders often exhibit atypical language patterns including delay of speech onset, literal speech interpretation, and poor recognition of social and emotional cues in speech. We acquired fMRI images during an auditory language task to evaluate for systematic differences in language network activation between control and high-functioning autistic populations.
41 right-handed male subjects (26 high-functioning autistic, 15 control) were studied using an auditory phrase recognition task, and areas of differential activation between groups were identified. Hand preference, verbal IQ, age, and language function testing were included as covariables in the analysis.
Control and autistic subjects showed similar language activation networks, with two notable differences. Control subjects showed significantly increased activation in the left posterior insula compared to autistic subjects (p<0.05, FDR), and autistic subjects showed increased bilaterality of receptive language compared to control subjects. Higher receptive language score on standardized testing was associated with greater activation of the posterior aspect of left Wernicke’s area. Higher verbal IQ was associated with greater activation of bilateral Broca’s area and involvement of prefrontal cortex and lateral premotor cortex.
Control subjects showed greater activation of the posterior insula during receptive language, which may correlate with impaired emotive processing of language in autism. Autism subjects showed greater bilateral activation of receptive language areas that was out of proportion to differences in hand preference in autism and control populations.
Language disturbances are among the most pronounced and clinically significant features and strongest predictors of outcome in autism(1, 2). Auditory and language disturbances in autism include delayed onset of speech(3), widely varying impairments in comprehension and spoken language, stereotypical or idiosyncratic speech patterns(4), and hypersensitivity to distracting or unexpected auditory stimuli(5). Though much less studied, language impairment in autism may also involve reading ability(6). Qualitatively similar, though milder, delays in speech, spoken language, and reading are also found in family members of children with autism(7). The neural basis of language-related impairments in autism is critical to understanding brain mechanisms driving the clinical impairments, developing interventions to improve function and prognosis of affected individuals, and identifying genes and other risk factors involved.
There is converging evidence from multiple modalities that language in autistic subjects shows atypical hemispheric lateralization in the brain. It has long been established that metrics of hand preference show decreased right-hand dominance in autistic populations(8–10). Dichotic listening tasks show reduced right-ear advantage for speech processing in autism(11). SPECT and PET imaging shows decreased relative cerebral blood flow lateralization in language-related cortex for autistic subjects(12, 13). An EEG examination found reduced left lateralization of temporal lobe mu rhythms in an autistic population(14). Anomalous age-related changes in lateralization of language areas during later childhood were found for autistic subjects in an MEG study(15). Although abnormal volumetric asymmetry of frontal but not superior temporal language areas have been consistently found in autism (16–18), functional asymmetry of the superior temporal gyrus has been observed repeatedly (15, 19–26).
Lateralization of language has been examined directly using fMRI. In an expressive language letter fluency task, 14 high-functioning adolescent and adult males with ASD were found to have less left-lateralized activation in frontal language areas than 14 controls (22). A separate study using a semantically-based response naming task demonstrated a larger difference in percent signal change between Broca’s area and its right homolog in 12 controls compared to 12 high-functioning males with ASD(27).
We investigated receptive language processing using an auditory language task, with a paradigm that examined phrase- and sentence-levels of language processing in order to determine whether differences in activation outside core receptive and expressive language regions were seen between autism and typically developing groups.
Twenty-six high-functioning males with autism were compared to15 male normal volunteer subjects, group matched by age. Table 1 compares group demographics of age, handedness, receptive language function, verbal IQ, and performance IQ of the autism and control populations. There was a slight trend toward decreased right-handedness in the autism group that was not statistically significant in our sample. One autism participant was left-handed and two controls were ambidextrous. The participants had no history of hearing problems and all had English as their first language. All controls had normal language function. As expected, language function was impaired in the autism participants as a group. Verbal and performance IQ scores showed small, but significant, decreases in the autism group. Experimental procedures were approved by University of Utah Institutional Review Board. Informed consent was obtained for all subjects.
Diagnosis of autism was established by the Autism Diagnostic Interview-Revised (ADI-R) (28), Autism Diagnostic Observation Schedule-Generic (ADOS-G) (29), DSM-IV (30) and ICD-10 criteria, under the direction of a board-certified child psychiatrist. Participants were excluded if medical causes of autism were identified by participant history, Fragile-X gene testing, karyotype, and observation.
Control participants underwent tests of IQ, language function and standardized psychiatric assessments (31) and were assessed with the ADOS-G (29) to confirm typical development. Controls with any history of developmental, learning, cognitive, neurological, or neuropsychiatric conditions were excluded.
The Edinburgh Handedness Inventory (32), a standardized assessment of hand preference, was obtained for each subject. This inventory consists of a numerical score between −100 and 100, where −100 represents strong left-handedness and 100 represents strong right-handedness.
