Toward a unified framework for understanding the various symptoms and etiology of autism and Williams syndrome

Authors

  • Toshio Inui

    Corresponding author
    • Kyoto University
    Search for more papers by this author
    • The author wishes to thank Drs. Kenji Ogawa, Chiyoko Nagai, and Chisato Yoshida for their helpful comments.

Correspondence concerning this article should be sent to: Toshio Inui, Department of Intelligence Science and Technology, Graduate School of Informatics, Kyoto University, Yoshida-honmachi, Sakyo-ku, Kyoto 606-8501, Japan. (E-mail: inui@i.kyoto-u.ac.jp)

Abstract

To date, the unifying pathogenesis, or etiology, of autism spectrum disorders (ASDs) and Williams syndrome (WS) remains unknown, partly because of the broad variation of phenotypes and the heterogeneity of syndrome expression. In particular, in order to comprehend the etiological mechanisms of their characteristic behaviors, great importance should be placed on realizing how the neural networks of individuals with autistic disorders and WS are formed and work. As such, in this paper, cortical network abnormalities, based on data from a variety of research fields, are presented: psychopathological, histopathological, and clinicopathological studies, as well as structural (i.e., morphological) and functional magnetic resonance imaging studies, including functional connectivity analysis. Based on the structure of the network, we propose an etiology for ASD and WS. Finally, we explain a variety of symptoms of these two disorders, including social and nonsocial dysfunction, based on our proposed neural network.

Autism spectrum disorders (ASDs) and Williams syndrome (WS), two well-known neurodevelopmental disorders, have been considered to be very helpful in understanding how human social cognitive functions develop and work. Although they contrast in a few features behaviorally and morphologically, these two disorders have been regarded as complementary to each other (Deacon, 1997). However, according to a large amount of empirical and clinical data accumulated over the past two decades, such a dichotomic understanding has not been found to be applicable to many of the symptoms of the two disorders. Both disorders present some disabilities of the same cognitive competence, ranging from regulation of social interaction to nonsocial cognition, such as face recognition and response to facial expression. At this point, it is necessary anew to provide an overview of the clinical evidence acquired so far and to reconsider holistically the characteristic symptoms of both ASD and WS.

The clinical bases of ASD and WS have been revealed through different approaches. Autism was first described by Kanner (1943) and Asperger (1991) as a pervasive developmental disorder with a wide variety of phenotypes. It is a biologically based disorder characterized by a severe impairment of social and cognitive functions. The symptoms of autism result from abnormal brain development, which is probably caused by genetic factors (Hallmayer, Cleveland, Torres, Phillips, Cohen, Torigoe, Miller, Fedele, Collins, Smith, Lotspeich, Coren, Ozonoff, Lajonchere, Grether, & Risch, 2011; Pickles, Starr, Kazak, Bolton, Papanikolaou, Bailey, Goodman, & Rutter, 2000). Clinically, the disorder is defined by a triad of deficits that comprise impaired social interaction and communication skills, restricted interests, and repetitive behaviors, as described by the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV; American Psychiatric Association, 1994). In recent decades, several theories regarding autistic disorders have been proposed. These were summarized by Gergely and Watson (1999) and Belmonte, Allen, Beckel-Mitchener, Boulanger, Carper, and Webb (2004) as: “a missing drive for global coherence” (Frith, 1989); a lack of a “theory of mind” (Baron-Cohen, Leslie, & Frith, 1985; Leslie, 1994); a deficient “eye-tracking module” (Leekam, Baron-Cohen, Perrett, Milders, & Brown, 1977); a deficit of empathy (Baron-Cohen, 2002); a deficient “attention switching mechanism” ( Courchesne, Townsend, Akshoomoff, Saitoh, Yeung-Courchesne, Lincoln, James, Haas, Schreibman, & Lau, 1994); an “executive function deficit” (Ozonoff, Pennington, & Rogers, 1991; Russell, 1996); or a deficient “imitation mechanism” (Meltzoff & Gopnik, 1993). All of these theories are plausible in regard to the target disabilities; however, none of them can comprehensively explain the wide range of symptoms.

A number of theories have also been proposed at the neural level. For example, Baron-Cohen, Ring, Bullmore, Wheelwright, Ashwin, and Williams (2000) proposed an “amygdala theory of autism,” wherein they explain that a part of various symptoms of autism can be explained by an abnormality of the amygdala and resulting dysfunction. Several autism symptoms can be explained by an abnormality of the amygdala that would render it unable to serve the function for which the intact amygdala should be responsible. In contrast, Hadjikhani, Joseph, Snyder, and Tager-Flusberg (2006) and Hadjikhani (2007) proposed dysfunction of the mirror neuron system in autism. However, Hadjikhani et al. (2006) additionally pointed out that some of the behavioral changes observed in autism might be the outcome of dysfunctions of the distributed neural circuitry for social cognition, apart from the mirror neuron system (MNS) and the amygdala. In contrast, many other researchers have found that several other brain areas also show some functional and morphological abnormalities (as shown later in this article) and that functional connectivities also deteriorate or develop abnormalities between brain areas or within specific brain areas. The question as to why these types of neural impairments would result in a wider distribution over cortical and subcortical area remains unanswered.

At present, few computational models have dealt with symptoms of ASD: one model was based on the adaptive resonance theory (Grossberg & Seidman, 2006) and another model was based on a self-organizing map (SOM; Gustafsson & Papliński, 2004). However, these models focused only on a few symptoms, and therefore cannot answer the question posed above. The unifying pathogenesis, or etiology, of this disease is unknown, partly because of the broad variation of phenotypes and the heterogeneity of syndrome expression.

