Hypometabolic mismatch with atrophy and tau pathology in mixed Alzheimer’s and Lewy body disease

Abstract

Polypathology is a major driver of heterogeneity in the clinical presentation and extent of neurodegeneration (N) in patients with Alzheimer’s disease (AD). Beyond amyloid (A) and tau (T) pathologies, over half of patients with AD have concomitant pathology such as α-synuclein (S) in mixed AD with Lewy body disease (LBD). Patients with multiple aetiology dementia such as AD + LBD have faster progression and potentially differential responses to targeted treatments, although the diagnosis of AD + LBD can be challenging given the overlapping clinical and imaging features. Development and validation of improved in vivo biomarkers are required to study relationships between N and S and identify imaging patterns reflecting mixed AD + LBD pathologies.

We hypothesized that individual proteinopathies, such as T and S, are associated with commensurate levels of N. Thus, we assessed biomarkers of A, T, N and S with PET, MRI and CSF seeding amplification assay (SAA) data to determine molecular presentations of mixed A+S+ versus A+S– cognitively impaired patients from the Alzheimer’s Disease Neuroimaging Initiative (ADNI).

Strikingly, A+S+ patients had parieto-occipital ¹⁸F-fluorodeoxyglucose hypometabolism (a measure of N) disproportionate to the degree of regional atrophy or T burden, highlighting worse hypometabolism associated with S+ status on SAA. Following up on this hypometabolic mismatch with CSF metabolite and proteome analyses, we found that A+S+ patients exhibited lower CSF levels of dopamine metabolites and synaptic markers like neuronal pentraxin-2 (NPTX2), suggesting that altered neurotransmission and neuron integrity contribute to this dissociation between metabolic PET and MRI. Potential confounders exist when studying relations between N, AD and LBD pathologies, including neuroinflammation and other non-Alzheimer’s pathologies in mixed dementia, although our findings imply posterior hypometabolic mismatch is related more to S than vascular or TDP-43 pathology.

Overall, A+S+ patients had posterior mismatch with excessive ¹⁸F-fluorodeoxyglucose hypometabolism relative to atrophy or T load, possibly reflecting impaired neuronal integrity. Further research must disentangle the impact of multiple proteinopathies and clinicopathologic factors on hypometabolism and atrophy. Cumulatively, patients with mixed AD + LBD aetiologies harbour a unique metabolic PET mismatch signature.

By assessing imaging and CSF biomarkers of amyloid, tau, α-synuclein and neurodegeneration, Duong et al. demonstrate that patients with mixed Alzheimer’s and Lewy body disease show a signature pattern of posterior hypometabolism that is disproportionate to the degree of atrophy on MRI and tau pathology as measured by PET.

Introduction

Over half of patients with Alzheimer’s disease (AD) have concomitant non-Alzheimer’s pathologies at autopsy,^1-6 including α-synuclein (S) and TDP-43. Such patients with mixed pathologies are classified as multiple aetiology dementia. AD is often an amnestic disorder characterized by amyloid-β plaques and tau-based neurofibrillary tangles (NFTs) that can be detected in vivo by amyloid (A) and tau (T) biomarkers. NFTs are more highly associated with neuronal and synaptic loss or neurodegeneration (N).^7,8 Conversely, Lewy body disease (LBD) is a disorder with movement and neuropsychiatric symptoms linked to α-synuclein Lewy bodies.^9,10 LBD encompasses a spectrum of neuronal α-synucleinopathies, including dementia with Lewy bodies (DLB) and Parkinson’s disease (PD) based on recent biological frameworks.^11,12 Furthermore, patients with multiple aetiology dementia tend to have faster progression of cognitive presentation than patients with fewer pathologies.⁷ While some treatments, such as anti-cholinesterase medications, may overlap between AD and DLB,⁹ it is uncertain the degree to which copathology influences the efficacy of emerging anti-amyloid treatment or other therapies in clinical trials. Therefore, systematic characterization of neural biomarkers in multiple aetiology dementia is paramount to determining accurate diagnosis and treatment response.

Promising α-synuclein biomarkers are now available, including the seeding amplification assay (SAA) used to detect α-synuclein aggregates from CSF¹³ and peripheral skin and salivary glands.¹⁴ α-synuclein SAA has been validated in patients with PD and parkinsonian disorders,^13,15,16 cognitively impaired patients (including mixed AD + LBD)¹⁷ and cognitively unimpaired older adults.^16-18 SAA determines binary α-synuclein status, but does not currently provide information on regional distribution or relative load of α-synuclein pathology throughout the brain.

