We first performed a flow cytometry screen of surface molecules to identify antigens that may enable isolation of reactive astrocytes for downstream analyses. We differentiated hiPSCs to neural progenitor cells that self-assembled into neurospheres and then plated the spheres onto poly-ornithine/laminin-coated dishes to allow migration and differentiation into astrocytes (Barbar et al., 2020a; Figure 1A; Supplementary Figure 1). After 66–75 days, the cells that migrated out of the neurospheres formed a monolayer composed of about 50% astrocytes, 40% oligodendrocyte lineage cells, and 10% neurons. Next, we stimulated these cultures with TIC for 24 h, leaving some wells as unstimulated controls. Through immunofluorescent staining for glial fibrillary acidic protein (GFAP), we observed a drastic change in morphology of TIC-induced reactive astrocytes when compared to the unstimulated controls (Figure 1B) as we have previously reported (Barbar et al., 2020a). Of note, elevated GFAP staining was found in both unstimulated and TIC-induced reactive astrocytes, supporting previous findings attributing increased GFAP levels to in vitro culture conditions (Molofsky et al., 2012). We then screened 361 surface molecules, gating on the CD49f⁺ population (Supplementary Figure 2) and selected six surface molecules that were enriched in the TIC-induced reactive astrocytes for further investigation: VCAM1, BST2, ICOSL, HLA-E, PD-L1, and PDPN (Figure 1C). Using histogram plots we showed the enrichment of each of these markers within the TIC-induced reactive astrocyte population compared to the unstimulated cells (Figure 1D).

To confirm screening results, we next performed an independent immunofluorescence assay on a second batch of differentiations; in this experiment, we used CD49f-sorted astrocytes rather than mixed cultures to eliminate any non-astrocyte responses to the inducing stimuli. As shown in Figure 2A, all six proteins were detected in the TIC-stimulated condition. Quantification of immunofluorescence staining for VCAM1, BST2, HLA-E, PD-L1, and PDPN confirmed a significantly greater number of positive cells upon stimulation across astrocytes from three independent hiPSC lines from healthy donors (Figure 2B). Next, leveraging our available RNA-sequencing datasets (Barbar et al., 2020a), we probed for each of the six genes encoding for the markers to assess whether they were upregulated at the mRNA level. All genes were upregulated within the three hiPSC-line previously used (Figure 3A); notably, these findings were confirmed when we extended the analysis to an independent cohort of an additional 13 hiPSC lines that were differentiated to astrocytes in a parallel study (Figure 3B). We also probed for these six genes within our available scRNA-sequencing datasets of TIC-induced reactive versus unstimulated astrocyte cultures (Barbar et al., 2020a; Figure 3C). Focusing on the astrocyte clusters, we found that expression of each gene is distinctly higher in the TIC-induced reactive condition, which confirms that VCAM1, BST2, ICOSL, HLA-E, PD-L1, and PDPN are upregulated upon stimulation, with BST2 being the most upregulated and ICOSL being the least (Figure 3D).

We therefore conclude that VCAM1, BST2, ICOSL, HLA-E, PD-L1, and PDPN are highly expressed by hiPSC-derived astrocytes in response to TIC stimulation and we propose them as putative, novel markers of TIC-induced reactive astrocytes.

We have previously shown that human and rodent cells have a highly conserved transcriptomic signature upon TIC stimulation (Barbar et al., 2020a). Although transcriptomic analyses are powerful tools, gene expression does not always correlate with protein abundance (Leitner et al., 2021; Johnson et al., 2022). Thus, we extended the characterization of our human model to include proteomic profiling of TIC-induced reactive versus unstimulated astrocytes derived from five independent hiPSC lines from healthy donors (Figure 4A). We confirmed high protein quality and no loading bias (Supplementary Figure 3A) and performed Tandem Mass Tag-based (TMT) liquid chromatography/liquid chromatography-mass spectrometry/mass spectrometry (LC/LC-MS/MS; Bai et al., 2017), which identified 10,649 unique proteins. Of these, 601 proteins were differentially expressed (DE; 400 upregulated and 201 downregulated; Supplementary Table 1) using the thresholds of p-value < 0.05 and absolute z-score > 2. Furthermore, principal component analysis (Figure 4B) of the top 10% DE proteins showed a clear separation of the TIC-treated and unstimulated astrocytes, indicating a dramatic change to the astrocyte proteome upon TIC stimulation. Hierarchical clustering of the DE proteins and samples emphasized the robust and highly consistent response to proinflammatory stimuli across all lines, with some line-to-line variability (Supplementary Figure 3B). We also examined the six markers identified through flow cytometry and further confirmed their increased abundance in TIC-induced reactive astrocytes (Figure 4C), with BST2 and ICOSL showing the greatest fold change. We then probed a recent proteomic dataset of TIC-induced reactive rat astrocytes (Guttenplan et al., 2021) and found that VCAM1 and RT-BM1 [RT1.A1(F) domain; ortholog of HLA-E] were more abundant in rat TIC-induced reactive astrocytes compared to unstimulated cells, indicating these two as potentially conserved markers (Supplementary Figure 4). The other human markers were not detected in the rat dataset using this methodology.

