Integrated Stress Response Signatures Drive Monocyte Dysfunction in GBA1- and LRRK2-Linked Parkinson’s Disease

TL;DR

This study shows that monocytes in Parkinson’s disease show impaired function linked to stress responses and proteostasis issues, especially in cases with specific genetic mutations.

Contribution

The study identifies shared and mutation-specific monocyte dysfunction signatures in genetic and idiopathic Parkinson’s disease.

Findings

Monocytes in GBA1- and LRRK2-linked PD show impaired lysosomal, proteasomal, and mitochondrial function.

Integrated stress response and ER stress signatures are enriched in PD monocytes.

Phagocytosis and mitochondrial dynamics are defective, particularly in genetic PD cases.

Abstract

Monocytes are increasingly implicated in Parkinson’s disease (PD) pathogenesis, with idiopathic cases showing mitochondrial and lysosomal dysfunction. However, the impact of PD-associated mutations on monocytes remains unclear. To address this, we investigated transcriptomic and functional disturbances in peripheral monocytes from patients with GBA1- and LRRK2-associated PD and idiopathic PD. Transcriptomic data revealed shared and mutation-specific signatures, including those related to immune dysregulation, and consistent defects in lysosomal, proteasomal and mitochondrial pathways. Network and pathway analyses further uncovered downregulation in protein translation and enrichment of integrated stress response (ISR) signatures, alongside aberrant expression of genes linked to ER stress, mitophagy and type-I interferon signaling. Protein levels of heat-shock proteins and ISR effectors were elevated at baseline and following α-synuclein exposure, consistent with impaired proteostasis. Live-cell assays demonstrated defects in lysosomal function, mitochondrial dynamics, and phagocytosis, most pronounced in GBA1- and LRRK2-associated PD but evident across all PD groups. Together, these findings define a PD-associated myeloid state of immunodegeneration, marked by impaired clearance, proteostasis failure, and mitochondrial dysfunction across genetic and idiopathic PD.

Figures

C](#F1), [D](#F1)) we observed an upregulation of type-I interferon (IFN-I) signaling genes in both groups ([Fig. 2A](#F2), Supplementary Fig. 2A). This included upregulation of endogenous double stranded RNA (dsRNA) sensors (*RIG-I, ADAR, RNASEL*) and downregulation of dsRNA-binding proteins (*DHX9, ILF3*), consistent with prior PD blood transcriptome findings^[53](#R53)^ (Supplementary Fig. 2). We also observed dysregulation of genes coding for cytosolic DNA sensing pathway components (*ZBP1, TREX1, SLC19A1*)^[54](#R54)^. Additionally, interferon-stimulated genes such as *ISG15* and *IFITM1*

A](#F2). We hypothesize that impaired lysosomal and proteasomal degradation promote defective proteostasis, ER stress, and mitochondrial dysfunction in PD monocytes, driving an ISR-like “immunodegenerative” state reminiscent of degenerating neurons^[40](#R40)^ ([Fig. 6](#F6)). Furthermore, dysregulation of genes involved in mitochondrial ETC and mitophagy, paired with defective Ψ_m_, suggests accumulation of damaged mitochondria. Damaged mitochondria can leak nucleic acids that act as viral mimics and activate cytosolic sensors, thereby promoting IFN-I signaling and sustaining ISR activation^[

A-IV](#F3)). Conversely, younger onset was positively associated with the eigengenes of M6 (cell cycle, transcriptional regulation) and M8 (endosomal transport). PD diagnosis was negatively associated with the expression of module M2 (signal transduction, autophagy), while University of Pennsylvania Smell Identification Test (UPSIT) and Movement Disorder Society Unified Parkinson’s Disease Rating Scale (MDS-UPDRS, parts I-III) scores correlated with eigengenes of mitochondria and cell migration-related modules M10 and M11 ([Fig. 3A-IV](#F3)). Notably, sex-dependent expression differences were

B). The adjacency matrices were constructed using the average linkage hierarchical clustering of the topological overlap dissimilarity matrix (1-TOM). Co-expression modules were defined using a dynamic tree cut method with minimum module size of 20 genes and deep split parameter of 4. Modules highly correlated with each other, corresponding to a module eigengene (ME) correlation of 0.75 were merged, (Supplementary Fig. 4D). The genes were prioritized based on their module membership value, also known as eigengene-based connectivity (kME). The genes for each module and their gene-set enrichment

A), consistent with impaired lysosomal clearance. Per-timepoint rate comparisons showed lower initial DQ-BSA processing in iPD and higher in mutation carriers versus controls, with differences persisting over time, suggestive of exaggerated uptake and inefficient lysosomal processing (Supplementary Fig. 6). Consistently, LysoSensor measurements revealed reduced lysosomal acidification dynamics across all PD groups compared to controls ([Fig. 5.2A](#F5)–[B](#F5)). Consistently, LysoSensor measurements revealed decreased lysosomal acidification dynamics across all PD groups relative to controls

Full text

Background

Parkinson’s disease (PD) is the second most common neurodegenerative disorder and arises from a complex interplay of genetic and environmental factors^1^. Among the most prevalent genetic risk factors are variants in GBA1, which encodes the lysosomal enzyme glucocerebrosidase (GCase)^2–4^, and LRRK2, which encodes leucine-rich repeat kinase 2^5–7^. PD-associated GBA1 mutations, such as N370S, have been proposed to reduce GBA enzymatic activity, and impair lysosomal, autophagic, and mitochondrial functions in neurons, thereby promoting α-synuclein (α-syn) misfolding and enhancing Lewy body formation.^8^ In contrast, LRRK2 mutations, such as G2019S, result in elevated kinase activity and disruption of vesicle trafficking, lysosomal homeostasis, and mitochondrial dynamics.^8^

While most PD studies focus on neuronal dysfunction, both GBA1 and LRRK2 are highly expressed in myeloid cells, including brain-resident microglia and peripheral monocytes^9,10^. We and others have shown that PD risk variants modulate gene expression in myeloid cells^9–12^ and that monocytes from idiopathic PD (iPD) patients display mitochondrial, lysosomal/proteasomal, and immune pathway alterations^9,10,13^. Multiple studies have reported abnormal monocyte activation, impaired phagocytosis, and altered responses to α-syn in PD (reviewed in^14^). Dysfunctional monocytes may drive neurodegeneration through peripheral inflammation, altered neuroimmune signaling^14,15^, and CNS infiltration^16–18^. Indeed, elevated pro-inflammatory cytokines and chemokines, including the monocyte chemoattractant CCL2, have been observed in both blood and cerebrospinal fluid (CSF) of PD patients, correlating with motor and neuropsychiatric symptom severity^19^. Abnormal monocyte proportions in CSF from PD patients and preclinical models further support their role in PD-related CNS pathology^17,18,20–22^.

