Glucose metabolism alterations and Aβ deposition in AD and FTD are related to the distribution of neurotransmitter systems

Abstract

Objective

This study aimed to elucidate the spatial correlations among alterations in glucose metabolism, amyloid-beta (Aβ) deposition, and neurotransmitter systems across Alzheimer’s disease (AD), mild cognitive impairment (MCI) and frontotemporal dementia (FTD), while assessing their associations with clinical cognitive decline.

Methods

In this retrospective cohort study, 507 participants (261 AD, 111 MCI, 62 FTD and 73 normal controls) underwent multimodal neuroimaging, including ¹⁸F-FDG PET, ¹⁸F-AV45 Aβ PET, and structural MRI. Spatial co-localization of imaging alterations with neurotransmitter receptor/transporter distributions was assessed using the JuSpace toolbox. Spearman correlations evaluated associations between imaging-neurotransmitter co-localization and cognitive scores. False discovery rate (FDR) correction was used to control for P < 0.05 for all analyses.

Results

AD showed glucose hypometabolism in temporoparietal and frontal regions, while FTD was observed in the frontotemporal areas. Spatial co-localization analyses revealed subtype-specific neurotransmitter vulnerabilities: AD glucose hypometabolism correlated with serotonergic, γ-aminobutyric acidergic (GABAergic), dopaminergic, and glutamatergic systems, while FTD correlated with serotonergic, dopaminergic, and opioid receptors. Aβ deposition co-localized with 5HT2a receptor, γ-aminobutyric acid type A (GABAa) receptors, and noradrenaline transporter (NAT) in AD, as well as D1 receptor in MCI. In AD, FDG or Aβ PET-neurotransmitter correlations significantly associated with MMSE/MoCA scores, while Aβ-serotonin transporter (SERT) or Fluorodopa (FDOPA) correlations linked to cognitive decline in Aβ-positive MCI (P < 0.05).

Conclusion

This study demonstrates that AD and FTD exhibit unique spatial vulnerabilities in neurotransmitter systems, closely tied to glucose hypometabolism and Aβ pathology. The identification of disease specific neuroimaging-neurotransmitter signatures advances biomarker development and supports targeted therapeutic strategies tailored to molecular pathways.

Clinical trial number: not applicable.

Introduction

Alzheimer’s disease (AD) and frontotemporal dementia (FTD) are the two most common types of dementia with the highest prevalence. They are characterized by progressive cognitive and behavioral impairment, which seriously affects daily functioning, learning ability, occupational performance and social participation. Mild cognitive impairment (MCI) constitutes prodromal stages of AD, wherein individuals exhibit measurable cognitive deficits that remain subthreshold for dementia diagnosis. According to the main clinical manifestations, FTD is divided into three subtypes: behavioral variant of FTD (bvFTD), non-fluent variant of primary progressive aphasia (nfvPPA), and semantic variant of PPA (svPPA) [1–3]. The substantial symptomatic overlap during early disease stages creates diagnostic ambiguity, complicating timely clinical intervention. Therefore, understanding the underlying mechanisms leading to cognitive decline is essential to improve diagnostic accuracy and identify therapeutic targets.

Neuroimaging modalities have emerged as indispensable tools for both diagnostic evaluation and pathophysiological investigation of cognitive disorders. Fluorine-18-fluorodeoxyglucose positron emission tomography (¹⁸F-FDG PET), which quantifies regional cerebral glucose metabolism through standardized uptake value ratio (SUVR) measurements, serves as the most common, widely used, or well-established biomarker for assessing neuronal metabolic activity and functional integrity in cognitive disorders [4]. Besides, the accumulation of amyloid-β (Aβ) plaques is recognized as a principal neuropathological driver in AD, initiating cascades of neuronal degeneration that ultimately manifest as clinical symptomatology. Aβ PET enables in vivo visualization of Aβ deposition in the brain by specific binding of radiotracer to reflect AD pathology [5]. The multimodal integration of ¹⁸F-FDG PET and Aβ PET imaging data enables comprehensive characterization of disease pathophysiology, facilitating the identification of presymptomatic biomarkers for cognitive deterioration.

