The approach: autism as a developmentally emergent difference of neural connectivity

Autism, defined behaviourally in terms of deficits in social, communicative, and stereotyped or repetitive activities, and described psychologically in terms of deficits in central coherence, executive function, or empathising, has always been a behavioural syndrome in search of a biological foundation. My work on the cognitive neurophysiology of autism has contributed to the basis for an emerging theory in which abnormally strong coupling within local neural networks occurs in combination with abnormally weak activity-dependent development of longer-range connections. The resulting weakened coordination of anatomically distributed neural processes has the potential to explain and to unify these disparate psychological descriptions of autism, and has received support from physiological studies both in people with autism and in animal models.

During the past several years autism has entered the scientific limelight, and with this visibility has come a strengthened research focus on autism’s most obvious, most debilitating, and most diagnostic symptoms, namely, its social and communicative deficits. Although these overt symptoms urgently demand exploration, it’s important to approach autism in its complete context as a developmental disorder — developmental not only in terms of nosology but also in terms of ætiology. Autism’s complex abnormalities of social cognition may be most productively considered not as wholly modular capacities disrupted by the functional equivalent of a lesion, but as emergent capacities that are progressively disrupted as brain and cognitive development proceeds. Efforts to unravel autism therefore have much to teach us not only about how such emergent cognitive capacities go wrong during autistic development, but also about how they go right during normal development — and how innate information processing capacities may be elaborated by interactive specialisation.

Recent findings: social and non-social traits within and beyond the autism spectrum

This entwinement of higher-order social cognitive skills with lower-level capacities is illustrated by a couple of our recent results. The first of these findings arises in children with autism spectrum conditions (ASC), in whom a non-social task of visual spatial attention evokes abnormally prolonged frontal activations, the degree of which correlates with psychometric measures of social and communicative deficits. Interestingly, clinically unaffected siblings of children with ASC are not exempted from this frontal abnormality, and its correlation with (mild and subclinical) autistic traits. What does seem to distinguish affected from unaffected siblings, though, is a decrease in long-range functional connectivity between brain regions. This combination of results reinforces the idea that autism may arise when a localised (frontal) abnormality becomes developmentally translated into a generalised disruption of brain connectivity, and suggests an eventual target for preventive therapies.

The second result relating social and non-social aspects arises not within the autism spectrum, but in the context of autistic cognitive traits within the normal population. In undergraduates sampled from a range of major fields associated with non-social, ‘systemising’ traits such as perceptual disembedding and attention to detail (e.g. mathematics, physics), or with social, ‘empathising’ traits such as theory-of-mind and emotion recognition (e.g. literature, theatre), social and non-social sets of autistic traits are partially dependent, and this dependence shows not only in psychometric profiles but in experimental behavioural measures. (An interesting footnote is that whereas women entering systemising fields are characterised by superiorities in systemising, men entering systemising fields are more defined by deficits in empathising.)

In parallel with this main line of work on high-functioning autism, I’ve been using behavioural methods to investigate “low-functioning” (non-speaking) cases of autism, and this work suggests that many people with autism who lack speech are not as “low-functioning” as they may seem. Without any physical support or response-specific prompting, non-speaking children can develop significant degrees of communication by pointing amongst alternative responses, often seeming to use the non-social, scripted activity of stereotypic behaviour as a way to relieve anxiety associated with joint social communication. Practical constraints have confined most studies of autism — and essentially all physiological studies — to the “high-functioning” subpopulation, and it’s unclear to what extent such studies’ results may generalise to non-speaking cases. Behavioural studies that engage non-speaking people with autism therefore are a crucial component of autism research.

Current work: behavioural and physiological investigations across domains and levels of cognition

I’ve recently begun work funded by my NSF Faculty Early Career Development Award — a proposal ranked first amongst all those considered by this past fall’s NSF Development and Learning Sciences panel — to build on these ideas with tests spanning domains of perception, attention, and social cognition within a single sample of children with autism spectrum conditions, using a combination of behavioural and electrophysiological methods. To test this broad range of levels and domains of cognition, I embed a large number of experimental tasks within a suite of video games — a vehicle that combines the precise stimulus timing necessary for experimental control with an entertaining and engaging format the preserves ecological validity. Children take the game home on a laptop computer and play it at leisure, avoiding the all too frequent experimental confound of autistic behaviours with the anxiety induced by novel tasks and environments.

I also use the game in the laboratory, to examine high-density EEG with multivariate analyses in time and frequency domains. EEG’s superior temporal resolution allows examination of functional connectivity on a much finer time scale, and the high-density data that we’re acquiring will yield new information not only about autism but about the cortical generators of evoked potentials in normal individuals. The hypothesis of increased local frontal activity in autism families but decreased functional connectivity in autism probands predicts a corresponding increase in induced gamma EEG power within frontal generators in probands and sibs, but a decrease in phase coherence between generators in probands only. Data acquisition and analysis are proceeding.

Future plans: relating autistic and normal cognitive variations, and genetic analyses

My recent work shows that evaluating unaffected siblings can help illuminate autism by revealing primary, familial abnormalities which aren’t masked by autism’s complex developmental sequelae. In the same way, evaluating autistic traits in the normal population can help discern the relationships amongst these traits, by gleaning information from the entire population distribution rather than just its autistic extreme. In this regard, studies of autistic traits in normal cognitive variation can provide information complementary to that gained from studies within the autism spectrum. As we have a large set of measures and experimental protocols developed in order to examine ASC, applying these to large populations of normal subjects carries very little marginal cost. Our game-based behavioural paradigm, in particular, can be easily downloaded and completed by large numbers of subjects, and I intend to pursue such a large-scale study.