Inflammatory changes have also been observed in GBA1- and LRRK2-associated PD, including early evidence of immune pathway dysregulation, albeit with distinct features^9,23–27^. Functional studies using patient monocyte-derived macrophages and iPSC-derived microglia (iMGLs) have shown that PD-associated GBA1 (N370S, L444P) and LRRK2 (G2019S) mutations lead to abnormal transcriptional states, immune responses, and phagocytic activity^28–31^. We recently demonstrated that G2019S iMGLs exhibit transcriptional deregulation and enhanced myelin phagocytosis, and that monocytes from GBA1- and LRRK2-PD patients show distinct transcriptional profiles compared to iPD, particularly in immune and α-synuclein degradation pathways^9,32^.

Despite the growing evidence of altered monocyte state in PD, anti-inflammatory therapies have shown limited clinical efficacy^33^, suggesting that immune activation changes likely represent downstream effects of disrupted cellular homeostasis and underscoring the need for deeper mechanistic understanding. Monocytes from genetically defined PD cases represent a valuable model to dissect variant-specific effects and uncover convergent, targetable pathways^9^. Importantly, GBA1 and LRRK2 function may also be disrupted in iPD^34–37^, suggesting that insights gained from genetic forms may be broadly applicable across PD subtypes. In the present study, we profiled monocytes from LRRK2-PD and GBA1-PD cases, all from Ashkenazi Jewish (AJ) individuals, a genetically homogeneous population that minimizes confounding from population stratification. Alongside transcriptomic profiling, we performed protein expression and functional analyses of monocytes, linking genetic and transcriptional alterations to defects in proteostasis and cellular clearance along with mitochondrial and phagocytic functions. This integrated design enhances our ability to define disease- and mutation-specific mechanisms of monocyte dysfunction in PD and informs the development of targeted immunomodulatory therapies.

Results

Shared and Divergent Transcriptomic Features of Genetic and idiopathic PD

We recruited 52 individuals with PD carrying the LRRK2 G2019S mutation, 57 GBA1 variant carriers with PD, 124 individuals with iPD, and 41 neurologically and immunologically healthy controls from the Tom and Bonnie Strauss Movement Disorders Center and from the Bendheim Parkinson and Movement Disorders Center at Icahn School of Medicine in New York City^38,39^ (Table 1; Supplementary Fig. 1). All participants were of genetically confirmed AJ ancestry, with a mean age of 68.8 years at the time of blood collection (n = 274, see methods).

CD14^+^ monocytes were isolated from each participant for bulk RNA sequencing (RNA-seq; Supplementary Fig. 1). Following normalization and correction for known biological and technical covariates, we identified 1,595 differentially expressed genes (DEGs) in GBA1-PD monocytes compared to iPD monocytes (752 upregulated and 843 downregulated at FDR < 0.05; |log_2_-fold change (FC)| >0; Fig. 1A). In LRRK2-PD, we found 2,085 DEGs, including 1,164 upregulated and 921 downregulated genes (FDR < 0.05; Fig. 1B, |log_2_-fold change (FC)| >0; Supplementary Table 1). Comparison of DEG discovery in GBA1- and LRRK2-PD monocytes revealed 970 shared DEGs, with 583 upregulated and 388 downregulated genes.

Gene ontology (GO) analysis revealed that upregulated DEGs in both GBA1-PD and LRRK2-PD monocytes were enriched for biological processes (BPs) such as phospholipid metabolism, viral processing, cell differentiation, and apoptotic signaling (Fig. 1C, D). In addition, LRRK2-PD upregulated pathways included mitophagy, GTPase regulation, and TOR signaling (Fig. 1D, E, Supplementary Table 1). Among downregulated genes, both genetic PD subtypes showed strong enrichment for protein translation, RNA splicing, lysosomal function, and type I interferon (IFN-I) signaling, and the integrated stress response (ISR) (Fig. 1C–E, Supplementary Table 1). GBA1-PD monocytes exhibited downregulation in glycolysis, phagocytosis, viral response pathways, and Golgi vesicle transport (Fig. 1C, E, Supplementary Table 1), while LRRK2-PD monocytes showed distinct downregulation of oxidative phosphorylation (OXPHOS), ER stress, ubiquitin-proteasome system (UPS), macroautophagy, as well as downregulation of translation and splicing (Fig. 1D–E, Supplementary Table 1).

Transcriptional signatures of impaired clearance, ISR, and organelle stress mark GBA1- and LRRK2-PD monocytes

Among the most dysregulated processes in our pathway analysis was a consistent signature of downregulation of translation initiation and regulation genes in both GBA1- and LRRK2-PD monocytes, indicative of ISR activation^40^ (Fig. 1C–E, Supplementary Table 1). Specifically, among core ISR components, EIF2A was downregulated and its kinase EIF2AK1 was upregulated in both GBA1- and LRRK2-PD monocytes, along with other core ISR genes (Fig. 1A, B, F; Supplementary Fig. 2B). In line with stress-induced translational repression^41,42^, we further observed broad downregulation of ribosomal protein genes (e.g., RPL5, RPS13) (Supplementary Fig. 2) and translation regulators, including reduced expression of the PD-risk gene EIF4G1 in GBA1-PD monocytes^43^. The ISR effector ATF4 was selectively downregulated in LRRK2-PD monocytes (Fig. 1F, Supplementary Fig. 2), consistent with previous findings in iPD monocytes.^44^ To investigate this more accurately, we curated gene sets for specific sub-functions pertinent to the biological processes enriched amongst the DEGs (Fig. 1) using GO terms and the literature (see Supplementary Table 2). Genes linked to lysosome biogenesis (e.g., SORT1) and degradative enzymes (e.g., CTSB) were more prominently downregulated in GBA1-PD monocytes, consistent with prior observations in iPSC-derived neural cells from human GBA1 (N370S) carriers.^45^ On the other hand, lysosomal acidification genes (e.g., ATP6V1F, ATP6V0E1) were specifically downregulated in LRRK2-PD (Fig. 2A, Supplementary Fig. 2A), aligning with findings in neurons from G2019S murine models.^46^ RAB GTPases (e.g., RAB13) were reduced in GBA1-PD, while RAB effector genes (e.g., SYTL1) were selectively upregulated in LRRK2-PD. Pertaining the UPS, genes coding for E1/E2 enzymes and deubiquitinating enzymes (DUBs) were downregulated, while Tripartite Motif (TRIM) and RING finger (RNF) E3 ligases and adaptors (e.g., PD-risk gene FBXO7^47^) were upregulated; proteasome subunit genes showed the strongest downregulation in LRRK2-PD (Fig. 2A, Supplementary Fig. 2A), suggesting impaired UPS clearance as seen in neural cells with LRRK2 or GBA1 mutations.^48,49^ Dysregulation in genes linked to cellular clearance were accompanied by downregulation in ER chaperones (e.g., SEC63, CANX)^50^ and upregulation in ER-stress genes such as OGA, a regulator of ER-stress and ISR^51^, (Fig. 2, Supplementary Fig. 2).