In addition to FDG and Aβ PET, the role of neurotransmitters in cognitive disorders is increasingly recognized as an essential aspect of understanding disease progression. The brain comprises different types of neurons to regulate various physiological functions in the human body. Neurotransmitters are chemical messengers secreted by neurons and are involved in the propagation of nerve impulses from one neuron to another, thus sending accurate information that enables effector organs to carry out specific functions [6, 7]. Based on the type of neurotransmitter secreted by neurons, neurotransmitters have been classified into different systems: cholinergic, glutamatergic, γ-aminobutyric acidergic (GABAergic), dopaminergic, and serotonergic [8]. Each neurotransmitter system has a particular role in the nervous system and is essential for sustaining healthy physiological and behavioural functions in the body. Emerging evidence suggests that neurotransmitter redistribution may mediate the pathophysiological coupling between cerebral hypometabolism and Aβ accumulation, underscoring their tripartite interaction in disease progression [9–12]. Understanding how neurotransmitter systems interact with brain metabolism and amyloid plaques is key to gaining a deeper understanding of cognitive decline and identifying potential biomarkers for early diagnosis and therapeutic targets.

The multimodal integration of neurotransmitter receptor mapping with macroscopic neuroimaging parameters offers a powerful framework for dissecting molecular pathways underlying physiological aging and pathological neurodegeneration. However, the cost of PET and the absence of in vivo radioligands have hindered extensive case-control imaging studies for various receptor types in disease populations. Nevertheless, template-based approaches using normative neurotransmitter maps derived from healthy controls have demonstrated construct validity, showing significant spatial concordance with macroscopic imaging biomarkers across clinical populations [13]. Advanced neuroimaging approaches capable of integrating data from different modalities are necessary for investigating these intricate interactions. JuSpace is an advanced software platform designed for the processing of multimodal neuroimaging data, offering an extensive perspective on the brain’s functional and structural alterations in cognitive disorders. Utilizing sophisticated algorithms, JuSpace can synchronize and examine these datasets, enabling researchers to investigate the geographical and temporal correlations among glucose metabolism, amyloid deposition, and neurotransmitter distribution [14–17]. The objective of this study was to investigate the relationship between altered glucose metabolism and the spatial distribution of Aβ deposition in relation to particular neurotransmitter systems in AD and FTD, and to determine if these relationships correlate with cognitive function.

Materials and methods

This was a retrospective study approved by the Ethics Committee of Xuanwu Hospital, Capital Medical University, China (Ethics Number:[2023]044). Written informed consent was obtained from all participants or legally authorized representatives prior to study enrollment.

Participants

Between March 2020 and February 2024, we included 507 participants through the Neurodegenerative Disease Cohort at Xuanwu Hospital, comprising 261 AD, 111 MCI, 62 FTD, and 73 age-/gender-matched normal controls (NC). All the dataset comprised of the ¹⁸F-FDG PET and paired high-resolution T1-weighted imaging (3D T1WI), and ¹⁸F-AV45 PET data were also included for AD and MCI participants. All Aβ-PET and ¹⁸F-FDG PET scans were performed within a 7-day interval to minimize physiological variance. Clinical diagnosis was established based on a standard dementia screening, including medical history review, physical and neurological examinations, laboratory tests, neuropsychological tests, and brain ¹⁸F-FDG PET/MR scans. The diagnosis of AD was based on the National Institute on Aging-Alzheimer’s Association workgroups guidelines for probable AD and Jack et al. 2024 revised criteria for the diagnosis and staging of Alzheimer’s disease in the presence of Aβ positive as confirmed by Aβ PET or CSF examination [18, 19]. The diagnosis of MCI was based on diagnostic criteria defined by Petersen [20], encompassing the following: (1) self-reported cognitive decline noted by the individual, family members, or an informant; (2) MoCA scores at least 1.5 standard deviations below the mean for their age and educational level; (3) predominantly intact cognitive function; and (4) exclusion of dementia or any physical or mental conditions that could impact cognitive function. The FTD patients consisted of 48 bvFTD and 14 svPPA. BvFTD diagnoses required fulfillment of the International bvFTD Criteria Consortium (FTDC) diagnostic criteria [21], including progressive behavioral disinhibition, apathy, and loss of empathy. Participants diagnosed with svPPA met the criteria proposed by Gorno-Tempini [22]. The NC participants were age- and gender-matched to patients and had no cognitive decline complaints, depression, or anxiety, with the Mini-Mental State Examination (MMSE) score ≥ 24. All patients received thorough assessments from specialists in neurology and radiology and nuclear medicine to confirm the diagnosis. Exclusion criteria were diabetes, severe white matter injury (Fazekas scores higher than 2) and other neurologic, psychiatric, or brain parenchyma diseases (e.g., stroke, tumors, and trauma) potentially leading to cognitive impairment. Furthermore, participants with a history of alcohol or substance abuse, significant visual or auditory impairments that could hinder the completion of neuropsychological assessments or adherence to examination directives, or medical conditions precluding PET-MR scans were excluded.

