The SN dysfunction hypothesis is a neurocognitive model based on a systematic approach that combines findings in the field of neuroanatomy, neurophysiology, and psychology. This hypothesis is based on a model that has successfully described the confluent relations of multiple neural circuits by analyzing the networks between brain regions by using recently developed functional MRI analytical methods. These neural circuits encode the saliency of sensory experiences predominantly associated with the AI, and exclusively control both the DMN, which is associated with self-awareness, and the ECN, which is associated with cognitive processing (see Figure 1). One proposal introduced by Uddin and Menon (2009) describes the SN (particularly the AI) as dysfunctional in ASD and dysregulated in specific cognitive processes that enable adaptation to a social context. Moreover, the AI has many functional roles for multimodal sensory processing, emotional cognition, and social cognition, among others. Thus, the SN dysfunction in ASD might be associated with abnormal sensory processing and may secondarily induce atypical cognitive processes, including social cognition. There is a strong expectation that the SN dysfunction hypothesis may become an important pathophysiological hypothesis in ASD. In this section, we take a closer look at the proposal provided by Uddin and Menon (2009) and present supportive evidence from multiple perspectives.
The insular cortex encodes the intensity of sensory experience
The SN has a neuronal connection with a central focus on the AI. We present several findings of the AI regarding its intensity in sensory experience and its integration in multimodal sensory processing. The insular cortex is located deep within the lateral sulcus between the frontal, temporal, and parietal lobes and processes multimodal and cognitive information, including visceral sensory, visceral motor, vestibular, pain, temperature, language, visual, auditory, and tactile information (Augustine, 1996). The insular cortex encodes the intensity of a sensory experience. Bornhovd, Quante, Glauche, Bromm, Weiller, and Buchel (2002) demonstrated that the response of the insular cortex to a pain stimulus showed a positive linear relation with stimulus intensity. Recently, several studies have discussed that the insular cortex might be “a multimodal magnitude estimator,” similar to a salience detector that triggers subjective evaluation and attention (Baliki, Geha, & Apkarian, 2009; Moayedi & Weissman-Fogel, 2009). In particular, the AI, which is an anterior part of the insular cortex, has been recognized as a center of integration between emotional expression and sensory experience (Kober, Barrett, Joseph, Bliss-Moreau, Lindquist, & Wager, 2008). For example, the insular cortex contributes to the shaping of emotional experience from the bodily state, similar to the autonomic nervous system (Damasio, 1996).
The AI contributes to emotion recognition and social cognition
A meta-analysis of a great majority of functional neuroimaging studies indicates that the AI is consistently activated during the expression of emotion, including various negative and positive emotions such as anger, sadness, fear, disgust, happiness or joy, trust, and surprise (Kober et al., 2008). It has been recognized that the AI integrates multimodal sensory information and conscious emotional recognition. In the field of social neuroscience, several studies have demonstrated that the AI is involved in the affective component of empathy and social pain, which are processes required to understand another's emotion by sharing in their own affective states (Bernhardt & Singer, 2012; Jackson, Meltzoff, & Decety, 2005; Singer, Seymour, O'Doherty, Kaube, Dolan, & Frith, 2004). In association with this finding, Singer et al. (2004) showed that while the posterior insula was activated when subjects received painful stimulation, the AI and anterior cingulate cortex were both activated when the subject received pain and when the subject witnessed a loved one receiving pain. The study concluded that the AI constituted a common neural base for our understanding of the feelings of others (empathy) and ourselves. Very few attempts have been made at identifying the neural substrate of emotional recognition in the insular cortex in ASD studies. For example, Silani, Bird, Brindley, Singer, Frith, and Frith (2008) indicated that difficulties in emotional awareness were related to the hypoactivity of the AI in both patients with ASD and typical developmental controls. The poorer the awareness of one's emotions, as well as the emotions of others, the weaker the activity observed in the AI (Silani et al., 2008). They interpreted this difficulty in emotion recognition as alexithymia and subsequently revealed a correlation between hypoactivity of the AI to emotional stimulus and alexithymic traits (Silani et al., 2008). Alexithymia is a state of deficiency in the understanding of one's emotions and in the ability to express emotions. Individuals with alexithymia lack the ability to imagine and, despite dwelling on their own situation, they find it difficult to verbally express the emotions that accompany such situations; this affects their interpersonal relationships. Furthermore, Bird, Press, and Richardson (2011) replicated these findings and demonstrated that the response of the AI to social pain was correlated with the degree of alexithymia, but was not an autistic trait.
