Cell Type Clustering and Annotation

The clustering results for the pre-processed scRNA-seq datasets for AD and PD and the annotations for the corresponding cell type clusters are shown in Fig. 2. For the PD dataset, we confirmed the obtained annotations by visualizing cell type-specific marker genes derived from the CellMarker database and manually inspecting the literature on the marker genes for all cell type clusters visible in the UMAP representation (see Supplementary Fig. 1). These markers exhibit significant overexpression in the respective clusters, as confirmed by assessing differential gene expression between the clusters. For the AD dataset, we validated the annotations from the original source publication [10] by performing a differential expression analysis comparing each cell type cluster against all others and comparing the resulting cell cluster-specific DEGs against known marker genes in the CellMarker database.

Fig. 2

Two-dimensional cluster visualization of the scRNA-seq datasets for (a) AD, showing the cell type assignments from the original SEA-AD study (validated against the CellMarker database), and for (b) PD, generated using the UMAP dimension reduction approach and highlighting the added cell type annotations (see Methods)

In the PD dataset, 10 clusters were identified as the optimal number of clusters using the Silhouette width evaluation. High-confidence cell type annotations were obtained for each of these clusters when applying the ScType algorithmic annotation approach using data from the CellMarker database. Overall, 8 main distinct cell types were detected for both the AD and PD datasets (see Fig. 2), including cell type annotations mapping to multiple point clusters, indicating the potential to investigate finer subdivisions of cell populations in future follow-up studies. To focus on the most robust data patterns with statistical support from large cell counts, we merged cell types for smaller clusters with overlapping differential gene markers and similar annotations for both the AD and PD datasets. Specifically, the clusters labeled as different subcategories of excitatory or inhibitory neurons were merged under the generic terms "Excitatory neuron" and "Inhibitory neuron". Overall, we identified several relevant cell types in the data from both cohorts that have been reported previously to display molecular alterations in neurodegenerative diseases, including neurons, astrocytes, oligodendrocytes, and microglial cells [25], among others. Therefore, the statistical analysis and cellular pathway and network analysis described below focus on these common cell types with confirmed disease relevance and enough cells for robust analysis.

Gene-Level Comparative Analysis

The statistical analysis of gene-level differential expression between AD vs. HC and PD vs. HC revealed numerous significantly differentially expressed genes (DEGs) for both diseases. The DEGs are categorized into three groups as defined in the Methods section: Shared DEGs (highlighted in red for increased expression and blue for decreased expression), disease-specific DEGs (shown in yellow for AD and purple for PD) and contrasting DEGs (highlighted in green). The numbers and overlaps of DEGs are indicated for the five main cell types of interest: excitatory and inhibitory neurons, astrocytes, oligodendrocytes, and microglia. Several significant DEGs were identified for each category (see Fig. 3). In general, more AD-specific than PD-specific DEGs were detected, which may be explained by the disparity in the cell counts between the two datasets (≈ 1.3M nuclei for AD vs. 80k nuclei for PD), although other factors may also contribute to the overall difference in DEG counts, such as differences between the disease stages covered and disease-specific variations in the extent of cortical molecular changes. Additionally, technical differences such as different methods for nuclei isolation between the AD and PD datasets (see Methods) can impact the yield and quality of isolated nuclei. Despite the potential noise introduced by these methodological differences, their impact on the variability in the integrity and yield of isolated nuclei is considered manageable, allowing for meaningful downstream analyses and comparisons of scRNA-seq results.

Fig. 3

Venn diagram visualizing the intersections between differentially expressed genes (DEGs) in Parkinson’s disease (PD) and Alzheimer’s disease (AD): Disease-specific DEGs are shown in yellow (AD) and purple (PD), shared DEGs with joint increased expression in red, joint decreased expression in blue, and shared significance but opposite direction of the change in green. Intersections are shown for the combination for the five specific cell types of interest (excitatory and inhibitory neurons, astrocytes, oligodendrocytes, and microglial cells)