Image acquisition, preprocessing

All data were acquired using the simultaneous PET/MR 3.0-Tesla system (uPMR 790, United Imaging Healthcare, Shanghai, China; Signa, GE Healthcare, Waukesha, WI, USA). A 19-channel or 24-channel head-neck phased array coil ensured simultaneous PET-MR acquisition with motion correction via optical tracking. Participants underwent scanning in the resting state (eyes closed, no auditory stimulation) after 8-hour fasting to standardize metabolic conditions. MRI sequence parameters were as follows: Sagittal T1-weighted three-dimensional (3D) turbo field echo, repetition time (TR)/echo time (TE) = 8.5 ms/3.2 ms, flip angle = 15°, voxel size = 1 × 1 × 1 mm3, several slices = 188, matrix = 256 × 256, field of view (FOV) = 256 mm. Ten-min ¹⁸F-FDG PET data were acquired simultaneously 40 min after 5.6–8.2 mCi ¹⁸F-FDG tracer injection with 3D list-mode. Patients with AD and MCI underwent a 20-minute Aβ PET scan 50 min following the injection of 10 mCi of ¹⁸F-AV45. Corrected PET data were obtained using a time-of-flight, point spread function, ordered subset expectation maximization (TOF-PSF-OSEM) algorithm with 8 iterations and 32 subsets, and a 3-mm cut-off filter. The image matrix was 192 × 192, the field of view was 35 × 35cm², and the pixel size was 1.82 × 1.82 × 2.78mm³.

The PET and structural MRI data were further processed by partial volume correction (PVC) and spatial normalization, both using Statistical Parametric Mapping (SPM12, Wellcome Department of Imaging Neuroscience, Institute of Neurology, London, UK) in the MATLAB R2020b (The MathWorks Inc., Natick, MA), as reported in our earlier studies [23, 24]. All the ¹⁸F-FDG PET and ¹⁸F-AV45 Aβ PET images were coregistered to the individuals’ structural MRI images and normalized to standard Montreal Neurologic Institute (MNI) space using an MRI template. Both were filtered to remove the background and skull. Then, the ¹⁸F-FDG PET images were transformed into maps representing the SUVR using pons as a reference region [25–28], and SUVR map of ¹⁸F-AV45 PET was calculated relative to the cerebellum [29]. Finally, all images were smoothed using an isotropic Gaussian kernel at full width at half maximum (FWHM) of 8 mm in all directions. SUVR maps were used for comparison between groups in subsequent statistical analysis. The correlation between ¹⁸F-FDG and ¹⁸F-AV45 Aβ PET t-maps and neurotransmitters was analyzed by the JuSpace toolbox after Fisher’s z-transformation.

The JuSpace toolbox was employed to examine if variations in ¹⁸F-FDG and Aβ PET SUVR in patients, compared to normal controls, correlate with various neurotransmitter systems [14]. The JuSpace toolkit facilitates cross-modal spatial correlations between diverse neuroimaging modalities and nuclear imaging data regarding the relative density distribution of various neurotransmitter systems. Specifically, we aimed to evaluate whether the spatial configuration of ¹⁸F-FDG and ¹⁸F-AV45 Aβ PET SUVR maps in patients compared to NC resembles the distribution of neurotransmitter maps derived from nuclear imaging in independent healthy volunteer cohorts, including 5-HT1a receptor [30], 5-HT1b receptor [30], 5-HT2a receptor [30], serotonin transporter (SERT) [30], D1 receptor [31], D2 receptor [32], dopamine transporter (DAT) [33], Fluorodopa (FDOPA) [34], γ-aminobutyric acid type A (GABAa) receptors [33, 35], vesicular acetylcholine transporter (VAChT) [36], metabotropic glutamate receptor 5 (mGluR5) [36], µ-opioid (MU) receptors [37], kappa-opioid (KappaOp) receptors [38], and noradrenaline transporter (NAT) [39].

Statistical analysis

Statistical analyses were performed using Statistical Package for Social Science version 26.0. Continuous variables with normal distribution were presented as mean ± standard deviation and compared between groups by independent T-test, while categorical variables were presented as count (with percentages) and tested using the Chi-square test. The voxel-based two-sample T-tests were performed using SPM12 software to compare the difference in ¹⁸F-FDG PET SUVR images between patients and NC with age, sex and education years as factors. Uncorrected P < 0.001 at voxel level and false discovery rate (FDR) corrected using the Benjamini-Hochberg procedure at P < 0.05 were considered statistically significant.