Mitochondrial gene-set analysis revealed widespread downregulation of mitochondrial complex I–V genes, most pronounced in LRRK2-PD, and upregulation of coenzyme Q (CoQ) genes, potentially compensating for electron transport chain deficits. We further observed downregulation of tricarboxylic acid (TCA)-cycle, glycolysis, and lipid metabolism genes, suggesting immunometabolic reprogramming (Fig. 2A). Mitophagy-related genes were also elevated (e.g., BNIP3, OPTN, MAP1LC3A)^52^, particularly in LRRK2-PD monocytes, including the PD-risk gene PINK1 (Supplementary Fig. 2).

Furthermore, related to the enrichment in BPs such as viral response (Fig. 1C, D) we observed an upregulation of type-I interferon (IFN-I) signaling genes in both groups (Fig. 2A, Supplementary Fig. 2A). This included upregulation of endogenous double stranded RNA (dsRNA) sensors (RIG-I, ADAR, RNASEL) and downregulation of dsRNA-binding proteins (DHX9, ILF3), consistent with prior PD blood transcriptome findings^53^ (Supplementary Fig. 2). We also observed dysregulation of genes coding for cytosolic DNA sensing pathway components (ZBP1, TREX1, SLC19A1)^54^. Additionally, interferon-stimulated genes such as ISG15 and IFITM1, were also upregulated, (Supplementary Fig. 2A). Thus, by further parsing biological processes enriched in DEGs of both GBA1- and LRRK2- monocytes, we uncovered pathways that may underlie an ISR-activated state in these cells, including impaired lysosomal clearance/autophagy, mitochondrial dysfunction, ER stress and activation of viral-response pathways.^40,55^

Finally, phagocytosis pathways displayed shared downregulation of phagocytic receptors, with adaptors and cytoskeleton-modulating genes upregulated in LRRK2-PD but downregulated in GBA1-PD, alongside increased expression of genes coding for negative regulators of phagocytosis and elevated phagocytic signaling kinases in both (Fig. 2).

These transcriptional changes were accompanied by enrichment in genes linked to apoptotic signaling (e.g., BCL2, FOXP1) in both GBA1 and LRRK2-PD groups (Fig. 1F, Supplementary Fig. 2A).

Network analysis reveals mitochondrial, immune and translation modules distinguishing GBA1- and LRRK2-PD monocytes

To identify functional gene modules within gene coexpression networks, we applied Weighted Gene Co-expression Network Analysis (WGCNA, Supplementary Fig. 4) to monocyte RNASeq data from iPD, GBA1-PD, LRRK2-PD, and controls. This analysis yielded 53 modules, 13 of which (M1–M13) had eigengenes significantly associated with Parkinson’s disease diagnosis. (Fig. 3A–I). These modules were enriched for protein translation (M1, M3-M5), proteasomal functions (M1), stress response (M12) and mitochondrial functions, (M1, M3, M4, M7 and M10) (Fig. 3A–I, Supplementary Table 3). Modules M11-M13 were enriched for genes linked to cell migration, interferon signaling, immune response and cholesterol metabolism.

We next assessed how these modules relate to gene expression changes in GBA1-PD and LRRK2-PD by testing the enrichment of up- and downregulated DEGs (Fig. 3A-II). Module M13 genes were significantly enriched with downregulated DEGs of GBA1-PD monocytes. This module contained genes related to glial activation, ER metabolic pathways and cholesterol metabolism. The distribution of other modules across the DEG sets exhibited consistent patterns with our gene ontology analyses findings. Module M2, which contained genes relevant to autophagy and proteasomal functions, were enriched with up- and downregulated DEGs in LRRK2-PD, while M6 (transcription and cell cycle regulation) contained LRRK2-PD downregulated DEGs. Modules related to oxidative phosphorylation (M7, M9, M10) were enriched with up- and downregulated DEGs in GBA1- and LRRK2-PD. Module M12 genes, enriched for immune activation and IFN-I signaling functions, included downregulated DEGs in both groups. To further refine the modules’ biological relevance, we annotated gene modules with curated gene sets for metabolism, lysosomal, and mitochondrial pathways used in Fig. 2, (Fig. 3A-III, Supplementary Table 2). Mitochondrial complex I/III and lysosomal acidification genes were overrepresented in modules strongly dysregulated in LRRK2-PD, whereas lysosomal enzyme genes and mitochondrial complex I/IV components were prominent in GBA1-PD-associated modules.

We used module eigengenes to associate coexpression modules to clinical features. Older age of onset correlated positively with M5 and M7 expression (translation, mitochondrial complex I/III) and correlated negatively with Montreal Cognitive Assessment (MoCA) scores (Fig. 3A-IV). Conversely, younger onset was positively associated with the eigengenes of M6 (cell cycle, transcriptional regulation) and M8 (endosomal transport). PD diagnosis was negatively associated with the expression of module M2 (signal transduction, autophagy), while University of Pennsylvania Smell Identification Test (UPSIT) and Movement Disorder Society Unified Parkinson’s Disease Rating Scale (MDS-UPDRS, parts I-III) scores correlated with eigengenes of mitochondria and cell migration-related modules M10 and M11 (Fig. 3A-IV). Notably, sex-dependent expression differences were observed in M1, M4, M12, and M13, consistent with our previous findings of mutation specific sex-differences in risk for PD.^56^ Finally, comparison of module eigengenes between groups revealed PD-subtype specific differences (Fig. 3B, Supplementary Fig. 5). The eigengene of module M6, which contains genes relevant to transcriptional regulation and cell cycle, was decreased in LRRK2-PD monocytes relative to iPD, while the eigengene of module M10, enriched for OXPHOS genes was increased in both genetic PD-subtypes as compared to iPD. Module M12, linked to immune response and IFN-I signaling, was selectively downregulated in GBA1-PD monocytes relative to iPD (Fig. 3B, Supplementary Fig. 5).

Proteostasis vulnerability in PD monocytes is accompanied by impaired uptake of pathogenic α-synuclein.

