Background
Autism spectrum disorder (ASD) is characterized by developmental impairments in social communication and restricted and repetitive behaviors [
1], which often include differences in sensory responsiveness to environmental stimuli1 [
2,
3]. Children with ASD may exhibit patterns of altered sensory reactivity such as hypo-responsiveness [
4‐
6], hyper-responsiveness [
7,
8], and/or sensory “seeking”—unusual interest in or fascination with specific sensory stimuli [
9]. While these patterns manifest with considerable inter-individual variability and are likely to have different underlying genetic [
10] and neural [
11,
12] mechanisms, recent work has begun to describe how these patterns of behavioral reactivity to sensory stimuli co-occur within individuals [
13].
Efforts toward untangling the neural basis of sensory reactivity patterns have focused on auditory and visual modalities, possibly because of their primacy to verbal and nonverbal communication skills affected by ASD. Far less work has been done to understand more proximal sensory systems such as touch, proprioception, and interoception. However, these systems are also of critical importance during social interactions, providing important cues about emotion, attachment, compliance, and intimacy [
14]. Tactile responsiveness patterns specifically are linked to social deficits in ASD [
6] and may have a central role in our earliest experiences of social communication. Specifically, infant–caregiver interactions that lay the foundation for social reward and secure attachment heavily involve the sense of touch [
15,
16], which is at an advanced stage of development relative to other sensory systems in neonates [
17]. It is during this early window in the first year of life that neural differences in ASD begin to emerge [
18]. Thus, the developmental primacy of touch for behaviors relevant to ASD warrants further investigation.
While sensory reactivity is typically assessed with clinical observational measures, discriminative touch has also been increasingly measured in ASD using rigorous psychophysical methods that often give insight into neurobiological mechanisms. While there is not a clear consensus for impaired or enhanced discriminatory ability, some themes have emerged from this work (for a review, see Mikkelsen et al. [
19]). Impaired performance on vibrotactile static detection and discrimination tasks suggests impaired lateral inhibition in the somatosensory cortex [
20,
21], as do impairments in amplitude discrimination and the absence of expected effects of a habituating stimulus [
20,
22]. These sensory indications of impaired cortical inhibitory mechanisms have been tied directly to reduced GABA in the sensorimotor cortex in ASD using spectroscopy [
23] and support the excitatory/inhibitory imbalance hypothesis of ASD [
24,
25]. The well-established autism candidate gene GABRB3 is associated with differences in tactile sensitivity [
26,
27], providing further support for a role of GABA dysfunction in the somatosensory differences common in individuals with ASD.
Distinct but overlapping neural pathways support the discriminative and affective aspects of touch. Extensive work suggests that discriminative touch (e.g., touch used to determine the shape or texture of an object) is primarily mediated by the thalamocortical projections to the somatosensory cortex [
28‐
30]. However, a separate system that is believed to mediate affective touch (e.g., touch used to elicit an emotional response or to communicate social affiliation) has recently been described. In this system, small-diameter, unmyelinated peripheral fibers, known as C-tactile (CT) afferents, respond preferentially to slow, stroking touch with light to moderate pressure, suggesting they are “tuned” to social/affective (in contrast to discriminative) touch [
31‐
33]. Functional magnetic resonance imaging (fMRI) studies in patients lacking large-diameter myelinated tactile afferents demonstrate that CT fibers project selectively to posterior insular cortex [
32,
34], a cortical target of multimodal sensory input that is associated with changes in affective state, including visceral sensation, temperature, and pain, further distinguishing this affective touch system from discriminative touch.
The insula is part of a complex cortical structure with a heterogeneous functional anatomy along its anterior-posterior axis. While the posterior insula receives somatosensory and visceral input, the anterior insula comprises heavily reciprocal projections with prefrontal and limbic regions [
35,
36]. This functional organization may be the basis of a caudo-rostral hierarchical processing stream by which sensory cues are received and integrated with emotional signals to form progressively higher-order representations of proximal (tactile and interoceptive) sensory information and its affective significance [
37]. This model is supported by fMRI evidence of strong connectivity between the anterior and posterior insula [
38]. There is growing evidence that the anterior insula, whose role as a hub of the salience network is to engage neural networks in response to emotionally important sensory stimuli [
39], is affected in ASD [
40,
41]. However, the connectivity between the affective anterior insula and sensory posterior insula has not been examined in ASD, nor has the relation between the insula and sensory-affective sequelae that are increasingly considered foundational to the disorder [
42,
43].
