Identification of markers for neurescence  through transcriptomic profiling of postmortem human brains

Identification of markers for neurescence through transcriptomic profiling of postmortem human brains | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Identification of markers for neurescence through transcriptomic profiling of postmortem human brains Shiva Kazempour Dehkordi, Sogand Sajedi, Amirreza Heshmat, Miranda E. Orr, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5903682/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 01 Jul, 2025 Read the published version in npj Aging → Version 1 posted 10 You are reading this latest preprint version Abstract Neuronal senescence (i.e., neurescent) is an important hallmark of aging and neurodegeneration, but it remains poorly characterized in the human brain due to the lack of reliable markers. This study aimed to identify neurescent markers based on single-nucleus transcriptome data from postmortem human prefrontal cortex. Using an eigengene approach, we integrated three gene panels: a) SenMayo, b) Canonical Senescence Pathway (CSP), and c) Senescence Initiating Pathway (SIP), to identify neurescent signatures. We found that paired markers outperform single markers; for instance, by combining CDKN2D and ETS2 in a decision tree, a high accuracy of 99% and perfect specificity (100%) were achieved in distinguishing neurescent. Differential expression analyses identified 324 genes that are overexpressed in neurescent. These genes showed significant associations with important neurodegeneration-related pathways including Alzheimer’s disease, Parkinson’s disease, and Huntington’s disease. Interestingly, several of these overexpressed genes are linked to mitochondrial dysfunction and cytoskeletal dysregulation. These findings provide valuable insights into the complexities of neurescent, emphasizing the need for further exploration of histologically viable markers and validation in broader datasets. Biological sciences/Cell biology/Senescence Health sciences/Diseases/Neurological disorders Health sciences/Diseases/Neurological disorders/Dementia Health sciences/Diseases/Neurological disorders/Dementia/Alzheimers disease Health sciences/Biomarkers Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Introduction Cellular senescence is a complex, multi-step process characterized by stable cell cycle arrest 1 , changes in cell morphology 2 , 3 , changes in gene expression 4 – 8 , and a pro-inflammatory secretory phenotype 9 , 10 . Senescence was first identified in fibroblasts 11 , is heterogeneous in tissues throughout the body 12 – 14 , and has been studied in the context of the human brain 15 , 16 . Increasing evidence implicates cellular senescence in brain aging and its links to neurodegenerative disorders 17 , 18 , including Alzheimer's disease (AD) 19 , 20 . The presence of senescent cells triggers a pro-inflammatory environment 21 and contributes to neuron loss, tissue dysfunction and cognitive impairment in animal models of AD 22 , 23 . Recent studies using AD mouse models and human postmortem brain tissues directly demonstrate that neuronal senescence associated with tau pathology drives neuronal loss and neurodegeneration 18 , and that senescent human astrocytes can induce neuronal death through excitotoxicity in co-culture systems 24 . In the human brain, senescent cells have been associated with the known AD pathologies e.g., amyloid plaques and neurofibrillary tangles 18 , 25 , 26 . Despite their significance, defining and identifying senescent cells in the human brain remains challenging due to their heterogeneous nature 27 . Although reliable markers have been utilized to identify senescent cells in specific tissues such as adipose tissue, retinal endothelial cells, and fibroblast 12 , 28 – 33 , a universally applicable set of senescence markers across diverse tissues remain poorly defined. This challenge is especially pronounced in neurons, where historically defined senescence markers have not been robustly validated 34 . This knowledge gap has limited our understanding of when senescent cells first appear in the adult brain, inferring their contribution to neurodegenerative disease pathophysiology and our ability to develop potential therapies to modulate or remove senescent cells from the human brain 25 , 35 . One recent attempt to generate a comprehensive list of senescent markers resulted in the SenMayo gene list 6 , which includes 125 genes that are highly correlated with age and the expression level of p16 and p21, cyclin-dependent kinase inhibitors that are upregulated in many senescent cells. The majority of genes in the SenMayo panel are senescence-associated secretory phenotype (SAPS) factors. However, cellular senescence has other aspects, including activation of senescent cell anti-apoptotic pathways (SCAPs) 36 , organelle dysfunction, and morphology changes. To provide a comprehensive and multidimensional perspective for neuronal senescence (neurescent), we employed an eigengene approach based on a) the SenMayo gene list, in addition to our two previously published lists 16 including b) Canonical Senescence Pathway (CSP) with 22 genes, which reflect cell cycle arrest 1 , 37 , 38 , and c) Senescence Initiating Pathway (SIP) panel with 48 genes, which are upregulated in early senescence and activate SCAPs. Technically, an eigengene is computed as a weighted average expression of all genes in a given list 39 , 40 derived using principal component analysis on those particular genes. In this study, we computed an eigengene 40 for each of the SenMayo, CSP, and SIP gene lists and used all these eigengenes for three main analyses: 1) identifying senescent and non-senescent cells, 2) identifying the most accurate markers for neuronal senescence, referred to as neurescence 34 , and 3) identifying differentially expressed genes in senescent vs non-senescent neurons. Results We used four independent single-nucleus RNA sequencing (snRNA-seq) datasets of the dorsal prefrontal cortex from postmortem human brains were used, which are here referred to as Mathys 2019 41 (n = 80,000), Zhou 2020 42 (n = 70,000), Xiong 2023 43 (n = 400,000), and Mathys 2024 44 (n = 255,000), respectively (Table 1). Mathys 2019 served as the discovery dataset to identify markers and the other three datasets were used to validate the performance of the identified markers. Performing the eigengene approach 16 on the Mathys 2019 discovery dataset, we identified 475 senescent and 16,871 non-senescent neurons (Table 1). Specifically, the identified 475 neurescent cells expressed each of the three CSP, SIP, and SenMayo eigengenes more than the corresponding mean plus three times the standard deviation (mean + 3 s.d.). In contrast, if a neuron expressed each of these eigengenes less than the corresponding mean, then that neuron was considered non-senescent (N = 16,871). The remaining 53,288 borderline neurons were excluded from our analysis. We used the identified neurescent and non-senescent neurons to train decision trees and to perform differential expression (DE) analysis. Of note, only excitatory neurons displayed consistent expression across all three senescence eigengenes, and subsequent differential expression analysis and marker identification were restricted exclusively to excitatory neurons. To further assess the robustness of our method and the reproducibility of the eigengene-derived senescence markers, we conducted a reciprocal validation analysis. We used Mathys 2024 as the discovery dataset to independently derive the eigengene weights for CSP, SIP, and SenMayo, and then applied these derived weights to Mathys 2019 as a validation dataset. Comparing the weights from the two independent discovery datasets (Mathys 2019 vs. Mathys 2024) (Supplementary Table 1) revealed high concordance: specifically, correlations of 94% for CSP weights, 88% for SIP weights, and 88% for SenMayo weights, respectively (Supplementary Fig. 3). Furthermore, when using the Mathys 2024-derived eigengene weights to identify senescent cells in the Mathys 2019 dataset, we identified 413 senescent neurons, closely matching the original 475 neurons that were identified using Mathys 2019 as discovery. Notably, 370 neurons were identified as senescent in both approaches, representing a highly significant overlap (hypergeometric test p-value \(\:<1{0}^{-1714}\) ). Collectively, these results demonstrate the robust consistency and reproducibility of our eigengene-based senescence identification framework, regardless of the initial discovery dataset selection. Identification of specific markers for neurons among genes generally associated with senescence To identify potential markers for neurescence, we fitted relatively small decision trees 45 to the discovery dataset. To ensure the selected genes were specific, each tree was allowed to use no more than two genes from the pool of gene sets known to be associated with cellular senescence 6 , 16 , as indicated on the third row of Table 2. Using the original counts to quantify gene expression levels, the best marker among the 180 senescence-associated genes was MAP2K1 leading to a single-gene tree with the highest accuracy of 99%, a sensitivity of 80%, and a specificity of 99% (Table 2, column 4). Adding CDKN2D as the second marker increased sensitivity to 93% with a negligible (i.e., < 0.1%) effect on accuracy. The corresponding decision tree, which was based on MAP2K1 and CDKN2D , had the best accuracy among all of our trees that could use two of the 180 senescence-associated genes. This suggested that these two genes could serve as effective markers and complement each other in identifying neurescence. In the corresponding tree, the thresholds for the expression levels of these genes were 2 and 1, respectively (Fig. 1 a), suggesting relatively low expression of these putative marker genes. These thresholds are undesirable for histology markers because distinguishing between levels of expression under the microscope is practically challenging. One would prefer markers that are totally absent in negative cells and ideally, have more than one transcript per cell in senescent neurons to ensure the observable signal is above background. To provide greater confidence in these lowly expressed marker genes, we repeated our analysis using a binary transformation of the discovery data, where any expression level above zero was converted to 1. Using these binary values, no single gene was found to accurately classify the senescent and non-senescent cells. This suggests that there may not be a single marker in our gene lists specifically expressed in neurescence. Interestingly, the best tree with two genes still included CDKN2D , but MAP2K1 was replaced with ETS2 (Fig. 1 b, last row of Table 2). The binary transformation did not change the accuracy but dropped the sensitivity from 93–83%, while increasing the specificity to 100%. After removing CDKN2D and ETS2 from the analysis, E2F3 and RB1 were selected in the second best tree (Table 2, column 5), leading to an accuracy of 99%, specificity of 100%, but a relatively low sensitivity of 67%. Differentially expressed genes To investigate differentially expressed genes in neurescence compared to non-senescent, we performed a DE analysis on the Mathys 2019 discovery dataset. Taking an agnostic approach, we included all 10,768 genes that had non-negligible expression in neurons of the discovery dataset (Methods). We employed two state of the art methods for DE analysis of scRNA-Seq data. The first method, MAST 46 , resulted in 375 differentially expressed genes, whereas using the second method, SigEMD 47 , we identified 576 differentially expressed genes (Supplementary Table 2). The two analyses shared 324 differentially expressed genes (Fig. 2 , Supplementary Table 2), which represented a significant overlap (p-value \(\:<1{0}^{-205}\) , hypergeometric test). These 324 genes were overexpressed in the 475 neurescent cells compared to other neurons (Fig. 2 ). In this study, we primarily focused on these 324 genes because they were identified by both DE analysis methods. Pathway analysis To understand the functional significance of the 324 differentially expressed genes, we performed a pathway analysis using Kyoto Encyclopedia of Genes and Genomes (KEGG) 48 and identified 18 significantly enriched pathways (p-value < 0.05, Fig. 3 ). Interestingly, the top pathways were related to neurodegeneration including Parkinson disease 49 , 50 , Amyotrophic lateral sclerosis 51 , Alzheimer’s disease 52 , 53 , and Huntington's disease 54 , suggesting common underlying biological mechanisms that might be associated with neuronal senescence. The top 5 enriched pathways shared 16 DE genes, which is a significant overlap (p-value < \(\:1{0}^{-8}\) , hypergeometric test, Fig. 4 ). These genes included COX6C, COX7A2L, COX7C, CYCS, KIF5A, NDUFA4, NDUFS1, PSMA7, TUBA4A, TUBB2A, TUBB4A, TUBB4B, UQCRB, UQCRC2, UQCRH , and UQCRHL . These 16 genes are mainly involved in mitochondrial function (COX, NDUF families) 55 , 56 and cytoskeletal structure (TUBB family) 57 , 58 . Mitochondrial dysfunction leads to impaired energy metabolism and increased production of reactive oxygen species 59 . It is a hallmark of both cellular senescence 60 and AD 61 . The deregulation of these genes supports prior work demonstrating mitochondrial dysfunction in neurescence, and their role in the neurodegeneration 18 . Identification of specific markers for senescent neurons among differentially expressed genes We hypothesized that there might be some specific markers for neurescence beyond the genes generally known to be associated with senescence (e.g., CSP, SIP, and SenMayo). To test this hypothesis, we extended our decision tree analysis to include the 324 differentially expressed genes identified based on the discovery dataset. Using the original count data, the top gene was DPYSL2 resulting in a relatively accurate classification (i.e., accuracy: 99%, sensitivity: 95%, and specificity: 99%) and when the gene ATP6V1H was added to make a two-gene tree, sensitivity increased to 99% (Table 2, column 6). However, for these trees to be useful, meaning that they identify molecules that can serve as biomarkers in histological assays, one must be able to determine whether these genes were expressed at relatively high levels in a neuron (i.e., above 4 and 3, respectively) (Fig. 5 a). To address this practical issue, we used binary expression levels leading to selection of UQCRHL together with ETS2 (Fig. 5 b). The recurrent selection of ETS2 (Table 2, columns 4 and 6) highlighted its importance even when additional differentially expressed genes were added to the analysis. Since the inclusion of UQCRHL led to very low sensitivity in the validation cohorts (Table 3), we considered removing this gene. Excluding UQCRHL and ETS2 led to selection of CDKN2D and RB1 again and increasing the specificity of senescence classification to 100% in the discovery dataset. Overall, adding differentially expressed genes did not seem to be helpful as the only reasonable tree in columns 6 and 7 was based on CDKN2D and RB1 , which were already in our CSP gene lists. This observation falsified the hypothesis that additional differentially expressed genes could identify specific markers for senescent neurons. Identification of alternative markers for senescent neurons The relatively low expression of the selected markers, combined with historical challenges in developing antibodies specific to senescence markers, necessitates that in silico studies identify multiple candidate molecules for subsequent in situ validation. We performed a systematic elimination analysis. We pooled all genes from CSP, SIP, and SenMayo lists with DE genes and fitted decision trees on the binary expression values in the discovery dataset, Mathys 2019 (Table 3). Our systematic approach involved iteratively removing the top-performing markers to evaluate their impact on model performance and to identify the next best set of markers. Including UQCRHL in decision trees led to poor sensitivity in the validation datasets. Thus, we excluded UQCRHL and found CDKN2D and ETS2 to be the second best pair. If reagents to detect ETS2 are unavailable or lack specificity, then CDKN2D and RB1 would be the next best pair of markers for neurescence. Our results showed that removing CDKN2D generally led to decreased accuracy, sensitivity, and specificity in validation datasets. This drop in model performance highlights the critical role CDKN2D plays in the classification. However, it is important to note that the currently available antibodies for p19, the protein product of CDKN2D , may lack specificity, which could complicate its use in histological applications. Discussion Identifying senescent neurons in the human brain is challenging due to the inherent heterogeneity of senescence, neuronal subtypes and the current lack of reliable markers specific to neurescence in the human brain 63 . To tackle this challenge, we systematically and unbiasedly analyzed scRNA-seq data from four datasets of the dorsal prefrontal cortex from postmortem human brains. Our eigengene-based method allowed us to focus on three key gene lists: SenMayo, CSP, and SIP, each reflecting distinct aspects of cellular senescence. While the recently published SenMayo list is a valuable resource, it primarily captures the inflammatory aspects of senescence. The majority of genes in this list are associated with Senescence-associated secretory phenotype and other important characteristics of neural senescence may not be represented by SenMayo. Inflammation is a critical component, particularly in the brain, where it has been linked to neurodegeneration 64 , 65 . The accumulation of senescent cells can lead to a persistent pro-inflammatory state, driven by mitochondrial dysfunction 66 , 67 , and concomitant elevated levels of actors like IL-6, IL-1β, and TNF-α, which are characteristic of SASP 68 – 70 . This chronic inflammation can impair tissue homeostasis and contribute to neuronal dysfunction 71 . However, inflammation alone does not fully capture the multifaceted nature of cellular senescence, where other mechanisms such as cell cycle arrest, DNA damage, mitochondrial dysfunction, and metabolic alterations also play critical roles 63 , 72 , 73 . Moreover, inflammation can occur independent of senescence. This inherent limitation of the SenMayo list underscores the need for a more comprehensive approach. Accordingly, in this study, we included CSP and SIP gene panels to capture other important senescence features such as cell cycle arrest and early stress responses. Our approach leveraged eigengenes derived from these three panels to classify neurons into senescent and non-senescent groups. Our integrative approach using multiple senescence-associated gene panels aims to improve specificity by capturing their combined activity rather than relying on any single pathway. Our findings demonstrated that a single marker gene is insufficient for accurately classifying neurescence. Decision tree analysis, incorporating multiple senescence panels and both continuous and binary expression data, revealed that combining markers significantly improves classification performance. For example, using ETS2 with CDKN2D enhanced model sensitivity and specificity significantly (Table 3, Supplementary Fig. 2), highlighting the complex signature of neurescence that cannot be captured by a single gene. While finding reliable markers for neurescence remains a difficult task, our paired-marker strategy, particularly using genes like CDKN2D and ETS2 , can offer a promising direction for future studies on neurescence. As expected, pathways associated with neurodegeneration are upregulated in neurescence. Interestingly, we identified 16 DE genes that are overexpressed in all of these upregulated pathways. These 16 genes are related to mitochondrial function and cytoskeletal structure. While the association between mitochondrial function and neurodegeneration has been known 59 , 74 – 80 , our contribution is to show that these pathways are enriched specifically in neurescence. Follow-up studies to elucidate the biological mechanisms linking these genes and mitochondrial function and neurodegeneration could facilitate the discovery of novel therapeutic targets for AD. Our findings can help such studies focus on mitigating the impact of cellular senescence on neurodegeneration.