Recent research published in natural neuroscience provide evidence that the biological diversity of autism can be classified into different brain connectivity subtypes. After analyzing brain scans from both mouse models and human participants, scientists found that people on the autism spectrum tend to have abnormally high or abnormally low levels of communication between different brain regions. These two patterns appear to be driven by very different biological mechanisms, suggesting new ways to understand and potentially support people with autism.
Autism spectrum disorders are known for their wide variety of characteristics. Some people with autism experience significant difficulties with language and motor skills, while others do not. This apparent variation is often thought to reflect different underlying biological causes.
It is very difficult to find direct evidence linking specific behaviors to specific biological causes. Only a small percentage of people with autism have known genetic mutations that scientists can easily identify and study. This makes it difficult to group people biologically using genetics alone.
“This study started with a very simple but long-standing question: Why is autism so diverse?” said Alessandro Gozzi, senior researcher and director of the Functional Neuroimaging Laboratory at the Center for Neuroscience and Cognitive Systems at the Italian Institute of Technology Rovereto. “We know that people with autism can vary greatly in their symptoms, abilities, and support needs, but it is much more difficult to understand whether this variation reflects differences in the underlying biological mechanisms.”
To fill this gap, the authors turned to functional magnetic resonance imaging. This technology, also known as fMRI, measures brain activity by detecting changes in blood flow. If different regions of the brain show synchronized changes in blood flow when humans or animals are at rest, these regions are considered to be functionally connected.
The rationale for the study was to see whether the various genetic and environmental factors associated with autism produce recognizable patterns of functional connectivity. Starting with genetically modified mice, the research team aimed to map specific brain patterns, then look for those exact same patterns in human brain scans.
Previous findings using fMRI have been largely inconsistent. “In brain imaging studies, autism is often associated with very mixed results, with some studies reporting decreased connectivity between brain regions, while others reporting increased connectivity,” Gozzi said. “Rather than treating this variation as noise, we wanted to test the idea that this variation may contain biological information. In other words, differences in brain connectivity patterns may reflect differences in the biological subtypes of autism.”
The researchers first examined fMRI data from 20 different autism mouse models. These include 17 models with specific genetic changes, two models with changes in the immune system, and one specially bred strain of mice. We compared each model to a control group of typical mice to see how specific biological changes affect the brain’s functional connectivity.
When scientists grouped the whole-brain fMRI results from these 20 models, they noticed two major patterns. “What surprised us most was that two opposing connectivity patterns clearly emerged across species,” Gozzi said. “We found associated hypo- and hyper-connectivity patterns in mouse models and human autism datasets, and these patterns were associated with different biological pathways.”
Eleven of the mouse models showed decreased connectivity. This means that there is much less intercommunication between mouse brain regions than expected. The other nine models showed overconnectivity. This means that the brain regions communicated much more than expected.
The team then used computational techniques to examine which biological pathways were associated with these two patterns. They looked at the genes involved in each mouse model and mapped how those genes interacted with other proteins. This network of interacting genes is known as the interactome.
“In our study, we identified two major subtypes defined by connectivity,” Gozzi said. “One was characterized by decreased communication between brain regions and was associated with synaptic mechanisms central to neuronal communication. The other was characterized by increased communication between brain regions and was associated with changes in immune-related mechanisms and gene regulation.”
Synapses are tiny gaps where nerve cells send chemical signals to each other, facilitating brain communication. The hyperconnectivity pattern, on the other hand, was linked to the immune system and how cells translate genetic instructions into proteins.
Building on these findings in mice, the researchers examined a large collection of human fMRI data. The dataset included resting-state brain scans from 940 individuals diagnosed with autism and 1,036 neurotypical individuals. Participants ranged in age from 5 to 30 years old, and scans were collected at 38 different research centers.
The research team focused on specific brain regions that are evolutionarily conserved. These are brain regions that are anatomically and functionally similar in both mice and humans. By examining these specific regions, researchers were able to identify the same two functional connectivity subtypes in human participants.
