|Home | About | Journals | Submit | Contact Us | Français|
Functional magnetic resonance imaging (fMRI) is a powerful tool for studying two fundamental functional properties of the brain. First, it informs us about how the brain responds to environmental stimuli, or more simply, evoked function. As an evoked technique fMRI has confirmed long-held beliefs grounded in lesion studies: language processing is largely lateralized to the left side of the brain (Dehaene-Lambertz et al., 2006a) and face processing to the right (Haxby et al., 2000; Kanwisher et al., 1997); the amygdala is integral to processing fearful stimuli (Adolphs et al., 1995) and the hippocampus to memory retrieval (Wais, 2008) and so on. It has also brought new information such as the discovery that there are specific brain regions integral to the processing body parts (Peelen and Downing, 2005), places (Park and Chun, 2009) numerosity (Dormal et al.), and even the experience of love (Beauregard et al., 2009). Second, fMRI is also capable of studying fundamental underlying neural organizational networks, or more simply, intrinsic function. It is now widely accepted that the brain is organized via multiple networks each with its own frequency during rest, a characteristic evident even while in a coma (Boly et al., 2008). This discovery can be traced back to an fMRI study conducted by Biswal and colleagues in 1995 (Biswal et al., 1995) and pioneered thereafter by Raichle and others (Raichle et al., 2001). The so-called “resting state functional connectivity” approach has opened many opportunities for discovery because it taps into the previously unstudied intrinsic functioning of the brain that is thought to be independent of task or environment.
Autism is a developmental disorder impacting one out of every 100 children born today. It is a disorder that affects how the brain grows and works, yet the functional brain characteristics of autism during the time when symptoms first appear, namely 12-36 months, is almost completely unknown. This is because fMRI studies have been conducted almost exclusively with high functioning adolescents and adults with autism (for Review see (Minshew and Keller). The reason for this major gap in knowledge is that despite its power to map brain function, fMRI cannot be successfully used with awake, alert toddlers, whether autistic or typically developing. This is due to the strong requirement for subjects to remain still during an entire fMRI experiment, a task beyond that of infants and toddlers.
Recently, however, scientists have discovered a new way to understand brain function in infants and toddlers by examining the brain’s functional signature using fMRI during natural sleep. When considering the study of autism, this method opens considerable doors because it eliminates biases of past studies that only sampled from high-functioning, older populations. This method, heretofore referred to as “sleep fMRI”, was first used to map aspects of brain functioning in typically developing infants and toddlers in the early 2000s (Dehaene-Lambertz et al., 2002; Dehaene-Lambertz et al., 2006b). Although the brain’s signal is attenuated during natural sleep in contrast to the awake state, the distribution is surprisingly similar (Dehaene-Lambertz et al., 2002; Wilke et al., 2003).
The sleep fMRI method enables both fundamental functional properties, evoked and intrinsic, to be examined in the very young developing autistic brain. As an evoked technique, examining language processing using sleep fMRI is ideal because defects in the emergence of language are among the earliest warning signs of autism. In contrast, most typically developing infants speak their first word by 12-months (Fenson et al., 1994) and can discriminate their native language from others within the first months of life (Moon et al., 1993), suggesting that the brain is eager to process language very early in development and its signature should be clearly evident. As an intrinsic technique it can be used to study resting networks in early development in autism because it is widely assumed that connectivity is abnormal in autism (Assaf et al.; Cherkassky et al., 2006; Courchesne et al., 2007; Ebisch et al.; Kleinhans et al., 2008b; Noonan et al., 2009).
Here I describe the application of sleep fMRI as a way to study both extrinsic and intrinsic brain function in autism between 12-36 months. I consider the merits of using sleep fMRI for biomarker discovery and discuss caveats as well. While the sleep fMRI literature is small, it gains some clarity when placed side by side with the more well-understood structural brain development of autism; in particular the idea that autism is characterized by brain overgrowth during the first year or two of life (Courchesne et al., 2003; Courchesne, 2004; Courchesne et al., 2007; Dementieva et al., 2005; Elder et al., 2008). This paper concludes by considering the idea that great strides can be made, including changing patterns of functional brain activity, if only autism could be identified and treated consistently prior to the emergence of full-blown symptoms.
