Region-specific and age-related differences in astrocytes in the human brain

Astrocyte heterogeneity and its relation to aging in the normal human brain remain poorly understood. We here analyzed astrocytes in gray and white matter brain tissues obtained from donors ranging in age between the neonatal period to over 100 years. We show that astrocytes are differently distributed with higher density in the white matter. This regional difference in cellular density becomes less prominent with age. Additionally, we confirm the presence of morphologically distinct astrocytes, with gray matter astrocytes being morphologically more complex. Notably, gray matter astrocytes morphologically change with age, while white matter astrocytes remain relatively consistent in morphology. Using regional mass spectrometry-based proteomics, we did, however, identify astrocyte specific proteins with regional differences in abundance, reflecting variation in cellular density or expression level. Importantly, the expression of some astrocyte specific proteins region-dependently decreases with age. Taken together, we provide insights into region-and age-related differences in astrocytes in the human brain.

Astrocytes are plastic cells that dynamically respond to changes in their local environment throughout life.Recent findings indicate that these cells develop a unique gene expression profile during aging.Astrocytes undergo changes in their homeostatic functions and display a more pronounced inflammatory phenotype with increasing age in both rodents and humans (Boisvert et al., 2018;Clarke et al., 2018;Matias et al., 2019;Orre et al., 2014;Soreq et al., 2017;Verkerke et al., 2021).Importantly, these age-dependent changes were highly dependent on location in the brain, suggesting regional heterogeneous responses of astrocyte to aging processes (Boisvert et al., 2018;Clarke et al., 2018;Matias et al., 2019;Orre et al., 2014;Soreq et al., 2017;Verkerke et al., 2021).Many studies on astrocyte aging have so far focused on mapping their transcriptional profiles within or between brain regions (Boisvert et al., 2018;Clarke et al., 2018;Soreq et al., 2017).Our insight into these region-specific and age-related differences at the protein level remains, however, poor.
Recently, Kovacs and coauthors identified the aging-related tau astrogliopathy (ARTAG) (Kovacs et al., 2016).The two major cytomorphologies of ARTAG are thorn-shaped astrocytes (TSA) and astrocytes with granular or fuzzy tau immunoreactivity (GFA) characterized by pathological accumulation of abnormally phosphorylated tau protein.TSA are localized mainly in subpial and subependymal regions, with a peculiar scenario of phospho-tau positive end-feet of astrocytic processes surrounding small vessels (Kovacs et al., 2017).Predilection site of ARTAG is the medial temporal lobe, while the cerebellum is not involved, both of healthy older individuals or as co-pathology in several neurodegenerative conditions including Alzheimer disease and chronic traumatic encephalopathy.
In the present study, we aimed at extending our current understanding about the heterogeneous nature of astrocytes in the aging human brain.We compared the density and morphology of astrocytes in brain tissue obtained from a cohort of donors ranging in age between the neonatal period to over 100 years.The frontal lobe and cerebellum were selected for our studies.Within the frontal lobe, we analyzed the gray and white matter areas, whereas in the cerebellum we focused on the white matter only.Next, we extended our analysis and set out to identify regional astrocyte specific markers.To this end, we performed a quantitative proteomics analysis obtained from 4 donors aged 24-35 years.These donors were selected to avoid changes related to development or neurodegeneration.Proteins with differential expression across brain regions were then assigned to specific brain cell types to reveal regional differences in astrocyte specific protein expression.Finally, results were further analyzed by immunohistochemistry and the impact of aging on these differentially expressed proteins was examined.

Human brain tissues
Postmortem brain tissues from 23 donors without confounding structural and neuropathological abnormalities were obtained from the Amsterdam UMC, the Netherlands Brain Bank, and the Fondazione IRCCS Istituto Neurologico Carlo Besta.To enlarge the case series, we used archive formalin-fixed paraffin-embedded (FFPE) material.The Netherlands Brain Bank provided us with frozen tissue of 4 adult donors.Participants were divided into 3 groups based on their age at death: (1) infants and donors up to 11 years of age, (2) donors between 21 and 79 years of age, and (3) donors of 100+ years of age.Demographics, including causes of death are listed in Table 1.The postmortem time was between 6 and 15 h; the donors of 100+ years of age underwent a postmortem examination within 24 h, although the exacted time was not recorded.In this study, we investigated the gray matter of the middle frontal gyrus, the deep frontal hemispheric white matter, and the cerebellar white matter.Informed consent was obtained in all cases.The study was approved by the institutional Medical Ethical Committee of the Amsterdam UMC location VUmc and conducted according to the declaration of Helsinki.

