Introduction
Autism spectrum disorder (ASD) is a neurodevelopmental disorder characterized by social, behavioral and cognitive impairments (American Psychiatric Association,
2013). The classical categorical system of diagnosing pervasive developmental disorders (i.e. autistic disorder, Asperger’s disorder, pervasive developmental disorder not otherwise specified, childhood disintegrative disorder, and Rett’s disorder) as found in the 4th edition of
Diagnostic and Statistical Manual of Mental Disorders (DSM-IV-TR) (American Psychiatric Association,
2000), was collapsed into a single dimensional diagnosis of ASD in the 5th edition (DSM-5). The rationale behind this dimension collapse is that the core symptoms exhibited by individuals with ASD are shared across the previous categories, but within a severity degree spectrum.
Core symptoms exhibited by an individual with ASD consist of restricted, repetitive, and stereotyped patterns of behavior, as well as impairment in social communication and interaction. A diagnosis of ASD follows a set of criteria based on non-biological, clinically evaluated symptoms (American Psychiatric Association,
2013). There are a number of major cognitive hypotheses for ASD: the ‘Theory of Mind (ToM) dysfunction’ hypothesis (Baron-Cohen et al.,
1985), the ‘executive dysfunction’ hypothesis (Ozonoff et al.,
1991), the ‘weak central coherence’ hypothesis (Frith,
1989), and the ‘empathizing-systemizing’ hypothesis, also known as the ‘extreme male brain’ theory (Baron-Cohen,
2009). To master ToM, joint attention and cognitive and emotional empathy are required (Baron-Cohen et al.,
1985). Impaired executive function (such as inhibition control, working memory, cognitive flexibility, and planning) is thought to aggravate non-social symptoms (Ozonoff et al.,
1991). A more prominent low-level rather than high-level processing system might explain ASD individuals’ improved ability in quantitative tasks, relative to those requiring central coherence, such as visuospatial, auditory-verbal and perceptual tasks (Frith,
1989). Finally, sex differences in ASD prevalence, with the male to female ratio in ASD being 3:1 (Loomes et al.,
2017), and in the behavior of typically developing controls (TC) correlate with ASD features. Namely, higher systemizing and lower empathizing ability is common in ASD (vs. TC) and in TC males (vs. TC females). As such, it is hypothesized that individuals with ASD present a shift in the ‘empathizing-systemizing’ continuum towards the systematizing ability (i.e. having a brain more similar to an ‘extreme’ TC male brain) (Baron-Cohen,
2009).
To validate and biomark cognitive-behavioral features of ASD, neuroscientific hypotheses supported by neuroimaging have also been put forward. The emerging, and putatively overarching, ‘disrupted connectivity’ (neuroscientific) hypothesis of ASD (Vasa et al.,
2016) proposes that clinical symptoms exhibited by ASD individuals have their origin in the way the brain organizes and synchronizes its regions, and is well poised to account for the four above-mentioned cognitive hypotheses and other neuroscientific hypotheses of ASD, such as the ‘salience network dysfunction’ hypothesis (Toyomaki & Murohashi,
2013). This hypothesis postulates that the disrupted connectivity between the salience network (responsible for stimuli salience attribution) and the systems receiving processed stimuli information [the default mode (DM) and executive control networks] leads to social impairments in ASD. Disrupted connectivity can be tested in a resting-state functional magnetic resonance imaging (rs-fMRI) study, such as the present one. A resting-state approach is essential for an unbiased and more comprehensive perspective on brain function which is free from the influence of task-specific confounders, such as task performance differences between ASD and TC (Bressler & Menon,
2010). In this approach, functional brain connectivity (FBC) is measured as the correlation between the spontaneous activity of several brain regions and is compared between ASD and TC. More specifically, one can measure FBC within individual resting-state networks (RSNs) which resemble known spatial topographies of brain activation attributed to task approaches, for example: salience, visual, language, and DM networks (Damoiseaux et al.,
2006; Shirer et al.,
2012; Smith et al.,
2009). Individuals with ASD, across the lifespan, have shown abnormal resting-state brain connectivity within the DM network, predominantly decreased, but also sometimes increased (Hull et al.,
2017; Nair et al.,
2020). The anterior salience (AS) network within-connectivity has been reported to be increased in children, but decreased in adolescents and adults (Uddin,
2015), which is consistent with the salience network dysfunction (neuroscientific) hypothesis of ASD (Toyomaki & Murohashi,
2013; Uddin & Menon,
2009). Additionally, there is some evidence for reduced long-range and increased short-range connectivity across the lifespan (Rane et al.,
2015), consistent with the weak central coherence (cognitive) hypothesis of ASD. Most of these studies performed a within-network FBC analysis using a seed-based (hypothesis-based) approach.
