Astrocyte Senescence Impairs Synaptogenesis due to Thrombospondin‐1 Loss

The accelerated senescence SAMP8/TaHsd model and the senescence‐resistant control SAMR1/TaHsd strain (Takeda et al. 1981) were used in this study. Mice were purchased from Inotiv Inc. and housed in the animal facility of the Biomedicine Institute of Valencia. Experiments were performed in male young (2 months, 2‐m), middle‐aged (6 months, 6‐m), and old (10 months, 10‐m) animals. For hippocampal neuron culture, RjOrl:SWISS wild‐type mice were purchased from Janvier Labs, and timed‐pregnant mice (E18.5) were bred in the animal facility of the Biomedicine Institute of Valencia. Mice were sacrificed by cervical dislocation for primary cultures, and by anesthetic overdose prior to perfusion for immunohistochemistry techniques. Animals were bred under controlled temperature conditions, 12 h light/dark cycle and water and food ad libitum. All experimental procedures and handling were in accordance with the European Union Council guidelines (2010/63/EU) and the Spanish regulation (RD53/2013). Procedures were approved by the Ethics and Animal Welfare Committee (CEEA) of the Biomedicine Institute of Valencia and CSIC (CEEA references: 2023‐VSC‐PEA‐0212 and 2024‐VSC‐PEA‐0094).

2, 6, and 10‐m SAMR1 and SAMP8 mice were euthanized by cervical dislocation and hippocampi were dissected from the brain. Hippocampal cell suspension was obtained using the Adult Brain Dissociation Kit (Miltenyi, 130‐107‐677) in combination with the gentleMACS Octo Dissociator (Miltenyi, 130–095‐937). ACSA‐2⁺ cells (astrocytes) were isolated by magnetic cell separation (MACS) method following the manufacturer's procedure and protocols (Miltenyi Biotec, 130‐097‐679). ACSA‐2⁺ cells were then seeded (200,000 cells) in 24‐well poly‐D‐lysine (Sigma‐Aldrich, 27964‐99‐4) and laminin‐coated plate (Sigma‐Aldrich, L2020), as recommended by Miltenyi Biotec for primary astrocyte cultures from the adult brain. Astrocytes were maintained in the defined commercial medium AstroMACS (Miltenyi Biotec, 130‐117‐031) supplemented with 2 mM L‐glutamine (Lonza, 17‐605C) and 100 μ/mL penicillin–streptomycin (Lonza, 17‐603E). Astrocytes were maintained for 14 days in vitro (DIV) and every other day, half of the medium was replaced. On the final day, cells were fixed with 2% paraformaldehyde (PFA) (PanReac, 141451). The cultures were maintained in an incubator at 37°C in controlled humidity at 5% CO₂.

For the primary culture of hippocampal neurons, 18.5‐day old embryos (E 18.5) from wild‐type mice were used. Embryos were decapitated and hippocampal cell suspension was obtained using the Neural Tissue Dissociation Kit (Miltenyi, 130‐092‐628) in combination with the gentleMACS Octo Dissociator (Miltenyi, 130‐095‐937). Neurons were isolated by MACS using the Neuron Isolation Kit (Miltenyi Biotec, 130‐115‐390). Neurons were seeded (150,000 cells) onto poly‐D‐lysine and laminin‐coated 24‐well plate and maintained for 11 DIV. Neurobasal medium (Thermo scientific, 21103049) supplemented with B‐27 (1×) (Thermo scientific, 17504044), 2 mM L‐glutamine and 100 μ/mL penicillin–streptomycin was used. On the first day post‐plating, AraC (2 μM) was added to reduce non‐neuronal contamination. The cultures were maintained in an incubator at 37°C in controlled humidity at 5% CO₂.