Sample preparation for laser capture microdissection
Twenty-µm frozen brain tissue sections mounted on polyethylene naphthalate-coated glass slides (Leica Microsystems) were fixed in 100% ethanol for 20 min, air-dried, and rehydrated twice in sterile H 2 O for 1 min.Tissue sections were stained with toluidine blue (1% w/v in sterile H 2 O) for 1 min, washed 3 times in sterile H 2 O for 1 min, and dehydrated in 100% ethanol twice for 3 min.Gray and white matter areas of interest were then microdissected using a Leica LMD6500 system (Leica Microsystems).Microdissected tissue (100 mm 3 per sample) was collected into adhesive caps (Zeiss).

Mass spectrometry analysis
Data-dependent acquisition mass spectrometry (MS) analysis was performed on an Orbitrap Q-Exactive HF-X mass spectrometer (Thermo Fisher Scientific) coupled to an UltiMate 3000 RSLCnano System (Thermo Fisher Scientific).About 1.5 µg of peptides per sample (dissolved in 2% formic acid) were loaded on a reversed phase C18 trap column (300 µm i.d.x 5 mm, 5 µm particle size, Thermo Fisher Scientific) and passed through an in house-made analytical column.Peptides were separated with solvent A (0.1% formic acid) and B (80% acetonitrile, 0.1% formic acid) at a flow rate of 300 nL/min as follow: 92% solvent A at 2.5 min; 8% solvent B at 157.5 min; 35% solvent B at 161 min; 100% solvent B at 165 min; 92% solvent A at 175 min.The full MS1 scan (m/z range = 315-1500, resolution = 60,000) was performed at a target of 3×10 6 ions.The top 15 precursors were selected for MS/MS acquisition with higher-collision dissociation (target ions = 1×10 5 , max ion fill time = 50 ms, isolation window = 1.4 m/z, normalized collision energy = 27%, resolution = 30,000).Dynamic exclusion was set at 16 s and precursor ions with unassigned charge state and charge state of 1, 6, or higher were not fragmented.

MS data analysis
MS/MS RAW files were processed using MaxQuant (Tyanova et al., 2016) (version 1.6.6.0) with the integrated Andromeda search engine.Spectra were searched against the Swiss-Prot human reference proteome database (version August 2019, 20,431 entries).Peptide N-terminal acetylation and methionine oxidation were set as variable modifications, while carbamidomethylation of cysteine was set as fixed modification.For peptide and protein identification, "match between runs" was enabled, the minimum peptide length was set at 7 amino acids, and maximum of 25 amino acids, and the false discovery rate (FDR) threshold at 1%. Peptide mass tolerance was set at 20 ppm and fragmentation mass tolerance at 0.5 Da.A maximum of two missed trypsin cleavages was allowed.Label-free quantification (LFQ) was performed with the MaxLFQ algorithm with a minimal ratio count of 2. Statistical analysis was performed with R software (version 4.2.1).Proteins that were identified as contaminants, only by site modification or in decoy reverse database, were excluded from analysis.LFQ intensities were Log 2 -transformed.Missing data were imputed based on random draws from a Gaussian distribution with the minimal mean value being the lowest observed expression value and the standard deviation being the mean standard deviation of all detected proteins.Principal component analysis (PCA) was performed using the singular value decomposition method from FactoMineR R package (Lê et al., 2008).ComplexHeatmap R package was used for hierarchical clustering analysis (Gu et al., 2016).For this, LFQ intensities were z-scored and clustered using the Euclidean and average methods as distance and clustering measures, respectively.For differential protein expression analysis, DEP R package was used (Zhang et al., 2018).Multiple comparison was adjusted by Benjamini-Hochberg's (BH) test and statistical significance was defined as p < 0.05.EWCE R package was employed to assign cell type specific protein expression (Skene and Grant, 2016).For the analysis, single-nucleus droplet-based sequencing (snDrop-seq) datasets of the frontal cortex and cerebellar hemisphere of 6 controls were used (GSE97930) (Lake et al., 2018).The frontal cortex and cerebellum datasets consisted of 10319 and 5602 nuclei, respectively (Lake et al., 2018).A cell type specificity matrix was compiled using the EWCE R package and matched against our protein expression dataset.Only proteins with a specificity value for a specific cell type greater than 0.5 were considered highly enriched in that cell type.