Verbal IQ (vIQ) and performance IQ (pIQ) were measured with the Wechsler Adult Intelligence Scale (WAIS III) or Wechsler Abbreviated Scale of Intelligence (WASI).
Clinical Evaluation of Language Fundamentals- Third Edition (CELF-3) (33), was used to assess language skills. It is a comprehensive and nationally-normalized clinical assessment tool that provides a quantitative measure of language level. The CELF-3 includes subtests that measure grammar, syntax, semantics, and working memory for language and provides an overall assessment of higher-order receptive and expressive language and total language level. We used the Receptive subtest score as a covariable in our analysis because our fMRI task was primarily designed to measure receptive language function.
Images were acquired on Siemens 3 Tesla Trio scanner. All fMRI subjects were fitted with MRI-compatible lenses to allow comfortable reading of 8-point text within the scanner. Subject alertness was monitored throughout examination by real-time eye tracking using infrared camera mounted on a 12-channel head coil (Siemens).
The scanning protocol consisted of initial 1 mm isotropic MPRAGE acquisition for an anatomic template. BOLD echoplanar images (TR= 2.0 s, TE = 28 ms, GRAPPA parallel acquisition with acceleration factor = 2, 40 slices at 3 mm slice thickness) were obtained during auditory language task described below. Prospective motion correction was performed during BOLD imaging with PACE sequence.
We chose phrase and sentence-level tasks rather than single-word tasks because the former are known to produce less variable and more sensitive activation maps compared to lexical-level tasks(34, 35). Auditory language task consisted of a block design alternating 20 seconds of auditory stimuli with 20 seconds of no stimuli. Auditory stimuli were delivered to sound-blocking pneumatic headphones (Avotec, Silent Scan SS-3100), and consisted of phrases that described a common word. Examples of phrases include “Jewelry we wear around our neck,” “The funny guys at the circus,” and “Water falling from the sky.” Subjects were instructed to think in their mind a word that each phrase describes. Six phrases were presented during each block, with 1 second pause between phrases for subjects to think of an appropriate word. During epochs with no stimuli, subjects were instructed to rest but keep their eyes open. A single run of the task was performed, of 4 minute duration, in all of the subjects.
The visual Language task consisted of a similar 20 second block design. Visual stimuli were displayed via LCD projector onto a screen in the bore of the scanner, and viewed by a mirror mounted on the top of the 12-channel head coil. Stimuli consisted of sentences with a blank at the end. Subjects were instructed to think in their mind of an appropriate word to complete the sentence. Example sentences include “She put the dishes in the _____,” and “He took a shortcut to go _____.” 8 sentences were presented per 20-second block. During epochs without sentences, subjects were instructed to fixate on a high contrast mark in the center of the screen. A single run of the task was performed, of 4 minute duration, in 14 autism and 9 control subjects.
Offline post-processing was performed in Matlab using SPM8b software. Field map sequence was acquired for each subject for distortion correction, and all images were motion corrected using realign and unwarp procedure. No difference in head motion was seen between autism and control groups during retrospective motion correction. Using maximal detected motion in x, y, and z directions by retrospective motion correction algorithm, we computed sqrt(x2+y2 +z2) for each subject as an index of head motion. For the autism group, this index measured 0.8 mm +/− 0.32 mm s.d. (range: 0.32 – 1.3 mm). For the control group, maximal head motion measured 0.71 mm +/− 0.31 mm (range: 0.24 – 1.2 mm). Findings were not significantly different using two-tailed t-test (p=0.35). BOLD images were coregistered to MPRAGE anatomic image sequence for each subject. All images were normalized to MNI template brain (T1.nii in SPM8b) and smoothing with 8 mm kernel was performed on all images.
Activation maps were generated for each subject using general linear model to obtain T-contrast image. No minimal cluster size was specified in any of the analyses. Second level random effects analysis was then performed for autism and control subjects on T-contrast images for each subject. Age, handedness, receptive language ability (CELF-3), and verbal IQ were included as covariables in the second-level analysis. Activation maps were obtained separately for the control group and for the autism group, each thresholded at an acceptable false discovery rate (FDR) of p<0.05 (Figure 1).
Visual sentence completion task was analyzed using a similar procedure and auditory > visual contrast across all subjects was thresholded at an acceptable FDR p<0.05 to identify brain regions active preferentially during the auditory task, which is shown for typically developing controls in Figure 2, blue. This auditory > visual contrast yielded activation maps of bilateral superior temporal gyri in expected locations of primary auditory cortex. To identify receptive language clusters in each subject, this auditory > visual contrast was used as a mask to exclude voxels in primary auditory cortex from the analysis. Maximal clusters were selected bilaterally of activated voxels posterior to primary auditory cortex in superior temporal, middle temporal, supramarginal and angular gyrus regions using p<0.05, FDR, as threshold for each subject. The number of activated voxels in left and right receptive language clusters was used to calculate fMRI laterality index using the formula (left − right)/(left+right) to obtain a measurement between −1 (strongly right dominant) and 1 (strongly left dominant).