As a general observation, the analysis reveals numerous overlapping DEGs between the diseases, with an intersection set size larger than expected by chance according to Fisher’s exact test (p < 0.05 for all cell type-specific analyses). While most of these shared significant DEGs display alterations in the same direction, interestingly, a subset shows changes in opposite directions (illustrated in green). These contrasting DEGs (see definition in the Methods section on "Gene-level analysis of disease-associated changes"), along with the identified disease-specific DEGs, may serve as candidate markers to discriminate between the different neurodegenerative conditions, and provide starting points to investigate disease-specific molecular and cellular mechanisms (see also the following pathway and network analysis). To highlight the main shared and distinct DEGs identified, Table 2 lists the top 5 most significant DEGs for each category, indicating the direction of the change (blue for decreased expression and red for increased expression; log fold changes, p-values and detailed annotations for these genes are provided in the Supplementary Tab. 1). For the contrasting DEGs, arrows indicate whether their expression increases (↗) or decreases (↘) in PD (left arrow) or AD (right arrow). All DEGs for all shared cell types are reported by category, with detailed statistics on the accompanying GitLab website (https://gitlab.com/uniluxembourg/lcsb/biomedical-data-science/bds/comparison_ad_pd_single-cell.git).

Table 2 Overview of the most significant differentially expressed genes (DEGs), categorized by disease-specificity (AD-specific, PD-specific, shared DEGs, and contrasting DEGs that share only significance but not the direction of the change) and the cell types considered (excitatory and inhibitory neurons, astrocytes, oligodendrocytes, and microglia). DEGs are highlighted in blue for decreased expression and in red for increased expression. For the contrasting DEGs with opposite directionality in PD and AD, arrows indicate whether their expression increases (↗) or decreases (↘) in PD (left arrow) and in AD (right arrow). An asterisk indicates that the DEG is annotated as a marker gene for the cell type in question in the CellMarker database, and its differential expression may therefore reflect a different representation of the cell type subpopulation between conditions Gene Function Analysis

Among the most significant DEGs listed in Table 2, we note several key functional groups:

1.

Cell Adhesion and Cell-Cell Communication: Genes such as CDH19, CAV1, ITGA2, BCAS1, CCDC9B, ARHGAP42, CCDC80, and SELL play important roles in cell adhesion and cell-cell communication. Disruptions in these processes can lead to impaired neuronal connections and signaling, which are central to the progression of both AD and PD [26,27,28].

2.

Transcriptional Regulation: This group includes genes such as JUNB, NR4A1, ETS1, FOSB, and FOSL2. Abnormal transcriptional regulation can affect numerous cellular functions and is a common feature in the pathology of neurodegenerative diseases, contributing to the dysregulation of gene expression [29,30,31,32].

3.

Neurotransmitter Transport and Signaling: Key genes include SLC1A3, SVOP, GABRA3, and GLT1D1. Dysfunctions in neurotransmitter transport and signaling are hallmark features of both AD and PD, resulting in impaired synaptic communication and neuronal death [33].

4.

Cellular Stress Response and Apoptosis: Genes such as PPP1R14A, SERPINB1, BAG3, TBC1D3L, JUNB, DDIT4, and HSPA5 are important for managing cellular stress and apoptosis. Increased cellular stress and improper apoptosis contribute to neuronal loss and the progression of neurodegeneration [6, 34].

5.

Developmental Processes and Growth: This group includes FGF10, CRH, SEMA3E, and IGFBP5. Aberrations in developmental processes and growth factors can impact neuronal development and regeneration, which are affected in AD and PD [35,36,37,38].

6.

Immune Response and Inflammation: Genes such as CTSZ, GAS1, and IL6R are involved in the immune response and inflammation. Chronic inflammation is considered a significant factor in the pathogenesis of both diseases, driving further neuronal damage [39].

7.

Vesicle Trafficking and Intracellular Transport: ST18, LRP2, PROS1, and TMPRSS9 are involved in vesicle trafficking and intracellular transport. Disruptions of these processes can lead to impaired protein and organelle transport, contributing to the cellular dysfunction observed in neurodegenerative diseases [40].

8.

Cellular Signaling and Regulation of Cellular Processes: Genes such as RFTN2, DUSP5, GNB3, PLCB3, CASS4, and SH3RF3-AS1 are involved in cellular signaling and regulation. Alterations in these pathways can disrupt normal cell functions and contribute to the pathophysiology of AD and PD [41, 42].