At 11 DIV, neuron medium was totally replaced by ACM from Ast‐Diff or half replaced by ACSA‐2⁺ ACM during 3 h. Gabapentin (GBP, MedChemExpress, HY‐A0057) and TSP‐1 (MedChemExpress, HY‐P701325) dosages were based on Cheng et al. (2016) procedures. GBP (32 μM) was added to neurons 30 min before ACM treatment and maintained 3.5 h with the neurons. TSP‐1 (250 ng/mL) was added directly diluted with the ACM during 3 h. Finally, neurons were fixed with 2% paraformaldehyde (PFA) and synapses were quantified.

SAMR1 and SAMP8 NSCs were isolated from the hippocampal dentate gyrus of 2‐m mice and expanded in the presence of mitogens following Babu's protocol (Babu et al. 2011). These cells were maintained in adherent culture with Neurobasal medium supplemented with B‐27 (1×), L‐glutamine (2 mM), penicillin–streptomycin (100 μ/mL), fibroblast growth factor 2 (FGF2 20 ng/mL) (PeproTech, 100‐18B) and epidermal growth factor (EGF 20 ng/mL) (PeproTech, AF‐315‐09). For cellular assays, experiments were conducted using comparable passage (P) number for both models, and cells were discarded at approximately P26. To obtain SAMR1 and SAMP8 Diff‐Astrocytes, NSCs were seeded (150,000 cells) in differentiation medium, consisting of Neurobasal medium supplemented with B‐27 (1×), L‐glutamine (2 mM), penicillin–streptomycin (100 μ/mL) and Fetal Bovine Serum (FBS at 5%) (Thermo scientific, 10100147) for a period of 8–11 days. On Day 4, half of the medium was replaced with fresh differentiation medium. Cultures were maintained in an incubator at 37°C in controlled humidity at 5% CO₂. For immunofluorescence staining, 8 DIV cells were fixed on coverslips in 2% PFA for 10 min.

SAMP8 Diff‐Astrocytes were co‐transfected at day 11 with pcDNA3 mTSP1 and pMaxGFP, or with empty pcDNA3 and pMaxGFP as a control, using Lipofectamine 2000 (Fisher scientific, 15338‐030). pcDNA3 mTSP1 was a gift from Paul Bornstein (Addgene plasmid #12405; http://n2t.net/addgene:12405↗; RRID: Addgene_12405). 24 h after the transfection, the medium was changed to remove Lipofectamine. Cells were processed for RNA extraction or astrocyte conditioned medium collection 2 days later.

Diff‐Astrocyte cultures were seeded in 24‐well plates and maintained 11 DIV. The culture medium was then completely removed and replaced with fresh differentiation medium for a 24‐h conditioning. After this time, the conditioned medium was collected and centrifuged at 300 g for 5 min to remove cellular traces. ACM from Diff‐astrocytes and transfected Diff‐astrocytes was immediately applied for 3 h to the hippocampal neurons (via full media change), or stored at −20°C.

On the other hand, ACM collection from primary ACSA‐2⁺ astrocyte cultures required a different procedure. Astrocytes were seeded in 24‐well plates and half of the AstroMACS medium was collected and replaced every 48 h for a period of 14 days. The medium was collected from each well, pooled, and centrifuged at 300 g for 5 min to remove cellular traces. For neuronal treatment, half of the neuronal culture medium was replaced during 3 h with ACSA‐2⁺ SAMR1 or SAMP8 ACM.

The SA‐β‐gal reaction was performed following the protocol established by Debacq‐Chainiaux et al. (2009) with minor modifications. This reaction was performed in the same way for hippocampal tissues and for cell culture models. For cell cultures, the cells were fixed with 2% formaldehyde and 0.2% glutaraldehyde, followed by 3 washes with 0.1 M phosphate buffer (PB). The β‐galactosidase staining solution was then prepared and the samples were incubated with it for 5 h at 37°C with humidity. Primary antibodies were then applied to label astrocytes and nuclei with DAPI (Sigma‐Aldrich, D9542). Finally, the number of SA‐β‐gal positive cells was quantified as a percentage of the total cells or astroglial cells photographed per field. For tissue, the sections were completely immersed in the staining solution for 5 h and subsequently photographed. SA‐β‐gal signal intensity was normalized to the quantification area (μm²).

