Graphical Abstract

   graphic file with name fx1.jpg
   [35]Open in a new tab

Highlights

     * •
       bFGF enhances the neuronal release or extracellular vesicles (EVs)
     * •
       bFGF induces proteomic changes in the neuronal cell lysate and in
       neuronal EVs
     * •
       bFGF incites changes in Lewy body pathology (LBP)-linked protein
       networks
     * •
       bFGF-induced molecular interactions may endorse LBP
     __________________________________________________________________

   Biological Sciences; Molecular Neuroscience; Proteomics

Introduction

   Synucleinopathies are characterized by the gradual appearance of
   intraneuronal inclusion bodies, termed Lewy bodies (LBs). These
   conditions encompass Parkinson's disease (PD) as well as PD dementia
   (PDD) and dementia with Lewy bodies (DLB) ([36]Galasko, 2017). The
   latter two entities are characterized by the presence of LBs in
   hippocampal neurons and associated memory circuits ([37]McKeith et al.,
   2017). In accord with such a distribution, cognitive deficits including
   impaired learning and memory are important clinical features of PDD and
   DLBs. LBs are intraneuronal protein aggregates composed of crowded
   organelles and lipid membranes and the protein alpha-synuclein (α-Syn)
   ([38]Shahmoradian et al., 2019; [39]Spillantini et al., 1997). The
   exact mechanism of α-Syn-induced neuronal dysfunction and death remains
   elusive, but prior research implicated changes in neurotrophic factor
   signaling and neuronal energy metabolism during pathological changes.
   In addition, cell-to-cell transmission of α-Syn through extracellular
   vehicles (EVs) likely contributes to the spread of pathology in these
   conditions. The molecular factors that control neuronal EV release are
   therefore likely to contribute to LB progression.

   FGFs are a family of pleiotropic growth and differentiation factors
   that regulate CNS homeostasis in health and disease. Although FGFs are
   best known for their roles in the early steps of patterning the neural
   primordium and proliferation of neural progenitors, they have equally
   important roles in the adult brain, where they regulate neuronal
   calcium homeostasis and plasticity further promoting neuroprotection
   and repair in response to neural tissue damage. In addition to these
   physiological roles, basic fibroblast growth factor (bFGF) has been
   linked with responses to neuronal injury ([40]Fagel et al., 2009;
   [41]Timmer et al., 2004; [42]Yoshimura et al., 2001; [43]Guillemot and
   Zimmer, 2011) or to psychiatric conditions ([44]Turner et al., 2012;
   [45]Deng et al., 2019). In addition to these well-known roles of bFGF,
   we have recently demonstrated bFGF-controlled release of EVs from
   hippocampal neurons ([46]Kumar et al., 2020a). Because of these
   functions, bFGF may likewise affect LB-associated pathological changes.

   Here, we investigated the cell lysate (CL) and EV proteome changes of
   hippocampal neurons in response to bFGF. Using high-resolution mass
   spectrometry (MS), we quantified differentially expressed
   proteins (DEPs) in the CL and EV fractions of bFGF-treated hippocampal
   neurons. To capture the protein interactions among these DEPs, we
   adapted a weighted protein co-expression network analysis (WPCNA)
   methodology and probed the subsequent co-expression modules for
   LBP-associated proteins. This approach allowed us to extract
   LBP-enriched modules and revealed the molecular interactions between
   bFGF signaling and LBP-associated molecular changes. We specifically
   identified LBP-related RNA-binding proteins (PD-RBPs), numerous ion
   channels and receptor proteins, and α-Syn interacting proteins as
   interactome components that connect bFGF signaling to α-Syn pathology.
   Therefore, our results will support the investigation of bFGF signaling
   in α-Syn-associated pathological changes.