⁤ This study provides promising markers for identifying neurescence, albeit some limitations need attention. Using the original counts to fit decision trees led to non-zero thresholds, which are undesirable for histology markers because distinguishing between levels of expression under the microscope is practically challenging. One would prefer markers that are totally absent in negative cells and ideally, have more than one transcript per cell in senescent neurons to ensure the observable signal is above background. Therefore, we used the binary values. Positive signals using antibodies against cyclin-dependent kinase inhibitors associated with senescence have historically been unreliable 81 . CDKN2D , cyclin-dependent kinase 4 inhibitor, encodes p19 INK 4 D , which we found to be elevated in NFT-bearing neurons in a prior study 16 However, that study revealed that not all neurescent cells expressed p19 INK 4 D , and not all p19 INK 4 D -positive cells were neurescent. We hypothesize that co-staining with antibodies against p19 INK 4 D and ETS2 may increase specificity of neurescence identification, but this remains to be experimentally determined. Another limitation working with snRNA-seq data lies in the inherent variability among the datasets on hand, despite all being derived from the same region of postmortem human brains. For instance, the distribution of total expression of senescence markers per cell varies considerably across the four datasets (Fig. 6 ). In particular, Mathys 2019 and Mathys 2024 display a more even distribution of gene expression across different levels, while Zhou 2020 and Xiong 2023 exhibit a more skewed distribution towards sparser expression. These differences highlight the challenge of comparing datasets generated by different research groups and under different experimental conditions. The observed variability may stem from technical error, different sample handling procedures, instruments, or biological factors such as the heterogeneity of sampled brain regions and variable progression of neurodegenerative diseases in donors. Also, while our study focused on neuronal senescence, it is important to note that other brain cell types, such as astrocytes, microglia, and oligodendrocyte-lineage cells, have also been reported to exhibit senescent-like phenotypes 82 – 84 . Given the likely variation in senescence marker profiles across these cell types, future studies applying similar approaches to characterize non-neuronal senescence will be essential for the development of targeted senolytics in the brain. Addressing these challenges could pave the way for expanding this study to include more diverse datasets, including those from earlier disease stages and other brain regions. Such an approach would help validate the robustness of the markers identified here. Methods The snRNA-Seq datasets We used four snRNA-Seq datasets generated by Mathys et al. 41 , Zhou et al. 42 , Xiong et al. 43 , and Mathys et al. 44 , which are accessible through the Accelerating Medicines Partnership–AD (AMP–AD) 85 with synapse IDs syn18485175, syn2112584, syn52293424, and syn52293442, respectively. The former dataset, referred to as Mathys 2019, was used in all analyses in this study as the discovery (i.e., train) dataset. The latters, referred to as Zhou 2020, Xiong 2023, and Mathys 2024 respectively, were used to validate the robustness and sensitivity of the proposed markers and our decision tree models. All samples were originally generated by longitudinal clinical-pathologic cohort studies of aging and Alzheimer’s disease (AD) from the Religious Order Study (ROS) and the Rush Memory and Aging Project (MAP) 86 . We employed the Synapser ( https://r-docs.synapse.org/articles/synapser.html ) R package 87 (Version 0.6.61) and a custom R script (Version 4.4.1) to download the four snRNA-seq datasets 41 – 44 . We downloaded clinical data from the corresponding publication pages. The snRNA-seq datasets included approximately 80,000, 70,000, 400,000, and 1.3 million single nuclei samples, with median postmortem interval (PMI) of 7, 6, 6, and 6 hours, respectively. We included only excitatory and inhibitory neurons in our analysis (Table 1). The first three datasets used samples from the dorsal prefrontal cortex of 48, 32, and 92 postmortem human brains, respectively. The Mathys 2024 dataset expanded the scope to include samples from entorhinal cortex (EC), hippocampus (HC), anterior thalamus (TH), angular gyrus (AG), midtemporal cortex (MT), and prefrontal cortex (PFC) regions of the same brain across 48 postmortem human samples. In this study, we used only the PFC samples, which included approximately 255,000 single nuclei transcriptomes. Eigengene analysis An eigengene for a given set is the first principal component (PCA) 88 , which is the weighted average of expression of all genes in the set. For each of the three SenMayo, CSP, and SIP gene lists, we used the compute.pigengene() function from the Pigengene package (Version 1.30.0) 40 to compute an eigengene. We addressed the challenge of cell type imbalance by implementing weighted PCA using the weight.pca() function from the Pigengene package to ensure that each cell type contributed equitably to the analysis. Specifically, each cell was weighted by dividing the total number of cells by the frequency of the corresponding cell type. This approach assigns higher weights to rarer cell types and lower weights to more abundant ones, thus balancing their influence in the PCA. We used the project.eigen() function from the Pigengene package to infer eigengene values in the validation datasets. This function computes each eigengene in the validation dataset using the same weights learned from the discovery dataset. After the weighted average is computed, the inferred eigengene was normalized to have the same Euclidean norm as the original eigengene. Cell labeling Three eigengenes were computed based on three independent gene sets associated with senescence: (1) Canonical Senescence Pathway (CSP) 16 , (2) Senescence Initiating Pathway (SIP) 16 , and (3) SenMayo 6 , which consisted of 22, 48, and 125 genes, respectively. The actual number of genes contributed to our eigengene analysis was slightly lower because the discovery dataset included only 22, 44, and 116 of these genes, respectively. We calculated the mean and standard deviation of each eigengene based on the empirical distribution in the discovery dataset. Any cell expressing an eigengene more than mean plus three standard deviations (i.e, mean + 3 s.d.) was considered “overexpressing” the eigengene, and was labeled as CSP + , SIP + , or SenMayo + , depending on their respective gene set. In contrast, any cell expressing the eigengene below the mean was labeled as CSP − , SIP − , and SenMayo − , respectively. Cells overexpressing all the three eigengenes were labeled as “senescent”, indicating a consensus across the three gene sets. In contrast, cells were labeled “non-senescent” when they expressed all the three eigengenes less than the corresponding means (Table 1). Other cells that did not meet either of these two criteria were considered borderline and excluded from our analysis. We used the phyper() function from the stats R package (Version. 4.4.1) 87 to perform a hypergeometric test with the null hypothesis that the number of senescent cells observed in a cell type is more than what would be expected at random. Differential expression analysis Our DE analysis was based on the senescent neurons that were overexpressing all three eigengenes (above mean + 3 s.d.) compared to the non-senescent neurons (below mean) (Table 1). We also filtered out 7,158 genes that did not have the minimum expression of 1 in at least 200 neurons. We normalized the data by multiplying all nuclei counts by the total library size of 1 million and transformed it to logarithmic space in base 2 (log 2 ). Given the significant imbalance between the groups, the smaller senescent class was upsampled by repeating each neuron 36 times so that the the number of non-senescent and senescent neurons appeared roughly equal for the DE analysis. We performed the differential expression analysis for senescent and non-senescent neurons, using two popular methods developed for scRna-seq, and the overlapped genes were chosen as the final differentially expressed genes. The first method was carried out using the MAST package (Version 1.30.0) 46 , 89 , which implemented a hurdle model 89 for analyzing scRNA-seq data that consists of a two-part generalized linear model. Considering the bimodality characteristic in single-cell expression data, MAST jointly models the positive mean expression (continuous) and the rates of expression (discrete) values. In this method, genes with a false discovery rate (i.e., adjusted p-value) less than 0.01 and an absolute value of log 2 fold change above 6, were declared as differentially expressed genes. The second method, SigEMD (Version 0.21.1) 47 , 90 , which is a custom R script that uses the nonparametric Earth Mover’s Distance (EMD) 91 . EMD is a special case of Wasserstein metric 92 , and measures the distance between gene expression distributions. Accordingly, a predefined adjusted p-values under 0.01 and an EMD score more than 30 were set to identify differentially expressed genes in senescent vs non-senescent neurons. Heatmap of differentially expressed genes was generated using the pheatmap.type() function from the Pigengene R package 40 , 93 . Decision trees Our primary objective in fitting decision trees to the discovery dataset was to identify the most predictive markers. We used the C50 package (Version 0.1.8) 94 in R to construct multiple decision trees based on different criteria. These trees differ in two ways: a) the eigengenes that served as the basis for neuron labeling and b) the selection of genes used as features in each tree (Table 2). To prevent overfitting, we set the parameter minCases of the function C5.0() to relatively large values to control the number of genes that contribute to a decision tree. This parameter specifies the minimum number of samples required in at least two splits of the tree, thus a larger minCases value results in a smaller decision tree. In this way, none of our trees was allowed to select more than two genes from the pool of genes available to it. We assessed whether the marker genes selected by the decision tree trained on the discovery dataset could accurately classify neurons in the validation datasets. The predict() function was applied on validation datasets to derive performance metrics of the decision trees fitted to the discovery dataset. This function predicted the senescence status of each neuron based on the trees fitted to the discovery dataset and the expression levels in the validation datasets. Particularly, the same cutoff values on gene expression values that were learned from the discovery dataset were used for the validation dataset. The predictions by decision trees, which used a few single genes, were then compared to the eigengene-based senescence labels. This comparison was done in each of the validation sets to compute accuracy, sensitivity, and specificity 95 of the trees (Table 3). Specifically, the neurons that were identified senescent by both a decision tree and eigengene approaches were considered true positives. In contrast, true negatives were the neurons that were not senescent based on both approaches. In each dataset, a) accuracy was defined as the number of true positives over the total number of neurons in the dataset, b) sensitivity was defined as the number of true positives over the number of neurescents based on our eigengene approach, and c) specificity was the ratio of true negatives over the total number of neurons that were not senescent based on our eigengene approach. The trees varied in the labels used to train them and also in the pool of genes they could select from. Initially, for each of the CSP, SIP, and SenMayo gene lists, we calculated their corresponding eigengenes, along with the positive and negative labels. Note that the resulting labelings could differ across gene lists. Using these three eigengenes and their corresponding gene lists, we trained three independent decision trees (Table 2, columns 1–3). Each tree specifically represented the unique characteristics derived from one of the gene lists. We trained a fourth decision tree using the senescent and non-senescent neuron labels unanimously defined based on all of our three eigengenes (Methods, Cell labeling) (Table 1). The fourth tree was allowed to use any of the 180 unique genes from the pooled set of all the genes in the CSP, SIP, and SenMayo lists (Table 2, columns 4 and 5). To train the fifth tree, we used the same senescent and non-senescent labels utilized in the fourth tree, but expanded the gene pool to include the 324 differentially expressed genes in addition to the previously combined CSP, SIP, and SenMayo gene lists. This led to including a total of 499 unique genes for training the fifth tree (Table 2, column 6 and 7). Furthermore, our methodology involved training the decision trees twice: first using the original counts directly from the expression matrix that quantified expression levels, and secondly by converting the expression levels into binary values, where any expression level above zero was considered 1. Pathway enrichment analysis We performed a pathway enrichment analysis based on the Kyoto Encyclopedia of Genes and Genomes 48 (KEGG) database using the get.enriched.pw() function from the Pigengene R package (Version 1.30.0) 40 , 96 , 97 . In this way, we identified the pathways that were overrepresented (adjusted p-values of the hypergeometric test < 0.05) by the genes that were differentially expressed between senescent and non-senescent neurons. Furthermore, we used the supertest() function from SuperExactTest R package (Version 1.1.0) 62 to assess the statistical significance of overlapping genes among multiple pathways. Declarations Data Availability The snRNA-Seq data analyzed in this study are available from https://www.synapse.org/ with synapse IDs: syn18485175, syn2112584, syn52293424, and syn52293442 for Mathys 2019, Zhou 2020, Xiong 2023, and Mathys 2024, respectively. Accessing these data requires submitting a Data Use Certificate through the AMP–AD website. Clinical data were available in the corresponding publication pages. Code availability Our R scripts can be used to fully reproduce our results. Our code is publicly available at https://bitbucket.org/habilzare/alzheimer/src/master/code/senescence/Shiva/ Acknowledgements: The authors would like to thank Xuehan “Victoria” Sun and Timothy Orr for their critical insights on wet lab validation. We obtained ROSMAP data from the AD Knowledge Portal (https://adknowledgeportal.synapse.org). Study data were provided by the Rush Alzheimer’s Disease Center, Rush University Medical Center, Chicago. Data generation was supported by National Institute on Aging (NIA, Grant Numbers: RF1AG57473, P30AG10161, R01AG15819, R01AG17917, U01G46152, U01AG61356, and RF1AG059082). Additional phenotypic ROSMAP data can be requested at https://www.radc.rush.edu. M.E.O. is supported by the Alzheimer’s Drug Discovery Foundation (GC-201908-2019443), Cure Alzheimer’s Fund, Hevolution/American Federation for Aging Research, NIA (R01AG068293, R01AG065839, U54AG079754, R24AG073199), National Institute of Neurological Disorders and Stroke (NINDS, R21NS125171), the Rainwater Charitable Foundation and US Department of Veterans Affairs (I01BX005717). H.Z. was supported by NIA (R01AG057896, 1RF1AG063507, R01AG068293, 1R01AG0665241A, 1R01AG065301, P30 AG066546, and 1R21AG087907) and NINDS (RF1NS112391, U19NS115388, and R015R01NS017950-39). S.K.D was supported by NIA (R01AG068293), the Alzheimer’s Association International Conference Travel Award 2024, and the Graduate School of Biomedical Sciences summer Travel Award 2024 from the University of Texas Health Science Center at San Antonio. We acknowledge the Texas Advanced Computing Center (TACC) at The University of Texas at Austin for providing high-performance computing (HPC) resources: http://www.tacc.utexas.edu. Author contributions: All authors edited and approved the final manuscript. In addition, S.K.D. developed code, designed and performed in silico experiments, prepared all tables and figures 1, 5, 6, and drafted the manuscript; S.S. conducted differential expression and pathway enrichment analyses, prepared figures 2, 3, and 4; A.H. provided expertise on various methods of differential expression analysis on snRNA-Seq data; M.E.O. conceived the study, acquired funding, and interpreted bioinformatic data; H.Z. contributed to the study design, interpreting the results, drafting and submitting the manuscript, oversaw S.K.D., S.S., and A.H. on the in silico experiments, acquired access to datasets, reviewed code, and confirmed bioinformatics data analyses. Competing interests: Dr. Orr has a patent on detecting and treating conditions associated with neuronal senescence with application number 18/694166. References Kumari, R. & Jat, P. Mechanisms of cellular senescence: Cell cycle arrest and senescence associated secretory phenotype. Front. Cell Dev. Biol. 9, 645593 (2021). Tsai, C.