Cerebellar hypermetabolism disrupts fronto-cerebellar resting-state functional connectivity and associated executive compensation

Abstract

Cerebellar hypermetabolism measured with FDG-PET is interpreted as maladaptive plasticity, while increased cerebellar functional connectivity reported in fMRI studies indicates a compensatory role. Using Alcohol Use Disorder (AUD) as a neurobiological model, the combination of PET and fMRI examinations can extend our understanding of cerebellar mechanisms underlying brain reorganization within the fronto-thalamo-cerebellar circuit (FCC) supporting executive functions. The aim of the present study was to investigate resting-state functional connectivity (rs-FC) of the cerebellum and to examine its relationship with cerebellar metabolism, thalamic grey matter volume (GM) as a key node of the FCC, and executive functioning. In AUD patients, stronger negative rs-FC was found between the cerebellar lobule VIII seed and voxels in the left superior frontal gyrus compared with HC. Path analysis conducted in AUD patients indicated that cerebellar hypermetabolism was positively related to fronto-cerebellar rs-FC, and that fronto-cerebellar rs-FC was positively related to inhibition performance. In this model, after controlling for thalamic GM abnormalities, cerebellar hypermetabolism negatively impacted inhibition performance through rs-FC. Cerebellar hypermetabolism disrupts negative fronto-cerebellar rs-FC, resulting in desynchronization within the fronto-cerebellar loop that compromises compensation for executive deficits. Cerebellar hypermetabolism may represent a biomarker of alcohol-related brain dysfunction, as an initial mechanism in the cascade linking fronto-cerebellar desynchronization and executive impairment.

Introduction

Cerebellar hypermetabolism, as measured by FDG-PET, has been frequently reported in several neurological and psychiatric conditions, including temporal lobe epilepsy and interictal psychosis [1], Huntington’s disease [2], traumatic brain injury [3], amyotrophic lateral sclerosis [4–6], COVID-19-related encephalopathy [7], opioid use disorder [8], Alcohol Use Disorder (AUD) [9, 10], and Wernicke-Korsakoff syndrome [11]. Across these clinical populations, cerebellar hypermetabolism has mainly been interpreted as reflecting maladaptive plasticity [2, 3, 5, 7, 10, 11], supported by its correlation with frontal and parietal hypometabolism [10], and related motor [2, 11] and cognitive deficits [10].

In contrast to metabolic measurements, several activation and resting-state fMRI studies have shown higher functional connectivity between the cerebellum and frontal cortices in patients with various pathologies, including multiple sclerosis [12], Parkinson’s disease [13–15], and AUD [16–18]. These patients also exhibited preserved cognitive and motor functioning, suggesting cerebellar compensatory mechanisms in these conditions.

Taken together, findings from FDG-PET and fMRI studies converge to highlight the cerebellum as a key structure involved in various neuropsychiatric conditions. However, these studies describe contrasting roles: cerebellar hypermetabolism is often interpreted as a marker of maladaptive plasticity, whereas increased cerebello-cortical functional connectivity is generally associated with compensatory mechanisms. These discrepancies may be explained by the fact that FDG-PET and fMRI rely on different physiological signals, each capturing distinct aspects of brain function. While both FDG-PET and fMRI studies aim at better characterize the brain’s functional integrity, they provide different yet complementary information. Glucose metabolism measured by FDG-PET is known to reflect local synaptic activity [19, 20], whereas fMRI measures focus on the blood-oxygen-level-dependent (BOLD) signal, which indirectly reflects neuronal activity at rest [21]. Functional connectivity reflects the temporal synchronicity between the BOLD signals of distinct brain regions.

The combination of FDG-PET and resting-state functional connectivity (rs-FC) investigations may significantly advance our understanding of the brain’s pathophysiological mechanisms by reconciling the apparent discrepancy between the maladaptive role of the cerebellum observed in FDG-PET studies and its compensatory role in fMRI studies. Combined FDG-PET/rs-FC fMRI investigations have been conducted in Parkinson’s disease [22, 23], Alzheimer’s disease [24], primary progressive aphasia [25], childhood epilepsy [26], and opioid use disorder [8]. However, none of these studies have specifically examined cerebellar metabolism and its resting-state functional connectivity with respect to other brain regions. The role of the cerebellum in brain reorganization has attracted considerable interest but remains poorly understood.

AUD is a valuable model for studying the neurobiological mechanisms underlying cerebellar function. Indeed, both its compensatory [27] and maladaptive [10] roles have been proposed based on rs-fMRI and FDG-PET studies, respectively. A specific brain circuit particularly affected in AUD, is the fronto-cerebellar circuit (FCC), which underlies executive and motor dysfunctions [28, 29]. The FCC also includes the thalamus, a key node that may contribute to brain reorganization, as its metabolism has been repeatedly reported to remain preserved in AUD [10, 30, 31], despite marked structural shrinkage [30]. Given the brain and cognitive deficits associated with AUD, a combined FDG-PET/ fMRI-rs-FC approach would allow for a more in-depth investigation of the cerebellum’s role in brain reorganization mechanisms.

The overall goal of the present study was thus to understand the nature of the different cerebellar mechanisms involved in the brain reorganization observed in AUD. More specifically, we first aimed to investigate the rs-FC of the cerebellum in AUD patients compared to control subjects, and then to examine its relationships with cerebellar metabolism, thalamic grey matter (GM) volume, and executive functions.

