|Home | About | Journals | Submit | Contact Us | Français|
Various animal models of Alzheimer's disease (AD) have been created to assist our appreciation of AD pathophysiology, as well as aid development of novel therapeutic strategies. Despite the discovery of mutated proteins that predict the development of AD, there are likely to be many other proteins also involved in this disorder. Complex physiological processes are mediated by coherent interactions of clusters of functionally related proteins. Synaptic dysfunction is one of the hallmarks of AD. Synaptic proteins are organized into multiprotein complexes in high-density membrane structures, known as lipid rafts. These microdomains enable coherent clustering of synergistic signaling proteins. We have used mass analytical techniques and multiple bioinformatic approaches to better appreciate the intricate interactions of these multifunctional proteins in the 3xTgAD murine model of AD. Our results show that there are significant alterations in numerous receptor/cell signaling proteins in cortical lipid rafts isolated from 3xTgAD mice.
Alzheimer's disease (AD) is one of the most prevalent neurodegenerative disorders amongst adults of advanced age, and it is the most common form of dementia and cognitive impairment [1, 2]. The behavioral abnormalities in AD result from dysfunction and death of neurons in brain regions involved in cognition and mood, such as the hippocampus, amygdala, and cortical regions. Progressive short-term and eventual long-term memory loss and reduced cognitive capacity are associated with two primary neurodegenerative lesions, that is, extra- and intracellular β-amyloid plaques, as well as neurofibrillary tangles (NFTs) composed of the microtubule protein tau [3–5]. In addition to the effects of amyloid plaques and NFTs, the lipid trafficking molecule, apolipoprotein E4 (apoE4), has also been demonstrated to be a genetic risk factor for AD [6, 7]. The AD characteristic extracellular plaques, found in both the hippocampus and cortex of AD patients, consist of 39–42 amino acid long amyloid-β (Aβ) peptides. These extracellular peptides are generated by digestion of a transmembrane amyloid precursor protein (APP). Proteolysis of the transmembrane APP by a set of intramembrane enzymes, β- (also known as BACE-1) and γ-secretases, is thought to be responsible for toxic Aβ creation . The discovery of familial mutations in the APP gene that were strongly correlated with the presentation of AD reinforced the importance of Aβ processing in this disorder. A growing body of evidence indicates that changes in lipid and cholesterol homeostasis can influence AD progression and specifically Aβ production. One of the prime sub cellular regions of amyloidogenic APP processing is thought to be cholesterol-enriched membrane microdomains, termed lipid rafts . Cellular organization of protein signaling complexes, to enhance the magnitude and fidelity of transmembrane signaling receptors, is facilitated by variations in the lipid constituents of the plasma membrane. Lipid rafts represent discontinuous regions of the plasma membrane that form functional microdomains, which constrain the association of proteins in a coherent and advantageous manner with respect to neurotransmissive signaling . Disruption of the correct stoichiometry of signaling complexes within lipid rafts may underpin the etiology of many different neurodegenerative disorders [10–12]. The hypothesis that changes in the lipid composition of rafts contribute to AD pathology has gained considerable support. For example, ApoE4 has been strongly correlated with the generation of AD symptomatology. Both of the amyloidogenic processing enzymes (β- and γ-secretase), as well as APP, are all enriched in lipid raft membranes [13, 14]. Reinforcing the connection between lipid density levels and Aβ production, increasing cholesterol levels elevate the activity of both β-secretase (BACE-1) and γ-secretase [14, 15]. In addition, ganglioside lipids, which are also enriched in lipid rafts, can control the assembly of amyloid-β proteins [16, 17]. Changes in ganglioside composition, similar to those noted in human AD patients, are also observed in different transgenic mouse models of AD . In addition to a role of the lipid components of lipid rafts in controlling amyloidogenesis, these raft environments may also affect NFTs as well. It has been demonstrated that Aβ can induce activation of the tyrosine kinase Fyn in neuronal cells, that is then recruited to lipid rafts which catalyzes phosphorylation of tyrosine residue 18 on tau [19, 20]. Association of Aβ plaques to lipid rafts can mediate recruitment of excess Fyn to the rafts, as well as further recruitment and phosphorylation of tau. These activities are thought to induce neurotoxicity via the effects of tau-induced changes in the actin cytoskeleton and receptor/cellular signaling pathways . Therefore, the potential changes in the lipid composition of lipid rafts, caused by exposure to cytotoxic activities characteristic to AD, can induce profound changes in cellular signal transduction and thereby induce intracellular changes that lead to the development of AD. The complexity of protein complexes within the lipid raft environments raises considerable challenges to understanding the molecular mechanisms of AD pathophysiology in both the hippocampus and cortex of animals. Therefore, we have employed a shotgun proteomics approach, allied to advanced bioinformatic functional profiling, to gain a broad and detailed appreciation of the alterations in signaling proteins in lipid rafts in the triple-transgenic (3xTgAD) model of AD . Our study demonstrates that cortical lipid rafts are profoundly affected in the 3xTgAD mice and that many of the neurophysiological deficits characteristic of AD (impaired synaptic strength, impaired learning and memory, and increased oxidative stress) can be strongly linked to changes in receptor and cell signaling events in the lipid rafts in these animals. Therefore, the lipid raft environments can be seen as one of the most important pathophysiological loci of this disorder.
Animal care and experimental procedures followed NIH guidelines and were approved by the National Institute on Aging Animal Care, and Use Committee. Experiments were performed using male 3xTgAD [22–24] and control male C57-BL6 mice that were maintained under a 12-hour light/12-hour dark cycle with continuous access to food and water. Water maze testing took place using a modified version of the methodology described previously . Briefly, animals (n = 10 per group, control male C57-BL6 or male 3xTgAD, on a C57-BL6 background) received 8 days of acquisition training using a nonvisible target platform, consisting of four trials per day, with an intertrial interval of approximately 10 minutes. Each trial lasted until the animal found the platform, or for a maximum of 60 seconds; animals that failed to find the platform within 60 seconds were guided there by the experimenter. On each trial, mice were placed into the pool, facing the wall, with start locations varied pseudorandomly. Distance swam to escape the water, escape time, and swim speed were measured for either control or 3xTgAD mice using a HVS2020 automated tracking system (HVS Image, UK).
The mice were anesthetized with isoflurane, decapitated, and the brain was microdissected on ice. After removal of the cortex, the tissue was split into left and right hemisphere, half for mass spectrometry raft analysis and half to prepare lipid raft tissues for Western blot analysis. The hemicortices were washed twice in ice-cold phosphate buffered saline (PBS) and then transferred into a Tris-saline buffer supplemented with a cocktail of protease and phosphatase inhibitors (50mM Tris-HCl, 150mM NaCl, 5mM EDTA, and Roche Complete-Mini (Roche Diagnostics Inc.) protease and phosphatase inhibitor cocktail, pH 7.4). Crude tissue disruption was then rapidly achieved (at 4°C) using a sonic dismembrator (Fisher Scientific Model 100) followed by a brief centrifugation (4°C, 1000 ×g, 10 minutes) to pellet cell nuclei and unbroken cells. The resultant supernatant was removed and Triton X-100 (Sigma Aldrich, USA) was added to the Tris-saline buffer to a final concentration of 1%. The supernatant membranes were then incubated at 4°C for 60 minutes in the Triton X-100 Tris-saline solution. After incubation, the supernatant solution was then added to a discontinuous gradient of 30% and 60% OptiPrep (Iodixanol, Sigma Aldrich, and U.S.A.) before centrifugation at 200000×g for 16 hours at 4°C. After centrifugation, a detergent-resistant lipid band was evident in the vertical solution column. Multiple fractions of 300μl volumes were then removed from the vertical centrifugation column. Proteins were then extracted from these fractions using a proprietary ProteoExtract (EMD Biosciences) kit, according to the manufacturer's instructions. Isolated protein pellets were then prepared for mass spectrometric analysis.