-H. et al. Up-regulation of cofilin-1 in cell senescence associates with morphological change and p27kip1 -mediated growth delay. Aging Cell 20, e13288 (2021). Oliva, J. L., Caino, M. C., Senderowicz, A. M. & Kazanietz, M. G. S-Phase-specific activation of PKC alpha induces senescence in non-small cell lung cancer cells. J. Biol. Chem. 283, 5466–5476 (2008). Purcell, M., Kruger, A. & Tainsky, M. A. Gene expression profiling of replicative and induced senescence. Cell Cycle 13, 3927–3937 (2014). Hudgins, A. D. et al. Age- and tissue-specific expression of senescence biomarkers in mice. Front. Genet. 9, 59 (2018). Saul, D. et al. A new gene set identifies senescent cells and predicts senescence-associated pathways across tissues. Nat. Commun. 13, 4827 (2022). Xu, P. et al. The landscape of human tissue and cell type specific expression and co-regulation of senescence genes. Mol. Neurodegener. 17, 5 (2022). Nagano, T. et al. Identification of cellular senescence-specific genes by comparative transcriptomics. Sci. Rep. 6, 31758 (2016). Deursen, V. & Jan, M. The role of senescent cells in ageing. Nature 509, 439–446 (2014). Rodier, F. & Campisi, J. Four faces of cellular senescence. J. Cell Biol. 192, 547–556 (2011). Hayflick, L. & Moorhead, P. S. The serial cultivation of human diploid cell strains. Exp. Cell Res. 25, 585–621 (1961). Cohn, R. L., Gasek, N. S., Kuchel, G. A. & Xu, M. The heterogeneity of cellular senescence: insights at the single-cell level. Trends Cell Biol. 33, 9–17 (2023). Bryant, A. et al. Endothelial cells are heterogeneous in different brain regions and are dramatically altered in Alzheimer’s disease. J. Neurosci. 43, 4541–4557 (2023). Kirschner, K., Rattanavirotkul, N., Quince, M. F. & Chandra, T. Functional heterogeneity in senescence. Biochem. Soc. Trans. 48, 765–773 (2020). Ogrodnik, M. et al. Whole-body senescent cell clearance alleviates age-related brain inflammation and cognitive impairment in mice. Aging Cell 20, e13296 (2021). Dehkordi, S. K. et al. Profiling senescent cells in human brains reveals neurons with CDKN2D/p19 and tau neuropathology. Nat Aging 1, 1107–1116 (2021). Diniz, B. S. et al. Mild cognitive impairment and major depressive disorder are associated with molecular senescence abnormalities in older adults. Alzheimers. Dement. 7, e12129 (2021). Musi, N. et al. Tau protein aggregation is associated with cellular senescence in the brain. Aging Cell 17, e12840 (2018). Herdy, J. R. et al. Increased post-mitotic senescence in aged human neurons is a pathological feature of Alzheimer’s disease. Cell Stem Cell 29, 1637–1652.e6 (2022). Ting, K. K. et al. Vascular senescence and leak are features of the early breakdown of the blood-brain barrier in Alzheimer’s disease models. GeroScience 45, 3307–3331 (2023). Acosta, J. C. et al. A complex secretory program orchestrated by the inflammasome controls paracrine senescence. Nat. Cell Biol. 15, 978–990 (2013). Xu, M. et al. JAK inhibition alleviates the cellular senescence-associated secretory phenotype and frailty in old age. Proc. Natl. Acad. Sci. U. S. A. 112, E6301–10 (2015). Schafer, M. J. et al. The senescence-associated secretome as an indicator of age and medical risk. JCI Insight 5, (2020). Limbad, C. et al. Astrocyte senescence promotes glutamate toxicity in cortical neurons. PLoS One 15, e0227887 (2020). Zhang, P. et al. Senolytic therapy alleviates Aβ-associated oligodendrocyte progenitor cell senescence and cognitive deficits in an Alzheimer’s disease model. Nat. Neurosci. 22, 719–728 (2019). Oliveri, D., Moschetti, G., Griego, A. & Scarpa, E. Endothelial cellular senescence and tau accumulation: An interplay full of opportunities? Ibrain 10, 225–230 (2024). Neri, F. et al. Senescent cell heterogeneity and responses to senolytic treatment are related to cell cycle status during cell growth arrest. bioRxiv (2024) doi: 10.1101/2024.06.22.600200 . Avelar, R. A. et al. A multidimensional systems biology analysis of cellular senescence in aging and disease. Genome Biol. 21, 91 (2020). Crespo-Garcia, S. et al. Pathological angiogenesis in retinopathy engages cellular senescence and is amenable to therapeutic elimination via BCL-xL inhibition. Cell Metab. 33, 818–832.e7 (2021). Wang, L. et al. Targeting p21Cip1 highly expressing cells in adipose tissue alleviates insulin resistance in obesity. Cell Metab. 34, 75–89.e8 (2022). Shimi, T. et al. The role of nuclear lamin B1 in cell proliferation and senescence. Genes Dev. 25, 2579–2593 (2011). Omori, S. et al. Generation of a p16 reporter mouse and its use to characterize and target p16high cells in vivo. Cell Metab. 32, 814–828.e6 (2020). Hickson, L. J. et al. Senolytics decrease senescent cells in humans: Preliminary report from a clinical trial of Dasatinib plus Quercetin in individuals with diabetic kidney disease. EBioMedicine 47, 446–456 (2019). Hudson, H. R., Riessland, M. & Orr, M. E. Defining and characterizing neuronal senescence, ‘neurescence’, as GX arrested cells. Trends Neurosci. 47, 971–984 (2024). Aguado, J. et al. Senolytic therapy alleviates physiological human brain aging and COVID-19 neuropathology. Nat. Aging 3, 1561–1575 (2023). Zhu, Y. et al. The Achilles’ heel of senescent cells: from transcriptome to senolytic drugs. Aging Cell 14, 644–658 (2015). Ashraf, H. M., Fernandez, B. & Spencer, S. L. The intensities of canonical senescence biomarkers integrate the duration of cell-cycle withdrawal. Nat. Commun. 14, 4527 (2023). Gorgoulis, V. et al. Cellular senescence: Defining a path forward. Cell 179, 813–827 (2019). Oldham, M. C., Horvath, S. & Geschwind, D. H. Conservation and evolution of gene coexpression networks in human and chimpanzee brains. Proc. Natl. Acad. Sci. U. S. A. 103, 17973–17978 (2006). Foroushani, A. et al. Large-scale gene network analysis reveals the significance of extracellular matrix pathway and homeobox genes in acute myeloid leukemia: an introduction to the Pigengene package and its applications. BMC Med. Genomics 10, 16 (2017). Mathys, H. et al. Single-cell transcriptomic analysis of Alzheimer’s disease. Nature 570, 332–337 (2019). Zhou, Y. et al. Author Correction: Human and mouse single-nucleus transcriptomics reveal TREM2-dependent and TREM2-independent cellular responses in Alzheimer’s disease. Nat. Med. 26, 981 (2020). Xiong, X. et al. Epigenomic dissection of Alzheimer’s disease pinpoints causal variants and reveals epigenome erosion. Cell 186, 4422–4437.e21 (2023). Mathys, H. et al. Single-cell multiregion dissection of Alzheimer’s disease. Nature 632, 858–868 (2024). Quinlan, J. R. Induction of decision trees. Mach. Learn. 1, 81–106 (1986). Odessa Press. MAST: A Flexible Statistical Framework for Assessing Transcriptional Changes and Characterizing Heterogeneity in Single-Cell RNA Sequencing Data . (CreateSpace Independent Publishing Platform, 2016). Wang, T. & Nabavi, S. SigEMD: A powerful method for differential gene expression analysis in single-cell RNA sequencing data. Methods 145, 25–32 (2018). Kanehisa, M. & Goto, S. KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res. 28, 27–30 (2000). Bobela, W., Aebischer, P. & Schneider, B. L. Αlpha-synuclein as a mediator in the interplay between aging and Parkinson’s disease. Biomolecules 5, 2675–2700 (2015). Ludtmann, M. H. R. & Abramov, A. Y. Mitochondrial calcium imbalance in Parkinson’s disease. Neurosci. Lett. 663, 86–90 (2018). Kume, K. et al. CGG repeat expansion in LRP12 in amyotrophic lateral sclerosis. Am. J. Hum. Genet. 110, 1086–1097 (2023). Takuma, K., Yan, S. S., Stern, D. M. & Yamada, K. Mitochondrial dysfunction, endoplasmic reticulum stress, and apoptosis in Alzheimer’s disease. J. Pharmacol. Sci. 97, 312–316 (2005). Rai, S. N., Singh, C., Singh, A., Singh, M. P. & Singh, B. K. Mitochondrial dysfunction: A potential therapeutic target to treat Alzheimer’s disease. Mol. Neurobiol. 57, 3075–3088 (2020). Bossy-Wetzel, E., Petrilli, A. & Knott, A. B. Mutant huntingtin and mitochondrial dysfunction. Trends Neurosci. 31, 609–616 (2008). Wang, C. et al. Novel role of COX6c in the regulation of oxidative phosphorylation and diseases. Cell Death Discov. 8, 336 (2022). Spielmann, N. et al. Knockout of the Complex III subunit Uqcrh causes bioenergetic impairment and cardiac contractile dysfunction. Mamm. Genome 34, 229–243 (2023). Santiago-Mujika, E., Luthi-Carter, R., Giorgini, F. & Mukaetova-Ladinska, E. B. Tubulin isotypes and posttranslational modifications in vascular dementia and Alzheimer’s disease. J. Alzheimers Dis. Rep. 6, 739–748 (2022). Bittermann, E. et al. Differential requirements of tubulin genes in mammalian forebrain development. PLoS Genet. 15, e1008243 (2019). Bustamante-Barrientos, F. A. et al. Mitochondrial dysfunction in neurodegenerative disorders: Potential therapeutic application of mitochondrial transfer to central nervous system-residing cells. J. Transl. Med. 21, 613 (2023). Miwa, S., Kashyap, S., Chini, E. & von Zglinicki, T. Mitochondrial dysfunction in cell senescence and aging. J. Clin. Invest. 132, (2022). Ashleigh, T., Swerdlow, R. H. & Beal, M. F. The role of mitochondrial dysfunction in Alzheimer’s disease pathogenesis. Alzheimers. Dement. 19, 333–342 (2023). Wang, M., Zhao, Y. & Zhang, B. Efficient Test and Visualization of Multi-Set Intersections. Sci. Rep. 5, 16923 (2015). Herdy, J. R., Mertens, J. & Gage, F. H. Neuronal senescence may drive brain aging. Science 384, 1404–1406 (2024). Amor, S., Puentes, F., Baker, D. & van der Valk, P. Inflammation in neurodegenerative diseases. Immunology 129, 154–169 (2010). Merelli, A., Repetto, M., Lazarowski, A. & Auzmendi, J. Hypoxia, oxidative stress, and inflammation: Three faces of neurodegenerative diseases. J. Alzheimers. Dis. 82, S109–S126 (2021). Wiley, C. D. et al. Mitochondrial dysfunction induces senescence with a distinct secretory phenotype. Cell Metab. 23, 303–314 (2016). Victorelli, S. et al. Apoptotic stress causes mtDNA release during senescence and drives the SASP. Nature 622, 627–636 (2023). Taipa, R. et al. Proinflammatory and anti-inflammatory cytokines in the CSF of patients with Alzheimer’s disease and their correlation with cognitive decline. Neurobiol. Aging 76, 125–132 (2019). Albrecht, D. S. et al. Early neuroinflammation is associated with lower amyloid and tau levels in cognitively normal older adults. Brain Behav. Immun. 94, 299–307 (2021). Shafqat, A. et al. Cellular senescence in brain aging and cognitive decline. Front. Aging Neurosci. 15, 1281581 (2023). Lopez-Rodriguez, A. B. et al. Acute systemic inflammation exacerbates neuroinflammation in Alzheimer’s disease: IL-1β drives amplified responses in primed astrocytes and neuronal network dysfunction. Alzheimers. Dement. 17, 1735–1755 (2021). Riessland, M. et al. Loss of SATB1 induces p21-dependent cellular senescence in post-mitotic dopaminergic neurons. Cell Stem Cell 25, 514–530.e8 (2019). Traxler, L. et al. Warburg-like metabolic transformation underlies neuronal degeneration in sporadic Alzheimer’s disease. Cell Metab. 34, 1248–1263.e6 (2022). Schon, E. A. & Area-Gomez, E. Is Alzheimer’s disease a disorder of mitochondria-associated membranes? J. Alzheimers. Dis. 20 Suppl 2, S281–92 (2010). Xu, H., Liu, Y.-Y., Li, L.-S. & Liu, Y.-S. Sirtuins at the Crossroads between mitochondrial quality control and neurodegenerative diseases: Structure, regulation, modifications, and modulators. Aging Dis. 14, 794–824 (2023). Tang, Y., Huang, Y., Wan, Z., Zhou, B. & Wu, Z. Mitochondrial quality control links two seemingly unrelated neurodegenerative diseases. Autophagy 18, 2495–2497 (2022). Yan, X., Wang, B., Hu, Y., Wang, S. & Zhang, X. Abnormal mitochondrial quality control in neurodegenerative diseases. Front. Cell. Neurosci. 14, 138 (2020). Franco-Iborra, S., Vila, M. & Perier, C. Mitochondrial quality control in neurodegenerative diseases: Focus on Parkinson’s disease and Huntington's disease. Front. Neurosci. 12, 342 (2018). Mitochondrial Dysfunction in Alzheimer’s Disease: Role of Oxidative Phosphorylation Deficits . Mitochondrial Transplantation as a Novel Therapeutic Strategy for Neurodegenerative DiseasesMitochondrial Transplantation as a Novel Therapeutic Strategy for Neurodegenerative Diseases . Ogrodnik, M. et al. Guidelines for minimal information on cellular senescence experimentation in vivo. Cell 187, 4150–4175 (2024). Cohen, J. & Torres, C. Astrocyte senescence: Evidence and significance. Aging Cell 18, e12937 (2019). Schlett, J. S. et al. NF-κB is a critical mediator of post-mitotic senescence in oligodendrocytes and subsequent white matter loss. Mol. Neurodegener. 18, 24 (2023). Rim, C., You, M.-J., Nahm, M. & Kwon, M.-S. Emerging role of senescent microglia in brain aging-related neurodegenerative diseases. Transl. Neurodegener. 13, 10 (2024). Hodes, R. J. & Buckholtz, N. Accelerating Medicines Partnership: Alzheimer’s Disease (AMP-AD) Knowledge Portal Aids Alzheimer's Drug Discovery through Open Data Sharing. Expert Opin. Ther. Targets 20, 389–391 (2016). Bennett, D. A. et al. Religious Orders Study and Rush Memory and Aging Project. J. Alzheimers. Dis. 64, S161–S189 (2018). R Core Team, R. & Others. R: A language and environment for statistical computing. (2013). Pearson, K. On Lines and Planes of Closest Fit to Systems of Points in Space . (1901). Cragg, J. G. Some Statistical Models for Limited Dependent Variables with Application to the Demand for Durable Goods . (1969). Wang, T. & Nabavi, S. Differential gene expression analysis in single-cell RNA sequencing data. in 2017 IEEE International Conference on Bioinformatics and Biomedicine (BIBM) (IEEE, 2017). doi: 10.1109/bibm.2017.8217650 . Rubner, Y., Tomasi, C. & Guibas, L. J. A metric for distributions with applications to image databases. in Sixth International Conference on Computer Vision (IEEE Cat. No.98CH36271) (Narosa Publishing House, 2002). doi: 10.1109/iccv.1998.710701 . Kantorovič, L. V. Mathematical Methods of Organizing and Planning Production . (1960). Kolde, R. pheatmap: Pretty Heatmaps. CRAN: Contributed Packages The R Foundation https://doi.org/10.32614/cran.package.pheatmap (2010). Kuhn, M. & Johnson, K. Classification Trees and Rule-Based Models. in Applied Predictive Modeling (eds. Kuhn, M. & Johnson, K.) 369–413 (Springer New York, New York, NY, 2013). doi: 10.1007/978-1-4614-6849-3_14 . Yerushalmy, J. Statistical problems in assessing methods of medical diagnosis, with special reference to X-ray techniques. Public Health Rep. 62, 1432–1449 (1947). Yu, G., Wang, L.-G., Han, Y. & He, Q.