Abstract

Background

   With the increase of age, multiple physiological functions of people
   begin gradually degenerating. Regardless of natural aging or
   pathological aging, the decline in cognitive function is one of the
   most obvious features in the process of brain aging. Brain aging is a
   key factor for several neuropsychiatric disorders and for most
   neurodegenerative diseases characterized by onset typically occurring
   late in life and with worsening of symptoms over time. Therefore, the
   early prevention and intervention of aging progression are particularly
   important. Since there is no unified conclusion about the plasma
   diagnostic biomarkers of brain aging, this paper innovatively employed
   the combined multi-omics analysis to delineate the plasma markers of
   brain aging.

Methods

   In order to search for specific aging markers in plasma during cerebral
   cortex aging, we used multi-omics analysis to screen out differential
   genes/proteins by integrating two prefrontal cortex (PFC)
   single-nucleus transcriptome sequencing (snRNA-seq) datasets and one
   plasma proteome sequencing datasets. Then plasma samples were collected
   from 20 young people and 20 elder people to verify the selected
   differential genes/proteins with ELISA assay.

Results

   We first integrated snRNA-seq data of the post-mortem human PFC and
   generated profiles of 65,064 nuclei from 14 subjects across adult
   (44–58 years), early-aging (69–79 years), and late-aging (85–94 years)
   stages. Seven major cell types were classified based on established
   markers, including oligodendrocyte, excitatory neurons, oligodendrocyte
   progenitor cells, astrocytes, microglia, inhibitory neurons, and
   endotheliocytes. A total of 93 cell-specific genes were identified to
   be significantly associated with age. Afterward, plasma proteomics data
   from 2,925 plasma proteins across 4,263 young adults to nonagenarians
   (18–95 years old) were combined with the outcomes from snRNA-seq data
   to obtain 12 differential genes/proteins (GPC5, CA10, DGKB, ST6GALNAC5,
   DSCAM, IL1RAPL2, TMEM132C, VCAN, APOE, PYH1R, CNTN2, SPOCK3). Finally,
   we verified the 12 differential genes by ELISA and found that the
   expression trends of five biomarkers (DSCAM, CNTN2, IL1RAPL2, CA10,
   GPC5) were correlated with brain aging.

Conclusion

   Five differentially expressed proteins (DSCAM, CNTN2, IL1RAPL2, CA10,
   GPC5) can be considered as one of the screening indicators of brain
   aging, and provide a scientific basis for clinical diagnosis and
   intervention.

   Keywords: brain aging, single-nuclear transcriptome sequencing,
   proteome, biomarker, plasma

1. Introduction

   Nowadays, human life expectancy in general has increased dramatically,
   and the phenomenon of population aging becomes one of the most
   significant public health problems worldwide. It is well known that the
   function of the brain will gradually decline during the aging process,
   mainly manifested by barriers in learning and memory, attention,
   decision-making speed, sensory perception and motor coordination
   ([33]Alexander et al., 2012). [34]Mendonca et al. (2017) suggested that
   the decline of brain function roughly parallels the time course of
   decline in other organ functions and accelerates significantly after
   age 50. The current aging trend greatly depresses the average quality
   of life of the elderly, but also increases the pressure on family and
   society for providing for the aged. Brain aging is mainly divided into
   normal aging and pathological aging, and pathological brain aging
   represents as one leading factor of increasing incidence of
   neurodegenerative diseases including Alzheimer’s disease (AD),
   Parkinson’s disease (PD), etc ([35]Isaev et al., 2019). However, there
   is currently no substantial diagnostic and therapeutic methods for the
   prevention of brain aging and the relief of symptoms.

   Single-cell/nucleus RNA sequencing (scRNA-seq/snRNA-seq) techniques
   have unraveled the transcriptional alterations underlying the
   heterogeneous process of aging at individual cell-type-specific
   resolution in multiple organs ([36]Zhang et al., 2021). Compared with
   scRNA-seq, snRNA-seq can directly extract nuclei from frozen tissues to
   perform sequencing ([37]Whytock et al., 2022). Moreover, snRNA-seq
   provides unique advantages in analyzing tissues difficult to be
   dissociated, such as the neuron-rich brain tissues, thus reducing the
   bias in cellular capture and the accompanied transcriptional artifacts
   ([38]Zhang et al., 2021). At present, many scholars are more inclined
   to use snRNA-seq for scientific research on brain tissue ([39]Preissl
   et al., 2018; [40]Zhong et al., 2020; [41]Hao et al., 2022). The plasma
   proteome is a research analysis method that focuses on the study of
   proteins in blood. The plasma proteome profile changes with age
   throughout the human life cycle, as does the aging of organs
   ([42]Lehallier et al., 2019), and has played a significant role in the
   study and exploration of neurological-related diseases ([43]He et al.,
   2021; [44]Sorek et al., 2021), cardiovascular diseases ([45]Roth et
   al., 2020; [46]Wang et al., 2021), tumors ([47]Riemondy et al., 2022),
   and ocular retinopathy ([48]Menon et al., 2019). This method provides
   new ideas for the diagnosis, prevention and subsequent research of
   various diseases by utilizing the differential expression of plasma
   proteome in different physiological states and disease states, and also
   shows some prospective and application prospects in clinical practice.