Results

Characteristic CL and EV Proteome Changes Are Induced by bFGF

   In order to examine bFGF-induced proteome changes in the CL and in EVs,
   we treated hippocampal neurons for 24 h with bFGF (50 ng/mL) and
   subjected the CL and EV fraction to high-resolution mass spectrometry
   (MS) (see [47]Transparent Methods for experimental details)
   ([48]Figure 1A). Primary rat hippocampal neurons from E18 CD (Sprague
   Dawley) rat embryos were used for all experiments. All experiments were
   performed by using at least biological triplicates and exhibited very
   strong Pearson correlation coefficients (r ∼ 0.98–1) among the
   replicates, demonstrating a high reproducibility in both CL and EV
   datasets ([49]Figures S1A and S1B). All proteins were detected at least
   thrice in the technical replicates of control and bFGF-treated samples.
   These criteria identified n = 5,314 and n = 2,258 proteins for the CL
   and EV fraction, respectively. We next performed an unsupervised
   clustering using principal component analysis (PCA) on CL and EV
   proteomic datasets. As expected, we found a clear separation between
   the two conditions ([50]Figure 1B), and Euclidian distance-based
   hierarchical clustering confirmed these findings ([51]Figure 1C). Next,
   we analyzed the differential expression of proteins (DEPs) to define
   the proteomic signature from CL and EV. This analysis yielded a set of
   n = 1,660 and n = 650 DEPs in CL and EV fractions ([52]Tables S1A and
   S1B). Taken together, our results demonstrate that bFGF extensively
   affects the expression of a large number of proteins in the CL and EVs
   from primary hippocampal neurons.

Figure 1.

   [53]Figure 1
   [54]Open in a new tab

   Evaluation of Proteomic Changes in bFGF-Treated Rat Brain Hippocampal
   Primary Neurons

   (A) Schematic representation of proteomic data collection from rat
   hippocampal primary neurons cell lysate (CL) and extracellular vesicles
   (EV). The protein expression data were used to perform network and Lewy
   body pathology enrichment analysis.

   (B) Principal component analysis (PCA) based on CL and EV pellet
   proteome datasets. Total three-dimensional PCA plotting as 86.8% of
   variance (PC1 = 82.03%, PC2 = 3.47%, PC3 = 1.37%).

   (C) The dendrogram represents the hierarchical clustering based on the
   Euclidean distances computed from log[2] LFQ intensities.

   See also [55]Figure S1.

Co-expression Analysis Organized Proteome-wide Changes in CL and EV into
Modules

   Next, we performed WPCNA on proteins differentially expressed in
   response to bFGF ([56]Figure S2). Co-expression module analysis yielded
   nine CL (M[CL]1–M[CL]9) and four modules EV (M[EV]1–M[EV]4) modules
   from CL (n = 5,314) and EV (n = 2258) proteome datasets, respectively
   ([57]Figures 2A and 2B). To further examine the bFGF-induced proteome
   signature, the statistically significant, differentially expressed
   proteins (sDEPs) were mapped to the co-expression modules. Among the CL
   modules, M[CL]2 and M[CL]8 depicted the highest sDEP signature
   ([58]Figure 2C), and of the four EV modules, M[EV]1, M[EV]2, and M[EV]3
   were found with the strongest sDEPs signature ([59]Figure 2D). These
   sDEP modules were considered for further analysis ([60]Tables S2A and
   S2B). In WPCNA analysis, modules are represented by a weighted
   expression profile (Eigenprotein) of co-expressed proteins. To test the
   expression pattern of each sDEP module, we computed module Eigenprotein
   values and found an increased expression of proteins in M[CL]2 and
   M[EV]2 and 3 modules, whereas modules M[CL]8 and M[EV]1 showed a
   decreased expression in response to bFGF treatment ([61]Figures 2E and
   2F). An additional module preservation analysis demonstrated that
   co-expression modules from both datasets were strongly preserved, with
   a Z-summary score above 10, in comparison with random modules
   ([62]Figures S3A and S3B). Next, we sought to identify the common
   modules and computed Jaccard similarity co-efficient between the CL and
   EV. This analysis indicated a higher overlap between the M[EV]1 and the
   M[CL]8 module ([63]Figure 2G), and both modules had down-regulated
   proteins in response to bFGF treatment. Next, we explored the common
   interactions of key regulatory proteins (so-called hubs) in the M[CL]8
   and M[EV]1. Hub proteins play an important role in regulating
   biological functions and are identified in common modules by
   calculating the module membership (MM) of each protein combined with
   protein interaction network analysis (see [64]Transparent Methods
   section). These analyses identified Gpi, Glul, Aldoc, Ldhp, Pygp, and
   Prdx6 (we used the rat gene symbol nomenclature for each protein
   throughout the text) as the six main down-regulated hub proteins
   ([65]Figure 2H). The main interaction partners of hub proteins in the
   corresponding CL modules were Lrrc47, Cd44, Cald1, Xpo1, Pja2, and
   Micu2. A pathway enrichment analysis of the common modules (M[CL]8 and
   M[EV]1) revealed a huge number of biological functions for these
   proteins ([66]Figure 2I). In sum, these results provided an initial
   modular assessment of the bFGF-incited proteomic changes in hippocampal
   neurons.

