Anxious individuals shift emotion control from lateral frontal pole to dorsolateral prefrontal cortex

The study addresses why anxious individuals struggle to control emotional action tendencies that lead to maladaptive avoidance. Prior work implicates the human lateral frontopolar cortex (FPI) in coordinating control over emotional actions via interactions with parietal, sensorimotor cortices, and the amygdala. The authors hypothesize that altered functional, structural, and neurochemical properties of FPI explain inter-individual differences in emotional action control between high-anxiety individuals and non-anxious peers. They situate the question in the challenge of translating rodent-based anxiety models to humans due to human-specific expansion and connectivity of granular prefrontal cortex, particularly FPI, which lacks a rodent homologue and has unique access to amygdala inputs.

Rodent models suggest hippocampal-amygdala inputs to agranular medial frontal areas drive avoidance and fear, whereas medial prefrontal recurrent signals can reduce threat responses and enable approach. Translational gaps arise from human-specific prefrontal expansion; notably, human FPI has extensive medial and lateral cortical connections and direct amygdala access via the amygdalofugal bundle, unlike macaques where amygdala projections largely target medial PFC. Prior human work implicates FPI in selecting emotional actions by modulating sensorimotor cortex and shows that FPI perturbation impairs emotional action selection, while FPI recruitment predicts resilience to later emotional disorders. Patients with emotional disorders often show failed FPI recruitment during emotion control. These insights motivate examining FPI-centric circuits as a candidate mechanism underlying anxiety-related control deficits.

Design: Comparative study combining neurochemical (MRS), structural (diffusion-weighted imaging/tractography), and functional (fMRI) measures with behavioral performance on a social-emotional approach-avoidance (AA) task. Participants: High-anxiety group (N=52; 14 males; LSAS >30) versus a convenience non-anxious peer group (N=41; all males) from an earlier dataset. Independent trait anxiety (STAI Y-2) confirmed higher anxiety in the high-anxiety group (t(84)=5.5, p<0.001). Mean ages: non-anxious 23.8±3.4 (18–34), high-anxious 25.66±4.4 (20–39). Ethical approval (CMO2014/288); informed consent obtained. Task: Social AA task in fMRI. Participants pushed/pulled a joystick to approach/avoid happy and angry faces. Congruent blocks (approach happy, avoid angry) alternated with incongruent blocks (approach angry, avoid happy). Each face 100 ms; max response 2000 ms. Two fMRI sessions (≈576 trials total per participant across two days). Behavioral primary metric: error rate difference (incongruent vs congruent). MRS: 3T MEGA-PRESS sequence. Voxels: right FPI (primary), left SMC, right occipital cortex (control). GABA (edited at 3.0 ppm) and Glx (from unedited spectra) quantified with LCModel; excitability index defined as GABA/Glx ratio (lower ratio indicates higher excitability). Quality control via CRLB, SNR, FWHM, visual inspection; exclusions reported. fMRI: 3T, multiband EPI (2 mm isotropic, TR=1000 ms). Preprocessing: motion correction, distortion correction (TOPUP), registration (FLIRT/FNIRT), 5 mm smoothing, high-pass filter, ICA denoising. First-level GLM modeled approach/avoid × emotion separately for three stimulation conditions; incongruent > congruent contrast defined emotion-control. Second-level combined sessions with fixed effects; group-level with FLAME1. Cluster correction z>2.3; family-wise error controlled (whole-brain or frontal-lobe-restricted as specified). DWI/tractography: High-angular DWI (b=2500 s/mm², 256 dirs). Preprocessing with TOPUP/EDDY; BedpostX for crossing fibers. ProbtrackX reconstructed amygdalofugal pathway using seed in extended amygdala/substantia innominata, y=22 waypoint, CSF/contralateral exclusions, and anterior constraint. Connection strength normalized/log-transformed; tract entry counts into FPI (and other frontal regions) extracted using white-matter border masks. Statistics: Behavioral mixed-effects (Bayesian, brms) with factors Group and Congruency; extended models added FPI and SMC GABA/Glx and amygdalofugal-FPI strength in interaction terms; significance by 95% credible intervals not including zero. Spearman/Pearson correlations as post hoc with Bonferroni where applicable. fMRI whole-brain and ROI analyses linked neural congruency effects to MRS/DWI measures. Bayesian t-tests used to assess evidence for absence of FPI recruitment in the high-anxiety group within FPI ROI from controls.