The majority of functional connectivity studies in ASD are based on samples composed mainly, or only, of males, most likely due to the unbalanced sex ratio of ASD (Loomes et al.,
2017). Nevertheless, the few studies which include both sexes have shown sex-specific differences (i.e. between ASD and TC) in regional functional connectivity, thereby providing growing evidence of an overall hyper-connectivity in females and hypo-connectivity in males, compared to TC (Alaerts et al.,
2016; Lawrence et al.,
2020; Smith et al.,
2019; Ypma et al.,
2016), albeit the opposite has also been reported once (Yang & Lee,
2018). For example, when comparing ASD with TC, the connectivity between the cerebellum and several cortical regions (i.e. the bilateral fusiform, the middle occipital, the middle frontal, the precentral gyri, the cingulate cortex, and the precuneus) was found to be increased in females and decreased in males across the lifespan (Smith et al.,
2019). However, when compared to TC males, ASD males have shown
increased connectivity between brain regions involved in mentalizing processes (i.e. the bilateral temporal-parietal junction), whereas ASD females have shown decreased connectivity (i.e. the medial prefrontal cortex, precuneus, and right temporal-parietal junction) (Yang & Lee,
2018), in a sample of adolescents. The within DM network connectivity has also been shown to be decreased across lifespan in ASD females, when compared to TC females (Ypma et al.,
2016). Furthermore, when analyzing diagnosis-specific sex effects on seed-based functional connectivity in children and adolescents, ASD girls have shown increased connectivity between the posterior cingulate cortex (that belongs to the DM network) and the left posterior parietal cortex (that belongs to the central executive network) compared to ASD boys, but with no difference between sexes in TC (Lawrence et al.,
2020). Interestingly, TC girls have shown decreased connectivity between the right frontoinsular cortex and the anterior cingulate cortex (that belong to the salience network) compared to TC boys, with no difference between sexes in ASD (Lawrence et al.,
2020). Moreover, the within DM network connectivity has been shown to be decreased in TC males when comparing to TC females—with the decrease in connectivity associated with a poorer performance on a mentalizing task (Ypma et al.,
2016) in a sample composed of children, adolescents and adults. Furthermore, the main effect of sex on within-network functional connectivity using a sample of children and adolescents with ASD and TC has been explored once, with girls showing increased connectivity within the DM network compared to boys (Olson et al.,
2020).
More specifically, a few authors have examined FBC
between-RSNs using a hypothesis-
free approach (Bos et al.,
2014; Cerliani et al.,
2015; Nomi & Uddin,
2015; Oldehinkel et al.,
2019; Olson et al.,
2020; von dem Hagen et al.,
2013), as we have done in this study. Compared to TC, ASD boys have shown decreased FBC between the executive control network and a network including the cingulate gyrus, in a sample composed of male children and young adolescents (Bos et al.,
2014). FBC between salience and DM networks has also been shown to be decreased in male adults with ASD (von dem Hagen et al.,
2013). When using a mixed-sex sample, ASD has shown decreased FBC between DM and precuneus (in children) and basal ganglia (BG) networks (in adolescents), whereas no differences were found in between-RSN FBC in adults (Nomi & Uddin,
2015). Increased FBC between BG and primary sensory [such as primary visual (PV), auditory and sensorimotor] networks and decreased FBC between auditory and sensorimotor networks were also shown in male children, adolescents and adults with ASD (Cerliani et al.,
2015). When using a mixed-sex sample of children, adolescents and adults, the FBC between visual network and somatosensory and motor networks was shown to be decreased in ASD, whereas the FBC between cerebellar and sensory (including auditory, language, visual, and somatosensory networks) and motor networks was shown to be increased in ASD (Oldehinkel et al.,
2019). Finally, in a study developed in parallel to ours, the main effect of ASD diagnosis and sex and the ASD diagnosis by sex interaction effect on the FBC between-RSNs were explored using a sample of children and adolescents (Olson et al.,
2020). All effects were reported to be non-significant (i.e. after correction for multiple comparisons).
