Alzheimer’s disease (AD) lacks protein biomarkers that reflect its multiple underlying pathophysiology, hindering the progress of diagnosis and treatment. Here, we use comprehensive proteomics to identify cerebrospinal fluid (CSF) biomarkers that represent a broad range of AD pathophysiology. Multiplex mass spectrometry identified approximately 3,500 and approximately 12,000 proteins in AD CSF and the brain, respectively. The network analysis of the brain proteome resolved 44 biodiversity modules, 15 of which overlapped with the cerebrospinal fluid proteome. The CSF AD markers in these overlapping modules are folded into five protein groups, representing different pathophysiological processes. The synapses and metabolites in the AD brain decrease, but the CSF increases, while the glial-rich myelination and immune groups in the brain and CSF increase. The consistency and disease specificity of the panel changes were confirmed in more than 500 additional CSF samples. These groups also identified biological subgroups in asymptomatic AD. Overall, these results are a promising step towards web-based biomarker tools for clinical applications in AD.
Alzheimer’s disease (AD) is the most common cause of neurodegenerative dementia worldwide and is characterized by a wide range of biological system dysfunctions, including synaptic transmission, glial-mediated immunity, and mitochondrial metabolism (1-3). However, its established protein biomarkers still focus on detecting amyloid and tau protein, and therefore cannot reflect this diverse pathophysiology. These “core” protein biomarkers that are most reliably measured in cerebrospinal fluid (CSF) include (i) amyloid beta peptide 1-42 (Aβ1-42), which reflects the formation of cortical amyloid plaques; (ii) total tau, a sign of axon degeneration; (iii) phospho-tau (p-tau), a representative of pathological tau hyperphosphorylation (4-7). Although these cerebrospinal fluid biomarkers have greatly facilitated our detection of “marked” AD protein diseases (4-7), they only represent a small part of the complex biology behind the disease.
The lack of pathophysiological diversity of AD biomarkers has led to many challenges, including (i) the inability to identify and quantify the biological heterogeneity of AD patients, (ii) insufficient measurement of disease severity and progression, especially in the preclinical stage, And (iii) the development of therapeutic drugs that failed to completely solve all aspects of neurological deterioration. Our reliance on landmark pathology to describe AD from related diseases only exacerbates these problems. More and more evidences show that most elderly people with dementia have more than one pathological characteristic of cognitive decline (8). As many as 90% or more of individuals with AD pathology also have vascular disease, TDP-43 inclusions, or other degenerative diseases (9). These high proportions of pathological overlap have disrupted our current diagnostic framework for dementia, and a more comprehensive pathophysiological definition of the disease is needed.
In view of the urgent need for a variety of AD biomarkers, the field is increasingly adopting the “omics” method based on the overall system to discover biomarkers. The Accelerated Pharmaceutical Partnership (AMP)-AD Alliance was launched in 2014 and is at the forefront of the program. This multidisciplinary effort by the National Institutes of Health, academia, and industry aims to use system-based strategies to better define the pathophysiology of AD and develop biodiversity diagnostic analysis and treatment strategies (10). As part of this project, network proteomics has become a promising tool for the advancement of system-based biomarkers in AD. This unbiased data-driven approach organizes complex proteomics data sets into groups or “modules” of co-expressed proteins that are associated with specific cell types, organelles, and biological functions (11-13). Almost 12 information-rich network proteomics studies have been conducted on the AD brain (13-23). Overall, these analyses indicate that the AD brain network proteome maintains a highly conserved modular organization in independent cohorts and multiple cortical regions. In addition, some of these modules show reproducible changes in AD-related abundance across data sets, reflecting the pathophysiology of multiple diseases. Collectively, these findings demonstrate a promising anchor point for the discovery of the brain network proteome as a system-based biomarker in AD.
In order to transform the AD brain network proteome into clinically useful system-based biomarkers, we combined the brain-derived network with the proteomic analysis of AD CSF. This integrated approach led to the identification of five promising sets of CSF biomarkers that are associated with a wide range of brain-based pathophysiology, including synapses, blood vessels, myelination, inflammation, and dysfunction of metabolic pathways. We successfully validated these biomarker panels through multiple replication analyses, including more than 500 CSF samples from various neurodegenerative diseases. These validation analyses include examining group targets in the CSF of patients with asymptomatic AD (AsymAD) or showing evidence of abnormal amyloid accumulation in a normal cognitive environment. These analyses highlight the significant biological heterogeneity in the AsymAD population and identify panel markers that may be able to subtype individuals in the earliest stages of the disease. Overall, these results represent a key step in the development of protein biomarker tools based on multiple systems that can successfully solve many of the clinical challenges faced by AD.
The main purpose of this study is to identify new cerebrospinal fluid biomarkers that reflect various brain-based pathophysiology that lead to AD. Figure S1 outlines our research methodology, which includes (i) a comprehensive analysis driven by preliminary findings of AD CSF and the network brain proteome to identify multiple brain-related CSF disease biomarkers, and (ii) subsequent replication These biomarkers are in several independent cerebrospinal fluid cohorts. The discovery-oriented research started with the analysis of the differential expression of CSF in 20 cognitively normal individuals and 20 AD patients at Emory Goizueta Alzheimer’s Disease Research Center (ADRC). The diagnosis of AD is defined as a significant cognitive impairment in the presence of low Aβ1-42 and elevated levels of total tau and p-tau in the cerebrospinal fluid [Mean Montreal Cognitive Assessment (MoCA), 13.8 ± 7.0] [ELISA (ELISA)]] (Table S1A). The control (mean MoCA, 26.7 ± 2.2) had normal levels of CSF biomarkers.
Human CSF is characterized by a dynamic range of protein abundance, in which albumin and other extremely abundant proteins can prevent the detection of proteins of interest (24). To increase the depth of protein discovery, we removed the first 14 highly abundant proteins from each CSF sample before mass spectrometry (MS) analysis (24). A total of 39,805 peptides were identified by MS, which were mapped to 3691 proteomes in 40 samples. Protein quantification is performed by multiple tandem mass tag (TMT) labeling (18, 25). In order to resolve the missing data, we included only those proteins that were quantified in at least 50% of the samples in the subsequent analysis, thus finally quantifying 2875 proteomes. Due to the significant difference in total protein abundance levels, a control sample was statistically considered an outlier (13) and was not included in the subsequent analysis. The abundance values of the remaining 39 samples were adjusted according to age, gender, and batch covariance (13-15, 17, 18, 20, 26).