-Y. clusterProfiler: an R package for comparing biological themes among gene clusters. OMICS 16, 284–287 (2012). Wu, T. et al. clusterProfiler 4.0: A universal enrichment tool for interpreting omics data. Innovation (Camb) 2, 100141 (2021). Tables Tables 1 to 3 are available in the Supplementary Files section. Additional Declarations Competing interest reported. Dr. Orr has a patent on detecting and treating conditions associated with neuronal senescence with application number ​​18/694166. Supplementary Files Table1.docx Table 1. Number of identified senescent neurons in the discovery dataset (Mathys 2019), and the three external validation datasets: Zhou 2020, Xiong 2023, and Mathys 2024. Table2.docx Table 2. The top selected genes for identification of senescent neurons based on different criteria. *The accuracy of the corresponding tree is written in parentheses. **When no tree with two markers were found, a tree with three markers is shown. ***If no single gene was found, both selected markers would be removed in the elimination analysis. Table3.docx Table 3. Validation performance of top-performing two-gene decision trees that were trained using the discovery dataset (Mathys 2019). Each model was evaluated based on three independent validation datasets: Zhou 2020, Xiong 2023, and Mathys 2024, using the same eigengene thresholds identified in the discovery cohort. Accuracy, sensitivity, and specificity were calculated by comparing the model's predictions to ground-truth labels derived from eigengene expression thresholds. On each row, the genes in the “Removed” column are excluded leading to selection of the pair of genes in the “Selected” column. SupplementaryfiguresSKD.pdf Supptable1.xlsx Supptable2.xlsx Cite Share Download PDF Status: Published Journal Publication published 01 Jul, 2025 Read the published version in npj Aging → Version 1 posted Editorial decision: Revision requested 23 Apr, 2025 Reviews received at journal 21 Apr, 2025 Reviews received at journal 21 Apr, 2025 Reviewers agreed at journal 07 Apr, 2025 Reviews received at journal 07 Apr, 2025 Reviewers agreed at journal 07 Apr, 2025 Reviewers agreed at journal 07 Apr, 2025 Reviewers invited by journal 07 Apr, 2025 Submission checks completed at journal 05 Apr, 2025 First submitted to journal 28 Mar, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-5903682","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":439727322,"identity":"33d44204-88cd-4565-a32d-cbd212b58f5d","order_by":0,"name":"Shiva Kazempour Dehkordi","email":"","orcid":"","institution":"The University of Texas Health Science Center at San Antonio","correspondingAuthor":false,"prefix":"","firstName":"Shiva","middleName":"Kazempour","lastName":"Dehkordi","suffix":""},{"id":439727323,"identity":"b0e91330-a599-425a-bf80-6fd89db04121","order_by":1,"name":"Sogand Sajedi","email":"","orcid":"","institution":"The University of Texas MD Anderson Cancer Center","correspondingAuthor":false,"prefix":"","firstName":"Sogand","middleName":"","lastName":"Sajedi","suffix":""},{"id":439727324,"identity":"ea25882b-5c01-4f2a-9233-1a7c285dd51b","order_by":2,"name":"Amirreza Heshmat","email":"","orcid":"","institution":"The University of Texas MD Anderson Cancer Center","correspondingAuthor":false,"prefix":"","firstName":"Amirreza","middleName":"","lastName":"Heshmat","suffix":""},{"id":439727325,"identity":"d306e508-2892-472b-b22a-de1b66c43709","order_by":3,"name":"Miranda E. Orr","email":"","orcid":"","institution":"Washington University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Miranda","middleName":"E.","lastName":"Orr","suffix":""},{"id":439727326,"identity":"2ae45337-ecc5-4704-aa45-57557b4c932c","order_by":4,"name":"Habil Zare","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAu0lEQVRIiWNgGAWjYLCCDwwMCVAmM3E6GGcwGJCohZmHJC3y7WcfPrZt+5Mn33784QeGCuvEBoKO6kk3Ns5tMyg2OJNjLMFwJp2wFmaGNDZpoJbEDRI8DBKMbYcJa2Hjf8b+2xKoZf4M9sc/GP8RoYVHIo2NmRGopeEGg5kEYwMRWiQknjFL9pwzTtxwJsfMIuFYujFBLfL9aYwffpTJJc5vP/74xocaa1mCWlBBAmnKR8EoGAWjYBTgAgA91jmnxTxvCQAAAABJRU5ErkJggg==","orcid":"","institution":"The University of Texas Health Science Center at San Antonio","correspondingAuthor":true,"prefix":"","firstName":"Habil","middleName":"","lastName":"Zare","suffix":""}],"badges":[],"createdAt":"2025-01-25 21:53:06","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5903682/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5903682/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41514-025-00235-y","type":"published","date":"2025-07-01T15:57:54+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":80215400,"identity":"4f0fbb73-1412-4606-87c0-6058201ae81f","added_by":"auto","created_at":"2025-04-09 09:29:08","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":294069,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eThe decision tree analysis based on CSP, SIP, and SenMayo identified senescence markers in neurons.\u003c/strong\u003e The trees show the classification of neurons into senescent (red) and non-senescent (blue). Each node represents a decision point based on \u003cstrong\u003e(a)\u003c/strong\u003e gene expression \u003cstrong\u003elevels\u003c/strong\u003e of\u003cstrong\u003e the \u003c/strong\u003eoriginal counts, and \u003cstrong\u003e(b)\u003c/strong\u003e the binary transformed expression. This binary tree reached zero false positive and thus a specificity of 1.v\u003c/p\u003e","description":"","filename":"11.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5903682/v1/6637b51083ce712c568b68cc.jpg"},{"id":80214447,"identity":"bc84e8eb-52ca-4dfd-a4fa-f3e14717d186","added_by":"auto","created_at":"2025-04-09 09:21:08","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":707811,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eThe heatmap of differentially expressed genes identified by comparing senescent versus non-senescent neurons. \u003c/strong\u003eThe heatmap shows the expression profile of 324 genes (columns) in the 17,346 cells (rows) in the discovery dataset. These genes were consistently identified as differentially expressed by both MAST and SigEMD methods.\u003c/p\u003e","description":"","filename":"12.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5903682/v1/091508a5dc64a6890f6268c0.jpg"},{"id":80215780,"identity":"c56d49dc-0c21-4044-ae2c-890aac7e3867","added_by":"auto","created_at":"2025-04-09 09:37:08","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":414045,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePathways overrepresented by genes that are differentially expressed in senescent neurons. \u003c/strong\u003eThe\u003cstrong\u003e \u003c/strong\u003ebubble chart displays the significantly enriched pathways identified from the pathway analysis of 324 differentially expressed genes. Each bubble represents a pathway, with the size indicating the number of DE genes in that pathway and the color denoting the adjusted p-value for overlap significance.\u003c/p\u003e","description":"","filename":"13.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5903682/v1/4136c998edb88c6baf245bbb.jpg"},{"id":80214456,"identity":"8de4c290-7d6c-4f99-9ccb-357696209186","added_by":"auto","created_at":"2025-04-09 09:21:08","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":347416,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eNumber of DE genes in the top enriched pathways. \u003c/strong\u003eThe number of DE genes in each pathway is listed on the right. The UpSet\u003ca href=\"https://paperpile.com/c/1d46cO/Lkl8\"\u003e\u003csup\u003e62\u003c/sup\u003e\u003c/a\u003e plot shows the sizes of the overlaps among DE genes and the top five significantly enriched pathways named on rows. In particular, the last column shows 16 DE genes are members of all these five pathways, which is a significant overlap (p-value \u0026lt; 10\u003csup\u003e-8\u003c/sup\u003e).\u003c/p\u003e","description":"","filename":"14.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5903682/v1/f49bf18095fa4a8596eb7518.jpg"},{"id":80214457,"identity":"c2ceaf41-b5b6-4b10-b25f-3aec1f22366b","added_by":"auto","created_at":"2025-04-09 09:21:08","extension":"jpg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":299205,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDecision tree analysis based on CSP, SIP, SenMayo, and DE genes identified novel senescence markers in neurons.\u003c/strong\u003e The trees show the classification of neurons into senescent (red) and non-senescent (blue) using a pair of selected markers from a pool of differentially expressed genes merged with CSP, SIP, and SenMayo gene lists. Each node represents a decision point based on \u003cstrong\u003e(a) \u003c/strong\u003egene expression thresholds of original counts\u003cem\u003e,\u003c/em\u003e and\u003cstrong\u003e (b)\u003c/strong\u003e binary transformed expression.\u003c/p\u003e","description":"","filename":"15.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5903682/v1/e59293a3bebe8f6e3191e588.jpg"},{"id":80214458,"identity":"558d1b3a-bae7-4567-a271-ccdcc415e447","added_by":"auto","created_at":"2025-04-09 09:21:08","extension":"jpg","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":318169,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDistribution of total expression of 499 senescence marker genes. \u003c/strong\u003eThe plot illustrates the inherent variability in gene expression distributions among datasets derived from postmortem human brain samples (Table 1). The x-axis represents the sum of expression values of all marker genes in each cell. The x-axis is grouped into bins of size 1000. While Mathys 2019 (green) and Mathys 2024 (red) exhibit a more even distribution of gene expression, Zhou 2020 (purple) and Xiong 2023 (blue) display skewed distributions towards sparser expression.\u003c/p\u003e","description":"","filename":"16.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5903682/v1/ef80835f3364764930cae401.jpg"},{"id":86179057,"identity":"5178cf2d-cc88-404c-8133-0ff15d2ad5b5","added_by":"auto","created_at":"2025-07-07 16:15:07","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3425098,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5903682/v1/042b2717-a3be-4dda-ae8a-39d635018826.pdf"},{"id":80214449,"identity":"e40c814d-6c18-47ec-9068-61ff3301c4bb","added_by":"auto","created_at":"2025-04-09 09:21:08","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":7819,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eTable 1.\u003c/strong\u003e Number of identified senescent neurons in the discovery dataset (Mathys 2019), and the three external validation datasets: Zhou 2020, Xiong 2023, and Mathys 2024.\u003c/p\u003e","description":"","filename":"Table1.docx","url":"https://assets-eu.researchsquare.com/files/rs-5903682/v1/2ceefad2598f5ed1f273aaa1.docx"},{"id":80214451,"identity":"8c398e44-dde2-4bcb-b252-f4a8d02e17fc","added_by":"auto","created_at":"2025-04-09 09:21:08","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":8684,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eTable 2\u003c/strong\u003e. The top selected genes for identification of senescent neurons based on different criteria.\u003c/p\u003e\n\u003cp\u003e*The accuracy of the corresponding tree is written in parentheses.\u003c/p\u003e\n\u003cp\u003e**When no tree with two markers were found, a tree with three markers is shown.\u003c/p\u003e\n\u003cp\u003e***If no single gene was found, both selected markers would be removed in the elimination analysis.\u003c/p\u003e","description":"","filename":"Table2.docx","url":"https://assets-eu.researchsquare.com/files/rs-5903682/v1/e5786f3218c4ce76db1006b9.docx"},{"id":80215401,"identity":"62ab3541-b429-45f4-8c3f-4a1d28ba89e2","added_by":"auto","created_at":"2025-04-09 09:29:08","extension":"docx","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":29285,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eTable 3.\u003c/strong\u003e Validation performance of top-performing two-gene decision trees that were trained using the discovery dataset (Mathys 2019). Each model was evaluated based on three independent validation datasets: Zhou 2020, Xiong 2023, and Mathys 2024, using the same eigengene thresholds identified in the discovery cohort. Accuracy, sensitivity, and specificity were calculated by comparing the model's predictions to ground-truth labels derived from eigengene expression thresholds. On each row, the genes in the “Removed” column are excluded leading to selection of the pair of genes in the “Selected” column.\u003c/p\u003e","description":"","filename":"Table3.docx","url":"https://assets-eu.researchsquare.com/files/rs-5903682/v1/a17f20a8c0a3843cf10f1b2a.docx"},{"id":80215405,"identity":"2e0ad994-26db-4dee-92cd-317cd4995e52","added_by":"auto","created_at":"2025-04-09 09:29:08","extension":"pdf","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":1217299,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryfiguresSKD.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5903682/v1/3d68a091cd551efcd871c68d.pdf"},{"id":80215406,"identity":"8501c5b0-b6f4-4f3a-af44-4aa66700caf6","added_by":"auto","created_at":"2025-04-09 09:29:08","extension":"xlsx","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":20006,"visible":true,"origin":"","legend":"","description":"","filename":"Supptable1.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-5903682/v1/75c62b95f06daad75ba64932.xlsx"},{"id":80215782,"identity":"edd7a147-f5fe-4dde-b7f2-5c02f8828902","added_by":"auto","created_at":"2025-04-09 09:37:09","extension":"xlsx","order_by":6,"title":"","display":"","copyAsset":false,"role":"supplement","size":79028,"visible":true,"origin":"","legend":"","description":"","filename":"Supptable2.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-5903682/v1/c5e5942aad4d4dd5e0c5c3d5.xlsx"}],"financialInterests":"Competing interest reported. Dr. Orr has a patent on detecting and treating conditions associated with neuronal senescence with application number ​​18/694166.","formattedTitle":"Identification of markers for neurescence through transcriptomic profiling of postmortem human brains","fulltext":[{"header":"Introduction","content":"\u003cp\u003eCellular senescence is a complex, multi-step process characterized by stable cell cycle arrest\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e, changes in cell morphology\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e,\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e, changes in gene expression\u003csup\u003e\u003cspan additionalcitationids=\"CR5 CR6 CR7\" citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e, and a pro-inflammatory secretory phenotype\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e,\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e. Senescence was first identified in fibroblasts\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e, is heterogeneous in tissues throughout the body\u003csup\u003e\u003cspan additionalcitationids=\"CR13\" citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e, and has been studied in the context of the human brain\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e,\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e. Increasing evidence implicates cellular senescence in brain aging and its links to neurodegenerative disorders \u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e,\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e, including Alzheimer's disease (AD)\u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e,\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e. The presence of senescent cells triggers a pro-inflammatory environment\u003csup\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e and contributes to neuron loss, tissue dysfunction and cognitive impairment in animal models of AD\u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e,\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e. Recent studies using AD mouse models and human postmortem brain tissues directly demonstrate that neuronal senescence associated with tau pathology drives neuronal loss and neurodegeneration\u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e, and that senescent human astrocytes can induce neuronal death through excitotoxicity in co-culture systems\u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e. In the human brain, senescent cells have been associated with the known AD pathologies e.g., amyloid plaques and neurofibrillary tangles\u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e,\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e,\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eDespite their significance, defining and identifying senescent cells in the human brain remains challenging due to their heterogeneous nature\u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e. Although reliable markers have been utilized to identify senescent cells in specific tissues such as adipose tissue, retinal endothelial cells, and fibroblast\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e,\u003cspan additionalcitationids=\"CR29 CR30 CR31 CR32\" citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e\u003c/sup\u003e, a universally applicable set of senescence markers across diverse tissues remain poorly defined. This challenge is especially pronounced in neurons, where historically defined senescence markers have not been robustly validated \u003csup\u003e\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e. This knowledge gap has limited our understanding of when senescent cells first appear in the adult brain, inferring their contribution to neurodegenerative disease pathophysiology and our ability to develop potential therapies to modulate or remove senescent cells from the human brain\u003csup\u003e\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e,\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eOne recent attempt to generate a comprehensive list of senescent markers resulted in the SenMayo gene list\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e, which includes 125 genes that are highly correlated with age and the expression level of p16 and p21, cyclin-dependent kinase inhibitors that are upregulated in many senescent cells. The majority of genes in the SenMayo panel are senescence-associated secretory phenotype (SAPS) factors. However, cellular senescence has other aspects, including activation of senescent cell anti-apoptotic pathways (SCAPs)\u003csup\u003e\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e\u003c/sup\u003e, organelle dysfunction, and morphology changes. To provide a comprehensive and multidimensional perspective for neuronal senescence (neurescent), we employed an eigengene approach based on a) the SenMayo gene list, in addition to our two previously published lists\u003csup\u003e\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e including b) Canonical Senescence Pathway (CSP) with 22 genes, which reflect cell cycle arrest\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e,\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e,\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e\u003c/sup\u003e, and c) Senescence Initiating Pathway (SIP) panel with 48 genes, which are upregulated in early senescence and activate SCAPs.