In contrast, WS is a rare disorder caused by hemizygous microdeletion of approximately 28 genes on chromosome 7q11.23. It has long intrigued neuroscientists, with its unique combination of striking behavioral abnormalities, such as hypersociability and characteristic neurocognitive profile (Meyer-Lindenberg, Mervis, & Berman, 2006). The absence of multiple genes in individuals with WS results in many effects on the brain, including abnormalities in the cerebellum, right parietal cortex, and left frontal cortical regions. At present, no unified and specific explanations have been proposed for these kinds of social and cognitive disorders. In general, WS shows hypersociability, while ASD shows disability in social interaction. We will show that a similar hypothesis holds for WS, which has been considered as the clinical complement or opposite of autism; although, as described later, anxiety symptoms have been reported in both ASD and WS. Our attempt is to position the contradistinctive features as well as the common features of ASD and WS to conform to a common framework. This should be an effective approach for comprehensive realization of these two dissimilar developmental disorders.

One objective of this paper is to propose a hypothesis regarding the emergence mechanism of these two distinctive neural dysfunctions. In other words, our aim is to propose a hypothesis regarding which brain area shows maldevelopment or abnormality in the early period of development, including the fetal period, and what kind of mechanism causes the underdevelopment of several brain areas and connections among the areas and results in the functional and morphological abnormalities. We will also explain various dysfunctions described below in autism based on this hypothesis.

In order to comprehend the etiological mechanisms of their characteristic behaviors in particular, it is of great importance to realize how the neural networks of ASD and WS are formed and work. Thus, in this paper, we aim to form a basis for the etiology of autism and WS. First, we review the details of the known brain network abnormalities, based on data from a variety of research fields: psychopathological, histopathological, and clinicopathological studies and structural (i.e., morphological) and functional magnetic resonance imaging (fMRI) studies, including functional connectivity analysis. From this overview of structural and functional impairment, we mainly focused on six areas in ASD and WS brains: the amygdala, the hippocampus, the orbitofrontal cortex (OFC), the inferior frontal gyrus (IFG), the posterior parietal cortex (PPC) including the TPJ and IPS, and the posterior STS. Two reasons prompted us to focus intensively on these six components. First, rich neurological data are available for ASD and WS; second, the function of these areas is understood in considerable detail and we can explain several symptoms found in ASD and WS based on these data.

We also discuss, in detail, the structures reported in these disorders. Third, we propose the presumable neural networks of ASD and WS to provide an etiology of the disorders. Finally, we apply these networks to explain the following symptoms of these disorders, which include social and nonsocial dysfunction. The core clinical features of ASD include: (a) impaired reciprocal social interaction skills; and (b) the presence of restricted and stereotyped behaviors and interests. In this paper, we focused on the abnormalities in eye contact and body language, or deficits in understanding and use of nonverbal communication in autism as social dysfunction, and atypical categorical perception as nonsocial dysfunction. We also focused on the hypersociability in WS as social dysfunction and impairment in copying line drawings in WS as nonsocial dysfunction.

Apart from the core features, anxiety symptoms frequently prevail in ASD and WS. For example, Kanner (1943) suggested that many of the core features of ASD, particularly the insistence on sameness and the repertoire of fixed behaviors, routines and obsessions, were derived from anxiety. Therefore, we also dealt with anxiety and deterioration of mood regulation in ASD.

Morphometric abnormalities of the brain in ASD and WS

ASD was first defined and diagnosed exclusively by criteria consisting of overt behavioral characteristics; hence, various types of behavioral examinations were designed and conducted, particularly in the field of psychology, and a large amount of empirical evidence of ASD was spontaneously collected. With the development and prevalence of human neuroimaging technologies in recent decades, detailed data regarding brain anatomy, physiology, genetics, and biochemistry have increasingly been obtained. As such, ASD has been characterized across different levels, from molecular to behavioral aspects: at the molecular level (Kleinhans, Richards, Weaver, Liang, Dawson, & Aylward, 2009; Lam, Aman, & Arnold, 2006; Martin-Ruiz, Lee, Perry, Baumnn, Court, & Perry, 2004; Penn, 2006), the local neural network level (Belmonte et al., 2004; Rubenstein & Merzenich, 2003), the large-scale neural network level (Barttfeld, Wicker, Cukier, Navarta, Lew, & Sigman, 2011), and the behavioral level (Baron-Cohen, Leslie, & Frith, 1986; Blair, Frith, Smith, Abell, & Cipolotti, 2002; Hamilton, Brindley, & Frith, 2009).

WS, in contrast, was first described as a series of developmental and physical features, including the specific facial features known as elfin faces (for details, see Lashkari, Smith, & Graham, 1999, as a review). The pathogenesis of this disease was investigated intensively in the 1990s, and it was revealed that WS could be characterized by contiguous gene deletions. For both ASD and WS, however, such a wealth of data was not understood yet on an integrated basis, and, therefore, could not be integrated into one unified theory or hypothesis.

In order to achieve our aim of understanding the unified etiological mechanisms of the cognitive disabilities manifested in ASD and WS, we should hypothesize a brain network and, based on that, explain how such a variety of characteristic behaviors are present in both disorders. In this section, we organize the morphometric characteristics, which are specifically important in considering the formation and function of the brain network. In particular, we describe in detail the abnormalities observed in the following six structures: the amygdala, the hippocampus, the OFC, the IFG, the PPC, and the STS (Table 1).

Table 1. Abbreviations
ASDAutism spectrum disorders
E/IExcitation/inhibition
fMRIFunctional magnetic resonance imaging
IFGInferior frontal gyrus
IPSIntraparietal sulcus
LOFCLateral orbitofrontal cortex
MIPMedial intraparietal sulcus
MOFCMedial orbitofrontal cortex
OFCOrbitofrontal cortex
PPCPosterior parietal cortex
STSSuperior temporal sulcus
TPJTemporoparietal junction
WSWilliams syndrome

Amygdala

ASD. 