Using statistical t-test analysis to evaluate differential expression on the regression data set, this analysis identified proteins whose abundance levels were significantly changed (P <0.05) between the control and AD cases (Table S2A). As shown in Figure 1A, the abundance of a total of 225 proteins in AD was significantly reduced, and the abundance of 303 proteins was significantly increased. These differentially expressed proteins include several previously identified cerebrospinal fluid AD markers, such as microtubule-associated protein tau (MAPT; P = 3.52 × 10−8), neurofilament (NEFL; P = 6.56 × 10−3), growth-related Protein 43 (GAP43; P = 1.46 × 10−5), Fatty Acid Binding Protein 3 (FABP3; P = 2.00 × 10−5), Chitinase 3 like 1 (CHI3L1; P = 4.44 × 10−6), Neural Granulin (NRGN; P = 3.43 × 10−4) and VGF nerve growth factor (VGF; P = 4.83 × 10−3) (4-6). However, we also identified other very important targets, such as GDP dissociation inhibitor 1 (GDI1; P = 1.54 × 10-10) and SPARC-related modular calcium binding 1 (SMOC1; P = 6.93 × 10-9) . Gene Ontology (GO) analysis of 225 significantly reduced proteins revealed close connections with body fluid processes such as steroid metabolism, blood coagulation, and hormone activity (Figure 1B and Table S2B). In contrast, the significantly increased protein of 303 is closely related to cell structure and energy metabolism.
(A) The volcano plot shows the log2 fold change (x-axis) relative to the -log10 statistical P value (y-axis) obtained by the t-test, which is used to detect differential expression between the control (CT) and AD cases of the CSF proteome Of all proteins. Proteins with significantly reduced levels (P <0.05) in AD are shown in blue, while proteins with significantly increased levels in disease are shown in red. The selected protein is labeled. (B) The top GO terms related to protein are significantly reduced (blue) and increased (red) in AD. Shows the three GO terms with the highest z-scores in the fields of biological processes, molecular functions, and cellular components. (C) MS measured MAPT level in CSF sample (left) and its correlation with sample ELISA tau level (right). The Pearson correlation coefficient with the relevant P value is displayed. Due to the lack of ELISA data for one AD case, these figures include values for 38 of the 39 analyzed cases. (D) Supervised cluster analysis (P <0.0001, Benjamini-Hochberg (BH) adjusted P <0.01) on the control and AD CSF found samples using the 65 most significantly changed proteins in the data set. Standardize, normalize.
The proteomic level of MAPT is closely related to the independently measured ELISA tau level (r = 0.78, P = 7.8 × 10-9; Figure 1C), supporting the validity of our MS measurement. After trypsin digestion at the level of amyloid precursor protein (APP), the isoform-specific peptides mapped to the C-terminus of Aβ1-40 and Aβ1-42 cannot be ionized efficiently (27, 28). Therefore, the APP peptides we identified have nothing to do with ELISA Aβ1-42 levels. In order to evaluate the differential expression of each case, we used differentially expressed proteins with P <0.0001 [false discovery rate (FDR) corrected P <0.01] to perform a supervised cluster analysis of the samples (Table S2A). As shown in Figure 1D, these 65 highly significant proteins can correctly cluster samples according to disease state, except for one AD case with control-like characteristics. Of these 65 proteins, 63 increased in AD, while only two (CD74 and ISLR) decreased. In total, these cerebrospinal fluid analyses have identified hundreds of proteins in AD that may serve as disease biomarkers.
Then we performed an independent network analysis of the AD brain proteome. The brain cohort of this discovery included dorsolateral prefrontal cortex (DLPFC) from control (n = 10), Parkinson’s disease (PD; n = 10), mixed AD/PD (n = 10) and AD (n = 10) cases. ) Sample. Emery Goizueta ADRC. The demographics of these 40 cases have been previously described (25) and are summarized in Table S1B. We used TMT-MS to analyze these 40 brain tissues and the replication cohort of 27 cases. In total, these two brain data sets produced 227,121 unique peptides, which were mapped to 12,943 proteomes (25). Only those proteins that were quantified in at least 50% of cases were included in subsequent investigations. The final discovery data set contains 8817 quantified proteins. Adjust protein abundance levels based on age, gender, and post-mortem interval (PMI). The differential expression analysis of the data set after regression showed that >2000 protein levels were significantly changed [P <0.05, analysis of variance (ANOVA)] in two or more disease cohorts. Then, we performed a supervised cluster analysis based on the differentially expressed proteins, and P <0.0001 in AD/control and/or AD/PD comparisons (Figure S2, A and B, Table S2C). These 165 highly altered proteins clearly depict cases with AD pathology from the control and PD samples, confirming the strong AD-specific changes in the entire proteome.
We then used an algorithm called Weighted Gene Co-expression Network Analysis (WGCNA) to perform network analysis on the discovered brain proteome, which organizes the data set into protein modules with similar expression patterns (11-13). The analysis identified 44 modules (M) co-expressed proteins, sorted and numbered from the largest (M1, n = 1821 proteins) to the smallest (M44, n = 34 proteins) (Figure 2A and Table S2D) ). As mentioned above (13) Calculate the representative expression profile or characteristic protein of each module, and correlate it with the disease state and AD pathology, that is, establish the alliance of Alzheimer’s Disease Registry (CERAD) and Braak Score (Figure 2B). Overall, 17 modules were significantly related to AD neuropathology (P <0.05). Many of these disease-related modules are also rich in cell type-specific markers (Figure 2B). As mentioned above (13), cell type enrichment is determined by analyzing the module overlap and the reference list of cell type-specific genes. These genes are derived from published data in isolated mouse neurons, endothelial and glial cells. RNA sequencing (RNA-seq) experiment (29).
(A) Discover the WGCNA of the brain proteome. (B) Biweight midcorrelation (BiCor) analysis of the modular signature protein (the first major component of modular protein expression) with AD neuropathological characteristics (top), including CERAD (Aβ plaque) and Braak (tau tangles) scores. The intensities of positive (red) and negative (blue) correlations are shown by a two-color heat map, and asterisks indicate statistical significance (P <0.05). Use Hypergeometric Fisher’s Exact Test (FET) (bottom) to assess the cell type association of each protein module. The intensity of the red shading indicates the degree of cell type enrichment, and the asterisk indicates statistical significance (P <0.05). Use the BH method to correct the P value derived from the FET. (C) GO analysis of modular proteins. The most closely related biological processes are shown for each module or related module group. oligo, oligodendrocyte.