   In this study, we have constructed the transcriptome profile of PFC
   aging by integrating two datasets from two independent snRNA cohort
   study ([49]Brenner et al., 2020; [50]Lau et al., 2020), and explored
   new biomarkers about the PFC aging through comprehensive integrated
   analysis of plasma proteome and cortical single-cell transcriptome
   ([51]Liu and Trapnell, 2016). We finally identified five candidate
   proteins associated with brain aging in plasma and verified them as
   potential biomarkers of brain aging by ELISA test. This study provides
   novel diagnostic basis and research ideas for the prevention and
   treatment of a variety of neurodegenerative diseases related to aging
   in the future.

2. Materials and methods

2.1. Datasets and subjects

   Two snRNA-seq datasets ([52]GSE141552 and [53]GSE157827) were obtained
   from Gene Expression Omnibus (GEO).^[54]1 These datasets were obtained
   using the 10X Genomics platform and NovaSeq 6000 sequencing platform
   ([55]Brenner et al., 2020; [56]Lau et al., 2020). The data of plasma
   proteome were obtained from the “Aging Plasma Proteome^[57]2”
   ([58]Lehallier et al., 2019).

2.2. SnRNA-seq analysis

2.2.1. Preprocessing, quality control, and data integration

   The gene barcode matrices for each sample were loaded into the R using
   the Read 10X function in the Seurat R package ([59]Nagy et al., 2020).
   The Seurat Object, corresponding to each sample was created using the
   Create Seurat Object function with the input gene barcode matrix
   provided as the raw data. The datasets were integrated using the method
   of [60]Stuart et al. (2019). Data quality was controlled prior to
   integration. The number of genes per sample, unique molecular identity
   counts, and the percentage of mitochondrial genes were controlled. To
   exclude potential dead cells and cell debris from the dataset, we
   filtered out nuclei with ≤200 genes, ≥7500 unique molecular
   identifiers, or ≥5% mitochondrial genes. In total, 65,064 high-quality
   nuclei were obtained for subsequent analyses. For the integration
   analysis, the highly variable features of each sample were identified
   using the FindVariableFeatures function with selection.method = vst,
   nfeatures = 2000. To integrate all the samples, the
   FindIntegrationAnchors function was used to anchor the features of the
   samples with dims = 1:30. All samples were integrated using the
   IntegrateData function with the parameter dims = 1:30.

2.2.2. Data dimension reduction and clustering analysis

   Subsequently, we scaled the expression matrix and performed a linear
   dimension reduction using the RunPCA function with the parameter npcs =
   50. The P-value distribution of each major component was visualized
   using the JackStrawPlot function and selected to perform graph-based
   clustering using the first 30 principal components. We performed
   K-nearest neighbor (KNN) clustering using the FindClusters function
   with the parameter resolution = 1 and UMAP clustering using the RunUMAP
   function with the parameter dims = 1:30, which initially yielded 22
   cell clusters. Differentially expressed genes (DEGs) in each cell
   cluster were identified by the Wilcoxon rank-sum test using the
   FindAllMarkers function with the parameters logfc.threshold = 0.25 and
   test.use = wilcox. We then assigned a cell-type identity to each cell
   cluster according to the expression of known cell-type markers and
   identified additional cell type-specific marker genes by the Wilcoxon
   rank-sum test using the FindAllMarkers function with the parameters
   logfc.threshold = 0.25 and test.use = wilcox. For cell-type markers,
   the level of statistical significance was set at an adjusted P-value <
   0.05.

2.2.3. Examination of cell type-specific transcriptomic changes

   To examine cell type-specific transcriptome changes in the cerebral
   cortex of the aging group, we used the FindMarkers function and
   parameters logFc.threshold = 0.25 and test.use = Wilcox by Wilcoxon
   rank sum test. The level of statistical significance of cell
   type-specific transcriptome changes was set to adjusted P & LT; 0.1 and
   log2 multiples change ≥0.25 or ≤ −0.25.

2.2.4. Gene-set enrichment analysis (GSEA)

   Gene-set enrichment analysis was applied to identify a priori-defined
   gene sets that show statistically significant differences between two
   given clusters. The expression file was set as input, and gene sets of
   KEGG pathways and Gene Ontology (GO) were implied, which were collected
   in Molecular Signatures Database (MSigDB) ([61]Subramanian et al.,
   2005; [62]Zhong et al., 2018).