From the Alzheimer’s Disease Neuroimaging Initiative (ADNI) study, we assessed the neurodegeneration reflecting underlying α-synuclein SAA status to determine the systematic patterns of neurodegeneration attributable to latent α-synuclein pathology. Non-Alzheimer copathologies can be inferred by measuring cellular responses as seen by changes in neurodegeneration not adequately accounted for by AD pathology alone, particularly in multiple aetiology dementia. While tau burden is used to stage AD and is tightly linked to neurodegeneration, measured as structural atrophy (N_S) or lowered metabolism (N_M) in AD,^19-23 we hypothesized that coexisting non-Alzheimer pathologies may independently drive neurodegeneration and thus distort the relationship between neurodegeneration and tau burden, creating a mismatch.^10,24-27 Hence, comparing the observed atrophy on MRI or hypometabolism on ¹⁸F-fluorodeoxyglucose (¹⁸F-FDG) PET with the expected neurodegeneration based on the degree of AD pathology (tau PET) can highlight the impact of α-synuclein SAA status on neurodegeneration. Particularly, we aim to identify the regional hypometabolism profile attributable to SAA-based S+ status, where hypometabolism is disproportionate to the extent of tau pathology in A+S+ patients compared to A+S– patients. To this point, our group has shown in a series of studies that a mismatch of neurodegeneration relative to tau load (‘TN mismatch’) may predict the presence of non-Alzheimer’s pathologies, such as TDP-43-related markers.^24,26,27 Now, we extend this work for α-synuclein SAA in mixed AD + LBD.

In addition to TN_M mismatch, we explore the ‘N_SN_M mismatch’ between MRI and ¹⁸F-FDG PET that may occur in the setting of mixed AD + LBD. Cortical thickness and glucose metabolism on MRI and ¹⁸F-FDG PET typically decrease in tandem in AD²³ but not in LBD.^28,29 In fact, compared to AD, patients with LBD may not exhibit as many overall structural changes.^28-30 Posterior parieto-occipital hypometabolic measures, such as the cingulate island sign on ¹⁸F-FDG PET are prototypical molecular signatures associated with LBD^31-35 that are not matched by a corresponding degree of atrophy.^28-30,33 Indeed, deficits in glucose metabolism seemingly precede reductions in cortical thickness and neurotransmission changes as detected by MRI²⁸ and dopamine transporter molecular imaging,³⁶ respectively. This convergence of evidence pointing to mismatch between MRI and ¹⁸F-FDG PET led us to explore this discordance in A+S+ patients.

Here, we analysed data from 246 cognitively impaired older adults from ADNI with amyloid, tau and SAA α-synuclein status, of which 185 had both MRI and ¹⁸F-FDG PET measures. We compared regional levels of TN_M and N_SN_M mismatches to study the hypometabolic patterns linked with α-synuclein that are not explained by tau or structural atrophy, respectively. Initially, we found that α-synuclein markers were associated with patients with amyloid pathology. Based on the literature, we hypothesized that A+S+ patients have posterior hypometabolism disproportionate to the degree of tau or atrophy. Finally, we predicted CSF metabolomic and proteomic studies would point to potential mechanisms connecting α-synuclein to altered neurotransmission and ‘neuronal integrity,’ a term encompassing neuronal dysfunction, as well as death, and therefore representing visible and occult changes on MRI. These hypotheses could account for how hypometabolism on ¹⁸F-FDG PET may potentially precede atrophy on MRI and support our goal to identify a unique hypometabolic signature in AD + LBD.

Materials and methods

Patient cohort for biomarkers

From the ADNI cohort database (http://adni.loni.usc.edu), we included 246 cognitively impaired patients with a measure of amyloid (A), tau (T) and α-synuclein (S) status (last accessed 12/2023). We found 180 participants with a diagnosis of mild cognitive impairment (MCI), and 66 patients with dementia (Table 1). Evaluation of A status utilized amyloid PET (¹⁸F-florbetapir or ¹⁸F-florbetaben by ADNI AMYLOID_STATUS, n = 242, UCBERKELEY_AMY_6MM, accessed 1/2024) or Elecsys CSF assay³⁷ (CSF amyloid-β₄₂ < 980 pg/ml, n = 4, UPENNBIOMK_MASTER_FINAL_21Dec2023, accessed 1/2024). T status was determined by ¹⁸F-flortaucipir PET images, obtained within 1 year from A testing. S status was obtained via Amprion α-synuclein SAA^13,38 obtained within 2 years of imaging (AMPRION_ASYN_SAA, accessed 1/2024). This range was extended to include additional patients with S– SAA results >2 years after imaging or S+ SAA results >2 years before imaging, since CSF SAA rarely converts from S+ to S– status.^14,16,39 From the whole cohort, 185 patients were found to have MRI and ¹⁸F-FDG PET within 1 year of tau PET. Median time between tau versus amyloid PET was 14 days (84% of cases within 3 months). Median time between tau PET versus α-synuclein CSF SAA was 34 days (63% within 3 months). Median time between tau versus ¹⁸F-FDG PET was 13 days (91% within 3 months). For cognitive data, we selected ADNI clinical testing sessions closest to imaging for the AD Assessment Scale-Cognition 13 item (ADAS, where higher score is worse, accessed 1/2024) and Neuropsychiatric Inventory (NPI, accessed 1/2024) for patients who presently had or developed hallucinations (NPI-B).

Table 1.

