The cholinergic hypothesis suggests that the progressive degeneration of cholinergic neurons in the basal forebrain (BF) of Alzheimer’s disease (AD) patients is responsible for memory loss, disorientation, and behavioral and personality changes [1]. Donepezil, a cholinesterase inhibitor, increases acetylcholine levels and improves cognition in AD, but its precise effects on BF pathways remain unclear [2].

Insights into donepezil’s therapeutic mechanisms in Alzheimer’s disease have been gained through functional magnetic resonance imaging (fMRI) [3, 4, 5, 6]. After 12 weeks of treatment, Goveas et al. [3] found that cognitive improvements in mild AD patients were linked to specific patterns of hippocampal functional connectivity. They found that the inferior frontal gyrus, dorsolateral prefrontal cortex, and left parahippocampal gyrus were the key regions underlying cognitive improvement. Cheng et al. [4] also found that AD patients had lower regional homogeneity in the right gyrus rectus, right precentral gyrus, and left superior temporal gyrus after 24 weeks of donepezil treatment. Zheng et al. [5] reported that after 6 months of donepezil treatment, AD patients showed improved cognitive and non-cognitive symptoms, which were associated with the increased amplitude of low-frequency fluctuations in cerebellar and frontostriatal regions. White-matter imaging also showed enhanced connectivity in the nucleus accumbens shell and anterior limb of the internal capsule, correlating with acetylcholinesterase inhibitors (AChEIs) exposure [6].

Although those studies have provided valuable insights into the effects of donepezil on distributed brain networks, they have not specifically targeted the basal forebrain cholinergic system—the primary site of cholinergic neuron degeneration in AD and the direct pharmacological target of cholinesterase inhibitors. This gap is significant because understanding how donepezil modulates the functional architecture of its primary target system would provide more direct evidence of its therapeutic mechanism.

A key cholinergic source is the nucleus basalis of Meynert (NBM) in the BF, the atrophy of which has been shown to be correlated with AD severity [7, 8, 9]. Grothe et al. [8] found cholinergic BF atrophy rates were higher than ageing-related global brain shrinkage rates in AD patients. Schumacher et al. [9] reported higher mean diffusivity in NBM-originating white-matter tracts; this was linked to increased dementia risk and global cognitive impairment. Those structural findings suggested that the NBM and its projection network are critically compromised in AD, raising the question of whether pharmacological intervention with donepezil can modulate the functional integrity of this system. However, to date, the specific impact of donepezil treatment on functional connectivity within NBM-centered cholinergic networks remains poorly understood.

The present longitudinal, resting-state fMRI (rs-fMRI) study investigated donepezil-induced changes in cholinergic pathway networks in AD. Using BF subregions as regions of interest, we tested the hypothesis that donepezil treatment modulates functional connectivity within BF-centered networks. Specifically, based on previous evidence [3, 4, 5, 6, 7, 8, 9], we predicted increased functional connectivity between the NBM and prefrontal/cingulate regions, and that these changes were correlated with cognitive improvements.

Using rs-fMRI and functional connectivity analysis, this study examined changes in BF networks in AD patients pre- vs. post-treatment with donepezil. Compared to baseline, post-treatment patients showed decreased BF functional connectivity in the left cerebellar lobule VI. Relative to HCs, treated patients exhibited lower BF functional connectivity in the right PoG and higher functional connectivity in the left cerebellum.

Studies have consistently implicated the PoG in AD. Diffusion imaging showed altered microstructure in the PoG, precentral gyrus, and superior temporal gyrus of AD patients [17]. Non-memory agnosia in AD has been correlated with episodic/semantic memory and gray matter volumes in regions including the anterior cingulate cortex, precentral gyrus, superior frontal gyrus, PoG, and lingual gyrus [18]. Network analyses have further revealed centrality loss in the precuneus, precentral gyrus, and PoG during AD progression [19]. Longitudinal studies have shown higher $\beta$-amyloid accumulation in the PoG of AD patients [20]. Reduced perfusion in both the PoG and basal forebrain has been shown to predict progression from mild cognitive impairment to AD [21]. A positron emission tomography (PET) study of galantamine (another cholinesterase inhibitor) combined with citalopram showed increased metabolism in the right middle frontal gyrus, right PoG, right superior temporal gyrus, and right cerebellum [22]. Animal research has provided mechanistic insight. Using 2-photon imaging in mice, Shanazz et al. [23] demonstrated that basal forebrain stimulation (60–130 Hz) increased cortical acetylcholine release. Donepezil administration potentiated that response in a dose-dependent fashion, thereby enhancing the integrated response by 57% (2 mg/kg) and 126% (4 mg/kg). Those findings from human neuroimaging and animal models have shown the relevance of the PoG in AD pathophysiology and treatment response, supporting the functional connectivity changes observed in our study after donepezil treatment.

