Identification of phosphorylated tau protein interactors in progressive supranuclear palsy (PSP) reveals networks involved in protein degradation, stress response, cytoskeletal dynamics, metabolic processes, and neurotransmission

Abstract Progressive supranuclear palsy (PSP) is a late‐onset neurodegenerative disease defined pathologically by the presence of insoluble phosphorylated‐Tau (p‐Tau) in neurons and glia. Identifying co‐aggregating proteins within p‐Tau inclusions may reveal important insights into processes affected by the aggregation of Tau. We used a proteomic approach, which combines antibody‐mediated biotinylation and mass spectrometry (MS) to identify proteins proximal to p‐Tau in PSP. Using this proof‐of‐concept workflow for identifying interacting proteins of interest, we characterized proteins proximal to p‐Tau in PSP cases, identifying >84% of previously identified interaction partners of Tau and known modifiers of Tau aggregation, while 19 novel proteins not previously found associated with Tau were identified. Furthermore, our data also identified confidently assigned phosphorylation sites that have been previously reported on p‐Tau. Additionally, using ingenuity pathway analysis (IPA) and human RNA‐seq datasets, we identified proteins previously associated with neurological disorders and pathways involved in protein degradation, stress responses, cytoskeletal dynamics, metabolism, and neurotransmission. Together, our study demonstrates the utility of biotinylation by antibody recognition (BAR) approach to answer a fundamental question to rapidly identify proteins in proximity to p‐Tau from post‐mortem tissue. The application of this workflow opens up the opportunity to identify novel protein targets to give us insight into the biological process at the onset and progression of tauopathies.


| INTRODUC TI ON
Progressive supranuclear palsy (PSP) is a late-onset, fatal neurodegenerative syndrome with heterogeneous clinical manifestations . Because of this heterogeneous phenotype, definitive diagnosis of PSP can only be confirmed at autopsy by the pathological presentation of neurodegeneration and specific hyperphosphorylated, four-repeat (4R) Tau in neurons, and glia in anatomical regions corresponding to primary clinical symptoms (Kovacs et al., 2020). Because of the presence of pathological Tau, PSP is classified as a Tauopathy along with other neurodegenerative conditions such as corticobasal degeneration (CBD), Pick's disease (PiD), and Alzheimer's disease (AD) (Kovacs, 2015). However, despite the common Tau neuropathology, numerous studies have shown that Tau is processed distinctly in each of these conditions, which is reflected in differences in inclusion morphology, biochemical profile, and ultrastructure among the Tauopathies including PSP (Arakhamia et al., 2020;Falcon et al., 2019;Falcon, Zhang, Schweighauser, et al., 2018;Shi et al., 2021;Taniguchi-Watanabe et al., 2016). This is also supported by comparative interactome studies of human and rodent tau protein networks that show distinct differences in their aggresome profiles (Kavanagh et al., 2022), which potentially makes studying cellular mechanisms of the human condition difficult to recapitulate in rodent models.
Further highlighting the causative role of Tau in disease pathogenesis, mutations to the gene encoding Tau (MAPT) can result in clinical and neuropathological PSP (Chen et al., 2019;Forrest et al., 2018;Fujioka et al., 2015;Im et al., 2015). Moreover, genomewide association studies have identified variants in the MAPT locus in conferring the most significant genetic risk associated with developing PSP (Höglinger et al., 2011). Collectively, neuropathology, biochemistry, and genetics strongly link altered biochemical forms of Tau to PSP pathogenesis. Several studies have performed biochemical extraction of phosphorylated Tau (p-Tau) from PSP tissue to identify the specific features of insoluble pathogenic Tau. These studies have identified distinct structural changes in Tau such as aggregation, fibrillization, and post-translational modifications of which phosphorylation is most prominent (Taniguchi-Watanabe et al., 2016;Wray et al., 2008). Tau is ordinarily a highly phosphorylated protein, and disease-specific phosphosites have been identified (Arendt et al., 2016).
In addition to post-translational modifications of Tau, identifying proteins in proximity to Tau has provided further insights into disease progression and pathogenesis that are common in Tauopathies.
For example, several nuclear speckle proteins have been found to co-localize with p-Tau inclusions in tauopathies, including AD, frontotemporal dementia, and CBD, providing links between Tauopathies and defects in RNA processing (Lester et al., 2021). A classical approach used to identify Tau inclusion components from post-mortem brain tissue involves enriching the detergent-soluble aggregates from post-mortem brain tissue using biochemical fractionation methods, after which the detergent-insoluble proteins are detected using mass spectrometry (MS). While this approach has provided valuable insight into the detergent-insoluble aggregate composition, technical limitations include a high degree of non-specific detection and loss of proteins through harsh biochemical processing.
Another classical method is to co-immunoprecipitate Tau with interaction partners from the soluble and sarkosyl-insoluble fractions of post-mortem tissue. However, this relies on the stable interaction being maintained between Tau and interaction partners through the disruptive extraction process (Julien et al., 2012). Alternate techniques such as immunohistochemistry (IHC), biotinylation (Prikas et al., 2022), and proximity-ligation detection methods provide precise in situ labeling of individual protein components but are limited by antibody combination limits (Duraiyan et al., 2012;Söderberg et al., 2008).
Biotinylation by antibody recognition (BAR) is a recently described proximity-ligation method that is particularly well-suited for analyzing formalin-fixed, post-mortem tissue from patients (Bar et al., 2018). BAR uses a primary antibody that recognizes the target of interest in fixed samples. A secondary antibody conjugated to horseradish peroxidase (HRP) recognizes the primary antibody and, with the addition of biotin phenol and hydrogen peroxide, facilitates the rapid conjugation of biotin onto proteins within the vicinity of the antibody complex. In contrast to classical biochemical extractions, BAR enables labeling of the endogenous, aggregating proteins before tissue homogenization. Therefore, even if the aggregate is disrupted during the subsequent extraction process, the BAR-labeled proteins that have been separated from the aggregate can still be identified. Thus, this method significantly enhances the likelihood of identifying proteins within protein aggregates, including low-abundance proteins.
Proximity-ligation methods such as BAR, proximity-dependent biotin identification (BioID), and an engineered ascorbate peroxidase method (APEX) are all based on the covalent attachment of biotin to proteins in proximity to a protein of interest via enzymatic activity after which the biotinylated proteins are isolated using streptavidinconjugated beads and identified by MS (Hedl et al., 2019;Lobingier et al., 2017;Roux et al., 2012). Notably, these methods provide additional advantages in studying insoluble protein inclusions because of the strong interaction between biotin and streptavidin, which can withstand harsh extraction conditions, including buffers containing Grant/Award Number: AV20200003; Australian Research Council, Grant/Award Number: DP210103469