Clinical characteristics of the Alzheimer's disease neuroimaging initiative cohort with amyloid, tau and α-synuclein markers

AS Group	Disease (AD/LBD)	MCI/Dementia	Sex (F/M)	Age (years)	Education (years)	ADAS-Cog	Hallucinations (%)
A–S– (n = 81)	Neither AD nor LBD	74/7	32/49	71.3 (8.4)	16.3 (2.6)	14.0 (6.3)	0%
A–S+ (n = 19)	LBD without AD	16/3	4/15	78.3 (6.4)	16.3 (3.0)	16.2 (8.5)	5.3%
A+S– (n = 99)	AD without LBD	70/29	44/55	75.0 (7.9)	15.8 (2.6)	20.7 (9.7)	7.1%
A+S+ (n = 47)	Mixed AD + LBD	20/27	24/23	75.0 (7.6)	15.8 (2.4)	27.4 (11.8)**	10.6%

Frequencies are shown for disease severity [mild cognitive impairment (MCI)/dementia] and sex [female/male (F/M)]. Mean (and standard deviations) are shown for age, education and Alzheimer’s Disease Assessment Scale-Cognitive (ADAS-Cog). Percentages are shown for the proportion of patients with/developing hallucinations over time. Significant differences were calculated using likelihood ratio tests with the amyloid+ α-synuclein− (A+S–) group as the reference, covariates of age, sex and tau (T) burden in the inferior temporal gyrus. *P < 0.05 and **P < 0.005. LBD = Lewy body disease.

Imaging data

Post-processed PET images from the ADNI data archive (http://adni.loni.usc.edu/data-samples/access-data/) were obtained. Tau PET imaging was originally performed using the ADNI protocol with 30-min brain scans (six 5-min frames) starting 75 min after intravenous administration of ∼10.0 mCi ¹⁸F-flortaucipir. ¹⁸F-FDG PET imaging consisted of a 30-min scan (six 5-min frames) at 30 min after 5.0 mCi ¹⁸F-FDG injection. For amyloid PET, a 20-min brain scan (four 5-min frames) was performed 50 min after ∼10.0 mCi ¹⁸F-florbetapir or 90 min following ∼8.1 mCi ¹⁸F-florbetaben injection. Processed tau and ¹⁸F-FDG PET images with uniform isotropic resolution (8 mm full-width-at-half-maximum) were obtained with the ADNI archive description ‘Coreg, Avg, Std Img and Vox Size, Uniform Resolution.’ ADNI MRI included a T1-weighted structural scan (resolution 1.0 × 1.0 × 1.2 mm³) and fluid-attenuated inversion recovery (FLAIR) sequence scan acquired in the same session.

Image processing and PET regional analysis

MRI studies were processed using the ANTs pipeline⁴⁰ for inhomogeneity correction, brain extraction, template registration and cortical thickness measurement.^41,42 MRI scans were divided into cortical, subcortical, brainstem and cerebellar regions of interest (ROIs) with multi-atlas segmentation^43,44 (http://neuromorphometrics.com/ParcellationProtocol_2010-04-05.PDF). PET images were co-registered to T1-weighted MRI with ANTs for rigid-body transformation.⁴⁰ Standardized uptake value ratio (SUVR) maps were generated with reference regions specific for each tracer: inferior cerebellar cortex for ¹⁸F-flortaucipir⁴⁵ and cerebellar cortex for ¹⁸F-FDG.⁴⁶ Mean regional T, N_M and N_S measures were extracted from tau SUVR, ¹⁸F-FDG SUVR and cortical thickness maps, respectively. The cingulate island sign represents metabolic sparing of posterior cingulate cortex relative to precuneus and cuneus and has been associated with S+ status. It was quantified as the ratio of posterior cingulate/precuneus/cuneus ¹⁸F-FDG SUVR; higher cingulate island ratio is linked to LBD.^32,33 Likewise, the I/MTL/FSO ratio represents sparing of inferior temporal gyrus relative to medial temporal lobe and frontal supraorbital gyrus in ¹⁸F-FDG SUVR; higher I/MTL/FSO ratio is associated with TDP-43 disease.⁴⁷

Regional TN_M and N_SN_M mismatch

Spatial patterns of TN_M mismatch were investigated using a previously published residual method.^25,26 Briefly, robust linear regressions were performed comparing individual ¹⁸F-FDG SUVR versus a log transform of tau SUVR or cortical thickness (in mm) across all cohort patients in each of the grey matter ROIs. A bi-square weighting function minimized the influence of outliers in robust regression. Regional mean residuals were calculated from regressions and visualized in two-dimensional maps and three-dimensional renderings by ITK-SNAP⁴⁸ and MRIcroGL.⁴⁹ For general 3D rendering maps, the white colour reflected a mismatch residual value close to about 0.6 standard deviations (where about half of the patients reside) from the regression line, and blue colour denoted negative mismatch residual with N_M worse than expected. For significant difference renderings, white colour represented no significant difference between A+S+ and A+S– groups, and blue colour depicted a negative mismatch residual that significantly differs between groups.