Immunohistochemistry and image analysis
Formalin-fixed, paraffin-embedded tissue (5-µm-thick) was deparaffinized and incubated in 0.3% H 2 O 2 in H 2 O for 30 min.Heat-induced antigen retrieval was performed in 10 mM citrate buffer (pH 6.0) or Tris/EDTA buffer (pH 9.0).Tissue sections were thereafter stained according to standard protocols using the antibodies listed in Table 2. Immunopositivity was visualized using chromogen-bound Alexa Fluor® labeled secondary antibodies or 3,3'-diaminobenzidine (DAB) chromogen.Sections were counterstained with hematoxylin or Fluoromount-G® (ITK diagnositics).Images were acquired using a Leica DM4000B or light microscope or Leica DM5000B fluorescence microscope (Leica Microsystems).In the frontal cortex, images were taken from layers 4 through 6; in the cerebellum, from the dorsal peridentate area.To determine regional astrocytic density, GFAP immunopositive cell bodies were manually counted in the regions of interest using the cell counter plugin.Total number of GFAP + cells were counted in at least 5 standardized fields per subject using a 200× objective.Morphology of GFAP + astrocytes was assessed using the ImageJ Sholl analysis plug-in.Briefly, a minimum of 10 GFAP + cells of which the nucleus was entirely visible were manually traced in the regions of interest.These traces were then used for the Sholl analysis in which concentric rings with increasing sizes were placed around the cell starting from the cell soma.The number of intersections of cellular processes with concentric rings were measured to determine total number of intersections.Analyzed cells were averaged per subject.Expression in astrocytes of selected proteins was confirmed by double labeling with GFAP.As for the purpose of cell counting analysis, at least 5 images were captured at a 200× magnification, allowing to confirm that all GFAP + astrocytes also express the proteins of interest.Statistical analyses were performed using GraphPad Prism version 9.3.1.(GraphPad Software).Data distribution was assessed for normality using the Shapiro-Wilk test.Statistical significance was determined by a one-way ANOVA or Kruskal-Wallis test if not normally distributed.Confidence level was set at 95% (p < 0.05).