Using the JuSpace toolbox, mean values were extracted from both neurotransmitter and SUVR maps using gray matter regions from the neuromorphometrics atlas. We z-transformed the ¹⁸F-FDG PET t-maps after group comparisons and then entered the z-transformed t-maps into the JuSpace toolbox to test Spearman correlation coefficients between the ¹⁸F-FDG PET differential brain regions and their respective neurotransmitter maps for each patient group relative to NC. For ¹⁸F-AV45 PET, we first performed a one-sample t-test for the AD and MCI groups. Similar to ¹⁸F-FDG PET, z-transformed t-maps were used in the JuSpace toolbox to examine the Spearman correlation coefficients between the ¹⁸F-AV45 PET and the distribution of neurotransmitter maps in the AD and MCI groups. Exact permutation-based p values as implemented in JuSpace (10,000 permutations randomly assigning group labels using orthogonal permutations) were computed to test if the distribution of the observed Fisher’s z-transformed individual correlation coefficients significantly deviated from zero.

To test whether ¹⁸F-FDG and Aβ PET SUVR - neurotransmitter correlations are associated with symptoms of patients, we first z-transformed ¹⁸F-FDG and AV45 PET SUVR images and then input all SUVR images of each group into JuSpace toolbox to output the correlation between SUVR images and neurotransmitter profiles. Spearman correlation coefficients were subsequently calculated between significant SUVR-neurotransmitter correlations (Fisher’s z-transformed Spearman correlation coefficient from JuSpace toolbox output) and clinical scales. FDR correction was applied across all hypothesis families using the Benjamini-Hochberg procedure at P < 0.05.

Results

Demographics and clinical characteristics

A total of 507 subjects comprising 261 AD, 111 MCI, 62 FTD, and 73 NC participants were included in this study. The demographic characteristics of all participants are listed in Table 1. Significant group differences were found in education years, MMSE, and MoCA scores. All patient groups exhibited significantly lower MMSE and MoCA scores compared to NC subject (P < 0.001). Both AD and MCI patients included completed ¹⁸F-AV45 Aβ PET examination, and all AD patients were positive, while 66 of 111 MCI patients were positive.

Table 1.

Demographics of the cohort

	AD (N = 261)	MCI (N = 111)	FTD (N = 62)	NC (N = 73)	Group comparison P Patients vs. NC
					AD	MCI	FTD
Female (percentage)	178 (68.2%)	69 (62.2%)	37(59.7%)	46 (63.0%)	0.405	0.907	0.691
Age	63.93 (8.30)	64.54 (7.07)	62.71 (6.42)	62.77 (9.58)	0.301	0.153	0.968
Education years	10.20 (5.32)	12.70 (3.35)	9.18 (4.13)	13.22 (2.97)	< 0.001	0.285	< 0.001
MMSE	15.60 (6.13)	25.45 (2.83)	16.65 (7.34)	28.52 (1.61)	< 0.001	< 0.001	< 0.001
MoCA	11.21 (5.85)	21.09 (3.60)	9.76 (5.94)	25.63 (2.99)	< 0.001	< 0.001	< 0.001
Aβ positive (percentage)	261(100.0%)	66(59.5%)	-	-	-	-	-

Data were presented with mean (standard deviation). Group comparisons: independent T-test (age, education, MMSE, and MoCA), chi-square test (gender). MMSE Mini-mental state examination; MoCA Montreal cognitive assessment. Statistical significance set at P < 0.05

Voxel-wise based group differences of ¹⁸F-FDG PET SUVR

Initially, we assessed group differences in ¹⁸F-FDG PET SUVR between NC and patients. The voxel-wise based analysis showed significantly reduced ¹⁸F-FDG PET SUVR in the bilateral parietotemporal regions, frontal lobes, insular and part of occipital cortex, cingulate cortex, and subcortical gray matter areas, including amygdala and caudate nucleus in AD patients compared with NC participants. In MCI patients, a pattern of reduced ¹⁸F-FDG PET SUVR in the bilateral frontal lobe, cingulate cortex, precuneus, insular lobe, and temporal lobe including temporal pole and hippocampus. The FTD group displayed a pattern of significantly reduced ¹⁸F-FDG PET SUVR in the extensive frontotemporal cortex, insula cortex, part of the parietal-occipital cortex and subcortical nucleus compared with NC. The comparison between groups without PVC showed that the hypometabolism range of MCI group and FTD group was reduced. FDR correction was applied across all hypothesis families using the Benjamini-Hochberg procedure at P < 0.05. Voxel-wise spatial maps of the ¹⁸F-FDG PET SUVR with and without PVC showing the respective regional patterns of hypometabolism are presented in Fig. 1.

Fig. 1.