In conclusion, the AI contributes to an integrated emotional awareness based on multimodal sensory information and applies these common neural substrates to understand one's and other's emotional state. The degree of evaluating the intensity of a sensory experience in the AI might affect the cognitive aspects of causal attribution in the emotional state and subsequently contribute to one part of the difficulty in empathizing faced by patients with ASD.
Supportive evidence for dynamic switching of the SN
The SN, which Uddin and Menon (2009) had initially proposed, consists of mainly the AI and anterior cingulate cortex, which was identified by a network analysis using fMRI. Seeley, Menon, Schatzberg, Keller, Glover, Kenna, Reiss, & Greicius (2007) conducted independent component analyses of resting-state fMRI data and extracted an independent brain network consisting of the AI, anterior cingulate cortex, and subcortical structures, including the amygdala, substantia nigra / ventral tegmental area, and thalamus. Uddin and Menon (2009) called this network the SN on the basis of the functional role of the AI. The SN is distinct from the two other large-scale brain networks, ECN and DMN. The SN first evaluates any changes in the physiological homeostasis that occurs as a result of body sensation or other sensory processing and triggers the appropriate adaptive behavior (Eckert, Menon, Walczak, Ahlstrom, Denslow, Horwitz, & Dubno, 2009).
The ECN is a functional integration of brain regions focused on both the dorsolateral prefrontal cortex and the posterior parietal cortex, and is a network involved in various cognitive processes, such as working memory and behavioral control. The ECN has shown a strong intrinsic functional coupling and strong coactivation during various cognitive tasks associated with information processing cognitive functions, including goal-oriented behavior, and the flexible switching of working memory and problem solving (Koechlin & Summerfield, 2007; Miller & Cohen, 2001; Muller & Knight, 2006; Petrides, 2005).
The DMN is a functional integration focused on the ventromedial prefrontal cortex (VMPFC) and the posterior cingulate cortex (PCC) and is characterized by a coherent neural oscillation at a low frequency rate in the resting state (Deco, Jirsa, & McIntosh, 2011). The precise functions of the DMN are still largely unknown; however, the individual brain regions comprising it are hypothesized to be involved in the integration of autobiographical, self-monitoring, and related social cognitive functions (Spreng, Mar, & Kim, 2009). The DMN includes the medial temporal lobes, angular gyrus, the PCC, and the VMPFC. The PCC is activated during tasks that involve autobiographical memory and self-referential processes (Buckner & Carroll, 2007); the VMPFC is associated with social cognitive processes that are related to self and others (Amodio & Frith, 2006); the medial temporal lobe is engaged in episodic and autobiographical memory (Cabeza, Prince, Daselaar, Greenberg, Budde, Dolcos, LaBar, & Rubin 2004), and the angular gyrus is implicated in semantic processing (Binder, Desai, Graves, & Conant, 2009). During the performance of cognitively demanding tasks, the ECN typically shows increases in activation, whereas the DMN shows decreases in activation (Greicius, Krasnow, Reiss, & Menon, 2003; Greicius & Menon, 2004; Raichle, MacLeod, Snyder, Powers, Gusnard, & Shulman, 2001). As the activation of the DMN temporally fluctuates with a negative correlation to signals from within the region that includes the network stimulated during cognitive problem solving, the DMN and the ECN might contribute to different aspects of both social and nonsocial cognitive processing. Currently, the DMN has garnered much attention in different fields and is being extensively studied in areas including psychology, neuroscience, and psychiatry (Broyd, Demanuele, Debener, Helps, James, & Sonuga-Barke, 2009). Among these studies, ASD studies have shown multiple cases of reduced DMN activity (Ebisch, Gallese, Willems, Mantini, Groen, Romani, Buitelaar, & Bekkering, 2011). According to Assaf, Jagannathan, Calhoun, Miller, Stevens, Sahl, O'Boyle, Schultz, and Pearlson (2010), the strength of the DMN integration within cases of ASD has weakened, thereby correlating it with a measure of sociability.
Regarding the integration of the three functional networks described above, studies have found confluent relations between brain regions using recent functional imaging and have shown that the AI is not only stimulated prior to reaching the ECN and the DMN, but is also involved in driving these networks, where the ECN and the DMN are exclusively activated. Sridharan, Levitin, and Menon (2008) have shown that across three independent data sets, the right AI plays a critical and causal role in switching between two other major networks (the ECN and the DMN), which are known to demonstrate competitive interactions during cognitive information processing (Fox, Snyder, Vincent, Corbetta, Van Essen, & Raichle, 2005; Greicius et al., 2003). The study used Granger causality analyses to examine the directionality of the effect of the AI and anterior cingulate cortex nodes of the SN on other brain regions. Granger causal analyses enabled the detection of causal interactions between brain regions by assessing the extent to which signal changes in one brain region can predict signal changes in another brain region (Goebel, Roebroeck, Kim, & Formisano, 2003). Across stimulus modalities, the right AI plays a critical and causal role in activating the ECN and deactivating the DMN (Sridharan et al., 2008). This study also showed that the right AI is involved in switching between brain networks across task paradigms and stimulus modalities, and thus acts as a causal outflow hub that coordinates between two major large-scale networks. Latency analysis, which includes measures of the time to peak, further confirmed that the right AI activity temporally precedes the activity in the other nodes of the ECN and the DMN. This new understanding of the right AI as a critical node for initiating network switching provides a key insight into the core functions of the AI.