2.3. GO and KEGG signaling pathway enrichment analysis

   In this study, DEGs were all mapped to the GO terms in the Gene
   Ontology database,^[63]3 and the number of genes were calculated for
   each term. Pathway-based analysis was used to characterize the
   biological functions of the genes. Pathway enrichment analysis
   identified significant signal transduction pathways in the KEGG
   database.^[64]4 In GO and KEGG enrichment analysis, R software version
   3.8.1,^[65]5 and multiple R packages, such as clusterProfiler,
   org.Hs.eg.db, enrichplot and ggplot2, were used to generate the bars,
   bubble maps, and signaling pathway maps.

2.4. PPI network analysis

   The STRING database^[66]6 was used for DEG-associated protein
   interaction analysis and production of PPI networks. Cytoscape 3.8.0
   was used^[67]7 to construct the cell differential expression network.

2.5. Participant information, inclusion criteria and ELISA assay

   Participant in the age range of 20–90 years were recruited. All of them
   met the following inclusion criteria to be considered in this study: no
   head injury at time of death, lack of developmental disorder, no recent
   cerebral stroke, no history of other psychiatric or neurological
   disorders, and no history of intravenous or polydrug abuse. Twenty
   healthy adult volunteers aged between 20 and 27 years were selected as
   the adult group, and 20 elderly volunteers aged between 64 and 84 years
   were arranged into the elderly group. All volunteers were informed of
   the purpose of the study and signed informed consent forms. Blood
   samples collected from participants were coagulated and centrifuged.
   Afterward, plasma was collected and stored in Eppendorf tubes until
   use. The levels of CA10, DSCAM, GPC5, IL1RAPL2, CNTN2 and SPOCK3
   proteins in plasma samples were detected by ELISA kit. The operation
   and detection were performed strictly according to the kit instructions
   as previously described ([68]Yu et al., 2022). A spectrophotometer
   (Thermo Fisher Scientific, Vantaa, Finland) was applied to measure the
   absorbance [optical density (OD) value] of each sample at a wavelength
   of 450 nm for 15 min. Finally, the linear regression equation of the
   standard curve was calculated by using the concentration and OD of the
   standard product. Then, the concentration of each protein in the plasma
   was calculated.

2.6. Statistical analysis and data visualization

   snRNA-seq data were analyzed using the Seurat package in R. P-values
   were adjusted based on Bonferroni correction. For ELISA assays,
   differences between aging and adult were analyzed using Student’s
   t-test (SPSS v26.0, IBM, USA). P < 0.05 was considered statistically
   significant. We visualized the data using Cytoscape (version 3.8.0) and
   ggplot2 package in R.

3. Results

3.1. Cell-specific transcriptional profiles of the prefrontal cortex of human
brain aging

   To investigate how molecular and cellular features are altered during
   brain aging in healthy individuals, we integrated the data from two
   snRNA-seq datasets ([69]Grubman et al., 2019; [70]Brenner et al., 2020;
   [71]Lau et al., 2020) using the method of [72]Stuart et al. (2019)
   (Seurat3). Seurat3 uses canonical correlation analysis (CCA) and mutual
   nearest neighbor (MNN) for data integration, which can effectively
   remove batch effect between samples. The analysis flow is shown in
   [73]Figure 1A, and two datasets containing snRNA-seq data of prefrontal
   tissues from 14 normal healthy individuals across young adults to
   nonagenarians were analyzed. These samples were divided into three
   groups according to age, namely stage1 (adult, age 53.5 ± 6.4), stage2
   (early-aging, age 75.0 ± 4.5), and stage3 (late-aging, age 89.7 ± 3.4)
   ([74]Supplementary Table 1). After standardized integration and quality
   control, a total of 65,064 cells (stage1: 10343, stage2: 28897, stage3:
   25824) were obtained for subsequent downstream processing
   ([75]Supplementary Table 2).

FIGURE 1.

   [76]FIGURE 1
   [77]Open in a new tab

   Integration analysis of snRNA-seq data for prefrontal cortex during
   aging. (A) Integrated analysis workflow. (B,C) Unbiased recognition of
   cell-type heterogeneity in the human prefrontal cortex. (B) The UMAP
   plot shows distribution of different cluster (left) and cell types
   (right) in the PFC. (C) The distribution of different cell types in the
   stage1 (left), stage2 (center) and stage3 (right) PFC. (D) Bubble dot
   plots of the top cluster-specific marker genes. The size of the dot
   indicates expression percentage and the darkness of the color indicates
   average expression. (E) The heatmap of top 10 DEGs in each cell types.
   (F) GSEA analysis of seven cell types. (G) Total captured cell number
   for each cell type. (H) The bar plots show the proportion of the seven
   cell types found in the PFC samples during aging. Oligos,
   oligodendrocyte; Ex, excitatory; OPC, oligodendrocyte progenitor cells;
   Astros, astrocytes; Micro, microglia; Inhib, inhibitory; Endo,
   endotheliocyte.