K E Y W O R D S
Biotinylation, mass spectrometry, neuropathology, progressive supranuclear palsy, tau, Tauopathy ionic detergents and chaotropic reagents such as sodium dodecyl sulfate (SDS) and urea. However, the fundamental constraints of other proximity-ligation methods, such as BioID and APEX, are that they must be performed in living cells and require genetic alteration or transfection. As a result, BAR represents a novel method for analyzing proteinprotein interactions within patient post-mortem tissue samples.
In order to answer one of the most fundamental questions in tau pathology and proteomics, as a proof-of-concept, we have applied BAR followed by MS to identify proteins that are proximal to phosphorylated-Tau (p-Tau) inclusions in post-mortem cases of PSP.
We demonstrate the utility of this method to rapidly identify proteins beyond Tau that aggregate in the neurons and glia of PSP patients.
Notably, we were able to identify previously reported Tau phosphorylation sites as well as p-Tau interaction partners. Furthermore, analysis of the p-Tau-aggregate interactome also revealed multiple proteins in proximity to p-Tau, which may reveal further insights into the molecular pathways that may be affected in PSP.

| Post-mortem tissue
As a proof-of-concept to apply the BAR experimental approach to characterize the p-Tau aggresome, we have used formalin-fixed post-mortem tissue blocks limited to 4 cases with PSP (Table 1) were obtained from the South Australian Brain Bank under ethics approvals MSC/16/11/HREC and HREA 5201600387. An exclusion post-mortem interval (PMI) criteria of >24 were used to select tissue samples that had minimal amounts of protein degradation (Blair et al., 2016). The blocked motor cortex of 4 PSP patients was cut to fit within a 12 mm 2 template before being serially sectioned on a Leica VTS1200 vibratome at 50-85 μm into 24-well plates. Individual sections were weighed before subsequent immunohistochemistry, imaging, or proteomics. As this was a proof-of-concept study to identify p-Tau proximal proteins using BAR, blinding and sample power calculations were not performed for this study. Other proofof-concept studies that have developed new techniques for tissue proteomics have used similar small sample sizes (Blair et al., 2016;Drummond et al., 2015).

| Biotin-streptavidin pull-downs
In order to isolate biotinylated proteins, biotinylated tissue was first homogenized using a hand-held douncer in BAR lysis buffer (1% SDS (Sigma, Catalog number: L3771), 1% sodium deoxycholate (Merck, Catalog number: D6750)). After homogenization, tissue was TA B L E 1 Details of patient post-mortem tissue used for study.

| Immunoblotting
Equal amounts of protein were separated on a 4-15% gradient CLx Imaging System at the appropriate wavelengths. Images were analyzed using Image Studio Lite Software (RRID: SCR_002285).