Figure 2.

   [67]Figure 2
   [68]Open in a new tab

   WPCNA Analysis of Cell Lysate (CL) and Extracellular Vesicles (EV) bFGF
   Induced Proteomes

   (A) Dendrogram clusters CL proteins (n = 5,314) into nine modules.

   (B) The dendrogram groups EV proteins (n = 2,258) into three
   co-expression modules.

   (C and D) The bar plots represent the enrichment of differentially
   expressed proteins (DEP) signatures in CL and EV co-expression modules;
   x and y axis denote the modules and percentage overlap of the DEP
   signature (∗∗∗p ≤ 0.01, negative log[10] Benjamini-Hochberg adjusted p
   values; Fisher's exact test).

   (E and F) Module expression profiles of two CL and three EV modules
   (Wilcoxon test p value: 0.0034, 0.00094), respectively.

   (G) Module resemblance between each set of modules was assessed by the
   Jaccard similarity co-efficient between their sets of CL and EV
   modules. See also [69]Figures S2 and [70]S3.

   (H) The module plot shows the common module the M[EV]1 and M[CL]8
   modules based on the high Jaccard similarity coefficient.

   (I) The most significant (negative log[10] p ≤ 0.05) biological
   processes gene ontology (GO) terms of common module between CL and EV.
   Note: red and blue colors indicate up- and down-regulation,
   respectively. Gene symbols corresponding to proteins are used as labels
   in the module plot.

Proteins Linked to Lewy Body Pathologies Are Enriched in CL and EV
Co-expression Modules

   We next mined the proteins that associate to Lewy body pathologies
   (LBPs) from the widely published PD literature and intersected them
   with our co-expression modules in order to explore LBP-related proteins
   among the bFGF-exhorted modules ([71]Figure 1A). We endorse that
   associative likelihood of these results may be more for LBP due to the
   aforementioned clinical features of PD and primary hippocampal neurons
   as a source of datasets used in this study. We found M[CL]2 and M[CL]8
   and M[EV]1 and M[EV]3 modules to be significantly (BH corrected p value
   ≤ 0.05) enriched with the LBP molecular signature ([72]Figures 3A and
   [73]S4). The assimilation of markers from genome-wide association
   studies (GWASs) and proteome-wide data permits the identification of
   potential molecular mechanisms in disease modules. Thus, we compiled 15
   statistically significant LBP-GWAS studies (see [74]Transparent
   Methods) and intersected them with bFGF-incited sDEP modules. Again, we
   found a higher degree of overlap with GWAS datasets in M[CL]2 and 8 and
   in M[EV]1 and 3 ([75]Tables S3A and S3B). All LBP modules consisted of
   up- and down-regulated proteins further contributing to a substantial
   level of heterogeneity among the interacting proteins, where M[CL]8 and
   M[EV]3 were more heterogeneous as compared with M[CL]2 and M[EV]1 in
   terms of up- or down-regulated proteins ([76]Figure S4). For a better
   understanding of this heterogeneity we conducted a module-wise pathway
   enrichment analysis revealing module-specific up- and down-regulated
   pathways ([77]Figure 3B).

Figure 3.

   [78]Figure 3
   [79]Open in a new tab

   Enrichment of Lewy Body Pathology (LBP) Linked RNA-Binding Proteins
   (RBPs) in Cell Lysate (CL) and Extracellular Vesicles (EV)
   Co-expression Modules

   (A) The bar plots show Lewy body pathology proteins enrichment in CL
   and EV co-expression modules. See also [80]Figure S4 (Note: The y axis
   of bar plot denotes negative log[10] Benjamini-Hochberg adjusted p
   values, Fisher's exact test, dotted line represents the statistical
   significance 1.31, is comparable with the p values ≤ 0.05).

