Connecting autism’s many human faces with invisible biological factors underlying its diversity of symptoms is a grand challenge facing the research community. As noted by a team of investigators led by a BBRF grantee, it is widely assumed that behind the many clinical manifestations of autism are likely disturbances and abnormalities in many different biological pathways.

Poignantly, however, “causal evidence in support of this hypothesis is lacking,” writes the team, led by 2017 BBRF Independent Investigator Alessandro Gozzi, Ph.D., of the Center for Neuroscience and Cognitive Systems, Italian Institute of Technology, Italy, in collaboration with Adriana Di Martino M.D., and her team at the Child Mind Institute. Given this problem, it has made sense for researchers to try to identify robust autism subtypes, in which clusters of clinical symptoms might be matched with specific sets of signals from genetic or neuroimaging studies. To the extent such signals can be reproduced in different studies, they might suggest distinct pathobiological mechanisms underlying each putative subtype. Still, direct evidence in research of this kind “remains elusive,” Dr. Gozzi and colleagues write.

They have taken a distinct “cross-species” research path, to bridge what they call the “critical knowledge gap.” Leveraging recent technical advances in imaging and comparing patterns of functional neural connections across different species, they and other researchers have recently identified “remarkably conserved fMRI connectivity alterations,” i.e., very similar changes in connectivity, in scans of the brains of people with autism compared with those of mice modeling a variety of autism syndromes.

Conducting “large-scale functional neuroimaging across multiple mouse autism models” provided the team with “an unprecedented opportunity to biologically decode autism heterogeneity into causally distinct dysconnectivity signatures,” the team explained. These signatures in turn, provided a basis for the researchers to tentatively identify autism “brain subtypes.”

In explaining the potential import of their new findings, Dr. Gozzi noted that prior attempts to glean biological signals from functional brain-scan data in autism, while important, have generated ambiguous results, showing both increased and decreased neural connectivity in different brain areas. “We think our new study renders the heterogeneity of prior fMRI data from a nuisance into a biologically meaningful signal. Across mice and humans, we have observed two large-scale connectivity patterns that are not reducible to just ‘more’ or ‘less’ connectivity, but that track distinct directions of circuit biology.”

The new paper’s most important finding, reported in Nature Neuroscience, is that fMRI connectivity alterations in 20 mouse models of autism can be clustered into two dominant subtypes: hypoconnectivity and hyperconnectivity. In hypoconnected circuitry, signaling is significantly lower than in the typical brain; in hyperconnected circuitry, signaling is significantly higher. Hypoconnected circuits in the mouse models were traced to a range of biological processes affecting synapses, which connect brain cells. The hyperconnected circuits were related to immune system-related pathways.

Dr. Gozzi’s team found that these two “dominant” types of connectivity changes seen in the fMRI scans of the 20 mouse models mapped directly on to the same kinds of functional connectivity changes seen in fMRI scans of a cohort of 940 individuals with autism.

This may shed new light on a long-held theory of autism pathology. EEG technology, which traces brainwaves made by neurons oscillating at different frequencies, can be used as an index of the balance between excitatory vs. inhibitory neural activity in the brain (often referred to as the E/I balance). Past EEG data has shown both decreased and increased excitatory neural activity in autism patients, a measure that has varied with different study cohorts. This inconsistency, like some of the fMRI data, has perplexed researchers.

Dr. Gozzi explained that his team’s new findings “offer a concrete biological handle on ‘autism is heterogeneous.’ We move from a generic ‘some brains are more connected, some less’ to the two dominant connectivity patterns we saw, involving hypo- and hyperconnectivity.” The cross-species analysis reported in the paper “provides compelling evidence that atypical functional connectivity is a defining pathophysiological hallmark of autism and is associated with distinct signaling pathways. The hypoconnectivity subtype implicates a central role of synaptic dysfunction in altering large-scale fMRI connectivity.” In mouse lines with the hypoconnectivity subtype, observations of decreased density of dendritic spines—nodes of connection between neurons—and specifically in excitatory neurons, “broadly supports our hypothesis and suggests that alterations in synaptic regulation may represent a plausible neurocellular marker for the observed hypoconnectivity.”

Importantly, this is a testable hypothesis—it can be the basis for experiments in autism mouse models, with potentially translatable implications for people with autism, given that the same hypoconnectivity patterns have been identified in fMRI scans of autism patients. Observations in the study linking hyperconnectivity patterns with immune-related signaling have similar potential implications. The suggestion that various immune-mediated pathways converge to drive excessive or aberrant activity in the mammalian brain (i.e., in both mice and humans) points, the team suggests, to mechanisms including immune-mediated alterations of excitatory or inhibitory function, microstructural white matter abnormalities, microglial-induced alterations in axonal wiring and synaptic pruning, and immune-related disruption of synapse formation. “This framework provides a testable mechanistic account for prior human fMRI evidence of hyperconnectivity linked to social deficits in people with autism,” the team noted, and “supports the translational relevance of our findings.”

More broadly, “Our results provide a systems-level framework that aligns with, and extends, the longstanding theory of E/I imbalance in autism.” The possible coexistence of “contrasting excitatory dysfunction within the autism spectrum would be consistent with recent identification of EEG-based opposed autism subtypes characterized by increased and decreased excitability.”

If so, the researchers say, “E/I imbalance may not be a unitary phenomenon but rather a multidimensional construct whose expression can vary across individuals and developmental stages,” potentially explaining the inconclusive and heterogeneous results of past autism-related brain connectivity studies. “This hypothesis warrants further empirical testing in both rodents and humans and could have important implications for autism stratification or therapy.”

Simplifying this, Dr. Gozzi explained: the study’s results “suggest how to stratify autism patients mechanistically, not just in terms of their symptoms. If a person’s fMRI profile looks ‘synaptic-hypoconnected,’ it is plausible that interventions aimed at restoring synaptic function or E/I balance will be more relevant. If it looks ‘immune-hyperconnected,’ approaches targeting neuroimmune signaling or microglial function may be more rational. Even an imperfect ability to enrich patients along these lines would already be valuable for future clinical trials.”

Because the same hypo- and hyperconnected subtypes exist in people and mice, as the study demonstrates, “we can now ask very precise questions: if we normalize synaptic density or E/I balance in a ‘hypoconnected’ mouse model, does the fMRI pattern move toward normal? If we dampen pathological immune activation in a ‘hyperconnected’ mouse model, do we see a corresponding normalization of hyperconnectivity in the fMRI scans? In this way this research can be the basis of a causal validation platform for mechanistically defined autism subtypes.”

Finally, Dr. Gozzi commented: “While no one would claim immediate clinical utility for this work, the trajectory is clear, at least to us: we can now use fMRI not just to show that autism brains are different—which is what we have been seeing for many years—but to carve the spectrum into biologically distinct circuit-level conditions that can be targeted differently. In this sense, we see this as a potential building block for future precision psychiatry in autism, rather than just another set of imaging results.”