Materials and methods

Participants

Twenty-seven patients with severe AUD (without Korsakoff’s syndrome) and 25 healthy controls (HC) were included in this study.

All participants were aged between 18 and 70 years and were native French speakers. None had a history of neurological disorders, endocrine or other infectious diseases (diabetes, HIV, or hepatitis, as confirmed by blood analysis), mental illness (psychiatric disorders assessed using the Mini International Neuropsychiatric Interview), or other forms of substance use disorder (except tobacco). Additionally, none were taking psychotropic medications (such as benzodiazepines only used during alcohol withdrawal) that could affect cognitive functioning. All participants were assessed for signs of lacunar stroke, small vessel disease, or overt vascular damage. This evaluation was systematically performed using FLAIR and T2* sequences. The images were reviewed by a physician, who confirmed the participant’s eligibility for the study after excluding any individuals with neurological alterations unrelated to the pathophysiology of AUD.

AUD patients were recruited by clinicians while receiving withdrawal treatment as inpatients at Caen University Hospital (France). Patients were included based on the DSM-IV criteria for alcohol dependence and the DSM-5 criteria for alcohol use disorder. All patients met the criteria for severe AUD according to DSM 5 (≥6 criteria), based on a posterior harmonization of the DSM-IV alcohol dependence criteria. Although AUD patients were in the early stages of abstinence at inclusion, none of the them exhibited physical symptoms of alcohol withdrawal, as assessed by the Cushman’s scale [32], and they had not taken psychotropic medication (benzodiazepines) for at least the past 48 h. Information regarding benzodiazepine prescriptions during detoxification, including duration, was systematically recorded. They were assessed using the Alcohol Use Disorders Identification Test (AUDIT [33]), a semi-structured interview [34], and questions accompanying the Structured Clinical Interview for DSM-IV-TR (SCID [35]), which included measures of the duration of alcohol use disorder (in years), the number of detoxifications (including the current one), and daily alcohol consumption over the month preceding treatment (in units, with one standard drink corresponding to a beverage containing 10 g of pure alcohol). Tobacco dependence was assessed using the Fagerstrom test (Table 1).

Table 1.

Main demographic and clinical features of the participants.

	AUD N = 27	HC N = 25	Statistical analyses
Men/Women	24/3	22/3	Chi² = 0.01; p = 0.92a
Age (years)	46.40 ± 9.84	43.70 ± 6.35	t(1,50) = 1.14; p = 0.26b
Range	26–63	31–55
Education (years of schooling)	12.00 ± 2.04	11.70 ± 1.67	t(1,50) = 0.54; p = 0.59b
Range	9–17	9–15
AUDIT	27.20 ± 6.07	2.76 ± 1.51	t(1,50) = 19.55; p < 0.001*b
Range	9–38	0-6
Days of sobriety before inclusion	9.00 ± 3.87	-	-
Range	4–21
Duration of benzodiazepine prescription (days)	4.41 ± 4.62	-	-
Range	0–15
Daily alcohol consumption in the month preceding treatment (standard drink)	17.80 ± 8.66	-	-
Range	0–40
Duration of alcohol use disorder (years)c	22.70 ± 10.20	-	-
Range	5–41
Tobacco use (yes/no)	20/7	6/19	Chi² = 11.1; p < 0.001a
Fagerstrom test	3.78 ± 2.91	0.80 ± 1.66	t(1,50) = 4.48; p < 0.001*b
Range	0–10	0–6

A standard drink corresponds to a beverage containing 10 g of pure alcohol.

Data are shown as means ± standard deviations; -: data not applicable.

AUD alcohol use disorder, HC healthy controls.

* significant at p ≤ 0.05.

^a Chi² (Yates’ correction applied).

^b Independent Student t-test.

^c one missing data point.

HCs were recruited to match the AUD patients in terms of sex, age, and education (Table 1). All HCs were assessed using the AUDIT questionnaire [33] to ensure that they did not meet the DSM-IV criteria for alcohol abuse or dependence (AUDIT < 7 for men and < 6 for women). A score below 129 on the Mattis Dementia Rating Scale [36] or a score of 19 or higher on the Beck Depression Inventory (BDI [37]) were exclusion criteria.

The main demographic and clinical characteristics of the participants are presented in Table 1.

All participants were informed about the study prior to inclusion and provided written informed consent, in accordance with the Declaration of Helsinki [38]. The Ethical Principles of Psychologists and Code of Conduct of the American Psychological Association [39] regarding the ethical treatment of human participants were adhered to for all participants. The study was approved by the local ethics committee of Caen University Hospital (CPP Nord Ouest III n° IDRCB: 2011-A00495).

Neuroimaging assessment

All participants underwent anatomical MRI, FDG-PET, and resting-state functional MRI examinations.

Neuroimaging acquisition and preprocessing

MRI assessment

For anatomical MRI, a high-resolution 1 mm³ T1-weighted image was acquired for each participant on a 3 T MR Philips scanner (Achieva 3.0 T TX) at the Cyceron neuroimaging center (Caen, France) using a three-dimensional fast-field echo sequence (repetition time = 20 ms, echo time = 4.6 ms, flip angle = 10°, 180 slices, slice thickness = 1 mm, field of view = 256 × 256 mm², matrix = 256 × 256).