Of the top 10% most DE proteins detected in our human system (Figure 4D), many were known chemokines (e.g., CXCL10, CXCL11) and interferon inducible proteins (e.g., IFI6, IFIT1, IFIT3). Pathway enrichment analysis based on the DE proteins revealed a total of 67 significantly enriched GO-terms (Supplementary Table 2), mostly related to immune responses, cytokine-associated signaling, and recruitment of peripheral immune cells to the CNS (Figure 4E).

Taken together, these results suggest that the proteome of hiPSC astrocytes undergoes rapid and robust changes upon TIC stimulation. The proteomic profiling of reactive astrocytes suggests a putative participation in immune responses and in the processes of neuroinflammation associated with CNS diseases, while independently validating the novel surface markers.

Next, we sought to confirm the presence of these markers in the astrocytes in human post-mortem brain samples. We selected a cohort of AD patient and age-matched non-symptomatic (NS) controls (Sadick et al., 2022) and stained formalin-fixed paraffin embedded frontal cortex samples for our putative reactive astrocyte markers. Broadly, we report no or minimal immunofluorescence for any marker in the NS brain. In the AD brain, we observed that PD-L1⁺ cells were almost exclusively GFAP⁺ in the white matter, suggesting this marker is largely restricted to astrocytes in this region (Figure 5 and Supplementary Figure 5). Another marker, PDPN, was also almost exclusively found in GFAP⁺ astrocytes in the AD brain, while some GFAP^– cells were positive for PDPN in the NS brain. Similarly, BST2 protein was abundant in the AD tissue compared to NS tissue; however, its presence was not limited to GFAP⁺ astrocytes. This could indicate the presence of BST2⁺ non-astrocytic cells in the human brain, but it must be noted that GFAP astrocytes exhibit regional heterogeneity and some of the BST2 signal observed in these regions could possibly be in these GFAP^– astrocytes. Finally, astrocytes in the AD tissue showed increased HLA-E toward the sub-pial regions of the layer 1 of the prefrontal cortex. However, HLA-E was also enriched near the blood vessels and outside GFAP⁺ astrocytes in these regions. Together, these preliminary investigations suggest that cell surface markers identified in hiPSC-derived reactive astrocyte cultures are able to detect populations of astrocytes in the human post-mortem brain, with enrichment in the AD brain.

In this study, we used six hiPSC lines from healthy donors, that were previously generated using the NYSCF Global Stem Cell Array^®, a fully automated platform for fibroblast and blood reprogramming (Paull et al., 2015). All lines passed stringent quality controls, including karyotyping, mycoplasma testing, pluripotency efficiency, and identity match. A certificate of analysis from one of the lines is shown in Supplementary Figure 6. Five of these lines were used in our previous studies on hiPSC-derived astrocytes (Barbar et al., 2020a,b). Table 2 contains the demographic information for the donors of the skin biopsies that were reprogrammed to hiPSCs.

All cultures were maintained in a 37°C incubator with 5% CO₂. For astrocyte differentiation, hiPSCs were plated at 1.5 × 10^5 cells per well in a 6-well geltrex-coated plate and fed with mTeSR1 maintenance medium with 10 μM Y27632 for the first 24 h after seeding (Marotta et al., 2021). Differentiation and FACS-purification of CD49f⁺ astrocytes were performed using a serum-free protocol that we have previously established; differentiation media and patterning reagents are fully detailed in our previous publications and summarized in a schematic of the protocol in Supplementary Figure 1 (Barbar et al., 2020a,b). After enzymatic dissociation of neural cultures for fluorescence-activated cell sorting on days 66–75, neurospheres were replated and isolation of CD49f⁺ astrocytes was repeated on Day 95–105 to isolate “second round spheres,” as previously detailed (Barbar et al., 2020a). To model a neurotoxic reactive astrocyte state, mixed neural cultures and sorted CD49f⁺ astrocytes around day 65 were stimulated for 24 h with human recombinant proteins TNF (30 ng/mL, R&D Systems; 210-TA-020), IL1α (3 ng/mL, Sigma; 3901), and C1q (400 ng/mL, MyBioSource; MBS143105), which we defined as “TIC.”