In this study, we hypothesized that the integrity of structural connectivity in somatosensory and insular regions associated with discriminative and affective touch, respectively, would relate to observed sensory behaviors in a standardized assessment of tactile discrimination and affective responses to touch in young children with ASD. Given the functional posterior-anterior gradient of the insula and the clear differences in affective touch perception [
8,
44] and interoception [
45‐
47] in ASD, we specifically examined the structural integrity of intrainsular white matter as a neural substrate for affective response to touch in young children with ASD. As a comparison, we also examined a thalamocortical tract between the ventroposterolateral (VPL) thalamus and primary somatosensory cortex (SI), which is associated with discriminative, rather than affective, touch processing. With a growing body of literature describing differences in discriminative [
19] as well as affective response [
44] to touch in ASD, we hypothesized that structural connectivity in multiple tracts might differ in ASD [
48,
49]. However, given the functional roles of these regions, we predicted that connectivity between the anterior and posterior insula would specifically relate to aberrant
affective response to touch in individuals with ASD, while connectivity between the VPL thalamus and SI would uniquely relate to aberrant touch
discrimination in individuals with ASD.
Discussion
The goal of this study was to assess the integrity of structural connectivity in somatosensory regions associated with affective and discriminative touch [
31] in young children with ASD and to relate it to observed sensory behaviors in a standardized assessment of discriminative and affective responses to touch (Baranek, 2010, unpublished manual). Using the TDDT-R, we observed widespread differences in responses to both affective and discriminative touch in young children with ASD, as has been reported previously [
6,
8,
10,
11,
20‐
22,
71‐
73]. In the ASD group, we found reduced integrity (increased MD) of the white matter tract connecting the posterior and anterior insula. Based on the roles of these two regions in somatic sensation [
32] and affective evaluation [
39], respectively, we hypothesize that this tract is likely important for the emotional evaluation of somatic sensory input. The posterior insula receives input from CT afferents [
32], a peripheral system that mediates affective touch by responding preferentially to slow, gentle stroking touch on hairy skin. CT afferents are absent from the palmar surface of the hand, suggesting they do not have a role in discriminative touch [
74]. Emerging research suggests that individuals with ASD or heightened autistic traits show diminished response to affective touch all along the anterior-posterior axis of the insula [
75‐
77]; the reduced integrity of the intrainsular tract shown by the current study may be a structural basis for this diminished response.
We noted disparate relationships in the two groups between sensory seeking and the integrity of the intrainsular tract, such that higher levels of tactile seeking were associated with
more intact (lower MD) intrainsular white matter in the ASD group and with
less intact (higher MD) in the TD group. The frequency of sensory seeking is low in the TD group, and thus, the association with compromised white matter in the insula may reflect aberrant affective responses to touch in typical children and is likely to be of an entirely different etiology than sensory seeking in ASD. The link between sensory seeking and better white matter integrity in the ASD group is intriguing given the debate about the valence of affect in sensory seeking behavior in ASD [
9]. While the repetitive nature of the behavior suggests that it is intrinsically reinforcing and there is empirical [
78] and autobiographical [
79] evidence for association with positive affect, there is also evidence linking the behavior with negative [
80] or neutral [
9] affect. The association of tactile seeking with greater integrity of a neural pathway that supports positive affective touch supports, but does not prove, the view that sensory seeking reflects positive affect.
It has been hypothesized that sensory seeking behaviors may reflect a compensatory strategy for reduced sensory input [
81] or, alternatively, may serve as a coping strategy in response to overwhelming sensory experiences [
82]. The possibility that tactile seeking is a compensatory strategy for diminished sensory input may be consistent with association between increased seeking and reduced integrity of the intrainsular tract in the TD group. However, in the ASD group, increased seeking was associated with
greater integrity of the intrainsular tract. This unexpected profile in ASD—of greater sensory seeking behaviors relating to better integrity of the intrainsular tract—suggests that a more typical level of throughput of sensory information to the salience network may increase the salience of the input disproportionately in ASD, leading to repetitive engagement with a sensory stimulus. Thus, for those individuals with ASD for whom the intrainsular tract is relatively more intact, seeking behaviors could be more effective at increasing sensory input, and thus more self-reinforcing—manifesting as increased intensity or frequency of seeking in a positive feedback loop. This is consistent with the ideas that sensory seeking may overlap with repetitive behaviors more broadly [
6,
83] and that many repetitive behaviors appear to be reinforced by affective and reward circuitry [
84,
85]. The relatively binary distribution of seeking behavior and its association with a neurobiological variable in our sample suggests its potential for distinguishing meaningful subtypes based on sensory reactivity. Assessing such a subtype in a larger sample might provide additional information about the adaptive function of sensory seeking.