Activation maps for control and autism populations during the auditory language task is shown in Figure 1. All subjects showed activated clusters in expected language regions (Wernicke’s area, Broca’s area, lateral premotor cortex, supplementary motor area, left dorsolateral prefrontal cortex), when thresholded at acceptable FDR p < 0.05. Population activation maps showed remarkable similarity in spatial distribution of activation, with similar T-scores and spatial coordinates of local maxima in core language regions (Table 2). A few notable differences are seen. There is a focus of activation in the left posterior insula much greater in the control map (Figure 1, white arrow), and activation in the right Wernicke’s homolog extends further posterior in the autism map than in the control map (Figure 1, black arrows). A trend towards increased activation in right lateral premotor cortex in the autism map (Figure 1, black arrow) was not statistically significant in our sample. We detect no significant difference in magnitude of auditory stimulus responses in primary auditory cortical regions between groups. Cerebellar activation was also noted for most subjects, but the infratentorial brain was not consistently included in the field of view for all subjects and does not appear on activation maps.
The visual task showed similar spatial distribution of language activation, with additional posterior parietal, lateral temporo-occipital, and visual cortical areas. The difference between auditory and visual task activation, thresholded at FDR p<0.05, is shown in blue in Figure 2, rendered on MNI template brain image, with remaining auditory task activation from all subjects shown in red. The auditory task showed significant increases in activation over the visual task in bilateral superior temporal gyrus corresponding to primary auditory cortex, with additional activation of left greater than right posterior insula. No significant posterior insular activation was seen during visual task.
To compare autism and control populations, a second-level, 2-sample t-test design was performed and control > autism and autism > control contrasts were evaluated. Figure 3 illustrates control > autism contrast, shown for acceptable FDR p<0.05. Activity was seen for this contrast only in the left posterior insula, with peak activity at MNI coordinates x= −39, y= −28, z= 16, and T-score of 5.7, corresponding to FDR corrected p-value of 0.0087. No other areas of significantly differential activation were seen.
Autism > control contrast demonstrated a cluster in right posterior middle temporal gyrus, along the posterior inferior aspect of Wernicke’s homolog at MNI coordinates x= 48, y= −55, z= 4, with T-score 4.1. To further evaluate this region, a small volume correction was performed by limiting evaluation to right Wernicke’s homolog, given our a priori hypothesis that autistic subjects would show increased right-hemispheric activation of language regions. We limited evaluation by searching within a 2 cm diameter sphere that enclosed all significantly activated voxels in right Wernicke’s homolog region on the group activation map obtained from all subjects. With this correction, this cluster was significant at p=0.004, FDR. No other foci of significantly different activation were seen for autism > control contrast.
To test for significant differences in language lateralization between autism and control samples, we identified receptive language clusters in the left and right hemisphere for each subject. Receptive language regions were selected because our task was designed primarily for receptive language activation.
Left and right hemispheric clusters in Wernicke’s region and homolog were identified for each subject after masking out the primary auditory cortex using auditory > visual contrast. Laterality indices are shown in Figure 4, with population statistics listed in Table 1. There was significantly greater relative activation of right-sided language regions in the autism group that was out of proportion to slight differences in hand preference as measured by Edinburgh Handedness Inventory, indicating that language lateralization is not merely a consequence of known population biases towards decreased right-handedness in autism. Moreover, these differences in laterality were observed in data for which variance associated with hand preference was already included as a regressor in the group-level analysis.
Four covariables were included in the regression when evaluating group-level activation: age, handedness, verbal IQ, and CELF-3 receptive language function score. No significant foci of differential activation were associated with handedness. Increased proficiency on receptive language testing (CELF-3) was associated with increased activation in the posterior left Wernicke’s area (Figure 5A, red), significant at p<0.05, FDR. Increased verbal IQ scores were associated with a trend toward activation in bilateral posterior inferior frontal gyrus (left Broca’s area and right hemispheric homolog), right putamen, left dorsomedial prefrontal cortex, left lateral premotor cortex, and right anterior temporal pole (Figure 5A, blue). These clusters were all significant at p<0.001, uncorrected, but did not survive FDR correction at p=0.05. MNI coordinates of activation associated with CELF-3 and verbal IQ tests are listed in Table 3. With younger age, activation was seen in the region of bilateral nucleus accumbens, with peak activity at x= −3, y= −1, z= −5, T-score = 4.8, corresponding to acceptable FDR p-value of 0.031. This cluster is shown in Figure 5B. No significant differential activation was associated with older age, lower verbal IQ score, or lower CELF-3 score.