A variety of inherited and de novo chromosomal rearrangements has been reported in individuals with ASC, suggesting that in some cases dosage effects of defined genes can contribute to the phenotype. The proportion of people with ASC who harbour chromosomal anomalies detectable with either standard karyotyping or high-throughput technologies may be as great as 10% to 30% — an incidence much greater than previously thought and one that gives impetus to collection of whole-genome SNP data as an adjunct to autism studies of even modest size. Two general strategies have been successful in screening for genetic variants affecting autism susceptibility: one can work forward by beginning with specific candidate genes and screen a large population for the relevant alleles, or one can work backward from genome-wide screens for deletions, duplications and copy-number variations, identifying the known genes affected and the other genes with which they interact. We are beginning to collect data for this latter strategy, using the Illumina Omni chips. In an exploratory analysis, identified variants in individual genes and in networks of genes will be phenotypically characterised in terms of the behavioural and physiological measures that we have been collecting. Genetic studies have begun to associate specific aspects of the phenotype — such as delayed phrase speech, or repetitive behaviours — with specific loci of linkage. Many of these linkage sites contain identified candidate genes, and many of these tag functional gene networks. As polymorphisms within these genes are identified, their presence in autism probands and families can be related to specific behavioural and neural endophenotypes.

Autism genetics, like autism physiology, has suffered from the distraction of twentieth-century, univariate approaches. Single-gene studies of complex neuropsychiatric disorders are valuable insofar as they highlight points of entry into networks of interacting genes, but it's these network-level disruptions that are the meat of the matter. The risk for autism arises not only in the protein product of an individual gene, but rather in that protein product's many interactions with other genes and proteins which themselves carry individual differences. Many of these interacting susceptibility alleles vary greatly in their prevalance between distinct populations. For example, the G/T SNP rs3813034 in the 3' UTR of SLC6A4 has been selected for the T allele in African and European populations but for the G allele in Asian (Chinese, Japanese, and eastern Indian) populations (Guhathakurta et al., 2008), and in the case of RELN, 16% of Indians but only 6% of Europeans carry an expanded sequence of 11 or more CGG repeats in the 5' UTR (Dutta et al., 2007). These are but two examples of a phenomenon that is likely widespread amongst populations and amongst individual autism risk factors. Because autism susceptibility or protection alleles individually produce only small increments or multiples of risk, we can expect that the risk level and perhaps even the very identity of an allele as an autism risk factor, within a specific population, may depend on the population-genetic context consisting of all the genes with which that putative risk factor interacts. Contrasts between genotypically and (endo)phenotypically well characterised populations from different regions or other stratifications (e.g. castes; see Reich et al., 2009) therefore may have much to say about the nature of the genetic networks involved in autism susceptibility or protection. Candidate-gene studies in eastern Indian populations, for instance (Dutta et al., 2007, 2008; Guhathakurta et al., 2008, 2009; Sen et al., 2010), have revealed tantalising contrasts, both positive and negative, with findings from the more often studied populations of Europe and North America. In a project based in Kolkata, we are beginning to exploring this potential by examining detailed genotype-phenotype correlations in Bengali families affected by autism.

Related work in bioinformatics and in cognitive literary theory

This main line of research is complemented by a couple of further interests, drawing on my initial schooling in computer science and in literature. In computer science, I have a particular interest in algorithms for statistical analysis of biophysical time series and sequences. I’ve developed software to measure spatiotopic modulations of EEG phase-locking, to conduct visual psychophysics and single-unit recordings, to model the psychophysics of human auditory perception, to help assemble genomes from next-generation sequencing technologies, and to analyse fMRI data nonparametrically (part of the widely distributed AFNI software package). Most recently, I’ve directed a team of undergraduate programmers as they have implemented our video-game experimental paradigm. This close connection between neuroscience and applied computer science is a link from which both fields can draw inspiration.

My literary interests relate my neurobiological work to the growing field of cognitive literary studies. I’m interested in how autism’s abnormal neurophysiology affects the ability to construct narrative representations of perceptual experience, and how a great deal of the autistic syndrome, including withdrawal from social contact, impairment in pragmatic language use, and the substitution of repetitive for adaptive behaviours can be understood as compensatory or accommodative responses to this representational challenge. In this regard, much of autistic behaviour is most productively construed as the response of a rational human mind to an abnormal perceptual and cognitive environment. This recognition has profound implications for the relation of autistic cognitive development to normal cognitive development, and for the relation of people with autism to representations of autism in science and society. I’ve described people with autism as “human, but more so” since the extraordinary effort and deliberation with which they approach representational challenges can produce unusually deep insights, as exemplified in the developing genre of autistic memoir. The narrator who is more conscious of the effort of narration can, almost paradoxically, in the end achieve a more fundamental understanding of the characters and events surrounding him or her, precisely because (s)he is so impaired at automatic social perception and must concentrate harder to construct an explicit theory of reality, to piece a story together from perceptual fragments. In literature just as in science, understanding autism helps us to understand humanity.