CSF metabolite and proteomics analysis

Processed CSF proteomics and metabolomics results^50,51 for each protein and metabolite were available from ADNI in 544 participants with A and CSF SAA-based S markers within 2 years of each other (58% within 3 months) (CruchagaLab_CSF_SOMAscan7k_Protein_matrix_postQC, CruchagaLab_CSF_metabolomic_matrix_20230620, accessed 1/2024). This cohort included A–S– (n = 161), A–S+ (n = 35), A+S– (n = 246) and A+S+ (n = 102). Metabolites were assessed by targeted hypothesis-driven analysis of dopamine metabolites (homovanillic acid and 3-methoxytyramine sulfate) per literature.^52,53 We next investigated proteome differences between patients with mixed AD + LBD versus AD without LBD or LBD without AD. Groups were compared by ANOVA to establish group-wise differences (P < 0.05) followed by two-sample t-tests to assess hypothesis-driven pairwise differences between A+S+ versus A+S– and A+S+ versus A–S+ groups, with significant hits identified by a stringent false discovery rate (FDR) <0.005 threshold after Benjamini-Hochberg adjustment for multiple comparisons. Based on comparisons, there were 276 protein hits for A+S+ versus A+S– and 210 proteins for A+S+ versus A–S+ analyses, which were then input into the ontology functional database of the WebGestalt program (https://www.webgestalt.org/). Over-representation analysis was conducted for biological process and cellular compartment in the Homo sapiens protein-coding genome with FDR <0.05 after Benjamini-Hochberg adjustment and redundancy reduction by weighted set cover. Data were visualized by graphs of fold enrichment, protein count and –log₁₀(FDR) per ontology class.

Statistical analysis

Statistical analysis was performed in R (v4.1.2). All statistical tests were two-sided. Comparisons for variables such as tau and ¹⁸F-FDG SUVRs were performed with likelihood ratio tests by linear regression. Covariates included sex, age and tau SUVR in the inferior temporal gyrus, a region where T correlates with disease severity.^8,54 Multiple test adjustment by Benjamini-Hochberg correction with FDR <0.05 was conducted for pairwise comparisons of proteomics data. Relative frequency comparisons were performed with χ² tests.

Results

Amyloid, tau and α-synuclein markers can be compared in vivo

First, we evaluated in vivo SAA-based S status and its associations with A and T pathology from an ADNI cohort of 246 cognitively impaired participants with A, T and S status. A+S+ patients had significantly worse impairment compared to A+S– patients on the ADAS-Cog after adjusting for covariates of age, sex and T burden (P = 0.003), in keeping with more severe cognitive presentation in mixed AD + LBD (Table 1). A+S+ patients trended towards a higher frequency of hallucinations.

Consistent with autopsy cohorts,³ cognitively impaired patients with SAA-based S+ status were more likely to have AD pathology on amyloid PET, although this could further be affected by the recruitment criteria for ADNI (see the ‘Discussion’ section). In our ADNI cohort, 32% of A+ patients were S+, compared to just 19% of A– patients (P = 0.022).

Combining Alzheimer’s and Lewy body disease markers highlights a metabolic mismatch signature

From our ADNI cohort of cognitively impaired participants with known A, T and S status, we assessed the dissociation between N_M versus T (TN_M mismatch) and N_M versus N_S (N_SN_M mismatch) in the 185 patients with available MRI and ¹⁸F-FDG PET. TN_M mismatch is based on the hypothesis that the degree of molecular pathologies is linked to a commensurate level of cellular injury (Fig. 1A). Evaluating correlations across the entire sample together, greater T is associated with worse N_M (Fig. 1B), while structural and metabolic measures of N are positively correlated (Fig. 1C and D). However, the degree of variability in these TN_M and N_SN_M relationships, as measured by regression model residuals, can reflect the impact of non-Alzheimer pathology on N that is not accounted for by AD pathology alone. Here, S+ status is associated with TN_M mismatch where worse N_M is disproportionate to the extent of T burden, as illustrated in two example cases (Fig. 1E and F) and quantitative group analysis (Fig. 2). Moreover, different pathologies (T, S, TDP-43) may potentially impact individual neurodegenerative measures differently. For instance, imaging evidence implies that S+ status is linked more to hypometabolism on ¹⁸F-FDG PET than atrophy on MRI,^28,29 therefore leading to an N_SN_M mismatch that we study further.

Figure 1.

Imaging examples highlight mismatch residuals between tau and neuronal ¹⁸F-FDG PET hypometabolism, and MRI/¹⁸F-FDG PET mismatch relationships.

(A) Mismatch between tau pathology (T) and neuronal ¹⁸F-fluorodeoxyglucose (¹⁸F-FDG) hypometabolism (N_M) occurs in mixed Alzheimer’s and Lewy body disease (AD + LBD) since N_M may reflect a balance of both T and α-synuclein (S) pathologies. (B) Associations between worse ¹⁸F-FDG hypometabolism and greater T burden show outliers with worse N_M than expected based on T. (C) Mismatch between atrophy on MRI (N_S) and N_M occurs in mixed AD + LBD since S appears to impact N_M on PET moreso than N_S on MRI. (D) Associations between worse ¹⁸F-FDG hypometabolism and lower cortical thickness (mm) in the inferior temporal gyrus highlight outliers with worse N_M than expected based on N_S. Amyloid (A), T, S status, ¹⁸F-FDG PET and MRI are shown for two cognitively impaired older adults, whose residuals are highlighted in B and C. (D) A 70-year-old male with A+T+S+ markers has AD Assessment Scale-Cognitive (ADAS-Cog) score of 29 (higher is worse) and cingulate island sign, while a (F) 75-year-old female with A+T+S– markers has an ADAS-Cog score of 24 and higher metabolism, despite both having similar A and T burdens. Regression lines of best fit and standard deviation based-lines are shown.