   (B) Biological processes enrichment analysis on LBP Co-expression
   modules both CL (left-side) and EV (right-side).

   (C and D) Module plots representing the RNA-binding proteins and their
   interacting partners of LBP modules from both CL and EV up and down
   biological processes enrichment of each module. Note: Red and blue
   colors denote up and down-regulation, respectively. Gene symbols
   matching to proteins are used as labels in the LBP-RBPs modules.

bFGF Modulates the Abundance of LBP-Associated RNA-Binding Proteins

   Next, we examined the role of disease-associated pathological processes
   in these modules. Altered RNA metabolism is associated with familial PD
   ([81]Lu et al., 2014). Therefore, we screened our data for LBP-related
   RNA-binding proteins (LBP-RBPs). This analysis revealed LBP-RBPs in the
   CL (M[CL]2 = 8; M[CL]8 = 4) and EV (M[EV]1 = 3; M[EV]3 = 1) modules,
   thus allowing us to construct LBP-RBPs interaction modules in each
   dataset and determine hub-RBPs along with their respective enriched
   pathways ([82]Figures 3C and 3D). In the CL, Rps6, Eef2, Srp14, Gspt1,
   Ddx6, Lars, Atxn2, and Rps14 were found as the key up-regulated
   LBP-RBPs, whereas Hnrnpa2b1, Hnrnph3, Fus, and IIf2 were key
   down-regulated LBP-RBPs. In the EV fraction, Aco1, Eef2, and Hnrnpa2b1
   were key down-regulated LBP-RBPs and Rps14 was a key up-regulated
   LBP-RBP. From the cellular and subcellular location analysis of hub
   LBP-RBPs, we found up-regulated proteins like Rps6 localize mainly to
   mitochondria, Eef2 localizes to the plasma membrane and cytosol, Srp14
   to nucleoli, Gspt1 and Atxn2 to the cytosol, Ddx6 to cytoplasmic
   bodies, Lars to nuclear bodies, and Rps14 to the endosomal reticulum
   and cytosol. Most of the nucleoplasm LBP-RBPs are down-regulated except
   for Eef2 (localized to the plasma membrane and cytosol) and Aco1 (found
   in mitochondria and cytosol). In summary, these results suggest an
   effect of bFGF on a variety of LBP-associated RBPs located to distinct
   cellular compartments.

bFGF Predominantly Affects the Abundance of LBP-Associated Metabotropic
Receptors

   In addition to RBPs, neuronal ion channels and receptors contribute to
   α-Syn-associated pathological changes in LBP ([83]Surmeier and
   Schumacker, 2013). Therefore, we next investigated the effect of bFGF
   on metabotropic and ionotropic receptors in our data. Two sets of
   (ionotropic) AMPA receptor subunits were found within the up-regulated
   LBP-M[CL]2-module (n = 12) and in the down-regulated LBP-M[CL]8 (n =
   10) module with Nptx2 (M[CL]2) and Gria1 or Gria2 (M[CL]8) as
   corresponding intramodular hubs. Ionotropic NMDA receptors were
   detected for the down-regulated LBP-M[CL]8 (n = 7) module with Grin2b
   as intramodular hub. A number of up-regulated metabotropic adrenergic
   receptors were found in the LBP-M[CL]2 (n = 20) module with Nedd4 as an
   intramodular hub and down-regulated dopamine receptors were found in
   the LBP-M[CL]8 (n = 6) module with Cnr1 as an intramodular hub
   ([84]Figure 4A). Limited LBP-related glutamate receptors were detected
   in the down-regulated LBP-M[EV]1 (n = 2) module from the EV fraction,
   whereas LBP-related adrenergic and opioid receptors were found in the
   up-regulated LBP-M[EV]3 module (n = 7 and n = 14, respectively)
   ([85]Figure 4B). The strength between the protein interactions was
   determined by computing a connectivity score; metabotropic adrenergic
   receptors and ionotropic glutamatergic AMPA receptors were strongly
   connected in the CL modules ([86]Figure 4C). In the EV modules, a high
   connectivity was observed for metabotropic opioid and adrenergic
   receptors ([87]Figure 4D). Jaccard similarity analysis enabled the
   identification of an up-regulated M[CL]2 module and M[EV]3 module as a
   common module, suggesting a linearized protein exodus as EV content for
   LBP-related receptors ([88]Figure 4E). Only the down-regulated
   LBP-M[CL]8 module (n = 11) has shown ion channel enrichment with
   slightly less down-regulation of Scn2a as intramodular hub
   ([89]Figure 4F). Taken together, our analyses suggest that
   predominantly metabotropic receptor-associated molecules are pledged in
   response to treatment with bFGF along with a small subset of ionotropic
   glutamate receptors.