MRI datasets were pre-processed using SPM12 (Wellcome Department of Cognitive Neurology, Institute of Neurology, London, UK). Raw MRI data were spatially normalized to the Montreal Neurological Institute (MNI) space (voxel size = 1.5 mm³; matrix = 121 × 145 × 121) and segmented into grey matter (GM), white matter (WM), and cerebrospinal fluid (CSF). The normalised GM images were modulated by the Jacobian determinants to correct for non-linear warping only, so that the resulting brain volumes were adjusted for overall brain size. The GM volume of the thalamus was extracted for each participant using the AAL template [40].

FDG-PET assessment

FDG-PET data were acquired using a Discovery RX VCT 64 PET-CT scanner (GE Healthcare) with a resolution of 3.76 × 3.76 × 4.9 mm³ and an axial field of view of 157 mm. Participants had fasted for at least 6 h prior to scanning. FDG uptake was measured in the resting state, with eyes closed, in a quiet and dark environment. A bolus of 3–5 mci of FDG was injected at time 0, and a 10-min data acquisition period began 50 min post-injection, following the acquisition of a low-dose CT transmission scan (140 kV, 10 mA). Forty-seven planes were acquired with septa out (3D list-mode data acquisition), and the images were reconstructed using the Ordinary Poisson - Ordinary Subset Expectation Maximisation algorithm (OP-OSEM, 21 subsets, 2 iterations) with a voxel size of 1.95 × 1.95 × 3.2 mm (x y z).

The PET data were first corrected for CSF and WM partial volume effects (PVE) in GM using the voxel-by-voxel “modified Müller-Gartner” method. Using SPM12, the PVE-corrected PET datasets were then co-registered (rigid-body co-registration) to their respective native MRIs and normalized to MNI space by reapplying the normalisation parameters estimated from the VBM protocol described above (final voxel size: 2 mm³; matrix = 79 × 95 × 79). To control for inter-individual variations in PET measurements, semi-quantitative normalization was performed by scaling the PET images to the mean PET value of a cerebellar mask (including only cerebellar lobules III, IX and X), as described in detail in Ritz et al. [30] and in the Supplementary Material.

Resting-state functional MRI assessment

Resting-state functional MRI (Rs-fMRI) data were acquired using an interleaved bottom to top 2D T2* SENSitivity Encoding EPI sequence designed to reduce geometric distortions (2D-T2*-FFE-EPI, 80 × 80 × 52 axial slices, 2.8 mm thickness, TR = 2.90 s, TE = 30 ms, flip angle = 80°, field of view = 224 × 224, no gap, in-plane voxel size = 2.8 × 2.8 mm², 240 volumes, acquisition time = 11.59 min). Imaging parameters were selected to cover the whole brain, including the cerebellum. During data acquisition, participants were instructed to relax, keep their eyes closed without falling asleep, and let their thoughts flow freely.

Rs-fMRI data (i.e., EPI volumes) were processed following the procedure recommended by Whitfield-Gabrieli and Nieto-Castanon [41] using the functional connectivity toolbox (CONN toolbox; https://www.nitrc.org/projects/conn). The CONN toolbox operates within the framework of Statistical Parametric Mapping software (SPM12; Welcome Department of Cognitive Neurology, Institute of Neurology, London, UK) implemented in MATLAB (R2022) (MathWorks Inc., Natick, MA). Rs-fMRI volumes were first checked for artifacts due to head motion or abnormal variance distribution using the Artifact Detection Tools (ART tools; https://www.nitrc.org/projects/artifact_detect), implemented in the CONN toolbox. For each subject, the EPI volumes were corrected for slice timing and realigned on the first volume. Data were then spatially normalized to reduce geometric distortion effects [42]. This procedure included: 1) co-registration, including the alignment of the mean EPI and the non-EPI T2* volumes; 2) co-registration between the non-EPI T2* and T1 volumes; 3) normalization of the mean EPI volume to match the non-EPI T2* volume; 4) segmentation of the T1 volume using the unified segmentation and spatial normalization procedure to extract gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF) in SPM12 [43]); normalization of the co-registered T1, non-EPI T2*, and EPI volumes to corresponding anatomical regions in a common space (the Montreal Neurological Institute template; MNI), using the parameters obtained from the previously described T1-weighted segmentation routine (final dimensions: 79 × 95 × 79; final voxel size 2 × 2 × 2 mm³); and 6) smoothing of EPI volumes using a 4 mm FWHM (full-width at half-maximum) Gaussian kernel. Subsequent steps included temporal preprocessing for noise reduction (denoising), first-level (individual) analysis, and second-level (group) analysis.

The denoising step was applied to the previously pre-processed rs-fMRI data to minimize the influence of physiological sources (head motion, heart rate, and respiration) and reduce inter-subject variability. Noise reduction in the CONN toolbox was performed using the anatomical component-based noise correction method (aCompCor [44]). EPI volumes were filtered using covariates derived from the T1 volume segmentation, including the first five principal components of WM and the first five principal components of CSF. In addition, movement parameters generated by SPM (6 motion parameters and their first- and second-order derivatives, 18 parameters in total) and output variables generated by ART (scrubbing and framewise displacement) were included as covariates. After regression, temporal frequencies below 0.008 Hz or above 0.09 Hz were removed from the resulting Blood Oxygenation Level Dependent (BOLD) time series [45] to reduce noise and increase sensitivity of the measures. Preprocessing steps are detailed in the Supplementary Material.

Assessment of executive functioning

An examination of executive abilities was conducted in all participants. Flexibility was assessed using the Trail Making Test [46], inhibition using the Stroop test [47] and set-shifting using the Modified Card Sorting Test [48]. For all tasks, higher scores indicated poorer performance. Details of the executive scores and the results for the two groups are presented in Table 2.