Figure 2.

Imaging reveals a metabolic mismatch signature in patients with concomitant amyloid+synuclein pathology. Mean regional residual z-score maps show posterior mismatch (arrows) in amyloid+ α-synuclein+ (A+S+) patients compared with A+S– patients in [A(i)] mismatch residuals between tau and ¹⁸F-fluorodeoxyglucose (¹⁸F-FDG) PET (TN_M) and [B(i)] mismatch residuals between MRI and ¹⁸F-FDG PET (N_SN_M). Here, the white colour represents a mismatch residual value close to about 0.6 standard deviations from the regression line, and the blue colour denotes a negative mismatch residual with an N_M worse than expected. Mismatch maps of average residual z-scores in regions that significantly differ between A+S+ and A+S– groups are seen for [A(ii)] TN_M and [B(ii)] N_SN_M mismatch, after adjusting for age, sex and tau load. Here, white colour represents no significant difference between A+S+ and A+S– groups, and blue colours depict a negative mismatch residual that significantly differs between groups, where residual values (and hence the colour) in A(i) and B(i) are similar to those shown in A(ii) and B(ii), respectively.

Regional TN_M and N_SN_M mismatch residual maps in A+S+ patients reveal a posterior hypometabolism signature, where hypometabolism is greater than expected given the level of tau tracer uptake [Fig. 2A(i)] or atrophy [Fig. 2B(i)]. Conversely, regions outside the parieto-occipital cortex generally did not have significant deviations between observed and expected N_M. Maps visualizing only the regions where residuals significantly differ between A+S+ and A+S– groups (adjusting for age, sex and tau burden) also accentuate posterior metabolic mismatch for TN_M [Fig. 2A(ii)] and N_SN_M [Fig. 2B(ii)]. Significant differences in mismatch residuals are localized primarily in medial retrosplenial and posterolateral parieto-occipital areas, such as precuneus and occipital pole. Hence, compared to A+S– status, comorbid A+S+ pathology is associated with a posterior hypometabolic phenotype that is disproportionately worse than expected given the degree of either T or N_S.

Occipital hypometabolic mismatch does not correlate with vascular or TDP-43 markers

We investigated the role of non-Alzheimer’s, non-LBD aetiologies that may contribute to TN_M and N_SN_M mismatch. Posterior mismatch does not correlate significantly with in vivo white matter hyperintensity volumes on FLAIR MRI or a hypometabolic I/MTL/FSO ratio on ¹⁸F-FDG PET suggestive of TDP-43 disease⁴⁷ in A+S+ or A+S– patients (Supplementary Fig. 1). While additional autopsy validation in a large cohort is needed, this may support our hypothesis that posterior hypometabolic mismatch is related to α-synuclein more than cerebrovascular or TDP-43 pathology.

CSF markers reflect altered neurotransmission and neuronal integrity with S+ status

To further evaluate why there may be a more prominent phenotype on ¹⁸F-FDG PET than MRI in S+ pathology, we next performed exploratory analyses of CSF metabolite and proteomics data^50,51 from 544 ADNI participants with A and S biomarkers. The A+S+ group shows lower metabolism of dopamine, a neurotransmitter reduced in LBD due to hallmark dopaminergic neuron loss. Significant depletion was observed in A+S+ relative to A+S– groups for dopamine catabolites (Fig. 3A) such as homovanillic acid (P < 0.001) and 3-methoxytyramine sulfate (P = 0.004) after correcting for age and sex covariates (Fig. 3B and C). These findings are concordant with the literature on disrupted dopamine metabolism in LBD.^52,53

Figure 3.

CSF metabolite and proteome analysis reveal altered dopamine neurotransmission and neuron integrity in patients with concomitant amyloid+synuclein pathology. (A) Dopamine is metabolized by enzymes, including catechol-O-methyltransferase (COMT), monoamine oxidase (MAO), aldehyde dehydrogenase (ALDH) and sulfotransferase (SULT). The CSF metabolite study demonstrated decreased dopamine metabolites (B) homovanillic acid and (C) 3-methoxytyramine sulfate (normalized concentration) with amyloid+ α-synuclein+ (A+S+) status. (D) The synaptic protein neuronal pentraxin-2 (NPTX2) is significantly decreased in A+S+. RFU = relative fluorescence unit. Box plots: filled circles = data points; × = mean; mid-line = median; lower and upper edges of each filled box = first (Q1) and third (Q3) quartiles (interquartile range, IQR); whiskers = minimum/maximum points and outliers based on thresholds <Q1 – 1.5(IQR) or >Q3 + 1.5(IQR). (E) Ontology of CSF proteomics comparing A+S+ versus A+S– indicates that S+ status is associated with neuron signalling, synapse organization and vesicles. (F) Comparing A+S+ versus A–S+ highlights that A+ status is associated with immune responses, protein degradation and lipoparticles. False discovery rate correction, Benjamini–Hochberg adjustment and weighted set cover were performed. Comparisons were made using likelihood ratio tests with covariates of age and sex, where *P < 0.05 and **P < 0.005.