To perform immunofluorescence techniques, cells were pre‐fixed with 2% PFA for 10 min, followed by washing with 0.1 M PB. Subsequently, the samples were incubated at room temperature for 1 h with a blocking buffer composed of 0.1 M PB, 10% FBS, and 0.5% Triton X‐100. The primary antibody was then diluted in this blocking solution. This incubation was performed at 4°C for 24 h. The following day, cells were incubated with the secondary antibody, diluted in 0.1 M PB, for 2 h. DAPI was used to label nuclei. Primary antibodies: mouse anti‐GLAST (1:50; Miltenyi Biotec, 130‐119‐161), rabbit anti‐ATP1B2 (1:200; Alomone Labs, ANP‐012), guinea pig anti‐MAP2 (1:1000; Synaptic Systems, 188,004), rabbit anti‐VGLUT1 (1:300; GeneTex, GTX133148), mouse anti‐PSD95 (1:500; Thermo scientific, MA1045), rabbit anti‐GFAP (1:500; Sigma‐Aldrich, G3893), mouse anti‐TSP‐1 (1:50; Santa Cruz, sc‐59887) and guinea pig anti‐S100β (1:500; Synaptic Systems, 287004). Secondary antibodies: Alexa Fluor 488 anti‐mouse (1:500; Invitrogen, A21202), Alexa Fluor 555 anti‐mouse (1:500; Invitrogen, A31570), Alexa Fluor 488 anti‐rabbit (1:500; Invitrogen, A21206), Alexa Fluor 647 anti‐rabbit (1:500; Invitrogen, A31573), Alexa Fluor 488 anti‐guinea pig (1:500; Jackson, 706‐545‐148) and Alexa Fluor 647 anti‐guinea pig (1:500; Jackson, 706‐605‐148).

The brains were processed with a Leica VT 1200 vibratome. Serial 40 μm sections comprising the entire hippocampus were collected. Before immunostaining, the tissue was washed three times for 5 min with 0.1 M PB to remove azide from the storage medium. The tissue was then incubated at room temperature for 1 h with a blocking buffer composed of 0.1 M PB, 10% FBS, and 0.5% Triton X‐100. The primary antibody was then diluted in the blocking buffer, and the sections were completely immersed in the mixture and kept shaking at 4°C for 24 or 48 h. The next day, the slices were then incubated for 2 h with the secondary antibody diluted in 0.1 M PB, protecting them from light. DAPI was used to label nuclei. Primary antibodies: rabbit anti‐VGLUT1 (1:300; GeneTex, GTX133148), mouse anti‐PSD95 (1:500; Thermo scientific, MA1045), rabbit anti‐GFAP (1:500; Sigma‐Aldrich, G3893), mouse anti‐TSP‐1 (1:50; Santa Cruz, sc‐59887), and guinea pig anti‐S100β (1:500; Synaptic Systems, 287004). Secondary antibodies: Alexa Fluor 555 anti‐mouse (1:500; Invitrogen, A31570), Alexa Fluor 488 anti‐rabbit (1:500; Invitrogen, A21206), Alexa Fluor 647 anti‐rabbit (1:500; Invitrogen, A31573) and Alexa Fluor 488 anti‐guinea pig (1:500; Jackson, 706‐545‐148).

