Intrinsic functional clustering of the macaque insular cortex

Lotte Sypré, Saloni Sharma, Dante Mantini, Koen NelissenFrontiers in Integrative Neuroscience 2024 · DOI

TL;DR

Researchers used data-driven hierarchical clustering on resting-state fMRI data to map the macaque insular cortex, identifying a four-cluster solution as the most stable organizational scheme. By validating these clusters against task-based fMRI, they demonstrated that these "intrinsic" subdivisions correspond precisely to distinct functional domains: taste, social-visual cues, hand grasping, and vestibular processing.

What Problem Does This Paper Address?

The insular cortex is a hidden gem of the primate brain—literally tucked away inside the lateral sulcus—and serves as a massive integration hub for internal (interoceptive) and external (exteroceptive) signals. While we have detailed anatomical maps of the insula from invasive "gold standard" techniques like histology and tract-tracing, there is a growing trend in human neuroscience to use non-invasive, data-driven "connectivity-based parcellation" (CBP).

The problem? It’s often unclear if the clusters generated by these mathematical algorithms actually represent biologically meaningful functional units or if they are just statistical noise. This paper uses the macaque model to bridge that gap, testing whether subdivisions derived from "resting" brain activity actually predict how the brain responds during specific tasks.

Methods

The authors employed hierarchical clustering on awake resting-state fMRI (rs-fMRI) data from eight rhesus monkeys.

  1. Clustering: They calculated the partial correlation between the time-series of every voxel within the insula mask. This created a "dissimilarity matrix" used as input for an unweighted average-linkage hierarchical clustering algorithm.
  2. Validation: To find the "optimal" number of clusters (K), they used three distinct methods:
  3. Fingerprinting: Testing if clusters have significantly different connectivity profiles to 23 external "seed" regions.
  4. Silhouette Index: Measuring how similar a voxel is to its own cluster versus others.
  5. Elbow Method: Identifying the point where adding more clusters provides diminishing returns in explaining variance.
  6. Functional Mapping: Crucially, they took these resting-state clusters and looked at their activity during five different task-based fMRI experiments: taste (sucrose/citric acid), distaste (high-concentration sour), grasping (motor execution), social-visual (watching lip-smacking videos), and vestibular stimulation (galvanic current).

Datasets

  • Resting-State: rs-fMRI data from 8 rhesus macaques (Macaca mulatta).
  • Task-Based: A comprehensive suite of task-fMRI datasets previously described in Sypré et al. (2023), involving a subset of the same monkeys plus additional subjects for specific tasks (e.g., vestibular stimulation).

Key Insights and Ideas

  • The "Magic Number" Four: While the algorithm could generate anywhere from 2 to 10 clusters, all three validation methods converged on four clusters as the most robust representation of the macaque insula.
  • Form Follows Function: The four clusters weren't just random patches; they mapped onto specific functional "fields":
  • Anterior: Primarily gustatory (taste).
  • Mid-Ventral: Social and affiliative visual cues.
  • Mid-Dorsal: Somato-motor (hand/grasping).
  • Posterior: Vestibular and optic flow (self-motion).
  • The Hierarchy Matters: The study found that while four clusters are "optimal" for stability, higher-order solutions (like N=6) revealed even finer-grained specializations, such as splitting the anterior insula into distinct dorsal-taste and ventral-social zones. This suggests the insula is organized like a map that reveals more detail the closer you zoom in.
  • Hemispheric Asymmetry: Interestingly, the right hemisphere showed a slightly different spatial arrangement in the anterior section compared to the left, hinting at potential functional lateralization in primates that warrants further study.

Key Results

  • Statistical Convergence: The four-cluster solution was the highest dimensionality where every cluster still maintained a significantly unique "connectivity fingerprint" (p < 0.05).
  • Task Specificity: In the left hemisphere, Cluster 3 (mid-dorsal) showed the strongest response to hand grasping, while Cluster 1 (anterior) dominated taste responses. Cluster 4 (posterior) was the only one significantly activated by galvanic vestibular stimulation.
  • Connectivity Profiles: Seed-to-brain analysis showed that the anterior cluster is uniquely wired to mouth-related and gustatory regions (like the orbitofrontal cortex), while the posterior cluster connects to "optic flow" regions (like area MST) involved in navigating through space.

Limitations and Future Work

The authors acknowledge that clustering algorithms impose "hard borders" on brain regions, whereas the real biology likely involves smooth gradients of function. Furthermore, while the four-cluster solution is statistically optimal, it likely oversimplifies the true complexity of the insula, which contains over a dozen distinct cytoarchitectonic areas.

Future work should explore even higher-dimensional parcellations using larger datasets to see if rs-fMRI can eventually match the granularity of invasive anatomical studies. Additionally, investigating how these functional clusters change in disease models (like anxiety or depression, where the insula is often implicated) could provide new biomarkers for psychiatric research.

Summary updated

Abstract

The functional organization of the primate insula has been studied using a variety of techniques focussing on regional differences in either architecture, connectivity, or function. These complementary methods offered insights into the complex organization of the insula and proposed distinct parcellation schemes at varying levels of detail and complexity. The advent of imaging techniques that allow non-invasive assessment of structural and functional connectivity, has popularized data-driven connectivity-based parcellation methods to investigate the organization of the human insula. Yet, it remains unclear if the subdivisions derived from these data-driven clustering methods reflect meaningful descriptions of the functional specialization of the insula. In this study, we employed hierarchical clustering to examine the cluster parcellations of the macaque insula. As our aim was exploratory, we examined parcellations consisting of two up to ten clusters. Three different cluster validation methods (fingerprinting, silhouette, elbow) converged on a four-cluster solution as the most optimal representation of our data. Examining functional response properties of these clusters, in addition to their brain-wide functional connectivity suggested a functional specialization related to processing gustatory, somato-motor, vestibular and social visual cues. However, a more detailed functional differentiation aligning with previous functional investigations of insula subfields became evident at higher cluster numbers beyond the proposed optimal four clusters. Overall, our findings demonstrate that resting-state-based hierarchical clustering can provide a meaningful description of the insula's functional organization at some level of detail. Nonetheless, cluster parcellations derived from this method are best combined with data obtained through other modalities, to provide a more comprehensive and detailed account of the insula's complex functional organization.
BibTeX
@inproceedings{sypre2024intrinsic,
  author = {Lotte Sypré and Saloni Sharma and Dante Mantini and Koen Nelissen},
  title = {{Intrinsic functional clustering of the macaque insular cortex}},
  booktitle = {Frontiers in Integrative Neuroscience},
  year = {2024},
  doi = {10.3389/fnint.2023.1272529},
}
Figure