Table 2.

Executive function results in the participants.

	AUD N = 27	HC N = 25	Statistical analyses
TMT (part B - part A; time in seconds)	69.96 ± 53.78	39.00 ± 20.73	U = 174; p = 0.003*; rank-biserial correlation = 0.48
Range	25–232	7–98
Stroop (Interference - Naming; time in seconds)	57.78 ± 21.52	47.72 ± 15.57	U = 240; p = 0.06t ; rank-biserial correlation = 0.29
Range	21–126	23–86
Number of perseverative responses on the MCST	2.30 ± 3.11	1.36 ± 1.82	U = 300; p = 0.47; rank-biserial correlation = 0.11
Range	0–11	0–7

Statistical comparisons between AUD patients and HC were conducted using Mann-Whitney tests, as the data were not normally distributed (Kurtosis parameters > |2]).

Rank-biserial correlation = 0.10: small effect size; 0.30 : medium effect size; 0.50: large effect size.

Data are shown as means ± standard deviations.

AUD alcohol use disorder, HC healthy controls.

TMT trail making test; MCST modified card sorting test.

* significant at p ≤ 0.05.

^t trend toward significance 0.10 ≥ t ≥ 0.05.

The neuropsychological assessment was performed within 2 weeks of the neuroimaging sessions.

Statistical analyses

Rs-fMRI analyses

First-level analyses

Using the MarsBar toolbox, based on our previous study [10], a 5-mm-radius spherical seed was created, centered on the peak of the left cerebellar lobule VIII that showed hypermetabolism in AUD compared with HC (MNI coordinates: −28 −54 −52) (Fig. 1A). Although derived from a different imaging modality, this hypothesis-driven choice was intended to examine the cross-modal associations between cerebellar hypermetabolism and rs-FC, and should not be interpreted as a data-driven optimization within the same measurement space. Seed-to-voxel analyses were chosen because this approach provides higher reproducibility [49], and greater spatial localisation [50] compared with ROI-to-ROI analyses. For each participant, bivariate Pearson correlation coefficients were calculated using the CONN toolbox between the time course of the seed and the time course of each GM voxel. A Fisher z-transformation was then applied to the individual connectivity maps for use in second-level group analyses.

Fig. 1. The statistical analysis process.

FC: resting-state functional connectivity. GM: grey matter volume. A A spherical seed (5 mm radius) centred on the peak of the left cerebellar lobule VIII, which showed hypermetabolism in AUD compared with HC, was created. B In AUD and HC separately, positive and negative correlations between the seed and all GM voxels were computed using the CONN toolbox. C The averaged connectivity maps (for both positive and negative correlations in AUD and HC) were then combined to create a mask, which was subsequently applied in the following analyses. D Two-sample t-tests were conducted to identify patterns of lower (i.e., hypoconnectivity; AUD < HC) or higher (i.e., hyperconnectivity; AUD > HC) rs-FC of the cerebellar seed in AUD compared with HC (at p (uncorrected) ≤ 0.001 and p ≤ 0.05 FWE corrected at the peak voxel). E In AUD patients, correlations were performed to assess the relationship between these connectivity patterns and cerebellar hypermetabolism, thalamic GM volume, and executive functioning.

Second-level analyses

First, in AUD and HC separately, positive and negative correlations between the seed and all GM voxels were examined to assess whether the average connectivity within each group differed from zero, with a threshold set at p uncorrected ≤ 0.001 and p ≤ 0.05 FWE-corrected at the peak voxel (Fig. 1B). The averaged connectivity maps (for both positive and negative correlations in AUD and HC) were then combined to create a mask, which was applied in the subsequent analyses (Fig. 1C). These analyses were thus conducted only in voxels showing significant positive or negative correlations between the seed and GM voxels in AUD and HC. This mask was used to restrict the analyses to voxels functionally related to the seed at the individual level, independently of group membership, thereby avoiding comparisons in regions unrelated to the seed while preserving sensitivity to detect group-specific effects. Finally, to address our first objective, two-sample t-tests were conducted to identify patterns of lower (i.e., hypoconnectivity; AUD < HC) or higher (i.e., hyperconnectivity; AUD > HC) rs-FC of the cerebellar seed in AUD compared with HC (Fig. 1D). Results are reported at p ≤ 0.05 FWE-corrected for multiple comparisons, with a peak-voxel threshold of p < 0.001.

Relationship between rs-FC connectivity, cerebellar metabolism, thalamic GM, and executive functioning in AUD patients

In AUD patients only, rs-FC values in regions showing hypoconnectivity and hyperconnectivity compared with HC were extracted using the CONN Toolbox. Correlation analyses were then conducted between the extracted rs-FC values and, on the one hand, the metabolism of the cerebellar lobule VIII seed, thalamic GM volume, and measures of executive functioning (Fig. 1E). Thalamic GM volume was included in the analysis because this brain region exhibited severe atrophy without hypometabolism [30], and thalamic abnormalities have been identified as a key feature of alcohol-related brain dysfunction [51]. Executive functioning was examined in AUD patients because it has been correlated with cerebellar hypermetabolism [10] and is associated with cerebellar compensatory mechanisms in rs fMRI studies [16–18]. Due to the non-normal distribution of the executive data, Spearman’s correlations were used.