Babies seem to be born ready to process language. Studies have found that newborns are able to discriminate sentences in different languages (Meler et al., 1988), (Nazzi et al., 1998) and prefer to listen to their native language when given a choice (Moon et al., 1993).
Although various acoustic properties of speech are often processed bilaterally(Hickok and Poeppel, 2007) it has been hypothesized that it is the left side of the human brain that is anatomically poised to process and generate language (Dehaene-Lambertz et al., 2006a). As compared to the right, the left side of the brain has enlarged white matter underlying Heschl’s gyrus (Penhune et al., 1996), increased area measures (Lyttelton et al., 2009) larger pyramidal neurons (Hutsler, 2003) increased contact by afferent fibers (Seldon, 1981a), increased width of cortical columns (Seldon, 1981b), longer sylvian fissure (Geschwind and Levitsky, 1968) and thicker mylenated fibers (Anderson et al., 1999). Most of these studies have been conducted with adults, however, and it is unclear whether these asymmetries are the cause of language abilities in our species or only the consequence of heavy exposure to the particular acoustic properties of speech (Dehaene-Lambertz et al., 2006a; Price, 2010). However, some regions of temporal cortex, such as the left planum temporale, appear to be enlarged even in fetuses and infants (Chi et al., 1977; Galaburda and Geschwind, 1981; Witelson and Pallie, 1973).
Perhaps more provocative than structural asymmetries is a small literature suggesting that the left side of the brain is also more functionally responsive to speech and language sounds during the first months of life. For example, functional activation in response to both forward and backward speech has been shown to be significantly greater in the left planum temporale than the right in three month old typically developing infants (Dehaene-Lambertz et al., 2002). Enhanced functional responding in the left temporal lobe in contrast to the right has also been found using ERP technology wherein 4-month old infants showed larger amplitude of response in temporal cortex when exposed to syllables such as “ba” or “da” (Dehaene-Lambertz, 2000).
Dehane-Lambertz argues that the similarities between functionally immature infants and competent mature adults implies a strong genetic basis for speech processing (Dehaene-Lambertz et al., 2006a). If it is true that babies are, at some level, hard-wired for language processing and that left temporal cortex highly involved in this function, this raises important considerations for the study of autism. Is there evidence that left temporal cortex was once also poised for normal language processing in autism but became derailed across development? Or is it the case that abnormal processing of language in key regions such as the superior temporal gyrus (STG) can be found from the first months of life in autism and as such is a “core” feature of the disorder? These questions are essential to answer because they speak to biomarker discovery (can sleep fMRI results be used as putative biomarkers for autism in infants?), discerning the genes that cause the disorder (e.g., are hemisphere patterning genes involved in autism?) and may even offer insight into critical periods for early identification and treatment (at what point in development to patterns of functional activity become fixed and resistant to change?). The only way to begin to reveal the answers to these questions, however, is to probe the brain’s response to language at the earliest ages possible.
Because autism is generally not diagnosed and thus not studied until age 3 years, researchers are left with the almost impossible task of discerning early brain development prior to the onset of symptoms. The use of prospective methods that track infants from very early in development, even prior to diagnostic confirmation (because diagnosis can always be confirmed at a later age) are essential. One new method, called the 1-Year Well-Baby Check-Up Approach (Pierce et al., In Review), advocates the use of a broad band developmental screen at 12-months which detects a wide range of developmental disorders, including autism. In this way, the brain’s functional signature at the earliest ages can be examined. Almost all of the studies discussed in this paper have utilized the 1-Year Well-Baby Check-Up Approach to detect infant at-risk for autism and subsequently examine the neurofunctional profile of the disorder as young as 12-months (Pierce et al., In Review).