Figure 4.

   [90]Figure 4
   [91]Open in a new tab

   Association of Ionotropic, Metabotropic Receptors and Ion Channel in
   Lewy Body Pathology (LBP) Modules

   (A and B) The module plots show CL and EV LBP ionotropic and
   metabotropic receptor and their interacting partners. Note:
   Edges/interactions between the proteins (gene symbols corresponding to
   proteins are used as labels) represent the correlation between them.

   (C and D) Connection strength of ionotropic, metabotropic receptors
   (CL, EV) modules analyzed using connectivity score represented as a
   pairwise matrix.

   (E) Pairwise similarity of ionotropic, metabotropic receptors between
   the CL and EV LBP modules was evaluated using Jaccard similarity
   co-efficient.

   (F) The module view of the CL LBP ion channel and its interactions.

bFGF Influences the Molecular Assembly of α-Syn-Interaction in LBP

   Among the detected ion channel receptors and channels, we found Nedd4
   and Tln1 as key α-Syn interaction partners. Because of the α-Syn role
   in LBP the dataset was examined for additional α-Syn interaction
   partners ([92]Figure 5A). We identified α-Syn interacting proteins in
   our LBP modules using information reported in [93]Khurana et al.
   (2017); this protruded to a α-Syn interaction network (pSIN)
   ([94]Figure 5B). There were 20 overlapping proteins in all modules of
   which 3 overlapped between EV and pSIN, 8 between CL-pSIN, 106 between
   EV-CL PD-modules, and 9 (Vdac1, Abca1, Rab6a, Nucb1, Aldh2, Nedd4,
   Tln1, Mapk1, and Rps14) among all subsets ([95]Figure 5B) ([96]Table
   S4). For a comprehensive sketch of these protein interactions, we
   further examined 50 immediate interactions of the overlapping 115
   proteins, obtaining an α-Syn interaction network using STRING database
   ([97]Figures S5A and S5B). In order to further examine the interactome
   between LBP and bFGF signaling, we developed an overall composite
   module (CM) among LBP-associated receptors, ion channels, RBPs, α-Syn
   interacting partners, and key players in the EV-CL modules
   ([98]Figure 5C). The CM was heterogeneously composed of both
   up-/down-regulated proteins and key proteins that were extracted from
   CM by estimating various centrality dimensions ([99]Figure S6). This
   comprehensive centrality analysis has concluded the
   shortest-path-betweenness-centrality as a key parameter. This enabled
   us to mine the top 5% informative proteins, including Scrn1, Slc6a11,
   Slc1a2, Tnr (down-regulated) and Rps14, Vim, Slc44a1 (up-regulated)
   ([100]Figure 5D). The functional enrichment of the composite module was
   further characterized by pathway enrichment analysis ([101]Figure 5E).
   Based on these results, we concluded that these proteins represent key
   components of the molecular interface of bFGF-signaling and LBP.

Figure 5.

   [102]Figure 5
   [103]Open in a new tab

   Amelioration of Alpha-Synuclein (α-Syn) Protein and Construction of
   Composite LBP Module

   (A) Outline describing the workflow for generating common modules.

   (B) Venn diagram showing the overlap between the LBP module proteins
   from CL, EV, and alpha-synuclein protein interacting partners.

   (C) A composite LBP co-expression module (CM) illustrating the
   interactions between the common LBP proteins from common CL, EV,
   alpha-synuclein protein interacting partners and receptors, ion
   channel, RNA-binding proteins. See also [104]Figure S6.

   (D) The bar plot shows the top 5% of informative proteins from the
   composite module based on the shortest path betweenness centrality.