A set of five closely related astrocyte and microglia-rich modules (M30, M29, M18, M24, and M5) showed a strong positive correlation with AD neuropathology (Figure 2B). Ontology analysis links these glial modules with cell growth, proliferation, and immunity (Figure 2C and Table S2E). Two additional glial modules, M8 and M22, are also strongly upregulated in disease. M8 is highly related to the Toll-like receptor pathway, a signaling cascade that plays a key role in the innate immune response (30). At the same time, M22 is closely related to post-translational modification. M2, which is rich in oligodendrocytes, shows a strong positive correlation with AD pathology and an ontological connection with nucleoside synthesis and DNA replication, indicating enhanced cell proliferation in diseases. Overall, these findings support the elevation of glial modules that we have previously observed in the AD network proteome (13, 17). It is currently found that many AD-related glial modules in the network show lower expression levels in control and PD cases, highlighting their disease specificity that is elevated in AD (Figure S2C).
Only four modules in our network proteome (M1, M3, M10, and M32) are strongly negatively correlated with AD pathology (P <0.05) (Figure 2, B and C). Both M1 and M3 are rich in neuronal markers. M1 is highly related to synaptic signals, while M3 is closely related to mitochondrial function. There is no evidence of cell type enrichment for M10 and M32. M32 reflects the connection between M3 and cell metabolism, while M10 is highly related to cell growth and microtubule function. Compared with AD, all four modules are increased in control and PD, giving them disease-specific AD changes (Figure S2C). Overall, these results support the decreased abundance of neuron-rich modules that we have previously observed in AD (13, 17). In summary, the network analysis of the brain proteome that we discovered produced AD-specifically altered modules consistent with our previous findings.
AD is characterized by an early asymptomatic stage (AsymAD), in which individuals exhibit amyloid accumulation without clinical cognitive decline (5, 31). This asymptomatic stage represents a critical window for early detection and intervention. We have previously demonstrated strong modular preservation of AsymAD and AD brain network proteome across independent data sets (13, 17). In order to ensure that the brain network we currently discovered is consistent with these previous findings, we analyzed the preservation of 44 modules in the replicated data set from 27 DLPFC organizations. These organizations include control (n = 10), AsymAD (n = 8 ) And AD (n = 9) cases. Control and AD samples were included in the analysis of our discovery brain cohort (Table S1B), while AsymAD cases were unique only in the replication cohort. These AsymAD cases also came from the Emory Goizueta ADRC brain bank. Although cognition was normal at the time of death, amyloid levels were abnormally high (mean CERAD, 2.8 ± 0.5) (Table S1B).
TMT-MS analysis of these 27 brain tissues resulted in the quantification of 11,244 proteomes. This final count includes only those proteins quantified in at least 50% of the samples. This replicated data set contains 8638 (98.0%) of the 8817 proteins detected in our discovery brain analysis, and has nearly 3000 significantly changed proteins between the control and AD cohorts (P <0.05, after Tukey’s paired t test for analysis of variance) (Table S2F). Among these differentially expressed proteins, 910 also showed significant level changes between AD and brain proteome control cases (P <0.05, after ANOVA Tukey paired t-test). It is worth noting that these 910 markers are highly consistent in the direction of change between proteomes (r = 0.94, P <1.0 × 10-200) (Figure S3A). Among the increased proteins, the proteins with the most consistent changes between data sets are mainly members of the glial-rich M5 and M18 modules (MDK, COL25A1, MAPT, NTN1, SMOC1, and GFAP). Among the reduced proteins, those with the most consistent changes were almost exclusively members of the M1 module (NPTX2, VGF, and RPH3A) associated with the synapse. We further verified the AD-related changes of midkine (MDK), CD44, secreted frizzled-related protein 1 (SFRP1) and VGF by western blotting (Figure S3B). Module preservation analysis showed that about 80% of the protein modules (34/44) in the brain proteome were significantly conserved in the replication data set (z-score> 1.96, FDR corrected P <0.05) (Figure S3C). Fourteen of these modules were specially reserved between the two proteomes (z-score> 10, FDR corrected P <1.0 × 10−23). Overall, the discovery and replication of the high degree of consistency in differential expression and modular composition between the brain proteome highlights the reproducibility of the changes in AD frontal cortex proteins. In addition, it also confirmed that AsymAD and more advanced diseases have a very similar brain network structure.
A more detailed analysis of the differential expression in the brain replication data set highlights the significant degree of AsymAD protein changes, including a total of 151 significantly changed proteins between AsymAD and the control (P <0.05) (Figure S3D). Consistent with the amyloid load, APP in the brain of AsymAD and AD increased significantly. MAPT only changes significantly in AD, which is consistent with increased levels of tangles and its known correlation with cognitive decline (5, 7). The glial-rich modules (M5 and M18) are highly reflected in the increased proteins in AsymAD, while the neuron-related M1 module is the most representative of the decreased proteins in AsymAD. Many of these AsymAD markers show greater changes in symptomatic diseases. Among these markers is SMOC1, a glial protein belonging to M18, which is associated with brain tumors and the development of eyes and limbs (32). MDK is a heparin-binding growth factor related to cell growth and angiogenesis (33), another member of M18. Compared with the control group, AsymAD increased significantly, followed by a greater increase in AD. In contrast, the synaptic protein neuropentraxin 2 (NPTX2) was significantly reduced in the AsymAD brain. NPTX2 was previously associated with neurodegeneration and has a recognized role in mediating excitatory synapses (34). Overall, these results reveal a variety of different preclinical protein changes in AD that seem to progress with the severity of the disease.
Given that we have achieved a significant depth of protein coverage in the discovery of the brain proteome, we are trying to understand more fully its overlap with the network-level AD transcriptome. Therefore, we compared the brain proteome we discovered with the module we previously generated from the microarray measurement of 18,204 genes in AD (n = 308) and control (n = 157) DLPFC tissues (13). overlapping. In total, we identified 20 different RNA modules, many of which demonstrated the enrichment of specific cell types, including neurons, oligodendrocytes, astrocytes, and microglia (Figure 3A). The multiple changes of these modules in AD are shown in Figure 3B. Consistent with our previous protein-RNA overlap analysis using the deeper unlabeled MS proteome (about 3000 proteins) (13), most of the 44 modules in the brain proteome network we found are in the transcriptome network There is no significant overlap in. Even in our discovery and replication of the 34 protein modules that are highly retained in the brain proteome, only 14 (~40%) passed Fisher’s exact test (FET) proved to have a statistically significant overlap with the transcriptome (Figure 3A) . Compatible with DNA damage repair (P-M25 and P-M19), protein translation (P-M7 and P-M20), RNA binding/splicing (P-M16 and P-M21) and protein targeting (P-M13 and P- M23) does not overlap with modules in the transcriptome. Therefore, although a deeper proteome data set is used in the current overlap analysis (13), most of the AD network proteome is not mapped to the transcriptome network.
