Mitochondrial Disruption by Amyloid Beta 42 Identified by Proteomics and Pathway Mapping

Alzheimer’s disease (AD) is marked by chronic neurodegeneration associated with the occurrence of plaques containing amyloid β (Aβ) proteins in various parts of the human brain. An increase in several Aβ fragments is well documented in patients with AD and anti-amyloid targeting is an emerging area of therapy. Soluble Aβ can bind to various cell surface and intracellular molecules with the pathogenic Aβ42 fragment leading to neurotoxicity. Here we examined the effect of Aβ42 on network adaptations in the proteome of nerve growth factor (NGF) differentiated PC12 cells using liquid-chromatography electrospray ionization mass spectrometry (LC-ESI MS/MS) proteomics. Whole-cell peptide mass fingerprinting was coupled to bioinformatic gene set enrichment analysis (GSEA) in order to identify differentially represented proteins and related gene ontology (GO) pathways within Aβ42 treated cells. Our results underscore a role for Aβ42 in disrupting proteome responses for signaling, bioenergetics, and morphology in mitochondria. These findings highlight the specific components of the mitochondrial response during Aβ42 neurotoxicity and suggest several new biomarkers for detection and surveillance of amyloid disease.


Introduction
Alzheimer's disease (AD) is the most common cause of human dementia and a major public health problem. Commonly prescribed drugs for treatment include acetylcholinesterase (AChE) inhibitors, which lessen cognitive symptoms without halting neurodegeneration [1]. A recently approved monoclonal antibody, aducanumab (Aduhelm), which specifically targets Aβ oligomers, appears to slow disease progression [2]. In clinical studies, however, aducanumab was also associated with transient swelling of the brain and hemorrhaging, thus increasing the urgency for a better understanding of the cellular effects of Aβ.
Amyloid plaques and neurofibrillary tangles (NFT) of aggregated hyperphosphorylated tau are the primary histopathological lesions of AD [3]. Amyloid plaques consist of highly structured Aβ peptides generated by βand γsecretase cleavage of APP in cells [4]. High levels of Aβ 42 were shown to contribute to neural damage in early and late stages of AD [5]. Soluble, low molecular weight, Aβ 42 oligomers contribute to cell damage and neurotoxicity [6], with the pathogenicity of amyloid proteins developing well before the emergence of measurable cognitive symptoms. In fact, memory deficits in mice correlate with the early stages of Aβ 42 oligomer formation during Aβ deposition [7].
In this study, we examined whole cell proteomic adaptations to Aβ 42 (Aβ 42 proteome) in PC12 cells, which are an established model system for amyloid cytotoxicity [14]. Our findings underscore the role of mitochondrial responses to Aβ 42 and delineate specific pathways for ion transport, calcium signaling, and energy production that may contribute to the early stages of AD.

Liquid-Chromatography Electrospray Ionization Mass Spectrometry
Sample preparation for proteomics was conducted as described in [18]. Briefly, 100 µg of protein was precipitated by the addition of acetone 100 µL at 4 • C for 5 min and then centrifuged for 3 min at 16,000× g. The pellet was sequentially treated with 8 M urea, 1 M dithiothreitol, 0.5 M iodoacetamide, and 2 µL (0.5 µg/µL) trypsin in 0.5 M ammonium bicarbonate, then incubated at 37 • C for 5 h to denature, reduce, alkylate, and digest proteins. The sample was desalted using C-18 ZipTips (Millipore), dehydrated in a SpeedVac for 18 min, and reconstituted with 0.1% formic acid for a final volume of 20 µL for liquid-chromatography electrospray ionization mass spectrometry (LC-ESI MS/MS). LC-ESI MS/MS was performed using an Exploris Orbitrap 480 equipped with an EASY-nLC 1200HPLC system (Thermo Fischer Scientific, Waltham, MA, USA). Peptides were separated using a reversed-phase PepMap RSLC 75 µm i.d by 15 cm long with a 2 µm particle size C18 LC column (Thermo Fisher Scientific, Waltham, MA, USA), eluted with 0.1% formic acid and 80% acetonitrile at a flow rate of 300 nL/min. After one full scan (60,000 resolution) from 300 m/z to 1200 m/z, high abundance peptides were selected in a data-dependent fashion for fragmentation by high-energy collision dissociation (HCD) with a normalized collision energy of 28%. Enabled filters included EASY-IC internal mass calibration, monoisotopic precursor selection, and dynamic exclusion (20 s). Peptide precursor ions with charge states from +2 to +4 were included. All samples were run in triplicate.