Meta-analyses confirmed significant cerebellar atrophy in AD patients, which was correlated with cognitive decline [24]. Notably, early-onset AD patients exhibited substantial amyloid-beta (A$\beta$) deposits in the cerebellum [25]. Post-mortem tissue studies revealed a more than fourfold higher level of both A$\beta$ expression and load in the cerebellums of AD patients than in those of controls. Furthermore, the cerebellar microglia of ADs showed significantly elevated ionized calcium-binding adapter molecule 1 (IBA1) protein expression and load (91% and 69% increases, respectively), accompanied by a reduction in the length and branching complexity of their processes [26]. Those established pathological alterations in the cerebellum support our functional connectivity findings. The involvement of the cerebellum extends across the spectrum of Alzheimer’s disease and related dementias (ADRD). A recent large-scale study demonstrated that distinct patterns of cerebellar atrophy were observed in all ADRD subtypes, with significant gray matter changes appearing in the early stages of most subtypes—including the very early stages of AD, Lewy body disease (LBD)-AD, frontotemporal lobar degeneration (FTLD)-transactive response DNA binding protein (FTLD-TDP), and progressive supranuclear palsy [27]. Furthermore, cortical atrophy was found to positively predict cerebellar atrophy across all subtypes, suggesting that cerebellar degeneration is closely coupled with cortical pathology [27]. These findings reinforce the notion that the cerebellum is not spared in neurodegenerative diseases and may serve as an early indicator of widespread network dysfunction.

The observed functional connectivity changes between the NBM and the cerebellum/PoG in the present study warrant discussion. This was initially unexpected because basal forebrain cholinergic neurons do not provide direct monosynaptic input to these regions. However, we should clarify that functional connectivity is not the same as direct anatomical connectivity. Functional connectivity reflects statistical dependencies that can arise through polysynaptic pathways, intermediary regions, or common modulatory inputs. This perspective is supported by growing evidence implicating the cerebellum in cognition—including attention, executive function, and working memory—beyond its traditional motor role [28]. The connections of the cerebellum with associative cortical regions provide a substrate for its cognitive contributions. Thus, altered NBM-cerebellar functional connectivity may have reflected modulation of cerebello-cerebral circuits via indirect pathways (e.g., brainstem, thalamus, or cholinergically-innervated cortical regions). Similarly, PoG connectivity changes were likely to involve multi-synaptic mechanisms. Anatomically, cholinergic input to the sensory cortex originates from the NBM, whereas cerebellar cholinergic afferents arise from the pedunculopontine tegmental nucleus (PPN), a key regulator of movement, gait, and postural control [29]. The PPN may therefore serve as a functional interface linking cerebellar motor circuits with basal forebrain cognitive networks [30]. Notably, gait and balance impairments are recognized as early manifestations of AD and correlate with cognitive decline [31, 32]. Donepezil has been demonstrated to improve gait in AD patients [33]. Our findings suggested that donepezil alleviated AD symptoms, in part, by modulating cholinergic pathways involved in motor integration. The observed connectivity changes involving the cerebellum and PoG may have reflected enhanced sensorimotor integration, a process fundamental to both motor function and cognition. The mirror neuron system (MNS), which includes the PoG and receives cerebellar modulation via inhibitory interneurons, is central to this integration, enabling the updating of internal sensorimotor models based on sensory feedback [34, 35]. Cholinergic modulation influences MNS activity [35], and our findings of increased NBM-cerebellum and altered PoG connectivity may have represented enhanced cholinergic drive to these circuits. This interpretation aligns with evidence that donepezil improves gait in AD patients [33], a function heavily dependent on intact sensorimotor integration, and suggests that treatment benefits may extend beyond cognition to the real-time calibration of action-perception loops. Collectively, these findings indicated the distributed effects of cholinergic enhancement beyond direct projection targets. Previous literature is limited, with only one study reporting NBM-cerebellum connectivity in healthy controls [34]. Future studies using effective connectivity methods (e.g., dynamic causal modeling) or molecular imaging could help clarify the pathways underlying these network-level changes and test the hypothesis of PPN-mediated connectivity.