Bauman and Kemper (1994) found that the amygdala and entorhinal cortex (EC) showed reduced neuronal cell size and increased bilateral cell packing density in individuals with autistic disorder when compared with controls. In contrast, Abell, Krams, Ashburner, Passingham, Friston, Frackowiak, Happé, Frith, & Frith (1999) found relative increases in the amygdaloid and periamygdaloid regions using voxel-based morphometry. The enlargement in the amygdaloid region in individuals with autistic disorder was also detected by Howard, Cowell, Boucher, Brocks, Mayes, Farrant, and Roberts (2000).

WS. 

In 2006, Meyer-Lindenberg et al. showed that amygdala reactivity to threatening-relevant social stimuli (angry and fearful facial expressions) in individuals with WS was significantly diminished, while reactivity to nonsocial, threatening-relevant stimuli (threatening and fearful scenes) was abnormally increased. This result suggests that the amygdala in individuals with WS is not structurally damaged, but that there is dysfunction, particularly in processing social stimuli. Therefore, we can conclude that social behavioral abnormality in WS is not due to a deficit of the amygdala itself, but rather an impairment of its function through the amygdalo-OFC network, as suggested by Meyer-Lindenberg et al. (2006).

Hippocampus

ASD. 

Bauman and Kemper (2005) found reduced cell size and increased cell packing density in the hippocampus in postmortem brains of individuals with autistic disorder. Raymond, Bauman, and Kemper (1996) found reduced complexity and extent of dendritic arbors in CA1 and CA4. A later study (Nicolson, DeVito, Vidal, Sui, Hayashi, Drost, Williamson, Rajakumar, Toga, & Thompson, 2006), which used a computational mapping strategy to examine the three-dimensional profile of hippocampal abnormalities in autism, suggested that autism may be associated with subtle regional reductions in the size of the hippocampus.

WS. 

Structural changes in the hippocampus have not been detected. However, a multimodal neuroimaging study by Meyer-Lindenberg, Mervis, Sarpal, Koch, Steele, Kohn, Marenco, Morris, Das, Kippenhan, Mattay, Weinberger, and Berman (2005) showed a profound reduction in resting blood flow in the anterior hippocampal formation.

Meda, Pryweller, and Thornton-Wells (2012) found a reduced volume of the hippocampus and reduced surface area of the parahippocampal gyrus in WS patients. They pointed out that this might be consistent with the results presented by Meyer-Lindenberg et al. (2005), who observed a profound reduction in blood flow to the hippocampal formation in WS.

OFC

The OFC is an important brain region for emotional and social behavior, as it is responsible for the processing of rewards and punishments. A mediolateral distinction is known to exist: medial OFC (MOFC) activity is related to the monitoring of reward values of many different reinforcers. At the same time, lateral OFC (LOFC) activity is related to the evaluation of punishers, which may lead to a change in ongoing behavior (Kringelbach & Rolls, 2004).

ASD. 

Pardini, Garaci, Bonzano, Roccatagliata, Palmieri, Pompili, Coniglione, Krueger, Ludovici, Floris, Benassi, and Emberti Gialloreti (2009) used the techniques of diffusion tensor imaging and fiber tractography to investigate white matter in low-functioning autism and the relation between white matter and cognitive impairment. They showed that white matter abnormalities were significant in the white matter areas that surrounded the bilateral cortical surfaces of the inferior and medial frontal gyri, which correspond to the LOFC and MOFC, respectively. In this case, the MOFC corresponds to the orbital medial frontal cortex; the Talairach z-coordinate is less than zero (Amodio & Frith, 2006).

Jiao, Chen, Ke, Chu, Lu, and Herskovits (2010) found decreased cortical thickness in the left medial orbitofrontal gyrus. In contrast, Girgis, Minshew, Melhem, Nutche, Keshavan, and Hardan (2007) found decreased gray matter volume in the right LOFC in children with autism (age range 8.1–12.7 years).

WS. 

Empirical evidence obtained by meta-analyses of functional imaging data has shown abnormalities in the LOFC of individuals with WS (Meyer-Lindenberg et al., 2006).

Inferior frontal cortex

ASD. 

Abell et al. (1999) used a voxel-based whole brain analysis and found decreases in gray matter in the left IFG. Cortical thinning found bilaterally in the IFG pars opercularis was correlated with Autism Diagnostic Interview-Revised (ADI-R) combined social and communication diagnostic algorithm scores, which are based on the parental report of an individual's behaviors between the ages of 4 and 5 years (Hadjikhani et al., 2006).

Posterior parietal lobe

ASD. 

Cortical thinning has been found in the bilateral inferior parietal lobule and also has been correlated with ADI-R combined social and communication diagnostic algorithm scores (Hadjikhani et al., 2006). The superior temporal sulcus (STS) at the temporoparietal junction (TPJ) shows less activation compared with normal subjects when viewing animations that elicited mentalizing (Castelli, Frith, Happe, & Frith, 2002). Furthermore, Kana, Keller, Cherkassky, Minshew, and Just (2009) also found less activation in theory of mind tasks. They found that, in the autism group, an independent psychometric assessment of theory of mind ability and activation in the right temporoparietal junction were positively correlated.

WS. 

Gray matter volume reduction has been found in the region of the dorsal parietooccipital/vertical part of the intraparietal sulcus (IPS) around V6/V6A (Meyer-Lindenberg, Kohn, Mervis, Kippenhan, Olsen, Morris, & Berman, 2004).

Functional activation studies and functional connectivity in ASD and WS

This section provides a brief summary of the recent findings regarding brain activation and functional or effective connectivities among brain regions of interest in ASD and WS.