   To construct taxonomic maps of cell populations, all data were
   subjected to principal component analysis (PCA) and uniform manifold
   approximation and projection (UMAP) clustering analysis and visualized,
   which yielded 22 unique cell clusters ([78]Figures 1B, C). Comparing
   differential genes across all cell types, the specific genes expressed
   by each cell type were revealed to identify cell phenotype ([79]Figures
   1D, E). Seven major cell types were classified based on their
   respective transcriptomic expression profiles and the previously
   reported cell type marker genes ([80]Mathys et al., 2019; [81]Lau et
   al., 2020): Astrocytes (Astros, C0, C17, C18, and C21; marked by
   AQP4/GFAP, 13.3% of the total), endotheliocytes (Endo, C15; EBF1/RGS5,
   1.2%), excitatory neuron (Ex, C3, C4, C6, C10, C11, and C14;
   CAMK2A/SATB2, 34.3%), inhibitory neuron (Inhib, C8, C9, C12, and C13;
   GAD1 and GAD2, 14.5%), microglia (Micro, C7, C19 and C20; PTPRC/CSF1R,
   6.6%), oligodendrocyte (Oligos, C1, C2 and C16; marked by
   MOG/PLP1/ST18, 22.7%), and OPC (OPC, C5; VCAN/PDGFRA, 7.6%)
   ([82]Figures 1B–G and [83]Supplementary Tables 2, [84]3). GSEA of these
   cell-type-specific marker genes demonstrated functional characteristics
   of the corresponding cell type ([85]Figure 1F). For example, “synapse
   assembly and organization” was enriched for the top marker genes
   (log2FC > 0.25) for Ex and Inhib, “myelination and glia cell
   development” for Oligos, “leukocyte activation” for microglia,
   “angiogenesis” for End, etc ([86]Figure 1F). In order to investigate
   how molecular and cellular characteristics of prefrontal cortex are
   altered during aging in healthy individuals, we compared the proportion
   of different cell types across three stages. We found that the
   proportion of Ex was sharply decreased with age, while the OPC, Oligos,
   Micro was increased ([87]Figure 1H and [88]Supplementary Table 3). The
   proportion of Astros was increased in the early-aging ([89]Figure 1H
   and [90]Supplementary Table 3). Then, To identify overall
   transcriptomic changes for each cell type across three stage, we
   compared the transcriptomic profiles of individual cell types across
   three age groups. A total of 93 differentially expressed genes (DEGs)
   were identified for all cell type during aging. Among them, genes
   upregulated with age include 5 astrocytes, 13 endotheliocytes, 5
   microglia; excitatory neurons, inhibitory neurons, oligodendrocytes,
   and OPC are not significantly upregulated with age ([91]Supplementary
   Table 4). Genes down-regulated with age include 2 astrocytes, 6
   endotheliocytes, 16 excitatory neurons, 23 inhibitory neurons, 11
   microglia, 6 oligodendrocytes, 6 oligodendrocyte precursors
   ([92]Supplementary Table 4).

3.2. Acquisition of differential genes/proteins between prefrontal cortex and
plasma associated with aging

   In order to investigate how molecular and cellular characteristics of
   prefrontal cortex are altered during aging in healthy individuals, we
   compared the transcriptomic profiles of individual cell types across
   three age groups. A total of 870 DEGs were identified for each cell
   type during aging ([93]Supplementary Tables 5–[94]18). Then, we
   compared levels of gene expression in cells across all stage by cell
   type, and identified 93 unique DEGs that is constantly changing in all
   stages ([95]Supplementary Table 4).

   In order to identify the cell-specific plasma markers of brain aging,
   we further collected and analyzed the plasma proteomics data from 2,925
   plasma proteins across 4,263 young adults to nonagenarians (18–95 years
   old). There were 1,376 differential proteins in the plasma that changed
   significantly with age ([96]Figure 2). By integrated analysis of
   snRNA-seq data and proteomic data, we found 12 differentially expressed
   genes/proteins (GPC5, CA10, DGKB, ST6GALNAC5, DSCAM, IL1RAPL2,
   TMEM132C, VCAN, APOE, PYH1R, CNTN2, SPOCK3, [97]Table 1) from
   astrocytes ([98]Figure 2A), excitatory neurons ([99]Figure 2B),
   inhibitory neurons ([100]Figure 2C), oligodendrocyte progenitors
   ([101]Figure 2D), endothelial cells ([102]Figure 2E), and microglia
   ([103]Figure 2F).