Regional and age-related differences in astrocyte density, morphology, and phospho-tau immunoreactivity
To gain insight into region-specific and age-related differences in astrocyte heterogeneity, we first assessed the density of these cells in the frontal lobe (gray and white matter) and the cerebellum (white matter) across ages.For this, FFPE tissues sections were stained with glial fibrillary acid protein (GFAP) and immunopositive cells were counted (Fig. 1a-i).We observed differences in number of GFAP + astrocytes between brain gray and white matter areas, especially the cerebellar white matter, in infants and children up to 11 years of age and donors between 21 and 79 years (Fig. 1j).Notably, number of GFAP + astrocytes decreased with age in the white matter, but not in the gray matter (Fig. 1k).
We then sought to determine regional and age-related differences in astrocyte morphology (Fig. 2a).Astrocytes typically differ in morphology depending on their location in the gray and white matter (Garcia-Marin et al., 2007;Kohler et al., 2021;Miller and Raff, 1984;Sofroniew and Vinters, 2010).In the gray matter, astrocytes show a more complex morphology with extensively branched and bushy processes.By contrast, white matter astrocytes are more simple in shape with smaller and elongated cell bodies and long fiber-like processes (Sofroniew and Vinters, 2010).Morphometric analysis confirmed a clear difference between gray and white matter astrocytes, with astrocytes in the frontal gray matter being more complex in morphology (Fig. 2b).We also observed differences in morphology across white matter regions, with cerebellar white matter astrocytes being morphologically less complex, but only in infants and children up to 11 years of age and donors between 21 and 79 years (Fig. 2b).Notably, astrocytes in the frontal gray matter became more complex in morphology with increasing age, while the morphology of white matter astrocytes remained relatively consistent (Fig. 2c).We finally investigated whether the immunostaining for hyperphosphorylated-tau protein in astrocytes varied with age in our series of neurologically intact donors.As expected, no astrocytic phospho-tau expression was observed in infants and children up to 11 years of age (Fig. 3a).Astrocytes were also immunonegative for phospho-tau in the 21-79 years old donors (Fig. 3b).Remarkably, most of the 100+ years old donors were immunonegative for phospho-tau, except for one (Fig. 3c-d).In the frontal white matter, only a couple of GFAP + astrocytes co-localized with phospho-AT8.Focal areas of ARTAG were identified in the frontal cortex appearing as cluster of sub-pial TSA associated with rare granular astrocytes with plump cytoplasmic phospho-AT8 accumulation (Fig. 3d).In this subject, more severe ARTAG changes under the form of perivascular TSA were present in the sub-ependymal temporal areas (Fig. 3d).

Regional proteome analysis to identify astrocyte specific proteins
To extend our analysis on astrocyte heterogeneity in the brain, we aimed at identifying regional astrocyte specific markers.We first conducted a quantitative proteome analysis on multiple brain regions.The frontal lobe (gray and white matter) and cerebellum (white matter) from 4 donors aged 24-35 years were obtained using laser capture microdissection for mass spectrometry analysis (Table 1 and Fig. 4a).These donors were selected to avoid changes related to development or neurodegeneration.We identified a total of 2577 proteins across brain regions (Supplemental Table 1).PCA revealed that gray matter samples from all donors clearly clustered separately from white matter samples (Fig. 4b).Additionally, samples from different white matter regions formed 2 distinct clusters, where samples from biological replicates region-dependently aggregated together (Fig. 4b).To visualize regional protein expression signatures, a hierarchical clustering analysis of 1408 proteins with statistical differential expression across the different brain regions (p < 0.05) was performed (Supplemental Table 1).Notably, this again showed a clear separation of gray and white matter samples as well as moderate distinction between samples from different white matter regions (Fig. 4c).Indeed, the heatmap highlighted clusters of proteins highly expressed in either the gray or white matter and also revealed proteins with differential expression across each of the regions (Fig. 4c-d).
Astrocytic markers were then identified by integrating our regional proteome data with 2 snDrop-seq datasets from 6 human adult frontal cortex and cerebellum (GSE97930) (Lake et al., 2018).The snDrop-seq datasets consist of a total of 15,921 nuclei, of which 10,319 from the frontal cortex and 5602 from the cerebellum (Lake et al., 2018).These were resolved into neuronal and non-neuronal cell types (Lake et al., 2018).Genes from the snDrop-seq data were classified as cell type specific based on specificity values (greater than 0.5 in a certain cell type, less than 0.5 in all other cell types).These were then cross-referenced with our regional proteome dataset to identify astrocyte specific proteins.Amongst the 1408 proteins with statistical differential expression across the 3 brain regions, a total of 19 proteins were considered highly specific for astrocytes (Supplementary Table 2 and Fig. 5).For example, we identified 14 astrocyte specific proteins showing differential abundance between gray and white matter regions (Fig. 5, pink cluster).Five were found differentially expressed in the frontal lobe compared to cerebellum (Fig. 5, green cluster).Interestingly, no astrocyte specific proteins were found with differential abundance across white matter regions only.