| Statistical analysis of protein identifications
The MS data were further processed using a Label-Free Quantitation workflow featuring the peptide spectrum match (PSM) Minora Feature node. This made use of the percolator node, enabling an estimation of false discovery rates (FDR) at the protein and PSM levels.
Protein identifications were validated employing a q-value of 0.01.
Label-free quantitation (LFQ) using intensity-based quantification was carried out. LFQ was carried out according to default parameter settings in the Proteome Discoverer 2.4 Software (Thermo). Briefly, peptide spectral matches (PSM) were filtered using a maximum delta Cn of 0.05, rank of 0, and delta mass of 0 ppm. PSMs and peptides were, respectively, validated using a strict FDR for PSMs of 0.01 and 0.05 for a relaxed FDR. Peptides shorter than 6 amino acids were filtered out. PSMs were chromatographically aligned for each input file in a sample set with a mass tolerance of 10 ppm and a maximum retention time (RT) shift of 10 min. Peptide groups used for protein quantification were analyzed using the default parameters which set a peptide as unique if it is included in only one protein group.
The quantification was processed using unique and razor peptides (peptides shared among multiple proteins group or proteins) with the precursor abundance based on the intensity. Protein abundance was calculated as a sum of the individual peptide group abundances, and the ratio was based on pairwise ratio using a geometric median of the peptide group ratios. An ANOVA test was used for the hypothesis test and uses the background population of ratios for all peptides and proteins to determine whether any given single peptide or protein is significantly changing relative to that background (Proteome Discoverer User Guide Software Version 2.4, Thermo).

| Bioinformatics analysis
Ingenuity Pathway Analysis (IPA) (Qiagen, RRID: SCR_008653) was used to analyze the MS results. "Diseases and Disorders" function was employed to create pathways and interaction for "Neurological Disease" and "Molecular and Cellular Functions" if they contained 10 or more proteins found associated with p-Tau from PSP samples were present. Canonical pathway analysis was employed to find pathways for multiple proteins associated with p-Tau from PSP samples.

| Statistical analysis
For immunoblot analysis, GraphPad Prism (Version 9.4.1, RRID: SCR_002798) was used to determine significance. A one-tailed ttest was used. Results were considered statistically significant if the p-value was <0.05. Data were not assessed for normality, and no test for outliers was conducted.
For bioinformatics analysis, ingenuity pathway analysis was used to determine significance. Specifically, a right-tailed Fisher's exact test was used and results were considered statistically significant if the p-value was <0.05. Data were not assessed for normality, and no test for outliers was conducted.

| Comparison of BAR-identified tau proximal proteins to literature and the human protein atlas
Proximal proteins to p-Tau that were identified in our study were compared to published studies of the Tau interactome in mouse and human brain samples. Published studies using mouse/rat brains

| Enhanced tissue expression and subcellular location
In order to assess if the proteins found associated with p-Tau had enhanced tissue expression in the brain were compared to normalized Tissue-Specific RNA expression obtained from V20.1 of the Human protein atlas (RRID:SCR_006710) (Atlas). To understand the normal subcellular distribution of proteins found proximal to p-Tau in PSP, we compared our significantly enriched list to the subcellular location of these proteins also from V20.1 of the Human protein atlas (Atlas; Thul Peter et al., 2017). Only proteins with an annotation of "approved," "enhanced," or "supported" under the "Reliability (IF)" category were included for subcellular location analysis. The cell graphic used in Figure 7b was also retrieved from the Human Protein Atlas.

| Enhanced tissue expression and subcellular location
In order to assess if the proteins found associated with p-Tau had enhanced tissue expression in the brain were compared to normalized tissue-specific RNA expression obtained from V20.1 of the Human protein atlas (Atlas; Uhlén et al., 2015). To understand the normal subcellular distribution of proteins found proximal to p-Tau in PSP, we compared our significantly enriched list to the subcellular location of these proteins also from V20.1 of the Human protein atlas (Atlas; Thul Peter et al., 2017). Only proteins with a "Reliability (IF)" annotation of "approved," "enhanced," or "supported" were included for subcellular location analysis.

| Cell type enrichment
Genes encoding for the proteins identified using BAR were assigned cell-type specificity based on transcriptomic datasets from the human temporal and motor cortex (Bakken et al., 2020;Zhang et al., 2016). Each gene's mean expression level was standardized across the different cell types, and genes were classified as enriched to a particular cell type with a Z-score 1.5 and above. Hierarchical clustering analysis with heatmap representation was performed on the standardized datasets in R using "heatmaply" (Galili et al., 2018).