The present study had several limitations. First, the sample size in this study was relatively small, which represented a key limitation. The modest sample size increased the risk of type II errors (missing true effects) and potential overestimation of observed effects, and limited the generalizability of our findings. Second, this study did not include biomarker confirmation for AD, such as CSF biomarkers or PET imaging due to hospital condition limitations and the difficulty in obtaining cerebrospinal fluid samples, we would add plasma biomarkers to future sample collection. Third, we acknowledge a methodological limitation concerning the duration of our rs-fMRI acquisition. Each scan lasted 6 min, which, although consistent with many clinical studies involving AD patients, was shorter than the durations recommended for optimal test-retest reliability of functional connectivity estimates. Our 6-min protocol was necessitated by practical considerations related to patient tolerability and motion control in an elderly AD cohort. Nonetheless, we recognize that this may have increased variability in our connectivity estimates. Future studies should incorporate longer scan durations or multi-session acquisitions to enhance reliability, particularly if individual-level analyses are pursued. Fourth, the clinical symptoms are mainly assessed using the MMSE, but without evaluations of olfaction, sleep, and other aspects, it cannot fully reflect the efficacy of the drug and its connection to brain neural mechanisms.

It is worth noting that AD progression and treatment responses vary considerably among individuals. Longitudinal studies are essential to determine whether the observed connectivity improvements persist over time and how they relate to cognitive outcomes. Also, AD is a heterogeneous disease due to the presence of different genetic phenotypes, which likely contributed to the variable therapeutic responses. Future investigations should therefore stratify patients based on genetic markers (e.g., apolipoprotein E (APOE) genotype), phenotypic characteristics, and clinical variables, to better understand the determinants of treatment response and to facilitate the development of more personalized therapeutic approaches.

This exploratory study based on real-world clinical data has uncovered the neural mechanisms of donepezil intervention in Alzheimer’s disease, which to some extent deepens our understanding of both disease pathophysiology and therapeutic effects. These findings also provide a direction for future research. However, further studies with more rigorous designs—including randomized controlled trials, biomarker-based patient stratification, and multimodal neuroimaging—are still needed to validate and extend these findings. In particular, given recent evidence suggesting that the cerebellum may serve as a synergistic therapeutic target in other neurodegenerative conditions such as frontotemporal dementia (FTD) [36], future studies could explore whether combining cholinergic therapy with cerebellar neuromodulation yields enhanced clinical benefits in AD.

In the present study, we compared BF functional connectivity among AD patients at baseline and after donepezil treatment, as well as with HCs. The results revealed intervention-related functional alterations in the right PoG and left cerebellar lobule VI. BF–PoG connectivity may serve as a potential neuroimaging marker for evaluating the therapeutic effect of donepezil treatment, though its utility as a specific treatment target requires further validation.

The datasets used and analysed during the current study are available from the corresponding author upon reasonable request.

YG and FW contributed to the acquisition of the data and drafted the manuscript. ZZ, HL, and BC collected the data, XL and ZG conceived, analyzed the data and revised the manuscript. All authors contributed to editorial changes in the manuscript. All authors read and approved the final manuscript. All authors have participated sufficiently in the work and agreed to be accountable for all aspects of the work.

This study was performed in accordance with the principles of the Declaration of Helsinki. Approval was granted by the ethics committee of Tongde Hospital of Zhejiang Province (approval no. 2017-11-12). Written informed consent was obtained from all participants.

We thank Sarina Iwabuchi, PhD, from Liwen Bianji (Edanz) for editing the language of a draft of this manuscript.

This research was supported by the General Project of the Department of Science and Technology of Zhejiang Province (2017KY109 and 2020358406 to XL), the General Project of the Department of Science and Technology of Zhejiang Province (2018KY031 and 2024KY873 to ZG), and the National Leading Medical Specialty Development Project-Department of Geriatrics, Tongde Hospital of Zhejiang Province (Project Number: [2024]90662), and Zhejiang Provincial Alliance of Traditional Chinese Medicine Advantage Specialty for Geriatric Diseases (Project Number: [2024]10).

The authors declare no conflicts of interest.