Given the transcriptional signs of impaired proteostasis and ISR accompanied by signatures of altered mitochondrial and ER homeostasis in GBA1- and LRRK2-PD monocytes, we examined ISR and proteostasis markers in a subset of monocyte samples by Western blot relative to iPD and controls (CTR) (Fig. 4A–C). Specifically, we quantified levels of the mitochondrial chaperonin HSP60, key for mitochondrial proteostasis found altered in DA-neurons from PD cases^57^, along with phosphorylated HSP27 (pHSP27), a cytosolic stress-responsive chaperone implicated in the prevention of α-synuclein aggregation.^58^ Furthermore, we quantified protein levels of core ISR transcription factors ATF4 and ATF6, the latter being particularly important for ER-UPR.^40^

All PD groups exhibit elevated proteostasis stress markers at baseline. HSP60 levels were markedly increased in iPD, GBA1-PD and LRRK2-PD relative to controls (Fig. 4A). ATF4 showed an even stronger induction across PD groups (F(3,12) = 46.76, p < 0.0001; all PD vs CTR q < 0.0001), with GBA1-PD displaying higher ATF4 than iPD (q = 0.0049). Phospho-HSP27 was also elevated in all PD subgroups compared to controls (F(3,20) = 3.84, p = 0.0255; each PD group vs CTR q < 0.05). Notably, ATF6 was not detected in freshly isolated monocytes in the present experiment. To test whether this proteostasis vulnerability is maintained and exacerbated under proteotoxic challenge, we cultured and differentiated a subset of the same monocytes into monocyte-derived macrophages (MDMs, see Methods), and treated them with ATTO594-labeled α-synuclein pre-formed fibrils (PFF) for 40h, to investigate for 1) differences in PFF phagocytosis and 2) to assess stress-response proteins by Western blot in response to PFF uptake, (Fig. 4D–F). PFF phagocytosis was quantified by live-cell imaging in an IncuCyte S3 apparatus. Fluorescence was adjusted for technical and biological confounders, scaled and analyzed via generalized additive models (GAMs) and linear mixed-effects models to estimate group-specific uptake dynamics. All PD subgroups displayed reduced PFF uptake rates relative to controls, with the strongest effect in LRRK2-PD MDMs and iPD (Fig. 4D). Consistently, pairwise comparison of uptake rates confirmed reductions in LRRK2-PD versus controls (FDR = 10e^−6^), as well as in iPD and GBA1-PD (FDR = 3.95×10^−3^ and 4.1×10^−6^ respectively; Fig. 4E). Following 40h of PFF phagocytosis, HSP60 was strongly upregulated in MDMs from all PD groups relative to controls (Fig. 4G). ATF4 was similarly elevated across PD groups (Fig. 4H), with higher levels in GBA1-PD compared to iPD (q = 0.0049). ATF6 was also increased in MDMs from all PD subtypes (Fig. 4I), with GBA1-PD showing higher ATF6 than both iPD and LRRK2-PD (q ≤ 0.030). Uncropped western blot bands and Ponceau S staining are presented in Supplementary File 1.

PD monocyte-derived macrophages display functional impairments in cellular clearance, mitochondrial membrane potential and phagocytic functions.

Given the transcriptional deregulation in gene-sets linked to cellular clearance, mitochondrial and phagocytic functions (Fig. 2A), we next tested whether these changes were reflected in MDM function via live-cell fluorescence imaging of MDMs from a subset of controls, iPD, GBA1-PD and LRRK2-PD donors, (Fig. 5). For each assay, raw fluorescence was adjusted for technical and biological covariates (see Methods), scaled, and analyzed as covariate-adjusted fluorescence (arbitrary units, a.u.) over time using generalized additive models (GAMs) and linear mixed-effects regression (LMER) with group, time and their interaction as fixed effects and donor as a random effect. The complete statistical summary results for each assay described below can be found in Supplementary Data 1–3.

Lysosomal proteolysis was assessed using dye-quenched bovine serum albumin-Red (DQ-BSA), which is internalized by endocytosis and cleaved by lysosomal hydrolases such as CTSB^59^ to become fluorescent (Fig. 5.1B). Over the first 4h, when the signal mainly reflects uptake and early proteolysis, fluorescence increased similarly across groups (Fig. 5.1A). At later time points, when DQ-BSA is fully processed and the signal decays, mutation carriers showed a slower decline in fluorescence than controls (group×time GBA1 p = 0.016; LRRK2 p = 9.2×10^− 7^, Fig. 5.1A) and higher overall group- and subject-level signal emission rates, (Fig. 5.1C; Supplementary Fig. 6A), consistent with impaired lysosomal clearance. Per-timepoint rate comparisons showed lower initial DQ-BSA processing in iPD and higher in mutation carriers versus controls, with differences persisting over time, suggestive of exaggerated uptake and inefficient lysosomal processing (Supplementary Fig. 6). Consistently, LysoSensor measurements revealed reduced lysosomal acidification dynamics across all PD groups compared to controls (Fig. 5.2A–B). Consistently, LysoSensor measurements revealed decreased lysosomal acidification dynamics across all PD groups relative to controls (Fig. 5.2A–B). We then measured proteasomal hydrolysis rates through the fluorescent probe TAS2^60^ (Fig. 5.3A–B). Notably, all PD subgroups showed markedly reduced hydrolysis rates, in line with transcriptional deregulation in proteasomal and UPS genes in mutation carriers (Fig. 2A) and iPD monocytes^10^.

Mitochondrial membrane potential (Ψ_m_) dynamics were assessed via the potentiometric dye tetramethyrhodamine, methyl ester (TMRM^57^; Fig. 5.4). LRRK2-PD MDMs exhibited a significantly lower baseline TMRM signal (Fig. 5.4A; p = 0.0029), and slower spontaneous Ψ_m_ depolarization rates (p = 0.015), (Fig. 5.4A; Supplementary Fig. 7A), with a similar trend displayed by GBA1-PD MDMs (p = 0.07). Consistently, analysis of the area under the curves (AUCs) revealed significantly lower TMRM fluorescence in LRRK2-PD and a similar trend in GBA1-PD MDMs, (Supplementary Fig. 7B; Kruskal-Wallis H(3) = 11.23, p = 0.011; post-hoc FDR = 0.016 for LRRK2-PD and FDR = 0.055 for GBA1-PD vs controls), suggestive of a more depolarized Ψ_m_ baseline state in mutation carriers. Upon stimulation with the chemotactic, damage-associated signal adenosine diphosphate^61^ (ADP), LRRK2-PD MDMs displayed slower mitochondrial depolarization over the ADP-response window (LRRK2×time, p = 0.02; Fig. 5.4B). Per-time-point Wilcoxon tests on the rates highlighted significantly slower mitochondrial hyperpolarization at peak ADP-response time points in all PD subgroups, and significantly lower rates of subsequent mitochondrial depolarization (Supplementary Fig. 7C), indicative of reduced respiratory chain-driven proton pumping, and slower utilization of the proton motive force upon ADP stimulation.