| RE SULTS
3.1 | BAR selectively biotinylated proteins closely associated with p-tau in post-mortem tissue from PSP patients In order to selectively biotinylate proteins in a small radius to phosphorylated Tau in PSP patient post-mortem tissue, we first targeted phosphorylated forms of Tau using a primary mouse antibody that recognizes p-Tau at serine 202 (S202) and threonine 205 (T205).
p-Tau pathology has previously been extensively characterized in post-mortem tissue from PSP patients using the AT8 antibody (Goedert et al., 1995;Malia et al., 2016;Pikkarainen et al., 2010;Williams et al., 2007). Following wash steps, an anti-mouse secondary antibody conjugated to poly-HRP was used to bind the primary antibody. Next, biotin-XX-tyramide and hydrogen peroxide were added, enabling the biotinylation of proximal proteins by poly-HRP ( Figure 1a). In the negative control group, tissues were processed in the same way except without the addition of the primary antibody.
The comparison between the Tau

| BAR identifies pathological forms of tau in PSP patients
In order to identify proximal proteins labeled with biotin, the samples were further processed, and antibody-targeted Tau was isolated from the complex tissue ( Figure 2a). To do this, labeled samples (both control and Tau-targeted samples) were homogenized and then reverse cross-linked. The resulting lysates were used to determine whether the biotinylated proteins could be detected by immunoblotting ( Figure 2b). Notably, immunoblot analysis of the patient lysates, with streptavidin-conjugated to a fluorophore, revealed an apparent increase in the levels of biotinylated proteins when compared to the controls ( Figure 2c).
To isolate biotinylated proteins, equal amounts of patient samples were incubated with magnetic beads conjugated to streptavidin.
The beads were subjected to stringent wash procedures which involved sequential wash steps using high-salt, high-pH, and high-urea buffers to reduce non-specific binding (Hung et al., 2016). Finally, the isolated proteins were subject to on-bead trypsin digestion and detected using LC-MS/MS, followed by bioinformatic analyses. To exclude false negatives (non-specific binding to the beads), we only assessed proteins with at least two unique peptides identified and increased at least 2-fold compared to the negative control. In total, 117 proteins fit this criterion. To validate our approach in identifying components associated with p-Tau, we first checked explicitly for amino acid sequences belonging to Tau. As we would expect in tau-related PSP tissue, we noted an enrichment of Tau peptide sequences compared to the control ( Figure 2d). These data clearly indicate the selectivity of this approach.
The MS analysis further allowed us to screen the identified peptides for potential phosphorylation sites. Notably, phosphorylation sites on p-Tau were found at T403/S404 and T181 ( Figure S1) and at T217 (Figure 2e and Figure S1). Notably, all phosphorylation sites differed from the phosphosites recognized by the antibody used (AT8) (Figure 2e).

| BAR-MS identifies known and novel interaction partners of tau
We next compared our putative list of proximal and/or interacting   Table 2). These data demonstrate the ability and application of our BAR-optimized approach to identifying likely interaction partners of Tau.

| Bioinformatic analysis of proteins in proximity to tau is enriched for cellular assembly and organization
In order to gain further insight into the functionality of proteins in proximity to p-Tau, we analyzed the 117 BAR-MS identified proteins (stringent criteria of ≥2 fold-change compared to the control and 2 or more unique peptides) using IPA. Notably, the most significant "Diseases and Disorders" function associ- Residues highlighted in blue represent the phosphorylation sites recognized by the p-Tau antibody used. Residues in black boxes refer to phosphorylation sites that were unambiguously assigned. Residues in gray boxes (and with an * above) refer to phosphorylation sites that were assigned to one of two closely positioned amino acids. PSMs -peptide spectrum matches. A 0.01 or 100-fold abundance respectively indicates the absence or presence of the proximal protein.
associated with PSP, such as dyskinesia (1.01E-15) and cognitive impairment (2.06E-3) (Figure 3), also indicated enrichment of proteins known to associate with Tau and with involvement in neurological dysfunction.
Notably, the most statistically significant molecular and cellular functions identified by IPA were "Cellular Assembly and Organization" (59 Molecules, p-value range = 3.02 E-03 -3.54 E-11). Accordingly, we further analyzed the proteins associated with TA B L E 2 Proteins in proximity to p-Tau using Bar-MS with at least five additional MS-based published datasets. "Additional validation" refers to potential interaction partners with Tau that orthogonal methods have validated. "Insoluble in disease states" indicates previously identified proteins in insoluble fractions in neurodegenerative disease states or animal models of Tauopathies. Proteins not found in any MS datasets or previously described as a Tau interactor are italicized. A 0.01 or 100-fold abundance respectively indicates the absence or presence of the proximal protein.