Region-and age-related differences in localization and distribution of selected astrocytic proteins
Differential expression of cell type specific proteins could reflect regional as well as age-related differences in cell numbers.Based on antibody availability and regional differences in protein abundance, we selected 3 proteins for further immunohistochemical analysis, that is, glutamine synthetase (GLUL), aquaporin-1 (AQP1), and acyl-CoAbinding domain-containing protein 7 (ACBD7) (Figs. 5 and 6).Expression in astrocytes was confirmed by double labeling with GFAP (Fig. 6).No immunopositivity for GLUL, AQP1 or ACBD7 was noticed in GFAPnegative cells.
Results were analyzed by counting the number of immunopositive cells per mm 2 .Protein expression of GLUL was more abundant in the gray than in the white matter (Fig. 6a).Labeling for GLUL showed localization in astrocytic cell body and processes in the gray matter, whereas in the white matter it was limited to the cell bodies (Fig. 7a-i).We observed different results depending on age (Fig. 7a-i).In infants and children up to 11 years of age, GLUL expression was significantly higher in the white compared to the gray matter (Fig. 7j).In 21-79 years old donors, we observed the same trend, however with further higher expression in the frontal compared to the cerebellar white matter (Fig. 7j).Donors of 100+ years of age showed no significant differences amongst brain regions (Fig. 7j).Analysis by brain region showed a decreasing GLUL expression with increasing age in the cerebellar white matter only (Fig. 7k).
Proteome analysis revealed that AQP1 expression was higher in the white matter (Fig. 6e).Labeling for AQP1 analyzed by age demonstrated higher density of immunopositive cells in the white matter in all age groups (Fig. 8a-j).Notably, results were not significant in donors including infants and children up to 11 years of age, probably due to a wide variation between them (Fig. 8j).Analysis by region showed progressive decrease of AQP1 expression with increasing age, however only in the frontal white matter (Fig. 8k).
Protein expression of ACBD7 showed higher abundance in the cerebellar white matter (Fig. 6f).Labeling for ACBD7 revealed increased expression in the white compared to the gray matter, however in donors with age between 21 and 79 and 100+ years only.No difference between white matter regions was detected (Fig. 9a-j).Analysis by region showed no significant differences with increasing age (Fig. 9k).Taken together, findings indicate that differential expression of some astrocyte specific proteins coincides with cell numbers, whereas others did not.