(A) Hypergeometric FET demonstrates the enrichment of cell type-specific markers in the RNA module of the AD transcriptome (top) and the degree of overlap between the RNA (x-axis) and protein (y-axis) modules of the AD brain (bottom) . The intensity of the red shading indicates the degree of enrichment of cell types in the top panel and the intensity of the overlap of the modules in the bottom panel. Asterisks indicate statistical significance (P <0.05). (B) The degree of correlation between the characteristic genes of each transcriptome module and AD status. The modules on the left are the most negatively correlated with AD (blue), and those on the right are the most positively correlated with AD (red). The log-transformed BH-corrected P value indicates the degree of statistical significance of each correlation. (C) Significant overlapping modules with shared cell type enrichment. (D) Correlation analysis of the log2 fold change of the labeled protein (x-axis) and RNA (y-axis) in the overlapping module. The Pearson correlation coefficient with the relevant P value is displayed. Micro, microglia; celestial bodies, astrocytes. CT, control.
Most overlapping protein and RNA modules share similar cell type enrichment profiles and consistent AD change directions (Figure 3, B and C). In other words, the synapse-related M1 module of the brain proteome (PM1) is mapped to three neuronal-rich homologous RNA modules (R-M1, R-M9 and R-M16), which are in AD Both showed a reduced level. Similarly, the glial-rich M5 and M18 protein modules overlap with RNA modules rich in astrocytes and microglial markers (R-M3, R-M7, and R-M10) and are highly involved in diseases Increase. These shared modular features between the two data sets further support the cell type enrichment and disease-related changes we have observed in the brain proteome. However, we observed many significant differences between the RNA and protein levels of individual markers in these shared modules. Correlation analysis of the differential expression of the proteomics and transcriptomics of the molecules within these overlapping modules (Figure 3D) highlights this inconsistency. For example, APP and several other glial module proteins (NTN1, MDK, COL25A1, ICAM1, and SFRP1) showed a significant increase in the AD proteome, but there was almost no change in the AD transcriptome. These protein-specific changes may be closely related to amyloid plaques (23, 35), highlighting the proteome as the source of pathological changes, and these changes may not be reflected in the transcriptome.
After independently analyzing the brain and CSF proteomes we discovered, we conducted a comprehensive analysis of the two data sets to identify AD CSF biomarkers related to the pathophysiology of the brain network. We must first define the overlap of the two proteomes. Although it is widely accepted that CSF reflects neurochemical changes in the AD brain (4), the exact degree of overlap between the AD brain and the CSF proteome is unclear. By comparing the number of shared gene products detected in our two proteomes, we found that nearly 70% (n = 1936) of the proteins identified in the cerebrospinal fluid were also quantified in the brain (Figure 4A). Most of these overlapping proteins (n = 1721) are mapped to one of 44 co-expression modules from the discovery brain data set (Figure 4B). As expected, the six largest brain modules (M1 to M6) exhibited the greatest amount of CSF overlap. However, there are smaller brain modules (for example, M15 and M29) that achieve an unexpectedly high degree of overlap, larger than a brain module twice its size. This motivates us to adopt a more detailed, statistically driven method to calculate the overlap between the brain and cerebrospinal fluid.
(A and B) The proteins detected in the discovery brain and CSF data sets overlap. Most of these overlapping proteins are associated with one of the 44 co-expression modules of the brain co-expression network. (C) Discover the overlap between the cerebrospinal fluid proteome and the brain network proteome. Each row of the heat map represents a separate overlap analysis of the hypergeometric FET. The top row depicts the overlap (gray/black shading) between the brain module and the entire CSF proteome. The second line depicts that the overlap between brain modules and CSF protein (shaded in red) is significantly up-regulated in AD (P <0.05). The third row shows that the overlap between brain modules and CSF protein (blue shading) is significantly down-regulated in AD (P <0.05). Use the BH method to correct the P value derived from the FET. (D) Folding module panel based on cell type association and related GO terms. These panels contain a total of 271 brain-related proteins, which have meaningful differential expression in the CSF proteome.
Using single-tailed FETs, we assessed the importance of protein overlap between the CSF proteome and individual brain modules. The analysis revealed that a total of 14 brain modules in the CSF data set have statistically significant overlaps (FDR adjusted P <0.05), and an additional module (M18) whose overlap is close to significance (FDR adjusted P = 0.06) (Figure 4C, top row). We are also interested in modules that strongly overlap with differentially expressed CSF proteins. Therefore, we applied two additional FET analyses to determine which of (i) CSF protein was significantly increased in AD and (ii) CSF protein was significantly decreased in AD (P <0.05, paired t test AD/control) Brain modules with meaningful overlap between them. As shown in the middle and bottom rows of Figure 4C, these additional analyses show that 8 of the 44 brain modules significantly overlap with the protein added in AD CSF (M12, M1, M2, M18, M5, M44, M33, and M38). ), while only two modules (M6 and M15) showed a meaningful overlap with the reduced protein in AD CSF. As expected, all 10 modules are in the 15 modules with the highest overlap with the CSF proteome. Therefore, we assume that these 15 modules are high-yield sources of AD brain-derived CSF biomarkers.
We folded these 15 overlapping modules into five large protein panels based on their proximity in the WGCNA tree diagram and their association with cell types and gene ontology (Figure 4D). The first panel contains modules rich in neuron markers and synapse-related proteins (M1 and M12). The synaptic panel contains a total of 94 proteins, and the levels in the CSF proteome have changed significantly, making it the largest source of brain-related CSF markers among the five panels. The second group (M6 and M15) demonstrated the close connection with endothelial cell markers and vascular body, such as “wound healing” (M6) and “regulation of humoral immune response” (M15). M15 is also highly related to lipoprotein metabolism, which is closely related to endothelium (36). The vascular panel contains 34 CSF markers related to the brain. The third group includes modules (M2 and M4) that are significantly related to oligodendrocyte markers and cell proliferation. For example, the top-level ontology terms of M2 include “positive regulation of DNA replication” and “purine biosynthesis process”. Meanwhile, those of M4 include “glial cell differentiation” and “chromosome segregation”. The myelination panel contains 49 CSF markers related to the brain.