FIGURE 2.

   [104]FIGURE 2
   [105]Open in a new tab

   VENNY map of cerebral cortex and plasma proteins in naturally aging
   patients. (A) Venn diagram shows overlap of DEGs of astrocytes and DEPs
   of plasma. (B) Venn diagram shows overlap of DEGs of excitatory neurons
   and DEPs of plasma. (C) Venn diagram shows overlap of DEGs of
   inhibitory neurons and DEPs of plasma. (D) Venn diagram shows overlap
   of DEGs of OPCs and DEPs of plasma. (E) Venn diagram shows overlap of
   DEGs of endotheliocyte and DEPs of plasma. (F) Venn diagram shows
   overlap of DEGs of microglia and DEPs of plasma.

TABLE 1.

   Differential gene expression information.
   Source Gene name Protein name Gene expression
   Astrocyte GPC5 Glypican proteoglycan 5 Down
   Excitatory neuron CA10 Carbonic anhydrase 10 Down
   Excitatory neuron DGKB Diacylglycerol kinase beta Down
   Excitatory neuron ST6GALNAC5 ST6 N-acetylgalactosaminide
   alpha-2,6-sialyltransferase 5 Down
   Inhibitory neuron DSCAM DS cell adhesion molecule Down
   Inhibitory neuron IL1RAPL2 Interleukin 1 receptor accessory protein
   like 2 Down
   Inhibitory neuron TMEM132C Transmembrane protein 132C Down
   OPC VCAN Versican Down
   Endotheliocyte APOE Apolipoprotein E Up
   Endotheliocyte PTH1R Parathyroid hormone 1 receptor Up
   Microglia CNTN2 Contactin 2 Up
   Microglia SPOCK3 SPARC (osteonectin), Cwcv and kazal like domains
   proteoglycan 3 Up
   [106]Open in a new tab

3.3. GO, KEGG, and PPI analysis of differential genes

   In order to elucidate the biological functions of these 12 molecules,
   GO functional enrichment analysis and KEGG signaling pathway analysis
   were performed ([107]Figure 3). The 12 genes/proteins were mainly
   enriched in these biological processes (BP): synapse organization and
   aminoglycan metabolic process ([108]Figure 3A), molecular function
   (MF): glycosaminoglycan binding and cell-cell adhesion ([109]Figure
   3B), cell component (CC): anchored component of Plasma membrane,
   Glutamatergic synapse, Lysosomal Lumen ([110]Figure 3C). KEGG analysis
   showed that these crucial genes were mainly enriched in cell adhesion
   molecules ([111]Figure 3D). In order to identify whether there was a
   relationship among these 12 molecules, the protein-protein interaction
   analysis was performed and the PPI network was obtained ([112]Figure
   3E). The PPI network showed that there was extensive direct and
   indirect communication, physical and genetic interactions among these
   12 molecules.

FIGURE 3.

   [113]FIGURE 3
   [114]Open in a new tab

   The functional enrichment analysis. (A–C) GO functional enrichment
   analysis. (A) The top 20 GO enrichments in BP. (B) The top 20 GO
   enrichments in CC. (C) The top 20 GO enrichments in MF. Each node
   signaled a GO term, and its size represented the gene number. The color
   indicates the p-value. (D) KEGG enrichment analysis and pathway
   mapping. (E) Differential Gene PPI Network Analysis diagram. Each
   circle in the diagram represents a differentially expressed gene. The
   larger the circle, the higher the expression of DEG. The line
   connecting the DEGs represents the correlation between them. The
   thicker the connection line, the stronger the correlation between DEGs.

4. Verification of the brain aging markers

   To verify the results obtained above, we further collected plasma from
   20 young individuals and 20 elderly individuals ([115]Supplementary
   Table 19). By consulting relevant literature, 6 genes related to
   neurodevelopmental and neurological diseases (GPC5, CA10, DSCAM,
   IL1RAPL2, CNTN2, and SPOCK3) were verified by ELISA method
   ([116]Supplementary Table 20). The ELISA results demonstrated that
   compared with the adult group, GPC5, CA10, DSCAM, IL1RAPL2, and SPOCK3
   were down-regulated in the aging group ([117]Figures 4A–D, F). CNTN2
   showed an up-regulation trend with significant differences ([118]Figure
   4E). The expression trends of GPC5, CA10, DSCAM, IL1RAPL2, and CNTN2
   were consistent with snRNA-Seq and proteomic outcomes. Moreover,
   correlation analysis of these genes with age revealed that there are
   significantly age-related change trends for these molecules
   ([119]Figure 4G).

FIGURE 4.