Protein pellets were dissolved into an ammonium bicarbonate buffer (100mM, pH 8.5) and then reduced with dithiothreitol (500mM: Pierce Biotechnology), alkylated with iodoacetamide (800mM: Sigma Aldrich) and then digested with modified trypsin (5–10μg) (Promega) at 37°C for 17 hours. Proteolysis was terminated by the addition of glacial acetic acid. Tryptic peptides were then loaded onto a desalting column (360 × 200μm fused silica packed with 15cm of C18 beads (YMC ODS-AQ, Waters)), washed with 0.1% acetic acid and eluted into sample tubes with 80% acetonitrile in 0.1% acetic acid. Sample volume was reduced to usable volumes under vacuum on a Savant SpeedVac. Samples were then transferred onto a PicoFrit (75 × 100mm) column packed with ProteoPep II C18, 300Å, 5μm particles (New Objective) connected to a nanoliquid chromatography system (Dionex, Sunnyvale, CA) online with an LTQ ion trap mass spectrometer (Thermo Finnigan, San Jose, CA). The peptides were eluted using a linear gradient of 0–65% acetonitrile over 90 minutes at a flow rate of 250nl/min directly into the mass spectrometer, which was operated to generate collision-induced dissociation spectra (data-dependent MS/MS mode). The resultant tandem mass spectrometry data were processed using the BioWorks suite, and multiple collected spectra were used to interrogate the NCBI nonredundant mouse and Swiss-Prot protein sequence databases, using the computer algorithm SEQUEST to generate accurate protein identities. Protein genpept accession identities were then converted to Official Gene Symbol terms using NIAID-DAVID v. 6.7 (http://david.abcc.ncifcrf.gov/). The statistical analysis and validation of the search results were performed using MASCOT (Matrix Science). For protein identification, a maximum of three missed tryptic cleavages was used, including fixed modification of carbamidomethylation and variable modifications of oxidized methionine and N-terminal glutamine conversion to pyroglutamic acid in the search. Only proteins with at least two validated peptides and a total score 25 and a confidence of identification of at least 95% were considered valid for reporting. Where required, additional spectral counting was performed to determine simplistic relative quantitation in conjunction with the reported number of identified unambiguous peptides per protein. Three lipid raft fraction samples (fractions 2, 3, and 4) from each three control (nontransgenic gender/age matched C57-BL6) or Alzheimer's disease (3xTgAD) were pooled and then run in an individual random order. Proteins identified based on two unambiguous peptides that were present in at least two out of the three individual animals were employed for further expression pattern analysis.
Digitized images of centrifugal vertical fluid columns were obtained using a Canon Digital camera and were converted from Joint Photographic Experts Group (JPG) files to a TIFF (Tagged Image File Format) form using L-Process v. 2.2 (image handling software: Fuji-Film). Image densitometry was then performed using Fuji-Film Image Gauge v. 4.2. Lipid raft band intensity was represented as a relative absorbance unit (AU) value with background (B) subtraction per square pixel (px2) (AU-B/px2).
For the examination of specific proteins in cortical cell samples (both lipid raft and nonlipid raft), aliquots were removed from centrifugal fractions from Section 2.2 and their protein concentration was determined with a standard BCA protocol. Aliquot samples for western blotting analysis were then mixed with an equal volume of Laemmli sample buffer . Samples were resolved using one-dimensional gel electrophoresis (SDS-PAGE), followed by electrotransfer to polyvinylenedifluoride (PVDF: PerkinElmer, Waltham, MA). PVDF membranes were blocked for one hour at room temperature in 4% nonfat milk (Santa Cruz; Santa Cruz, CA) before application of specific primary antisera in the same nonfat milk. The presence of primary antibody reactivity with the PVDF membrane was detected by the application of a 1:5000 dilution of a species-specific alkaline phosphatase-conjugated secondary antibody (Sigma, St. Louis, MO). PVDF-bound immune complexes of secondary and primary antibodies were subsequently detected using enzyme-linked chemifluorescence (ECF: GE Healthcare; Pittsburgh, PA). Chemifluorescent signals from the membranes were captured and quantified using a Typhoon 9410 phosphorimager (GE Healthcare, Pittsburgh, PA). Specific primary antisera used were obtained from the following sources: flotillin-1, proline-rich tyrosine kinase 2 (Pyk2), focal adhesion kinase (FAK), G protein-coupled receptor kinase interactor-1 (GIT-1), and paxilin antibodies were obtained from BD Bioscience, San Jose, CA; Janus kinase 2 (Jak2), v-Crk avian sarcoma virus CT10 oncogene homolog (Crk), and insulin receptor substrate-1 (IRS1) antibodies were obtained from Santa Cruz Biotechnology Corporation, CA; caspase-7, FKBP12-rapamycin complex-associated protein 1/mammalian target of rapamycin (FRAP1/mTOR), and Fyn and IGF-1 receptor beta antibodies were obtained from Cell Signaling Technology, Danvers, MA); G protein-regulated inducer of neurite outgrowth 2 (Grin2) antibody was obtained from Sigma Aldrich. For the identification of nonspecific total proteins in each sample the highly sensitive protein dye, SYPRO Ruby (Invitrogen Corporation) was employed. Fixed SDS-PAGE gels were immersed in SYPRO-Ruby for 1 hour and then washed in deionized water before scanning using a Typhoon 9410 phosphorimager (GE Healthcare, Pittsburgh, PA).
Protein identities were converted to standard gene symbol nomenclature for simplicity of usage with the batch conversion tool of NIH Bioinformatics Resources DAVID v. 6.7 (http://david.abcc.ncifcrf.gov/). Primary protein sets (containing consistently identified lipid raft extract proteins) were organized into functional signaling pathway groups and then analyzed for their differential significance of population of these canonical signaling pathways. To compare the relative degree of association of specific signaling pathways with the control or 3xTgAD protein sets, the difference between the signaling pathways “hybrid scores” was calculated (control subtracted from 3xTgAD). The magnitude of the “hybrid score” is indicative of strength and significance of association of the input protein set with the specific signaling pathway. Signaling pathway hybrid scores were generated using a process that takes into account the significant population and potential activation of that pathway by multiplying the pathway enrichment ratio (percentage of proteins in a designated pathway that were also found in the experimental dataset) and the probability (P) that the respective pathway is significantly associated with the experimental dataset. However, to create a simple numerical value, the hybrid pathway score is calculated by multiplication of the ratio with the negative log − 10 of the P value. Each signaling pathway considered was required to contain at least two unique proteins from either control or 3xTgAD datasets and possess a P value of ≤.05. Unbiased network analysis was also performed on subsets of the primary protein sets that were specifically limited to transmembrane receptor proteins. The networks generated create predictions of the most likely functional interactions between proteins in a complex dataset . Networks are created to indicate the most significant series of molecular interactions. The networks with the highest predictive “scores” possess the highest number of statistically significant “focus molecules”: “focus molecules” are proteins that are present in the most statistically-likely predicted functional network and are present in the input experimental dataset. The network “score” is a numerical value used to rank networks according to their degree of relevance to the input dataset. The “score” accounts for the number of experimental focus molecules (proteins) in the network and its size, as well as the total number of proteins in the Ingenuity Knowledge Base that could potentially be included in the specific networks. The network “score” is based on the hypergeometric distribution and is calculated with the right-tailed Fisher's Exact Test. Specific scientific textual associations between filtered protein sets (transmembrane receptor proteins IPA analysis) and Alzheimer's disease processes were created using latent semantic indexing (LSI) algorithms using GeneIndexer (Computable Genomix, Incorporated: https://www.computablegenomix.com/geneindexer). GeneIndexer correlates the strength of association between specific factors (proteins) in a dataset with a user-defined interrogation term. GeneIndexer employs a 2010 murine or human database of over 1 × 106 scientific abstracts to perform text-protein correlation analysis. LSI facilitates the specific textual interrogation of an input dataset with a specific term, that is, Alzheimer's disease, to ascertain which of the input dataset proteins are explicitly associated with the interrogation term. Using LSI algorithms, not only is the direct interrogation term used to analyze the input dataset but also closely correlated additional terms, implicitly associated with the user-defined interrogation term, are also employed in the search patterns. A latent semantic indexing correlation score indicates the strength of association of the interrogation term and the specific proteins in the dataset. A highly relevant protein-term correlation yields a large number of explicitly/implicitly associated proteins with high LSI correlation scores. Therefore, a strong correlation between the proteins in a dataset and a specific user-defined interrogation term yields a large number of correlated proteins with high LSI correlation scores.
Statistical analysis on multiple samples was performed using a standard nonparametric two-tailed Student's t-test using 95% confidence limits. Analyses were computed using built-in software in GraphPad Prism v. 3.0a (GraphPad Software Inc., La Jolla, CA). Results are expressed as means ± SE. P ≤ .05 was considered statistically significant. For statistical analysis using Ingenuity Pathway Analysis v. 8.5 of signaling pathways and interaction network analysis, Fisher's Exact test was employed with a P ≤ .05 cutoff. Network interaction scores were generated using a right-tailed Fisher's Exact Test.