\u003c/p\u003e \u003cp\u003eTechnically, an eigengene is computed as a weighted average expression of all genes in a given list\u003csup\u003e\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e,\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e\u003c/sup\u003ederived using principal component analysis on those particular genes. In this study, we computed an eigengene\u003csup\u003e\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e\u003c/sup\u003e for each of the SenMayo, CSP, and SIP gene lists and used all these eigengenes for three main analyses: 1) identifying senescent and non-senescent cells, 2) identifying the most accurate markers for neuronal senescence, referred to as neurescence\u003csup\u003e\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e, and 3) identifying differentially expressed genes in senescent vs non-senescent neurons.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eWe used four independent single-nucleus RNA sequencing (snRNA-seq) datasets of the dorsal prefrontal cortex from postmortem human brains were used, which are here referred to as Mathys 2019\u003csup\u003e41\u003c/sup\u003e (n\u0026thinsp;=\u0026thinsp;80,000), Zhou 2020\u003csup\u003e42\u003c/sup\u003e (n\u0026thinsp;=\u0026thinsp;70,000), Xiong 2023\u003csup\u003e43\u003c/sup\u003e (n\u0026thinsp;=\u0026thinsp;400,000), and Mathys 2024\u003csup\u003e44\u003c/sup\u003e (n\u0026thinsp;=\u0026thinsp;255,000), respectively (Table\u0026nbsp;1). Mathys 2019 served as the discovery dataset to identify markers and the other three datasets were used to validate the performance of the identified markers.\u003c/p\u003e \u003cp\u003ePerforming the eigengene approach\u003csup\u003e\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e on the Mathys 2019 discovery dataset, we identified 475 senescent and 16,871 non-senescent neurons (Table\u0026nbsp;1). Specifically, the identified 475 neurescent cells expressed each of the three CSP, SIP, and SenMayo eigengenes more than the corresponding mean plus three times the standard deviation (mean\u0026thinsp;+\u0026thinsp;3 s.d.). In contrast, if a neuron expressed each of these eigengenes less than the corresponding mean, then that neuron was considered non-senescent (N\u0026thinsp;=\u0026thinsp;16,871). The remaining 53,288 borderline neurons were excluded from our analysis. We used the identified neurescent and non-senescent neurons to train decision trees and to perform differential expression (DE) analysis. Of note, only excitatory neurons displayed consistent expression across all three senescence eigengenes, and subsequent differential expression analysis and marker identification were restricted exclusively to excitatory neurons.\u003c/p\u003e \u003cp\u003eTo further assess the robustness of our method and the reproducibility of the eigengene-derived senescence markers, we conducted a reciprocal validation analysis. We used Mathys 2024 as the discovery dataset to independently derive the eigengene weights for CSP, SIP, and SenMayo, and then applied these derived weights to Mathys 2019 as a validation dataset. Comparing the weights from the two independent discovery datasets (Mathys 2019 vs. Mathys 2024) (Supplementary Table\u0026nbsp;1) revealed high concordance: specifically, correlations of 94% for CSP weights, 88% for SIP weights, and 88% for SenMayo weights, respectively (Supplementary Fig.\u0026nbsp;3). Furthermore, when using the Mathys 2024-derived eigengene weights to identify senescent cells in the Mathys 2019 dataset, we identified 413 senescent neurons, closely matching the original 475 neurons that were identified using Mathys 2019 as discovery. Notably, 370 neurons were identified as senescent in both approaches, representing a highly significant overlap (hypergeometric test p-value \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\u0026lt;1{0}^{-1714}\\)\u003c/span\u003e\u003c/span\u003e). Collectively, these results demonstrate the robust consistency and reproducibility of our eigengene-based senescence identification framework, regardless of the initial discovery dataset selection.\u003c/p\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eIdentification of specific markers for neurons among genes generally associated with senescence\u003c/h2\u003e \u003cp\u003eTo identify potential markers for neurescence, we fitted relatively small decision trees\u003csup\u003e\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e\u003c/sup\u003e to the discovery dataset. To ensure the selected genes were specific, each tree was allowed to use no more than two genes from the pool of gene sets known to be associated with cellular senescence\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e,\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e, as indicated on the third row of Table\u0026nbsp;2.\u003c/p\u003e \u003cp\u003eUsing the original counts to quantify gene expression levels, the best marker among the 180 senescence-associated genes was \u003cem\u003eMAP2K1\u003c/em\u003e leading to a single-gene tree with the highest accuracy of 99%, a sensitivity of 80%, and a specificity of 99% (Table\u0026nbsp;2, column 4). Adding \u003cem\u003eCDKN2D\u003c/em\u003e as the second marker increased sensitivity to 93% with a negligible (i.e., \u0026lt;\u0026thinsp;0.1%) effect on accuracy. The corresponding decision tree, which was based on \u003cem\u003eMAP2K1 and CDKN2D\u003c/em\u003e, had the best accuracy among all of our trees that could use two of the 180 senescence-associated genes. This suggested that these two genes could serve as effective markers and complement each other in identifying neurescence. In the corresponding tree, the thresholds for the expression levels of these genes were 2 and 1, respectively (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ea), suggesting relatively low expression of these putative marker genes. These thresholds are undesirable for histology markers because distinguishing between levels of expression under the microscope is practically challenging. One would prefer markers that are totally absent in negative cells and ideally, have more than one transcript per cell in senescent neurons to ensure the observable signal is above background.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTo provide greater confidence in these lowly expressed marker genes, we repeated our analysis using a binary transformation of the discovery data, where any expression level above zero was converted to 1. Using these binary values, no single gene was found to accurately classify the senescent and non-senescent cells. This suggests that there may not be a single marker in our gene lists specifically expressed in neurescence. Interestingly, the best tree with two genes still included \u003cem\u003eCDKN2D\u003c/em\u003e, but \u003cem\u003eMAP2K1\u003c/em\u003e was replaced with \u003cem\u003eETS2\u003c/em\u003e (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eb, last row of Table\u0026nbsp;2). The binary transformation did not change the accuracy but dropped the sensitivity from 93\u0026ndash;83%, while increasing the specificity to 100%. After removing \u003cem\u003eCDKN2D\u003c/em\u003e and \u003cem\u003eETS2\u003c/em\u003e from the analysis, \u003cem\u003eE2F3\u003c/em\u003e and \u003cem\u003eRB1\u003c/em\u003e were selected in the second best tree (Table\u0026nbsp;2, column 5), leading to an accuracy of 99%, specificity of 100%, but a relatively low sensitivity of 67%.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eDifferentially expressed genes\u003c/h3\u003e\n\u003cp\u003eTo investigate differentially expressed genes in neurescence compared to non-senescent, we performed a DE analysis on the Mathys 2019 discovery dataset. Taking an agnostic approach, we included all 10,768 genes that had non-negligible expression in neurons of the discovery dataset (Methods). We employed two state of the art methods for DE analysis of scRNA-Seq data. The first method, MAST\u003csup\u003e\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e\u003c/sup\u003e, resulted in 375 differentially expressed genes, whereas using the second method, SigEMD\u003csup\u003e\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e\u003c/sup\u003e, we identified 576 differentially expressed genes (Supplementary Table\u0026nbsp;2). The two analyses shared 324 differentially expressed genes (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, Supplementary Table\u0026nbsp;2), which represented a significant overlap (p-value \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\u0026lt;1{0}^{-205}\\)\u003c/span\u003e\u003c/span\u003e, hypergeometric test). These 324 genes were overexpressed in the 475 neurescent cells compared to other neurons (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). In this study, we primarily focused on these 324 genes because they were identified by both DE analysis methods.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e\n\u003ch3\u003ePathway analysis\u003c/h3\u003e\n\u003cp\u003eTo understand the functional significance of the 324 differentially expressed genes, we performed a pathway analysis using Kyoto Encyclopedia of Genes and Genomes (KEGG)\u003csup\u003e\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e\u003c/sup\u003e and identified 18 significantly enriched pathways (p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05, Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). Interestingly, the top pathways were related to neurodegeneration including Parkinson disease\u003csup\u003e\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e,\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e\u003c/sup\u003e, Amyotrophic lateral sclerosis\u003csup\u003e\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e\u003c/sup\u003e, Alzheimer\u0026rsquo;s disease\u003csup\u003e\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e,\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e\u003c/sup\u003e, and Huntington's disease\u003csup\u003e\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e\u003c/sup\u003e, suggesting common underlying biological mechanisms that might be associated with neuronal senescence.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe top 5 enriched pathways shared 16 DE genes, which is a significant overlap (p-value \u0026lt; \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:1{0}^{-8}\\)\u003c/span\u003e\u003c/span\u003e, hypergeometric test, Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). These genes included \u003cem\u003eCOX6C, COX7A2L, COX7C, CYCS, KIF5A, NDUFA4, NDUFS1, PSMA7, TUBA4A, TUBB2A, TUBB4A, TUBB4B, UQCRB, UQCRC2, UQCRH\u003c/em\u003e, and \u003cem\u003eUQCRHL\u003c/em\u003e. These 16 genes are mainly involved in mitochondrial function (COX, NDUF families)\u003csup\u003e\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e,\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e\u003c/sup\u003e and cytoskeletal structure (TUBB family)\u003csup\u003e\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e,\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e\u003c/sup\u003e. Mitochondrial dysfunction leads to impaired energy metabolism and increased production of reactive oxygen species\u003csup\u003e\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e\u003c/sup\u003e. It is a hallmark of both cellular senescence\u003csup\u003e\u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e\u003c/sup\u003e and AD\u003csup\u003e\u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e61\u003c/span\u003e\u003c/sup\u003e. The deregulation of these genes supports prior work demonstrating mitochondrial dysfunction in neurescence, and their role in the neurodegeneration\u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e\n\u003ch3\u003eIdentification of specific markers for senescent neurons among differentially expressed genes\u003c/h3\u003e\n\u003cp\u003eWe hypothesized that there might be some specific markers for neurescence beyond the genes generally known to be associated with senescence (e.g., CSP, SIP, and SenMayo). To test this hypothesis, we extended our decision tree analysis to include the 324 differentially expressed genes identified based on the discovery dataset. Using the original count data, the top gene was \u003cem\u003eDPYSL2\u003c/em\u003e resulting in a relatively accurate classification (i.e., accuracy: 99%, sensitivity: 95%, and specificity: 99%) and when the gene \u003cem\u003eATP6V1H\u003c/em\u003e was added to make a two-gene tree, sensitivity increased to 99% (Table\u0026nbsp;2, column 6). However, for these trees to be useful, meaning that they identify molecules that can serve as biomarkers in histological assays, one must be able to determine whether these genes were expressed at relatively high levels in a neuron (i.e., above 4 and 3, respectively) (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ea). To address this practical issue, we used binary expression levels leading to selection of \u003cem\u003eUQCRHL\u003c/em\u003e together with \u003cem\u003eETS2\u003c/em\u003e (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eb). The recurrent selection of \u003cem\u003eETS2\u003c/em\u003e (Table\u0026nbsp;2, columns 4 and 6) highlighted its importance even when additional differentially expressed genes were added to the analysis. Since the inclusion of \u003cem\u003eUQCRHL\u003c/em\u003e led to very low sensitivity in the validation cohorts (Table\u0026nbsp;3), we considered removing this gene. Excluding \u003cem\u003eUQCRHL\u003c/em\u003e and \u003cem\u003eETS2\u003c/em\u003e led to selection of \u003cem\u003eCDKN2D\u003c/em\u003e and \u003cem\u003eRB1\u003c/em\u003e again and increasing the specificity of senescence classification to 100% in the discovery dataset. Overall, adding differentially expressed genes did not seem to be helpful as the only reasonable tree in columns 6 and 7 was based on \u003cem\u003eCDKN2D\u003c/em\u003e and \u003cem\u003eRB1\u003c/em\u003e, which were already in our CSP gene lists. This observation falsified the hypothesis that additional differentially expressed genes could identify specific markers for senescent neurons.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e\n\u003ch3\u003eIdentification of alternative markers for senescent neurons\u003c/h3\u003e\n\u003cp\u003eThe relatively low expression of the selected markers, combined with historical challenges in developing antibodies specific to senescence markers, necessitates that \u003cem\u003ein silico\u003c/em\u003e studies identify multiple candidate molecules for subsequent \u003cem\u003ein situ\u003c/em\u003e validation. We performed a systematic elimination analysis. We pooled all genes from CSP, SIP, and SenMayo lists with DE genes and fitted decision trees on the binary expression values in the discovery dataset, Mathys 2019 (Table\u0026nbsp;3). Our systematic approach involved iteratively removing the top-performing markers to evaluate their impact on model performance and to identify the next best set of markers.\u003c/p\u003e \u003cp\u003eIncluding \u003cem\u003eUQCRHL\u003c/em\u003e in decision trees led to poor sensitivity in the validation datasets. Thus, we excluded \u003cem\u003eUQCRHL\u003c/em\u003e and found \u003cem\u003eCDKN2D\u003c/em\u003e and \u003cem\u003eETS2\u003c/em\u003e to be the second best pair. If reagents to detect \u003cem\u003eETS2\u003c/em\u003e are unavailable or lack specificity, then \u003cem\u003eCDKN2D\u003c/em\u003e and RB1 would be the next best pair of markers for neurescence. Our results showed that removing \u003cem\u003eCDKN2D\u003c/em\u003e generally led to decreased accuracy, sensitivity, and specificity in validation datasets. This drop in model performance highlights the critical role \u003cem\u003eCDKN2D\u003c/em\u003e plays in the classification. However, it is important to note that the currently available antibodies for p19, the protein product of \u003cem\u003eCDKN2D\u003c/em\u003e, may lack specificity, which could complicate its use in histological applications.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIdentifying senescent neurons in the human brain is challenging due to the inherent heterogeneity of senescence, neuronal subtypes and the current lack of reliable markers specific to neurescence in the human brain\u003csup\u003e\u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e63\u003c/span\u003e\u003c/sup\u003e. To tackle this challenge, we systematically and unbiasedly analyzed scRNA-seq data from four datasets of the dorsal prefrontal cortex from postmortem human brains. Our eigengene-based method allowed us to focus on three key gene lists: SenMayo, CSP, and SIP, each reflecting distinct aspects of cellular senescence.