Fronto-striatal organization: Defining functional and microstructural substrates of behavioural flexibility

Laurel Morris, Prantik Kundu, Nicholas G. Dowell, Daisy J. Mechelmans, Pauline Favre, Michael A. Irvine, Trevor W. Robbins, Nathaniel D. Daw, Edward T. Bullmore, Neil A. Harrison, Valerie VoonCortex 2015 · DOI

TL;DR

This study maps the functional and microstructural connections between the prefrontal cortex and the striatum, revealing distinct circuits for goal-directed versus habitual behaviors. By combining advanced fMRI and diffusion imaging, they demonstrate how these brain networks underpin our ability to flexibly adapt our actions and learn from experience.

What Problem Does This Paper Address?

Understanding how the brain controls behavior, particularly the ability to switch between flexible, goal-oriented actions and automatic, habitual responses, is a fundamental challenge. While it's known that the frontal cortex and the striatum (part of the basal ganglia) are crucial for this, the precise organization and underlying neural substrates of these "fronto-striatal circuits" remain complex. Previous research has often used broad regions of interest or focused on specific aspects of behavior. This paper aims to provide a more detailed map of these circuits, linking specific functional and microstructural properties to different types of behavioral control, including goal-directed learning (model-based) and habit formation (model-free).

Methods

The researchers employed a multi-pronged approach:

  • Advanced fMRI (Multi-echo ICA): They utilized a novel multi-echo resting-state functional MRI (ME-ICA) sequence. This technique significantly improves the signal-to-noise ratio, allowing for higher spatial resolution and more precise mapping of functional connectivity between brain regions, particularly in subcortical areas like the striatum.
  • Region-of-Interest (ROI) based functional connectivity: They defined specific prefrontal cortical regions based on their known functions (e.g., medial orbitofrontal cortex for value processing, dorsolateral prefrontal cortex for executive control) and then mapped their functional connections to different parts of the striatum. They also examined connectivity from striatal subregions to other basal ganglia structures and the thalamus.
  • Behavioral Tasks: Participants completed tasks designed to assess:
  • Model-based vs. Model-free learning: A two-step choice task where participants learn to associate choices with outcomes, allowing for the quantification of how much their decisions are driven by a model of the environment (model-based) versus learned stimulus-response associations (model-free).
  • Attentional Shifting: An extradimensional (ED) set-shifting task to measure cognitive flexibility.
  • Probabilistic Reversal Learning: Tasks involving learning and then adapting to changing reward contingencies.
  • Neurite Orientation Dispersion and Density Imaging (NODDI): This advanced diffusion MRI technique was used to assess the microstructural properties of white and grey matter. Specifically, it measures neurite density (the complexity of neuronal processes) and orientation dispersion (how varied the orientation of these processes is), providing insights into the structural integrity and complexity of neural pathways.

Datasets

The study used data from healthy volunteers. Specific numbers of participants are mentioned for each study component:

  • Study 1 (fMRI): 66 healthy volunteers.
  • Study 2 (Behavioral Tasks): 37 healthy volunteers for the probabilistic reversal learning task, and the same cohort as Study 1 for the model-based/model-free task and ED set-shifting.
  • Study 3 (NODDI): 38 healthy volunteers.

No new benchmarks were introduced; rather, existing tasks and imaging techniques were applied and analyzed in novel ways to address the research questions.

Key Insights and Ideas

  • Fronto-striatal circuits are highly specialized: The research clearly demonstrates that different parts of the prefrontal cortex connect to distinct sub-regions of the striatum, forming specialized circuits for different cognitive functions. This goes beyond a general "fronto-striatal loop" to a more nuanced understanding of parallel processing pathways.
  • Distinct neural signatures for goal-directed vs. habitual behavior: The study provides strong evidence that model-based (goal-directed) learning is associated with functional connectivity between the medial orbitofrontal cortex (mOFC) and the ventral striatum (VS), and also with higher neurite density in the mOFC. Conversely, model-free (habitual) learning is linked to connectivity between the supplementary motor area (SMA) and the putamen, and potentially with neurite complexity in the putamen.
  • Microstructure matters for learning: The use of NODDI reveals that the physical structure of neurons (neurite density and complexity) in specific brain regions is directly related to an individual's learning strategy. This suggests that the "hardware" of the brain plays a critical role in how we learn and adapt.
  • Behavioral flexibility has specific neural correlates: Attentional shifting (ED set-shifting) was negatively correlated with connectivity between the dorsolateral prefrontal cortex (dlPFC) and the ventral striatum (VS), highlighting the role of this pathway in cognitive flexibility. Similarly, reversal learning errors were linked to connectivity patterns involving the lateral OFC and VS.