Using the nonvisible Morris water maze trial and 16-month-old male control (C57-BL6) and 3xTgAD animals (n = 10 for both) we noted that the 3xTgAD mice demonstrated a significant reduction in their ability to find the location of the hidden platform (Figure 1). The 3xTgAD mice demonstrated significantly longer escape latencies and distances traveled compared to the control mice, while not showing any significant difference in calculated swim speed. Retention testing (three trials one week after the initial training) of these animals (control and 3xTgAD) also demonstrated a reduced cognitive capacity of the 3xTgAD mice compared to control (data not shown).
Employment of the lipid raft isolation process described in the Methods section resulted in the clear visible isolation of a detergent-resistant lipid layer comprising centrifugal fractions 2–4 (Figure 1(a)). The lipid raft marker protein, flotillin-1, was demonstrated to be specifically enriched in these centrifugal fractions (2–4) (Figure 2(a)). The visual lipid density (absorbance units-background/square pixel) of the raft layers was quantified using Fuji-Film Image Gauge. Compared to control, both 8-month-old (Figure 2(c)) and 16-month-old (Figure 2(d)) 3xTgAD-derived centrifugal raft layers demonstrated a significant (8 months old P = .027, n = 3; 16 months old P = .031, n = 3) increase in buoyant detergent-insoluble density. This 3xTgAD increase in raft size, compared to control animals, demonstrated a strong association with a significant increase in expression of flotillin-1 in the raft fractions of 3xTgAD mice, especially in centrifugal fraction 2 (Figure 2(e), P = .017, n = 3). Equal levels of total protein (measured using BCA and also SYPRO gel staining) were employed for each Western blot of the raft extracts. Quantification of fraction 2 was chosen, as this reliably indicated the greatest enrichment of this lipid raft marker. Similar quantitative alterations in expression of flotillin-1 between control and 3xTgAD mice were also seen in the additional lipid raft centrifugal fractions, that is, 3 and 4. Qualitatively similar results, with respect to 3xTgAD mouse lipid raft density and flotillin-1 expression were noted in parallel experiments carried out with age-matched female mice. In addition we also noted a similar qualitative lipid raft expression of flotillin-1 trend in male human cortex tissue (data not shown). These latter data and their significance to our current data will be further addressed in subsequent manuscripts.
Using an un-biased proteomic analysis of replicate lipid raft extracts, we were able to identify (from at least two individual nonambiguous peptides) multiple proteins in both control and 3xTgAD cortical extracts (control, Appendix A; 3xTgAD, Appendix B). When comparing the relative differences in lipid raft protein expression, only a small minority (17%: Figure 3(a)) of identified proteins were substantively identified in both control and 3xTgAD raft samples; however many of these common proteins identified were differentially detected (see Supplementary Table 1 in Supplementary Material available online at doi:10.4061/2010/604792). To verify the relative differential expression of multiple proteins in the control versus 3xTgAD lipid raft extracts, we also performed multiple Western blot analyses of raft centrifugal fraction-2 (F-2) samples. With loading of total equal protein quantities (50μg: assessed in an unbiased manner with SYPRO-Ruby: Figure 3(b)) of either control or 3xTgAD F-2 samples, we assessed the relative differential expression of multiple proteins (Figures 3(c)–3(n)). From the Western blot analysis it was consistently demonstrated that differential qualitative protein detection in control versus 3xTgAD raft samples strongly correlated with differential semiquantitative protein expression. Hence, the absence of consistent MS-based detection of Pyk2 (Figure 3(d)), Jak2 (Figure 3(f)), Fyn (Figure 3(h)), paxilin (Figure 3(i)), IRS-1 (Figure 3(j)), caspase 7 (Figure 3(k)), mTOR/FRAP1 (Figure 3(l)) and IGF-1R (Figure 3(n)) in control raft sample correlated to their significantly lower expression in control raft F-2 samples, compared to that in 3xTgAD samples. Conversely, the absence of consistent MS-based detection of FAK (Figure 3(c)), GIT-1 (Figure 3(e)), Crk (Figure 3(g)), and Grin2 (Figure 3(m)), correlated to their significantly lower expression in 3xTgAD raft F-2 samples, compared to that in control samples.
As our MS-based multidimensional protein identification process identified several hundred proteins from each control or 3xTgAD lipid raft sample, we employed a bioinformatic analysis process to assess the relative functionalities of both the control versus 3xTgAD raft protein lists. As the majority of cellular signaling processes are mediated and regulated by multiple groups of proteins interacting with each other, we clustered, in a statistically significant manner, proteins in control or 3xTgAD animal raft samples into functional signaling groups. To assess the relative changes in regulation of classical signaling pathways, we applied a subtractive approach for the pathway “hybrid” scores (indicative of the “activity” of the specific signaling pathway: calculated by significant expression enrichment ratio of proteins in that pathway multiplied by the negative log10 (−log10) of the probability of the pathway enrichment). For each specific common signaling pathway, our mathematical approach subtracted control pathway “hybrid” scores from the pathway “hybrid” scores generated from the 3xTgAD protein set. Hence, a positive result of this subtraction indicates a greater activity of this functional pathway in 3xTgAD animals, and a negative score indicates a greater activity of this functional pathway in the control animals. Analysis of pathways involved in cellular signaling (Figure 4(a): proteins and scores in associated Appendix C) demonstrated that pathways commonly associated with cell stress responses were highly activated in 3xTgAD rafts, for example “p53 signaling”, “p38 mitogen-activated protein kinase (MAPK) signaling”, and “stress-activated protein kinase (SAPK)/JNK pathways”. In a stark contrast, prosurvival synaptic connectivity and neurotransmissive pathways were more profoundly activated in the control mice, for example, “tight junction signaling”, “calcium signaling”, “PTEN signaling”, and “Wnt/β-catenin signaling”. To investigate the specific neuronal functional effects of these disparate signaling activities, we next studied the significant clustering of raft proteins into neuron-functional pathways (Figure 4(b): Appendix D). As one would expect, the 3xTgAD mice raft protein clustering revealed a considerably greater (relative to control) activation of multiple neurodegenerative neuronal processes including: “amyloid processing”, “Amyotrophic lateral sclerosis”, “Huntington's disease signaling” and “Parkinson's signaling”. In addition to the greater activation of these degenerative processes, the 3xTgAD mice also demonstrated profound changes in the significant clustering of proteins into cytoskeletal remodeling groups (“actin cytoskeleton signaling”, regulation of actin-based motility by Rho') compared to the control mice. In accordance with our demonstration of the significant diminution of the learning and memory ability of the 3xTgAD mice, it was striking to notice the profoundly greater activation of neuron-functional pathways that control synaptic learning-dependent processes (i.e., “synaptic long-term depression”, “axonal guidance signaling”,and “synaptic long-term potentiation”) in the control mice compared to the 3xTgAD mice. Considerable evidence from multiple experimental studies has recently underlined the importance of the regulation of energy metabolism in controlling the aging process and neurodegenerative disorders [28–30]. Upon inspection of the relative differences in the activation of energy-regulatory pathways created by clustering of control raft proteins versus 3xTgAD raft proteins, a profound functional distinction was noted (Figure 4(c): Appendix E). In control animals versus 3xTgAD, there was a considerably stronger activation of energy-generating pathways connected to the use of the primary metabolic substrate, that is, sugars (“amino sugars metabolism”, “glycolysis/gluconeogenesis”, and “pentose phosphate pathway”). In contrast, the energy regulatory pathways that were more strongly associated with the 3xTgAD animals involved energy derivation from alternative energy sources, for example, “synthesis and degradation of ketone bodies”, “butanoate metabolism”, and “fatty acid biosynthesis”. Many of the alterations in energy regulation in aging and degenerative disorders are thought to be associated with adaptive responses to the induction of cellular stresses, potentially through toxic effects of Aβ, NFTs and accumulated oxidative damage . When the raft proteins from control and 3xTgAD mice were clustered into functional stress response pathways, again a stark contrast in the control- or 3xTgAD-associated pathways was demonstrated (Figure 4(d): Appendix F). In the 3xTgAD mice raft clustering it was noted that the association of energy-associated stressful and neuronal damage-related pathways (“PPARα/RXRα activation”, “hypoxia signaling”, “apoptosis signaling”, and “endoplasmic reticulum stress pathway”) was considerably stronger than in the control mice. Indicating a correlated connection between stress response capacity and AD pathology, there was a considerably greater association of the “Nrf2-mediated oxidative stress response pathway” in control mice compared to the 3xTgAD. Therefore, the 3xTgAD mice may demonstrate excessive neuronal stress and damage due to the attenuated activation of such stress response pathways in lipid rafts of AD synapses.