As with most behavioral and biological profiles of autism, what we know about language processing comes mainly from studies of adults and older children with the disorder. In typically developing adults, language processing is generally left hemisphere dominant, with the superior temporal gyrus (STG; or “Wernike’s area”) and the inferior frontal gyrus (IFG or “Broca’s area”), being strongly involved. In adults with autism, functional brain imaging studies generally report hypoactivity in frontal and temporal regions in response to language (Anderson et al., 2010; Groen et al., 2010; Just et al., 2004) and have even shown the unexpected pattern of right hemisphere dominance (Boddaert et al., 2004; Kleinhans et al., 2008a; Knaus et al.; Knaus et al., 2010; Mason et al., 2008; Muller et al., 1999; Tesink et al., 2009; Wang et al., 2006). Thus, not only is the adult autistic brain seemingly under-responsive in response to language, but also exhibits patterns of abnormal laterality.
How can we interpret the adult autism fMRI literature in terms of what it can tell us about early functional brain development in autism? The answer is that it may not be possible to directly do so. That is, while fMRI studies with adults with autism tell abundance about the mature autistic brain, such a functional signature represents a relative end state in development. This end state is necessarily impacted by a lifetime of experience that comes from living with a severe developmental disorder, which in turn impacts the establishment, refinement, and maintenance of functional connections.
In contrast to adult fMRI studies, which examine brain function after the establishment of mature functional circuitry, four new studies have utilized the sleep fMRI method to examine the brain’s response to language within the first 12-36 months of life. This method, described extensively elsewhere (Nordahl et al., 2008; Redcay et al., 2007), can be summarized by 4 simple steps. First, parents are instructed to eliminate all naps from their child’s typical routine. Second, parents are asked to keep their child awake while at home and arrive at the scanner 1-hour past their child’s normal bedtime. Third, the child either falls asleep during the car ride to the scanner or falls asleep in their parent’s arms at the scanner. Sleep onset time is documented. Fourth, in order to attempt to standardize stages of sleep during scanning, babies are placed on the scanner bed approximately 20 minutes after sleep onset. Our autism center (www.autismsandiego.org), has successfully scanned > 100 infants and toddlers with autism ranging in age between 12-48 months. What follows are four of the first studies, that include examinations of both evoked and intrinsic activity, to come out of this effort.
In 2008 Redcay & Courchesne (Redcay and Courchesne, 2008) used sleep fMRI to examine the autistic toddler’s brain response to a popular children’s bed time story written by Mem Fox called “It’s Time For Bed.” In this study functional activation patterns in 12 ASD toddlers (mean age 34 months) was compared with functional activation patterns in two groups of typically developing toddlers, those that were matched to the ASD group based on mental age (mean age 19 months), and those that were matched to the ASD group based on chronological age (mean age 36 months). In comparison to mental age matched subjects, those with an ASD showed reduced activation throughout many regions of cortex, including the left superior temporal gyrus. When compared with CA matched typical peers, however, several interesting effects demonstrating the emergence of atypical patterns of laterality emerged. First, in comparison to typical children of the same age, children with autism recruited greater right hemisphere frontal lobe activity. Second, in a paired hemisphere comparison, the ASD group showed a trend towards greater recruitment of both right frontal and right temporal regions in comparison to CA-matched toddlers. In contrast, the CA-matched typical toddlers showed a trend towards the expected greater left hemisphere dominance. Third, positive correlations with language ability and right hemisphere activity, not left, was found in the ASD group suggesting that the right hemisphere is playing a strong role in language processing in autism. See Figure 1.
The findings of Redcay and Courchesne (2008), based mainly on 3-year-olds with autism, raised an important question: Are the observed functional abnormalities, particularly the trend for reversed asymmetry, a core deficit inherent to autism, or a sequeale of impoverished receptive and expressive language use during the toddler years?