   (E) The statistically significant (negative log[10] p value ≤ 0.05,
   hypergeometric test from METASCAPE) biological processes enrichment of
   the composite LBP module. Note: Red and blue color denote up- and
   down-regulation, respectively. Gene symbols related to proteins are
   used as labels in the CM plot.

bFGF-Enhanced EV-Release Possibly Supplements LBP Progression

   Because of the sequential progression of pathology to defined brain
   regions in LBP, we finally mapped the presence of LBP module proteins
   to major brain regions suggested in [105]Sharma et al. (2015)
   ([106]Figures 6A and [107]S7). LBP modules from CL and EV were
   appreciated in brain regions like the optic nerve, cerebellum, corpus
   callosum, olfactory lobe, brain stem and hippocampus. Conversely, we
   did not find a significant amount of proteins matched for the
   prefrontal cortex, striatum, and thalamus. We attributed this to sample
   collection from hippocampal neurons and, possibly, to variations in rat
   and mouse proteomic homologs. Within LBP proteins of various brain
   regions, we examined the coherent candidates in CL-EV modules
   ([108]Figure 6B). For the cerebellum, we found a total number of 30
   proteins in both CL-EV LBP-modules, and among these, two were shared
   with an overall partaking candidature of 6.67% for the designated brain
   region. Similarly, in the corpus callosum (25 proteins; 4 shared in the
   EV-CL PD-modules), optic nerve (72 proteins; 4 shared), hippocampus (23
   proteins; 2 shared), and olfactory bulb (13 proteins; 2 shared) have
   earned an overall candidature of 16%, 5.56%, 8.7%, and 15.38%
   respectively. We made use of these shared candidates ([109]Table S5) to
   develop a composite protein interaction module (CPIM) along with the
   α-Syn interacting partners and other common PD proteins of EV-CL
   ([110]Figures S8A and S8B). By assessing the centrality parameters
   ([111]Figure S8C) we identified the top 5% informative proteins of CPIM
   based on shortest path betweenness centrality. Scrn1, Slc6a11, Slc1a2,
   and Tnr were among the top down-regulated and Slc44a1, Rps14 were the
   most informative up-regulated proteins in the CPIM ([112]Figure S8D).
   Submodular analysis revealed Aldh2, Mcat, Abca1, Ppp3ca, Slc32a1,
   Mapk1, and Prdx2 as top α-Syn interacting proteins, supporting a
   bilateral interaction with the most informative top down-regulated
   proteins identified in CPIM and Atxn2, Gmps, Nedd4 as α-Syn-related
   interactions with top most informative up-regulated proteins
   ([113]Figures 6C and [114]S8E). We counter-validated the implications
   of bFGF-induced expression in our study and found up-regulated modules
   M[CL]2 and M[EV]3 significantly enriched for pathology associated
   hub-molecules identified in postmortem brain tissues of patients with
   LB disease ([115]Figures S9A and S9B). Taken together, these results
   imply that, in a disease state an EV-mediated cross talk among various
   brain regions is supported from these results, which could involve
   transport of misfolded proteins via EVs and can affect the protein
   metabolism beyond the host at recipient brain region eventually
   claiming shutdown of neuronal molecular mechanisms and
   neurodegeneration.

Figure 6.

   [116]Figure 6
   [117]Open in a new tab

   Influence of α-Syn Protein Interactions in Brain-Specific Regions

   (A) Schematic representation of the workflow to generate brain
   region-specific LBP modules.

   (B) Doughnut plot shows the percentages of common brain region-specific
   proteins between the CL and EV LBP modules. See also [118]Figure S7.

   (C) The module characterizes interactions between top 5% informative
   proteins of common brain region-specific module and their interactions
   with alpha-synuclein protein interacting partners. Note: Gene symbols
   associated to proteins are used as labels in the modules.

   See also [119]Figure S8.