Protein Identification and Quantification
Proteome Discoverer v2.4 (Thermo Fisher Scientific, Waltham, MA, USA) was used to identify proteins using the SEQUEST HT search engine to compare MS spectra to the Rattus norvegicus Rat_NCBI2016 database under the following parameters: mass tolerance for precursor ions = 5 ppm; mass tolerance for fragment ions = 0.05 Da; and cut-off value for the false discovery rate (FDR) in reporting peptide spectrum matches (PSM) to the database = 1%. The abundance ratios and p-values were calculated by precursor ion quantification in Proteome Discoverer v2.4. Non-treated NGF differentiated PC12 cells were used as a control.

Gene Set Enrichment Analysis (GSEA) and Genomic Enrichment Analysis (GEA)
Bioinformatic analysis was performed on filtered dataset proteins obtained from Proteome Discoverer v2.4 with a quantifiable spectra signal profile in at least two out of the three replicates. Proteins differentially impacted by the treatment condition were identified based on the calculated protein abundance ratio (fold-change) p < 0.05 using Proteome Discoverer. Proteins were identified by Entrez ID gene identifiers in NCBI.
For GSEA or GEA, protein identifiers, abundance ratios, and individual p-values were uploaded into Pathway Studio (PS) (www.pathwaystudio.com, accessed 2 March 2021 through 20 June 2021). GSEA was performed against the Gene Ontology (GO) pathways stored in the PS environment.

Statistics
One-way analysis of variance (ANOVA), followed by Benjamini-Hochberg post-hoc, was used to calculate p-values within Proteome Discoverer v2.4. Statistical significance of GO pathways and GEA, analyzed in Pathway Studio, was obtained using a Mann-Whitney U test. p-values of < 0.05 were considered significant. The data was organized and presented using the R package, ggplot2 [19], and Excel.

Whole-Cell Proteome Analysis of PC12 Cells Exposesd to Aβ 42
To uncover molecular mechanisms of Aβ 42 mediated neurotoxicity, NGF differentiated PC12 cells were treated with 100 nM Aβ 42 prepared and presented in a manner consistent with the formation of pathogenic oligomeric amyloid peptides [20]. PC12 cells were exposed to three days of Aβ 42 , then analyzed using LC-ESI-MS/MS. In these experiments, NGF differentiated cells not exposed to Aβ 42 were used as the control condition. Using MS proteomic spectra identification and label-free precursor ion quantification for relative abundance [21] in Proteome Discoverer v2.4, we identified proteins significantly impacted by the presence of Aβ 42 . Pathway Studio (PS) and Gene Ontology (GO) informatic analysis was used to deduce significant molecular pathways and cellular components of the Aβ 42 response ( Figure 1).

Gene Set Enrichment Analysis (GSEA) and Genomic Enrichment Analysis (GEA)
Bioinformatic analysis was performed on filtered dataset proteins obtained from Proteome Discoverer v2.4 with a quantifiable spectra signal profile in at least two out of the three replicates. Proteins differentially impacted by the treatment condition were identified based on the calculated protein abundance ratio (fold-change) p < 0.05 using Proteome Discoverer. Proteins were identified by Entrez ID gene identifiers in NCBI.
For GSEA or GEA, protein identifiers, abundance ratios, and individual p-values were uploaded into Pathway Studio (PS) (www.pathwaystudio.com, accessed 2 March 2021 through 20 June 2021). GSEA was performed against the Gene Ontology (GO) pathways stored in the PS environment.

Statistics
One-way analysis of variance (ANOVA), followed by Benjamini-Hochberg post-hoc, was used to calculate p-values within Proteome Discoverer v2.4. Statistical significance of GO pathways and GEA, analyzed in Pathway Studio, was obtained using a Mann-Whitney U test. p-values of < 0.05 were considered significant. The data was organized and presented using the R package, ggplot2 [19], and Excel.