   [120]FIGURE 4
   [121]Open in a new tab

   ELISA test verified the expression of brain aging markers. Plasma
   expression level of CA10 (A), DSCAM (B), GPC5 (C), IL1RAPL2 (D), CNTN2
   (E), SPOCK3 (F) in young and elderly groups. (G) Correlation analysis
   of the six molecules with age. *p < 0.05, **p < 0.01, ***p < 0.001.

5. Discussion

   It is well known that the degeneration process of various physiological
   functions caused by aging is particularly marked by the decline in
   cognitive ability of the brain. Aging is a major risk factor for human
   neurodegenerative disorders, such as Alzheimer’s disease (AD), which is
   rapidly rising in prevalence with age and has become the most pressing
   challenge to be tackled ([122]Zhang et al., 2021). The brain aging
   process is mainly reflected in the shrinkage of the whole brain, the
   reduction of gray and white matter, and the enlargement of the
   ventricle ([123]Drayer, 1988). Magnetic resonance imaging (MRI) studies
   have shown that age-related gray matter loss is most prominent in the
   temporal and frontal lobes ([124]Jack et al., 1997). Histological
   analysis has shown that brain atrophy is the result of a combination of
   physiological degeneration such as dendritic degeneration and neuronal
   death ([125]Dumitriu et al., 2010). Thus, the degree of brain atrophy
   during aging can broadly predict the onset of cognitive impairment and
   the degree of cognitive decline ([126]Jack et al., 2005). It has also
   been suggested that although there are individual differences in the
   rate of brain atrophy during aging, brain imaging data can still be
   used to initially determine the “biological age” of a person’s brain
   ([127]Cole and Franke, 2017), but this method has limitations for
   primary clinical screening. The resolution and accuracy of MRI in the
   prediction of brain aging remains modest, failing to predict the
   changes of brain structure or function more accurately on early onset
   Previous exploratory analysis of the brain at the cellular and
   molecular levels has revealed many distinct features of brain aging,
   including mitochondrial dysfunction, intracellular accumulation of
   oxidatively damaged proteins, nucleic acids and lipids, dysregulated
   energy metabolism, impaired cellular “waste disposal” mechanisms
   autophagy lysosome and proteasome function), impaired adaptive stress
   response signaling, impaired DNA repair, abnormal neural network
   activity, dysregulated neuronal Ca2+ processing, stem cell depletion
   and neuroinflammation, etc ([128]Mattson, 2000; [129]Shimura et al.,
   2001; [130]López-Otín et al., 2013; [131]Chow and Herrup, 2015;
   [132]Cohen et al., 2015). However, it is still difficult to establish
   and widely promote the above-mentioned aging indicators and methods in
   clinical practice. So it is especially important to find and establish
   a new authoritative screening index and method related to brain aging.
   Recent studies utilized single-nucleus transcriptome analysis to
   investigate the transcriptomic changes in aging brains and have
   revealed molecular alterations at the single-cell level using readily
   available frozen brain tissues ([133]Lau et al., 2020). However,
   snRNA-seq has not been applied to explore the cellular and molecular
   alterations of brain aging in the human. At present study, we first
   performed snRNA-seq analyses of the post-mortem human PFC and generated
   profiles of 65,064 nuclei from 14 subjects across adult (44–58 years),
   early-aging (69–79 years), and late-aging (85–94 years) stages. A total
   of 93 cell-specific genes were identified to be significantly
   associated with age. Combined with proteomics data, we screened 12
   differential genes/proteins (GPC5, CA10, DGKB, ST6GALNAC5, DSCAM,
   IL1RAPL2, TMEM132C, VCAN, APOE, PTH1R, CNTN2, SPOCK3) associated with
   aging. Among these genes, we found decreased expression of GPC5 derived
   from astrocytes, VCAN derived from OPC, DSCAM, IL1RAPL2, TMEM132C
   derived from inhibitory neurons and CA10, DGKB, ST6GALNAC5 derived from
   excitatory neurons. Meanwhile, the expression of APOE and PTH1R derived
   from endothelial cells, and CNTN2 and SPOCK3 derived from microglia
   were increased. Finally, six genes/proteins related to
   neurodevelopmental and neurological diseases (GPC5, CA10, DSCAM,
   IL1RAPL2, CNTN2, and SPOCK3) were verified by ELISA.

   In the Elisa sample collection process, we selected samples in the age
   range of 18–95 years. The age span of the young group increased
   relative to the age of the sample in the snRNA-seq analysis, and we
   used this to more broadly verify the differences and trends of the
   predicted proteins between the young and old groups. Referring to the
   sample size verified by Elisa experimental data in previous articles of
   this kind, we selected 40 subjects of different ages for sampling and
   subsequent Elisa verification and analysis.