As one of the most important functions of synaptic lipid rafts is to congregate transmembrane or juxtamembrane receptor systems , we next performed an in-depth investigation of the significant differential functional clustering of receptor signaling pathways between control and 3xTgAD rafts. Upon functional clustering of the raft proteins into receptor signaling pathways, strong differences in pathways association between control and 3xTgAD mice were noted (Figure 5: Appendix G). Some of the strongest differences were noted by the considerably poorer activation of growth factor-related signaling (“PDGF signaling”, “EGF signaling”, and “FGF signaling”), structural trans-synaptic receptor signaling (“Neuregulin signaling” and “Ephrin receptor signaling”), excitatory signaling (glutamate receptor signaling) and neurodevelopmental signaling (Sonic hedgehog signaling) pathways in the 3xTgAD mice, compared to the control mice. The receptor signaling profile of the 3xTgAD mice demonstrated a more profound association compared to control mice for pathways linked to inhibitory synaptic signaling (“GABA receptor signaling” and “Aryl hydrocarbon receptor signaling”: ), amyloid processing (“Notch signaling” and “Integrin signaling”: ), and neuronal stress (glucocorticoid receptor signaling). Transmembrane receptor signaling by systems including receptor tyrosine kinases or G protein-coupled receptors (GPCRs) represents one of the most important signaling mechanisms of neuronal synaptic regulation . However, the activation of transmembrane receptor systems and the stimulation of their intracellular signaling cascades, especially for GPCRs, is now considered to be far more complex and intricate than initially proposed by two-state receptor models [34, 35]. Much of this additional signaling diversity is thought to arise from the additional complexity of receptor-accessory scaffolding protein modification of receptor signaling . To appreciate the multiple connections between receptor (and GPCR in particular) activity and the presence of neurophysiological deficits in AD, we performed a multidimensional analysis of the proteins present in control or 3xTgAD mice. Using the novel, un-biased, bioinformatic GeneIndexer latent semantic indexing (LSI) process (https://www.computablegenomix.com/geneindexer.php), explicit correlations can be made between protein/gene factors from input datasets and their linkage (in over 1 × 106 curated scientific abstracts) to a specific interrogation term, for example, Alzheimer's. The semantic indexing algorithms of GeneIndexer also allow for multidimensional correlations to be measured for terms significantly related to the interrogation term, hence providing a flexible, intelligent query process. For the control and 3xTgAD datasets, we employed multiple interrogation terms targeted to demonstrate differences between control and 3xTgAD datasets. Using the following interrogation terms: Alzheimer's, oxidation, neurodegeneration, synaptic transmission, neurogenesis, scaffolding, and GPCR, we demonstrated that, at a multidimensional interactive level, there is minimal functional cross-over between control and 3xTgAD raft samples (Figure 6: Appendix H). Only proteins that demonstrated explicit correlations (latent semantic indexing score of ≥0.1) to at least two of the interrogation terms are denoted in the multidimensional heatmap (Figure 6: Appendix H). Therefore, each of the proteins identified in the heatmap are strongly correlated with many of the connected interrogation terms and therefore show potential synergistic activity. Analysis of identified proteins in this manner, that is, selected specifically for multidimensional neurological roles, allows for an un-biased focusing on proteins that may possess keystone-like functions in the molecular signaling networks involved in neurodegeneration. With respect to specific interrogation terms, a strong validation of the technique is demonstrated by the fact that using the Alzheimer's interrogation term, 47 3xTgAD-unique multidimensional proteins were indicated while only 10 such multidimensional proteins were shown in control rafts (Figure 6). Confirming a strong role of oxidative damage, 21 3xTgAD-unique multidimensional proteins were present in the oxidation results while only 6 such control-unique were demonstrated. Interestingly and perhaps suggestive of a potential future line of AD research was the observation that considerably fewer multidimensional 3xTgAD-unique proteins were associated with the process of neurogenesis (15), compared to the 32 control-unique multidimensional proteins associated with this neuroprotective mechanism. In accordance with the important role of receptor systems in AD, we additionally noted that more 3xTgAD-unique proteins were associated with GPCRs (27) compared to control-unique GPCR-related multidimensional proteins (16).
From our investigation of the multidimensional nature of raft proteins with respect to neurodegenerative processes, it is clear that there are many factors that are highly likely to work together in complex and intricate functional networks. To investigate the nature of the most statistically likely functional network interactions, we employed IPA network analysis of a receptor-filtered (using IPA-knowledge base data filtering, IPA v. 8,5) subset of the control or 3xTgAD raft protein datasets (Tables (Tables11 and and2,2, resp.). Using these datasets, un-biased network analysis is able to predict the most likely series of functional interactions (based on empirically derived experimental evidence) that take place between the receptor-associated raft proteins. The control and 3xTgAD receptor-specific filtered datasets demonstrated a relatively minimal overlap, that is, 11.6% commonality, indicating that substantial alterations of these proteins may occur in the rafts of 3xTgAD animals (Figure 7(a)). The most statistically likely interaction network that was predicted to occur in control mice centered on neuroprotective and neurotransmissive factors such as phosphoinositide-3-kinse (PI3K), Akt-1, muscarinic GPCRs (Chrm5), and glutamate receptors (Grm5) (Figure 7(a): Appendix I). In contrast, the highest statistically scoring network from 3xTgAD receptor-specific proteins was centered on lipid-regulating factors and stress-related factors including: p38 MAPK; Jnk (c-jun N-terminal kinase), Lrpap1 (low-density lipoprotein receptor-related protein associated protein 1); LDL (low density lipoprotein), and NF-κB (Figure 7(b): Appendix J). Therefore, at the level of functional interaction of receptor-related proteins in the lipid rafts, it is clear that these membrane microdomains are a strong functional locus of this degenerative disease. Reinforcing this AD-relevant microcosm effect in the lipid rafts we analyzed, using latent semantic indexing (LSI) interrogation of these receptor-specific datasets (Table 2—control; Appendix I-3xTgAD), the Alzheimer's disease correlation of these receptor-specific proteins. In Figure 8, we demonstrate that almost twice as many proteins in the 3xTgAD dataset (21) explicitly correlated with the interrogation term Alzheimer's disease compared to the proteins in the control dataset (11). The degree of correlation of the interrogation term (Alzheimer's) to each specific protein is indicated by the LSI score. In Figure 8(b) the cumulated LSI score for the 3xTgAD Alzheimer's-related proteins was 2.92 while for the control proteins only 1.21 (Figure 8(b)), demonstrating that a much stronger correlation of receptor-associated proteins existed for the proteins in 3xTgAD rafts, compared to control. Upon comparison of the phylogenetic relationships of the receptor-specific proteins that form the most likely functional networks (from Figure 7(b)), it is clear that for the control-set proteins only three of these in this network are highly correlated to reports of Alzheimer's disease (27.2%: Figure 8(c)) while the most coherent interaction network of 3xTgAD raft proteins possessed a much higher percentage of Alzheimer's disease-related proteins (62%: Figure 8(d)). Therefore, this suggests that one of the important pathological loci of AD could be the disruption of interactivity of receptor-associated proteins in the lipid raft microdomains and that such a finding can be uncovered in a completely un-biased informatic format from extremely large and difficult to interpret datasets.
Multiple informatic techniques were employed to investigate and elucidate the nature of functional protein interactions that occur differentially in the lipid raft microdomains of control versus 3xTgAD Alzheimer's disease mice. Cognitively impaired male 3xTgAD mice showed profound differences in the protein constituents of lipid raft extracts compared to age-matched control mice, that is, only 17% of raft proteins were similar between the two groups of mice. Our microdomain proteomic approach was able to specifically assist in the identification of altered proteins that are highly characteristic of AD-related raft pathophysiology, for example, the Src-family tyrosine kinase Fyn (Figure 3(h)) [19, 20].
Currently, biological scientists are often faced with complex biological issues concerning the interpretation of their results due to the development of facile mass data acquisition technologies, that is, extracting relevant and illuminating information from large datasets is often extremely challenging. However, with the application of multiple, sequential un-biased informatic processes, we were able to identify specific lipid raft involvement in altered pathways that can control multiple degenerative mechanisms, for example, attenuation of the Wnt/β-catenin signaling pathway, profound reduction of the important Nrf2 stress response pathway (Figure 4(d)), the loss of neurogenesis association, and the important involvement of synaptic GPCR-systems (Figure 6) in the 3xTgAD mice [36–38].