Discussion

   Here we performed a global proteomic evaluation of bFGF effect induced
   to the CL and EV fractions from cultured hippocampal neurons and
   reasoned out its relevance in LBP ([120]Figure 1). Instead of
   individual protein analysis, we rather applied a system-level
   comprehensive approach to examine bFGF-regulated DEPs using WPCNA
   ([121]Figure 2). Because of the role of neurotrophic factors in
   hippocampal pathology, we screened our WPCNA-derived modules for
   proteins linked to LBP-associated molecular changes. This strategy
   enabled us to capture LBP modules ([122]Figures 3A and [123]Tables S3A
   and S3B) and allowed intersecting bFGF- with LBP-related proteome
   changes. Our data identified n = 532, bFGF-induced DEPs associated to
   LBP molecular changes thus revealing a molecular network of
   LBP-associated proteins and their modulation by bFGF in hippocampal
   neurons. Our results will therefore support the investigation of
   neurotrophic signaling in LBP pathology onset and progression.

   The top 5% up-regulated α-Syn interacting proteins shown in our results
   were Rps14, Vim, and Slc44a1 ([124]Figure 5D). Ribosomal protein Rps14
   has been reported to be differentially regulated in mitochondria of
   neural stem cells from patients with PD; under ribosomal stress this
   protein contributes to mitochondrial fragmentation ([125]Zhou et al.,
   2015). Similar to PD-GWASs ([126]Iwaki et al., 2019) our analysis
   confirms the up-regulation of the Slc44a1 (choline transporter-like
   protein-1) in response to bFGF treatment. It is shown that Slc44a1
   might be involved in mitochondrial energy metabolism ([127]Michel and
   Bakovic, 2009). Remarkably, a decrease of choline uptake reported in
   old adults ([128]Cohen et al., 1995) and in α-Syn overexpressing
   animals ([129]Wassouf et al., 2019) further establishes the association
   of bFGF-induced proteomic changes to LBP. Finally, up-regulation of Vim
   in response to bFGF is correlated with PD ([130]van den Berge et al.,
   2012). Scrn1, Slc6a11, Slc1a2, and Tnr were among the top
   down-regulated CPIM proteins. In line with its role in PD, Scrn1 is
   another α-Syn interaction partner involved in synaptic vesicle
   recycling, ER modulation, and calcium homeostasis ([131]Lindhout
   et al., 2019). Scrn1 interactions with the vesicle-associated membrane
   protein (VAMP)-associated protein (VAP) and association of VAP low
   levels in PD further suggest a pathological relevance for Scrn1
   ([132]Murphy and Levine, 2016). In summary, these proteins are strongly
   associated with PD and their modulation by bFGF may provide more
   molecular substrates for the effect of bFGF in LBP. Slc1a2 has been
   investigated for polymorphic associations with PD ([133]Appenzeller
   et al., 2013). Slc6a11 has been demonstrated to modulate basal ganglia
   neuronal networks ([134]Chazalon et al., 2018), and Tnr may likewise
   contribute to PD pathology ([135]Tsai et al., 2014). In conclusion,
   these results will provide the molecular basis for further studies to
   address the role of bFGF-induced proteomic changes in LBP.

   In addition to the CPIM-derived top 5%, numerous additional candidate
   genes from our data have a potential pathogenic relevance for LBP. For
   instance, our data demonstrated an up-regulation of Nptx2
   ([136]Figure 4A). This protein is involved in excitatory synapse
   formation. It also plays a role in clustering of AMPA-type glutamate
   receptors at established synapses, resulting in non-apoptotic cell
   death of dopaminergic nerve cells. In accord with the relevance of
   glutamate receptors, we found a down-regulation of AMPA (Gria1 and
   Gria2) and NMDA (Grin2b) receptors, which is associated with the
   accelerated aging of neurons ([137]Dryanovski et al., 2013). Likewise,
   we found a down-regulation of the glutamate transporter 2 gene Slc1a2,
   possibly further contributing to a dysregulated glutamate metabolism.
   Notably, Nptx2 has been found to be present in LBs of patients with PD
   ([138]Moran et al., 2008), possibly implicating bFGF signaling in LB
   formation. Another example comes from the alterations of the mTOR and
   growth factor signaling pathways ([139]Lynch-Day et al., 2012) and
   impaired autophagy in LBP. Earlier, we confirmed the role of
   Nedd4-family E3 ligases in mTOR signaling ([140]Hsia et al., 2014) and
   it has been reported that the ubiquitination of Nedd4 by E3-ligases
   controls autophagy mechanisms ([141]Sun et al., 2017) and activation of
   the inflammasome ([142]Liu et al., 2019). Up-regulation of Nedd4 in our
   data ([143]Figures 4A and 4B) thus implicates the role of bFGF in
   modulating autophagy during LBP. In accord, Nedd4 ubiquitination has
   been shown to suppress autophagy in neuroblastoma cells ([144]Xu
   et al., 2017) possibly impairing neuronal clearance mechanisms.
   Furthermore, Nedd4 ligases rapidly promote ubiquitination of α-Syn
   ([145]Mund et al., 2018), further supporting a role of bFGF in
   α-Syn-mediated pathogenic mechanisms. In concurrence to such
   hypothesis, we observed new and formerly unknown interactions of Tln1
   (up-regulated) with Nedd4 and Nptx2 ([146]Figure 4A). Tln1 is known as
   integrin-associated cytoskeletal protein and has binding sites for
   other cytoskeletal proteins such as α-Synemin, helping them to unfold.
   In addition, Tln1 regulates functions like cell proliferation,
   survival, and migration ([147]Roberts and Critchley, 2009). The
   upstream and downstream status of these proteins requires a further
   validation and may support the understanding of protein misfolding
   mechanisms ([148]Moran et al., 2008).