CSF proteome changes related to S+ status may involve altered neuronal integrity. Prior biomarker studies revealed that CSF levels of synaptic neuronal pentraxin-2 (NPTX2) are reduced in neurodegenerative diseases such as AD⁵⁵ and PD, where NPTX2 correlates with dopaminergic pre-synaptic integrity characterized by dopamine transporter scan.⁵⁶ Here, we find that A+S+ patients have significantly attenuated NPTX2 CSF concentrations than patients with A+S– (P = 0.002) and A–S+ (P = 0.005) status after adjusting for age and sex covariates (Fig. 3D), suggesting that AD + LBD polypathology may lead to worse neuronal integrity than in AD or LBD alone.

From a proteomic screen of over 7000 proteins, significant differences in protein expression were enriched in A+S+ for certain pathway ontologies. To better isolate the effect of α-synuclein in mixed AD + LBD pathology versus AD alone, we identified the protein ontologies by over-representation analysis reflecting differential CSF protein levels in A+S+ versus A+S– groups after multiple comparison adjustment (Fig. 3E). Compared to A+S–, the A+S+ group had CSF protein changes related to biological processes of synapse organization, cell-cell signaling and stimulus response. Dysregulated CSF proteins generally localized to the apparatus of neurotransmission and transport (receptor complex, cell surface/junction) and endolysosomal pathway (secretory/cytoplasmic vesicles, endoplasmic reticulum and Golgi complex).

We next evaluated the role of amyloid-β in the A+S+ group by contrasting proteome expression in mixed AD + LBD versus LBD alone (Fig. 3F). Compared to A–S+, A+S+ status had CSF protein changes related to biological processes of proteolysis, inflammation and immune mechanisms (wound and defence responses). Proteins with aberrant CSF expression were enriched for biochemical compartments including serum lipoparticles, myelin sheath and extracellular matrix. Together, these exploratory CSF studies suggest that the hypometabolic mismatch phenotype on imaging in A+S+ status may reflect selective injury of dopamine pathways and impaired neuron integrity seen with α-synuclein rather than the neuroimmune changes that may perhaps be more widespread with amyloid-β.

Discussion

Polypathology is increasingly recognized as a cause of clinical and biological variability in AD and multiple aetiology dementia. The confluence of amyloid, tau and α-synuclein pathologies in mixed AD + LBD can lead to accelerated neurodegeneration and worse outcomes,^2,3,7 raising questions on whether targeted therapies have similar efficacy in ‘pure’ versus mixed disease. While various SAA-based tests are available^13,38 and specific PET markers of α-synuclein are being developed,^57,58 much can be learned from the cellular responses to α-synuclein pathology by comparing MRI, ¹⁸F-FDG PET and measures of AD and LBD pathology. From a cohort of cognitively impaired ADNI participants with markers of amyloid, tau, α-synuclein, cortical thickness and glucose metabolism, we found that S+ status was associated with A+ status, similar to autopsy studies.³ Next, we uncovered that A+S+ patients have a posterior parieto-occipital pattern of reduced glucose metabolism that is disproportionate to both the degree of AD pathology (NFTs) and the degree of atrophy in these patients. This mismatch could be used as a ‘virtual biomarker’ for AD + LBD that does not correlate with cerebrovascular or TDP-43 biomarkers. In probing why this dissociation between ¹⁸F-FDG PET and MRI exists in AD + LBD, exploratory CSF analysis suggested this hypometabolic mismatch may be related to aberrant neuronal integrity with disrupted metabolism and death of critical neurons in dopamine-mediated pathways^51,52 (altering levels of dopamine catabolites and synaptic marker NPTX2), which may be better detected by ¹⁸F-FDG PET than by MRI.

Posterior hypometabolic mismatch is consistent with but not entirely explained by the cingulate island sign on ¹⁸F-FDG PET, the prototypical hypometabolic signature in LBD.^31-35 Cingulate island sign emphasizes the lack of hypometabolism in the posterior cingulate relative to the extensive hypometabolism in the cuneus. With the cingulate island of LBD, neither the spared posterior cingulate cortex nor the hypometabolic cuneus have corresponding atrophy on MRI.^28-30,33 Conversely, our TN_M and N_SN_M mismatch measures highlight greater parieto-occipital hypometabolism relative to local tau load and atrophy. These additional measures of tau and atrophy are not explicitly incorporated into the cingulate island sign.