Spatial maps of the voxel-wise based analysis with the patterns of reduced ¹⁸F-FDG PET SUVR after (a) and before PVC (b) in patients with AD, MCI, and FTD compared with NC participants (P_FDR < 0.05). Colors indicate t scores

Spatial correlation with neurotransmitter maps

We performed correlation analyses to test if ¹⁸F-FDG PET SUVR alterations in patients significantly co-localize with the spatial distribution of specific neurotransmitter systems (P_FDR<0.05, Fig. 2). In AD patients as compared to NC, ¹⁸F-FDG PET SUVR alterations after PVC were significantly associated with the spatial distribution of 5-HT1a (r=−0.737), 5-HT2a (r=−0.483), GABAa (r=−0.342), mGluR5 (r=−0.498), and D2 (r=−0.280). In MCI patients as compared to NC, ¹⁸F-FDG PET SUVR alterations were significantly associated with the spatial distribution of 5-HT1a (r=−0.521), 5-HT2a (r=−0.220), 5-HT1b (r = 0.362), SERT (r=−0.302), mGluR5 (r=−0.336), D2 (r=−0.349), FDOPA (r=−0.279), NAT (r=−0.362), KappaOp (r=−0.371) and VAChT (r=−0.228). In FTD patients as compared to NC, ¹⁸F-FDG PET SUVR alterations were significantly associated with the spatial distribution of 5-HT1a (r=−0.592), 5-HT4 (r=−0.409), SERT(r=−0.302), D1 (r=−0.389), D2 (r=−0.495), NAT (r = 0.229), MU (r=−0.675), and KappaOp (r=−0.433).

Fig. 2.

Results of spatial correlation analyses of ¹⁸F-FDG PET SUVR after (a) and before PVC (b) with spatial distribution of neurotransmitter systems in patients with AD, MCI, and FTD (FDR corrected, *: P_FDR<0.05, **: P_FDR<0.01, ***: P_FDR<0.001)

Furthermore, we examined if similar co-localization patterns are observed with Aβ PET SUVR in AD and MCI groups (P_FDR<0.05, Fig. 3). In AD patients, Aβ PET SUVR were significantly associated with the spatial distribution of 5-HT2a (r = 0.301), GABAa (r = 0.277), and NAT (r = 0.316). Only the spatial distribution of D1 (r=−0.316) was significantly associated with Aβ PET SUVR of MCI patients. Considering that 66 of 111 MCI patients were Aβ positive, we then observed the correlation between Aβ deposition and spatial distribution of neurotransmitters in Aβ negative and positive group respectively. As for Aβ positive MCI group, Aβ deposition were significantly associated with the spatial distribution of SERT (r=−0.257), D1 (r=−0.347), DAT (r=−0.283), and FDOPA (r=−0.414). In Aβ negative MCI group, Aβ PET SUVR was significantly associated only with the spatial distribution of mGluR5 (r = 0.418).

Fig. 3.

Results of spatial correlation analyses of Aβ PET SUVR with spatial distribution of neurotransmitter systems in patients with AD (a), MCI (b), Aβ positive MCI (c), and Aβ negative MCI (d) (FDR corrected, *: P_FDR<0.05, **: P_FDR<0.01)

Relationship to clinical symptoms

In addition, we evaluated if the significant ¹⁸F-FDG and Aβ PET SUVR-neurotransmitter correlation coefficients are also connected with clinical scales of patients. After FDR correction, the strength of ¹⁸F-FDG PET SUVR after PVC co-localization with 5-HT1a, 5HT2a, GABAa, and mGluR5 distribution exhibited significant associations with MMSE and MoCA in AD patients (Fig. 4). The strength of ¹⁸F-FDG PET SUVR before PVC co-localization with 5-HT1a, 5HT2a, D2, DAT, FDOPA, GABAa, SERT, and mGluR5 distribution exhibited significant associations with MMSE and MoCA in AD patients (Fig. 5). For Aβ PET SUVR, co-localization with 5-HT2a, GABAa, and NAT distribution exhibited significant associations with MMSE and MoCA in AD patients (Fig. 6a and b). Only the strength of Aβ PET SUVR co-localization with SERT and FDOPA distribution were significantly associated with MoCA in Aβ positive MCI patients (Fig. 6c).

Fig. 4.

Correlations of MMSE (a) and MoCA (b) with ¹⁸F-FDG PET SUVR–neurotransmitter strength of association after PVC with 5-HT1a, 5HT2a, GABAa, and mGluR5 in AD patients (P < 0.001). Symbols represent individual Fisher’s z-transformed spearman correlation coefficients between ¹⁸F-FDG PET SUVR–neurotransmitter correlations after PVC and each neuropsychological scale

Fig. 5.