Elucidation of the crucial signaling relationships in this degenerative disorder, and the identity of the proteins that mediate them, could lead to the more rational development of novel therapeutics for Alzheimer's disease. Using our informatic receptor-targeted approach, we were also able to reinforce the validity of our discovery process by also identifying the importance of energy-related insulin/insulin-like growth factor (IGF) signaling in AD (Figures 3(j), 3(n), ,5,5, and 8(d)) that has recently become more widely appreciated by other researchers [39–42]. In addition to IGF receptor activity, we noted a potential implication of the presynaptic latrophilins (lphn) in AD pathophysiology (Figure 8(d)). These unique receptors form the high affinity target of α-latrotoxin and may be able to integrate presynaptic calcium regulation with an ability to potentially physically interact with the postsynaptic neuron . Another receptor signaling system that our analysis revealed to possess a potential role in AD is the Sonic Hedgehog (Shh) systems (Figures (Figures55 and 8(d)). With a specific correlation to AD pathophysiology, it has been demonstrated that Shh can act in a synergistic manner with nerve growth factor to act on central nervous system cholinergic neurons , and cholinergic insufficiency has been strongly associated with cognitive decline and Alzheimer's-related pathophysiology .
The combined use of broad range proteomic analysis, with sequential multidimensional informatic analysis enables the determination of important correlations between pathophysiology and functional protein differences. Using discrete proteomes, that is, lipid raft extracts, versus whole-cell/tissue proteomes, facilitates an improved ability to understand the complex interactivity between transmembrane receptor protein systems in a disease setting. The appreciation and clustering of protein datasets into coherent groups greatly increases our capacity to focus upon potentially important therapeutic target networks. The knowledge gained of protein activity networks may greatly assist in the future development of network-targeting therapeutics that possess a multidimensional efficacy at several interacting proteins.
Supplementary Table 1: the table illustrates the top 50 most upregulated and top 50 most downregulated (italics) proteins (commonly identified in 3xTgAD and control WT) in 3xTgAD cortical raft extracts compared to WT.
Ranking was based on relative expression ratios generated by spectral counting for 3xTgAD:WT proteins that displayed a >95% identity confidence.
Proteins ranked 1-50 for 3xTgAD>WT (increased in 3xTgAD compared to WT) represent those with expression ratios (3xTgAD:WT) greater than unity while proteins ranked 1-50 for WT>3xTgAD (decreased in 3xTgAD compared to WT: italics) represents those with expression ratios (3xTgAD:WT) less than unity. In each case the most up or downregulated proteins are ranked 1.
This work was carried out with the support of the Intramural Research Program of the National Institute on Aging, National Institutes of Health. The authors thank Dr. LaFerla for kindly donating 3xTgAD breeding pairs to the National Institute on Aging.
Lipid raft-derived extracted proteins from control mouse 16-month-old cortical tissue. Specific proteins isolated from control mouse cortex were identified with multidimensional protein identification technology (MudPIT) LC MS/MS. Each protein was identified in at least two out of three experimental replicate animals and from at least two isolated peptides per protein. The following are the protein symbol and its corresponding definitions.
Lipid raft-derived extracted proteins from Alzheimer's disease (3xTgAD) mouse 16-month-old cortical tissue. Specific proteins isolated from 3xTgAD mouse cortex were identified with multidimensional protein identification technology (MudPIT) LC MS/MS. Each protein was identified in at least two out of three experimental replicate animals and from at least two isolated peptides per protein. The following are the protein symbol and its corresponding definition.
Cellular signaling pathway clusters of extracted proteins from primary cortical tissue. Specific cellular signaling pathway clusters generated in an un-biased manner using Ingenuity Pathway Analysis (v. 8.5). The relative score generated for the degree of pathway population by proteins from the respective input sets (control or 3xTgAD) is shown in bold. The following are different types of signaling pathways.
PPP2R2A, CRK, PPP1R14B, DUSP2, PTK2, SHC1, PAK1, PPP1R10, PPP1R7, PIK3CG, STAT1, PRKCA, ETS1, SRC, PAK2, YWHAB, CRKL, PRKAR2A, STAT3, PLA2G4C, FOS, PPP2CB, PPP2R1A, PRKAR2B, PRKCI, PAK3, PPP2R2B, PRKACA, PPP2R5E, ELK1, PPP2R1B, PRKAR1A, PRKCB.
PXN, YWHAH, PTK2B, RAC1, PRKAR2A, PLA2G2A, STAT3, MYC, YWHAQ (includes EG:10971), PPP2R1A, PAK3, PIK3C3, PRKCD, PRKACA, PIK3CB, PIK3R2, ELK1, FYN, PRKAR1A.
PDIA3, OCRL, PAK1, PIK3CG, PRKAA2, PLCB1, PLCL1, PI4KA, PMPCA, ATM, GRK4, PRKCQ, PAK2, CDK7, CDK6, PLK1, GRK5, ITPKA, SYNJ2, PIP5K1A, PAK3, SYNJ1, PIP5K1C, GRK6, PIP4K2A, PIP4K2C, CDK2.
PLK1, INPP5D, SYNJ2, INPP4A, INPP4B, INPP4B, PAK3, PIM1, PRKCD, PIK3C3, PRKAA2, PIK3CB, PIK3R2, PIP4K2A, PLCD4, PI4KA, ATM.
PPP2R2A, SOX10, CSNK1E, WIF1, WNT7A, TGFB1, SMO, CSNK2A1, CSNK2B, CTNNB1, SRC, SFRP4, GJA1, CSNK1G2, DVL1, FZD9, PPP2CB, CDH2, CDH1, PPP2R1A, PPP2R2B, TLE3, PPP2R5E, UBC, PPP2R1B, WNT5A.
CSNK1G1, CSNK1G3, DVL1, ACVR1, APC, MYC, CSNK1E, PPP2R1A, CDH2, CDH1, WIF1, TGFB1, GNAO1, TLE3, UBC, CTNNB1, SOX5 (includes EG:6660).
TRPC2, GRIN2A, TNNI2, GRIA1, RCAN2, CALM2, CHRNB4, TRPC5, TNNT3, RYR3, GRIK1, GRIN2B, PRKAR2A, TRPC4, PPP3CC, TRPC7, CHRNG, PRKAR2B, TPM3, CAMKK1, PRKACA, TNNI1, CHRNB3, CAMKK2, PRKAR1A.
RAP2B, MYH10, CALR, MYH6, ATP2B1, PRKAR2A, TPM4, PPP3CC, ATP2B2, HDAC6, TRPV6, HTR3A, CAMKK1, PRKACA, MYH9, SLC8A1, ACTA1, PRKAR1A.
TJP2, CDC42, TJP1, PPP2R2A, CLDN19, VAPA, PRKAR2A, PRKCZ, FOS, PPP2CB, PPP2R1A, PRKCI, PRKAR2B, CLDN4, CLDN1, TGFB1, PPP2R2B, PRKACA, PPP2R5E, STX4, CTNNB1, PPP2R1B, PRKAR1A.
MYH10, MYH6, CDC42, ACTB, HSF1, CLDN18, VAPA, RAC1, PRKAR2A, MPDZ, PPP2R1A, TGFB1, RHOA, CEBPA, PRKACA, MYH9, CTNNB1, ACTA1, PRKAR1A.
PRKCQ, EGF, PRKCZ, TLR9, TLR4, TLR5, PIK3CG, TLR6, PRKACA, CSNK2A1, PDGFRA, TLR3, CSNK2B, PRKCB, EGFR.
IL1R2, TLR5, BCL10, PIK3C3, PRKACA, IL1B, PIK3CB, PIK3R2, NFKB1, EGFR.
PTK2, SHC1, CDC42, PIK3CG, CSNK2A1, PDGFRA, CSNK2B, CDKN1B, PRKCZ, FASLG, EGFR.
CDC42, YWHAH, RAC1, PIK3CB, PIK3R2, CDKN1B, NFKB1, INPP5D, EGFR, MAGI3.
SHC1, CDC42, GADD45A, CRKL, PIK3CG, CRK, ELK1, GNG7.
CDC42, GADD45A, IRS1, PIK3C3, RAC1, PIK3CB, PIK3R2, ELK1, HNRNPK.
TSC1, PPP2R2A, YWHAB, PRKCZ, PPP2CB, SHC1, PPP2R1A, PIK3CG, TSC2, PPP2R2B, PPP2R5E, CDKN1B, CTNNB1, PPP2R1B.