The fourth group contains the most modules (M30, M29, M18, M24, and M5), and almost all modules are significantly rich in microglia and astrocyte markers. Similar to the myelination panel, the fourth panel also contains modules (M30, M29, and M18) that are closely related to cell proliferation. The other modules in this group are highly related to immunological terms, such as “immune effect process” (M5) and “immune response regulation” (M24). The glial immune group contains 42 CSF markers related to the brain. Finally, the last panel includes 52 brain-related markers on the four modules (M44, M3, M33, and M38), all of which are on the body related to energy storage and metabolism. The largest of these modules (M3) is closely related to mitochondria and is rich in neuron-specific markers. M38 is one of the smaller module members in this metabolome and also exhibits moderate neuron specificity.
Overall, these five panels reflect a wide range of cell types and functions in the AD cortex, and collectively contain 271 brain-related CSF markers (Table S2G). In order to evaluate the validity of these MS results, we used the proximity extension assay (PEA), an orthogonal antibody-based technology with multiplexing capabilities, high sensitivity and specificity, and reanalyzed the cerebrospinal fluid samples we found A subset of these 271 biomarkers (n = 36). These 36 targets demonstrate the change in the AD multiple of PEA, which is closely related to our MS-based findings (r = 0.87, P = 5.6 × 10-12) , Which strongly verified the results of our comprehensive MS analysis (Figure S4).
The biological themes emphasized by our five groups, from synaptic signaling to energy metabolism, are all related to the pathogenesis of AD (1-3). Therefore, all 15 modules containing these panels are related to the AD pathology in the brain proteome that we discovered (Figure 2B). The most notable is the high positive pathological correlation between our glial modules and the strong negative pathological correlation between our largest neuronal modules (M1 and M3). The differential expression analysis of our replicated brain proteome (Figure S3D) also highlights M5 and M18-derived glial proteins. In AsymAD and symptomatic AD, the most increased glial proteins and M1-related synapses The protein is reduced the most. These observations indicate that the 271 cerebrospinal fluid markers we identified in the five groups are related to disease processes in the AD cortex, including those that occur in the early asymptomatic stages.
In order to better analyze the change direction of the panel proteins in the brain and spinal fluid, we drew the following for each of the 15 overlapping modules: (i) found the module abundance level in the brain data set and (ii) the module protein The difference is expressed in the cerebrospinal fluid (Figure S5). As mentioned earlier, WGCNA is used to determine the module abundance or characteristic protein value in the brain (13). The volcano map is used to describe the differential expression of modular proteins in the cerebrospinal fluid (AD/control). These figures show that three of the five panels show different expression trends in the brain and spinal fluid. The two modules of the synapse panel (M1 and M12) show a decrease in the abundance level in the AD brain, but significantly overlap with the increased protein in the AD CSF (Figure S5A). The neuron-related modules containing the metabolome (M3 and M38) showed similar brain and cerebrospinal fluid expression patterns inconsistent (Figure S5E). The vascular panel also showed different expression trends, although its modules (M6 and M15) were moderately increased in the AD brain and decreased in the diseased CSF (Figure S5B). The remaining two panels contain large glial networks whose proteins are consistently up-regulated in both compartments (Figure S5, C and D).
Please note that these trends are not common to all markers in these panels. For example, the synaptic panel includes several proteins that are significantly reduced in the AD brain and CSF (Figure S5A). Among these down-regulated cerebrospinal fluid markers are NPTX2 and VGF of M1, and chromogranin B of M12. However, despite these exceptions, most of our synaptic markers are elevated in AD spinal fluid. Overall, these analyses were able to distinguish statistically significant trends in brain and cerebrospinal fluid levels in each of our five panels. These trends highlight the complex and often different relationship between brain and CSF protein expression in AD.
Then, we used high-throughput MS replication analysis (CSF replication 1) to narrow our 271 set of biomarkers to the most promising and reproducible targets (Figure 5A). CSF copy 1 contains a total of 96 samples from Emory Goizueta ADRC, including control, AsymAD, and AD cohort (Table S1A). These AD cases are characterized by mild cognitive decline (mean MoCA, 20.0 ± 3.8), and changes in AD biomarkers confirmed in cerebrospinal fluid (Table S1A). Contrary to the CSF analysis we found, this replication is performed using a more efficient and high-throughput “single-shot” MS method (without off-line fractionation), including a simplified sample preparation protocol that eliminates the need for immunodepletion of individual samples. Instead, a single immune-depleted “enhancement channel” is used to amplify the signal of less abundant proteins (37). Although it reduces the total proteome coverage, this single-shot method significantly reduces machine time and increases the number of TMT-labeled samples that can be analyzed viable (17, 38). In total, the analysis identified 6,487 peptides, which mapped to 1,183 proteomes in 96 cases. As with the CSF analysis we found, only those proteins quantified in at least 50% of the samples were included in the subsequent calculations, and the data was regressed for the effects of age and gender. This led to the final quantification of 792 proteomes, 95% of which were also identified in the CSF data set found.
(A) Brain-related CSF protein targets verified in the first replicated CSF cohort and included in the final panel (n = 60). (B to E) Panel biomarker levels (composite z-scores) measured in the four CSF replication cohorts. Paired t-tests or ANOVA with Tukey’s post-correction were used to evaluate the statistical significance of the changes in abundance in each replicate analysis. CT, control.
Since we are particularly interested in verifying our 271 brain-related CSF targets through comprehensive analysis, we will limit further examination of this replicated proteome to these markers. Among these 271 proteins, 100 were detected in CSF replication 1. Figure S6A shows the differential expression of these 100 overlapping markers between the control and AD replication samples. Synaptic and metabolite histones increase the most in AD, while vascular proteins decrease the most in disease. Most of the 100 overlapping markers (n = 70) maintained the same direction of change in the two data sets (Figure S6B). These 70 validated brain-related CSF markers (Table S2H) largely reflect the previously observed panel expression trends, that is, the down-regulation of vascular proteins and the up-regulation of all other panels. Only 10 of these 70 validated proteins showed changes in AD abundance that contradicted these panel trends. In order to generate a panel that best reflects the overall trend of the brain and cerebrospinal fluid, we excluded these 10 proteins from the panel of interest that we finally verified (Figure 5A). Therefore, our panel ultimately includes a total of 60 proteins verified in two independent CSF AD cohorts using different sample preparation and MS platform analysis. The z-score expression plots of these final panels in the CSF copy 1 control and AD cases confirmed the panel trend observed in the CSF cohort we found (Figure 5B).