Key Results

  • Functional Connectivity Mapping: The study successfully mapped distinct functional connectivity patterns between various prefrontal regions (e.g., mOFC, dlPFC, SMA) and specific striatal subregions (ventral striatum, caudate, putamen).
  • Model-Based Learning Correlates: Higher model-based learning (higher 'w' scores) was positively correlated with connectivity between the mOFC and VS (R = .32, p = .01) and with increased mOFC neurite density.
  • Model-Free Learning Correlates: Exploratory analysis showed that model-free learning scores were positively correlated with connectivity between the posterior putamen and SMA (R = .266, p = .033).
  • Attentional Shifting Correlates: ED set-shifting errors were negatively correlated with connectivity between the dlPFC and VS (R = -.298, p = .021).
  • Reversal Learning Correlates: Slower reversal learning (both reward and loss) was negatively correlated with connectivity between the VS and lateral OFC, and positively with the amygdala.
  • Microstructural Findings: Higher model-based learning ('w') was negatively correlated with neurite density in the left posterior putamen and right SMA. Model-based learning (MBC) also correlated negatively with left VS orientation dispersion.

Limitations and Future Work

  • Exploratory Nature: While the study aimed to confirm a priori hypotheses, some analyses (e.g., model-free correlates) were exploratory. This means findings should be interpreted with caution and require further validation.
  • Cross-sectional Design: The study uses a cross-sectional design, meaning it captures a snapshot in time. Longitudinal studies could provide deeper insights into how these circuits develop and change over time or with learning.
  • Generalizability: The study was conducted on healthy volunteers. Future work could investigate how these fronto-striatal circuits and their microstructural properties are altered in neuropsychiatric disorders characterized by deficits in behavioral flexibility, such as addiction, OCD, or schizophrenia.
  • Causality: While correlational findings are strong, establishing causality would require experimental manipulation (e.g., using TMS or lesion studies) to confirm the direct role of these circuits in specific behaviors.

The authors suggest that their findings provide a foundational understanding of the neural architecture supporting behavioral control, paving the way for further investigations into the specific roles of these circuits in both healthy function and disease.

Summary updated

Abstract

Discrete yet overlapping frontal-striatal circuits mediate broadly dissociable cognitive and behavioural processes. Using a recently developed multi-echo resting-state functional MRI (magnetic resonance imaging) sequence with greatly enhanced signal compared to noise ratios, we map frontal cortical functional projections to the striatum and striatal projections through the direct and indirect basal ganglia circuit. We demonstrate distinct limbic (ventromedial prefrontal regions, ventral striatum - VS, ventral tegmental area - VTA), motor (supplementary motor areas - SMAs, putamen, substantia nigra) and cognitive (lateral prefrontal and caudate) functional connectivity. We confirm the functional nature of the cortico-striatal connections, demonstrating correlates of well-established goal-directed behaviour (involving medial orbitofrontal cortex - mOFC and VS), probabilistic reversal learning (lateral orbitofrontal cortex - lOFC and VS) and attentional shifting (dorsolateral prefrontal cortex - dlPFC and VS) while assessing habitual model-free (SMA and putamen) behaviours on an exploratory basis. We further use neurite orientation dispersion and density imaging (NODDI) to show that more goal-directed model-based learning (MBc) is also associated with higher mOFC neurite density and habitual model-free learning (MFc) implicates neurite complexity in the putamen. This data highlights similarities between a computational account of MFc and conventional measures of habit learning. We highlight the intrinsic functional and structural architecture of parallel systems of behavioural control.
BibTeX
@inproceedings{morris2015fronto,
  author = {Laurel Morris and Prantik Kundu and Nicholas G. Dowell and Daisy J. Mechelmans and Pauline Favre and Michael A. Irvine and Trevor W. Robbins and Nathaniel D. Daw and Edward T. Bullmore and Neil A. Harrison and Valerie Voon},
  title = {{Fronto-striatal organization: Defining functional and microstructural substrates of behavioural flexibility}},
  booktitle = {Cortex},
  year = {2015},
  doi = {10.1016/j.cortex.2015.11.004},
}
Figure