Activation study

ASD. 

The human MNS, also known as the action observation network, is comprised of the PPC and the ventral premotor region. This system is known to play an essential role in the understanding of another's actions, as determined by findings regarding the mirror neurons of monkeys (for a review, see Iacoboni & Dapretto, 2006; Rizzolatti, Fogassi, & Gallese, 2001). A large number of studies with human participants have reported activation in the parietal and premotor areas during observation or imitation of the actions of others (Buccino, Binkofski, Fink, Fadiga, Fogassi, Gallese, Seitz, Zilles, Rizzolatti, & Freund, 2001; Iacoboni, Woods, Brass, Bekkering, Mazziotta, & Rizzolatti, 1999; Johnson-Frey, Maloof, Newman-Norlund, Farrer, Inati, & Grafton, 2003; Tanaka & Inui, 2002). It has been suggested that EEG oscillations in the mu frequency (8–13 Hz) over the sensorimotor cortex show the activation of mirror neurons (Ramachandran & Oberman, 2007; Ulloa & Pineda, 2007). Recently, Lepage & Theoret (2006) showed that mu rhythm attenuation occurs in children less than 11 years of age during observation of hand movements. The observation of goal/object-orientated movement reveals greater modulation of the mu rhythm than does observation of intransitive movement. In an ASD group, Oberman, Hubbard, McCleery, Altschuler, Ramachandran, and Pineda (2005) found significant mu suppression in response to self-performed hand movements, but not to observed hand movements, which supports the hypothesis of a dysfunctional MNS in high-functioning individuals with ASD.

WS. 

Meyer-Lindenberg et al. (2006) found a significant diminution in the reactivity of the amygdala in individuals with WS in response to threatening, socially relevant stimuli, while the reactivity to socially irrelevant stimuli was abnormally increased. This is consistent with the clinical profile of WS; namely, specific phobia. Therefore, in WS, a functional abnormality exists in the amygdala.

Functional connectivity

ASD. 

Local overconnectivity and long-range underconnectivity has been found in ASD patients (Belmonte et al., 2004; Herbert, Ziegler, Makris, Filipek, Kemper, Normandin, Sanders, Kennedy, & Caviness, 2004). Local overconnectivity is considered to cause hyperactivation in local brain areas, while long-range underconnectivity is considered to cause dysfunction of information integration. Just, Cherkassky, Keller, and Minshew (2004) also proposed an underconnectivity theory of autism based on the data from their sentence comprehension experiment. This theory targeted the underlying biological structures and processes of the weak central coherence theory of autism by Frith (1989) in order to explain the underfunctioning of integrative circuitry and emergent cognitive, perceptual, and motor abilities in autism. Minshew, Sweeney, and Bauman (1997) considered that ASD is selectively impaired in complex information processing that does not involve visuospatial processing and that it is consistent with a late information processing disorder with sparing of early information processing.

An example of long-range underconnectivity was presented by Grèzes, Wicker, Berthoz, and de Gelder (2009), who revealed a weaker connectivity from the amygdala to the lateral IFG, STS, and premotor area in ASD subjects when compared with normal subjects using dynamic causal modeling (DCM) analysis on activation for the emotional expression of observed actions. They also found that effective connectivities from IFG to premotor, and from premotor to STS, are weak in ASD subjects.

WS. 

Meyer-Lindenberg et al. (2004) applied structural equation modeling (SEM) to an activation study regarding visual attention by adults with WS and age-, IQ-, and gender-matched controls. The study included the following nodes: early visual areas, the most activated ventral stream region, the structurally changed IPS region, and a dorsal stream location of pronounced between-group activation difference. The only difference noted between groups was in the path from IPS to the later dorsal stream region.

Anatomical connections between the OFC and the limbic system

The cells of origin for the extensive projections from the basal amygdaloid nucleus to different areas of the MOFC form distinct clusters within the nucleus. In addition, the subiculum projects to the medial orbital cortex and gyrus rectus (Carmichael & Price, 1996). Projections also extend from the accessory basal and lateral nuclei of the amygdala within the LOFC. Most of the amygdaloid connections are reciprocated by cortico-amygdaloid fibers. In addition, bidirectional connections also link the perirhinal and entorhinal cortex and the central and posterior LOFC (Carmichael & Price, 1996).

Structural and functional abnormalities of the neural networks in ASD and WS

Before proposing our hypothesis, we briefly summarized the various data mentioned in the previous sections, as shown in Table 2. We then synthesized these findings, as well as some additional data particularly pertaining to anatomical connections (Carmichael & Price, 1996; Kringelbach & Rolls, 2004; Rizzolatti & Matelli, 2003), and the functional connectivity described in the previous section, into a neural network model for ASD and WS. These are shown in Figures 1 and 2, respectively.

Figure 1.

Structural abnormality, functional abnormality, and abnormal connections between regions in autism spectrum disorder. AMG = the amygdala; HC = hippocampus; IFG = inferior frontal gyrus; LOFC = lateral orbitofrontal cortex; MOFC = medial orbitofrontal cortex; SMG = supramarginal gyrus; STS = superior temporal sulcus; TPJ = temporoparietal junction.

Figure 2.

Structural abnormality, functional abnormality, and abnormal connections between regions in Williams syndrome. AMG = the amygdala; HC = hippocampus; LOFC = lateral orbitofrontal cortex; MIP = medial intraparietal sulcus; MOFC = medial orbitofrontal cortex.