Among these 60 proteins, there are molecules known to be associated with AD, such as osteopontin (SPP1), which is a pro-inflammatory cytokine that has been associated with AD in many studies (39-41), and GAP43, A synaptic protein that is clearly linked to neurodegeneration (42). The most fully verified proteins are markers related to other neurodegenerative diseases, such as amyotrophic lateral sclerosis (ALS) related superoxide dismutase 1 (SOD1) and Parkinson’s disease related desaccharase (PARK7) . We have also verified that many other markers, such as SMOC1 and brain-rich membrane attachment signaling protein 1 (BASP1), have limited previous links to neurodegeneration. It is worth noting that due to their low overall abundance in the CSF proteome, it is difficult for us to use this high-throughput single-shot detection method to reliably detect MAPT and certain other AD-related proteins (for example, NEFL and NRGN) ( 43, 44).
We then checked these 60 priority panel markers in three additional replicate analyses. In CSF Copy 2, we used a single TMT-MS to analyze an independent cohort of 297 control and AD samples from Emory Goizueta ADRC (17). CSF replication 3 included a reanalysis of available TMT-MS data from 120 control and AD patients from Lausanne, Switzerland (45). We detected more than two-thirds of the 60 priority markers in each dataset. Although the Swiss study used different MS platforms and TMT quantification methods (45, 46), we strongly reproduced our panel trends in two repeated analyses (Figure 5, C and D, and Tables S2, I, and J) . To evaluate the disease specificity of our group, we used TMT-MS to analyze the fourth replication data set (CSF replication 4), which included not only control (n = 18) and AD (n = 17) cases, but also PD (n = 14)), ALS (n = 18) and frontotemporal dementia (FTD) samples (n = 11) (Table S1A). We successfully quantified nearly two-thirds of the panel proteins in this cohort (38 out of 60). These results highlight the AD-specific changes in all five biomarker panels (Figure 5E and Table S2K). The increase in the metabolite group showed the strongest AD specificity, followed by the myelination and glial group. To a lesser extent, FTD also shows an increase between these panels, which may reflect similar potential network changes (17). In contrast, ALS and PD showed almost the same myelination, glial, and metabolome profiles as the control group. Overall, despite differences in sample preparation, MS platform, and TMT quantification methods, these repeated analyses show that our priority panel markers have highly consistent AD-specific changes in more than 500 unique CSF samples.
AD neurodegeneration has been widely recognized several years before the onset of cognitive symptoms, so there is an urgent need for biomarkers of AsymAD (5, 31). However, more and more evidences show that the biology of AsymAD is far from homogeneous, and the complex interaction of risk and resilience leads to large individual differences in subsequent disease progression (47). Although used to identify AsymAD cases, the levels of core CSF biomarkers (Aβ1-42, total tau and p-tau) have not proven to be able to reliably predict who will progress to dementia (4, 7), indicating more It may be necessary to include holistic biomarker tools based on multiple aspects of brain physiology to accurately stratify the risk of this population. Therefore, we subsequently analyzed our AD-validated biomarker panel in the AsymAD population of CSF copy 1. These 31 AsymAD cases showed abnormal core biomarker levels (Aβ1–42/total tau ELISA ratio, <5.5) and complete cognition (mean MoCA, 27.1 ± 2.2) (Table S1A). In addition, all individuals with AsymAD have a clinical dementia score of 0, indicating that there is no evidence of a decline in daily cognitive or functional performance.
We first analyzed the levels of the validated panels in all 96 CSF replicates 1, including the AsymAD cohort. We found that several panels in the AsymAD group had significant AD-like abundance changes, the vascular panel showed a downward trend in AsymAD, while all other panels showed an upward trend (Figure 6A). Therefore, all panels showed a highly significant correlation with ELISA Aβ1-42 and total tau levels (Figure 6B). In contrast, the correlation between the group and the MoCA score is relatively poor. One of the more striking findings from these analyses is the large range of panel abundances in the AsymAD cohort. As shown in Figure 6A, the panel level of the AsymAD group usually crosses the panel level of the control group and the AD group, showing relatively high variability. To further explore this heterogeneity of AsymAD, we applied Multidimensional Scaling (MDS) analysis to 96 CSF replication 1 cases. MDS analysis allows to visualize the similarity between cases based on certain variables in the data set. For this cluster analysis, we only use those validated panel markers that have a statistically significant change (P <0.05, AD/control) in the CSF discovery and replication 1 proteome (n = 29) (Table S2L) level. This analysis produced clear spatial clustering between our control and AD cases (Figure 6C). In contrast, some AsymAD cases are clearly clustered in the control group, while others are located in AD cases. To further explore this AsymAD heterogeneity, we used our MDS map to define two groups of these AsymAD cases. The first group included AsymAD cases clustered closer to the control (n = 19), while the second group was characterized by AsymAD cases with a marker profile closer to AD (n = 12).
(A) The expression level (z-score) of the CSF biomarker group in all 96 samples in the CSF replication 1 cohort, including AsymAD. Analysis of variance with Tukey’s post-correction was used to evaluate the statistical significance of panel abundance changes. (B) Correlation analysis of panel protein abundance level (z-score) with MoCA score and total tau level in ELISA Aβ1-42 and CSF copy 1 samples. The Pearson correlation coefficient with the relevant P value is displayed. (C) The MDS of 96 CSF copy 1 cases was based on the abundance levels of 29 validated panel markers, which were significantly changed in both the discovery and CSF copy 1 data sets [P <0.05 AD/control (CT)]. This analysis was used to divide the AsymAD group into control (n = 19) and AD (n = 12) subgroups. (D) The volcano plot shows the differential expression of all CSF replication 1 proteins with log2 fold change (x-axis) relative to the -log10 statistical P value between the two AsymAD subgroups. The panel biomarkers are colored. (E) CSF replication 1 abundance level of the selection group biomarkers are differentially expressed between AsymAD subgroups. Tukey’s post-adjusted analysis of variance was used to assess statistical significance.
We examined the differential protein expression between these control and AD-like AsymAD cases (Figure 6D and Table S2L). The resulting volcano map shows that 14 panel markers have changed significantly between the two groups. Most of these markers are members of the synapse and metabolome. However, SOD1 and myristoylated alanine-rich protein kinase C substrate (MARCKS), which are members of the myelin and glial immune groups, respectively, also belong to this group (Figure 6, D and E) . The vascular panel also contributed two markers that were significantly reduced in the AD-like AsymAD group, including AE binding protein 1 (AEBP1) and complement family member C9. There was no significant difference between the control and AD-like AsymAD subgroups in ELISA AB1-42 (P = 0.38) and p-tau (P = 0.28), but there was indeed a significant difference in the total tau level (P = 0.0031) (Fig. S7). There are several panel markers that indicate that the changes between the two AsymAD subgroups are more significant than the total tau levels (for example, YWHAZ, SOD1, and MDH1) (Figure 6E). Overall, these results indicate that our validated panel may contain biomarkers that can subtype and potential risk stratification of patients with asymptomatic disease.