TSC1, JAK1, YWHAH, JAK2, NOS3, INPP5D, YWHAQ (includes EG:10971), MTOR, PPP2R1A, HSP90AB1, PIK3CB, PIK3R2, CDKN1B, CTNNB1.
IL1R2, MYC, TGFB1, MAP3K7IP2, IL1B, PLA2G2A, EEF2K, MAPKAPK2, ELK1.
GADD45A, FASN, PIK3CG, SNAI2, CTNNB1, CDK2, ATM, SERPINE2.
CCND2, GADD45A, PIK3C3, SNAI2, CABC1, PIK3CB, PIK3R2, BAX, CTNNB1, ATM.
SHC1, SOCS1, PTPN11, PIK3CG, STAT3, STAT1.
STAT5A, MTOR, JAK1, PTPN11, PIK3C3, PIK3CB, STAT3, PIK3R2, JAK2.
Neuronal function pathway clusters of extracted proteins from primary cortical tissue. Specific neuronal function pathway clusters generated in an un-biased manner using Ingenuity Pathway Analysis (v. 8.5). The relative score generated for the degree of pathway population by proteins from the respective input sets (control or 3xTgAD) is shown in bold. The following are different types of signaling pathways.
GRIN2B, GRIN2A, PRKCQ, PPP1R1A, GRIA1, PRKAR2A, CALM2, CACNA1C, PPP3CC, GRM4, PPP1R14B, PRKCZ, GRM5, PRKCI, PRKAR2B, PPP1R10, PPP1R7, PPP1R14D, PRKACA, PLCB1, PRKD1, PRKCA, PRKAR1A, PRKCB
NOS1, GUCY1B2, PRKCQ, GUCY1A3, GUCY2D, PPP2R2A, GRID2, GRIA1, GRM4, PRKCZ, GRM5, PPP2CB, PPP2R1A, PRKCI, PPP2R2B, RYR3, GUCY1A2, PLCB1, PPP2R5E, PPP2R1B, PRKD1, PRKCA, PRKCB.
NOS1, GUCY2D, GNAI1, PLA2G2A, GNAZ, NOS3, PRDX6, GNAI2, PPP2R1A, PRKCD, GNAO1, IGF1R, GUCY2F, ADCY8, NPR2.
NOS1, GUCY1B2, PRKCQ, GUCY1A3, GUCY2D, PPP2R2A, GRID2, GRIA1, GRM4, PRKCZ, GRM5, PPP2CB, PPP2R1A, PRKCI, PPP2R2B, RYR3, GUCY1A2, PLCB1, PPP2R5E, PPP2R1B, PRKD1, PRKCA, PRKCB.
NOS1, GUCY2D, GNAI1, PLA2G2A, GNAZ, NOS3, PRDX6, GNAI2, PPP2R1A, PRKCD, GNAO1, IGF1R, GUCY2F, ADCY8, NPR2.
UCHL1, PARK7, SNCA.
UCHL1, PARK7, SNCA.
EGF, GNG7, PRKCZ, TGM2, CTSD, SHC1, PIK3CG, VAMP3, PLCB1, PRKD1, PRKCA, EGFR, SDHA, GRIN2B, PRKCQ, YKT6, SH3GL3, STX1A, SNAP25, TAF9B, GRM5, PRKCI, UBC, GOSR2, SNCA, PRKCB.
VTI1A, HSPA14, REST, GNB2L1, HDAC6, NSF, MTOR, PIK3C3, IGF1R, PIK3R2, EGFR, CAPN5, CAPN6, CLTC, BAX, STX1A, SNAP25, DNM1, DNAJC5, ATP5B, CAPN1, PRKCD, HAP1, DCTN1, PIK3CB, UBC, SNCA, CASP7.
PIP5K1A, WIPF1, PAK1, PAK2, CDC42, CFL1, PAK3, PIP5K1C, PIP4K2A, PIP4K2C, PI4KA.
CDC42, CFL1, ACTB, RAC1, RHOJ, ROCK1, PAK3, RHOB, RHOA, ARHGDIA, PIP4K2A, ACTA1, PI4KA.
NOS1, GRIN2B, GRIN2A, SOD1, GRIA1, PGF, VEGFA, PAK1, GRIK4, PIK3CG, CAT, GLUL, SSR4.
NOS1, CAPN5, CAPN6, SOD1, RAB5A, RAC1, NEFH, BAX, CCS, VEGFA, CAPN1, PIK3C3, NEFM, PIK3CB, PIK3R2, CASP7.
PAK2, CFL1, CDC42, CRKL, EGF, CRK, TTN, PTK2, SHC1, PAK1, PIP5K1A, PAK3, PIP5K1C, FGF18, PIK3CG, EZR, FGF23, PIP4K2A, PIP4K2C, GIT1.
MYH10, MYH6, CDC42, ROCK2, PIK3C3, EZR, PIK3R2, LBP, ACTA1, PXN, ARHGEF12, CFL1, ACTB, RAC1, RDX, APC, TTN, ROCK1, PAK3, RHOA, ARHGEF6, MYH9, VAV1, PIK3CB, ACTN4, PIP4K2A, MSN.
CSNK1E, PRKAR2B, CSNK2A, PRKAR2A, PRKACA, NCSTN, CSNK2B, PRKAR1A.
CSNK1E, CAPN5, CAPN6, CAPN1, MARK1, PRKAR2A, PRKACA, NCSTN, BACE1, APP, PRKAR1A.
Energy regulation/Metabolism pathway clusters of extracted proteins from primary cortical tissue. Specific energy regulation/metabolism pathway clusters generated in an un-biased manner using Ingenuity Pathway Analysis (v. 8.5). The relative score generated for the degree of pathway population by proteins from the respective input sets (control or 3xTgAD) is shown in bold. The following are different types of signaling pathways.
FMO3, PDE7A, PDE10A, GM2A, CYB5R1, PDE4A, PDE4B, PDE1A, PDE4D, GFPT2, PDE1C, PDE7B, GALK1, CYB5R3.
HK1, CYB5R4, PDE10A, GALK1, CYB5R3, PDE4B.
GPI, PGD, TKT, TALDO1, PRPSAP1, PGM1, PRPSAP2, PFKP, PFKL, PDHB, PFKM.
GPI, TKT, PRPSAP2, PFKL.
FMO3, GPT, GLUL, GCLC, GLUD1, GOT1, GOT2, GFPT2, EPRS.
ENPP3, CRMP1, BCAT1, DPYD, DPYSL3, ENPP5, ENPP2.
CRMP1, DPYD, ENPP2.
PKM2, PGK1, ENO3, ENO2, GAPDH (includes EG:2597), PGM1, ALDH1L1, PFKP, PFKL, PDHB, PFKM, GPI, GALK1, DLAT, PGAM1, DLD.
PKM2, HK1, ADH1A, GPI, ALDH1A3, DLAT, GALK1, ENO2, PFKL, ACSL1.
SDHA, PC, CS, SUCLG1, DLD, PCK1.
GLA, GALT, GALK1, GAA, PGM1, PFKP, PFKL, PFKM.
HK1, GLA, GALK1, GAA, PFKL, AKR1B1.
GLA, ST3GAL1, GM2A.
RPN2, ST6GAL2, MGAT5, MGAT1.
UCHL1, B3GAT1, HPSE, DCXR, AKR1B1.
GPAM, PCYT1B, PLD3, CDS1, PDIA3, GNPAT, DGKB, DGKG, PLCB1, PLCL1.
AGPAT4, CDS1, GNPAT, CHKA, AGPAT1, DGKG, PLA2G2A, CLCF1, PHKA1, PLCD4, PRDX6.
ENTPD8, PDE7A, PDE4A, DNTT, PDE1A, PDE4D, PSMC5, PDE7B, VCP, DHX16, ENPP2, PRPSAP2, PKM2, TJP2, GUCY1B2, GUCY1A3, GUCY2D, PDE10A, ENPP5, PDE4B, PDE1C, ENPP3, PSMC1, ENTPD2, PRPSAP1, GUCY1A2, PDE5A, DDX1, GMPR2, PSMC3.
MYH6, ENTPD8, NME2, DNTT, WRNIP1, ABCC1, VCP, KATNA1, ATP5H (includes EG:10476), ENPP2, PRPSAP2, ADCY8, ATP5F1, ATP5I, ATP5J, PKM2, ATP1B1, GUCY2D, PDE10A, AMPD1, ATP5A1, AK3, PSMC4, PDE4B, DDX19B, ATP5B, GUCY2F, MYH9, DDX1, NPR2.