We also observed an association within the TD group between intrainsular tract integrity and overall tactile defensiveness. Somatic input to posterior insula is not limited to the pleasant touch described by Olausson and colleagues; the region responds to interoceptive input [
86], unpleasant and painful touch as well [
87‐
89]. Higher defensiveness in the TD group was associated with lower MD, which reflects greater integrity between the posterior and anterior insula. Thus, it is plausible that for TD children, enhanced throughput of unpleasant feelings from the posterior insula to the anterior insula resulted in greater tactile defensiveness. This result should be interpreted cautiously, however, since defensiveness was infrequent in the TD group relative to the ASD group and we did not observe a similar association in the ASD group. The specificity of association between intrainsular connectivity and (presumed) pleasant affective response to touch (seeking) in the ASD group to sensory seeking is of note, suggesting that different neural pathways may contribute to the perception of touch as either pleasant or unpleasant in the altered sensory experience of individuals with autism.
Finally, we did not find any associations between discriminative touch (form perception) and either thalamocortical or intrainsular MD. Indeed, we would not have expected to observe associations between discriminative touch and intrainsular integrity, given that somatosensory projections to the posterior insula are limited to affectively relevant inputs, such as social touch [
31] and interoception [
37]. While we did predict that group differences in this kind of discriminative touch may be modulated by thalamocortical tract integrity, other neural correlates of form discrimination have also been described [
90]. A recent study implicates the serotonergic system in the differentiation of affective versus discriminative touch [
91]. Given the importance of serotonin for modulating sensory cortical responses [
92,
93], the implication of serotonin in ASD [
94,
95], and evidence that the variation in this system specifically impacts somatosensory processing in ASD [
10], the relations between autism, serotonin, and affective responses to touch, merit further investigation. Given previous work implicating GABA in altered touch perception [
23,
26] in ASD and the modulatory role of serotonin on GABA signaling [
96], the interaction of these two neurotransmitter systems in the context of somatosensory perceptual differences in ASD also warrants further study.
Our study had several important strengths, including the use of a standardized observational measure of tactile responsiveness that included quantification of both discrimination and affective response, a high angular resolution DTI sequence that is optimal for tractography, a rigorous QA procedure, and a relatively large and young sample of children with ASD. We utilized a narrow age band of 5–8 years, which is well before pubertal changes in white matter, increases the homogeneity of the sample, and is the earliest developmental look at these tracts in school-age children with ASD. Our study had limitations as well, including the inability to extrapolate our results beyond this narrow age band or to individuals with ASD who could not complete an MRI scan. Ceiling effects or limited variability in some TDDT-R scores for the TD group for some variables may have also hampered our ability to find additional brain–behavior correlations. Finally, while the narrow age band was a strength in some respects, it also limits the ability to extrapolate our findings to the broader population of individuals with ASD. Future research should focus heavily on characterizing behavioral and neural responses to affective touch earlier in development, as the primacy of touch in infancy for a foundation of social reward is well established [
97], but not well studied in infants at risk for developing ASD. Understanding how the intersection of perception and affect—both of which have fairly well-characterized neural circuitry—gives rise to the more complex behavioral symptoms of ASD will also depend on strong translational ties between basic and clinical neuroscience. Ultimately, the combination of prospective longitudinal studies of at-risk infants and better cross-talk between basic and clinical researchers is expected to have a strong impact on understanding the pathophysiology of ASD and advancing evidence-based treatment approaches [
43].
Acknowledgements
The authors thank Dr. Norman Farb for providing the intrainsular ROI tracing protocol and Dr. Grace Baranek for providing the laboratory for training and guidance in the use of the TDDT-R.