Correlations of MMSE (a) and MoCA (b) with ¹⁸F-FDG PET SUVR–neurotransmitter strength of association before PVC with 5-HT1a, 5HT2a, D2, DAT, FDOPA, GABAa, SERT, and mGluR5 in AD patients (P < 0.05). Symbols represent individual Fisher’s z-transformed spearman correlation coefficients between ¹⁸F-FDG PET SUVR–neurotransmitter correlations and each neuropsychological scale

Fig. 6.

Correlations of MMSE and MoCA with Aβ PET SUVR–neurotransmitter strength of association with 5HT2a, GABAa, and NAT in AD patients (a, b, P < 0.05), and SERT, FDOPA in Aβ positive MCI patients (c, P < 0.05). Symbols represent individual Fisher’s z-transformed spearman correlation coefficients between Aβ PET SUVR–neurotransmitter correlations and each neuropsychological scale

Discussion

This cross-disorder study systematically investigated ¹⁸F-FDG PET hypometabolic and Aβ deposition signatures across AD and FTD, as well as their spatial convergence with neurochemical architectures. We found that ¹⁸F-FDG PET changes generally colocalized with the distribution of specific receptors and transporters involved in serotonergic and dopaminergic neurotransmission in patients. In addition, we found colocalization of Aβ deposition with noradrenergic, dopaminergic and serotonergic receptors and transporters in AD and MCI patients. In AD patients, ¹⁸F-FDG and Aβ PET were significantly correlated with partial neurotransmitter association strength and clinical scale scores. Aβ PET-partial neurotransmitter association strength showed a slight correlation with MoCA scores only in Aβ positive MCI. By analyzing the spatial correlation between glucose metabolism and Aβ deposition and neurotransmitter receptor profiles, we identified specific neurotransmitter systems associated with regional hypometabolism and Aβ deposition that are unique to each condition. These findings showed alterations in all corresponding neurotransmitter systems from the imaging perspective [7, 15, 40–43].

As the brain’s obligate energetic substrate, glucose metabolism exhibits bidirectional regulatory coupling with neurotransmitter systems, collectively governing cognitive operations and neurohomeostatic regulation. To avoid partial volume effects due to atrophy, we performed a partial volume correction (PVC) using the Muller-Gartner approach implemented in SPM toolbox PET-PVE12 on the ¹⁸F-FDG PET images, and the overall patterns were approximately analogous to the voxel-wise results without correction, aligning with several prior research findings [44–46]. We found that the pattern of decreased glucose metabolism in the temporoparietal and frontal lobes of AD patients was significantly correlated with the brain distribution of 5-HT1a, 5-HT2a, GABAa, mGluR5, and D2. Most of our results of the spatial correlation analysis between PET and neurotransmitter maps showed negative correlations, indicating that regions with high physiological receptor/transporter density show more severe hypometabolism in patients, suggesting that these neurochemical centers have selective vulnerability. The serotonin system is known to influence glucose uptake and utilization in key brain regions, while the cholinergic system is directly involved in the regulation of glucose metabolism through acetylcholine receptors that modulate insulin secretion. Dopaminergic pathways in the prefrontal cortex support goal-directed behaviors through D1/D2 receptor signaling, regulating emotional responses, maintaining working memory, and coordinating movement. Impairments in this system are linked to cognitive decline and psychiatric symptoms in dementia [47, 48]. Whereas glutamatergic and gamma-aminobutyric acid ergic systems interact with glucose metabolism because both excitatory and inhibitory neurotransmission require substantial energy. Studies have shown that mGluR receptors of the glutamate system have emerged as key therapeutic targets in AD because of their protective roles in memory formation and deleterious roles in driving neurodegeneration through overactivation of nerve cells [49]. In contrast, GABAergic inhibition through GABAa receptors modulates cortical activity, affecting visuospatial processing, working memory networks, and chronic pain pathways—functional areas often impaired in neurodegenerative diseases [50].

Moreover, glucose metabolism in MCI patients after PVC remained significantly associated with serotonin, SERT, D2, FDOPA, NAT, mGluR5, and VaChT. The locus coeruleus-noradrenergic system modulates cognitive flexibility and memory consolidation via NA signaling and sustains synaptic homeostasis through presynaptic NAT. Currently, various small molecule drugs are employed clinically to elevate NA levels in the synaptic cleft by inhibiting NAT, enhancing signal transmission in noradrenergic neurons to treat AD [51]. Furthermore, the pathogenesis of AD involves dual cholinergic defects: presynaptic ACh depletion due to degeneration of the Meynert basal nucleus (Braak stages III-IV), and postsynaptic degeneration of muscarinic M1 receptors in the hippocampus -entorhinal circuit [52]. Compared with before PVC correction, PVC correction can remove the partial volume effect of brain atrophy and cerebrospinal fluid, and make the hypometabolic brain areas of MCI larger, resulting in more brain areas with SUVR difference compared with the spatial distribution of neurotransmitters, which may lead to a more significant correlation with neurotransmitters compared with without PVC. Additionally, we found more neurotransmitters associated with hypometabolism in MCI than in AD patients, which may suggest that the neurotransmitter system is impaired in early patients and therefore may be a potential early therapeutic target. Our findings are therefore consistent with previous studies, highlighting the importance of metabolic-neurotransmitter interactions in AD.