Abnormal SN processing, such as chronic decreased AI activity, may induce an inappropriate neural response and cognitive processing in response to a cognitively challenging task. In ASD, a comprehensive meta-analysis of functional neuroimaging studies of social processing demonstrated that, across a group of studies examining various aspects of social processing, one of the regions that consistently showed a significant hypoactivity in individuals with ASD was the right AI (Di Martino, Ross, Uddin, Sklar, Castellanos, & Milham, 2009). This may underlie the consequence of ineffective salience processing in the AI in response to reduced attention to social stimuli, a hallmark of ASD.
Psychological interpretation of SN dysfunction in ASD
From a psychological perspective, the SN is a process that motivates coping behavior in response to the environment by evaluating the intensity and saliency of the emotional response associated with sensory experiences. Uddin and Menon (2009) indicated that the SN chronically decreases and does not function to evaluate emotional salience in ASD. Moreover, because of the dysregulation resulting from a switch between the DMN and ECN, patients with ASD cannot execute appropriate coping behavior. Thus, we speculate how the SN dysfunction contributes to the clinical presentation of patients with ASD from a psychological perspective.
As previously mentioned, activation of the AI in response to an emotional stimulus is not always reduced in ASD, but rather reflects an alexithymic trait and difficulty to empathize (Bird et al., 2011; Silani et al., 2008). Although psychological studies have found that the processing of empathy is weaker in patients with ASD compared with neurotypical individuals, the activity of the AI does not predict the degree of the autistic trait. Empathy is divided between the cognitive processes that deduce the mental state of another individual and the processes of reproducing an emotional state. The AI responds to empathic processing by contributing to the coding of emotional responses, which simulates the state of another person. Thus, understanding one's own emotional state and understanding another individual's emotional state are both processed using common information processing and neural bases. Some studies have called these neural substrates “shared-network” models of empathy (Bird, Silani, Brindley, White, Frith, & Singer, 2010; Preston & de Waal, 2002). Clinically, individuals with ASD show difficulty in understanding the state of another individual and even have difficulty in understanding themselves. Alexithymia is a state of deficiency in the understanding of one's own emotions and in the ability to express these emotions. Alexithymia was originally thought to predominantly exist in patients with somatoform disorder; however, the difficulty in self-emotional awareness is relevant to ASD. An assessment of alexithymia is generally conducted using a self-report questionnaire known as the 20-item Toronto Alexithymia Scale (TAS-20; Bagby, Parker, & Taylor, 1994; Bagby, Taylor, & Parker, 1994). There have been a few studies using the TAS-20 in which patients with ASD have demonstrated a high tendency to exhibit an alexithymic trait (Berthoz & Hill, 2005; Hill, Berthoz, & Frith, 2004). However, these two studies did not quantitatively evaluate the autistic tendency and the correlation with the alexithymic trait was unclear.
An fMRI study conducted by Silani et al. (2008) also used the TAS-20 questionnaire. They compared the results of individuals with high-functioning autism / Asperger syndrome and a group of neurotypical individuals and revealed that the high-functioning autism / Asperger syndrome group demonstrated a higher tendency toward alexithymia. In addition, the alexithymia tendency of both groups showed a significant correlation with their empathetic abilities, with a negative correlation found in AI-stimulated individuals as assessed using fMRI in subjects viewing an emotion-evoking slideshow. The AI is involved in the monitoring of emotions, regardless of whether they are one's own or another's emotions, and thus, the AI does not properly function in individuals with ASD because of this fundamental problem. These individuals find it difficult to understand emotions in both themselves and in others.
Thus, the activity of the AI itself partially contributes to social difficulty in individuals with ASD, and the neural network (SN), including the AI, dynamically affects various aspects of social cognition. There is a strong expectation that hypoactivity of the AI contributes to difficulties in emotion recognition such as alexithymia, which is not an autistic trait, and the subsequent SN dysfunction induces a comprehensive autistic clinical representation due to a dysregulated ECN and/or DMN. From a psychological perspective, the SN dysfunction hypothesis may support hierarchical processes for emotion recognition.