Remarkably, the above between-RSN FBC findings are inconsistent, possibly due to different subject inclusion choices, such as using: (a) mixed age groups [children, adolescents and adults Cerliani et al.,
2015; Oldehinkel et al.,
2019), children and adolescents (Bos et al.,
2014; Olson et al.,
2020), or only children or adolescents or adults (Nomi & Uddin,
2015; von dem Hagen et al.,
2013)]; or (b) mixed sex groups (Nomi & Uddin,
2015; Oldehinkel et al.,
2019; Olson et al.,
2020) or only male subjects (Bos et al.,
2014; Cerliani et al.,
2015; von dem Hagen et al.,
2013). Indeed, previous between-RSN FBC studies have shown differences between ASD and TC populations across lifespan to be age-dependent, in particular, showing decreased subcortico-cortical connectivity with age (Cerliani et al.,
2015) and showing differences in between-RSN FBC (see above) in children and adolescents, but not in adults (Nomi & Uddin,
2015). Additionally, there is also growing evidence that FBC is influenced across the lifespan by sex, both in healthy subjects (Gong et al.,
2011; Stumme et al.,
2020; Zhang et al.,
2016) and in individuals with ASD (Lai et al.,
2017; Olson et al.,
2020). Therefore, it is important that age and sex are accounted for when performing group comparisons based on FBC measures. Furthermore, only one of the previous between-RSNs studies (Olson et al.,
2020), recently published, has explored if ASD-associated effects on between-RSNs FBC vary depending on the sex of the individuals.
In addition to the functional connectivity findings, task-based functional disturbances in ASD, compared to TC, have also been reported, such as: (a) decreased activation in the medial prefrontal cortex, the superior temporal sulcus, the anterior insula, the anterior cingulate cortex and the amygdala during social processing across the lifespan (Adriana Di Martino et al.,
2009; Hernandez et al.,
2015); (b) increased activation in BG during cognitive control in adults (Prat et al.,
2016), possibly as a compensatory mechanism for cortical malfunction (Subramanian et al.,
2017); and (c) a desynchronization of brain regions in language processing in adults (Dichter,
2012). Furthermore, some of the above effects have shown to be modeled by sex. In particular, the decreased activity in the posterior superior temporal sulcus during social processing is present in males, but not in females (when comparing ASD with TC in adults) (Kirkovski et al.,
2016). During an empathy task using a sample of adults, ASD males have shown increased activation in the medial frontal gyrus compared to ASD females (an effect not present in the TC group); and ASD females have shown decreased activation in the midbrain and limbic regions compared to TC females (an effect not present in males) (Schneider et al.,
2013).
In light of the overarching disrupted connectivity hypothesis of autism, and given the above lack of consistency in the literature, we sought to investigate: (1) the main effects of ASD (where we expect to replicate a few previous reports) and of sex on between-RSNs FBC [where we expect to replicate two studies—performed in parallel to our own, one in older TC adults (Stumme et al.,
2020) and one in children and adolescents with ASD and TC (Olson et al.,
2020)]; and (2) if and how sex influences ASD diagnosis effects (i.e. a diagnosis by sex interaction) on the between-RSNs FBC [attempting to corroborate work which has been performed once, in parallel and independently (Olson et al.,
2020), using a sample derived partially from the same original database (ABIDE; see below)]. In the present study, we carefully considered subject eligibility and data acquisition choices to avoid confounding or noise-contributing effects of age, intelligence quotient, handedness, and eye state (open vs. closed) at scan. We compared female and male samples of children, adolescents and adults with ASD with age-matched TC and applied independent component analysis (ICA) on rs-fMRI data. ICA is a fully data-driven method able to identify brain regions that function in a temporally synchronized manner which, in rs-fMRI studies with healthy subjects, correspond to RSNs that resemble task-based functional brain activation (e.g. PV, sensorimotor and executive control networks) (Damoiseaux et al.,
2006; Shirer et al.,
2012; Smith et al.,
2009).