Our findings are concordant with the literature exemplifying hypometabolism over atrophy as a feature of LBD, possibly reflecting a pathobiological basis of α-synuclein. Patients with LBD, even those with severe neocortical α-synuclein load, have fairly preserved cortical volume on ante-mortem and post-mortem assessment.^28,29,33,59 While LBD may be associated with some temporal atrophy,^29,30 this is much less than the degree of parieto-occipital hypometabolism.^28,31 In fact, occipital cortex thickness has even been reported to increase in patients with AD and hallucinations (who may be enriched for mixed AD + LBD).⁶⁰ Glucose hypometabolism on ¹⁸F-FDG PET also precedes neurotransmission changes seen on dopamine transporter molecular imaging.³⁶ From a network perspective, LBD displays broad metabolic abnormalities in the brain, which could reflect damage at key pathway nodes in dopaminergic neuronal ensembles that are susceptible to α-synuclein pathology,^52,53,59-64 and/or regional alterations in neuronal integrity that could appear on ¹⁸F-FDG PET despite the lack of change on structural MRI.^28-30 This selective neurodegeneration of vulnerable dopamine pathway neurons in LBD^9,56 may be associated with wider metabolic network changes^36,60-64 and disproportionate posterior ¹⁸F-FDG hypometabolism relative to atrophy when compared to a more generalized neurodegeneration and glial activation in AD.^7,65 At the cellular/molecular level, neuronal activity modulates the phosphorylation and aggregation kinetics of α-synuclein,⁶⁶ further supporting that pathophysiological manifestations of α-synuclein may be better detected by changes in neuronal integrity more associated with ¹⁸F-FDG PET than by cortical thickness on MRI. Intriguingly, NPTX2 colocalizes with α-synuclein aggregates in Lewy bodies and neurites and is connected to dopamine neuron dysfunction^56,67 and AMPA-mediated excitotoxicity.^67,68 Misregulated NPTX2 expression has been reported among patients with various neurodegenerative proteinopathies^55,56,68 and we found NPTX2 CSF levels were lower with A+S+ status than A+S– or A–S+ status, perhaps reflecting a greater impact on neuron integrity in mixed AD + LBD. Overall, we posit that converging findings implicate α-synuclein pathology in disrupted neuronal integrity, especially in the dopaminergic pathway, that is associated more with hypometabolism on ¹⁸F-FDG PET than changes seen on MRI.

In the absence of α-synuclein pathology, A+S– patients demonstrated a relative lack of dissociation between MRI and ¹⁸F-FDG PET. This consilience between neurodegenerative measures may stem from differences in metabolic activity in neurons versus glia in AD versus LBD. CSF proteome analyses comparing to A+S+ patients suggested that the presence of amyloid pathology favor an inflammatory or innate immune response phenotype, more so than the selective dopaminergic injury and neuronal integrity changes observed with the presence of α-synuclein. Considering AD pathology, numerous studies support the hypothesis that amyloid-β incites a strong neuroinflammatory response by astrocytes, microglia and immune cells⁶⁴ while tau hyperphosphorylation and NFTs are more closely associated with neuronal loss.^7,19 Because ¹⁸F-FDG PET detects both neuronal ‘hypo’metabolism in neurodegeneration and glial ‘hyper’metabolism in neuroinflammation, the extent of net hypometabolism visualized in AD is likely dampened in magnitude. This is supported by dual tracer studies with ¹⁸F-FDG PET and the microglia-related Translocator protein (TSPO) PET,⁶⁹ demonstrating that microglial inflammation may be present in AD and potentially contribute to measurable ¹⁸F-FDG PET uptake. Thus, neuronal hypometabolism on ¹⁸F-FDG PET could be masked by microglial activation in some regions. Differing degrees of neuroinflammation in AD versus LBD may also contribute to this mismatch between MRI and ¹⁸F-FDG PET. While α-synuclein pathology provokes a neuroimmune reaction,⁷⁰ including adaptive immune CD4+ T cells,^71,72 LBD may recruit a less robust or more confined neuroinflammatory response of glia and innate immunity than in AD,^73-75 especially at early stages of selective dopaminergic neuron injury. Indeed, microglial and astrocyte activation at autopsy is more closely associated with AD rather than LBD pathologies, even in patients with mixed AD + LBD.⁷³ For these reasons, ¹⁸F-FDG PET may be better suited to identify neuronal responses to LBD and mixed AD + LBD.