• Behavior: Across groups, participants made more correct responses in congruent than incongruent trials (main effect of congruency; Bayesian mixed-effects b=0.198, 95% CI [0.10, 0.29]). No group difference in behavioral congruency effect (b=0.03, CI [-0.06, 0.12]; BF01=4.3 supporting absence of group effect).
• fMRI: Emotional control (incongruent>congruent) activated bilateral FPI and deactivated bilateral SMC at the group level; coordinates include FPI [32 54 4; -30 56 8] and SMC [42 -26 68; -32 -26 68]. In non-anxious participants, FPI showed significant congruency effects (as previously reported); in high-anxious participants, there was moderate evidence for absence of FPI congruency effect within the control-defined FPI ROI (Bayesian t-test BF01=4.2). High-anxious > non-anxious showed greater congruency effects in dorsolateral frontal cortex (BA 8B/9/46d) at [24 30 34], frontal-lobe corrected. Across both groups, higher trait anxiety correlated with reduced FPI congruency effects (max z=4.24, p=0.0004 at [40 56 -4], whole-brain corrected). Across participants, lower FPI recruitment associated with higher dIPFC recruitment, r≈-0.22, p=0.038.
• MRS (excitability): High-anxious participants had lower FPI GABA/Glx ratios (greater excitability) than non-anxious (t(88)=2.3, p=0.02). No group differences in SMC or occipital cortex (both t<1.3, p>0.19). The relationship between FPI GABA/Glx and behavior differed by group (three-way interaction Congruency×Group×FPI GABA/Glx b=0.19, CI [0.10, 0.28]). In non-anxious, more excitable FPI predicted better control (smaller behavioral congruency effect), Spearman ρ=0.47, p=0.0025. In high-anxious, the relationship was reversed (ρ=-0.29, p=0.036; not surviving multiple-comparison correction across three voxels). In non-anxious with more excitable FPI, SMC excitability further related to behavior (four-way interaction b=-0.10, CI [-0.21, -0.0002]). Neural congruency in SMC correlated with FPI excitability in non-anxious but not high-anxious participants (whole-brain corrected effect in left SMC [-20 -20 66]). Analyses using GABA/Cr or Glx/Cr alone showed no group differences or behavioral correlations, underscoring the importance of the GABA/Glx ratio.
• DWI (amygdalofugal connectivity): High-anxious participants had stronger amygdalofugal projections to FPI vs non-anxious (t(90)=3.3, p=0.0014); and to area 46 (t(90)=3.0, p=0.0027). No group differences for amygdalofugal projections to medial FP (FPm), BA24, BA25 (t<1.08, p>0.28), indicating anatomical specificity. The behavior–connectivity relationship differed by group (three-way interaction Congruency×Group×DWI b=0.14, CI [0.02, 0.26]). In non-anxious, stronger amygdalofugal-FPI connectivity related to larger behavioral congruency effects (ρ=0.27, p=0.04, one-sided), whereas no significant relation in high-anxious (ρ=-0.22, p=0.12). Group differences extended to neural coupling: in non-anxious, amygdalofugal-FPI strength correlated with SMC congruency; in high-anxious, it correlated with mPFC/ACC congruency.
• Compensatory recruitment: Across both groups, stronger amygdalofugal projections to FPI correlated with greater dIPFC congruency effects (ρ=0.29, p=0.005). No direct correlation between FPI GABA/Glx and dIPFC congruency (ρ=-0.15, p=0.15), nor between amygdalofugal connectivity and FPI excitability (ρ≈0, p=0.99). However, in regions showing greater amygdalofugal dependence in high-anxious (mPFC/ACC), neural congruency correlated with FPI excitability (ρ=-0.27, p=0.0095).

The findings demonstrate that anxious individuals implement control over emotional action tendencies by recruiting dIPFC and medial prefrontal regions instead of FPI, unlike non-anxious peers. This reduced FPI engagement occurs despite high-anxious individuals having higher FPI excitability (lower GABA/Glx) and stronger amygdalofugal input specifically to FPI and nearby lateral prefrontal regions. The authors propose that heightened excitability and stronger amygdala afferents may saturate FPI’s dynamic range even under mild emotional challenge, limiting its capacity to finely regulate sensorimotor outputs and context-dependent action selection. This could drive compensatory reliance on dIPFC/mPFC circuits, preserving behavior under mild demands but potentially failing under heavier loads where FPI’s integrative functions (combining affective information with contextual rules) are essential. The results align with observed FPI dysfunction across emotional disorders and suggest that circuit degeneracy allows alternative prefrontal nodes to assume control, with potential vulnerabilities when those nodes are concurrently taxed (e.g., working memory, public speaking). The work refines mechanistic links between anxiety, amygdala–prefrontal circuitry, and excitatory/inhibitory balance in FPI, offering targets for intervention (e.g., neuromodulation to restore FPI–SMC communication).

This study shows a functional-anatomical shift in anxious individuals from FPI to dIPFC during emotional action control, alongside increased FPI excitability (lower GABA/Glx) and stronger, anatomically specific amygdalofugal inputs to FPI. The strength of amygdala–FPI projections predicts the degree of dIPFC involvement, indicating compensatory recruitment when FPI function is compromised. These results characterize a circuit-level vulnerability—overexcitable, heavily amygdala-innervated FPI—that can become a bottleneck for emotion-control. Future directions include testing neuromodulatory strategies to normalize FPI excitability and FPI–SMC communication, concurrently managing hyperactivation in dIPFC/mPFC; employing high-field MRS for individualized E/I assessments; conducting larger, sex-balanced samples to generalize findings; and delineating amygdala nucleus-specific inputs to FPI.

• The non-anxious group included only male participants, potentially limiting generalizability. Although prior work reports similar FPI engagement across sexes, sex effects cannot be ruled out.
• Correlational brain–behavior findings in modest samples may overestimate effect sizes; larger confirmatory studies are warranted.
• 3T MRS GABA/Glx estimates include glutamine in Glx and provide proxy E/I balance rather than absolute neurotransmitter concentrations; higher-field MRS could refine estimates.
• The precise amygdala nuclei projecting to FPI via the amygdalofugal bundle could not be resolved with DWI.
• Subjective components of anxiety and their relationship to measured circuit properties were not directly assessed.