PKM2, PC, DLAT, DLD, PCK1, PDHB, GLO1.
PKM2, ACACB, ALDH1A3, DLAT, ACAT1, ACACA, MDH1, AKR1B1, ACSL1.
ECHS1, ALDH1A3, AACS, ACAT1, MYO5B, DCXR, HMGCS1.
Cellular stress/damage pathway clusters of extracted proteins from primary cortical tissue. Specific energy cellular stress/damage pathway clusters generated in an un-biased manner using Ingenuity Pathway Analysis (v. 8.5). The relative score generated for the degree of pathway population by proteins from the respective input sets (control or 3xTgAD) is shown in bold. The following are different types of signaling pathways.
PRDX1, PPIB, GCLC, PRKCZ, GSTM2, SOD2, PIK3CG, VCP, DNAJA2, TXN, FKBP5, PRKD1, CBR1, PRKCA, GSTK1, SOD1, PRKCQ, TXNRD1, GSTO1, FOS, PRKCI, ERP29, STIP1, CAT, PRKCB, EPHX1.
USP14, SOD1, PRDX1, ACTB, GSTM3 (includes EG:2947), JUNB, DNAJC5, GSTM2, SOD2, ERP29, STIP1, PIK3C3, PRKCD, ABCC1, VCP, PIK3CB, FOSL1, PIK3R2, TXN, EIF2AK3, ACTA1.
CTSD, PPP2CB, FOS, PPP2R1A, PPP2R2A, PIK3CG, PPP2R2B, PPP2R5E, PPP2R1B.
OPLAH, PGD, GSTM2, GGT5, GCLC, GGT1, GSTO1, GSTK1, GGT7.
SULT1C2, FMO3, CPT1A, CHST7, ALDH1L1, NR1H3, GSTO1, SULT4A1, GSTM2, NR1I2, SREBF1, SULT1A1, CAT, CPT2, FABP4, XPO1, FABP1, FABP7, SULT1E1, FABP3, GSTK1, TLR4, FABP2.
PPARA, GSTM3 (includes EG:2947), ABCA1, IL1R2, ABCC2, SLC27A5, GSTM2, ALDH1A3, NR1I2, CPT2, ACSL4, FABP1, IL1B, NR5A2, CHST10, LBP, ABCC4, HMGCS1, ACSL1, MAOA.
ACIN1, ROCK1, CAPN5, CAPN6, CAPN1, BAX, CASP7, PARP1.
VEGFA, P4HB, HSP90AB1, UBE2L3, NOS3, UBC, ARNT, ATM.
PDIA3, PRKAR2A, NR2C2, SHC1, PRKAR2B, TGFB1, FASN, PRKACA, PLCB1, GOT2, PLCL1, PRKCB, PRKCA, PRKAR1A.
PPARA, ACVR1, PRKAR2A, CD36, IL6, JAK2, ABCA1, IL1R2, ACADL, HSP90AB1, TGFB1, IRS1, PRKACA, SMAD4, IL1B, ADCY8, PLCD4, PRKAR1A.
Receptor signaling pathway clusters of extracted proteins from primary cortical tissue. Specific receptor signaling pathway clusters generated in an un-biased manner using Ingenuity Pathway Analysis (v. 8.5). The relative score generated for the degree of pathway population by proteins from the respective input sets (control or 3xTgAD) is shown in bold. The following are different types of signaling pathways.
PPP2R2A, PRKAR2A, PPP1R14B, PPP2CB, PPP2R1A, PRKAR2B, PPP1R10, PPP1R14D, PPP1R7, PPP2R2B, PRKACA, FREQ, PPP2R5E, PPP2R1B, PRKAR1A, CALY.
PPP2R1A, FREQ, PRKAR2A, PRKACA, ADCY8, CALY, MAOA, PRKAR1A.
SRC, CRKL, CRK, STAT3, FOS, SHC1, PIK3CG, CSNK2A1, PDGFRA, CAV1, CSNK2B, ELK1, STAT1, PRKCB, PRKCA.
MYC, JAK1, PIK3C3, STAT3, PIK3R2, JAK2, ELK1, INPP5D.
SHC1, FOS, PIK3CG, CSNK2A1, EGF, CSNK2B, STAT3, STAT1, ELK1, PRKCA, EGFR.
JAK1, PIK3C3, PIK3CB, STAT3, PIK3R2, ELK1, EGFR.
SRC, PRKCQ, CRKL, EGF, CRK, ERBB3, PRKCZ, SHC1, PRKCI, PTPN11, ERBB4, CDKN1B, ELK1, PRKD1, PRKCB, EGFR, PRKCA.
MYC, STAT5A, MTOR, PTPN11, HSP90AB1, PRKCD, PIK3R2, CDKN1B, ELK1, EGFR.
GRM5, GRIN2B, GRIN2A, GRIK4, GRIA1, GRID2, CALM2, GLUL, GRIP1, GRM4, GNG7, GRIK1.
SHH, CCNB1, PRKAR2B, PRKAR2A, PRKACA, SMO, GLI1, PRKAR1A.
ADRBK1, PTCH1, PRKAR2A, PRKACA, PRKAR1A.
UCP2, RCAN2, UCP1, THRA, GRIP1, PFKP, PCK1, KLF9, SREBF1, DIO1, FASN, PIK3CG, STRBP, THRB, FGA, SYT12.
MTOR, PIK3C3, STRBP, ACACA, PIK3CB, PIK3R2, THRB.
GRIN2A, CDC42, EGF, CRK, GNG7, PGF, VEGFA, PTK2, SHC1, PAK1, SDC2, PIK3CG, EFNB1, EPHA7, GRIN2B, SRC, PAK2, CFL1, CRKL, EPHA3, STAT3, WIPF1, SDCBP, PTPN11, PAK3, EPHA5, EPHB3.
FYN, PXN, CFL1, CDC42, KALRN, GNB2L1, GNAI1, RAC1, JAK2, STAT3, GNAZ, ROCK1, VEGFA, GNAI2, ROCK2, ABI1, PTPN11, PAK3, RHOA, GNAO1.
PDE7A, PDE4A, PDE1A, PDE4D, SHC1, SYNGAP1, PDE7B, PIK3CG, PLCB1, PRKCA, SRC, GRK4, EDNRB, PDE10A, PRKAR2A, PDE4B, STAT3, GRM4, PDE1C, CHRM5, GRM5, OPRD1, PRKAR2B, PRKACA, PRKCB, PRKAR1A.
FYN, PTK2B, HTR4, ADRBK1, PDE10A, RGS7, PRKAR2A, GNAI1, RGS4, STAT3, PDE4B, GNAI2, HTR2C, PIK3C3, GNAO1, PRKACA, PIK3CB, PIK3R2, ADCY8, PRKAR1A.
PTK2, NOX1, SRC, FOS, CCL4, CFL1, PIK3CG, CALM2, PLCB1, PRKCA, PRKCB.
ROCK2, GNAI2, CFL1, PTK2B, RHOA, GNAI1.
CTSD, PPP2CB, FOS, PPP2R1A, PPP2R2A, PIK3CG, PPP2R2B, PPP2R5E, PPP2R1B, PRKCZ.
FGFR3, PTPN11, FGF18, CRKL, PIK3CG, FGFR1, FGF23, FGFR2, CRK, STAT3, PRKCA.
PTPN11, PIK3C3, HGF, RAC1, PIK3CB, STAT3, PIK3R2, MAPKAPK2.
TLR4, FOS, TLR5, TLR6, TLR3, ELK1, TLR9.
PPARA, TLR5, MAP3K7IP2, LBP, ELK1.
GUCY1B2, GUCY1A3, GUCY2D, CALM2, PRKAR2A, PGF, VEGFA, PRKAR2B, PIK3CG, GUCY1A2, CAV1, PRKACA, PRKAR1A.
VEGFA, HSP90AB1, GUCY2D, PRKCD, PIK3C3, GUCY2F, PRKAR2A, PRKACA, PIK3CB, PIK3R2, NOS3, PRKAR1A.
GRK4, SRC, CNGA4, PDE7A, PDE10A, PRKAR2A, CALM2, PDE4A, PPP3CC, STAT3, PDE4B, GRM4, PDE1A, PDE4D, CNGA1, PDE1C, CHRM5, OPRD1, PRKAR2B, PDE7B, PRKACA, PKIA, PRKAR1A.