9.

Regulation of Neuroplasticity / Fos Family Genes: FOSB and FOSL2, members of the Fos family, are involved in transcriptional regulation linked to neuroplasticity. These genes are associated with cognitive dysfunction, as they influence processes such as memory formation and synaptic plasticity [43].

10.

Maintenance of Neuronal Excitability / Excitotoxicity-Related Genes: The genes SLC1A3 and GABRA3 are both associated with the regulation and maintenance of neuronal excitability. SLC1A3 gene encodes a glutamate transporter, which is important for regulating glutamate levels in synaptic regions. Its dysfunction can lead to excitotoxicity and has been reported to contribute to neuronal damage in AD and PD [44]. GABRA3 encodes a subunit of the GABA-A receptor, playing important roles in GABAergic signaling and in the general maintenance of neuronal excitability and prevention of excitotoxicity. Alterations in GABAergic signaling are implicated in various brain diseases, including PD, where disrupted inhibitory signaling can contribute to motor and cognitive symptoms [45].

These diverse groups of the top significant DEGs illustrate the complex and multifaceted nature of the molecular changes involved in AD and PD. Overall, they highlight relevant cellular processes, regulatory factors, and mechanisms that have previously been implicated in the pathogenesis of AD, PD, or other neurodegenerative disorders.

Neuroprotective and Neurotrophic Genes

In addition to potential shared disease susceptibility genes, we also investigated the occurrence of genes with neurotrophic or neuroprotective functions among the DEGs. A corresponding curated collection of neurotrophic/neuroprotective genes associated with in vitro and in vivo evidence from biomedical literature has been assembled in the public database NeuroProDB (neuroprodb.net). The identified neurotrophic/protective DEGs are listed in Supplementary Table 2.

A protective DEG of interest with a contrasting change between AD and PD in astrocytes and inhibitory neurons is MT3 (metallothionein 3), which encodes a metal-binding protein induced under hypoxic conditions. MT3 has been reported to protect against oxidative stress by contributing to the removal of reactive oxygen species [46] and its diverging expression alterations in AD (under-expressed in both cell types) and PD (over-expressed in both cell types) indicate that it may be involved in the two diseases through different mechanisms. However, according to the CellMarker database, MT3 is also a marker gene for astrocytes and astrocyte sub-populations [47]. Thus, diverging alterations in MT3 expression may also reflect differential representations of subpopulations of cells and need to be interpreted with caution.

An example of a shared DEG with neuroprotective functions detected across multiple cell types is VEGFA (Vascular Endothelial Growth Factor A), a growth factor involved in the regulation of vascularization and angiogenesis [48]. VEGFA shows significantly increased expression in both neurodegenerative diseases in astrocytes, inhibitory neurons, and oligodendrocytes and may therefore represent a protective mechanism with broad relevance across different degenerative disorders. VEGFA's neuroprotective role is underscored by its involvement in promoting angiogenesis and mitigating neuronal damage in the context of AD and PD, and previous studies have suggested potential therapeutic applications aimed at enhancing its expression or function [49, 50].

Comparative Pathway Analysis

Pathway enrichment analysis was performed for the main cell types of interest (see Methods), and comprehensive ranking tables for all pathways and DEG categories are provided on a dedicated GitLab webpage (https://gitlab.com/uniluxembourg/lcsb/biomedical-data-science/bds/comparison_ad_pd_single-cell.git). Since a detailed coverage of the pathway analysis results for all cell types and all categories of DEGs would extend beyond the scope of this study, we present here as a representative example the enriched pathways for shared and contrasting DEGs for the five main cell types of interest, which displayed the largest numbers of DEGs. For each of these cell types, an overview of the top 3 most significant molecular functions and biological processes in the Gene Ontology database, and pathways in the KEGG database enriched for shared and contrasting DEGs is provided in Table 3.