Astrocytes from three lines were fixed using 4% paraformaldehyde for 10 min then washed 3× with DPBS. Cells were then incubated for one hour in blocking solution (DPBS + 5% donkey serum + 0.1% Triton-X-100) followed by overnight incubation with primary antibody at 4°C. After 24 h, cells were washed 3× with DPBS followed by incubation for 1 h with secondary antibody (Alexa Fluor) at 1:1000 dilution and washed 3× with DPBS. Cells were then incubated with HOECHST 33342 (1:1000 dilution) for 10 min and washed 3× with DPBS. Images were acquired on Zeiss scanning confocal microscope (LSM800). All antibodies used are listed in the Table 3.

Astrocytes fixed and stained for each marker were imaged on the Opera Phenix High-Content Screening System (PerkinElmer, Waltham, MA, United States). Images were acquired at 10× magnification and stitched together to capture nearly the entirety of each well (n = 9 fields per well) while maintaining consistent imaging settings for each marker across lines. Harmony analysis software was used to identify nuclei of live cells using size and HOECSHT 33342 intensity criteria to exclude apoptotic nuclei and debris. The nuclear region was then expanded by 50% to create a region of interest (ROI) and the Alexa Fluor intensity within this region in a separate channel was quantified within each ROI. Thresholds were set consistently for each marker across all lines, with ROIs exceeding the threshold identified as astrocytes positive for the marker. Results were reported as percentage of positive astrocytes out of all live cells (HOECHST nuclei) identified.

We performed a cell surface antigen screen on cells differentiated from hiPSC line 6 (Line ID 051106). Cultures at day 64 were enzymatically digested with accutase (ThermoFisher, Waltham, MA, United States; A1110501) for 20 min at 37°C and second round spheres were re-plated to continue differentiation as described (Barbar et al., 2020a). At day 95, cells were stimulated with TIC cocktail for 24 h. Unstimulated wells were also prepared as control. Cells were then dissociated using accutase for 30 min at 37°C, diluted in DMEM, and filtered using a 70 μm filter (STEMCELL Technologies; 27260) to eliminate spheres and cells aggregates. The cell suspension was centrifuged at 300 g for 5 min and resuspended in FACS buffer (DPBS + 0.5% BSA + 1% penicillin/streptomyocin + 2mM EDTA + 20 mM glucose). For barcoding, TIC-induced astrocytes were incubated for 15 min at room temperature (RT) with live/dead fixable near-IR (ThermoFisher; L34976), washed 2× with FACS Buffer, and centrifuged at 300 g (RT) for 5 min. Unlabeled (unstimulated) and barcoded (TIC-stimulated) cells were then combined and stained with CD49f-Pacific Blue (1:25 dilution, BioLegend; 313619) and incubated for 20 min at 4°C. Cells were washed 1× with FACS buffer and centrifuged at RT and 300 g for 5 min. Cells were distributed evenly into four 96-well plates of the LEGENDScreen™ Human PE kit (Biolegend; 700007), containing 361 PE-conjugated antibodies to cell surface markers. Plates were incubated for 20 min at 4°C, washed 1× with FACS buffer, and resuspended in each well. Samples were acquired on a 4-laser Attune NxT Flow Cytometer (405, 488, 561, 633, Thermo Scientific). Subsequent data analysis was performed in FlowJo v.10.

mRNA levels for the six markers of TIC-induced reactive astrocytes were retrieved from our previous bulk transcriptome analysis performed on CD49f⁺ TIC-induced reactive versus unstimulated astrocytes, differentiated from lines 3, 5, and 6 (Barbar et al., 2020a).

mRNA levels for the six markers of TIC-induced reactive astrocytes were also retrieved from an independent cohort of 13 hiPSC lines from MS patients and age/sex matched controls. These lines were reprogrammed using the NYSCF Global Stem Cell Array^®, from peripheral blood mononuclear cells and they all passed the quality controls described above. Human iPSC lines were differentiated to CD49f-astrocytes and stimulated to a TIC-induced reactive state as described above. Details can be found in a manuscript published back-to-back to the special issue “Glia-mediated neurotoxicity: uncovering the molecular mechanisms” (Smith et al., 2020). Cells were washed and lysed in RLT Plus buffer (Qiagen RNeasy Plus Micro Kit #74034) and lysates were frozen at −80°C. RNA was isolated following manufacturer’s recommended protocol. RNA quality was assessed on a Fragment Analyzer to proceed to library preparation. Total RNA libraries were generated using Illumina Stranded Total RNA Prep with Ribo-Zero Plus kit. Libraries were sequenced (100 bp paired end) on an Illumina NovaSeq. Sequencing data quality was checked with FastQC (Andrews, 2010) followed by MultiQC (Ewels et al., 2016). Transcript quantification was done with Salmon (v1.4.0) (Patro et al., 2017) against RefSeq GRCh38.p12 reference genome (Schneider et al., 2017) in a decoy aware fashion using the entire genome as the decoy sequence (Srivastava et al., 2020). These transcript quantifications were imported into R (4.1.2) and aggregated to gene level using tximeta (v1.12.3) (Love et al., 2020). Using the gene level quantifications, library size normalization and differential expression testing was done using DESeq2 (v1.34.0) (Love et al., 2014) where donor was a blocking factor and stimulation (unstimulated versus TIC-stimulated) was the comparison variable. P values were calculated by DESeq2 then adjusted for multiple comparisons using the Benjamini and Hochberg method.