   Interestingly, many of these LBP-associated proteins exhibited a
   similar behavior in the CL and EV fraction, i.e., up-regulated CL
   proteins were up-regulated in the EV fraction and vice versa. We
   attribute this finding to the mechanism of EV formation during the
   generation of multivesicular bodies (MVBs), where highly abundant
   proteins are more likely to be excreted by EVs ([149]Gruenberg et al.,
   1989). In accord with an EV-mediated transmission of LBP-associated
   proteins, our data suggest a high representation of the top 5% proteins
   in CPIM and CM for olfactory bulb (OB) and corpus callosum (CC)
   ([150]Figures 6B and 6C) ([151]Zapiec et al., 2017; [152]Niu et al.,
   2018), suggesting a connection of the pathogenic molecular network
   between these regions. In accord, the OB is considered as an entry for
   environmental pathogens that may induce LBP changes in the OB that are
   then transmitted to central brain areas, as an occurrence via EVs
   ([153]Braak and Del Tredici, 2009). Along these lines, the loss of CC
   volume in patients with PD has been associated with cognitive
   impairment ([154]Goldman et al., 2017).

   Taken together, our study identifies a number of proteomic network
   changes and individual protein candidates that result from growth
   factor signaling and their potential contribution to LBP. Future
   experimental studies should investigate these candidates and assess
   their contribution to Lewy body-associated pathological changes
   in vitro and in vivo. This will ultimately strengthen the connection
   between growth factor signaling and LBP and allow assessment of growth
   factor signaling as a potential therapeutic target in LBP. In summary,
   these results will support the examination of bFGF for modulating
   disease progression in LBP and its cell-to-cell spreading, which has
   become a common theme for understanding disease progression in
   neurodegenerative conditions.

Limitations of the Study

   A limitation of our study is the absence of wet-lab experimental data
   from an LBP disease model. However, there is currently no rat model
   that replicates Lewy pathology well in its complexity ([155]Rockenstein
   et al., 2002; [156]Hashimoto et al., 2003). Because our earlier results
   on the effect of bFGF ([157]Kumar et al., 2020b) were derived from rat
   neurons, using more abundant mouse models would add additional bias.
   Taken together, we thus believe that intersecting “real-world” human
   proteomic alterations derived from databases with the effects of growth
   factor treatment in healthy rat primary neurons allows for a less
   confounded investigation of its true cellular effects. Future studies
   should validate our results in experimental disease model in vitro and
   in vivo as well as in postmortem material from human cases.

Resource Availability

Lead Contact

   Any questions or requests should be addressed to the Lead Contact
   ([158]thomas.koeglsperger@dzne.de).

Materials Availability

   We used the proteome data as mentioned below. This study did not
   generate new reagents.

Data and Code Availability

   We deposited the proteome datasets in the ProteomeXchange Consortium
   via the PRIDE partner repository, CL dataset identifier PXD015969 and
   EV dataset identifier PXD014401. We used published algorithms without
   generating new computer code.

Methods

   All methods can be found in the accompanying [159]Transparent Methods
   supplemental file.

Acknowledgments