Gene
Abundance ratio (tau-specific/ control) MS datasets Additional validation Insoluble in disease states

TA B L E 2 (Continued)
indicating that our proximity-labeling method was reliable for identifying p-Tau alongside major complexes (and associated proteins) known to interact with Tau.
To gain insight into the protein network relationships in proximity to p-Tau, IPA was used to identify networks of proteins with known relationships. Notably, Tau was found in a network Next, the protein network was used to analyze known proteinprotein interactions (Figure 5c). This analysis revealed 12 proteins/ protein complexes interacting with Tau (CRYAB, HSPA1A/HSPA1B,

F I G U R E 3
Proteins in proximity to Tau are enriched for known Tau interactions and proteins linked to neurological dysfunction. Groups of at least 10 proteins found in proximity to p-Tau (ratio p-Tau/control ≥2-fold) that are known to be associated with neurological disorders -Frontotemporal degeneration spectrum disorder, Parkinson's disease, parkinsonism, Alzheimer disease, Huntington Disease, and disorder of the basal ganglia in red or neurological signs associated with PSP such as dyskinesia and cognitive impairment highlighted in gray. How Tau is related to these neurological dysfunctions and disorders is highlighted in green, while previously unreported proteins identified in close proximity to p-Tau from Figure 4a are highlighted in blue. Results were obtained from proteomic studies using tissue from 4 different patients.
Of the known interaction partners, further analysis of each proteinprotein interaction using IPA revealed a subset of proteins/protein complexes that have other known relationships with Tau ( Figure 5c).
These included components of the HSP70 complex, which could be a binding regulator of Tau, inhibit Tau function, and influence the localization and ubiquitination of Tau. In addition, proteins that form part of the HSP90 complex were also identified, influencing the localization of Tau and acting as a Tau binding regulator. Another known interaction partner of Tau in the network was YWHAZ, which is known to act as a binding regulator of Tau.

| Proximal proteins to p-Tau in PSP display enhanced brain expression and multiple subcellular localizations
Next, we sought to identify if proteins found proximal to p-Tau in PSP showed enhanced expression in the brain compared to other F I G U R E 4 Grouping of proteins associated with Cellular Assembly and Organization (a) Groups of at least 10 proteins found in proximity to p-Tau (ratio p-Tau/control ≥2-fold) associated with Cellular Assembly and Organization subfunctions, highlighted in blue is further broken down in (b). (b) Individual proteins in proximity to p-Tau (ratio p-Tau/control ≥2-fold) were identified in a network associated with Microtubule dynamics. MAPT is highlighted in green, while a previously unreported protein identified proximal to p-Tau from Figure 2a is highlighted in blue. Results were obtained from proteomic studies using tissue from 4 different patients. tissue types. We used a bioinformatic approach to extract RNA-seq data generated in the Human Protein Atlas (Atlas; Uhlén et al., 2015).
We found that 28.2% of the 117 significantly p-Tau-associated proteins display enhanced brain expression relative to multiple tissues sampled, indicating that these proteins may have an essential role in brain function ( Figure 6a). Additionally, we utilized the Human Protein Atlas cell immunofluorescence data to observe the subcellular location of the proteins associated with p-Tau in PSP (Figure 6b).
Eighty-six proteins had sufficient data to be allocated subcellular locations, and Figure 6b shows that proteins associated with p-Tau are from multiple subcellular compartments despite p-Tau pathology in PSP residing overwhelmingly in the cytoplasm (Kovacs, 2015).

| Enriched cell-type expression of proteins proximal to p-Tau in PSP
Next, to determine if any of the proteins identified using BAR could be associated with specific cell types, we compared our list of significantly associated p-Tau proteins with two RNA expression datasets from the human temporal and motor cortices (Bakken et al., 2020;Zhang et al., 2016). Comparing our list of p- Tau OPCs, and 2 in endothelial cells (Table S1) and visually displayed by Figure 7c. Figure 7d shows that the variance is described mainly

| Bioinformatic analysis shows pathways enriched for protein degradation, stress responses, cytoskeletal dynamics, metabolism, and neurotransmission
Finally, to determine additional canonical pathways that could be affected by proteins in proximity to p-Tau, IPA was used to identify statistically enriched molecular canonical pathways. Notably, subsets of canonical pathways with related functions could be identified. These identified canonical pathways involved in protein degradation systems (including the proteasome and autophagy), stress responses (UPR and EIF2 signaling), cytoskeletal dynamics (including proteins which are involved in Actin-based motility and signaling, synaptogenesis and endocytosis signaling pathways), metabolic processes, and neurotransmission. These data indicate potential perturbations in protein degradation systems, stress responses, cytoskeletal dynamics, metabolism, and neurotransmission associated with p-Tau aggregation (Figure 8).