To ensure that the study results were reliable, the researchers divided the human data into two separate groups. The first group served as the discovery dataset containing exactly 78.5 percent of the participants. The remaining 21.5 percent served as a replication dataset to validate the initial results.
Two subtypes appeared consistently in both datasets. Together, the hypo- and hyper-connectivity groups accounted for 25.1% of the human autism scans analyzed. The remaining scans did not fit neatly into either of these two extreme categories.
These findings help contextualize mixed results to date in human studies. “This was important because it suggests that the apparently contradictory findings in previous autism imaging studies may not simply reflect inconsistencies,” Gozzi said. “Some of them may reflect real biological differences between subgroups of individuals.”
These two human subtypes exhibited very different brain network structures. Individuals in the hyperconnectivity group showed significantly increased connectivity between deeper subcortical brain regions and the lateral cerebral cortex. Patients in the low connectivity group showed reduced connectivity between brain regions responsible for processing sensory and motor information.
Human subtypes also showed different behavioral profiles. Researchers looked at standardized symptom severity scores for a subset of participants. Individuals in the overconnected group tended to score slightly higher on social communication and interaction.
Finally, the scientists mapped human gene expression data to fMRI patterns to see if a biological cause matched the mouse model. They found very similar matches across both species. Brain regions that were poorly connected in humans were highly enriched in genes related to synaptic function.
At the same time, hyperconnected human brain regions became enriched with genes related to the immune system. “The main point is that autism diversity is not just symptom diversity,” Gozzi said. “At least in part, they also reflect biologically distinct patterns in how brain circuits communicate.”
Readers should not interpret these results to mean that all people with autism can be easily classified into these categories. “I think the broader message is that we shouldn’t assume that all autistic people share the same basic biology just because they fit the same diagnostic label,” Gozzi said. “While two people may appear clinically similar, the brain and molecular mechanisms contributing to their condition may be different.”
“At the same time, we would like to emphasize that our goal is not simply to create a new label,” Gozzi added. “The goal is to understand the biological structure underlying the autism spectrum so that future research, and ultimately clinical trials, can be better tailored to the mechanisms involved.”
The authors also note that current knowledge has several limitations. “The most important limitation is that this is not a clinical diagnostic tool,” Gozzi says. “We still cannot scan individuals and use this information to make clinical decisions.”
The two subtypes identified accounted for only about a quarter of the autistic individuals studied. “Another important point is that the two subtypes we identified only explain part of the heterogeneity of autism,” Gozzi noted. “This is not surprising because autism is so diverse, but it means that additional biological subtypes are almost certainly yet to be discovered.”
“This does not mean there are only two types of autism,” Gozzi continued. “Rather, it suggests that the autism spectrum may include biologically distinct subgroups, and that understanding these differences may ultimately help research move toward a more individualized approach.”
In the future, the team hopes to discover additional patterns within a broader spectrum. “We also want to improve the biological map,” Gozzi said. “While we identified two key features in this study, it is unlikely that autism can be explained by just two categories. Using richer mouse and human datasets, we hope to identify more detailed biologically defined subtypes and understand what hypocoupling and hypercoupling mean at the physiological level.”
Human data also need to be expanded to fully understand the impact of these subtypes on daily life. “The next major step is to understand what these brain-based subtypes mean in humans,” Gozzi said. “This requires large human datasets with deeper clinical and behavioral information, including cognition, sensory symptoms, development, adaptive function, genetics, and medical history.”
The study, “Subtypes of autism identified using cross-species functional connectivity analysis,” was authored by Marco Pagani, Valerio Zerbi, Silvia Gini, Filomena Grazia Alvino, Abhishek Banerjee, Andrea Barberis, M. Albert Basson, Yuri Bozzi, Alberto Galbusera, Jacob Elgood, Michela Fagiolini, and Jason P. Larch. Michela Matteoli, Caterina Montani, Davide Pozzi, Giovanni Provenzano, Maria Luisa Scatoni, Nicole Wenderoth, Ting Xu, Michael V. Lombardo, Michael P. Milam, Adriana Di Martino, Alessandro Gozzi.
