Systems Biology of Late-Onset Severe Regressive Autism

Background: Regression is commonly described in ASD.  However, the causes of regressive ASD in the vast majority of cases are unclear.  Furthermore, there is debate regarding the precise definition of regression and whether it is a distinct phenomenon within ASD.

Objectives: To pursue an intense, multimodal investigation of a rare, severe example of this phenomenon, ASD with late-onset severe Regression (ASDR).  Examining an extreme cohort will maximize our ability to identify neurobiological features relevant to regression.

Methods: In addition to meeting criteria for ASD, ASDR subjects met the following criteria:  (1) loss of language skills, (2) loss in at least one other domain:  social skills or adaptive behavior, bowel or bladder control, play, and motor skills, (3) loss must have occurred after 24 months of age, (4) must not have regained level of skill prior to loss, and (5) full-scale IQ<70.  To investigate the genetic contribution to regression, we performed whole-exome sequencing and CNV analyses on 15 ASDR probands and their unaffected siblings and parents to search for novel or rare de novo, recessive, and hemizygous mutations.  To study neural systems, we performed fMRI analyses involving a block design of alternating fearful face versus house images in 4 cohorts:  7 ASDR, 7 low-functioning ASD (LFASD), 14 high-functioning ASD (HFASD), and 19 typically-developing (TD) subjects.  To quantify the social phenotype, we collected eye-tracking data as our 4 cohorts viewed emotional faces:  5 ASDR, 7 LFASD, 40 HFASD, and 11 TD.

Results: Genetic analysis identified one or more novel or rare mutations in almost all of the probands and a diverse list of candidate genes.  Coexpression network analysis showed that a significant number of these genes converge in the cerebellum and thalamus during mid-late fetal development (P=0.009, permutation testing).  fMRI revealed abnormal hyperactivity in these brain regions compared to LFASD, HFASD, and TD subjects.  Eye-tracking results replicated prior findings of decreased fixation on the eyes and increased fixation on the mouth in HFASD relative to TD.  However, the amount of time subjects with LFASD and ASDR spend looking at the eyes is not significantly different from TD.  Among individuals on the autism spectrum, those with ASDR fixated on eyes the most.

Conclusions: The combination of genetic findings and coexpression network analysis predicted a distinct neuroimaging signature in ASDR subjects:  enhanced, face-evoked activity in the cerebellum and thalamus.  The cerebellum is involved in the use of visual social cues for making mental inferences.  Given the role of the thalamus in modulating cortical neural rhythms, abnormal hyperactivity may explain why individuals with ASDR generally have high levels of encephalographic abnormalities, anxiety, and stereotyped patterns of behavior.  The initial normal development of children with ASDR suggests that the early development of the social brain may follow a typical trajectory, raising the possibility that some aspects of social function will be preserved, as reflected by our eye-tracking results.  Although ASDR and other forms of ASD are, eventually, clinically similar, the natural history of ASDR marks a unique set of genetic, neural, and behavioral mechanisms.