F I G U R E 7
Predicted cell-type enrichment of proteins found in proximity to p-Tau (ratio p-Tau/control ≥2-fold) based on standardized human RNA-seq datasets. (a) and (b) were constructed using an RNA-seq dataset with cells sampled from the human temporal cortex . (a) hierarchical cluster analysis and heatmap representation for transcriptional expression of proteins found proximal to p-Tau in neurons, astrocytes, oligodendrocytes, oligodendrocyte precursor cells (OPCs), microglia, and endothelial cells. Z-scores above 1.5 with a predicted enrichment are shown in green-yellow. (b) principal component analysis (PCA) shows that the predicted expression variance of neurons and astrocytes is primarily explained in the first and second principal components. (c) and (d) were constructed using an RNAseq dataset with cells sampled from the human motor cortex (Bakken, Jorstad et al. 2020). (c) hierarchical cluster analysis and heatmap representation for transcriptional expression of proteins found proximal to p-Tau in neurons, astrocytes, oligodendrocytes, oligodendrocyte precursor cells (OPCs), microglia, endothelial and vascular and leptomeningeal cells (VLMC). Z-scores above 1.5 with a predicted enrichment are shown in green-yellow. (d) principal component analysis (PCA) shows the predicted expression variance in the first and second dimensions are primarily explained by neurons and oligodendrocytes indicated by cos2. Results were obtained from proteomic studies using tissue from 4 different patients.
F I G U R E 8 Selection of statistically significant canonical pathways associated with proteins found in proximity to p-Tau (ratio p-Tau/ control ≥2-fold). Pathways highlighted in green represent pathways associated with protein degradation systems, while those highlighted in red, blue, orange, and purple are associated with stress responses, cytoskeletal dynamics, metabolic processes, and neurotransmission. Particular pathways predicted to be enriched in neurons, astrocytes (Astro), and oligodendrocytes (o) are also displayed. Results were obtained from proteomic studies using tissue from 4 different patients.
Together, IPA analysis of enriched "Molecular and Cellular Functions" along with "Canonical Pathways" and interaction networks of Tau revealed clusters of proteins that were associated with cellular organization systems, protein clearance, stress response pathways, metabolic processes, and neurotransmission providing further insight into the individual proteins and protein complexes in proximity to pathological forms of Tau from the post-mortem tissue of PSP patients. Additionally, cell enrichment showed that some of these pathways are predicted to be enriched in neurons, astrocytes, and oligodendrocytes based on RNA-seq data.

| DISCUSS ION
As a proof-of-concept study to determine proximal or interacting proteins to p-Tau in tissue pathology, we applied the BAR method and MS for unbiased, in situ identification of pathological p-Tau aggregates and co-aggregating protein partners in a small subset of PSP patient tissue. We identified 117 proteins of which >80% of the proteins have previously been associated with Tau while an extensive network of proteins has previously been associated with Tauopathies and known complexes associated with Tau. Furthermore, higher confidence p-Tau proximal proteins included previously associated with Tau in multiple publications.
These included 15 proteins that were found associated with Tau in at least four other MS-based publications. Notably, in addition to being identified in four MS datasets, the interaction between Tau and YWHAZ (14-3-3ζ) was also found by additional publications (Hashiguchi et al., 2000;Jaesun et al., 2004;Li & Paudel, 2016;Matthews & Johnson, 2005;Nellist et al., 2002;Papanikolopoulou et al., 2018;Qureshi et al., 2013;Sadik et al., 2009;Sluchanko et al., 2009;Tugaeva et al., 2017;Yuan et al., 2004). This demonstrates the evident selectivity of this method to enrich interaction partners of a selected protein of interest, in this instance, p-Tau from PSP cases.
In addition to identifying known interaction partners of Tau, we were also able to identify proteins that have additional relationships with Tau, some of which directly relate to features of Tau pathology, including aggregation, ubiquitination, and Tau association with microtubules. Proteins that form part of HSP90/70 complexes and YWHAG (14-3-3γ) were found to have additional relationships with Tau. Dou et al. (2003) demonstrated that reduced expression of HSP70 and HSP90 using siRNA leads to the accumulation of aggregated, phosphorylated Tau and reduced binding of Tau with microtubules. HSP70 proteins have been reported to increase inhibition of phosphorylated active Tau and have also been associated with Tau ubiquitination as HSP70 proteins and CHIP can increase ubiquitination of Tau (Carrettiero et al., 2009;Muchowski & Wacker, 2005). YWHAZ has been reported to bind and mediate the phosphorylation of Tau (Yuan et al., 2004). Together, the identification of known interaction partners of Tau and modifiers of Tau pathology demonstrates the ability of this method to characterize the web of protein interactions associated with p-Tau, including proteins that may have a role in altering the pathological features of Tau.