To assess phagocytic functions, we used pHRodo-tagged myelin debris and synaptosomes (Fig. 5.5) as CNS-associated phagocytic stimuli relevant to neurodegeneration. pHRodo is a pH-sensitive dye that progressively emits fluorescence relative to uptake rate and phagolysosomal acidification.^62^ LRRK2-PD MDMs displayed significantly lower baseline myelin signal than controls (Fig. 5.5A, p = 0.009), while over time, both GBA1-PD and iPD MDMs displayed faster myelin pHRodo-myelin signal accumulation compared to control and LRRK2 MDMs, (Fig. 5.5B). Similarly, GBA1-PD and LRRK2-PD MDMs exhibited markedly steeper synaptosome trajectories relative to controls (Fig. 5.5C–D). Given that pHRodo signal reflects both particle uptake and phagolysosomal acidification, these kinetics suggest that PD MDMs accumulate myelin and synaptosome cargo more rapidly, with sustained fluorescence rate emission over time indicating less efficient cargo clearance, consistent with decreased LysoSensor signal and slower DQ-BSA processing.

Discussion

In this study, we provide a comprehensive transcriptomic, protein, and functional characterization of peripheral monocytes and monocyte-derived macrophages (MDMs) in iPD and genetic PD (GBA1- and LRRK2-PD) within a genetically homogeneous AJ cohort. By integrating bulk RNA-seq with Western blotting and live-cell imaging assays, we identify shared and mutation-specific disruptions in lysosomal, mitochondrial, and proteostasis pathways that converge on ISR activation and impaired innate immune function.

Across genetic PD subtypes, we observed downregulation of the translation machinery, lysosomal genes, proteostasis pathways, and interferon signaling, accompanied by ISR signatures. These alterations mirror prior findings in iPD neurons, monocytes and microglia^10,40,63^, and point to convergent defects in cellular clearance, metabolic homeostasis, and immune functions in PD. Over 970 DEGs were shared between GBA1- and LRRK2-PD monocytes, with partial our previous iPD monocytes dataset,^10^ involving oxidative phosphorylation, lysosomal and the UPS genes. We also identified distinct molecular signatures in GBA1- and LRRK2-PD monocytes. GBA1-PD was marked by deficits in lysosome biogenesis and enzyme genes, phagocytic receptors, and viral response pathways. In contrast, LRRK2-PD showed pronounced downregulation of UPS components, ER-stress regulators, ETC/OXPHOS and mitophagy-related genes. These patterns denote that GBA1-PD, as anticipated, primarily affects lysosomal enzyme-driven degradation, while LRRK2-PD additionally affects mitochondrial metabolism and proteostasis, aligning with prior reports in other cell types.^6,8^ Despite these differences, monocytes from both genetic forms appear to converge on a sustained ISR state.

ISR suppresses global protein synthesis via phosphorylation of the eukaryotic translation initiation factor eIF2α in response to impaired proteostasis, ER stress and mitochondrial dysfunctions.^40^ We found marked downregulation of translation-related genes, alongside reduced EIF2A expression and increased levels of the kinase EIF2AK1, a key ISR regulator implicated in α-synuclein-induced stress.^41,42^ While the ISR is increasingly investigated in neurons the context of neurodegeneration^40^, our results suggest similar mechanisms in genetic PD monocytes. Importantly, elevated HSP60, pHSP27, and ATF4 in monocytes from iPD and mutation carriers at baseline, and in MDMs after α-syn PFF exposure, support an active state of proteostatic weakness. Functional assays in patient MDMs confirmed defective lysosomal processing and acidification, paralleled by decreased proteasomal hydrolysis and aberrant mitochondrial Ψm dynamics. Notably, our results reveal a “graded” pattern of dysfunction, with iPD cells showing milder but consistent functional defects, while GBA1- and especially LRRK2-PD showing stronger phenotypes, consistent with the transcriptional results as displayed in Fig. 2A. We hypothesize that impaired lysosomal and proteasomal degradation promote defective proteostasis, ER stress, and mitochondrial dysfunction in PD monocytes, driving an ISR-like “immunodegenerative” state reminiscent of degenerating neurons^40^ (Fig. 6). Furthermore, dysregulation of genes involved in mitochondrial ETC and mitophagy, paired with defective Ψ_m_, suggests accumulation of damaged mitochondria. Damaged mitochondria can leak nucleic acids that act as viral mimics and activate cytosolic sensors, thereby promoting IFN-I signaling and sustaining ISR activation^64,65,66^. Indeed, we found enrichment of viral and IFN-related pathways, with upregulation of dsRNA and cytosolic DNA sensors (e.g., ADAR, RIG-I, ZBP1) and interferon-stimulated genes (Fig. 6).

ISR signatures have also been identified in microglia from Alzheimer’s disease (AD) brains and models^67^, and more recently in iPD microglia^63^, supporting conserved stress programs across neurodegenerative diseases and myeloid populations. However, the latter studies do not provide mechanistic insights into potential upstream causes of the microglial ISR in PD. Our data point to impaired clearance and proteostasis as candidate upstream drivers. Similar transcriptional programs have been reported in whole blood and patient-derived cells from idiopathic and LRRK2-PD cases^68,69^, further suggesting that translational repression may be a shared feature of PD. ISR-related signatures have also been reported in peripheral blood of prodromal and early-stage PD patients, and in postmortem substantia nigra from early and late PD stages^53,70,71^, underscoring its relevance across disease progression and compartments.

The presence of ISR signatures across brain-resident microglia and peripheral monocytes raises important questions about their origin and triggers. Microglia are long-lived and directly exposed to aggregated α-synuclein, a strong inducer of lysosomal damage, proteasomal dysfunction, and ISR.^40,72^ In contrast, classical monocytes are short-lived (1–3 days^73^) and not tissue-resident, making the detection of similar signatures in monocytes from both idiopathic and genetic PD particularly intriguing. Our results suggest that cell-autonomous defects in clearance and proteostasis may be partly driven by PD-linked mutations (e.g., GBA1, LRRK2) and, in iPD, by the intersection of common risk variants^10^ and environmental factors. Ex-vivo studies showed that pathogenic α-synuclein exposure induces pro-inflammatory activation and impaired autophagy in monocytes.^74–76^ Hence, future work should dissect how circulating α-synuclein and other peripheral factors contribute to monocyte dysfunction in PD.