For FTD patients, previous studies have found that significantly reduced low-frequency fluctuations in frontotemporal and frontoparietal regions are significantly correlated with the distribution of serotonin and GABAa receptors and noradrenaline transporter (NAT) [53]. The changes of gray matter volume in prodromal FTD and symptomatic FTD patients have also been reported to be related to dopamine, acetylcholine and serotonin pathways [54]. Lagarde et al. [55] studied decreased basal ganglia volume and locus ceruleus signal intensity in Meynert, suggesting impaired noradrenergic and cholinergic systems in FTD patients. These results are similar to those of our study applying ¹⁸F-FDG PET, suggesting that serotonergic, dopaminergic, and noradrenergic systems may be effective or potential therapeutic targets for FTD. Notably, we also found that decreased glucose metabolism in frontotemporal cortex was significantly associated with higher MU-opioid and KappaOp receptor density in FTD patients. This is similar to the recent study by Morais et al. [15], who found an inverse correlation between regional brain volume and MU-opioid receptors in patients with bvFTD [56]. Our results suggest that the change of opioid receptor level in FTD may be one of the potential pathogeneses.

The accumulation of Aβ plaques in AD disrupts multiple neurotransmitter systems through direct synaptic toxicity and indirect glia-mediated mechanisms. When Aβ plaques are is present, the levels of various neurotransmitters are unbalanced along with impaired receptor localization and expression [57, 58]. We conducted the same correlation analysis between Aβ deposition and neurotransmitter distribution similar to FDG PET in AD and MCI patients, revealing significant correlations with 5HT2a, GABAa, and NAT distribution in AD, and exclusively with D1 distribution in MCI patients. Given that 66 of 111 MCI patients were Aβ-positive, we also analyzed patients with Aβ-positive MCI and showed significant correlations between Aβ deposition and distribution of D1, DAT, FDOPA, and SERT. Interestingly, different trends were found for the correlation between Aβ and neurotransmitters in AD and positive MCI. The positive correlation in AD stage may reflect the synergistic degradation of Aβ and transmitter systems in advanced pathology. For example, 5HT2a receptors are aberrantly expressed in glial cells surrounding Aβ plaques, or GABAa receptors are regionally lost due to inhibitory neuron death spatially coupled to Aβ deposition [59, 60]. This positive correlation suggests that the breakdown of the neurotransmitter system is directly related to Aβ toxicity. The negative correlation of MCI stage may reflect the early compensatory mechanism. For example, in patients with Aβ-positive MCI, D1 receptors are downregulated in regions with Aβ deposition to mitigate glutamate excitatory toxicity (by reducing dopaminergic input), whereas compensatory DAergic neurons project to unaffected brain regions, resulting in reduced D1 density in regions with high Aβ (negative correlation). Similarly, the negative correlation of SERT may reflect that the 5-HTergic system enhances synaptic signaling by reducing reuptake, resisting Aβ-induced synaptic dysfunction [61–63].