If abnormal laterality is in fact a core feature of autism, then one would expect this signature to emerge within the first year of life as the brain readies itself for the commencement of language use. To examine this possibility, in a new study (Eyler, Pierce & Courchesne, In Review), twenty toddlers with an ASD ranging in age from 14-30 months and twenty typically developing toddlers ranging in age from 13-30 months were exposed to the identical bed time story used in Redcay & Courchesne (2008). Results from this study showed that at-risk infants and toddlers later diagnosed as autistic already display the abnormal right-lateralized temporal cortex response to language, some as young as 14-months in age. This defect was enduring across development and continued to be present even in autistic 3 and 4 year olds. Chronological age-matched typically developing subjects, by contrast, displayed bilateral (but left dominant) temporal cortex language activation at 13-15 months of age that became more strongly left lateralized by 3-4 years of age. These results suggest that failure to develop normal language in autism may be due to early abnormal right-lateralization of temporal cortical specialization for processing language. This right temporal cortex activity and failure of left hemisphere specialization during language comprehension in early life may not only delay basic language acquisition in infants and toddlers with autism, but it may also “crowd out” the development of social communication abilities such as gesture and gaze following, typically mediated by the right hemisphere (Friederici and Alter, 2004). This cortical lateralization defect is early and invariant and may reflect a core, fundamental neural developmental pathology in autism. As such this defect, namely abnormal right laterality in the STG in response to language, may be the first neurofunctional biomarker of autism. See Figure 2.
Social orienting deficits in autism can be traced back to the defining paper written by Leo Kanner in 1943 (Kanner, 1943). He noted that many children failed to respond to various social signals, including their name being called, thus leading some parents to suspect their child deaf. Although a few studies have raised the possibility of peripheral hearing loss in autism (Klin, 1993), we know now that children with autism are not deaf. Instead, there appears to be a selective deficit in responding to social environmental signals that are greater than deficits in response to non-social orienting signals. For example, consider a study by Dawson and colleagues that found that children with autism more regularly oriented to non-social sounds such as to the sound of a rattle than to more social sounds such as their name being called (Dawson et al., 1995).
If the brain’s response to language in autism is abnormal by as young as 14 months in age, and likely younger, then the brain response to more emotionally communicative speech, such as speech used to orient attention socially, may be even more easily identified in babies at-risk for autism. The superior temporal sulcus (STS), which runs along the inferior aspect of the STG and separates it from the middle temporal gyrus, is a key brain region involved in processing such stimuli.
In an effort to examine functional activity in the brain in toddlers with autism using the sleep fMRI method, Pierce & colleagues (Pierce et al., In Preparation) delivered both social and non-social orienting sounds to 30 ASD and 30 typical toddlers during natural sleep. The “social” orienting sounds consisted of each child hearing their name being called embedded in a wide range of socially orienting sentences. For example, one sentence was, “Johnny, look over here!” The non-social orienting sounds consisted of environmental noises such as the sound of an airplane engine or glass shattering. Results from this study, currently in preparation, have replicated the finding of abnormal functioning in the superior temporal gyrus noted by Redcay & Courchesne (2008) and Eyler, Pierce & Courchesne (Eyler et al., In Review) and detailed analyses regarding the specific role of the STS are underway. As such, this third study further confirms the idea that STG defects are early-emerging in autism.
As pointed out by Raichle (Raichle, 2010), the human brain accounts for approximately 2% of overall body weight, yet consumes approximately 20% of available energy (Clarke and Sokoloff, 1999). Consider that at most, evoked changes in percent signal strength in brain activity are no more than 5% (and usually much less, particularly in sleep fMRI studies) in response to a wide range of tasks including those related to language processing. The remaining energy requirements at rest must therefore nourish exceptionally potent resting state networks. To date, there have been at least 8 resting state networks (RSNs) proposed including the default mode network (DMN) which links posterior cingulate regions with medial frontal and bilateral inferior parietal regions and the “core” network that links bilateral insular regions and anterior cingulate cortex and several others. In addition to the presence of large scale networks, the brain is also characterized by smaller micro-networks such as strong synchronization found across corresponding contralateral locations in many brain regions, particularly in visual (Nir et al., 2006) and right and left auditory cortex (Nir et al., 2008).
Examining the adult autistic brain, several dozen functional connectivity studies report under connectivity (see (Pierce and Eyler, In Press) for Review), which is consistent with the degeneration phase postulated by Courchesne’s three-stage brain growth model (Courchesne et al., 2007) (i.e., early brain overgrowth during the first years of life, arrest of growth during childhood and degeneration in adulthood, see Courchesne et al., this volume). Early brain overgrowth as postulated during the first phase of the model could be the result of many factors such as an increased number of neurons, synapses and/or dendritic arbors. Given these possibilities it is conceivable that the development of resting state networks might show evidence of over connectivity as has been found in DTI studies that report increased fractional anisotropy during early development (Ben Bashat et al., 2007) or even elevated levels of the default network that we have found in early pilot studies (Pierce and Eyler, In Press).