Table 2. Morphological abnormality (MA) and functional abnormality (FA) in autism and Williams syndrome WS
 WSAutism
HippocampusFAMA
AmygdalaFAMA
Posterior parietal cortexMA (V6/V6a)FA (right temporoparietal junction)
Medial intraparietal sulcusFA 
Lateral orbitofrontal cortexMA 
Medial orbitofrontal cortex MA
Inferior frontal gyrus MA

The figures show the regions (depicted by gray boxes) with structural abnormalities and the regions (depicted by partially gray hatched boxes) with functional abnormalities. Structural abnormalities mean a reduction in the volume of gray matter, a reduction in cell size, or an increase in cell density. The arrows in the figure represent connections between regions, as confirmed by several physiological studies described above. Solid arrows indicate the intact connections, and dashed arrows depict significantly weaker connections in the disorder groups than in the control group. Several regions in the large-scale network, as shown in Figures 1 and 2, exhibit functional abnormalities, such as hypoactivation, without structural abnormalities. These functional abnormalities are located in the TPJ and STS in individuals with ASD and in the hippocampus, amygdala, and MIP in individuals with WS (see section “Posterior parietal lobe”). It is supposed that these abnormalities, shown in Figures 1 and 2, evoke various cognitive dysfunctions in these two developmental disorders. We will discuss the etiology and the neural basis mechanisms of various symptoms in detail in the following sections.

Hypothesis about pathophysiological mechanisms

Ontogenetic cytoarchitectural viewpoints

In general, the limbic system shows earlier maturation relative to the neocortex during fetal development. At 9 weeks' gestation, the hippocampal region contains four layers, and at 15–19 weeks, it can differentiate into anatomical components (Arnold & Trojanowski, 1996). Reciprocal connections between the entorhinal cortex, hippocampus, and subiculum also develop by 19 weeks (Hevner & Kinney, 1996). With regard to the amygdala, one of the other limbic structures, characteristic cytoarchitectonic modules, develops by 15 fetal weeks (Kostovic, Judas, & Petanjek, 2008).

Here, we supposed that an early developed system affects strongly (the development of) the late matured system through interaction, because normal development of the neural system needs appropriate input in general. According to the data mentioned above, the limbic system cytoarchitectonically develops earlier than other parts of the cerebral cortex. In addition, the interactive connection between the limbic system and other brain areas forms gradually. Supposing that there is some type of damage to the limbic system in the early stages of fetal life, the input from the limbic to the other connected systems is expected to be inappropriate, that is, deteriorated or nonexistent. In this case, the brain areas connected directly to the limbic system will not develop normally. Therefore, any damage to the limbic system during the developmental process of the fetus could be considered to cause extraordinary effects on the cytoarchitectural development of many brain areas in comparison with damage to other brain areas. In the next section, we discuss recent findings regarding the experience-dependent plasticity of the human brain as a mechanism of influence, by the appropriate input, to form networks.

Experience-dependent plasticity and abnormality

The importance of appropriate inputs for the formation of the brain network is suggested by the development of the visual system; therefore, ontogenetic development has been studied very extensively (Blakemore & Van Sluyters, 1974; Blakemore & Vital-Durand, 1986; Pettigrew, 1974). Recently, the plasticity of the human visual nervous system has been examined through the use of voxel-based morphometry. Mendola, Conner, Roy, Chan, Schwartz, Odom, and Kwong (2005) showed that adults and children with amblyopia have a decreased volume of gray matter in their visual cortical regions, including the primary visual cortex. Children with amblyopia also show further gray matter reduction in the parietal-occipital areas and in the ventral temporal cortex. Boucard, Hernowo, Maguire, Jansonius, Roerdink, Hooymans, and Cornelissen (2009) obtained high-resolution anatomical MRIs in subjects with foveal (age-related macular degeneration) and peripheral (glaucoma) retinal lesions, and found density reduction in the approximate retinal lesion projection zones of the visual cortex.

These results suggest that the absence of an appropriate input to a particular locus in the brain may alter the synaptic modification, and thereby cause changes in the morphological structure. The reduction of the gray matter volume, both in the visual cortex of amblyopia and retinal lesions, is possibly influenced by a reduced behavioral experience of visual perception. Regarding the physiological aspects, the low, weakened, or nonexistent visual inputs would affect the formation of normal structures in the visual and connected cortices. This neurophysiological factor in the formation of neural networks, experience-dependent plasticity, is applicable to other brain regions, and also to other developmental stages.

Hypothesis of the etiology

Based on the composition of the affected network, and on suggestions from ontogenesis and the plasticity of the neural network, we hypothesized that the etiology of both ASD and WS involves underdevelopment, or damage, to the fetal limbic system, particularly the amygdala. This would then result in dysfunction, or morphological abnormality of the brain areas connected directly with the limbic system, giving rise to the various phenotypes of ASD and WS, as described in the following section. In other words, deficits within the higher-order association areas, for example TPJ and OFC, are a secondary disorder, due to abnormal structures and/or functions caused by too few, or no appropriate inputs from the limbic cortex, such as the amygdala and hippocampus, because a large number of inputs as well as outputs converge to and diverge from those components. On the basis of this hypothesis, we attempt to provide concrete descriptions of symptoms evident with ASD and WS.

Implications of our hypothesis for various expressions

In this section, several symptoms described in the DSM-IV (American Psychiatric Association, 1994), and in the DSM-5 development (American Psychiatric Association, 2010) and also in clinical research papers, are examined and explained according to our hypothesis and according to the abnormalities seen in large neural networks, as shown in Figures 1 and 2.

Symptom 1 Social dysfunction

Abnormalities in eye contact and body language, or deficits in understanding and use of nonverbal communication in ASD

With ASD, several lines of evidence indicate a deterioration of perception of biological motion.