Whole-Cell Proteome Analysis of PC12 Cells Exposesd to Aβ42
To uncover molecular mechanisms of Aβ42 mediated neurotoxicity, NGF differentiated PC12 cells were treated with 100 nM Aβ42 prepared and presented in a manner consistent with the formation of pathogenic oligomeric amyloid peptides [20]. PC12 cells were exposed to three days of Aβ42, then analyzed using LC-ESI-MS/MS. In these experiments, NGF differentiated cells not exposed to Aβ42 were used as the control condition. Using MS proteomic spectra identification and label-free precursor ion quantification for relative abundance [21] in Proteome Discoverer v2.4, we identified proteins significantly impacted by the presence of Aβ42. Pathway Studio (PS) and Gene Ontology (GO) informatic analysis was used to deduce significant molecular pathways and cellular components of the Aβ42 response ( Figure 1).  MS proteomic analysis identified a total of 4206 known proteins within our cells; a complete list has been deposited in FigShare dataset (10.6084/m9.figshare.15157257). Of those, a subset of 316 proteins was identified as significantly altered by Aβ 42 relative to controls (p < 0.05) and are assigned the label "Aβ 42 proteome" in the study. As shown in Figure 2A (Supplementary Tables S1 and S2), within the Aβ 42 proteome 226 proteins (72%) were decreased, while 90 proteins (28%) were increased. GO analysis of the Aβ 42 proteome revealed an impact on mitochondrial processes ( Figure 2B-D) consistent with earlier findings in neurons [22]. MS proteomic analysis identified a total of 4206 known proteins within our cells; a complete list has been deposited in FigShare dataset (10.6084/m9.figshare.15157257). Of those, a subset of 316 proteins was identified as significantly altered by Aβ42 relative to controls (p < 0.05) and are assigned the label "Aβ42 proteome" in the study. As shown in Figure 2A (Supplementary Tables S1 and S2), within the Aβ42 proteome 226 proteins (72%) were decreased, while 90 proteins (28%) were increased. GO analysis of the Aβ42 proteome revealed an impact on mitochondrial processes ( Figure 2B-D) consistent with earlier findings in neurons [22].

Gene Set Enrichment Analysis of Amyloid-Associated Pathways
Using Gene Set Enrichment Analysis (GSEA) in Pathway Studio (https://www.pathwaystudio.com, accessed in 2 March 2021) [24], Aβ42 proteome functions were extracted and analyzed. A bioinformatic assessment of all significantly altered proteins within the Aβ42 proteome (Supplementary Table S5), identified 30 specific enriched GO pathways for the Aβ42 proteome ( Figure 4 and Table 1). A trend for engaging adenine nucleotide transmembrane transport (ADP/ATP) and mitochondrial ribosomal proteins (mitoribosome) in the presence of amyloids emerged. This bioinformatic analysis complements previous findings, which showed an impact of amyloid on mitochondrial protein expression [25] and suggests a specific role for proteins, such as SLC25A5, in oxidative phosphorylation (OXPHOS) amyloid toxicity.

Gene Set Enrichment Analysis of Amyloid-Associated Pathways
Using Gene Set Enrichment Analysis (GSEA) in Pathway Studio (https://www. pathwaystudio.com, accessed in 2 March 2021) [24], Aβ 42 proteome functions were extracted and analyzed. A bioinformatic assessment of all significantly altered proteins within the Aβ 42 proteome (Supplementary Table S5), identified 30 specific enriched GO pathways for the Aβ 42 proteome (Figure 4 and Table 1). A trend for engaging adenine nucleotide transmembrane transport (ADP/ATP) and mitochondrial ribosomal proteins (mitoribosome) in the presence of amyloids emerged. This bioinformatic analysis complements previous findings, which showed an impact of amyloid on mitochondrial protein expression [25] and suggests a specific role for proteins, such as SLC25A5, in oxidative phosphorylation (OXPHOS) amyloid toxicity.
Earlier studies demonstrated an ability of Aβ [26] and α-synuclein [27], to impact mitochondrial trafficking and function. Consistently, we detected significant changes in the expression of many mitochondrial proteins (95 down-regulated; and 13 up-regulated) within the Aβ 42 proteome dataset ( Figure 5A, Supplementary Tables S6 and S7). GO analysis showed that these proteins were important for mitochondrial membrane organization, inner structural scaffolding, and mitoribosome function ( Figure 5B). In particular, a significantly downregulated set of solute carriers (SLC) proteins, including SLC25A4 and SLC25A5, are known to mediate nucleotide (ADP/ATP) transport across the mitochondrial membrane [28]. These proteins are also known to regulate nuclear function ( Figure 5C, Supplementary Tables S8 and S9) [29].