Cells and tissue were observed with a Leica TCS SP8 confocal microscope and LSM 980 confocal microscope with ZEISS Airyscan detector. For the SA‐β‐gal reaction in cells, the images were obtained on a Leica DM6 B Automatic Upright Microscope (Thunder) and for tissue a Leica Aperio Versa Digital Scanner was used. Tissue images were captured at 1024 × 1024 resolution with a 40× focal lens, tracing vertical maps to obtain representative areas of all hippocampal regions. Cell photographs were taken with a 40× or 63× focal lens at 512 × 512 resolution, with at least 5 representative photographs of each coverslip. The colocalization synapse analysis was performed by neuronal immunostaining of pre‐ and post‐synaptic markers (VGLUT1 and PSD95 respectively) and normalized to the total number of neurons (MAP2⁺) of the image, using the Puncta Analyzer plugin and protocol of analysis (Ippolito and Eroglu 2010), with Fiji ImageJ software.

RNA was extracted from cell cultures using a column extraction kit (Cytiva, 25050071) and subsequently quantified using a NanoDrop spectrophotometer. cDNA was obtained using the PrimeScript RT Reagent Kit (Takara, RR037A), following the manufacturer's instructions. The qPCR reaction was performed using the TB Green Premix Ex Taq Kit (Takara, RR82LR) and a QuantStudio 5 thermal cycler (Applied Biosystems). The specific forward and reverse oligonucleotides were as follows:(5′–3′): Lmnb1: (F) CAACTGACCTCATCTGGAAGAAC, (R) TGAAGACTGTGCTTCTCTGAGC; Il1β: (F) CAGGCAGGCAGTATCACTCA, (R) TAATGGGAACGTCACACACC; Il6: (F) CAAAGCCAGAGTCCTTCAGAG, (R) TGGTCCTTAGCCACTCCTTC; Cdkn1a: (F) GGCAGACCAGCCTGACAGAT, (R) TTCAGGGTTTTCTCTTGCAGAAG; Cacna2d1: (F) CCAAATCTTCAGCCAAAGGAGC, (R) ATTGACAGGCGTCCATGTGT. mRNA expression levels were calculated using the 2‐ΔΔCT method, according to Livak and Schmittgen (2001). Sdha was used as a housekeeping gene for normalization: (F) AGAGGACAACTGGAGATGGCATT, (R) AACTTGAGGCTCTGTCCACCAA.

Free glutamate in the culture medium from NSCs and SAMR1 or SAMP8 Diff‐Astrocytes was quantified using the commercial Glutamate Assay kit, following the manufacturer's instructions (Sigma‐Aldrich, MAK004). Glutamate levels were normalized with an MTT (3‐(4, 5‐dimethylthiazolyl‐2)‐2, 5‐diphenyltetrazolium bromide) assay for cellular viability.

TSP‐1 levels were determined in 10‐m SAMR1 and SAMP8 hippocampi. Tissue lysates were obtained by diluting hippocampi with cold phosphate‐buffered saline (PBS) supplemented with protease inhibitors (Roche) and subjecting the samples to freeze–thaw cycles and to mechanical disruption with a homogenizer. The lysates were centrifuged at 5000 g for 10 min at 4°C, and the supernatants were collected and used. TSP‐1 protein measurements were determined using the Mouse Thrombospondin‐1 ELISA kit (CliniSciences, orb1807782‐48) following the manufacturer's instructions. Sample protein content normalization was based on total protein concentration determined for each sample by Bradford assay.

SAMR1 and SAMP8 Diff‐Astrocytes were washed with PBS and homogenized in lysis buffer, supplemented with a protease inhibitor cocktail (Roche). Cell lysates were centrifuged at 13200 rpm for 15 min at 4°C, and supernatant protein concentration was determined in a Bradford assay (Pierce). 50 μg of protein lysates were loaded and resolved on sodium dodecyl sulfate‐polyacrylamide gel (SDS‐PAGE), and transferred to a nitrocellulose membrane (Amersham). The membrane was stained with Ponceau for total protein determination and blocked in 5% skim milk. The membrane was incubated overnight with anti‐TSP1/2 (1:200, sc‐133061, Santa Cruz Biotechnology) and anti‐β‐Actin (1:5000, A‐5441, Sigma‐Aldrich) primary antibodies. The anti‐TSP1/2 primary antibody cross‐reacts with TSP1 and TSP2, which are structurally and functionally related and form a distinct subfamily of thrombospondins. IRDye 680LT anti‐mouse (1:5000, Licor, 925‐68020) secondary antibody was used. Proteins were detected with LI‐COR Odyssey and analyzed with Image Studio Lite.