To begin to address the question of intrinsic activity in infants and toddlers at-risk for autism, a new study by Dinstein et al.,(Dinstein et al., In Review) examined the synchronization of spontaneous activity across three diagnostic groups: toddlers with an ASD, language delay (LD) or typically developing (TD). Spontaneous fMRI activity was averaged across voxels of six “seed” regions of interest that included: lateral occipital area, anterior intraparietal sulucs, motor and somatosensory cortex, superior temporal gyrus, inferior precentral gyrus and lateral prefrontal cortex. The correlation of activity between each seed ROI and every voxel in the rest of the brain was computed. Results revealed that inter-hemispheric correlations were significantly weaker in the autism group in the superior temporal gyrus (STG), the same brain region found to be abnormal in the three preceding studies. This finding of reduced inter-hemispheric correlation in the STG is illustrated by a single representative ASD and typically developing subject. See Figure 3.
The resting state study by Dinstein et al., (In Review) is the first of many we envision emerging from our dataset. Autism as a disorder of connectivity, including both resting and evoked networks, has been postulated by multiple independent groups. Recently we (Pierce & Eyler, In Press) posited the “threshold theory ” suggesting that resting functional networks are established as faulty during the early brain overgrowth period in autism, a time when possibly excess numbers of neurons contribute to excessive neural connections. Although the true function of resting networks is unknown, one plausible idea is that they work to ready the brain to react and respond to the world (Raichle, 2010). As such they work in concert with evoked networks, as is evidenced by the fact that they are often anti-correlated with other networks (e.g., activity in the default network reduces while activity in other brain regions increase in response to certain tasks). These faulty resting networks may serve to explain why fMRI studies in adults with autism almost unanimously report hypoactivity during task conditions in widespread cortical and subcortical areas including (but not limited to) the fusiform gyrus, amygdala, hippocampus, precuneus, posterior cingulate, anterior cingulate, orbitofrontal cortex, medial frontal cortex, inferior temporal gyrus, middle temporal gyrus, and superior temporal sulcus. Specifically, we hypothesize that in many cases, the stimulus condition – be it during a face, language or emotion processing task - is unable to modulate the normal balance between resting and task-activated networks. That is, if resting state networks are poorly tuned and contain high levels of neural noise, then there is a higher “threshold” or barrier, which must be passed for normalized brain function to occur. The threshold theory is strongly supported via specific examples from the face processing literature. During most face processing fMRI experiments, the subject is presented with weakly engaging “stranger” faces and hypoactivity of the face processing region, the fusiform gyrus, generally ensues (Humphreys et al., 2008; Pierce, 2001; Schultz et al., 2000). These faces, however, have limited attentional resources assigned to them by the autistic subject. In contrast, when salient, attention grabbing stimuli are used, such as mother’s face, the face processing system has been shown to become more normally activated (Pierce et al., 2004; Pierce and Redcay, 2008). In the case of face processing, salient, attention grabbing stimuli may have allowed the necessary threshold to be exceeded, thus allowing for more normalized interactions between resting and evoked networks. Thus, if the threshold theory holds true, then it may not be the case that defects in individual brain regions per se are immediately responsible for the autism phenotype. Rather, defects in coordinated switching between resting and stimulus-evoked networks are more relevant to understanding behavioral deficits in autism. Indeed a recent paper by Kennedy and Courchesne (2008) wherein autistic adults failed to show deactivation of the default mode network during a cognitive task also supports this conclusion. See Figure 4.
Collectively, the four studies described above suggest that functional defects in the STG are both early emerging and a core defect in the disorder, with abnormalities found in our studies as young as 14-months. These defects largely take the form of reduced activation in the left STG relative to controls and a proclivity for greater right hemisphere activation relative to the left, the opposite of what is found in typically developing toddlers.