The statistical analysis process is summarized in Fig. 1.

Path analysis in AUD patients

To determine how rs-FC, cerebellar metabolism, and thalamic GM volume contribute to executive performance in AUD patients, path analyses were conducted [52, 53] using Mplus software. Only variables showing significant bivariate associations were entered into the path models to limit model complexity and ensure interpretability. Path analyses allow direct and indirect effects to be examined simultaneously through path coefficients involving multiple independent and dependent variables. Due to the non-normal distribution of the executive measures, path analyses were conducted using the Robust Maximum Likelihood estimator. Model fit was evaluated using the Standardized Root Mean Residual (SRMR; values of 0.08 or less indicate a good fit), and the Comparative Fit Index (CFI; values > 0.95 indicate a good incremental fit) [52, 54].

Three hypothesized models, each containing an equal number of paths and following the same methodological principles as Mander and colleagues [55], were tested and subsequently compared using model fit indices. The model with the lowest adjusted Bayesian Information Criterion (adj BIC) value was identified as the optimal solution, with differences in BIC values > 10 indicating a strong effect, 6–10 a moderate effect, and 2–6 a small effect [56].

Results

Rs-FC in AUD patients and HC

First-level: Correlations between the seed and GM voxels in AUD patients and HC separately

In AUD patients, positive correlations were observed between rs-FC in the cerebellar lobule VIII seed and voxels in the cerebellum (lobules VIII, IV, IV-V, Crus I and Vermis 8 and Vermis 7), the superior parietal gyrus (BA 5 and 7), and the precentral cortex (BA 6) bilaterally. Negative correlations were observed with rs-FC in the precuneus (BA 40), the superior frontal gyrus (BA 10), and the middle frontal gyrus (BA 9) bilaterally.

In HC, positive correlations were observed between rs-FC in the cerebellar lobule VIII seed and voxels in the right postcentral gyrus and cerebellar lobule VIII bilaterally. Negative correlations were observed with rs-FC in the parietal cortex (BA 7) bilaterally and in the middle frontal gyrus (BA 9).

These brain regions showing positive and negative correlations with rs-FC of the left cerebellar lobule VIII seed were used to create a mask, which was applied in the second-level analyses (Fig. 1C).

Second-level: Comparisons of rs-FC in AUD and HC

Compared with HC (AUD>HC), AUD patients showed higher rs-FC between the cerebellar seed and voxels in the left parietal gyrus (MNI coordinates: −18 −60 + 70; t₍₅₀₎₌3.26; p_(FWE) = 0.05; ke = 180 voxels; Fig. 2).

Fig. 2. Higher positive rs-FC connectivity and lower negative rs-FC in AUD patients compared with HC.

Rs-FC: resting-state functional connectivity. Green: seed in the cerebellar lobule VIII showing higher metabolism in AUD patients compared with HC. MNI coordinates were taken from a previous study reporting cerebellar hypermetabolism in AUD [10]. Red: AUD > HC contrast. AUD patients show stronger positive functional connectivity between the cerebellar lobule 8 seed and the parietal gyrus compared with HC. Blue: AUD > HC contrast. AUD patients show stronger negative functional connectivity between the cerebellar lobule 8 seed and the superior frontal gyrus compared with HC. Results are displayed at p uncorrected ≤ 0.001 p FWE-corrected ≤ 0.05 at peak clusters. * indicates statistical significance.

For the reverse comparison (AUD<HC), we observed lower rs-FC in AUD patients between the cerebellar seed and voxels in the left superior frontal gyrus (BA 10; MNI coordinates: −12 + 64 + 20; t₍₅₀₎₌3.26; p_(FWE) = 0.01; ke = 72 voxels; Fig. 2). This functional connectivity was negative - indicating that higher activity in the cerebellar seed was associated with lower activity in the frontal gyrus - and was significant only in AUD patients at the first-level analysis. These findings suggest a specific anticorrelation between the cerebellar seed and frontal regions in AUD.

Correlations between cerebellar metabolism, thalamic GM volume, and executive functioning in AUD patients

FDG metabolism in the cerebellar lobule VIII seed was higher in AUD patients (mean = 1.42 ± 0.26 sd) compared with HC (mean = 1.25 ± 0.16 sd; t₍₅₀₎₌2.86; p = 0.006; Cohen’s = 0.79, large effect size; Fig. 3). Thalamic GM volume was atrophied in AUD patients (mean = 0.27 ± 0.03 sd) compared with HC (mean = 0.31 ± 0.03 sd; t₍₅₀₎ = −4.82; p < 0.001; Cohen’s = 1.4, large effect size; Figure S1).

Fig. 3. Relationships between lower negative rs-FC (between the cerebellar lobule VIII seed and the frontal gyrus) and cerebellar hypermetabolism (top) and the performance (bottom) in AUD patients.

Cerebellar metabolism in lobule VIII was higher in AUD patients compared with HC (p = 0.006). A trend toward significance was observed on the Stroop task (p = 0.06), with AUD patients showing longer completion times (i.e., poorer performance) than HC. AUD alcohol use disorder. Rs-FC resting-state functional connectivity. * significant at p ≤ 0.05. t trend toward significance, 0.10 ≤ t ≤ 0.05.

In AUD patients only, rs-FC measures were extracted using the CONN Toolbox from regions showing higher positive connectivity (positively synchronized) and higher negative connectivity (negatively synchronized) with the cerebellar seed compared with HC.

