Distinct transcriptomic and hierarchical organization associated with brain hyperconnectivity and hypoconnectivity in autism spectrum disorder

The conventional model says hyperconnectivity and hypoconnectivity in autism are two ends of a single spectrum. This study finds they are actually two separate biological phenomena. Researchers report that these two states differ in their molecular signatures, their place in the brain's hierarchy, and their cognitive associations.

Traditionally, scientists treated hyperconnectivity (excessive signaling) and hypoconnectivity (reduced signaling) as opposing expressions of one circuit-level disturbance. This "single axis" model assumes a singular underlying failure in how neurons communicate. However, this view fails to explain why these patterns appear in different brain regions. It also struggles to explain how they relate to specific genetic drivers.

Beyond a single connectivity axis

Current research in ASD neurobiology often treats connectivity changes as a simple polarity shift. Under this view, the brain is either "too connected" or "not connected enough." This implies a singular failure in neural communication. While some animal models suggest distinct genetic origins for these two states, this has rarely been verified in humans.

Existing studies have looked at these elements in isolation. Some map the physical locations of connectivity changes. Others look at gene expression or the brain's functional hierarchy (the way the brain organizes itself from simple sensory processing to complex thought). Because these layers were studied separately, the community lacked a unified view. We did not know if a "high-connectivity" gene drives a different biological process than a "low-connectivity" gene.

Linking genes to macroscale brain maps

To resolve this, the authors developed a multi-scale framework. This framework bridges the gap between microscopic gene expression and macroscopic brain function. The researchers utilized a massive dataset from the Autism Brain Imaging Data Exchange (ABIDE I/II). This dataset contains resting-state fMRI (scans measuring spontaneous brain activity) from 1,737 individuals. They integrated this with gene-expression maps from the Allen Human Brain Atlas (AHBA). The AHBA provides a blueprint of which genes are active in specific brain regions.

The methodology follows a structured pipeline, as illustrated in .

The process begins by quantifying regional "node strength." This is a measure of how integrated a specific brain region is with the rest of the network. They then used Partial Least Squares Regression (PLSR). This is a statistical technique designed to find the maximum covariance (shared variation) between two high-dimensional datasets. In this case, they compared thousands of genes to hundreds of brain regions.

By applying this, they could identify "transcriptomic signatures." These are specific groups of genes whose spatial expression patterns mirror the patterns of altered connectivity in ASD. They further analyzed these signatures using cortical gradient analysis. This method maps the brain's transition from primary sensory areas (like vision) to "transmodal" association areas (complex centers for social and executive function). This allowed them to see where these genes sit in the brain's organizational hierarchy.

Two distinct biological identities

The results reveal a profound dissociation between the two connectivity states. The paper finds that hyperconnectivity is largely concentrated in higher-order cortical and cerebellar regions .

The authors noted a descriptive trend where hyperconnectivity appeared larger in adults than in children. However, this age-related difference did not reach statistical significance (P = 0.536). Conversely, hypoconnectivity is localized to subcortical and orbitofrontal systems. This pattern remains remarkably consistent across different age groups .

The molecular evidence is equally striking. The authors report that hyper- and hypoconnectivity-associated genes show partially distinct neurotransmitter profiles .

While there is some overlap, the genes driving these states are not identical. Furthermore, when these genes were projected onto the brain's functional hierarchy, they occupied different territories [, Figure 5].

Hyperconnectivity-associated genes were centered in transmodal cortical territories. These are the "command centers" of the brain. Hypoconnectivity-associated genes were more broadly distributed toward sensory and subcortical areas.

Most importantly, the study connects these biological signatures to actual behavior. Through meta-analytic decoding, the researchers found that these transcriptomic signatures map onto cognitive domains relevant to ASD. These include social cognition, attention, and reward processing .

This suggests that the "too much" and "too little" connectivity are tied to fundamentally different functional challenges.

Constraints of the molecular map

While the findings are robust, the authors highlight several critical constraints. First, the transcriptomic data was sourced from the Allen Human Brain Atlas. This atlas uses neurotypical adult donor brains. This means the study relies on a "normative" molecular reference. It describes how ASD-related connectivity aligns with a healthy brain. It does not necessarily capture the unique, disease-specific gene expressions occurring within an autistic brain.

Second, the study is correlational. The researchers demonstrate that certain genes and connectivity patterns occupy the same space. However, they cannot prove that a specific gene causes the connectivity change. Finally, the researchers note they used a volumetric mapping approach. This assigns genes to 3D brain volumes rather than just the outer surface. This choice was necessary to include subcortical structures like the thalamus. However, it means the results may not be directly comparable to studies focusing exclusively on the cortical surface.

A new framework for ASD neurobiology

The verdict is clear: the "single axis" model of ASD connectivity is insufficient. The authors demonstrate that hyper- and hypoconnectivity are separable components of a complex architecture. This dissociation holds steady across different sexes, ages, and levels of symptom severity. This suggests it is a fundamental feature of the disorder rather than a transient developmental quirk.

For researchers, this means moving away from treating ASD as a monolithic "connectivity deficit." Instead, it should be modeled as a collection of distinct, interacting biological processes. This work provides a roadmap for future studies. Researchers can stop asking if connectivity is altered. They can start asking which specific molecular pathways drive which specific functional disruptions.

Figures from the paper