Autism is a biological disorder, yet biomarkers with high sensitivity (the ability of a test or maker to detect true positives) and specificity (the ability of a test or marker to rule out true negatives) have not yet been discovered. All of the aforementioned studies, however, have made some strides in this area. In the Eyler, Pierce & Courchesne (In Review) study, conservatively, 75% of ASD toddlers displayed right hemisphere dominance patterns in response to language in comparison to only 20% of typically developing toddlers. In the Dinstein et al., (In Review) study, almost 100% of ASD toddlers could be accurately classified based in interhemispheric correlation strength. While the expense and technical expertise required to perform sleep fMRI studies limits the potential for population based screening studies to examine these findings further, we believe that there is potential for future studies to examine sleep fMRI as a follow-up examination for infants who test positive on other tests.
An important consideration of the sleep fMRI method is the issue of the impact of sleep stage on brain function. The duration of individual sleeps stages (stages I-III and REM sleep) as well as overall cycle length, changes with development. Generally, like adults, infants and toddlers take longer to enter REM during the first cycle and overall cycle length reduces across the night. For example, the average sleep cycle length in 3-5 year olds is 82 minutes during the first cycle, but reduces to 75 minutes by the 6th cycle (Montgomery-Downs et al., 2006). Furthermore, the amount of REM sleep increases with each successive sleep cycle (Montgomery-Downs et al., 2006). Regardless of age, REM onset generally does not begin until at least 60 minutes into the sleep cycle for infants, and it is for this reason that all of the aforementioned sleep fMRI studies occur during the first sleep cycle, prior to the onset of REM.
Beyond sleep cycle is the issue of normal fluctuations in physiology during sleep that may impact brain function. Sleep spindles, generated by the reticular nucleus of the thalamus, are reliably present by 3-months of age (Sankupellay et al.) and occur at a rate of approximately once per minute during the toddler years (Scholle et al., 2007). As such, an average 7-minute sleep fMRI experiment would expect the presence of approximately 7 sleep spindles. Sleep spindles are seen in the brain as a burst of activity immediately following muscle twitching and are viewed as the onset of stage II sleep, consisting of a burst of 11 to 15 hertz waves that last for .5 to 1.5 seconds (Rodenbeck et al., 2006). Also during stage II sleep is the presence of K-complexes. A K-complex is high voltage EEG activity that consists of a sharp downward component followed by a slower upward component and lasts more than .5 second, although the precise rate of K-complexes in infants is not well established.
If sleep spindles and K-complexes are present at the same rate in ASD and typically developing children, then the impact of these natural processes should balance equally between groups and pose little problem for the sleep fMRI researcher. However, children with developmental delays in general have been shown to exhibit a reduction in the number of sleep spindles, which may hold true for children with autism although this question has not specifically been addressed. Thus, the degree to which atypicalities in sleep architecture in general may impact fMRI recordings in autism is unknown.
The presence of differences in sleep architecture become less of an in issue when developing an early biomarker test for autism because all that is required is a reliable and reproducible signature of brain activity that can suggest risk status, regardless of the etiology of that signature.
While considered largely impossible just 20 years ago, the belief that autism can be greatly improved, “cured”, or is “reversible”, is a notion under serious consideration (Dawson, 2008; Helt et al., 2008). A recent review suggests that 3% to 25% of affected children can expect to achieve age appropriate functioning in cognitive and language behaviors following a period of intensive early behavioral treatment (Helt et al., 2008). The early treatment movement was spurred, in part, by the controversial study by Lovaas in 1987, that reported a 47% “recovery” rate for autistic children between the ages of 3 and 5 years following 40-hours per week of intensive behavioral therapy (Lovaas, 1987). More recently, Dawson and colleagues demonstrated that early intervention initiated by 30 months of age increased overall IQ by an average of 15 points.