AKAP12, HTR4, ADRBK1, PDE10A, RGS7, PRKAR2A, GNAI1, RGS4, AKAP3, PPP3CC, STAT3, PDE4B, CNGA1, GNAI2, AKAP4, GNAO1, PRKACA, CNGB1, PKIA, ADCY8, AKAP9, PRKAR1A.
TGFB1, ACVR1, SMAD4, HNF4A, UBC, INHBC.
SHC1, FOS, CDC42, PTPN11, PIK3CG.
NTRK2, CDC42, PTPN11, PIK3C3, PIK3CB, PIK3R2.
PRKAR2A, PTK2, SHC1, FOS, PRKAR2B, PTPN11, PIK3CG, PRKACA, CSNK2A1, CSNK2B, ELK1, PRKAR1A.
PXN, YWHAH, PRKAR2A, NEDD4, YWHAQ, PTPN11, IRS1, PIK3C3, PRKACA, IGF1R, IRS2, PIK3CB, PIK3R2, ELK1, PRKAR1A.
HTR2C, HTR4, HTR3A, MAOA.
STAT5A, JAK1, YWHAH, HSPA14, RAC1, IL6, PPP3CC, STAT3, JAK2, NFKB1, MNAT1, IL1R2, HSP90AB1, TGFB1, PIK3C3, CEBPA, PRKACA, SMAD4, IL1B, PIK3CB, PIK3R2, ELK1.
SRC, CDK6, ALDH1L1, GSTO1, TGM2, CTSD, FOS, GSTM2, TGFB1, FASN, CDKN1B, ESR2, CDK2, FASLG, ATM, GSTK1.
CYP1A1, NFIX, GSTM3, IL6, BAX, NFKB1, ARNT, MYC, AHRR, GSTM2, CCND2, HSP90AB1, TGFB1, ALDH1A3, IL1B, CDKN1B, AHR, MCM7, ATM.
TSC1, TRIP10, CRKL, PRKAR2A, CRK, VAMP2, PPP1R14B, PRKCZ, SHC1, PRKCI, PPP1R10, PTPN11, PIK3CG, TSC2, PRKACA.
TSC1, FYN, JAK1, PRKAR2A, JAK2, VAMP2, INPP5D, MTOR, PTPN11, IRS1, PIK3C3, PRKACA, EIF2B1, PIK3CB, IRS2, PIK3R2, PRKAR1A.
DLL1, CNTN1, NCSTN, DLL3.
NOTCH4, DLL1, NOTCH2, NCSTN, JAG1, NOTCH1, HEY1.
GABRG2, UBQLN1, GABBR1, GABARAP, UBC, GABRA3.
DNM1, NSF, SLC32A1, UBQLN1, GABBR1, MYO5B, GABRE, UBC, GABRA3.
VEGFA, PTK2, SHC1, PIK3CG, PGF, PRKCA, PRKCB.
EIF2S2, PXN, PTK2B, ACTB, NOS3, ARNT, ROCK2, VEGFA, ROCK1, PIK3C3, EIF2B1, PIK3CB, PIK3R2, ACTN4, ACTA1.
RAP2B, FYN, RALA, CDC42, TSPAN3, ARF6, RHOB, PIK3C3, ARF4, PIK3R2, ACTA1, CAPN5, CAPN6, PXN, ACTB, RALB, RAC1, ITGA6, RHOJ.
The individual position numbers are correlated to the specific multidimensional proteins represented by colored blocks in Figure 6. The following are the symbols of the proteins (the numbers between brackets refer to the position).
Receptor-specific protein interaction networks in lipid raft extracts from control animals. IPA-generated receptor-specific protein lists from control lipid raft samples were clustered into coherent functional interaction networks. Focus molecules (BOLD in “molecules in network”) denote the proteins that are present in the predicted reaction network as well as the experimental input protein set. The following are different types of interaction networks.
(1) Control: cell signaling, nucleic acid metabolism, small molecule biochemistry; score: 32; focus molecule: 15; molecules in network: ADCY, Akt, Beta-Arrestin, CD4, CHRM5, CNTFR, Creb, EDNRB, ERK1/2, FSHR, GABBR1, GCGR, Gpcr, GPR3, GPR12, GPR20, GPR34, GPR65, GPR161, GRM5, hCG, Mapk, OPRD1, OVGP1, P2RY6, P2RY11, PDGF BB, PI3K, Pkc(s), PTGER2, SFRP4, SSTR2, THBD, UNC5B, VIPR2.
(2) Control: infectious disease, antigen presentation, antimicrobial response; score: 31; focus molecule: 15; molecules in network: Ap1, CHRNB3, CHRNB4, CHRNG, CLDN4, CNR1, CXADR, G alphai, Ifn, IFN Beta, IFN TYPE 1, IgG, IKK (complex), IL-1R/TLR, IL12 (complex), Il12 (family), Interferon alpha, IRAK, IRF, NFkB (complex), NRG, OSMR, P38 MAPK, SMO, TACR1, Tlr, TLR3, TLR4, TLR5, TLR6, TLR9, TSHR, Ubiquitin.
(3) Control: cellular development tumor morphology, cell death; score: 18; focus molecule: 10; molecules in network: APOL5, ARR3, ARTN, ASB16, Cadherin (E,N,P,VE), CELSR2, CELSR3, CTNNB1, CTNNβ-CDHE/N, DAG1, EVX1, FZD9, GDNF, GFRA1, GFRA3, GLTSCR1, GPRC6A, GRB2, GRM4, HRH4, KCNN1, OPN1LW, OPN1SW, PALM2-AKAP2, PHACTR2, PRICKLE3, PTH2R, SAG, SEPN1, SHROOM2, SRC.
(4) Control: carbohydrate metabolism; score: 2; focus molecule: 1; molecules in network: CLEC4A, IL13.
(5) Control: cell-to-cell signaling and interaction, nervous system development and function; score: 2; focus molecule: 1; molecules in network: PRPH2, ROM1.
(6) Control: protein synthesis, molecular transport, protein trafficking; score: 2; focus molecule: 1; molecules in network: GABRR3, PRKCZ, SQSTM1.
Receptor-specific protein interaction networks in lipid raft extracts from 3xTgAD animals. IPA-generated receptor-specific protein lists from 3xTgAD lipid raft samples were clustered into coherent functional interaction networks. Focus molecules (BOLD in “molecules in network”) denote the proteins that are present in the predicted reaction network as well as the experimental input protein set. The following are different types of interaction networks.
(1) 3xTgAD: metabolic disease, endocrine system disorders, cell signaling; score: 45; focus molecule: 18; molecules in network: ADCYAP1R1, Ap1, CD36, CD86, CNR1, Creb, CREB-NFkB, CXADR, ERK, ERK1/2, GABBR1, GPR56, hCG, HTR4, HTR2C, IGF1R, IGF2R, IgG, IL12 (complex), IL1R2, Insulin, Jnk, LDL, LIFR, LRPAP1, Mapk, MIP1, NFkB (complex), NPR2, OVGP1, P38 MAPK, Pkc(s), TLR5, TNFRSF14, TSHR.
(2) 3xTgAD: cell signaling, nucleic acid metabolism, small molecule biochemistry; score: 13; focus molecule: 7; molecules in network: AATK, ABR, AKR1A1, BPI, CACNG5, CRYM, cyclic AMP, CYP26B1, DHRS3, DLG4, GH1, Histone h3, HSD17B11, KCTD11, KIF3C, LAP3, LPHN1, LPHN2, OPN1LW, OPN1SW, P2RY11, PTCH1, PTH2R, RLN3, RN5S, ROBO1, ROS1, RXFP1, RXFP2, SERPINB8, TMEM49, TNF, USP3.
(3) 3xTgAD: cell signaling, cellular function and maintenance, molecular transport; score: 2; focus molecule: 1; molecules in network: CFTR, FREQ, IL1RAPL1, MYD88.
(4) 3xTgAD: cell-to-cell signaling and interaction, cellular function and maintenance, cellular movement; score: 2; focus molecule: 1; molecules in network: GPR1, PAX3, PRDM5.
(5) 3xTgAD: behavior, digestive system development and function, cell morphology; score: 2; focus molecule: 1; molecules in network: NPY, NPY5R, PPY, PYY, SSB.
(6) 3xTgAD: cancer, reproductive system disease, gene expression; score: 2; focus molecule: 1; molecules in network: FOS, MIR103-1, MIR103-2, MIR107, MIRLET7G, MYC, OMG, RHOA, RTN4R.