There is an urgent need for system-based biomarker tools to better measure and target the various pathophysiology behind AD. These tools are expected to not only change our AD diagnostic framework, but also promote the adoption of effective, patient-specific treatment strategies (1, 2). To this end, we applied an unbiased comprehensive proteomics approach to AD brain and CSF to identify web-based CSF biomarkers that reflect a broad range of brain-based pathophysiology. Our analysis produced five CSF biomarker panels, which (i) reflect synapses, blood vessels, myelin, immune and metabolic dysfunction; (ii) demonstrate strong reproducibility on different MS platforms; ( iii) Show progressive disease-specific changes throughout the early and late stages of AD. Overall, these findings represent a promising step towards the development of diverse, reliable, web-oriented biomarker tools for AD research and clinical applications.
Our results demonstrate the highly conserved organization of the AD brain network proteome and support its use as an anchor for system-based biomarker development. Our analysis shows that two independent TMT-MS datasets containing AD and AsymAD brains have strong modularity. These findings extend our previous work, demonstrating the preservation of the powerful modules of more than 2,000 brain tissues from multiple independent cohorts in the frontal, parietal, and temporal cortex (17). This consensus network reflects various disease-related changes observed in current research, including the increase of glial-rich inflammatory modules and the decrease of neuron-rich modules. Like current research, this large-scale network also features significant modular changes in AsymAD, showing a variety of different preclinical pathophysiology (17).
However, within this highly conservative system-based framework, there is more fine-grained biological heterogeneity, especially among individuals in the early stages of AD. Our biomarker panel is able to depict two subgroups in AsymAD, which demonstrate the significant differential expression of multiple CSF markers. Our group was able to highlight the biological differences between these two subgroups, which were not obvious at the level of core AD biomarkers. Compared with the control group, the Aβ1-42/total tau ratios of these AsymAD individuals were abnormally low. However, only the total tau levels were significantly different between the two AsymAD subgroups, while the Aβ1-42 and p-tau levels remained relatively comparable. Since high CSF tau seems to be a better predictor of cognitive symptoms than Aβ1-42 levels (7), we suspect that the two AsymAD cohorts may have different risks of disease progression. Given the limited sample size of our AsymAD and the lack of longitudinal data, further research is needed to confidently draw these conclusions. However, these results indicate that a system-based CSF panel can enhance our ability to effectively stratify individuals during the asymptomatic stage of the disease.
Overall, our findings support the role of multiple biological functions in the pathogenesis of AD. However, dysregulated energy metabolism became the prominent theme of all our five validated labeling panels. Metabolic proteins, such as hypoxanthine-guanine phosphoribosyltransferase 1 (HPRT1) and lactate dehydrogenase A (LDHA), are the most robustly validated synaptic biomarkers, indicating that the increase in AD CSF is highly reproducible sex. Our blood vessels and glial panels also contain several markers involved in the metabolism of oxidative substances. These findings are consistent with the key role that metabolic processes play in the entire brain, not only to meet the high energy demand of neurons, but also to meet the high energy demand of astrocytes and other glial cells (17, 48). Our results support growing evidence that changes in redox potential and interruption of energy pathways may be the core link between several key processes involved in the pathogenesis of AD, including mitochondrial disorders, glial-mediated inflammation, and Vascular damage (49). In addition, the metabolic cerebrospinal fluid biomarkers contain a large number of differentially rich proteins between our control and AD-like AsymAD subgroups, suggesting that the disruption of these energy and redox pathways may be critical in the preclinical stage of the disease .
The different brain and cerebrospinal fluid panel trends we have observed also have interesting biological implications. Synapses and metabolomes rich in neurons show decreased levels in the AD brain and increased abundance in cerebrospinal fluid. Given that neurons are rich in energy-producing mitochondria at synapses to provide energy for their numerous specialized signals (50), the similarity of the expression profiles of these two neuron groups is expected. The loss of neurons and the extrusion of damaged cells can explain these brain and CSF panel trends in later disease, but they cannot explain the early panel changes we observe (13). One possible explanation for these findings in early asymptomatic disease is abnormal synaptic pruning. New evidence in mouse models suggests that microglia-mediated synaptic phagocytosis may be abnormally activated in AD and lead to early synapse loss in the brain (51). This discarded synaptic material may accumulate in CSF, which is why we observe the increase in CSF in the neuron panel. Immune-mediated synaptic pruning may also partially explain the increase in glial proteins we observe in the brain and cerebrospinal fluid throughout the disease process. In addition to synaptic pruning, overall abnormalities in the exocytic pathway may also lead to different brain and CSF expressions of neuronal markers. A number of studies have shown that the content of exosomes in the pathogenesis of AD brain has changed (52). The extracellular pathway is also involved in the proliferation of Aβ (53, 54). It is worth noting that suppression of exosomal secretion may reduce AD-like pathology in AD transgenic mouse models (55).
At the same time, the protein in the vascular panel showed a moderate increase in the AD brain, but significantly decreased in the CSF. The blood-brain barrier (BBB) dysfunction can partly explain these findings. Many independent postmortem human studies have demonstrated BBB breakdown in AD (56, 57). These studies confirmed various abnormal activities surrounding this tightly sealed layer of endothelial cells, including brain capillary leakage and perivascular accumulation of blood-borne proteins (57). This can provide a simple explanation for the elevated vascular proteins in the brain, but it cannot fully explain the depletion of these same proteins in the cerebrospinal fluid. One possibility is that the central nervous system is actively isolating these molecules to solve the problem of increased inflammation and oxidative stress. The reduction in some of the most severe CSF proteins in this panel, especially those involved in lipoprotein regulation, is related to the inhibition of harmful levels of inflammation and the neuroprotective process of reactive oxygen species. This is true for Paroxonase 1 (PON1), a lipoprotein binding enzyme responsible for reducing oxidative stress levels in the circulation (58, 59). Alpha-1-microglobulin/bikunin precursor (AMBP) is another significantly down-regulated marker of the vascular group. It is the precursor of the lipid transporter bikunin, which is also involved in inflammation suppression and neurological Protection (60, 61).