The possibility for early treatment and better outcome is the direct product of early identification. In the past, treatment generally began when a child entered school, around age 5 years. The advent of early identification tools, such as the MCHAT (Robins et al., 2001) and the 1-Year Well-Baby Check-Up Approach, (Pierce et al., In Review) are facilitating treatments as young as 12-18 months in age. This time window shift, from treatment beginning at 5 years and older, to treatment beginning at 3 years and younger has the powerful advantage of moving treatment initiation into an epoch of brain development characterized by high brain plasticity. Consider, for example, that children and infants between the ages of 1 month to 3 years who undergo a hemispherectomy (due to intractable seizures) achieve largely normal social, language, and cognitive functioning (Byrne and Gates, 1987). Similarly, infants and young children with focal lesions in Broca’s area, located in the inferior frontal gyrus, go on to achieve normal language function as well (Mosch et al., 2005; Wulfeck et al., 2004). The remarkable outcome experienced by such infants and children is largely due to the timing of the neural insult, namely, one that occurred very early in development. An essential point is that the insult occurs prior to the full establishment of neural circuitry and the mastery of complex skills such as language. By contrast, a similar neural insult in older children and adults rarely results in a full recovery (Mosch et al., 2005). Plasticity, then, is very tightly linked to critical periods in neural and behavioral development. Based on his research with guinea pigs, over a century ago John Beard described a critical period as a time in which there is sudden anatomic, biochemical, and functional maturation of the brain (Beard, 1896). In humans, there is no single critical period of brain development because developmental neural events are heterosynchronous–neurogenesis, neuronal migration, axon outgrowth, dendrite and synapse formation occur in different brain systems during different times throughout pre- and postnatal development. With the exception of dentate gyrus neurons in the hippocampus (Lee and Son, 2009; Suh et al., 2007), and interneurons in the olfactory bulb, neurogenesis is largely concluded by birth in humans. In contrast, synpatogenesis begins to accelerate during the last trimester of pregnancy and peaks between 2-3 years in age in prefrontal cortex, Heschel’s gyrus and Wernicke’s area, brain regions heavily involved in language development. Synaptogenesis continues at slower rates throughout childhood and adolescence. Huttenlocher (2002) stresses that the first two years of life, a time when synpatogenesis is at its peak, may be a particularly important time period for plasticity effects, and writes: “ Environmental factors become more important during the later phases of development, especially those related to synaptogenesis (pp 36)….The development and maintenance of synapses in the cerebral cortex differs from the earlier developmental steps (e.g., neurogenesis, neuronal migration), which are largely under genetic control. In contrast, environmental factors become important in synaptogenesis, especially in the stabilization of the initial synaptic contacts. Some of these early synapses are incorporated into functioning circuits and are stabilized (persist), while others are useless and are reabsorbed. It is here that the environmental input to the cerebral cortex becomes essential for further development.”
The period following peak synaptogenesis is mirrored by an explosion of skills for the normal infant. Between the ages of 12 to 24 months, an infants’ vocabulary changes from approximately 12 to > 300 words; learning an average of 3 new words per day (Fenson et al., 1994); joint attention emerges and is mastered, wherein infants can both respond to and direct the attention of others (Tomasello et al., 2005); an understanding of objects as permanent occurs (Spelke et al., 1995) and the ability to display clear signs of empathy in response to others in distress (Spinrad and Stifter, 2006; Young et al., 1999) emerges.
A new study by Nelson and colleagues (2007) suggests that the time window for massive changes in cognitive ability in human babies may be strongest between 12-18 months. This study examined the outcomes of institutionalized infants who were placed in foster care after various periods of time. Results indicated that institutionalized infants placed into foster care by 18 months had a mean developmental quotient (DQ) of 94 based on the Mullen Scales of Early Learning when tested at 42 months in age. In contrast, infants who remained in an institution until at least 2 years and were later placed in foster care later, displayed mean DQ scores of only 80 or less when tested at 42 months.
Currently, even with the advent of early identification techniques, most children with autism do not begin treatment until well after 18 months. While the Nelson study did not examine infants at-risk for autism, findings are nonetheless provocative because they suggest a critical period for cognition in humans that begins very early in development and raises the hypothesis that treatment for autism could have an even more beneficial impact if started at or before 18 months in age.