   Glypican proteoglycan 5 (GPC5) is a class of heparan sulfate
   proteoglycans bound to the outer surface of the plasma membrane, and
   all members of its family are expressed during the development of the
   nervous system. This protein plays an important role in the control of
   cell division and growth regulation involved in brain patterning,
   synapse formation, axon regeneration and guidance and has also been
   found in a dense network of active MS plaques (cerebral atherosclerotic
   plaques) and members of this family may be involved in the isolation of
   pro-inflammatory chemokines ([134]Kenny et al., 2017). [135]Luxardi et
   al. (2007) found that GPC5 plays an important regulatory role in cell
   signaling during embryogenesis, while Gpc5 is expressed in the ventral
   brain when neurogenesis begins. The presence of this protein in human
   and mouse brains involves in brain development and neural function,
   including neuron growth and repair ([136]Veugelers et al., 1997;
   [137]Taniuchi et al., 2002a). In adults, GPC5 is mainly expressed in
   neurons of brain tissue ([138]Kenny et al., 2017) and it interacts with
   growth factors, chemokines and extracellular matrix proteins. At
   present study, we found that the GPC5 expression of astrocytes was
   down-regulated with age. In combination with the ELISA experiment, GPC5
   showed a down-regulation trend in the elderly group, corresponding to
   the snRNA-seq and plasma results. Recent work has shown that in
   addition to being supportive cells, astrocytes are crucial in
   controlling vascular blood flow, axon guidance, synaptic formation,
   regulating synaptic transmission and neural circuits, which are
   essential for the development and function of the central nervous
   system ([139]Zhang et al., 2016; [140]Chai et al., 2017;
   [141]Verkhratsky and Nedergaard, 2018; [142]Huang et al., 2020).
   Although astrocytes promote neuron survival and synaptic formation,
   their ability to promote synaptic formation tends to decline with age
   ([143]Zhang et al., 2016). Moreover, GPC5 can stimulate the Shh
   signaling pathway to promote neuron growth and synaptic formation
   ([144]Li et al., 2011). Therefore, the down-regulation of GPC5 in
   astrocytes may indicate the downregulation of synaptic formation
   function in the brain, thus causing cognitive impairment.

   Carbonic anhydrases (CA) are present in many organisms where they are
   involved in several important biological processes. [145]Romeo et al.
   (2009) showed that this protein and its family members are mainly
   expressed on neurons, especially on neuronal axons. A research report
   related to glypican genes in neural progenitor cells and differentiated
   neurons also pointed out that the expression of CA was found in some
   mouse and human tissues, and immunohistochemical results showed that
   their expression was mainly concentrated in brain tissue compared to
   other organ ([146]Taniuchi et al., 2002a,[147]b). Some studies have
   also shown that CA10 mRNA expression levels are high in the cerebellum,
   frontal cortex and parietal cortex, low in the midbrain, and extremely
   low in the eye ([148]Aspatwar et al., 2010). Analysis of genome-wide
   genotyping BeadChip assays in clinically relevant patients revealed
   that when CA10 and its related genes involved in brain development and
   neurological processes are absent, patients exhibit mild intellectual
   disability, growth retardation, poor weight gain, microcephaly, long
   face, large beaked nose, thick lower lip, micrognathia and other
   dysmorphic features ([149]Preiksaitiene et al., 2012). Consistently, we
   found that the CA10 expression levels in excitatory neuron were
   down-regulated with age and is reflected in plasma levels. These
   results suggested that the decreased of CA10 in aging brain may lead to
   a series of cognitive decline problems.

   DSCAM is a member of the cell adhesion molecule (Ig-CAM) immunoglobulin
   superfamily and is involved in the development of the human central and
   peripheral nervous system. DSCAM is expressed in the nervous system of
   invertebrates and vertebrates, as well as a candidate gene for Down
   syndrome and congenital heart disease (DSCHD) ([150]Barlow et al.,
   2001; [151]Boulanger, 2009). DSCAM is involved in cell migration,
   synaptic targeting, axon guidance, dendrites and cell tiling, axon
   fascicles and branching, programmed cell death and synaptogenesis in
   the nervous system, all of which can influence the establishment of
   motor circuits during development and the effects are maintained into
   adulthood ([152]Garrett et al., 2012). There is growing evidence that
   DSCAM expressed in vertebrates also plays an important role in axonal
   growth, fasciculation and branching, dendritic arborization, mosaic
   spacing of cells, and synaptogenesis ([153]Garrett et al., 2012;
   [154]Montesinos, 2014, [155]2017; [156]Mitsogiannis et al., 2020), and
   that this protein promotes pyramidal neuron morphogenesis by regulating
   dendritic arborization and spine formation during cortical circuit
   development ([157]Agarwala et al., 2001; [158]Maynard and Stein, 2012).
   So far, we experimentally showed that DSCAM was significantly
   down-regulated in the elderly group. Combining previous research and
   analysis by other scholars, when it was decreased in adult brain, the
   axonal and dendritic developmental processes and intercellular
   interactions of various neuronal cells are also adversely affected,
   leading to a series of neurodegenerative diseases and behaviors
   ([159]Li et al., 2023). Accordingly, the down-regulation of this
   protein may also predict brain aging and cognitive disorder.