The STS region is activated by movements of the eyes, mouth, hands, and body, and it is involved in analysis of biological motion (Allison, Puce, & McCarthy, 2000). Certain neurons are sensitive to gaze direction in the STS. Dysfunction of the STS causes sensitivity deterioration of gaze direction (Campbell, Heywood, Cowey, Regard, & Landis, 1990). In contrast, the STS is a part of the network of the social brain (Brothers, 1997), of mentalizing (Amodio & Frith, 2006), and the MNS (Iacoboni, 2005). Castelli et al. (2002) used animations depicting two triangles moving around on a screen as the condition of moving interactively with implied intentions (coaxing, tricking), which elicited descriptions in terms of mental states that viewers attributed to the triangles (mentalizing). In this experiment, less activation was observed in the ASD group than in the normal group in the medial prefrontal cortex and STS, particularly at the TPJ and temporal pole/amygdaloid region (maximum difference between these two groups) (see Figure 2).

Castelli et al. (2002) considered that a lack of feedback from the temporal pole/amygdaloid region and/or medial prefrontal cortex to the STS results in a transmission failure between the V3 and STS, and thus in the inability to recognize the social significance of the moving triangles. They also pointed out the importance to visual processing of the feedback from the amygdala. In fact, Amaral, Behniea, and Kelly (2003) showed that the amygdala may have substantial modulatory control over sensory processing at all stages of the ventral-stream visual cortical hierarchy. Lack of this modulatory control by the amygdala would contribute to the dysfunction of attribution of mental states to animated shapes observed in ASD.

The neural network shown in Figure 1, which features impairments of the amygdala and the posterior STS (pSTS) or TPJ, could also explain the abnormality of eye contact and joint attention. Abnormality of gaze fixation is one of the typical features of ASD (Dalton, Nacewicz, Johnstone, Schaefer, Gernsbacher, Goldsmith, Alexander, & Davidson, 2005). Spezio, Huang, Castelli, and Adolphs (2007) showed that complete amygdala lesions result in a severe reduction in direct eye contact during conversations with real people, together with an abnormal increase in gaze to the mouth. Kawashima, Sugiura, Kato, Nakamura, Hatano, Ito, Fukuda, Kojima, and Nakamura (1999) found that discrimination of gaze direction significantly activated a region in the left amygdala during both eye-contact and no-eye-contact tasks to the same extent, while the right amygdala was specifically activated only during the eye-contact task. Based on these results, they considered that the left amygdala plays a general role in the interpretation of gaze direction, and that the activity of the right amygdala increases when another individual's gaze is directed towards the viewer. Recently, Redcay, Kleiner, and Saxe (2012) found that the pSTS, as well as the dorsal medial prefrontal cortex (dMPFC), is engaged for both IJA (Initiating Joint Attention) and RJA (Responding to Joint Attention) in a functional MRI experiment in which participants had to follow the experimenter's gaze to a target (RJA) or cue the experimenter to look at the target (IJA). Redcay, Dodell-Feder, Mavros, Kleiner, Pearrow, Triantafyllou, Gabrieli, and Saxe (2012) found significant differences between ASD and control groups in activation of social-cognitive brain regions, including the dMPFC and the right pSTS, during the RJA.

Furthermore, autistic children frequently show visually matched imitation instead of body-centered imitation: for example, autistic children wave “bye-bye” by presenting their palms to themselves, but not to the model. This suggests the dysfunction of a body-centered representation of the other's action in autism, which would be a function of the mirror system. It has been known that the fronto-parietal MNS is important for both the recognition and execution of sequential actions. The MNS is a network composed of the ventral premotor cortex/posterior IFG, the inferior parietal lobule, and the pSTS (Iacoboni & Dapretto, 2006). The pSTS is reciprocally connected to the inferior parietal lobule (Harries & Perrett, 1991; Seltzer & Pandya, 1994) which is in turn connected to the ventral premotor cortex/posterior IFG (Luppino, Murata, Govoni, & Matelli, 1999): These areas are hierarchically organized and reciprocally connected (Kilner, Friston, & Frith, 2007). Furthermore, it has been shown that the densest projections from the AMG to the prefrontal cortex terminate in the medial and lateral prefrontal cortex in monkeys, which includes the IFG in humans (Amaral & Price, 1984).

Recently, Ogawa and Inui (2011) investigated the neural representation of observed actions in the human parietal and premotor cortex, which comprise the action observation network, or the MNS, for action recognition. Participants observed object-directed hand actions, in which action, as well as other properties, were independently manipulated: action (grasp or touch), object (cup or bottle), perspective (first or third person), hand (right or left), and image size (large or small). Multivoxel pattern analysis was used to determine whether each feature could be correctly decoded from regional activities. The early visual area showed significant above-chance classification accuracy, particularly high in perspective, hand, and size, which was consistent with the pixel-wise dissimilarity of the stimuli. In contrast, the highest decoding accuracy for the action was observed in the anterior IPS and the ventral premotor cortex. Moreover, the decoder for action could be correctly generalized for images with high dissimilarity in the parietal and premotor region, but not in the visual area. This indicates that the parietal and premotor regions encode observed actions independently of retinal variations caused by combination of attributes of perspective, hand, and size, which may subserve the capacity for invariant recognition of the actions of others. Furthermore, as mentioned earlier, Grèzes et al. (2009) revealed a weaker connectivity from the amygdala to the lateral IFG, STS, and premotor area in individuals with ASD, compared with normal subjects. Hadjikhani (2007) also argued that deficiencies of empathy and imitation of another's body movements, including facial expression, are caused by dysfunctions of the MNS and the amygdala.

Therefore, the functional abnormality of the IFG, which would be caused by the mechanism discussed in the section “Hypothesis of the etiology” in IFG-amygdala interaction, results in a dysfunction both in the understanding and imitation of another's action.