For the child with autism, the brain undergoes a period of abnormal overgrowth in frontal and temporal regions (Carper et al., 2002) that extends across the first few years of life. Although the etiology of this overgrowth is unknown, one viable hypothesis is that it is the result of an excess number of neurons, which would in effect lead to an excess number of synapses formed. Figure 5 shows developmental mean synaptic density plots from Huttenlocher (Huttenlocher, 2002; Huttenlocher and Dabholkar, 1997) and illustrates a key point: children with autism experience abnormal brain overgrowth during the same time period in development that the brain is already experiencing a spike in synaptogenesis. It also illustrates the protracted course of development in the frontal lobes, which doesn’t peak in synapse numbers until approximately age 40 months. Early intervention, or environmental enrichment, during this time period would be essential to strengthen appropriate connections or eliminate those abnormal excess synapses that are non-functional. Left without targeted environmental input, it is easy to imagine how such a system could go awry.
Brain imaging technology offers the potential to measure changes in functional circuitry as the result of treatment. Changes in connectivity patterns have certainly been achieved following pharmacological intervention, such as increases in frontal lobe activity and concomitant increases in verbal fluency in healthy controls following administration of the performance-enhancing drug, erythropoietin (Miskowiak et al., 2008). Changes in functional connectivity as the result of strictly behavioral intervention has been largely studied with stroke patients, a research area that has shown considerable evidence for cortical reorganization and plasticity even in adults (Rossini et al., 2007). Such pre-post treatment imaging studies, however, have yet to happen in autism. Although laudable, such endeavors suffer from difficult if not impossible problems, such as challenges in controlling for cortical changes that may happen naturally during development regardless of environment, differing baseline functioning levels of participants, challenges inherent in ensuring that environmental stimulation (i.e., therapy) is equal across participants, and finally the ethics involved in having a no-treatment, or even a different treatment, contrast group.
Despite an absence of data and thorny methodological challenges, it is still essential to ask what brain changes may be occurring in response to treatment in autism, particularly those with a successful outcome? Do children change by way of compensatory mechanisms or strengthening existing but weak systems?
Our understanding of autism has indeed come far. Once viewed as a rare psychological disorder caused by poor parenting (Bettelheim, 1967), it is now considered one of the most common childhood disorders of genetic origin that brings with it varying biological signatures including a brain that grows too fast too soon (Courchesne et al., 2003; Redcay and Courchesne, 2005), genetic copy number variations (Cook et al., 1998; Fernandez et al.; Sebat et al., 2007), an amygdala that responds poorly to emotional stimuli (Baron-Cohen et al., 2000) and high levels of one of the brains most influential neurotransmitters – serotonin (Chugani et al., 1997; Chugani et al., 1999; Chugani, 2004) to name just a few. With the exception of early brain overgrowth, most of these signatures have been established by examining autistic adults and not infants and toddlers because of the relatively late age of diagnosis of the disorder. New early identification approaches, such as the 1-Year Well-Baby Check-Up Approach will help push the age of first diagnosis much lower and allow for the study and treatment of autism as young as 12-months or younger. As the future of autism early diagnosis, treatment, and etiology research moves at break neck pace, I believe that functional brain imaging, particularly the sleep fMRI method, has a place that stands firmly in the center of all of these efforts. Although the use of sleep fMRI is in its nacency, early studies suggest that a biomarker, possibly a signature of reduced STG and abnormal laterality in response to language, is no longer a distant goal. Considering trajectories of abnormal brain growth, as well as what is known about the timing of synapse proliferation and pruning, it is highly likely that early interventions will have a positive impact if they can begin prior to the establishment of mature brain circuitry. Fortunately, we now have the sleep fMRI to guide us on our way.
This work was funded by NIMH R01-MH080134 awarded to Karen Pierce and NIMH Autism Center of Excellence grant P50-MH081755 awarded to Eric Courchesne. A special thank you to Eric Courchesne for helpful comments on this manuscript.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Karen Pierce, Department of Neurosciences, School of Medicine, University of California San Diego, La Jolla, CA. Autism Center of Excellence, School of Medicine, University of California San Diego, La Jolla, CA.