Regarding the pattern of positively synchronized regions, no correlations were found between the rs-FC and measures of cerebellar metabolism, thalamic GM volume, or executive performance (all p values > 0.05). However, for the pattern of negatively synchronized regions (anticorrelations), rs-FC was correlated with cerebellar metabolism (rBP = 0.49 (95% CI [0.14–0.74]); p = 0.009; large effect size) and Stroop time (Spearman rho = 0.32; p = 0.04; medium effect size). Specifically, in AUD patients, this correlation indicates that the more negative the rs-FC between the cerebellar lobule VIII seed and the left superior frontal gyrus, the higher the cerebellar metabolism (i.e., lower hypermetabolism) and the better the Stroop performance (i.e., shorter time) (Fig. 3). No significant correlations were observed with thalamic volume or other executive measures.

Interestingly, Stroop time was also correlated with cerebellar hypermetabolism (Spearman rho = 0.55; p = 0.003; large effect size) and with thalamic GM volume (Spearman rho = −0.61; p < 0.001; large effect size). Poorer performance on the Stroop test was associated with greater cerebellar hypermetabolism and smaller thalamic GM volume (Figure S1).

Given the high prevalence of tobacco use in the AUD group (74% smokers), additional exploratory analyses were conducted to examine the influence of nicotine dependence on cerebellar metabolism, thalamic GM volume, and executive performance. These analyses revealed no significant associations between Fagerstrom scores and the main brain or cognitive measures after Bonferroni correction for multiple comparisons (p > 0.006 for 8 correlations).

Path analysis

Based on the results of the correlation analyses, cerebellar hypermetabolism, negative rs-FC between the cerebellum and frontal cortex, and thalamic GM were entered into path analyses to examine their association with Stroop performance (inhibition) in AUD patients (Fig. 4). Thalamic GM was not included as a mediator between cerebellar hypermetabolism and rs-FC because it did not significantly correlate with the rs-FC measure. The first model tested the direct effect of cerebellar hypermetabolism on inhibition, independently of rs-FC but controlling for thalamic GM volume. The second model tested the direct effect of thalamic GM volume on inhibition, independently of cerebellar hypermetabolism and rs-FC. The third model tested the indirect effect of cerebellar hypermetabolism on inhibition through rs-FC, after controlling for thalamic GM volume.

Fig. 4. Path models of the relationships between thalamic GM volume, cerebellar hypermetabolism, and negative rs-FC on inhibitory performance in AUD patients.

A Direct effect of cerebellar hypermetabolism on inhibition controlling for thalamic GM volume. B Direct effect of thalamic GM volume on inhibition independently of cerebellar hypermetabolism and rs-FC. C Indirect effect of cerebellar hypermetabolism on inhibition through rs-FC controlling for thalamic GM volume. Values are reported as β (standardized coefficients). Rs-FC resting-state functional connectivity; inhibition was assessed using the Stroop task (Interference – Naming; time). Data are presented as standardized coefficients. * significant path at p ≤ 0.05; ** significant at p ≤ 0.01; *** significant at p ≤ 0.001 after 95% CI percentile bootstrap (1000 replicates). A two-way arrow indicates a correlation, and a one-way arrow indicates a regression. The third model showed the best fit.

In the first model (Fig. 4A), the direct effect of cerebellar hypermetabolism on inhibition - independently of rs-FC but controlling for thalamic GM volume - was significant (β = 0.46 (95% CI [0.10; 0.82]; p = 0.04), but the model showed poor fit (SRMR = 0.11; CFI = 0.81). A significant negative correlation was observed between thalamic GM volume and cerebellar hypermetabolism (β = −0.47 (95% CI [−0.66; −0.28]; p < 0.001), whereas the direct path of rs-FC on inhibition was not significant (β = 0.17 (95% CI [−0.04; 0.39]; p = 0.178).

In the second model (Fig. 4B), the direct effect of thalamic GM volume on inhibition - independently of cerebellar hypermetabolism and rs-FC - was significant (β = −0.60 (95% CI [−0.83; −0.37]); p < 0.001) but showed the poorest model fit (SRMR = 0.16; CFI = 0.61). Thalamic GM volume negatively impacted inhibition performance in AUD patients. The direct paths of cerebellar hypermetabolism on rs-FC and rs-FC on inhibition were significant (β = 0.49 (95% CI [0.32; 0.67]; p < 0.001; β = 0.34 (95% CI [0.13; 0.59]; p = 0.007 respectively). However, the critical path was not significant (β = −0.10 (95% CI [−0.20; 0.04]; p = 0.09).

In the third model (Fig. 4C), path analysis indicated that thalamic GM volume was negatively correlated with cerebellar hypermetabolism (β = −0.49 (95% CI [−0.71; −0.26]; p < 0.001). Cerebellar hypermetabolism was positively related to fronto-cerebellar rs-FC (β = 0.49 (95% CI [0.32;0.67]; p < 0.001), and fronto-cerebellar rs-FC was positively related to inhibition performance (β = 0.46 (95% CI [0.29;0.63]; p < 0.001). This model showed the best fit (SRMR = 0.07; CFI = 0.94). The critical path indicated that, after controlling for thalamic GM abnormalities, cerebellar hypermetabolism negatively impacted inhibition performance through rs-FC (β = 0.23 (95% CI [0.09; 0.36]); p = 0.008). This third model had the lowest BIC value (adj BIC = 31.94) compared with the first and second models (adj BIC values = 40.87 and 34.07, respectively). When thalamic GM volume was included as a mediator in the path between cerebellar hypermetabolism and fronto-cerebellar rs-FC, the critical path was no longer significant (β = −0.01 (95% CI [−0.04; 0.04]; p = 0.83).