Symptom 2 Social dysfunction

Deterioration of mood regulation and anxiety in ASD

Individuals with ASD have a higher degree of anxiety than typically developing persons (Bradley, Summers, Wood, & Bryson, 2004). The OFC has been implicated in the pathophysiology of major depression and mood regulation; severity of depression correlates inversely with physiological activity in parts of the posterior LOFC and MOFC (Drevets, 2007).

Selective bilateral lesions of the amygdala in mature macaque monkeys result in a lack of fear responses to inanimate objects and a “socially uninhibited” pattern of behavior (Amaral, 2002). According to Holland and Gallagher (2004), while the amygdala appears to be a neural system that acts to detect the significance of objects or events for the individual, the OFC makes use of this information to guide goal-directed behaviors and to adjust behavior appropriately in accordance with changing conditions. Therefore, the OFC-amygdala is important for the adjustment of behavior. In other words, the amygdala codes for the valence of objects, actions, or events, whereas the OFC represents future outcomes. The MOFC respond more to positive events while the LOFC responds more to negative events.

Dysfunction in amygdala-cortical interactions results in anxiety disorders. The amygdala is implicated in generating fear responses, whereas the cortical regions, specifically the MOFC and the ventromedial prefrontal cortex, are implicated in fear extinction. In contrast to the MOFC, the LOFC has been associated with negative effects and obsessions (Milad & Rauch, 2007). Grèzes et al. (2009) considered that reduced connectivity between the amygdala and the OFC, or reduced activity in these areas, may impair decoding of the emotional coloring of events and could explain the flattening of emotions that is seen in ASD. Moreover, as described above, morphological abnormalities of the MOFC and functional connectivity are found in autism (Table 2); therefore, it is possible that deterioration of mood regulation and anxiety are caused by dysfunction of the amygdale-MOFC interaction.

Symptom 3 Nonsocial dysfunction

Atypical categorical perception and magnocellular visual pathway deficit in ASD

Atypical categorical perception has been found in ASD (Soulières, Mottron, Saumier, & Larochelle, 2007), which suggests that there is a diminished top-down effect on categorization tasks in persons with ASD.

It is well known that there are two parallel visual processing pathways from the retina to the visual cortex: the parvocellular pathway and the magnocellular pathway. In the parvocellular pathway, local, fine (i.e., with higher spatial frequency of components) visual information is processed and transmitted slowly, while in the magnocellular pathway, global, coarse (i.e., with lower spatial frequency of components) visual information is processed and transmitted rapidly. The function and pathways of magnocellular cells in a large-scale neural network were recently investigated using fMRI. Kveraga, Boshyan, and Bar (2007) found that magnocellular visual information activates the pathway from the visual cortex to the IFG (BA45) and LOFC (BA47). They also found that the activation is transmitted from these frontal areas to the fusiform gyrus in the inferotemporal gyrus as a top-down signal. Therefore, fast magnocellular projections, linking early visual and inferotemporal object recognition regions with the OFC, facilitate object recognition. This is consistent with the physiological finding that neural activity in the lateral prefrontal cortex of monkeys reflects visual stimulus categories (Freedman, Riesenhuber, Poggio, & Miller, 2001).

It is known that the sensitivity of the magnocellular pathway is impaired also in ASD (McCleery, Allman, Carver, & Dobkins, 2007). Furthermore, as mentioned above, IFG abnormalities have been found in ASD. Therefore, it is possible that magnocellular information is impaired and that the top-down signal from the frontal lobe to the temporal lobe is diminished.

Symptom 4 Social dysfunction

Individuals with WS exhibit indiscriminate approach behavior toward strangers in everyday life.

Damage to the OFC in humans is associated with impairment in emotional and social behavior, and is characterized by social disinhibition and social inappropriateness. In the human OFC, MOFC activity is related to monitoring the reward value of many different reinforcers, while LOFC activity is related to the evaluation of punishers in ongoing behavior (Kringelbach & Rolls, 2004). Similarly, O'Doherty, Kringelbach, Rolls, Honak, and Andrews (2001) found that the brain areas that respond to reward and punishment are the MOFC and the LOFC. Furthermore, O'Doherty, Winston, Critchley, Perrett, Burt, and Dolan (2003) found that attractive faces produced activation of the MOFC, a region involved in representing stimulus-reward value. Responses in this region were further enhanced by a smiling facial expression, suggesting that the reward value of an attractive face, as indexed by MOFC activity, is modulated by a perceiver-directed smile.

As previously mentioned, the OFC is connected to the amygdala bidirectionally. Through this interaction with the amygdala, the OFC integrates incentive information with other types of information in order to select appropriate behavioral responses (Schoenbaum, Setlow, Nugent, Saddoris, & Gallagher, 2003).

Therefore, simply stated, in the functioning of the amygdala-OFC network, the MOFC provides a “go” signal and the LOFC provides a “stop” signal based on the valence coding of the amygdala to the areas that are involved in the selection of behaviors (e.g., neucleus accumbence and ventral pallidum; Levine, 2005). In WS, as described earlier (see also Table 2), a morphological abnormality exists in the LOFC and a functional abnormality in the amygdala. Therefore, the amygdala-OFC network does not provide an appropriate stop signal based on the valence coding of the amygdala (see section “Symptom 2” for more detail about the function of the OFC and the amygdala).