Previous studies in mice have shown that the DA system dynamically controls Aβ levels and promotes the degradation of Aβ in vivo through the action of brain Aβ degrading enzyme neprilysin, and long-term L-DOPA treatment can improve Aβ pathology and memory function by increasing NEP expression in mice [62]. Moreover, the recycling of NA by NAT preserves the equilibrium of NA concentrations in the nervous system, hence enabling the normal physiological activities of the nervous system. Currently, several small molecule medicines are utilized in clinical practice to elevate levels in the synaptic cleft by blocking norepinephrine transporters, hence boosting the signaling of noradrenergic neurons to treat depression and attention deficit hyperactivity disorder [64]. Studies on the noradrenergic system have shown that Aβ oligomers are heterogenic ligands of α2A adrenergic receptors (α2AAR) and regulate NA signaling. Blockade of α2AAR reduced tau phosphorylation and ameliorated pathological and cognitive abnormalities in AD mouse models [65]. In AD studies, plasma NA levels were found to be significantly positively correlated with cerebrospinal fluid Aβ₁₋₄₂ levels [66]. This evidence suggests the potential of dopaminergic and norepinephrine systems as therapeutic targets for AD, especially for Aβ clearance. Different from glucose metabolism disorders, Aβ, as an early pathological event, may indirectly destroy transmitter homeostasis through global mechanisms such as synaptic toxicity and neuroinflammation [67–69]. However, its local colocalization with specific transmitter systems suggests that transmitter pathways have selective vulnerability or compensatory regulation to Aβ. This spatial specificity may result from direct interaction of key transmitter receptors such as 5-HT2a with Aβ oligomers or, for example, preferential damage of GABAergic interneurons to Aβ toxicity [70, 71]. In Aβ-positive MCI patients, Aβ deposition was significantly correlated with the spatial distribution of SERT and dopamine system receptors and transporters, whereas in negative patients, it was only correlated with mGluR5, consistent with previous studies in Alzheimer’s disease model mice [62, 72]. Our findings also suggest that future longitudinal studies are needed to track the dynamic changes of transmitter-Aβ correlation in patients with MCI to AD, and to explore the disease-stage specific molecular interaction mechanisms.

Furthermore, we tested if the significant ¹⁸F-FDG and Aβ PET SUVR–neurotransmitter correlation coefficients are also associated with MMSE or MoCA. For FDG PET, we found that the intensity of co-localization with neurotransmitters, including 5-HT1a, 5HT2a, GABAa, and mGluR5, was correlated with MMSE and MoCA only in AD patients. For Aβ PET, co-localization with 5-HT2a, GABAa, and NAT distribution exhibited significant associations with MMSE and MoCA in AD patients. Only the co-localization with SERT and FDOPA distribution demonstrated significant associations with MoCA scores in patients with Aβ positive MCI. The positive correlation coefficients suggest that greater receptor-density-associated hypometabolism predicts more severe clinical impairment. These findings point to a potentially more general role for corresponding neurotransmission in the cognitive decline observed in AD. In patients with FTD, the co-localization strength of PET and neurotransmitters did not correlate with MMSE and MoCA scores, likely because these scales assess overall cognitive function rather than specific behavioral and semantic performance. More detailed and specific scale scores are needed to investigate the role of different neurotransmitters.

The current study has several limitations. First, the lack of systematic medication records (e.g. neuromodulators) precludes definitive exclusion of pharmacological effects on neurotransmitter systems and functional activity, potentially confounding the interpretation of disease-specific alterations. Second, the neurotransmitter density maps were derived from nuclear imaging data of healthy populations, which may not fully capture dynamic pathological changes in receptor or transporter distributions during disease progression, possibly biasing vulnerability estimates. Third, the restricted sample size of FTD patients constrains the generalizability of the results, and there is a lack of more detailed neuropsychological tests in different domains for more specific correlation analysis. These limitations highlight the need for longitudinal, multimodal studies with larger, clinically stratified cohorts to unravel the dynamic interplay between neurotransmitter systems, molecular pathology, and clinical manifestations in neurodegenerative diseases.

Our findings suggest that PET-based in vivo neurotransmitter system mapping reveals distinct spatial vulnerabilities in AD and FTD, characterized by neurotransmitter spatial distribution associated with glucose hypometabolism, Aβ deposition, and clinical symptom severity. These patterns not only highlight the interplay between neurotransmitter dysregulation and metabolic dysfunction but also underscore their collective contribution to cognitive decline. The observed associations with Aβ pathology further suggest a potential mechanistic link between proteinopathy propagation and neurotransmitter network disruption. By bridging molecular imaging and clinical phenotypes, this work advances our understanding of dementia pathophysiology and provides a framework for developing multimodal biomarkers or pharmacological interventions tailored to individual neurotransmitter vulnerabilities.

Author contributions

All authors contributed to the study conception and design. Study was designed by Shaozhen Yan and Jie Lu. Material preparation, data collection and analysis were performed by Sheng Bi, Zhigeng chen, Yixia Li, Bixiao Cui, Yi Shan, Hongwei Yang. The first draft of the manuscript was written by Sheng Bi, Zhigeng chen, Yixia Li, and Shaozhen Yan, and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Funding

This study has received funding by the National Natural Science Foundation of China (Grant No. 82102010, 82394434, 62333002).

Data availability

The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study was performed in line with the principles of the Declaration of Helsinki. Approval was granted by the Ethics Committee of Capital Medical University (No. [2023]044]).

Consent for publication

Written informed consent was obtained from the parents.

Competing interests

The authors have no relevant financial or non-financial interests to disclose.