We assessed mRNA expression of the six surface markers from our previous analysis on astrocytes derived from lines 3 and 5 (Barbar et al., 2020a). We loaded the RStudio objects associated with Barbar et al. (2020a); see methods of that paper for detailed bioinformatic information. We then assessed mRNA expression of the six surface markers via the FeaturePlot and DotPlot commands in the Seurat software package (v4.0.1), separating cells by A0/A1 labels generated as part of Barbar et al. (2020a) and feature plots were generated.

Five hiPSC lines were differentiated for proteomics. Human iPSCs were differentiated into astrocytes as described above. FACS-purification of CD49f⁺ astrocytes was performed on D75 from all lines. Table 4 details yield of CD49f⁺ astrocytes per line. CD49f⁺ astrocytes were plated into two wells of a poly-Ornithine/g-coated 24-well Nunclon-Delta plate in PDGF medium. Cells were fed the day after sorting (Day 1 post-sort) with 2/3 media change of Glial medium. On Day 3 post-sort, 2/3 medium change with Glial medium was completed. On Day 5 post-sort, full media change with Glial medium for unstimulated condition and full medium change with Glial medium plus TIC cocktail for TIC-stimulated condition. After 24 h cells were washed once with ice-cold DPBS then lysed in the well using lysis buffer containing 50 mM HEPES (Thermo Scientific; 15640080, pH = 8.5), 8M Urea (Sigma, U5378), 0.5% sodium deoxycholate (Sigma; 30970), 1 cOmplete mini EDTA-free Protease Inhibitor Cocktail tablet (Sigma; 11836170001) per 10 mL and 1 PhosSTOP tablet (Sigma; 4906845001) per 10 mL. Lysates were collected in Protein LoBind tubes (Eppendorf; 022-43-108-1) and frozen at −80°C.

Protein concentration of the cell lysate was estimated by Coomassie-stained short SDS gel with BSA as a standard. Each sample was digested and TMT labeled using our previously optimized protocol (Xu et al., 2009; Wang et al., 2020) with Lys-C (Wako, 1:100 w/w) at 21°C for 3 h, then diluted with 50 mM HEPES (pH 8.5) to decrease urea concentration to 2 M, and further digested by trypsin (Promega, 1:50 w/w) at 21°C overnight. The digested peptides were reduced by dithiothreitol (DTT, 1 mM) for 30 min, alkylated by iodoacetamide (IAA, 10 mM) at dark for 30 min, and quenched with DTT (30 mM) for 30 min. The digested peptides were desalted with a C18 Micro SpinColumn (Harvard apparatus) and dried by speedvac. The dried samples were re-dissolved in 50 mM HEPES, pH 8.5 (∼2 μg/μL), reacted with 150 μg of TMTpro reagents for 30 min, then quenched with 0.5% hydroxylamine for 15 min. The labeled samples were pooled equally and desalted again with a solid-phase extraction cartridge. The desalted sample was dried before further LC/LC-MS/MS analysis.

The pooled TMT labeled peptides were fractionated by an offline basic pH RPLC with a XBridge C18 column (3.5 μm particle size, 4.6 mm × 25 cm, Waters; buffer A: 10 mM ammonium formate, pH 8.0; buffer B: 90% AcN, 10 mM ammonium formate, pH 8.0). The peptides were eluted in a 160 min gradient of 15–50% buffer B, and 160 fractions were collected every min and concatenated to 40 fractions. Each fraction was analyzed on a self-packed column (75 μm × 15 cm with 1.9 μm C18 resin) coupled with a Q Exactive HF Orbitrap MS (Thermo Fisher Scientific). The peptides were eluted at 0.25 μL/min flow rate with a 60 min gradient of 18–45% buffer B (buffer A: 0.2% formic acid, 5% DMSO; buffer B: buffer A plus 65% AcN). The mass spectrometer was operated in data-dependent mode with a MS1 scan in Orbitrap mass analyzer [450–1600 m/z; 60,000 resolution; 1 × 10⁶ automatic gain control (AGC) target; 50 ms maximum ion time] and 20 data-dependent MS2 scans (60,000 resolution, 1 × 10⁵ AGC target, 110 ms maximum ion time, 32% HCD normalized collision energy (NCE), 1.0 m/z isolation window, 0.2 m/z isolation offset, and 10 s dynamic exclusion).