Regarding hypersociability, Bellugi, Mills, Jernigan, Hickok, and Galaburda (1999) found an interesting difference in the responses shown by WS patients and those with bilateral amygdala damage in an experiment that examined the social judgments of unfamiliar individuals. Subjects with WS gave abnormally positive ratings across all stimuli, both those normally judged to be unapproachable and those normally judged approachable. Subjects with bilateral amygdala damage, in contrast, gave abnormally positive ratings only for those faces that normally receive the most negative ratings (normally judged to be the most unapproachable). The researchers' interpretation of these disparate results is that amygdala damage impairs the normal ability to judge unapproachability (to detect potentially dangerous or threatening people), whereas WS augments a strong social tendency that promotes approaching other individuals (Bellugi et al., 1999). These results are consistent with our scheme of an amygdala-OFC network with LOFC dysfunction in WS: the damage to the bilateral amygdalae might induce the inevitability of any of the stimuli to be avoided, and additional damage of the LOFC might induce, or enhance, the indiscriminate approachability to any kinds of social stimuli.

Symptom 5 Nonsocial function

Individuals with WS have an impaired ability to copy line drawings (Nagai, Inui, & Iwata, 2011; Ogawa, Nagai, & Inui, 2010); we consider the mechanism of this dysfunction, based on the above discussion.

To investigate the brain areas involved in copying line drawings, Ogawa and Inui (2009) used fMRI to measure brain activity during drawing, by having normal adults copy and trace. During a copying task, the IPS in the PPC showed, bilaterally, significantly greater activation. The observed IPS coordinate is close to the human medial intraparietal sulcus (MIP), which constitutes the human parietal reach region involved in hand/cursor movements. They considered that the role of the MIP is to coordinate the transformation between egocentric and allocentric representations of model figures, which is important in copying and construction tasks.

As stated previously, in WS, a reduction in gray matter volume was found in the region of the dorsal parieto-occipital/vertical part of the IPS (Meyer-Lindenberg et al., 2004), around V6/V6A (see Table 2). Furthermore, Meyer-Lindenberg et al. (2004) found hypoactivation of the IPS during a construction task.

We considered that this hypoactivation results in impairment of copying, and that the hypoactivation is caused by the morphological abnormality of V6/V6A, which is consistent with our hypothesis mentioned above: abnormality of V6/V6A causes abnormality in the area directly connected to it, because the output from the area V6 reaches the MIP and V6A (Rizzolatti & Matelli, 2003). The MIP is involved in coordinate transformation, as described above, and the V6A is important in eye-hand coordination (Caminiti, Genovesio, Marconi, Mayer, Onorati, Ferraina, Mitsuda, Giannetti, Squatrito, Maioli, & Molinari, 1999); both areas are very important for copying and construction.

Other important symptoms and future directions

Several important dysfunctions were not discussed in the previous section, in particular, repetitive behavior, which is a stereotypically known indicator of ASD. (DSM-IV, American Psychiatric Association, 1994; DSM-5 Development, American Psychiatric Association, 2010). This behavior has not been explained explicitly, and it may be explained by another factor, such as the type of neuromodulator, and its developmental change in concentration with age, for example, norepinephrine and acetylcholine. Computational neuroscience shows that this concentration is very important in the control of shifting and refocusing attention between sensory cues and in the adaptation to new predictive relationships (Yu & Dayan, 2005).

Another important characteristic of ASD is a delay in, or lack of, pointing behavior. Children with typical development show pointing behavior, as a communication of demands, at approximately 1 year of life. However, instead of using pointing behavior, children with ASD express their demands by using other persons' hands as their manipulator, known as the “crane phenomenon,” an abnormal use of another person as a kind of extension of the subject's arm. We recently showed that pointing to distant objects is a behavior that emerges through an autonomous learning process about the forward and inverse functions of one's own hand (Takemura & Inui, in press); as such, training for pointing behavior is not needed. Unfortunately, the reason individuals with autism do not exhibit pointing behavior is not understood at the physiological level.

Finally, we must mention levels of understanding in our framework. Marr (1982) pointed out that there are three levels of explanation, particularly in the human sciences: computation level, algorism and representation level, and implementation level. We hope our work leads to a breakthrough in the understanding of developmental disorders on the first two levels. To understand the implementation level, we must discuss the excitation/inhibition (E/I) ratio which would be caused by local overconnectivity in ASD, as described in the introduction.

Individuals with ASD often have a hypersensitivity to auditory and tactile stimuli. The hypersensitivity seen in autism is modeled by an increased E/I ratio (Rubenstein & Merzenich, 2003). Rubenstein and Merzenich proposed that the E/I ratio is increased in sensory, mnemonic, social, and emotional systems. However, this hypothesis has not been subjected to direct testing. Recently, Yizhar, Fenno, Prigge, Schneider, Davidson, O'Shea, Sohal, Goshen, Finkelstein, Paz, Stehfest, Fudim, Ramakrishnan, Huguenard, Hegemann, and Deisseroth (2011) developed novel optogenetic tools to causally investigate the cellular E/I balance hypothesis in freely moving mice. They showed that an elevated cellular E/I balance in the mPFC significantly impairs social behavior and conditioning, and that this deficit does not result from changes in anxiety or locomotion. Therefore, E/I balance is a critical pathogenesis for explaining various ASD phenotypes. Future research should address the relation between the E/I ratio and the network abnormalities discussed in this paper. Furthermore, the question of whether an increase in cell-packing density and a reduction in cell size are related to E/I ratio should be investigated.

Finally, great variations in the extent of learning disability and mental retardation are found in ASD. The degree of damage to the amygdala or hippocampus has been suggested to determine the severity of learning deficits and mental retardation in ASD (Bachevalier & Loveland, 2006). Although we do not discuss this problem further in this paper, the relation between the degree of brain damage and the severity or characteristics of symptoms of ASD or WS remains to be elucidated. It should be comprehensively analyzed how the activation level and morphology of brain areas, and functional connectivity among them are correlated with psychometrically measured abilities (e.g., perspective-taking, response inhibition, understanding the mental states of others).

Ancillary