An Analysis of Aβ 42 Proteome Associated Genes and Possible Relevance to Human Disease
Genomic enrichment analysis (GEA) of the human genome was performed on the Aβ 42 proteome dataset using PS. Genes that encode components of the Aβ 42 proteome were found randomly distributed throughout the human genome. Interestingly,~70% of the Aβ 42 proteome appeared to localize to gene regions of disease, including rare copy number variation (e.g., microdeletions and/or duplications). A complete list of the GEA results is presented in Supplementary Table S10 and lists all genes, chromosomal location, and human disease phenotype associated with the Aβ 42 proteome dataset. Figure 6, summarizes GEA of genes associated with mitochondrial proteins within the Aβ 42 proteome. As shown, our results suggest a link between amyloid-associated cellular responses and various genetic factors that contribute to human disease. For example, chromosome 1p36 deletion syndrome appears impacted by amyloid-sensitive proteins encoded by the SDHB and UQCRHL gene, which both reside within that gene region and can critically contribute to cell energetics. In fact, 1p36 deletion syndrome was shown to associate with congenital cardiac abnormalities and/or developmental and intellectual deficits [30]. Another genomic region of interest resides on chromosome 17q21, housing both the BRCA1 and PHB genes, whose products are estrogen receptor (ER) sensitive [31,32]. In this case, PHB encodes the mitochondrial protein prohibitin that is significantly decreased within our dataset, suggesting a link between estrogen and amyloid responses in cells. within the Aβ42 proteome dataset ( Figure 5A, Supplementary Tables S6 and S7). GO analysis showed that these proteins were important for mitochondrial membrane organization, inner structural scaffolding, and mitoribosome function ( Figure 5B). In particular, a significantly downregulated set of solute carriers (SLC) proteins, including SLC25A4 and SLC25A5, are known to mediate nucleotide (ADP/ATP) transport across the mitochondrial membrane [28]. These proteins are also known to regulate nuclear function ( Figure  5C, Supplementary Tables S8 and S9) [29].

Proteomic Mechanisms of Aβ 42 Cellular Calcium Signaling
Amyloid proteins were shown to increase intracellular calcium levels, leading to calcium-mediated neurotoxicity [22,33]. Aβ 42 mediates cellular calcium entry through binding to cell surface receptors, such as the α7 nACh [12,13]. Therefore, the subcellular distribution of calcium binding molecules (i.e., proteins with a known calcium binding domain) and more broadly, calcium associated molecules (i.e., those with known regulation by calcium) was analyzed within the Aβ 42 proteome. As shown in Figure 7, Aβ 42 treatment was associated with a noticeable impact on cellular calcium signaling through the regulation of both calcium binding and associated proteins. GO domain mapping indicated that impacted proteins localize to various subcellular compartments and can traffic between important organelles. For example, FKBP1A, which was significantly up-regulated (foldchange = 4.632, p = 2.37 × 10 −16 ) in the presence of Aβ 42 , is a calcium binding cis-trans isomerase that modulates ryanodine receptor (RyR) activity and protein folding in the ER. Here, FKBP1A appears to participate in RyR mediated calcium flux from the ER to the mitochondria at the mitochondrial associated membrane (MAM) [34,35] (Figure 7 and Supplementary Tables S11 and S12). Overall, our proteomic findings support existing evidence on the direct ability of amyloid peptides to disrupt intracellular calcium activity in neural cells [36], and suggest a multi-compartment model for the impact of amyloids on calcium homeostasis. letion syndrome appears impacted by amyloid-sensitive proteins encoded by the SDHB and UQCRHL gene, which both reside within that gene region and can critically contribute to cell energetics. In fact, 1p36 deletion syndrome was shown to associate with congenital cardiac abnormalities and/or developmental and intellectual deficits [30]. Another genomic region of interest resides on chromosome 17q21, housing both the BRCA1 and PHB genes, whose products are estrogen receptor (ER) sensitive [31,32]. In this case, PHB encodes the mitochondrial protein prohibitin that is significantly decreased within our dataset, suggesting a link between estrogen and amyloid responses in cells. Figure 6. Genomic analysis of mitochondrial protein impacted by Aβ42. Genome enrichment analysis (GEA) was used to examine the number and location of genes that encode mitochondrial components of the Aβ42 proteome. The red line indicates the total number of impacted genes/chromosome. Boxed are chromosomal regions of copy number variation that correlate with human disease. Chromosome 1, which contains 18 impacted genes, has been adjusted for size.