To account for the unbalanced sex distribution in the AUD group, sex was included as a covariate in the path analysis. Including sex did not alter model fit indices or the strength and significance of the estimated paths. Moreover, sex was not significantly associated with any variable in the model, and the critical paths remained comparable in magnitude.

Discussion

The present study examines the role of the cerebellum in brain reorganization observed in AUD, emphasizing the complementary nature of cerebellar mechanisms within the FCC. The current findings provide evidence that, after controlling for thalamic GM abnormalities, the lower the cerebellar hypermetabolism (i.e., cerebellar metabolic normalization, reflecting a tendency toward a normal level of metabolism), the stronger the fronto-cerebellar anticorrelation and the better the executive functioning. These novel associations point to complex underlying mechanisms. As the thalamus is a key relay in the fronto-cerebellar circuit, accounting for its structural abnormalities ensures that the observed effects are specifically related to cerebellar mechanisms. These findings provide a deeper understanding of the cerebellum’s particular role and help reconcile diverging evidence from PET and rs-fMRI data regarding its functional implications. Although the present study is cross-sectional and correlational, which inherently limits the establishment of causal relationships, these findings suggest that cerebellar hypermetabolism and the rs-FC anticorrelation between the cerebellum and frontal cortices may directionally influence executive performance.

Cerebellar hypermetabolism as a maladaptive plasticity phenomenon

The observed associations between cerebellar hypermetabolism, thalamic abnormalities, and lower inhibition performance support the view of cerebellar hypermetabolism as a maladaptive plasticity phenomenon, as previously described in substance use disorders [8, 11] and neurological diseases [1–7]. In addition, our data reveal a pathway linking cerebellar hypermetabolism to executive functioning through fronto-cerebellar rs-FC. Although thalamic atrophy was associated with cerebellar hypermetabolism, it did not directly contribute to the functional cascade. More specifically, the findings support a model in which a reduction in cerebellar hypermetabolism (i.e. metabolic normalization) triggers a stronger anticorrelation between the cerebellum and the frontal cortex, which, in turn, predicts better executive functioning. Conversely, higher cerebellar hypermetabolism is associated with reduced fronto-cerebellar anticorrelation (i.e. rs-FC values closer to zero) and poorer executive performance. This suggests that cerebellar hypermetabolism can be considered a pathological condition associated with less negative fronto-cerebellar anticorrelation, whereas more negative anticorrelation contributes to compensatory processes in executive performance.

Cerebellar rs-FC as a compensatory brain mechanism

Our findings emphasize that strong negative rs-FC between the cerebellum and frontal cortex can be considered a compensatory brain mechanism, as it is associated with better executive performance. Although cerebellar activity itself did not differ between AUD patients and controls (data not shown), the specific pattern of anticorrelation observed only in AUD suggests a reorganization of fronto-cerebellar dynamics that may support preserved executive functioning. Consistent with this interpretation, previous rs-FC fMRI studies have demonstrated the functional relevance and dynamic nature of posterior cerebellar connectivity. In particular, Brissenden et al. [57] showed that activity in cerebellar lobule VIII increases with working memory load and that rs-FC predicts task-related activation, indicating that rs-FC reflects the functional availability of cerebellar networks. Importantly, compensatory recruitment within large networks is not cost-free. Network reorganization has been shown to involve increased cognitive effort, reduced processing efficiency, and a trade-off between cost and efficiency, which may limit the availability of neural resources and lead to cognitive fatigue rather than uniform functional compensation [58, 59].

Our findings are consistent with previous fMRI studies reporting higher fronto-cerebellar FC in AUD patients [16–18] or in neurological diseases [12–14], interpreted as a compensatory brain mechanism supporting preserved motor and cognitive functioning. However, in these studies, fronto-cerebellar rs-FC was positive, with higher cerebellar neuronal activity coupled with higher neuronal activity in the frontal cortices. In the present study, a negative correlation between cerebellar lobule VIII and the frontal cortex was associated with preserved executive functioning and is thus considered a key feature of compensatory mechanism. Complementary analyses using other cerebellar seeds (Crus I and lobule VI) and examining group contrasts (AUD vs HC) revealed distinct patterns of fronto-cerebellar connectivity - predominantly positive and not associated with metabolic or cognitive measures – and did not show the reduced negative connectivity observed for lobule VIII (AUD < HC), supporting the specificity of the negative fronto-cerebellar coupling observed with cerebellar lobule VIII (see Supplementary material).