Limitations and future directions

Our study has several limitations and next steps. Examining the relationships between A, T, N and S is based on hypotheses of associations between proteinopathies and cellular responses (atrophy and/or hypometabolism), and therefore not causal in nature. Extended in vitro, in vivo and ex vivo investigations are required to pinpoint precise mechanisms of neuronal and glial injury from multiple pathologies. The clinical cohort available was based on patients with primarily amnestic phenotype in ADNI, such that the sample is more biased towards predominantly an AD-like presentation and underlying AD neuropathology, which would most likely be enriched for A+T+ status. Concomitant non-Alzheimer pathologies may be present but underrepresented in ADNI relative to the general population. ADNI largely excludes patients with substantial neurological diseases besides AD, systemic illnesses, substance use and primary psychiatric disorders and is thus not representative of a typical clinical population with comorbidities that could influence ¹⁸F-FDG PET and MRI. Additional cohorts such as patients with predominantly LBD-like presentation or community-based samples with more patients with typical presentations of multiple aetiology dementia should be examined to corroborate our signatures of hypometabolic mismatch. There are limitations to the SAA measure. Several SAA tests exhibit different sensitivities for α-synuclein pathology based on different stages, such as the lower detection rates in patients with lower α-synuclein load or olfactory involvement.^16,76 Hence, it may be more difficult to assess the role of hypometabolic mismatch in patients with low α-synuclein levels wherein the SAA test is insensitive. Future work may relate the relative degree of α-synuclein load based on quantitative SAA biofluid measures. Development of α-synuclein PET tracers,^57,58 specifically those sensitive to LBD, would likely allow for greater spatiotemporal resolution of α-synuclein spread, its relationship with neurodegeneration and the mismatch metrics examined here. Time differences between imaging and CSF SAA studies could contribute to more variability, although this was addressed by choosing scans and CSF collection times that were close together (median time was 34 days, with 63% within 3 months). Based on the goal to maximize power and increase sample size, the time interval of 2 years was selected within reason from the literature, similar to the time between diagnostic imaging and CSF collection at enrollment seen in some SAA studies^16,39 such as the Parkinson’s Progression Marker Initiative¹⁶ (though some studies of A and S markers in the ADNI cohort have applied a 6-month timeframe³⁸). The conversion rate from negative to positive result is slower for α-synuclein CSF SAA than for amyloid markers in cognitively unimpaired and impaired older adults,^18,38 which supports our selected intervals. Nevertheless, more studies may highlight ideal intervals between biomarker studies for cross-comparison. Additional temporal considerations include the limitation of cross-sectional studies to detect atrophy and hypometabolism due to individual variability, which could potentially be over- or under-called based on one set of scans. Future analyses of longitudinal change on MRI and PET can improve the ability to evaluate neurodegeneration robustly. Certain technical factors may saturate radiotracer retention, creating a ceiling effect for tau burden. Moreover, since each participant may be at a different disease stage, there may be patients with earlier-stage pathological burden where TN_M and N_SN_M mismatch may not have yet manifested despite the presence of early α-synuclein inclusions. Therefore, analysis with alternate measures of neurodegeneration and α-synuclein SAA, pathological classification and autopsy validation can strengthen results. Further inquiries into non-Alzheimer’s, non-LBD pathologies such as TDP-43 and cerebrovascular disease can be ascertained with exploratory biomarkers and neuropathology. Mismatch based on regressions performed within different subgroups of patients based on markers of amyloid-β, tau, α-synuclein and other pathologies can also be evaluated further, although we did not find major between-group differences in regressions across regions. The relationship and mismatch between tau PET, ¹⁸F-FDG PET and thickness on MRI may vary across disease stage or at higher levels of cognitive impairment. Non-Alzheimer’s copathologies may also increase at later stages of AD and LBD. Future studies with additional markers and longitudinal data can clarify such relations. Exploratory CSF proteome ontologies primarily found extracellular and secreted protein hits, so novel studies may aim to elucidate intracellular expression levels of protein/RNA from specific cell types. Proteome findings related to neuronal integrity could represent changes in both neuroglial structure and function, so additional investigations are required to obtain a more granular view linking microscopic and macroscopic features of α-synucleinopathy. This work may help contextualize emerging biological frameworks for AD^7,77 and LBD.^11,12

Overall, we found evidence supporting the hypothesis that ¹⁸F-FDG metabolism is lower than expected based on the degree of tau pathology or cortical thinning in the setting of mixed AD + LBD aetiologies. This posterior hypometabolism was consonant with CSF proteomics and metabolite studies highlighting altered dopamine transmission and neuronal integrity related to α-synuclein pathology and cognition but not vascular disease or TDP-43 pathology. While further validation is required, hypometabolic mismatch may represent a unique feature in the pathobiology of mixed AD + LBD that may assist in precise diagnosis and treatment.

Data availability

Raw and processed data, including scans and spreadsheets, are available on the Alzheimer’s Disease Neuroimaging Initiative (ADNI) data archive (http://adni.loni.usc.edu). Code for image processing is available at http://stnava.github.io/ANTs/ and code for analyses can be provided upon reasonable request.

Funding

Funding was provided by the National Institute on Aging (NIA) via a Ruth L. Kirschstein National Research Service Award (NIA F30 AG074524), research project grants (NIA R01 AG072796 and R01 AG069474) and a University of Pennsylvania Alzheimer’s Disease Core Center grant (NIA P30 AG072979).

Competing interests

L.X. received personal consulting fees from Galileo CDS, Inc. L.X. has become an employee of Siemens Healthineers since May 2022 but the current study was conducted during his employment at the University of Pennsylvania. D.A.W. reports grants from Merck, Biogen, Eli Lilly/Avid and additional fees from GE Healthcare, Functional Neuromodulation and Neuronix, all outside this work. I.M.N. reports fees from Biogen outside this work. The remaining authors report no competing interests.

Supplementary material

Supplementary material is available at Brain online.