Proteomic Mechanisms of Aβ42 Cellular Calcium Signaling
Amyloid proteins were shown to increase intracellular calcium levels, leading to calcium-mediated neurotoxicity [22,33]. Aβ42 mediates cellular calcium entry through binding to cell surface receptors, such as the α7 nACh [12,13]. Therefore, the subcellular distribution of calcium binding molecules (i.e., proteins with a known calcium binding domain) Figure 6. Genomic analysis of mitochondrial protein impacted by Aβ 42 . Genome enrichment analysis (GEA) was used to examine the number and location of genes that encode mitochondrial components of the Aβ 42 proteome. The red line indicates the total number of impacted genes/chromosome. Boxed are chromosomal regions of copy number variation that correlate with human disease. Chromosome 1, which contains 18 impacted genes, has been adjusted for size. and more broadly, calcium associated molecules (i.e., those with known regulation by calcium) was analyzed within the Aβ42 proteome. As shown in Figure 7, Aβ42 treatment was associated with a noticeable impact on cellular calcium signaling through the regulation of both calcium binding and associated proteins. GO domain mapping indicated that impacted proteins localize to various subcellular compartments and can traffic between important organelles. For example, FKBP1A, which was significantly up-regulated (foldchange = 4.632, p = 2.37 × 10 −16 ) in the presence of Aβ42, is a calcium binding cis-trans isomerase that modulates ryanodine receptor (RyR) activity and protein folding in the ER.
Here, FKBP1A appears to participate in RyR mediated calcium flux from the ER to the mitochondria at the mitochondrial associated membrane (MAM) [34,35] (Figure 7 and Supplementary Tables S11 and S12). Overall, our proteomic findings support existing evidence on the direct ability of amyloid peptides to disrupt intracellular calcium activity in neural cells [36], and suggest a multi-compartment model for the impact of amyloids on calcium homeostasis.

Discussion
Neurodegeneration in the course of AD is suspected to stem from alterations in amyloid precursor protein (APP) processing, resulting in the elevation of a pathogenic Aβ42 peptide in the brain and cerebrospinal fluid (CSF) [37]. We investigated the direct impact of this pathogenic amyloid peptide on cells using whole cell proteome analysis and GO