Discussion
Astrocytes are heterogeneous glial cells in terms of structure and function (Kohler et al., 2021;Miller, 2018;Pestana et al., 2020;Vasile et al., 2017;Westergard and Rothstein, 2020).These cells display region specificity within the brain and remain plastic throughout life (Boisvert et al., 2018;Clarke et al., 2018;Matias et al., 2019;Soreq et al., 2017;Verkerke et al., 2021).Surprisingly, the heterogeneous nature of astrocytes and its relation to aging in the normal human brain remain poorly understood.Gaining more insight into this is important to better understand the role of astrocytes in human health and disease.For this reason, we focused our study on the human brain.
In this study, we assessed differences in astrocytic density, morphology and protein expression in gray and white matter brain tissues obtained from a cohort of donors ranging in age between the neonatal period to over 100 years.Astrocytes vary in cellular density between distinct brain regions (Emsley and Macklis, 2006;Oberheim et al., 2012).In the gray and white matter, the density of these cells is highly dependent on the local environment, particularly the presence of specific transcription factors and extrinsic cues during development (Cai et al., 2007;Kohler et al., 2021;Vue et al., 2014).We confirm that the density of astrocytes differ between distinct brain gray and white matter areas, and found that these cells were higher in numbers in the white matter, especially in the cerebellar white matter.Notably, the density of astrocytes in the white matter starts to decline with increasing age, indicating region-specific changes during aging process.While it has been suggested that astrocyte cell numbers do not change with age (Fabricius et al., 2013;Jinno, 2011;Lindsey et al., 1979;Olabarria et al., 2010;Pelvig et al., 2008;Robillard et al., 2016), some studies have reported region-specific differences in the brain before (Hansen et al., 1987;Mansour et al., 2008;Wang et al., 2006).It should, however, be considered that a majority of these studies, including ours, focused on GFAP + astrocytic subpopulations only.GFAP is a classical astrocytic marker, the expression of which differs across brain regions and alters with age (Kohler et al., 2021;Palmer and Ousman, 2018;Verkerke et al., 2021).In the neocortex, in particular, cells inhabiting different layers show variable GFAP expression following a centrifugal gradient (Forrest et al., 2023).Future studies should therefore take other markers into considerations as well to improve our understanding about the distribution and number of astrocytes across different brain regions.
Astrocytes are also diverse in morphology based on their location in the brain.In the gray matter, astrocytes are highly branched cells with round cell bodies.By contrast, white matter astrocytes are less branched and have elongated cell bodies along the white matter tracts with long thin processes (Garcia-Marin et al., 2007;Kohler et al., 2021;Miller and Raff, 1984;Sofroniew and Vinters, 2010).Consistent with this and based on GFAP immunoreactivity, we found that astrocytes indeed showed regional differences in morphology with gray matter astrocytes being morphologically more complex than white matter astrocytes.We observed no differences in astrocyte morphology between different white matter regions.It should be noted however that GFAP only reveals the cytoskeletal intermediate filaments of astrocytes and does not reflect their full morphology.Additionally, in a subset of cortical astrocytes, GFAP expression is below the level of immunohistochemical detection (Verkerke et al., 2021).More refined techniques as dye-filling could not be employed in FFPE tissue.Although our study remains an underestimation, we confirmed the presence of distinct astrocyte populations in the gray and white matter, respectively.Notably, astrocytes undergo changes in their morphology with age (Palmer and Ousman, 2018;Verkerke et al., 2021).They morphologically change from long, thin cellular processes to cells with short, stubby processes (Cerbai et al., 2012;Cruz-Sanchez et al., 1998;Jyothi et al., 2015;Kanaan et al., 2010;Robillard et al., 2016).Our findings by contrast suggest that gray matter astrocytes become more complex in morphology with increasing age, as indicated by the increase in total cellular processes, while no age-related changes were observed in white matter astrocytes.These findings suggest region-specific response to aging, contrasting earlier findings.It should, however, be noted that most studies to date have only focused on morphological changes during aging in single brain regions.Indeed, region-dependent changes in morphology during aging have been reported in rodents before (Amenta et al., 1998;Bondi et al., 2021).Our understanding on possible region-specific differences across regions especially in humans however remains limited and warrants further studies.
It is generally agreed that astrocytes molecularly and functionally differ dependent on their regional environment (Bugiani et al., 2022;Escartin et al., 2021;Pekny and Pekna, 2016;Sofroniew, 2020;Vasile et al., 2017).With age, astrocytes undergo region-specific changes in their molecular profile.Among the changes at the molecular level, expression of molecules involved in processes such as inflammation, synaptogenesis and pruning, and neuronal transmission are altered (Matias et al., 2019;Palmer and Ousman, 2018;Verkerke et al., 2021).