Table 3 Overview of the top 3 most significant biological processes (BP) and molecular functions (MF) from the Gene Ontology database and pathways from the KEGG database with an overrepresentation of DEGs. Two categories of common significant DEGs between AD and PD were considered as input, those with shared (identical sign of the log fold change) and those contrasting expression changes (opposite sign of the log fold change). The labels "Up" and "Down" in the last column indicate the direction of the change when the majority of shared expression changes between AD and PD within a pathway/process display the same direction across different pathway members. The label "Equal" in the last column indicates that the same number of DEGs with increased or decreased expression was present in the corresponding pathway/process. For the contrasting DEGs with opposite directionality between PD and AD, arrows indicate whether their expression increases (↗) or decreases (↘) in PD (left arrow) and in AD (right arrow)

These results indicate that several cellular processes and pathways display significant alterations in both AD and PD, including both cell type-specific changes and changes shared across multiple relevant cell types. Six categories of significantly altered processes stand out in terms of prior evidence for their relevance in the context of neurodegenerative disorders:

Synaptic Dysfunction

We observe multiple pathways and molecular functions related to synaptic processes that display a significant overrepresentation in the contrasting DEGs in both astrocytes and excitatory neurons. The dysregulation of synapse organization is a well-documented hallmark of both PD and AD [51]. Notably, the biological process "Synapse organization" (GO:0050808) is enriched in contrasting DEGs for both astrocytes and excitatory neurons, and the pathway "Glutamatergic synapse" (KEGG, hsa04724) in contrasting DEGs for astrocytes. In astrocytes, the same contrasting pattern is also observed for other biological processes pertinent to the regulation and transmission of synaptic signaling, such as "Modulation of chemical synaptic transmission" (GO:0050804) and "Regulation of trans-synaptic signaling" (GO:0099177), as well as for molecular functions essential for synaptic activity, including "Calmodulin binding" (GO:0005516), "Monoatomic ion gated channel activity" (GO:0022839), and "Gated channel activity" (GO:0022836). Most of these pathways exhibit increased gene expression activity in PD while showing decreased activity in AD, potentially reflecting differential pathologic mechanisms underlying these neurodegenerative disorders.

Lipid Metabolism Dysregulation

The contrasting DEGs in oligodendrocytes display significant overrepresentation in biological processes associated with cholesterol metabolism (GO:0008203) and sterol metabolism (GO:0016125). Additionally, this alteration pattern is also observed in the KEGG pathways "Steroid biosynthesis" (KEGG, hsa00100) and "Protein processing in endoplasmic reticulum" (KEGG, hsa04141). This matches with the fact that the endoplasmic reticulum (ER) is critically involved in the synthesis of nearly all lipids, including cholesterol and phospholipids, which are essential for maintaining cellular membrane integrity and function. Furthermore, the results are in line with the previously documented dysregulation of lipid metabolism in multiple neurodegenerative disorders [52] indicating its broad relevance in these conditions. All pathways associated with lipid metabolism alterations display a decreased global expression activity in PD and an increased activity in AD, which could indicate distinct molecular and cellular mechanisms driving pathology in these disorders. In PD, the decreased activity may reflect neuronal loss and mitochondrial dysfunction leading to impaired lipid synthesis and processing capabilities. Conversely, in AD, the increased activity could be a compensatory response to amyloid-beta plaque accumulation, which disrupts membrane integrity and stimulates the need for enhanced lipid synthesis and repair processes [53].

Inflammation and Immune Response

Among the contrasting DEGs in microglial cells, we observed a significant enrichment of biological processes associated with immune responses, including "Myeloid cell differentiation" (GO:0030099), "Regulation of hemopoiesis" (GO:1903706), and "Regulation of leukocyte differentiation" (GO:1902105). Additionally, pathways related to inflammation, such as the "JAK-STAT signaling pathway" (KEGG, hsa04630) and the "PI3K-Akt signaling pathway" (KEGG, hsa04151), are also represented in this category. In both AD and PD, neuroinflammation and aberrant immune responses have been implicated in disease progression and pathology [54].

Cell Adhesion

The most significant pathways enriched in shared DEGs in astrocytes are associated with the processes "cell-substrate adhesion" (GO:0031589) and "cell-matrix adhesion" (GO:0007160). There is growing evidence that cell adhesion molecules (CAMs) play important roles in neurological disorders, influencing cell plasticity, neuroinflammation, vascular changes, and amyloid-beta (Aβ) metabolism [55,56,57]. In addition, alterations in CAM levels have been associated with AD in numerous studies by genetic association studies [58,59,60,61].