   IL1RAPL2 is a member of the interleukin 1 receptor family and located
   on the X chromosome in a region associated with X-linked non-syndromic
   cognitive impairment. [160]Ferrante et al. (2001) showed that IL1RAPL2
   is specifically expressed in the nervous system from embryonic day 12.5
   through in situ expression studies on RNA of mouse embryonic tissue
   sections at different developmental stages. IL1RAPL2 has a restricted
   and specific expression in the brain during development, which may
   point to a precise role of this gene in the development or function of
   the nervous system ([161]Ferrante et al., 2001). Because of its
   specific expression pattern, it is restricted to structures related to
   cognitive function, such as the PFC and hippocampus, which is also
   consistent with our experimental results that the gene is
   down-regulated in the old group. Therefore, it also suggests that
   IL1RAPL2 could be used as a candidate protein for brain aging to
   predict brain aging.

   Current research on specific protein markers for neurodegenerative
   diseases has evaluated 30 highly specific proteins associated with the
   brain as candidate biomarkers for the diagnosis of AD, Among these
   proteins, CNTN2 may also be indicators of disease progression, showing
   weak to moderate correlation with cognitive tests, which contain the
   CNTN2 protein of our interest ([162]Begcevic et al., 2018). Some
   scholars believe that CNTN2 is expressed in neurons and
   oligodendrocytes and is highly regulated in developing neurons,
   especially in axonal growth and guidance and neuronal migration
   ([163]Huang et al., 2014). On the other hand, CNTN2 expressed by
   oligodendrocytes interferes with the process of myelination, and its
   ablation leads to hypomyelination. [164]Wang et al. (2018) and [165]Zou
   et al. (2020) demonstrated through a series of experiments that CNTN2
   overexpression or knockdown could reverse the effects of miR-34a
   upregulation or downregulation on proliferation and migration of
   Schwann cells, respectively. Combined with our experimental results,
   CNTN2 showed an up-regulation trend in the elderly group, then we can
   speculate that this protein can also be one of the candidate blood
   markers of brain aging. However, the age-related CNTN2 we found was
   mainly derived from microglia, so the dysfunction caused by the rise of
   CNTN2 in microglia require further investigation. SPOCK3 is a heparan
   sulfate proteoglycan, also known as testosterone-3, expressed in the
   nervous system. Some scholars have also stated that SPOCK1 is a member
   of the Spock family, and Spock1 inhibits the attachment to the
   substrate and the neurite growth of neuronal cells, so SPOCK3 is also
   considered to be related to neuron development ([166]Yamamoto et al.,
   2014). At present study, SPOCK3 protein showed an increasing trend from
   the snRNA-seq and plasma proteomics, but the experimental results were
   contrary to it. This situation may be related to the multiple and
   complex regulatory roles of spock3 protein.

   In conclusion, this article presents a validation and analysis of
   biomarkers of brain aging in blood. The biomarkers of brain aging in
   blood were screened by the analysis of human cerebral cortex scRNA-seq
   data and plasma proteome. After validation of the predicted and
   screened candidate proteins by Elisa method using peripheral blood
   plasma from young and old groups, GPC5, CA10, DSCAM, IL1RAPL2, CNTN2
   have certain correlation with age. These proteins can be combined with
   imaging detection methods to provide certain reference value for the
   prevention and diagnosis of brain aging, and also provide new ideas and
   directions for future research on brain aging related topics.

Data availability statement

   The snRNA-seq data after integration and code reported in this
   manuscript have been deposited in the figshare
   ([167]https://figshare.com/), under [168]doi:
   10.6084/m9.figshare.22634395.

Ethics statement

   The studies involving human participants were reviewed and approved by
   the Medical Ethics Committee of Kunming Medical University (approval
   number: KMMU2022MEC092). The patients/participants provided their
   written informed consent to participate in this study. Written informed
   consent was obtained from the individual(s) for the publication of any
   potentially identifiable images or data included in this article.

Author contributions

   R-ZN and W-QF analyzed the data and drafted the manuscript. L-LS, Q-SY,
   and Q-MQ discussed and revised the manuscript. R-ZN and JL designed the
   research and collected the information. All authors read and approved
   the final manuscript.

Acknowledgments