The accumulation of hyperphosphorylated tau in astrocytes is typically observed in ARTAG (Kovacs, 2020;Kovacs et al., 2016).It is characterized by astrocytes with granular and fuzzy appearances in the brain gray matter, thorn-shaped astrocytes in the brain subependymal, subpial, perivascular, and white matter areas and it is frequently found in the brain of older individuals (Kovacs, 2020;Kovacs et al., 2016).Remarkably, we found no signs of astrocytic tau accumulation in the examined brain regions of our individuals, including donors between 21 and 79 and 100+ years of age, except one subject of a unique population (i.e.centenarians) who showed classical features of ARTAG neuropathology.Importantly, this study did not evaluate brain regions prone to show ARTAG, however, the lack of ARTAG in the regions examined supports the notion that ARTAG is not per se a marker of astrocyte senescence (Ameen-Ali et al., 2022;Bachstetter et al., 2021), but a pathology mostly seen in aging due to accumulating initiating events (Kovacs, 2020).
Many studies into the molecular changes in astrocytes during aging across different brain regions have focused on mapping the transcriptome profile (Boisvert et al., 2018;Clarke et al., 2018;Matias et al., 2019;Orre et al., 2014;Soreq et al., 2017;Verkerke et al., 2021), but relatively little is known about these changes at the protein level.Here, we identified astrocyte specific markers with regional differences in abundance across different brain gray and white matter areas at the protein level.For instance, GLUL, an enzyme involved in glutamate metabolism (Anlauf and Derouiche, 2013;Jayakumar and Norenberg, 2016;Rose et al., 2013), and AQP1, a water channel protein involved in brain water homeostasis (Badaut et al., 2014;Foglio and Rodella, 2010; Fig. 5. Differentially expressed astrocyte specific proteins identified across different brain regions.Heatmap shows a total of 19 differentially expressed astrocyte specific proteins across the frontal lobe (gray and white matter) and cerebellum (white matter).Pink protein clusters: proteins differentially expressed between gray and white matter regions.Green protein cluster: proteins differentially expressed between the frontal lobe and cerebellum.Potokar et al., 2016;Zelenina, 2010), were highly expressed in the gray and white matter, respectively, consistent with earlier findings (Arcienega et al., 2010;Hassel et al., 2003;Kohler et al., 2021;Nesic et al., 2008;Satoh et al., 2007).Another marker, that is, ACBD7, a paralog gene associated to diazepam-binding inhibitor/Acyl-CoA binding protein (Lanfray et al., 2016;Lanfray and Richard, 2017), showed differential expression across frontal versus cerebellar regions.To our knowledge, differential regional expression of ACBD7 has not been reported before.Immunohistochemistry was employed to confirm astrocyte specificity of differentially expressed proteins; and to verify if differential expression of these proteins was due to different cellular density or expression levels.Analysis revealed that the regional differences in expression of GLUL, AQP1, and ACBD7 detected by proteomics in part reflect variation in the amount of protein expressed per cell or the number of cells expressing the protein.Notably, the expression of some astrocyte specific proteins, that is, GLUL and AQP1, region-dependently decreases with age.Age-dependent changes in GLUL and AQP1 expression or function have been reported before (Kaiser et al., 2005;Olabarria et al., 2011;Squier et al., 2011).We now add to these findings and show that age-related changes in the expression of these markers is highly dependent on location in the brain.Taken together, our findings suggest that aging affects astrocytes from different regions differently.
Some limitations apply to this work.We grouped donors ranging 21 to 79 years of age to distinguish donors including infants and children up to 11 years of age from the "older" (21-79 years) and oldestaged (100+ years of age) groups.Given the broad age range, withingroup variations could occur.Nonetheless, we verified the absence of outliers in this group of 21-79 years old donors, suggesting that comparable data were obtained from these donors.Additionally, this is primarily an observational study.For the identification of regional astrocyte specific markers we only employed tissue from 21 to 35 years old donors, as this was the only frozen tissue available.Such selection, however, allowed us to avoid the possible interference of developmentand age-related pathology.Furthermore, only a few astrocyte specific proteins were selected for further immunohistochemical analysis, but many others still remain to be investigated.Notably, our findings give insights into region-specific differences of astrocytes in the human brain at a proteome level however focused on the frontocortical gray matter and cortical or cerebellar white matter.In order to further extent our understanding on regional heterogeneity, and more specifically in the context of astrocytes, analysis of additional brain regions using highthroughput methods is necessary for future research.In this respect, it must also be acknowledged that we quantitated astrocyte morphology using a densitometric approach.Further confirmatory studies are needed using design-based stereology.Lastly, the resolution of our regional proteome analysis does not mimic the single-cell type level.Future studies should extend to this and employ single-cell type proteomic approaches to further characterize region-specific differences astrocytes in the aging human brain.