Discussion
Neurodegeneration in the course of AD is suspected to stem from alterations in amyloid precursor protein (APP) processing, resulting in the elevation of a pathogenic Aβ 42 peptide in the brain and cerebrospinal fluid (CSF) [37]. We investigated the direct impact of this pathogenic amyloid peptide on cells using whole cell proteome analysis and GO analysis. We found that Aβ 42 exposure preferentially impacted proteins associated with the mitochondria including its known interactions with other cellular organelles. We also identified other pathways of known importance for amyloid protein processing including two subunits of the biogenesis of lysosome-related organelle complex 1 (BLOC1S2 and BLOC1S6) that are involved in endosome recycling, another mechanism implicated in AD [38][39][40]. These proteins are consistent with the concurrent activity of lysosome and autophagy pathways that known to be activated during by amyloid protein degradation in neural cells [38].
The Aβ 42 proteome reveals a mechanistic framework for understanding how amyloid peptides disrupt cell bioenergetics, calcium homeostasis, and the shape and localization of mitochondrion [41]. As depicted in Figure 8, our data proposes an important impact on mitochondrial cristae scaffolding proteins and disruption to energy management through proteins such as prohibitin. Specifically, the energy producing cellular machinery, including components of metabolite transport, tricarboxylic acid cycle (TCA), the electron transport chain (ETC) (complexes I-IV), and the ATP synthase, all appear to be impacted by Aβ 42 . The ETC, responsible for creating and maintaining the vital H + gradient across the mitochondrial membrane, upon exposure to Aβ 42 is likely perturbed at the H + generating complex (I, III and IV) as based on proteomic evidence of changes to components of complex II during Aβ 42 exposure. Here, a down-regulation in succinate dehydrogenase (SDBH) is also predicted to disrupt the TCA cycle and its link to the ETC [42]. This proposed failure of the ETC should yield a drop in mitochondrial ATP synthesis and drive ROS production, leading to disruption in membrane potential (∆ψ m ) [43]. This model is consistent with the observed features of amyloid toxicity [41]. The Aβ42 proteome reveals a mechanistic framework for understanding how amyloid peptides disrupt cell bioenergetics, calcium homeostasis, and the shape and localization of mitochondrion [41]. As depicted in Figure 8, our data proposes an important impact on mitochondrial cristae scaffolding proteins and disruption to energy management through proteins such as prohibitin. Specifically, the energy producing cellular machinery, including components of metabolite transport, tricarboxylic acid cycle (TCA), the electron transport chain (ETC) (complexes I-IV), and the ATP synthase, all appear to be impacted by Aβ42. The ETC, responsible for creating and maintaining the vital H + gradient across the mitochondrial membrane, upon exposure to Aβ42 is likely perturbed at the H + generating complex (I, III and IV) as based on proteomic evidence of changes to components of complex II during Aβ42 exposure. Here, a down-regulation in succinate dehydrogenase (SDBH) is also predicted to disrupt the TCA cycle and its link to the ETC [42]. This proposed failure of the ETC should yield a drop in mitochondrial ATP synthesis and drive ROS production, leading to disruption in membrane potential (Δψm) [43]. This model is consistent with the observed features of amyloid toxicity [41].
The Aβ42 proteome is also deficient in a number of other mitochondrial proteins essential for cell survival and synaptic function. For example, Aβ42 was found to decrease the expression of the pyruvate carrier MPC2, a component of the pyruvate dehydrogenase complex that transports pyruvate into the mitochondrial matrix [44]. In addition, Aβ42 was found to negatively impact the levels of proteins vital for the exchange of ATP/ADP, phosphate, and various metabolites across the mitochondrial membrane. Observed changes in several key SLC25 family transport proteins, including SLC25A4 and SLC25A5, necessary for the exchange of ATP/ADP [45], and SLC25A3, involved in phosphate transport during ATP production, suggest several mitochondrial biomarker proteins in early AD detection. Figure 8. A model of mitochondrial disruption by Aβ42 based on our proteomic findings of specifically altered mitochondrial proteins (bold). 1: bioenergetics (TCA and ETC); 2: pyruvate dehydrogenase complex; 3: ATP/ADP exchange; 4: inorganic phosphate carrier; 5: glutamate/aspartate carrier; 6: LETM1 calcium transport; 7: FKBP1a stems calcium influx from ER; 8: anti-apoptotic HAX1 factor localized to the inter-membrane space through interaction with PHB2. 9: OPAprohibitin complex stabilizes cristae; and 10: nucleoid stabilized to the IMM by PHB.
The Aβ 42 proteome is also deficient in a number of other mitochondrial proteins essential for cell survival and synaptic function. For example, Aβ 42 was found to decrease the expression of the pyruvate carrier MPC2, a component of the pyruvate dehydrogenase complex that transports pyruvate into the mitochondrial matrix [44]. In addition, Aβ 42 was found to negatively impact the levels of proteins vital for the exchange of ATP/ADP, phosphate, and various metabolites across the mitochondrial membrane. Observed changes in several key SLC25 family transport proteins, including SLC25A4 and SLC25A5, necessary for the exchange of ATP/ADP [45], and SLC25A3, involved in phosphate transport during ATP production, suggest several mitochondrial biomarker proteins in early AD detection.