\u003c/p\u003e \u003cp\u003eWhile the recently published SenMayo list is a valuable resource, it primarily captures the inflammatory aspects of senescence. The majority of genes in this list are associated with Senescence-associated secretory phenotype and other important characteristics of neural senescence may not be represented by SenMayo. Inflammation is a critical component, particularly in the brain, where it has been linked to neurodegeneration\u003csup\u003e\u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e64\u003c/span\u003e,\u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e65\u003c/span\u003e\u003c/sup\u003e. The accumulation of senescent cells can lead to a persistent pro-inflammatory state, driven by mitochondrial dysfunction \u003csup\u003e\u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e66\u003c/span\u003e,\u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e67\u003c/span\u003e\u003c/sup\u003e, and concomitant elevated levels of actors like IL-6, IL-1β, and TNF-α, which are characteristic of SASP\u003csup\u003e\u003cspan additionalcitationids=\"CR69\" citationid=\"CR68\" class=\"CitationRef\"\u003e68\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e70\u003c/span\u003e\u003c/sup\u003e. This chronic inflammation can impair tissue homeostasis and contribute to neuronal dysfunction\u003csup\u003e\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e71\u003c/span\u003e\u003c/sup\u003e. However, inflammation alone does not fully capture the multifaceted nature of cellular senescence, where other mechanisms such as cell cycle arrest, DNA damage, mitochondrial dysfunction, and metabolic alterations also play critical roles\u003csup\u003e\u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e63\u003c/span\u003e,\u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e72\u003c/span\u003e,\u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e73\u003c/span\u003e\u003c/sup\u003e. Moreover, inflammation can occur independent of senescence. This inherent limitation of the SenMayo list underscores the need for a more comprehensive approach. Accordingly, in this study, we included CSP and SIP gene panels to capture other important senescence features such as cell cycle arrest and early stress responses. Our approach leveraged eigengenes derived from these three panels to classify neurons into senescent and non-senescent groups. Our integrative approach using multiple senescence-associated gene panels aims to improve specificity by capturing their combined activity rather than relying on any single pathway.\u003c/p\u003e \u003cp\u003eOur findings demonstrated that a single marker gene is insufficient for accurately classifying neurescence. Decision tree analysis, incorporating multiple senescence panels and both continuous and binary expression data, revealed that combining markers significantly improves classification performance. For example, using \u003cem\u003eETS2\u003c/em\u003e with \u003cem\u003eCDKN2D\u003c/em\u003e enhanced model sensitivity and specificity significantly (Table\u0026nbsp;3, Supplementary Fig.\u0026nbsp;2), highlighting the complex signature of neurescence that cannot be captured by a single gene. While finding reliable markers for neurescence remains a difficult task, our paired-marker strategy, particularly using genes like \u003cem\u003eCDKN2D\u003c/em\u003e and \u003cem\u003eETS2\u003c/em\u003e, can offer a promising direction for future studies on neurescence.\u003c/p\u003e \u003cp\u003eAs expected, pathways associated with neurodegeneration are upregulated in neurescence. Interestingly, we identified 16 DE genes that are overexpressed in all of these upregulated pathways. These 16 genes are related to mitochondrial function and cytoskeletal structure. While the association between mitochondrial function and neurodegeneration has been known\u003csup\u003e\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e,\u003cspan additionalcitationids=\"CR75 CR76 CR77 CR78 CR79\" citationid=\"CR74\" class=\"CitationRef\"\u003e74\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR80\" class=\"CitationRef\"\u003e80\u003c/span\u003e\u003c/sup\u003e, our contribution is to show that these pathways are enriched specifically in neurescence. Follow-up studies to elucidate the biological mechanisms linking these genes and mitochondrial function and neurodegeneration could facilitate the discovery of novel therapeutic targets for AD. Our findings can help such studies focus on mitigating the impact of cellular senescence on neurodegeneration.⁤\u003c/p\u003e \u003cp\u003eThis study provides promising markers for identifying neurescence, albeit some limitations need attention. Using the original counts to fit decision trees led to non-zero thresholds, which are undesirable for histology markers because distinguishing between levels of expression under the microscope is practically challenging. One would prefer markers that are totally absent in negative cells and ideally, have more than one transcript per cell in senescent neurons to ensure the observable signal is above background. Therefore, we used the binary values.\u003c/p\u003e \u003cp\u003ePositive signals using antibodies against cyclin-dependent kinase inhibitors associated with senescence have historically been unreliable \u003csup\u003e\u003cspan citationid=\"CR81\" class=\"CitationRef\"\u003e81\u003c/span\u003e\u003c/sup\u003e. \u003cem\u003eCDKN2D\u003c/em\u003e, cyclin-dependent kinase 4 inhibitor, encodes p19\u003csup\u003eINK\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003eD\u003c/sup\u003e, which we found to be elevated in NFT-bearing neurons in a prior study \u003csup\u003e\u003cem\u003e\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/em\u003e\u003c/sup\u003e However, that study revealed that not all neurescent cells expressed p19\u003csup\u003eINK\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003eD\u003c/sup\u003e, and not all p19\u003csup\u003eINK\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003eD\u003c/sup\u003e-positive cells were neurescent. We hypothesize that co-staining with antibodies against p19\u003csup\u003eINK\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003eD\u003c/sup\u003e and \u003cem\u003eETS2\u003c/em\u003e may increase specificity of neurescence identification, but this remains to be experimentally determined.\u003c/p\u003e \u003cp\u003eAnother limitation working with snRNA-seq data lies in the inherent variability among the datasets on hand, despite all being derived from the same region of postmortem human brains. For instance, the distribution of total expression of senescence markers per cell varies considerably across the four datasets (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e). In particular, Mathys 2019 and Mathys 2024 display a more even distribution of gene expression across different levels, while Zhou 2020 and Xiong 2023 exhibit a more skewed distribution towards sparser expression. These differences highlight the challenge of comparing datasets generated by different research groups and under different experimental conditions. The observed variability may stem from technical error, different sample handling procedures, instruments, or biological factors such as the heterogeneity of sampled brain regions and variable progression of neurodegenerative diseases in donors. Also, while our study focused on neuronal senescence, it is important to note that other brain cell types, such as astrocytes, microglia, and oligodendrocyte-lineage cells, have also been reported to exhibit senescent-like phenotypes \u003csup\u003e\u003cspan additionalcitationids=\"CR83\" citationid=\"CR82\" class=\"CitationRef\"\u003e82\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR84\" class=\"CitationRef\"\u003e84\u003c/span\u003e\u003c/sup\u003e. Given the likely variation in senescence marker profiles across these cell types, future studies applying similar approaches to characterize non-neuronal senescence will be essential for the development of targeted senolytics in the brain. Addressing these challenges could pave the way for expanding this study to include more diverse datasets, including those from earlier disease stages and other brain regions. Such an approach would help validate the robustness of the markers identified here.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003eThe snRNA-Seq datasets\u003c/h2\u003e \u003cp\u003eWe used four snRNA-Seq datasets generated by Mathys et al.\u003csup\u003e\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e\u003c/sup\u003e, Zhou et al.\u003csup\u003e\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e\u003c/sup\u003e, Xiong et al.\u003csup\u003e\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e\u003c/sup\u003e, and Mathys et al.\u003csup\u003e\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e\u003c/sup\u003e, which are accessible through the Accelerating Medicines Partnership\u0026ndash;AD (AMP\u0026ndash;AD)\u003csup\u003e\u003cspan citationid=\"CR85\" class=\"CitationRef\"\u003e85\u003c/span\u003e\u003c/sup\u003e with synapse IDs syn18485175, syn2112584, syn52293424, and syn52293442, respectively. The former dataset, referred to as Mathys 2019, was used in all analyses in this study as the discovery (i.e., train) dataset. The latters, referred to as Zhou 2020, Xiong 2023, and Mathys 2024 respectively, were used to validate the robustness and sensitivity of the proposed markers and our decision tree models. All samples were originally generated by longitudinal clinical-pathologic cohort studies of aging and Alzheimer\u0026rsquo;s disease (AD) from the Religious Order Study (ROS) and the Rush Memory and Aging Project (MAP)\u003csup\u003e\u003cspan citationid=\"CR86\" class=\"CitationRef\"\u003e86\u003c/span\u003e\u003c/sup\u003e. We employed the Synapser (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://r-docs.synapse.org/articles/synapser.html\u003c/span\u003e\u003cspan address=\"https://r-docs.synapse.org/articles/synapser.html\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e)\u003c/span\u003e R package\u003csup\u003e\u003cspan citationid=\"CR87\" class=\"CitationRef\"\u003e87\u003c/span\u003e\u003c/sup\u003e (Version 0.6.61) and a custom R script (Version 4.4.1) to download the four snRNA-seq datasets\u003csup\u003e\u003cspan additionalcitationids=\"CR42 CR43\" citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e\u003c/sup\u003e. We downloaded clinical data from the corresponding publication pages. The snRNA-seq datasets included approximately 80,000, 70,000, 400,000, and 1.3\u0026nbsp;million single nuclei samples, with median postmortem interval (PMI) of 7, 6, 6, and 6 hours, respectively. We included only excitatory and inhibitory neurons in our analysis (Table\u0026nbsp;1). The first three datasets used samples from the dorsal prefrontal cortex of 48, 32, and 92 postmortem human brains, respectively. The Mathys 2024 dataset expanded the scope to include samples from entorhinal cortex (EC), hippocampus (HC), anterior thalamus (TH), angular gyrus (AG), midtemporal cortex (MT), and prefrontal cortex (PFC) regions of the same brain across 48 postmortem human samples. In this study, we used only the PFC samples, which included approximately 255,000 single nuclei transcriptomes.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eEigengene analysis\u003c/h2\u003e \u003cp\u003eAn eigengene for a given set is the first principal component (PCA)\u003csup\u003e\u003cspan citationid=\"CR88\" class=\"CitationRef\"\u003e88\u003c/span\u003e\u003c/sup\u003e, which is the weighted average of expression of all genes in the set. For each of the three SenMayo, CSP, and SIP gene lists, we used the compute.pigengene() function from the Pigengene package (Version 1.30.0)\u003csup\u003e\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e\u003c/sup\u003e to compute an eigengene. We addressed the challenge of cell type imbalance by implementing weighted PCA using the weight.pca() function from the Pigengene package to ensure that each cell type contributed equitably to the analysis. Specifically, each cell was weighted by dividing the total number of cells by the frequency of the corresponding cell type. This approach assigns higher weights to rarer cell types and lower weights to more abundant ones, thus balancing their influence in the PCA. We used the project.eigen() function from the Pigengene package to infer eigengene values in the validation datasets. This function computes each eigengene in the validation dataset using the same weights learned from the discovery dataset. After the weighted average is computed, the inferred eigengene was normalized to have the same Euclidean norm as the original eigengene.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eCell labeling\u003c/h2\u003e \u003cp\u003eThree eigengenes were computed based on three independent gene sets associated with senescence: (1) Canonical Senescence Pathway (CSP)\u003csup\u003e\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e, (2) Senescence Initiating Pathway (SIP)\u003csup\u003e\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e, and (3) SenMayo\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e, which consisted of 22, 48, and 125 genes, respectively. The actual number of genes contributed to our eigengene analysis was slightly lower because the discovery dataset included only 22, 44, and 116 of these genes, respectively. We calculated the mean and standard deviation of each eigengene based on the empirical distribution in the discovery dataset. Any cell expressing an eigengene more than mean plus three standard deviations (i.e, mean\u0026thinsp;+\u0026thinsp;3 s.d.) was considered \u0026ldquo;overexpressing\u0026rdquo; the eigengene, and was labeled as CSP\u003csup\u003e+\u003c/sup\u003e, SIP\u003csup\u003e+\u003c/sup\u003e, or SenMayo\u003csup\u003e+\u003c/sup\u003e, depending on their respective gene set. In contrast, any cell expressing the eigengene below the mean was labeled as CSP\u003csup\u003e\u0026minus;\u003c/sup\u003e, SIP\u003csup\u003e\u0026minus;\u003c/sup\u003e, and SenMayo\u003csup\u003e\u0026minus;\u003c/sup\u003e, respectively. Cells overexpressing all the three eigengenes were labeled as \u0026ldquo;senescent\u0026rdquo;, indicating a consensus across the three gene sets. In contrast, cells were labeled \u0026ldquo;non-senescent\u0026rdquo; when they expressed all the three eigengenes less than the corresponding means (Table\u0026nbsp;1). Other cells that did not meet either of these two criteria were considered borderline and excluded from our analysis. We used the phyper() function from the stats R package (Version. 4.4.1)\u003csup\u003e\u003cspan citationid=\"CR87\" class=\"CitationRef\"\u003e87\u003c/span\u003e\u003c/sup\u003e to perform a hypergeometric test with the null hypothesis that the number of senescent cells observed in a cell type is more than what would be expected at random.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eDifferential expression analysis\u003c/h2\u003e \u003cp\u003eOur DE analysis was based on the senescent neurons that were overexpressing all three eigengenes (above mean\u0026thinsp;+\u0026thinsp;3 s.d.) compared to the non-senescent neurons (below mean) (Table\u0026nbsp;1). We also filtered out 7,158 genes that did not have the minimum expression of 1 in at least 200 neurons. We normalized the data by multiplying all nuclei counts by the total library size of 1\u0026nbsp;million and transformed it to logarithmic space in base 2 (log\u003csub\u003e2\u003c/sub\u003e). Given the significant imbalance between the groups, the smaller senescent class was upsampled by repeating each neuron 36 times so that the the number of non-senescent and senescent neurons appeared roughly equal for the DE analysis. We performed the differential expression analysis for senescent and non-senescent neurons, using two popular methods developed for scRna-seq, and the overlapped genes were chosen as the final differentially expressed genes. The first method was carried out using the MAST package (Version 1.30.0)\u003csup\u003e\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e,\u003cspan citationid=\"CR89\" class=\"CitationRef\"\u003e89\u003c/span\u003e\u003c/sup\u003e, which implemented a hurdle model\u003csup\u003e\u003cspan citationid=\"CR89\" class=\"CitationRef\"\u003e89\u003c/span\u003e\u003c/sup\u003e for analyzing scRNA-seq data that consists of a two-part generalized linear model. Considering the bimodality characteristic in single-cell expression data, MAST jointly models the positive mean expression (continuous) and the rates of expression (discrete) values. In this method, genes with a false discovery rate (i.e., adjusted p-value) less than 0.01 and an absolute value of log\u003csub\u003e2\u003c/sub\u003e fold change above 6, were declared as differentially expressed genes. The second method, SigEMD (Version 0.21.1)\u003csup\u003e\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e,\u003cspan citationid=\"CR90\" class=\"CitationRef\"\u003e90\u003c/span\u003e\u003c/sup\u003e, which is a custom R script that uses the nonparametric Earth Mover\u0026rsquo;s Distance (EMD)\u003csup\u003e\u003cspan citationid=\"CR91\" class=\"CitationRef\"\u003e91\u003c/span\u003e\u003c/sup\u003e. EMD is a special case of Wasserstein metric\u003csup\u003e\u003cspan citationid=\"CR92\" class=\"CitationRef\"\u003e92\u003c/span\u003e\u003c/sup\u003e, and measures the distance between gene expression distributions. Accordingly, a predefined adjusted p-values under 0.01 and an EMD score more than 30 were set to identify differentially expressed genes in senescent vs non-senescent neurons. Heatmap of differentially expressed genes was generated using the pheatmap.type() function from the Pigengene R package\u003csup\u003e\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e,\u003cspan citationid=\"CR93\" class=\"CitationRef\"\u003e93\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003eDecision trees\u003c/h2\u003e \u003cp\u003eOur primary objective in fitting decision trees to the discovery dataset was to identify the most predictive markers. We used the C50 package (Version 0.1.8)\u003csup\u003e\u003cspan citationid=\"CR94\" class=\"CitationRef\"\u003e94\u003c/span\u003e\u003c/sup\u003e in R to construct multiple decision trees based on different criteria. These trees differ in two ways: a) the eigengenes that served as the basis for neuron labeling and b) the selection of genes used as features in each tree (Table\u0026nbsp;2). To prevent overfitting, we set the parameter minCases of the function C5.0() to relatively large values to control the number of genes that contribute to a decision tree. This parameter specifies the minimum number of samples required in at least two splits of the tree, thus a larger minCases value results in a smaller decision tree. In this way, none of our trees was allowed to select more than two genes from the pool of genes available to it.