Conclusions

In summary, our data support an ISR-linked immunodegenerative state in monocytes, accompanied by functional impairments in mitochondrial potential, cellular clearance, and phagocytosis, as a contributor to innate immune dysfunction in genetic and sporadic PD (Fig. 6). This points to new pharmacological entry points in PD, moving beyond anti-inflammatory or immunomodulatory approaches toward strategies that enhance cellular clearance and immunometabolic function. The cross-sectional design limits causal inference about whether monocyte dysfunction drives PD pathology or reflects a response to central or peripheral triggers. Medications such as L-DOPA, MAO-B inhibitors, and anti-inflammatory drugs, as well as acute infections, may also influence monocyte states and represent potential confounders. However, the close agreement between transcriptional changes in freshly isolated monocytes and functional deficits in MDMs differentiated ex vivo for seven days argues that at least part of this deregulation reflects PD-related, stable alterations rather than purely treatment effects. While the genetically homogeneous cohort reduces some sources of heterogeneity, replication in diverse populations is needed. Functional assays were performed in ex vivo MDM models, which may not fully capture in vivo immune states. Given the consistency of transcriptional signatures across PD subtypes, future studies should combine longitudinal sampling, single-cell resolution, and in vivo validation to map ISR dynamics and triggers over time, with the goal of informing biomarker development and therapeutic targeting.

Methods

Clinical centers and recruitment strategies

Subjects were enrolled from the Bonnie and Tom Strauss Movement Disorders Center and the New York Movement Disorder (NYMD) cohort at the Bendheim Parkinson and Movement Disorders Center (BPMD) at the Icahn School of Medicine at Mount Sinai. Additional subjects were also included from other PD and aging cohorts including Marlene and Paolo Fresco Institute for Parkinson’s and Movement Disorders at NYU Langone Health (New York); the Alzheimer’s Disease Research Center (ADRC) at Mount Sinai; and the Center for Cognitive Health (CCH) at Mount Sinai Hospital (New York). Each institution’s Institutional Review Board (IRB) approved the study protocol, recruitment procedures, and collection of data and biospecimens. Written informed consent was obtained from all participants prior to enrollment. PD diagnoses were confirmed by movement disorder specialists according to the United Kingdom Parkinson’s Disease Society Brain Bank Clinical Diagnostic Criteria^77^, but included those with family history. Healthy controls (CTRL) were individuals with no personal history of neurological disease and no first- or second-degree relatives affected by a neurological condition.

Blood Collection and PBMC, monocyte Isolation and RNA extraction

Blood samples were collected fresh during the morning of each research visit to minimize variability in sample composition and cell activation. Samples were drawn into Vacutainer tubes containing acid citrate dextrose (ACD) (BD Biosciences) and processed within 3–4 hours. DNA was extracted from 0.5 ml of whole blood using the QIAamp DNA Blood Midi Kit (Qiagen) following the manufacturer’s protocol, and DNA quality and concentration were assessed using a Nanodrop spectrophotometer. Blood samples were processed as previously described^10^. Briefly, processing consisted in isolation of peripheral blood mononuclear cells (PBMC) and subsequent CD14 + monocytes purification. For PBMC isolation, SepMate tubes (StemCell Technologies) were used. After dilution in 2-fold PBS (Gibco) tubes were filled with 15 ml of Ficoll-Plaque PLUS (GE Healthcare) and centrifuged at 1200g for 10 mins, followed by wash with PBS. 5 million PBMCs were immediately used for automated monocyte isolation through the AutoMacs sorter (Miltenyi) with human CD14 + magnetic beads (Miletenyi), according to manufacturer’s instructions. Sorted monocytes were stored at − 80°C in RLT buffer (Qiagen) + 1% 2-Mercaptoethanol (Sigma Aldrich). Isolated monocytes stored in RLT buffer were first thawed on ice. RNA was isolated with the RNeasy Mini kit (Qiagen) according to manufacturer’s instructions, including the DNase I optional step. RNA was then stored at − 80°C until library preparation. RNA integrity number (RIN) was assessed with TapeStation using Agilent RNA ScreenTape System (Agilent Technologies). RNA concentration was obtained with Qubit. The remaining PBMCs were cryopreserved in 90% FBS (Germini) + 10% dimethyl sulfoxide (DMSO, Sigma Aldrich) at a concentration of 10 million cells/ml in Nalgene cryogenic vials (ThermoScientific).

Monocyte isolation

Monocytes were isolated from cryopreserved PBMCs by Magnetic-Activated Cell Sorting (MACS; Miltenyi) using human CD14 antibody-conjugated magnetic microbeads (Miltenyi), following protocols previously established.^10^ Sorted monocytes were plated at a density of 25,000 cells per well in poly-D-lysine-coated 96-well plates in DMEM/F12 medium (Gibco) supplemented with 10% fetal bovine serum (FBS), Glutamax, and 1% penicillin-streptomycin. Differentiation into macrophages was induced by adding human recombinant M-CSF (50 ng/mL; Peprotech), with media changes performed every third day. Functional maturity was reached at day in vitro 7 (DIV7), all assays were thus performed at DIV7. Functional assays were conducted in triplicate using the IncuCyte S3 live-cell imaging system (Sartorius). Bright-field and fluorescent images were acquired every hour and analyzed using the Incucyte Software (v2024b). Integrated fluorescence intensity was normalized to cell confluency and used for downstream quantification. All assays were performed with the experimenter blinded to the group identity of the samples.

Functional assays

Mitochondrial membrane potential

In order to collect longitudinal data on the mitochondrial membrane potential, cells were stained with 25 nM tetramethylrhodamine, methyl ester (TMRM, ThermoFisher) and left in the media, as previously described^78^. At this concentration TMRM does not interfere with mitochondrial functions and its concentration within the mitochondrial matrix is low enough to ensure fluorescence intensity is proportional to the membrane potential, without self-quenching artifacts.^78^ Following 20h baseline measurements, 100 μM ADP (MilliporeSigma) were added to test the mitochondrial response to a chemotactic, damage-associated stimulus.^61^

Lysosomal proteolysis and proteasomal hydrolysis

Dye quenched bovine serum albumin-Red (DQ-BSA-Red, ThermoFisher), was employed to measure lysosomal proteolytic functions. Cells were loaded with 0.5 μg/ml for 30 minutes at 37°C according to manufacturer’s instructions. Following incubation, cells were washed with warm cell culture media. The fluorescence signal post-wash is proportional to the cellular DQ-BSA accumulation (endocytic capacity) and lysosomal degradation efficiency. For proteasomal hydrolysis, cells were incubated with TAS2 at 1 μM in culture media for 30min and washed with warm media before imaging. The TAS2 is a highly specific substrate for the eukaryotic proteasomal 20S core particle subunit.^60^ TAS2 emits fluorescence once hydrolyzed by the 20S core particle subunit, it is non-toxic and does not inhibit proteasomal functions, and is designed for long-term monitoring of proteasomal functions in live cells.^60^