Protein identification and quantification were performed using the JUMP search engine (Wang et al., 2014). The human protein database was generated by combining downloaded Swiss-Prot, TrEMBL, and UCSC databases and removing redundancy (83,955 entries). The target-decoy database was used to estimate the false discovery rate (FDR) (Peng et al., 2003). Major search parameters included precursor mass tolerance (±20 ppm) and product ion mass tolerance (±15 ppm), full trypticity, two maximal missed cleavages, static modification for TMTpro tags (+304.20715 on Lys and N-termini) and carbamidomethylation (+57.02146 on Cys), dynamic modification for oxidation (+15.99491 on Met). The resulting PSMs were filtered by mass accuracy and then grouped by precursor ion charge state followed by the cutoffs of JUMP-based matching scores (Jscore and ΔJn) to reduce FDR below 1% for proteins. When the same peptide was derived from numerous homologous proteins, the peptide was matched to the protein with the top PSM number, according to the rule of parsimony. Protein quantification was performed using TMT reporter ion intensities by a built-in program of JUMP software suite as previously described (Niu et al., 2017).

The differential expression (DE) analysis was performed following the previously reported method (Bai et al., 2020). Briefly, the ratios of all proteins between the replicates were modeled with a Gaussian distribution to evaluate standard deviation for z value analysis. The protein p values were calculated by using the moderated t-test (Phipson et al., 2016). The proteins with p-value < 0.05 and z > 2 were considered as DE proteins. All DE proteins are listed in Supplementary Table 1.

All 601 DE proteins from the proteomics analysis, including both up and downregulated proteins, were considered for pathway enrichment analysis. Functional enrichment analysis was performed on the STRING database (version 11.5), where each DE protein was ranked by its log₂fold change value. DE proteins were significantly enriched in a total of 67 pathways in the “biological processes” category. The downregulated proteins were enriched in a single pathway, “locomotory behavior (GO:0007626).” All the upregulated proteins were enriched in the remaining 66 pathways. All the significantly enriched pathways, GO-terms, enrichment score, pathway size, and the number of proteins mapped to each pathway are given in Supplementary Table 2.

We performed immunohistochemistry on post-mortem brain samples from AD and age-matched NS donors; their demographic and clinical information is listed in Table 5. The formalin-fixed paraffin-embedded human AD and NS tissues were sectioned at 5 μm and mounted onto glass microscope slides. The slides were placed on a 60°C hotplate for 30 min and moved through HistoChoice clearing aged (2 washes, 5 min each) to dewax the paraffin. Next, the sections were rehydrated by submerging the slides in the following solutions:100% ethanol (2 washes, 5 min each), 95% ethanol (1 wash, 5 min), 70% ethanol (1 wash, 5 min), 1×PBS (1 wash, 5 min). For the sections to be treated with BST2 and PDPN, antigen retrieval was performed by boiling the sections in a citrate buffer in the microwave for 3 min. The sections were then treated with a blocking buffer (10% normal goat serum and 0.4% Triton X-100 prepared in 1×PBS) for 1 h at room temperature in a humid chamber. Following blocking, they were treated with the primary antibodies and incubated overnight (16–20 h) at 4°C in a humid chamber. The following day, sections were washed with 1×PBS (3 washes, 10 min each) and treated with the secondary antibodies for 1 h at room temperature in a darkened humid chamber. To remove unbound secondary antibodies 3 10-min washes in 1×PBS was performed followed by a nuclear co-stain with DAPI (0.5 μg/mL prepared in PBS) for 1 min. The sections were then treated with TrueBlack Lipofuscin Autofluorescence Quencher (Biotium) for 1 min to reduce background fluorescence. Finally, sections were mounted with Fluoromount-G, and a 22 × 50 mm cover glass was placed onto the sections. Images were acquired on a Keyence BZ-X710 using a 60× oil objective. All antibodies used are listed in the Table 3. For quantification, 3 different tissue sections and 3 different ROI per section for each condition (AD and NS) were used. Unpaired t-tests were performed using the means from each tissue.