Ion Channel Activity Dysfunction

Across different cell types, we observed significant alterations in ion channel activity related processes. Specifically, the gene sets for "monoatomic ion gated channel activity" (GO:0022839) and "gated channel activity" (GO:0022836) display an overrepresentation of contrasting DEGs in astrocytes, and the processes "monoatomic ion channel activity" (GO:0005216) and "ATP hydrolysis activity" (GO:0016887) are enriched in shared DEGs in oligodendrocytes. Previous studies indicate that ion channel dysfunction in astrocytes is strongly associated with oxidative stress, neuroinflammation, and changes in pathological proteins associated with neurological disorders [62].

Network Analysis

To better understand gene regulatory mechanisms interlinking the identified DEGs and to determine important upstream regulators controlling these genes, we performed a gene regulatory network (GRN) analysis for the shared and contrasting DEGs (see Methods). This analysis was applied to astrocytes, oligodendrocytes, microglial cells, and excitatory neurons as the cell types with the largest number of DEGs. We built GRNs for each cell type, using the contrasting and shared DEGs. Table 4 shows each GRN top-ranked candidate regulator genes (see Methods). These candidate genes are also called perturbagens, and their activity modulation has the potential to reverse downstream pathologic gene expression changes in this network for both AD and PD (see Methods). We also identified multiple hub genes reported in Table 4 with a high network centrality (measured by the degree score) using a network topological analysis. All identified perturbagens and hub genes are listed with a brief description in Supplementary Tab 3. In addition, we searched for perturbagens overlapping across multiple cell types, to identify key regulators shared between distinct cell types. Among the perturbagens and hub genes shown in Table 4, the gene HIF1A stands out as a top-ranked perturbagen in the network for shared DEGs in microglial cells. HIF1A has a perturbation score of 4, representing the number of downstream targets DEGs whose expression can be reversed by modulating HIF1A activity. Moreover, HIF1A also displays high connectivity in multiple regulatory networks, including the network for shared DEGs in microglial cells (degree = 17, see Fig. 4) and the network for contrasting DEGs in astrocytes (degree = 23), further corroborating the relevance of HIF1A as a key regulator. HIF1A encodes a central transcriptional regulator in the HIF-1 signaling pathway, which is responsible for cellular and tissue adaptation to hypoxia. HIF1A also regulates genes involved in several other pathways with potential relevance in neurodegenerative disorders, such as apoptotic processes, iron and glucose metabolism, cell survival, and proliferation [63].

Table 4 Overview of the top perturbagens and top hub genes in networks derived from DEGs with shared or contrasting alteration patterns between AD and PD. These networks were built for 5 different cell types: astrocytes, oligodendrocytes, microglia cells, and excitatory neurons (degree: represents the total number of connections for a node; score: represents the number of downstream targets DEGs whose expression can be altered by modulating the activity of the top perturbagens)Fig. 4

Visualization of the gene regulatory network for the top DEGs with shared patterns between AD and PD identified in microglial cells. Activating interactions are highlighted in green, inhibiting interactions in red. The colored bar plots in the nodes represent the condition-specific gene expression changes, left in PD and right in AD; increases are shown in orange and decreases in blue

Apart from HIF1A, two additional top-scoring perturbagens identified include JUNB and FOS. They are both part of the AP-1 transcription factor family, which is involved in the regulation of cell proliferation and differentiation and, more specifically, in the regulation of inflammatory processes and T-cell signaling [64]. JUNB was identified as a key regulatory gene in the network of shared DEGs for oligodendrocytes (perturbation score: 238). Moreover, it was also identified as a highly connected node (degree = 66) in the regulatory network for contrasting DEGs in microglial cells, confirming its role as an important regulator of AD- and PD-associated DEGs across multiple cell types.

The gene FOS displayed high connectivity both in the regulatory networks for the shared DEGs in oligodendrocytes and in excitatory neurons. Furthermore, FOS was identified among the perturbagens in both networks for contrasting DEGs for astrocytes and microglial cells. In terms of known functional roles, FOS has been implicated both in neuronal survival pathways and in neuroprotective mechanisms by modulating cellular processes that enhance neuron resilience against stress [65].