At the network level, negative functional connectivity is not interpreted as a functional disconnection, but rather as an organized antagonism that supports functional segregation between brain networks [60, 61]. Fox et al. [60, 61] demonstrated that anticorrelations constitute a fundamental organizing principle of intrinsic brain architecture, serving to segregate neural processes that subserve competing or opposing functional goals. Within this framework, positive correlations support integration, whereas anticorrelations support functional differentiation or reduce interference between concurrent processes. Such functional segregation may be particularly relevant for executive functions that require the suppression of competing responses [62], such as inhibitory control, by limiting interference within the fronto-cerebellar loop. Recent evidence suggests that increased anticorrelation between the cerebellum and other functional networks in older adults may reflect adaptive reorganization processes supporting cognitive functions such as executive abilities [63]. In line with our findings, where cerebellar hypermetabolism was associated with reduced anticorrelation and poorer executive performance, a stronger anticorrelation (more negative) between the cerebellum and the frontal lobe may therefore reflect a more efficient, differentiated, and coordinated organization of the fronto-cerebellar loop. Such segregation is particularly relevant for inhibitory control, which critically depends on the ability to suppress competing representations, as in the Stroop task. Importantly, this antagonistic organization remains compatible with compensatory mechanisms, as effective compensation may rely on coordinated but differentiated interactions within the circuit, rather than on synchronous co-activation of cerebellar and frontal regions. Finally, cerebellar-cortical connectivity has been shown to be state-dependent and dynamically modulated by neuro-modulatory interventions [64, 65], supporting the interpretation that the observed connectivity patterns reflect functional reorganization rather than purely structural damage.

The cascade of disruption: cerebellar hypermetabolism hinders the fronto-cerebellar connectivity and related executive compensation

The present study is not only the first to combine FDG-PET and fMRI examinations, but also the first to attempt to describe the cascade of pathophysiological mechanisms and to propose a pathway in which maladaptive and compensatory processes arise within the cerebellum. The findings add to the evidence that abnormal synaptic activity in a specific brain region may be associated with compromised functional connectivity between that region and other brain structures, reflecting a desynchronization that may impair compensatory processing within the circuit. Previous PET/fMRI studies have shown that regional glucose metabolism is related to functional connectivity in healthy controls [66], and that hypometabolism is associated with disruption of intrinsic functional connectivity [24–26] or of brain networks [8] in patients. Our findings are the first to demonstrate that regional hypermetabolism may alter functional connectivity. More specifically, cerebellar hypermetabolism appears to disrupt rs-FC, resulting in a reduction (close to zero) of the negative connectivity between the cerebellum and frontal cortex and a functional disconnection between these two brain regions. This desynchronization may underlie executive dysfunction. Taken together, our findings suggest that cerebellar hypermetabolism is a sign of aberrant synaptic activity and is associated with reduced functional connectivity with the frontal cortex, disrupting the fronto-cerebellar functional connectivity loop. In line with this interpretation, cerebellar hypermetabolism may be viewed as a metabolic alteration involved in alcohol-related brain dysfunction, positioned upstream of functional desynchronization and executive impairment within the statistical cascade identified here. This cascade of pathological events may underlie the executive deficits observed in AUD patients. However, longitudinal studies are required to establish the temporal sequence of these alterations.

Strengths and limitations

A major strength of the present study is its use of a multimodal neuroimaging approach combining FDG-PET and rs fMRI, despite the considerable challenges of recruiting and clinically well-characterized AUD patients who met the strict inclusion criteria and quality control requirements for both imaging modalities.

A methodological limitation should be acknowledged regarding the delay between neuroimaging and neuropsychological assessments, which may introduce variability. Although longitudinal studies suggest limited recovery of complex executive abilities and functional connectivity during the first weeks of abstinence [67–69], short-term brain and cognitive changes in early abstinence cannot be fully excluded. In addition, patients were assessed shortly after withdrawal, which may represent a potential confounding factor [70–72]. Although the AUD group was predominantly male, additional analyses indicated that sex did not contribute to any path in the model. However, given the small number of female participants, the absence of detected sex effects should be interpreted with caution, and future studies including more balanced male and female samples are needed. Finally, tobacco use - a frequent comorbidity in AUD - may represent a potential confounding factor [73, 74]. Nevertheless, our results do not support a major contribution of tobacco use to the neural alterations observed in the present study.

Conclusion

The present combined PET and fMRI investigation was conducted in AUD as a model of cerebellar abnormalities. It provides novel insights into the synaptic, structural, and neuronal brain mechanisms occurring in the cerebellum. Our findings support the view that cerebellar hypermetabolism is a disruptive phenomenon that contributes to dysfunction of the fronto-cerebellar loop. Strong fronto-cerebellar anticorrelation, on the other hand, may serve as a compensatory mechanism related to executive functioning. Our path analysis reconciles these seemingly diverging findings and proposes a global framework in which normalization of cerebellar hypermetabolism promotes the emergence of compensatory mechanisms, as reflected by stronger negative fronto-cerebellar rs-FC and improved executive performance. Taken together, these multimodal functional insights pave the way for clarifying how each neuroimaging marker contributes to the understanding of underlying pathophysiological mechanisms and compensational patterns in neurological and psychiatric disorders.

Author contributions

Conception or design: Pitel. Acquisition of data: Ritz, Pitel. Statistical analyses: Morand, Ritz. Interpretation: Ritz, Segobin and Pitel. Drafting: Ritz. Resources: Cabé, Laniepce. Revision: Segobin and Pitel. Final approval: all authors.

Funding

This work was supported by the French National Institute for Health and Medical Research (INSERM), the French National Agency for Research (ANR) Postdoc Return (Retour Post-Doctorants, PDOC) program, the Regional council of Lower-Normandy, and the Mission Interministerielle de Lutte contre les Drogues Et les Conduites Addictives (Interministerial Mission for the Fight Against Drugs and Addictive Behaviors) (MILDECA). The funding sources had no role in the study design, data collection, management, analysis, or interpretation; manuscript preparation, review, or approval; or the decision to submit the manuscript for publication.

Data availability

Data are available upon request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41398-026-04217-w.