Myelin and synaptosome preparation: To create biologically representative samples, minimizing single-donor biases in myelin and synaptosome composition, the latter were extracted and pooled from frozen postmortem occipital cortex tissue of 8 male non-neurological control donors (mean age: 71 ± 14 years). Myelin was extracted via percoll gradient centrifugation following the protocol described in^79^. Briefly, brain tissue was thawed in ice-cold PBS (Gibco) + protease inhibitors (Roche cOmplete protease inhibitor cocktail) and dissociated in a dounce homogenizer (Millipore Sigma). Homogenized brain was pelleted, resuspended in 30% percoll in PBS, overlaid with PBS and centrifuged at 3000 xg, 10 minutes at 4°C to separate myelin from cellular and tissue debris. Myelin was collected between the layers, washed x2 in ultra-pure distilled water then washed 2x in PBS. Synaptosomes were purified from the same tissue samples using the Syn-PER^™^ Synaptic Protein Extraction Reagent (Thermofisher) + protease inhibitors, following manufacturer’s instructions. Stocks of myelin and synaptosomes were prepared in PBS with a final volume adjusted to obtain a final protein concentration of 1μg/μl. Endotoxin levels were measured using a commercial kit following manufacturer instructions (GenScript). Myelin and synaptosomes fragments were conjugated with 1:100 pHRodo red dye (Thermofisher) according to the manufacturer’s protocol and stored at −80°C.

Phagocytosis assay

2.5μg of pHRodo-tagged myelin and synaptosomes respectively were added to 150μl media/well and cells were imaged hourly over 24h. For α-synuclein PFF phagocytosis, 0.5μg of ATTO 594-conjugated human recombinant PFFs (Type 1, StressMarq Biosciences, cat. nr. SPR-322-A594) were added to 150μl media/well and imaged hourly over 40h.

Statistical analysis

Time-lapse fluorescence data (normalized/cell confluence) from each assay was analyzed in R (v2025.09.2 + 418) using mgcv, lme4/lmerTest, emmeans, and rstatix. For each assay, fluorescence values per well were merged with donor-level metadata (diagnostic group, age at blood draw, sex, ancestry, clinic, weeks of cryopreservation, and sample/assay dates). Fluorescence intensities were rescaled (×10^− 6^) for numerical stability; all analyses were performed on these scaled values. To obtain covariate-adjusted trajectories and rates of change, we first performed covariate screening with a linear model including time, group, and confounders. Covariates with p < 0.05 at the variable level were retained for the final models. For each assay, we then fit generalized additive models (GAMs; mgcv) with scaled fluorescence as the response, using group-specific smooths over time and subject-level random effects

$$[eqn]$$

where factor-smooth interactions (bs = “fs”) captured donor-specific temporal trajectories (or random-effect smooths if convergence failed). Models were fitted by REML. From each GAM, we extracted (i) covariate-adjusted fluorescence trajectories and (ii) instantaneous rates of change (dF/dt) computed via central finite differences on donor-specific fitted curves. Subject-level mean rates were calculated by averaging dF/dt across the assay window. We additionally fitted linear mixed-effects models (lme4/lmerTest) to quantify overall group differences:

$$[eqn]$$

Group-specific time slopes and pairwise slope differences (Δslope) were estimated using emmeans::emtrends and reported as scaled fluorescence per hour (×10^− 6^ a.u./h). Subject-level mean rates and area-under-curve (AUC) values were compared across groups using Kruskal–Wallis tests followed by Benjamini–Hochberg–corrected pairwise Wilcoxon tests.

Western Blotting

Cell lysates were generated using Roche cOmplete Lysis-M buffer (Millipore-Sigma) with protease and phosphatase inhibitor cocktails (Millipore-Sigma). Protein concentration was estimated using the BCA assay (Pierce). For Western blot analysis, 8–10 μg total protein was denatured under reducing conditions in NuPage LDS sample buffer (ThermoFisher) by heating for 5 min at 90°C before loading onto a 4–12% precast polyacrylamid (Bio-Rad), then transferred to a PVDF membrane (0.2 μm; Trans-blot turbo, Bio-Rad) using the iBlot 2 dry blotting system (Invitrogen). Membranes were blocked for 1h at RT in 5% BSA in TBS containing 0.1% Tween-20 (Fisher Scientific; TBS-T). Membranes were then incubated in the indicated primary antibody (in 5% BSA/TBS-T) overnight at 4°C. Primary antibodies: ATF-4 (D4B8, Cell Signaling #11815); AF6 (D4Z8V, Cell Signaling # 65880); HSP60 (abcam #ab190828); phospho-HSP27 (Ser82) (D1H2F6, Cell Signaling #9709). Following o/n incubation membranes were washed 4 times in TBS-T, incubated in species-specific HRP-conjugated secondary antibody (in 5% BSA/TBS-T, Cell Signaling # 7074) for 1h at RT, and then washed 4 times in TBS-T. SuperSignal West Femto Maximum Sensitivity Substrate (ThermoFisher) was used to detect target signals. Membranes were then washed once in TBS-T and stripped in Restore stripping buffer (ThermoFisher) with vigorous shaking to remove primary and secondary antibodies, washed three times in TBS-T, and blocked for 1 h (in 5% BSA/TBS-T) at RT before probing with the next primary antibody. To account for differences in protein loading, target band intensities for each sample were normalized by the total protein loaded for the same sample via Ponceau S staining as previously described.^80^ For staining of total protein load, membranes were incubated in 0.1% Ponceau S (Millipore-Sigma) in 1% acetic acid for 8 minutes at room temperature. Membranes were imaged in an Invitrogen iBright 1500 Imaging System. Ponceau S and target band intensities were measured as signal area under the curve via ImageJ (v1.53a).

RNA sequencing

RNA-seq libraries were prepared using the TruSeq Stranded Total RNA Sample Preparation Kit with the Low Sample (LS) protocol (Illumina) following the manufacturer’s instructions. RNA libraries were prepared by a commercial service (Azenta Inc., formerly Genewiz Inc.) using a standard RNA-seq protocol. For all samples, ribosomal RNA was removed using a ribo-depletion strategy. Sequencing was performed at Azenta Inc. on an Illumina HiSeq 4000 platform, generating 150-bp paired-end reads with an average depth of 60 million reads per sample. Libraries were processed in four independent sequencing batches.