Additionally, with datasets obtained from the Human Protein
Atlas, we assessed that 28.21% of proteins associated with p-Tau in PSP displayed enhanced RNA expression in the brain and typically are associated with various subcellular locations despite p-Tau pathology residing overwhelmingly in the cytoplasm (Atlas; Kovacs, 2015;Orr et al., 2017;Thul Peter et al., 2017;Uhlén et al., 2015). The presence of proteins typically associated in subcellular locations other than the cytosol indicates recruitment or sequestration of some of these proteins warrants further investigation as modulating this phenomenon may influence the development of pathology and dysfunction in PSP. Using a similar approach, we sought to identify if proteins associated with p-Tau were enriched in particular cell types by comparing them to two RNA-seq datasets from the human temporal and motor cortices (Bakken et al., 2020;Y. Zhang et al., 2016).
We identified that the transcriptomic profile indicates that these proteins are enriched particularly in neurons, astrocytes, and mature oligodendrocytes, reflecting the cell type Tau pathology of NFTs, TAs, and CBs seen in PSP (Ahmed et al., 2013;Kovacs et al., 2020).
An excellent example of this is FTL, where it is predicted to be enriched in microglia from the temporal cortex dataset, which is in line with IHC studies in humans showing it as a marker of microglia; however, it has been shown to be up-regulated in astrocytes in PSP patients with a subset of TAs displaying colocalization, which is also an independent validation of our finding (Ebrahim et al., 2011;Kaneko et al., 1989). Interestingly, TAs have been positively associated with the expression of the microglial gene-enriched immune network, so further investigation if astrocytes are directly influencing this by expressing these genes or having a complex interaction with microglia to have an impact on TA pathology and PSP pathogenesis (Allen et al., 2018).
We gained further insight using IPA into the canonical pathways associated with proteins in proximity to p-Tau. Using this workflow, we consistently identified groups of proteins affecting protein degradation systems, stress responses, cytoskeletal dynamics, metabolic processes, and neurotransmission. Notably, the protein degradation system and unfolded protein response pathways were also identified in the network of protein-protein interactions that connected back to Tau, demonstrating the ability to identify networks of protein-protein interactions that link to Tau and the downstream processes associated with these networks. The ability to identify and characterize networks of proteins associated with p-Tau has essential implications for understanding neurodegenerative processes as it may help identify changes to molecular pathways that occur early during pathogenesis.
Furthermore, identifying proteins and pathways is vital as the complete cellular mechanisms involved in p-Tau-induced onset, cellular dysfunction, and cell death are not fully understood (Orr et al., 2017).
IPA of proteins associated with p-Tau in PSP found canonical pathways associated with protein degradation, stress responses, metabolism, and neurotransmission. Two stress response pathways, UPR and EIF2 signaling have been associated with PSP via GWAS. A key regulator of UPR and EIF2 signaling, EIF2AK3, was identified as a GWAS risk factor giving rise to a hypomorphic allele, leading to increased neuronal vulnerability to endoplasmic reticulum stress and Tau accumulation (Höglinger et al., 2011;Stutzbach et al., 2013;Yuan et al., 2018). One of the components of the UPR, also involved in phagosome maturation which we found to associate with p-Tau in PSP, is VCP. VCP has been shown to regulate Tau accumulation via its disaggregase activity, and a missense mutation (D395G) leading to decreased disaggregase activity has been identified in patients with frontotemporal degeneration displaying prominent Tau pathology and vacuolation. Thus, indicating modulating the recruitment of VCP to p-Tau pathology in PSP may be a way of decreasing Tau pathology (Darwich et al., 2020). Another protein we found associated with p-Tau in PSP involved in the phagosome maturation pathway is NSF. NSF has been linked with PSP and Tau pathology because of an NSF exonic polymorphism within the MAPT H1 haplotype being associated with PSP, CBD, and FTD risk (Pastor et al., 2004;Yokoyama et al., 2017). This exonic polymorphism in NSF has been associated with a more significant Tau pathology in PSP, particularly Tau threads (Allen et al., 2018). NSF is required for vesiclemediated transport required for phagosome maturation, synaptic vesicle release, and neurotransmission, indicating it may be involved in multiple pathways implicated in PSP (Rizo & Xu, 2015). Additionally, the influence of alterations to neurotransmission and metabolic processes is highlighted by our finding of MAOB associated with p-Tau pathology, which previously has not been associated with p-Tau interaction via MS. MAOB is a primary amine oxidase that is important for the catabolism of dopamine and is primarily expressed by astrocytes and a subpopulation of neurons in the central nervous system (Ekblom et al., 1993). Quantitative western blot analysis showed MAOB is increased in PSP in regions associated with degeneration (Tong et al., 2015(Tong et al., , 2017. A randomized, placebo-controlled Phase-III clinical trial into a MAOB inhibitor rasagiline was conducted in a PSP cohort (Nuebling et al., 2016). While the trial was underpowered and failed to see a significant influence on the primary endpoint, post hoc analysis indicated a beneficial effect on limb motor function, which may implicate MAOB on influencing specific symptoms seen in PSP.
Some advantages of the BAR method over traditional fractionation methods and immunoprecipitations are the ability to specifically label proteins (Killinger et al., 2022) within a small labeling radius to the protein-of-interest (such as p-Tau), followed by homogenization that uses comparatively less sample amount. Using vibratome-prepared tissue also enhances antigenicity and reduces the chance of molecular extraction compared with paraffin-embedded samples and a validated p-Tau antibody resistant to over fixation, which can occur in human post-mortem samples (Pikkarainen et al., 2010;Ramos-Vara, 2005).
However, the technique is amendable to use paraffin-embedded samples (Bar et al., 2018). This unique characteristic significantly increases the chances of identifying proteins entangled or within proximity of the pathological Tau, which is not possible with other isolation and proteomic approaches such as laser microdissection. In addition, the covalent attachment of biotin to these proteins means that the proteins can later be isolated in more harsh buffers enabling extraction of proteins that would not be soluble using standard immunoprecipitation buffers. The ability to identify protein-interaction partners from fixed post-mortem tissue also means that these can be evaluated using tissue precisely from multiple central nervous system regions (allowing further interrogation of the unique aggresomes occurring throughout the central nervous system). Therefore, this methodology is highly applicable to other neurodegenerative diseases characterized by the aggregation of proteins such as α-synuclein and TDP-43, which have well-characterized antibodies available.
A limitation of the BAR technique is that labeling resolution does not allow direct interactors of p-Tau to those of proximal proteins, so independent methods such as IP may be required for confirmation (Bar et al., 2018). However, the labeling radius can be reduced by using smaller identifying antibodies such as nanobodies or antibodies directly conjugated with HRP. We sought to overcome this with the bioinformatic analysis outlined above to gain greater insight into potential direct interactors with p-Tau in PSP. Given the availability of motor cor-  Zhang et al., 2020). Thus, identifying novel aggregate components in human tissue is an important step that will help understand molecular pathways altered in specific Tauopathies. Together, identifying the common and unique aggregate components may help to reveal further insight into the converging and diverging mechanisms that underpin Tauopathies and other closely related neurodegenerative diseases.
Future studies comparing different Tauopathies in multiple brain regions would help elucidate these conserved and unique components and subsequent pathways, revealing molecular drivers of pathology and potential therapeutic approaches targeting these components.
Overall, this approach will serve as a complementary method to classical fractionation and immunoprecipitation, IHC, and proximityligation methods and may assist in identifying aggregating proteins beyond Tau, which is particularly pertinent in numerous neurodegenerative conditions. This, in turn, will assist in identifying other proteins (and molecular functions) involved in PSP and other forms of neurodegeneration.