\u003c/p\u003e \u003cp\u003eWe assessed whether the marker genes selected by the decision tree trained on the discovery dataset could accurately classify neurons in the validation datasets. The predict() function was applied on validation datasets to derive performance metrics of the decision trees fitted to the discovery dataset. This function predicted the senescence status of each neuron based on the trees fitted to the discovery dataset and the expression levels in the validation datasets. Particularly, the same cutoff values on gene expression values that were learned from the discovery dataset were used for the validation dataset. The predictions by decision trees, which used a few single genes, were then compared to the eigengene-based senescence labels. This comparison was done in each of the validation sets to compute accuracy, sensitivity, and specificity\u003csup\u003e\u003cspan citationid=\"CR95\" class=\"CitationRef\"\u003e95\u003c/span\u003e\u003c/sup\u003e of the trees (Table\u0026nbsp;3). Specifically, the neurons that were identified senescent by both a decision tree and eigengene approaches were considered true positives. In contrast, true negatives were the neurons that were not senescent based on both approaches. In each dataset, a) accuracy was defined as the number of true positives over the total number of neurons in the dataset, b) sensitivity was defined as the number of true positives over the number of neurescents based on our eigengene approach, and c) specificity was the ratio of true negatives over the total number of neurons that were not senescent based on our eigengene approach.\u003c/p\u003e \u003cp\u003eThe trees varied in the labels used to train them and also in the pool of genes they could select from. Initially, for each of the CSP, SIP, and SenMayo gene lists, we calculated their corresponding eigengenes, along with the positive and negative labels. Note that the resulting labelings could differ across gene lists. Using these three eigengenes and their corresponding gene lists, we trained three independent decision trees (Table\u0026nbsp;2, columns 1\u0026ndash;3). Each tree specifically represented the unique characteristics derived from one of the gene lists.\u003c/p\u003e \u003cp\u003eWe trained a fourth decision tree using the senescent and non-senescent neuron labels unanimously defined based on all of our three eigengenes (Methods, Cell labeling) (Table\u0026nbsp;1). The fourth tree was allowed to use any of the 180 unique genes from the pooled set of all the genes in the CSP, SIP, and SenMayo lists (Table\u0026nbsp;2, columns 4 and 5). To train the fifth tree, we used the same senescent and non-senescent labels utilized in the fourth tree, but expanded the gene pool to include the 324 differentially expressed genes in addition to the previously combined CSP, SIP, and SenMayo gene lists. This led to including a total of 499 unique genes for training the fifth tree (Table\u0026nbsp;2, column 6 and 7).\u003c/p\u003e \u003cp\u003eFurthermore, our methodology involved training the decision trees twice: first using the original counts directly from the expression matrix that quantified expression levels, and secondly by converting the expression levels into binary values, where any expression level above zero was considered 1.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003ePathway enrichment analysis\u003c/h2\u003e \u003cp\u003eWe performed a pathway enrichment analysis based on the Kyoto Encyclopedia of Genes and Genomes\u003csup\u003e\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e\u003c/sup\u003e (KEGG) database using the get.enriched.pw() function from the Pigengene R package (Version 1.30.0)\u003csup\u003e\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e,\u003cspan citationid=\"CR96\" class=\"CitationRef\"\u003e96\u003c/span\u003e,\u003cspan citationid=\"CR97\" class=\"CitationRef\"\u003e97\u003c/span\u003e\u003c/sup\u003e. In this way, we identified the pathways that were overrepresented (adjusted p-values of the hypergeometric test\u0026thinsp;\u0026lt;\u0026thinsp;0.05) by the genes that were differentially expressed between senescent and non-senescent neurons. Furthermore, we used the supertest() function from SuperExactTest R package (Version 1.1.0)\u003csup\u003e\u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e62\u003c/span\u003e\u003c/sup\u003e to assess the statistical significance of overlapping genes among multiple pathways.\u003c/p\u003e \u003c/div\u003e "},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eData Availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe snRNA-Seq data analyzed in this study are available from https://www.synapse.org/ with synapse IDs: syn18485175, syn2112584, syn52293424, and syn52293442 for Mathys 2019, Zhou 2020, Xiong 2023, and Mathys 2024, respectively. Accessing these data requires submitting a Data Use Certificate through the AMP\u0026ndash;AD website. Clinical data were available in the corresponding publication pages.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCode availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eOur R scripts can be used to fully reproduce our results. Our code is publicly available at https://bitbucket.org/habilzare/alzheimer/src/master/code/senescence/Shiva/\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements:\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors would like to thank Xuehan \u0026ldquo;Victoria\u0026rdquo; Sun and Timothy Orr for their critical insights on wet lab validation. We obtained ROSMAP data from the AD Knowledge Portal (https://adknowledgeportal.synapse.org). Study data were provided by the Rush Alzheimer\u0026rsquo;s Disease Center, Rush University Medical Center, Chicago. Data generation was supported by National Institute on Aging (NIA, Grant Numbers: RF1AG57473, P30AG10161, R01AG15819, R01AG17917, U01G46152, U01AG61356, and RF1AG059082). Additional phenotypic ROSMAP data can be requested at https://www.radc.rush.edu. M.E.O. is supported by the Alzheimer\u0026rsquo;s Drug Discovery Foundation (GC-201908-2019443), Cure Alzheimer\u0026rsquo;s Fund, Hevolution/American Federation for Aging Research, NIA (R01AG068293, R01AG065839, U54AG079754, R24AG073199), National Institute of Neurological Disorders and Stroke (NINDS, R21NS125171), the Rainwater Charitable Foundation and US Department of Veterans Affairs (I01BX005717). H.Z. was supported by NIA (R01AG057896, 1RF1AG063507, R01AG068293, 1R01AG0665241A, 1R01AG065301, P30 AG066546, and 1R21AG087907) and NINDS (RF1NS112391, U19NS115388, and R015R01NS017950-39). S.K.D was supported by NIA (R01AG068293), the Alzheimer\u0026rsquo;s Association International Conference Travel Award 2024, and the Graduate School of Biomedical Sciences summer Travel Award 2024 from the University of Texas Health Science Center at San Antonio. We acknowledge the Texas Advanced Computing Center (TACC) at The University of Texas at Austin for providing high-performance computing (HPC) resources: http://www.tacc.utexas.edu.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll authors edited and approved the final manuscript. In addition, S.K.D. developed code, designed and performed \u003cem\u003ein silico\u003c/em\u003e experiments, prepared all tables and figures 1, 5, 6, and drafted the manuscript; S.S. conducted differential expression and pathway enrichment analyses, prepared figures 2, 3, and 4; A.H. provided expertise on various methods of differential expression analysis on snRNA-Seq data; M.E.O. conceived the study, acquired funding, and interpreted bioinformatic data; H.Z. contributed to the study design, interpreting the results, drafting and submitting the manuscript, oversaw S.K.D., S.S., and A.H. on the \u003cem\u003ein silico\u003c/em\u003e\u0026nbsp; experiments, acquired access to datasets, reviewed code, and confirmed bioinformatics data analyses.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eDr. Orr has a patent on detecting and treating conditions associated with neuronal senescence with application number 18/694166.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eKumari, R. \u0026amp; Jat, P. Mechanisms of cellular senescence: Cell cycle arrest and senescence associated secretory phenotype. Front. Cell Dev. Biol. 9, 645593 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTsai, C.-H. \u003cem\u003eet al.\u003c/em\u003e Up-regulation of cofilin-1 in cell senescence associates with morphological change and p27kip1 -mediated growth delay. Aging Cell 20, e13288 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOliva, J. L., Caino, M. C., Senderowicz, A. M. \u0026amp; Kazanietz, M. G. S-Phase-specific activation of PKC alpha induces senescence in non-small cell lung cancer cells. J. Biol. Chem. 283, 5466\u0026ndash;5476 (2008).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePurcell, M., Kruger, A. \u0026amp; Tainsky, M. A. Gene expression profiling of replicative and induced senescence. Cell Cycle 13, 3927\u0026ndash;3937 (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHudgins, A. D. \u003cem\u003eet al.\u003c/em\u003e Age- and tissue-specific expression of senescence biomarkers in mice. Front. Genet. 9, 59 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSaul, D. \u003cem\u003eet al.\u003c/em\u003e A new gene set identifies senescent cells and predicts senescence-associated pathways across tissues. Nat. Commun. 13, 4827 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXu, P. \u003cem\u003eet al.\u003c/em\u003e The landscape of human tissue and cell type specific expression and co-regulation of senescence genes. Mol. Neurodegener. 17, 5 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNagano, T. \u003cem\u003eet al.\u003c/em\u003e Identification of cellular senescence-specific genes by comparative transcriptomics. Sci. Rep. 6, 31758 (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDeursen, V. \u0026amp; Jan, M. The role of senescent cells in ageing. Nature 509, 439\u0026ndash;446 (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRodier, F. \u0026amp; Campisi, J. Four faces of cellular senescence. J. Cell Biol. 192, 547\u0026ndash;556 (2011).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHayflick, L. \u0026amp; Moorhead, P. S. The serial cultivation of human diploid cell strains. Exp. Cell Res. 25, 585\u0026ndash;621 (1961).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCohn, R. L., Gasek, N. S., Kuchel, G. A. \u0026amp; Xu, M. The heterogeneity of cellular senescence: insights at the single-cell level. Trends Cell Biol. 33, 9\u0026ndash;17 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBryant, A. \u003cem\u003eet al.\u003c/em\u003e Endothelial cells are heterogeneous in different brain regions and are dramatically altered in Alzheimer\u0026rsquo;s disease. J. Neurosci. 43, 4541\u0026ndash;4557 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKirschner, K., Rattanavirotkul, N., Quince, M. F. \u0026amp; Chandra, T. Functional heterogeneity in senescence. Biochem. Soc. Trans. 48, 765\u0026ndash;773 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOgrodnik, M. \u003cem\u003eet al.\u003c/em\u003e Whole-body senescent cell clearance alleviates age-related brain inflammation and cognitive impairment in mice. Aging Cell 20, e13296 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDehkordi, S. K. \u003cem\u003eet al.\u003c/em\u003e Profiling senescent cells in human brains reveals neurons with CDKN2D/p19 and tau neuropathology. Nat Aging 1, 1107\u0026ndash;1116 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDiniz, B. S. \u003cem\u003eet al.\u003c/em\u003e Mild cognitive impairment and major depressive disorder are associated with molecular senescence abnormalities in older adults. Alzheimers. Dement. 7, e12129 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMusi, N. \u003cem\u003eet al.\u003c/em\u003e Tau protein aggregation is associated with cellular senescence in the brain. Aging Cell 17, e12840 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHerdy, J. R. \u003cem\u003eet al.\u003c/em\u003e Increased post-mitotic senescence in aged human neurons is a pathological feature of Alzheimer\u0026rsquo;s disease. Cell Stem Cell 29, 1637\u0026ndash;1652.e6 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTing, K. K. \u003cem\u003eet al.\u003c/em\u003e Vascular senescence and leak are features of the early breakdown of the blood-brain barrier in Alzheimer\u0026rsquo;s disease models. GeroScience 45, 3307\u0026ndash;3331 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAcosta, J. C. \u003cem\u003eet al.\u003c/em\u003e A complex secretory program orchestrated by the inflammasome controls paracrine senescence. Nat. Cell Biol. 15, 978\u0026ndash;990 (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXu, M. \u003cem\u003eet al.\u003c/em\u003e JAK inhibition alleviates the cellular senescence-associated secretory phenotype and frailty in old age. \u003cem\u003eProc. Natl. Acad. Sci. U. S. A.\u003c/em\u003e 112, E6301\u0026ndash;10 (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchafer, M. J. \u003cem\u003eet al.\u003c/em\u003e The senescence-associated secretome as an indicator of age and medical risk. JCI Insight 5, (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLimbad, C. \u003cem\u003eet al.\u003c/em\u003e Astrocyte senescence promotes glutamate toxicity in cortical neurons. PLoS One 15, e0227887 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang, P. \u003cem\u003eet al.\u003c/em\u003e Senolytic therapy alleviates Aβ-associated oligodendrocyte progenitor cell senescence and cognitive deficits in an Alzheimer\u0026rsquo;s disease model. Nat. Neurosci. 22, 719\u0026ndash;728 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOliveri, D., Moschetti, G., Griego, A. \u0026amp; Scarpa, E. Endothelial cellular senescence and tau accumulation: An interplay full of opportunities? \u003cem\u003eIbrain\u003c/em\u003e 10, 225\u0026ndash;230 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNeri, F. \u003cem\u003eet al.\u003c/em\u003e Senescent cell heterogeneity and responses to senolytic treatment are related to cell cycle status during cell growth arrest. bioRxiv (2024) doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1101/2024.06.22.600200\u003c/span\u003e\u003cspan address=\"10.1101/2024.06.22.600200\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAvelar, R. A. \u003cem\u003eet al.\u003c/em\u003e A multidimensional systems biology analysis of cellular senescence in aging and disease. Genome Biol. 21, 91 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCrespo-Garcia, S. \u003cem\u003eet al.\u003c/em\u003e Pathological angiogenesis in retinopathy engages cellular senescence and is amenable to therapeutic elimination via BCL-xL inhibition. Cell Metab. 33, 818\u0026ndash;832.e7 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang, L. \u003cem\u003eet al.\u003c/em\u003e Targeting p21Cip1 highly expressing cells in adipose tissue alleviates insulin resistance in obesity. Cell Metab. 34, 75\u0026ndash;89.e8 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShimi, T. \u003cem\u003eet al.\u003c/em\u003e The role of nuclear lamin B1 in cell proliferation and senescence. Genes Dev. 25, 2579\u0026ndash;2593 (2011).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOmori, S. \u003cem\u003eet al.\u003c/em\u003e Generation of a p16 reporter mouse and its use to characterize and target p16high cells in vivo. Cell Metab. 32, 814\u0026ndash;828.e6 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHickson, L. J. \u003cem\u003eet al.\u003c/em\u003e Senolytics decrease senescent cells in humans: Preliminary report from a clinical trial of Dasatinib plus Quercetin in individuals with diabetic kidney disease. EBioMedicine 47, 446\u0026ndash;456 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHudson, H. R., Riessland, M. \u0026amp; Orr, M. E. Defining and characterizing neuronal senescence, \u0026lsquo;neurescence\u0026rsquo;, as GX arrested cells. Trends Neurosci. 47, 971\u0026ndash;984 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAguado, J. \u003cem\u003eet al.\u003c/em\u003e Senolytic therapy alleviates physiological human brain aging and COVID-19 neuropathology. Nat. Aging 3, 1561\u0026ndash;1575 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhu, Y. \u003cem\u003eet al.\u003c/em\u003e The Achilles\u0026rsquo; heel of senescent cells: from transcriptome to senolytic drugs. Aging Cell 14, 644\u0026ndash;658 (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAshraf, H. M., Fernandez, B. \u0026amp; Spencer, S. L. The intensities of canonical senescence biomarkers integrate the duration of cell-cycle withdrawal. Nat. Commun. 14, 4527 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGorgoulis, V. \u003cem\u003eet al.\u003c/em\u003e Cellular senescence: Defining a path forward. Cell 179, 813\u0026ndash;827 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOldham, M. C., Horvath, S. \u0026amp; Geschwind, D. H. Conservation and evolution of gene coexpression networks in human and chimpanzee brains. \u003cem\u003eProc. Natl. Acad. Sci. U. S. A.\u003c/em\u003e 103, 17973\u0026ndash;17978 (2006).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eForoushani, A. \u003cem\u003eet al.\u003c/em\u003e Large-scale gene network analysis reveals the significance of extracellular matrix pathway and homeobox genes in acute myeloid leukemia: an introduction to the Pigengene package and its applications. BMC Med. Genomics 10, 16 (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMathys, H. \u003cem\u003eet al.\u003c/em\u003e Single-cell transcriptomic analysis of Alzheimer\u0026rsquo;s disease. Nature 570, 332\u0026ndash;337 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhou, Y. \u003cem\u003eet al.