SNP and GBA1/LRRK2 Genotyping

DNA samples were genotyped using the Illumina Infinium Global Screening Array (GSA), which includes a genome-wide backbone of 642,824 common variants and ~ 60,000 custom disease-related SNPs. Targeted genotyping for common GBA1 and LRRK2 mutations was performed at Dr. William Nichols’ laboratory (Cincinnati Children’s Hospital). For LRRK2, the G2019S variant was screened. For GBA1, 11 variants that are more frequent in the Ashkenazi Jewish population were analyzed: IVS2 + 1, 84GG, E326K, T369M, N370S, V394L, D409G, L444P, A456P, R496H, and RecNcil. Mutation frequencies were calculated for the entire cohort as well as for manifesting and non-manifesting carriers.

Genotyping QC and Ancestry Analysis

Quality control (QC) filters included minor allele frequency (MAF) > 5%, SNP and sample call rates > 95%, and Hardy–Weinberg equilibrium (HWE) P > 1 × 10^−6^. Duplicate or related samples were identified and removed using pairwise identity-by-descent (IBD) estimation in PLINK (PI_HAT = 0.99–1).

Genetic ancestry was assessed by principal component analysis (PCA) and multidimensional scaling (MDS), comparing the study cohort with Phase 3 reference samples from the 1000 Genomes Project^81^. For AJ ancestry, analyses were repeated using a custom AJ reference panel^82^. Non-AJ samples were excluded using Somalier v0.2.12^83^, with reference space derived from AMP-PD genotypes, following the approach of Iwaki et al.^84^.

Data processing and normalization of RNA-seq

For each cohort, gene-level expression was quantified using the GENCODE v30 genome reference and the RAPiD RNA-seq pipeline^85^. The RAPiD-nf pipeline, built on Mount Sinai’s APOLLO framework and implemented in Nextflow, processed paired-end FASTQ files through automated alignment, quantification, and quality control (QC)^86^. QC metrics were generated using Picard (v2.20)^87^ and FASTQC (v0.11.8)^88^. Trimming and alignment were performed with Trimmomatic (v0.36)^89^ and STAR (v2.7.2a)^90^, respectively. Gene and isoform abundance were estimated using RSEM (v1.3.1)^91^, and mapped reads were annotated to genomic features using featureCounts (v1.6.3) to generate read summarizations^92^.

Differential gene expression analysis

Differential gene expression analysis of monocytes was performed using edgeR (v4.0.16), limma (v3.58.1)^93^, and SVA (v3.5)^94^. Genes with a median TPM < 1 were excluded. Surrogate Variable Analysis (SVA) was applied to estimate and remove unknown sources of variation, while preserving mutation and disease status (Supplementary Fig. 3); 11 significant surrogate variables were included as covariates in the design matrix. Count data were normalized using the TMM method, incorporated into a limma object, and voom-transformed. Linear models were fitted for each gene, and contrasts were applied to identify differentially expressed genes for each group comparison.

Pathway Enrichment Analysis

Pathway enrichment was performed on the differentially expressed genes using clusterProfiler(v4.10.1). A filter of 0.05 was applied to the adjusted p-values (method = “BH”). The enrichment analysis was performed for upregulated (logFC > 0) and downregulated (logFC < 0) genes for the GO pathway databases using the function enrichGO(). A data frame containing term ID, pathway or category description, GeneRatio, and adjusted p-value was used.

To visualize gene- and pathway-level expression trends, expression data were normalized and batch-corrected using Surrogate Variable Analysis (SVA). For each pathway (lysosomal function, mitochondrial, ubiquitin–proteasome, interferon signaling, organelle stress, metabolism, phagocytosis), mean expression was calculated per group and center-scaled within pathways. Scaled values were plotted as a heatmap, enabling comparison of relative up- or downregulation across groups while minimizing interpathway variability.

Weighted Gene Co-Expression Analysis (WGCNA)

Weighted gene co-expression network analysis (WGCNA) was applied to corrected gene expression data in order to construct gene correlation networks and co-expression modules using the WGCNA(v1.72–5) package^95^. The counts were corrected using the same parameters as in the differential analysis for consistency. Using the sva_network package, we computed the SV loadings of the standardized expression matrix with singular value decomposition (SVD) and computed the residuals after regressing the top 12 SVs. Linear regression between the SVs and the covariates showed correlation mostly with technical covariates, including lane, batch, percentage of ribosomal bases and other sequencing metrics such as % of mRNA and intergenic bases (Supplementary Fig. 4A).

The co-expression network analysis was performed using the R package of Weighted Gene Correlation Network Analysis (WGCNA)^95^ following the standard pipeline to fit a scale-free topology (R2 > 0.8) and applying a Soft Threshold power of 5 into a signed network model (Supplementary Fig. 4B). The adjacency matrices were constructed using the average linkage hierarchical clustering of the topological overlap dissimilarity matrix (1-TOM). Co-expression modules were defined using a dynamic tree cut method with minimum module size of 20 genes and deep split parameter of 4. Modules highly correlated with each other, corresponding to a module eigengene (ME) correlation of 0.75 were merged, (Supplementary Fig. 4D). The genes were prioritized based on their module membership value, also known as eigengene-based connectivity (kME). The genes for each module and their gene-set enrichment analysis are shown in Supplementary Table 2. A Wilcoxon rank sum test was performed across modules comparing GBA1/iPD, LRRK2/iPD, GBA1/LRRK2, and iPD/Control groups to determine significance. The parameters used are paired is false (two-sided, Wilcoxon rank-sum test) and the p-adjust.method uses the Benjamini-Hochberg (BH) method. The significance criteria used was filtering at an adjusted p value less than 0.05.

To assess the enrichment of WGCNA modules within differentially expressed genes (Fig. 2A-II), and enrichment of custom gene-sets within modules (Fig. 2A-III), we performed Fisher’s exact tests to calculate enrichment ratios comparing the observed overlap of module genes with up- or down-regulated genes in GBA-PD and LRRK2-PD versus iPD or enrichment of custom gene-sets within module genes.

Visualization and Plots.

All plots were created using ggplot2(version 3.5.0) in R(v4.3.3), with ggrepel(v0.9.5), and ggfortify(v0.4.17) for additional layers of visualization.

Supplementary Material

Supplementary Files

This is a list of supplementary files associated with this preprint. Click to download.

• SupplementaryData1Clearance.xlsx
• SupplementaryData2Phagocytosis.xlsx
• SupplementaryData3Mitochondria.xlsx
• SupplementaryFile1.pdf
• SupplementaryTable1.xlsx
• SupplementaryTable2.xlsx
• SupplementaryTable3.xlsx
• SupplementaryFiguresMatteietal.pdf