| CON CLUS IONS
Finding and validating aggregate components has thus far been a slow process, and in many instances, their identification has been fortuitous. The identification of these components, however, is integral to understanding the molecular underpinnings of disease. Here, we demonstrate an unbiased approach for accurately labeling proteins within a small radius of p-Tau directly from PSP patient postmortem tissue facilitating multiple proximal proteins to be quickly identified by MS. Bioinformatic analysis showed the multi-cellular contribution to PSP pathway and highlighted canonical pathways including protein degradation, stress responses, cytoskeletal dynamics, metabolic process, and neurotransmission being related to p-Tau. This workflow is useful as a complementary approach, for rapidly identifying aggregate components in multiple neurodegenerative diseases. It may also be used to evaluate the heterogeneity of Tau pathology that exists between patients and between diseases characterized by Tauopathies, a feature that can only be accurately evaluated using the post-mortem tissue from patients.

CO N FLI C T O F I NTER E S T S TATEM ENT
The authors declare no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The mass spectrometry data can be found on the ProteomeXchange Consortium via the PRIDE partner repository (Perez-Riverol, Csordas, et al. 2019). The dataset file identifier is PXD028237.
PhosphoSite data can be accessed with the following. Project accession: PXD028770. Project DOI: Not applicable.

A PPROVA L A N D CO N S ENT TO PA RTI CI PATE
Griffith University MSC/16/11/HREC, Macquarie University (HREA 5201600387).

I N FO R M ED CO N S ENT S TATEM ENT
Informed consent was obtained from all subjects involved in the study.