Conclusions
In conclusion, our findings provide insights into region-specific and age-related differences in astrocyte density, morphology, and protein expression in the normal human brain.This study paves the way to a better understanding of astrocyte biology and their role in brain diseases.Building a database of similar data from normal human brains at different ages is warranted.

Fig. 1 .
Fig. 1.Regional and age-related differences in the density of astrocytes.Immunohistochemical stains showing GFAP + astrocytes in the (a-c) frontal gray matter, (d-f) frontal white matter, and (g-i) cerebellar white matter across ages.Depicted images are stains of donors 5, 12, and 23.Blue stain indicates nuclei.Scale bar in (i): 10 µm.(j-k) Quantification of GFAP + astrocytes.Graphs show GFAP + cell density (j) across regions per age group and (k) across ages per region.Data are presented as boxplots.Boxes highlight the 25th-75th percentiles, whereas whiskers show the minimum to maximum values.* p < 0.05, ** p < 0.01.

Fig. 2 .
Fig. 2. Regional and age-related differences in morphological complexity of astrocytes.(a) Illustrations showing representative traced astrocytes amongst brain gray and white matter areas across ages.Depicted images are astrocytes tracer from donors 2, 16, and 22. (b-e) Sholl analysis of traced astrocytes.Graphs show total number of intersections (b) across regions per age group and (c) across ages per region.Data are presented as boxplots.Boxes highlight the 25th-75th percentiles, whereas whiskers show the minimum to maximum values.* p < 0.05, ** p < 0.01, *** p < 0.001.

Fig. 3 .
Fig. 3. Phospho-tau immunostaining in human brain (a-b) AT8 immunoreactivity is absent in the frontal white matter of (a) infants and children up to 11 years old and (b) 21-79 years old donors.In the 100+ years old donors (c), only one of six harbored a couple of GFAP + cells were found co-expressing AT8 in the frontal white matter.(d) Immunohistochemical staining with AT8 in the frontal cortex of only this 100+ years old donor shows positivity of a cluster of sub-pial TSA and of rare granular astrocytes in the cortex.Depicted images are stains of donors 1, 12, and 19.Blue stain indicates nuclei.Scale bar: 10µm.

Fig. 4 .
Fig. 4. Proteins differentially expressed across different brain regions.(a) Workflow: gray and white matter of the middle frontal gyrus and white matter of the cerebellum from 4 donors were isolated using laser capture microdissection for mass spectrometry analysis.Black square highlights the middle frontal lobe.(b) PCA of the proteomes of the 3 brain regions.The first component accounts for 54.7% and segregates gray and white matter brain regions.Samples from different white matter regions also form distinct clusters.These samples are separated by the second component, which explains 16.1% of the variability.(d) Hierarchical clustering analysis of 1408 proteins with statistical differential expression across the 3 brain regions (p < 0.05).Heatmap shows a total of 6 protein clusters with their z-scored expression values (LFQ intensities).Dendrograms denoting distinct protein clusters are highlighted in different colors.(d) Plots showing the expression pattern of each protein cluster across regions.The number of proteins identified in each cluster is indicated above the plot.Thick line indicates the average expression pattern of all individual proteins identified in the cluster.GM gray matter, WM white matter, PC principal component, FrGM frontal gray matter, FrWM frontal white matter, CbWM cerebellar white matter.

Fig. 6 .
Fig. 6.Identification of astrocyte specific proteins.(a-c) LFQ intensity of the selected astrocyte specific proteins GLUL, AQP1 and ACBD7.(d-l) Astrocyte proteins selected for validation by immunohistochemistry. Astrocyte specificity was confirmed by double labeling of GLUL, AQP1, and ACBD7 in red with GFAP in green.Blue stain indicates nuclei.Scale bar: 5μm.Depicted images are stains of donor 10.FrGM frontal gray matter, FrWM frontal white matter, CbWM cerebellar white matter.

Fig. 9 .
Fig. 9. ACBD7 expression across gray and white matter regions in different age groups.Immunohistochemical stains showing ACBD7 + cells in (a-c) frontal gray matter, (d-f) frontal white matter, and (g-i) cerebellar white matter across ages.Blue stains indicate nuclei.Scale bar in (i): 10µm.Depicted images are stains of donors 1, 11, and 20.(j-k) Quantification of ACBD7 + cells.Graphs show ACBD7 + cell density (j) across regions per age group and (k) across ages per region.Data are presented as boxplots.Boxes highlight the 25th-75th percentiles, whereas whiskers show the minimum to maximum values.* p < 0.05, ** p < 0.01.

Table 1
Demographics of donors used in the study.

Table 2
List of antibodies used for immunohistochemistry.