Finally, one of the highest scoring regulators was the gene CREB1, the top-ranked perturbagen in both the network for contrasting DEGs in oligodendrocyte and in microglial cells (see Supplementary Fig. 2; perturbation scores: 100 for oligodendrocytes and 97 for microglial cells). CREB1 is the key transcription factor in the CREB signaling pathway, which regulates neuronal plasticity by facilitating gene expression necessary for long-term potentiation, memory formation, and synaptic strength. Activation of CREB1 through various signaling pathways, including cAMP/PKA and others, leads to the transcription of genes that are important for the structural and functional changes in neurons associated with learning and memory processes [66].

Cell-Cell Communication Analysis

We performed a cell-cell communication analysis to investigate shared patterns of altered communication events in AD and PD (see Methods). As a comprehensive analysis of all pairs of cell types is not feasible within the scope of the manuscript, we focus on the shared affected pathway results for astrocytes and microglial cells because they displayed profound and significant alterations and are of key interest in both diseases.

Astrocytes

When assessing disease-associated changes in cell-cell communication in astrocytes, we identified two biological processes that are commonly altered in both AD and PD, "response to hypoxia" (GO:0001666) and "positive regulation of miRNA transcription" (GO:1902893). Both displayed an overrepresentation in the target genes for the significant ligand-receptor (LR) pairs in astrocytes for both neurodegenerative diseases. Specifically, for this cell type we identified 36 significant LR pairs in the PD dataset and 45 LR pairs in the AD dataset. Notably, the process "response to hypoxia" contains the key hypoxia-associated regulatory gene HIF1A, which was already identified as significant in both the differential expression analysis and network analysis for astrocytes and microglia. Hypoxia has been increasingly recognized as a key factor in the pathogenesis of both AD and PD. It has been reported to accelerate the formation and accumulation of amyloid beta (Aβ) peptides, a hallmark of AD, through hypoxia-induced alterations in expression of the Aβ precursor protein (APP) and the secretase enzymes responsible for Aβ production [67,68,69]. Moreover, it has been described as increasing the expression and aggregation of alpha-synuclein, the key protein involved in the pathologic formation of Lewy bodies and whose aggregation contributes to the degeneration of dopaminergic neurons in PD [70].

Microglial Cells

In microglia, 26 biological processes were identified as significantly altered by cell-cell communication events (see complete list in Suppl. Tab. 4). Underlying these changes, 70 LR pairs were significant in the PD dataset, whereas the AD dataset contained 291 significant LR pairs. Among the biological processes enriched in the corresponding target genes, we identified a cluster of six processes jointly associated with inflammation/neuroinflammation, immune response, and apoptosis, that may reflect the contribution of immune dysregulation and cell death in the pathogenesis of AD and PD. They include the GO terms "microglial cell activation" (GO:0001774), "interleukin-6-mediated signaling pathway" (GO:0070102), "positive regulation of canonical NF-kappaB signal transduction" (GO:0043123), "positive regulation of superoxide anion generation" (GO:0032930), "extrinsic apoptotic signaling pathway" (GO:0097192) and "positive regulation of NF-kappaB transcription factor activity" (GO:0051092). In addition, a notable overlapping significant cellular process alteration between AD and PD was "growth hormone receptor signaling pathway via JAK-STAT" (GO:0060397). The JAK-STAT pathway is known to promote neuroinflammation in neurodegenerative diseases and has been proposed both as a target for pharmacological interventions in AD and as a potential predictive biomarker for AD [71]. This pathway was also highlighted as significant in the gene set enrichment analysis (see above), underscoring its relevance in the context of neurodegenerative pathology.

Overall, the cell-cell communication analysis identified shared affected cellular processes in AD and PD in astrocytes and microglia that align with the known molecular hallmarks of these diseases. Notably, the analysis also highlighted druggable pathways that have previously been proposed as potential therapeutic targets for AD or PD, and which may warrant further investigation for their applicability in broad-spectrum intervention strategies across multiple neurodegenerative conditions.