We find significant decrease in activation in the posterior insula during auditory language processing in autism compared to typically developing participants. This area showed striking activation during the language task for controls, with a T-score second only to primary auditory cortex, greater than scores for either Wernicke’s area or Broca’s area on the left. The posterior insula was activated only during our auditory and not our visual language task, which suggests an auditory processing function to this region. This idea is supported by a recent study where electrophysiological recordings in rhesus monkeys (which also show close proximity of posterior insula to Heschl’s gyrus) demonstrated selective responses in posterior insular neurons to sounds associated with vocal communication(36).
A longitudinal study in children ages 5–11 (younger than participants in our study) showed decreased left posterior insular activity with age in virtually identical coordinates to the focus we detected(37), a finding that may suggest an even more important role for this region in language development. Posterior insula has also been implicated in other features relevant to the autism phenotype, including emotive processing of stimuli such as experiencing pain(38), processing negative emotions such as disgust(39), and emotional responses to aversive facial stimuli(40).
We found similar activation magnitude in Broca’s area and in its right homolog in the autism and control groups. All autism studies measuring lateralization during language tasks using the fMRI BOLD response or blood flow velocity have found left lateralization in activation of Broca’s area compared to its right homolog in autism, although the degree of lateralization and activation in autism relative to controls has differed across studies(22, 27, 41, 42).
In contrast to frontal language regions, the pattern of activation in Wernicke’s area and its right homolog differed in our autism and typically developing control samples. Although the amount of activation in Wernicke’s area in the left hemisphere was similar in autism and control participants, activation in the right homolog was more extensive in autism, as measured both by population level activity in the right homolog, although this difference was small, as well as by population statistics on laterality index that showed greater right-sided spatial extent of activation. This difference in language laterality does not merely reflect the known differences in hand preference in our sample, as a quantitative metric of hand preference showed a slight, nonsignificant trend towards decreased right-handedness, while the language laterality observed with fMRI laterality index was significant and much more pronounced.
We noted several associations with standardized testing and our fMRI results. First, participants with higher scores for receptive language proficiency, as measured by CELF-3, showed increased spatial extent of left Wernicke’s area, with recruitment of more of the posterior middle temporal gyrus. This seems plausible given the known dominant receptive language function of left Wernicke’s area. Second, participants with higher verbal IQ scores demonstrated greater recruitment of frontal language areas, including left Broca’s area and its right-sided homolog, as well as dorsomedial prefrontal cortex and lateral premotor cortex. This finding is consistent with a more complex frontal language network in higher-IQ participants.
Finally, we saw activation of bilateral nucleus accumbens in younger participants. This area has been associated with neural processing of rewards(43) including novel stimuli and natural rewards(44). It is possible that younger subjects experience a greater sense of intrinsic reward when “solving” the auditory phrase task compared to older subjects.
We include a higher total number of autism participants in the study to better account for the known clinical heterogeneity of autism, but acknowledge that the observed findings may not be uniformly generalized to represent differing clinical autism phenotypes. We also elected not to have the subjects speak during the task to prevent head motion, which limited our ability to compare subject performance on the task between the groups. We observed very similar activation patterns, however, in the vast majority of activated regions, suggesting that language network activation was similarly achieved in the two populations. Finally, the language stimuli used were short phrases that lack the context of normal conversation, and may not generalize to conversational language.
High-functioning autistic subjects showed significantly decreased involvement of the left posterior insula during auditory language processing, a finding that may correlate with impaired perception of emotive content in language. We also find increased activation in the right hemispheric Wernicke’s homolog, consistent with prior reports of atypical functional lateralization in expressive language regions. These findings suggest targets for ongoing investigations of the neurophysiological mechanism of language abnormalities in autism.
The project described was supported by Grant Numbers RO1 MH080826 (JEL) from the National Institute of Mental Health and R01NS37483 (NL) from the National Institute of Neurological Disorders and Stroke and by the Ben B. and Iris M. Margolis Foundation (JSA).
The project described was supported by Grant Number RO1 MH080826 (JEL, ALA, NL, EDB) from the National Institutes of Health, an Autism Speaks Predoctoral Fellowship grant (JEL for MBD), and a University of Utah Multidisciplinary Research Seed Grant (JSA, JEL). Additional support came from NINDS R01 NS34783 (NL), NIMH P50 MH60450 (NL), and the Ben B. and Iris M. Margolis Foundation (JSA). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute of Mental Health, NINDS, or the National Institutes of Health. The authors appreciate the assistance of Dr. Jim Lee, Melody Johnson, and Henry Buswell, of the University of Utah Center for Advanced Imaging Research, for technical assistance in data acquisition and paradigm design. They also thank Barbara Young and Celeste Knoles of the Utah Autism Neuroscience Program, and express their sincere gratitude to the young people and their families who participated in the study.
Data have been presented in part at 2009 International Meeting for Autism Research (IMFAR).