Despite various interesting hypotheses, the inability to directly detect biochemical disease mechanisms is a well-known limitation of discovery-driven proteomics analysis. Therefore, further research is necessary to confidently define the mechanisms behind these biomarker panels. In order to move towards the development of MS-based clinical analysis, the future direction also requires the use of targeted quantitative methods for large-scale biomarker verification, such as selective or parallel reaction monitoring (62). We recently used parallel reaction monitoring (63) to validate many of the CSF protein changes described here. Several priority panel targets are quantified with significant accuracy, including YWHAZ, ALDOA, and SMOC1, which map to our synapse, metabolism, and inflammation panels, respectively (63). Independent Data Acquisition (DIA) and other MS-based strategies may also be useful for target verification. Bud et al. (64) It was recently demonstrated that there is a significant overlap between the AD biomarkers identified in our CSF discovery data set and the independent DIA-MS data set, which consists of nearly 200 CSF samples from three different European cohorts . These recent studies support the potential of our panels to transform into reliable MS-based detection. Traditional antibody and aptamer-based detection is also important for the further development of key AD biomarkers. Due to the low abundance of CSF, it is more difficult to detect these biomarkers using high-throughput MS methods. NEFL and NRGN are two such examples of low-abundance CSF biomarkers, which are mapped to the panel in our comprehensive analysis, but cannot be reliably detected using our single MS strategy. Targeting strategies based on multiple antibodies, such as PEA, may promote the clinical transformation of these markers.
Overall, this study provides a unique proteomics approach for the identification and verification of CSF AD biomarkers based on different systems. Optimizing these marker panels across additional AD cohorts and MS platforms may prove promising to advance AD risk stratification and treatment. Studies that evaluate the longitudinal level of these panels over time are also critical to determine which combination of markers best stratifies the risk of early disease and changes in disease severity.
Except for the 3 samples copied by CSF, all CSF samples used in this study were collected under the auspices of Emory ADRC or closely related research institutions. A total of four sets of Emory CSF samples were used in these proteomics studies. The CSF cohort was found to contain samples from 20 healthy controls and 20 AD patients. CSF copy 1 includes samples from 32 healthy controls, 31 AsymAD individuals, and 33 AD individuals. CSF copy 2 contains 147 controls and 150 AD samples. The multi-disease CSF replication 4 cohort included 18 controls, 17 AD, 19 ALS, 13 PD, and 11 FTD samples. According to the agreement approved by the Emory University Institutional Review Board, all Emory study participants obtained informed consent. According to the 2014 National Institute of Aging Best Practice Guidelines for Alzheimer’s Centers (https://alz.washington.edu/BiospecimenTaskForce.html), cerebrospinal fluid was collected and stored by lumbar puncture. Control and AsymAD and AD patients received standardized cognitive assessment at Emory Cognitive Neurology Clinic or Goizueta ADRC. Their cerebrospinal fluid samples were tested by INNO-BIA AlzBio3 Luminex for ELISA Aβ1-42, total tau and p-tau analysis (65 ). ELISA values are used to support the diagnostic classification of subjects based on established AD biomarker cut-off criteria (66, 67). Basic demographic and diagnostic data for other CSF diagnoses (FTD, ALS, and PD) are also obtained from Emory ADRC or affiliated research institutions. The summary case metadata for these Emory CSF cases can be found in Table S1A. The characteristics of the Swiss CSF replication 3 cohort have been previously published (45).
CSF found the sample. In order to increase the depth of our discovery of the CSF data set, immune consumption of high-abundance proteins was performed before trypsinization. In short, 130 μl of CSF from 40 individual CSF samples and an equal volume (130 μl) of High Select Top14 Abundance Protein Depletion Resin (Thermo Fisher Scientific, A36372) were placed in a spin column (Thermo Fisher Scientific, A89868) at room temperature Incubate). After spinning for 15 minutes, centrifuge the sample at 1000g for 2 minutes. A 3K ultracentrifugal filter device (Millipore, UFC500396) was used to concentrate the effluent sample by centrifuging at 14,000g for 30 minutes. Dilute all sample volumes to 75 μl with phosphate buffered saline. The protein concentration was evaluated by the bicinchoninic acid (BCA) method according to the manufacturer’s protocol (Thermo Fisher Scientific). The immunodepleted CSF (60 μl) from all 40 samples was digested with lysyl endopeptidase (LysC) and trypsin. In short, the sample was reduced and alkylated with 1.2 μl 0.5 M tris-2(-carboxyethyl)-phosphine and 3 μl 0.8 M chloroacetamide at 90°C for 10 minutes, and then sonicated in a water bath for 15 minutes. The sample was diluted with 193 μl 8 M urea buffer [8 M urea and 100 mM NaHPO4 (pH 8.5)] to a final concentration of 6 M urea. LysC (4.5 μg; Wako) is used for overnight digestion at room temperature. The sample was then diluted to 1 M urea with 50 mM ammonium bicarbonate (ABC) (68). Add an equal amount (4.5 μg) of trypsin (Promega), and then incubate the sample for 12 hours. Acidify the digested peptide solution to a final concentration of 1% formic acid (FA) and 0.1% trifluoroacetic acid (TFA) (66), and then desalt with a 50 mg Sep-Pak C18 column (Waters) as described above (25) . The peptide was then eluted in 1 ml of 50% acetonitrile (ACN). To standardize protein quantification across batches (25), 100 μl aliquots from all 40 CSF samples were combined to generate a mixed sample, which was then divided into five global internal standard (GIS) (48) samples. All individual samples and combined standards are dried by high-speed vacuum (Labconco).
CSF copies the sample. Dayon and colleagues have previously described immune depletion and digestion of CSF copy 3 samples (45, 46). The remaining replicate samples were not individually immunodepleted. Digest these unremoved samples in trypsin as described previously (17). For each repeated analysis, 120 μl aliquots of the eluted peptide from each sample were pooled together and divided into equal volume aliquots to be used as the TMT-labeled global internal standard (48). All individual samples and combined standards are dried by high-speed vacuum (Labconco). In order to enhance the signal of the low-abundance CSF protein, by combining 125 μl from each sample, an “enhanced” sample was prepared for each replicate analysis [ie, a biological sample that mimics the research sample, but the amount available is much larger (37, 69)] merged into a mixed CSF sample (17). The mixed sample was then immunoremoved using 12 ml of High Select Top14 Abundance Protein Removal Resin (Thermo Fisher Scientific, A36372), digested as described above, and included in the subsequent multiple TMT labeling.
Post time: Aug-27-2021