\u003c/em\u003e Author Correction: Human and mouse single-nucleus transcriptomics reveal TREM2-dependent and TREM2-independent cellular responses in Alzheimer\u0026rsquo;s disease. Nat. Med. 26, 981 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXiong, X. \u003cem\u003eet al.\u003c/em\u003e Epigenomic dissection of Alzheimer\u0026rsquo;s disease pinpoints causal variants and reveals epigenome erosion. Cell 186, 4422\u0026ndash;4437.e21 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMathys, H. \u003cem\u003eet al.\u003c/em\u003e Single-cell multiregion dissection of Alzheimer\u0026rsquo;s disease. Nature 632, 858\u0026ndash;868 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eQuinlan, J. R. Induction of decision trees. Mach. Learn. 1, 81\u0026ndash;106 (1986).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOdessa Press. \u003cem\u003eMAST: A Flexible Statistical Framework for Assessing Transcriptional Changes and Characterizing Heterogeneity in Single-Cell RNA Sequencing Data\u003c/em\u003e. (CreateSpace Independent Publishing Platform, 2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang, T. \u0026amp; Nabavi, S. SigEMD: A powerful method for differential gene expression analysis in single-cell RNA sequencing data. Methods 145, 25\u0026ndash;32 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKanehisa, M. \u0026amp; Goto, S. KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res. 28, 27\u0026ndash;30 (2000).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBobela, W., Aebischer, P. \u0026amp; Schneider, B. L. Αlpha-synuclein as a mediator in the interplay between aging and Parkinson\u0026rsquo;s disease. Biomolecules 5, 2675\u0026ndash;2700 (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLudtmann, M. H. R. \u0026amp; Abramov, A. Y. Mitochondrial calcium imbalance in Parkinson\u0026rsquo;s disease. Neurosci. Lett. 663, 86\u0026ndash;90 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKume, K. \u003cem\u003eet al.\u003c/em\u003e CGG repeat expansion in LRP12 in amyotrophic lateral sclerosis. Am. J. Hum. Genet. 110, 1086\u0026ndash;1097 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTakuma, K., Yan, S. S., Stern, D. M. \u0026amp; Yamada, K. Mitochondrial dysfunction, endoplasmic reticulum stress, and apoptosis in Alzheimer\u0026rsquo;s disease. J. Pharmacol. Sci. 97, 312\u0026ndash;316 (2005).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRai, S. N., Singh, C., Singh, A., Singh, M. P. \u0026amp; Singh, B. K. Mitochondrial dysfunction: A potential therapeutic target to treat Alzheimer\u0026rsquo;s disease. Mol. Neurobiol. 57, 3075\u0026ndash;3088 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBossy-Wetzel, E., Petrilli, A. \u0026amp; Knott, A. B. Mutant huntingtin and mitochondrial dysfunction. Trends Neurosci. 31, 609\u0026ndash;616 (2008).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang, C. \u003cem\u003eet al.\u003c/em\u003e Novel role of COX6c in the regulation of oxidative phosphorylation and diseases. Cell Death Discov. 8, 336 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSpielmann, N. \u003cem\u003eet al.\u003c/em\u003e Knockout of the Complex III subunit Uqcrh causes bioenergetic impairment and cardiac contractile dysfunction. Mamm. Genome 34, 229\u0026ndash;243 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSantiago-Mujika, E., Luthi-Carter, R., Giorgini, F. \u0026amp; Mukaetova-Ladinska, E. B. Tubulin isotypes and posttranslational modifications in vascular dementia and Alzheimer\u0026rsquo;s disease. J. Alzheimers Dis. Rep. 6, 739\u0026ndash;748 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBittermann, E. \u003cem\u003eet al.\u003c/em\u003e Differential requirements of tubulin genes in mammalian forebrain development. PLoS Genet. 15, e1008243 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBustamante-Barrientos, F. A. \u003cem\u003eet al.\u003c/em\u003e Mitochondrial dysfunction in neurodegenerative disorders: Potential therapeutic application of mitochondrial transfer to central nervous system-residing cells. J. Transl. Med. 21, 613 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMiwa, S., Kashyap, S., Chini, E. \u0026amp; von Zglinicki, T. Mitochondrial dysfunction in cell senescence and aging. J. Clin. Invest. 132, (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAshleigh, T., Swerdlow, R. H. \u0026amp; Beal, M. F. The role of mitochondrial dysfunction in Alzheimer\u0026rsquo;s disease pathogenesis. Alzheimers. Dement. 19, 333\u0026ndash;342 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang, M., Zhao, Y. \u0026amp; Zhang, B. Efficient Test and Visualization of Multi-Set Intersections. Sci. Rep. 5, 16923 (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHerdy, J. R., Mertens, J. \u0026amp; Gage, F. H. Neuronal senescence may drive brain aging. Science 384, 1404\u0026ndash;1406 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAmor, S., Puentes, F., Baker, D. \u0026amp; van der Valk, P. Inflammation in neurodegenerative diseases. Immunology 129, 154\u0026ndash;169 (2010).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMerelli, A., Repetto, M., Lazarowski, A. \u0026amp; Auzmendi, J. Hypoxia, oxidative stress, and inflammation: Three faces of neurodegenerative diseases. J. Alzheimers. Dis. 82, S109\u0026ndash;S126 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWiley, C. D. \u003cem\u003eet al.\u003c/em\u003e Mitochondrial dysfunction induces senescence with a distinct secretory phenotype. Cell Metab. 23, 303\u0026ndash;314 (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVictorelli, S. \u003cem\u003eet al.\u003c/em\u003e Apoptotic stress causes mtDNA release during senescence and drives the SASP. Nature 622, 627\u0026ndash;636 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTaipa, R. \u003cem\u003eet al.\u003c/em\u003e Proinflammatory and anti-inflammatory cytokines in the CSF of patients with Alzheimer\u0026rsquo;s disease and their correlation with cognitive decline. Neurobiol. Aging 76, 125\u0026ndash;132 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlbrecht, D. S. \u003cem\u003eet al.\u003c/em\u003e Early neuroinflammation is associated with lower amyloid and tau levels in cognitively normal older adults. Brain Behav. Immun. 94, 299\u0026ndash;307 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShafqat, A. \u003cem\u003eet al.\u003c/em\u003e Cellular senescence in brain aging and cognitive decline. Front. Aging Neurosci. 15, 1281581 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLopez-Rodriguez, A. B. \u003cem\u003eet al.\u003c/em\u003e Acute systemic inflammation exacerbates neuroinflammation in Alzheimer\u0026rsquo;s disease: IL-1β drives amplified responses in primed astrocytes and neuronal network dysfunction. Alzheimers. Dement. 17, 1735\u0026ndash;1755 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRiessland, M. \u003cem\u003eet al.\u003c/em\u003e Loss of SATB1 induces p21-dependent cellular senescence in post-mitotic dopaminergic neurons. Cell Stem Cell 25, 514\u0026ndash;530.e8 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTraxler, L. \u003cem\u003eet al.\u003c/em\u003e Warburg-like metabolic transformation underlies neuronal degeneration in sporadic Alzheimer\u0026rsquo;s disease. Cell Metab. 34, 1248\u0026ndash;1263.e6 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchon, E. A. \u0026amp; Area-Gomez, E. Is Alzheimer\u0026rsquo;s disease a disorder of mitochondria-associated membranes? J. Alzheimers. Dis. 20 Suppl 2, S281\u0026ndash;92 (2010).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXu, H., Liu, Y.-Y., Li, L.-S. \u0026amp; Liu, Y.-S. Sirtuins at the Crossroads between mitochondrial quality control and neurodegenerative diseases: Structure, regulation, modifications, and modulators. Aging Dis. 14, 794\u0026ndash;824 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTang, Y., Huang, Y., Wan, Z., Zhou, B. \u0026amp; Wu, Z. Mitochondrial quality control links two seemingly unrelated neurodegenerative diseases. Autophagy 18, 2495\u0026ndash;2497 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYan, X., Wang, B., Hu, Y., Wang, S. \u0026amp; Zhang, X. Abnormal mitochondrial quality control in neurodegenerative diseases. Front. Cell. Neurosci. 14, 138 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFranco-Iborra, S., Vila, M. \u0026amp; Perier, C. Mitochondrial quality control in neurodegenerative diseases: Focus on Parkinson\u0026rsquo;s disease and Huntington's disease. Front. Neurosci. 12, 342 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e\u003cem\u003eMitochondrial Dysfunction in Alzheimer\u0026rsquo;s Disease: Role of Oxidative Phosphorylation Deficits\u003c/em\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e\u003cem\u003eMitochondrial Transplantation as a Novel Therapeutic Strategy for Neurodegenerative DiseasesMitochondrial Transplantation as a Novel Therapeutic Strategy for Neurodegenerative Diseases\u003c/em\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOgrodnik, M. \u003cem\u003eet al.\u003c/em\u003e Guidelines for minimal information on cellular senescence experimentation in vivo. Cell 187, 4150\u0026ndash;4175 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCohen, J. \u0026amp; Torres, C. Astrocyte senescence: Evidence and significance. Aging Cell 18, e12937 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchlett, J. S. \u003cem\u003eet al.\u003c/em\u003e NF-κB is a critical mediator of post-mitotic senescence in oligodendrocytes and subsequent white matter loss. Mol. Neurodegener. 18, 24 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRim, C., You, M.-J., Nahm, M. \u0026amp; Kwon, M.-S. Emerging role of senescent microglia in brain aging-related neurodegenerative diseases. Transl. Neurodegener. 13, 10 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHodes, R. J. \u0026amp; Buckholtz, N. Accelerating Medicines Partnership: Alzheimer\u0026rsquo;s Disease (AMP-AD) Knowledge Portal Aids Alzheimer's Drug Discovery through Open Data Sharing. Expert Opin. Ther. Targets 20, 389\u0026ndash;391 (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBennett, D. A. \u003cem\u003eet al.\u003c/em\u003e Religious Orders Study and Rush Memory and Aging Project. J. Alzheimers. Dis. 64, S161\u0026ndash;S189 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eR Core Team, R. \u0026amp; Others. R: A language and environment for statistical computing. (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePearson, K. \u003cem\u003eOn Lines and Planes of Closest Fit to Systems of Points in Space\u003c/em\u003e. (1901).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCragg, J. G. \u003cem\u003eSome Statistical Models for Limited Dependent Variables with Application to the Demand for Durable Goods\u003c/em\u003e. (1969).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang, T. \u0026amp; Nabavi, S. Differential gene expression analysis in single-cell RNA sequencing data. in 2017 \u003cem\u003eIEEE International Conference on Bioinformatics and Biomedicine (BIBM)\u003c/em\u003e (IEEE, 2017). doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1109/bibm.2017.8217650\u003c/span\u003e\u003cspan address=\"10.1109/bibm.2017.8217650\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRubner, Y., Tomasi, C. \u0026amp; Guibas, L. J. A metric for distributions with applications to image databases. in \u003cem\u003eSixth International Conference on Computer Vision (IEEE Cat. No.98CH36271)\u003c/em\u003e (Narosa Publishing House, 2002). doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1109/iccv.1998.710701\u003c/span\u003e\u003cspan address=\"10.1109/iccv.1998.710701\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKantorovič, L. V. \u003cem\u003eMathematical Methods of Organizing and Planning Production\u003c/em\u003e. (1960).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKolde, R. pheatmap: Pretty Heatmaps. \u003cem\u003eCRAN: Contributed Packages\u003c/em\u003e The R Foundation \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.32614/cran.package.pheatmap\u003c/span\u003e\u003cspan address=\"10.32614/cran.package.pheatmap\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2010).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKuhn, M. \u0026amp; Johnson, K. Classification Trees and Rule-Based Models. in \u003cem\u003eApplied Predictive Modeling\u003c/em\u003e (eds. Kuhn, M. \u0026amp; Johnson, K.) 369\u0026ndash;413 (Springer New York, New York, NY, 2013). doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/978-1-4614-6849-3_14\u003c/span\u003e\u003cspan address=\"10.1007/978-1-4614-6849-3_14\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYerushalmy, J. Statistical problems in assessing methods of medical diagnosis, with special reference to X-ray techniques. Public Health Rep. 62, 1432\u0026ndash;1449 (1947).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYu, G., Wang, L.-G., Han, Y. \u0026amp; He, Q.-Y. clusterProfiler: an R package for comparing biological themes among gene clusters. OMICS 16, 284\u0026ndash;287 (2012).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWu, T. \u003cem\u003eet al.\u003c/em\u003e clusterProfiler 4.0: A universal enrichment tool for interpreting omics data. Innovation (Camb) 2, 100141 (2021).\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003eTables 1 to 3 are available in the Supplementary Files section.\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"npj-aging","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [npj Aging](https://www.nature.com/npjamd/)","snPcode":"41514","submissionUrl":"https://submission.springernature.com/new-submission/41514/3","title":"npj Aging","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"NPJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-5903682/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5903682/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eNeuronal senescence (i.e., neurescent) is an important hallmark of aging and neurodegeneration, but it remains poorly characterized in the human brain due to the lack of reliable markers. This study aimed to identify neurescent markers based on single-nucleus transcriptome data from postmortem human prefrontal cortex. Using an eigengene approach, we integrated three gene panels: a) SenMayo, b) Canonical Senescence Pathway (CSP), and c) Senescence Initiating Pathway (SIP), to identify neurescent signatures. We found that paired markers outperform single markers; for instance, by combining \u003cem\u003eCDKN2D\u003c/em\u003e and \u003cem\u003eETS2\u003c/em\u003e in a decision tree, a high accuracy of 99% and perfect specificity (100%) were achieved in distinguishing neurescent. Differential expression analyses identified 324 genes that are overexpressed in neurescent. These genes showed significant associations with important neurodegeneration-related pathways including Alzheimer\u0026rsquo;s disease, Parkinson\u0026rsquo;s disease, and Huntington\u0026rsquo;s disease. Interestingly, several of these overexpressed genes are linked to mitochondrial dysfunction and cytoskeletal dysregulation. These findings provide valuable insights into the complexities of neurescent, emphasizing the need for further exploration of histologically viable markers and validation in broader datasets.\u003c/p\u003e","manuscriptTitle":"Identification of markers for neurescence through transcriptomic profiling of postmortem human brains","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-04-09 09:21:03","doi":"10.21203/rs.3.rs-5903682/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-04-23T13:59:18+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-04-21T20:42:56+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-04-21T14:32:19+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"205782180021544356877936212708144228536","date":"2025-04-07T19:59:26+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-04-07T16:08:23+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"7511998332942860291085443913982260511","date":"2025-04-07T15:42:56+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"154930140225222849267580208745859796526","date":"2025-04-07T15:17:14+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-04-07T15:05:34+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-04-05T14:12:59+00:00","index":"","fulltext":""},{"type":"submitted","content":"npj Aging","date":"2025-03-28T17:05:22+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"npj-aging","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [npj Aging](https://www.nature.com/npjamd/)","snPcode":"41514","submissionUrl":"https://submission.springernature.com/new-submission/41514/3","title":"npj Aging","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"NPJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"4ba7c395-9b83-4e71-adf9-6dacdded0d60","owner":[],"postedDate":"April 9th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":46813747,"name":"Biological sciences/Cell biology/Senescence"},{"id":46813748,"name":"Health sciences/Diseases/Neurological disorders"},{"id":46813749,"name":"Health sciences/Diseases/Neurological disorders/Dementia"},{"id":46813750,"name":"Health sciences/Diseases/Neurological disorders/Dementia/Alzheimers disease"},{"id":46813751,"name":"Health sciences/Biomarkers"}],"tags":[],"updatedAt":"2025-09-06T06:02:30+00:00","versionOfRecord":{"articleIdentity":"rs-5903682","link":"https://doi.org/10.1038/s41514-025-00235-y","journal":{"identity":"npj-aging","isVorOnly":false,"title":"npj Aging"},"publishedOn":"2025-07-01 15:57:54","publishedOnDateReadable":"July 1st, 2025"},"versionCreatedAt":"2025-04-09 09:21:03","video":"","vorDoi":"10.1038/s41514-025-00235-y","vorDoiUrl":"https://doi.org/10.1038/s41514-025-00235-y","workflowStages":[]},"version":"v1","identity":"rs-5903682","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5903682","identity":"rs-5903682","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}