At least three different male SAMR1 and SAMP8 mice were used in all experiments. For in vitro culture assays, at least three independent experiments were performed. Error bars represent the standard error of the mean (SEM). All statistical analyses were performed using GraphPad Prism software version 10. Normality was assessed with Shapiro–Wilk test. Two‐way ANOVA, one‐way ANOVA, one‐sample t‐test (for RT‐qPCR), and Student's t‐tests were performed. In all analyses, a p < 0.05 was considered significant. Probabilities are represented as: * p < 0.05, ** p < 0.01, *** p < 0.001, and **** p < 0.0001. Statistical analyses are indicated in the figure legends.

Previously, SA‐β‐gal staining has been found to increase in the hippocampus of aged rodents, being reported in pyramidal neurons of the CA3 and CA1 hippocampal regions (Geng et al. 2010; Piechota et al. 2016). However, the age‐related accumulation of senescent SA‐β‐gal⁺ astrocytes throughout the hippocampus has been poorly explored. To evaluate the impact of astrocyte senescence on hippocampal function, SA‐β‐gal signal intensity was measured in brain sections of 2‐, 6‐, and 10‐month‐old (2, 6, and 10‐m) SAMR1 and SAMP8 animals. We evaluated both the entire hippocampus and a selection of hippocampal layers with high astrocyte content, according to the percentage of cells expressing the astrocyte marker GFAP (Glial Fibrillary Acidic Protein) (Figure 1A,B). When comparing animals of the same age, SA‐β‐gal signal intensity in the entire hippocampus of the SAMP8 strain increased significantly compared to the SAMR1 control strain (Figure 1C,D, p < 0.0001). SA‐β‐gal signal intensity also increased from 6 to 10‐m in both strains (Figure 1D). At an early age (2‐m), all the evaluated astrocyte‐enriched hippocampal layers (Stratum Oriens–SO, Stratum Radiatum–SR, Stratum Lacunosum Moleculare–SLM, Molecular Layer–MO, and Polymorphic Layer–PO) showed signs of senescence in SAMP8 versus SAMR1. For most layers, at 6 and 10‐m SA‐β‐gal signal intensity remained higher in SAMP8. Two‐way ANOVA analysis indicated that the effect of the strain was significant for the entire hippocampus and for all the analyzed layers, with an interaction between strain and age in the SR stratum (p < 0.05). The combination of SA‐β‐gal staining with immunofluorescence detection of the mature astrocyte markers S100β (S100 calcium‐binding protein beta) and GFAP confirmed the presence of SA‐β‐gal⁺ astrocytes (Figures 1E and S1). Together, these data allow us to conclude that the SA‐β‐gal senescence mark is precociously increased in SAMP8 hippocampal layers with high astrocyte content compared to SAMR1 and is detected in GFAP⁺S100β⁺ astrocytes.

Astrocytes are key regulators of neuronal synapse formation, maintenance, and function, so we next designed experiments to explore the impact of hippocampal senescent astrocytes on hippocampal neuronal synaptogenesis. To model astrocyte senescence, we developed a simple and efficient in vitro method using hippocampal neural stem cells (NSCs) isolated from 2‐m SAMR1 and SAMP8 animals (Figure 2A). We employed Neurobasal medium supplemented with serum to stimulate astrocyte differentiation (McCarthy and De Vellis 1980). Expression of the astrocyte markers GLAST (Glutamate and aspartate transporter), ATP1B2 (ATPase Na+/K+ transporting subunit beta 2), S100β, and GFAP was monitored in the cultures (Figure S2). Quantifications showed that nearly 100% of the differentiated SAMR1 and SAMP8 cells were GFAP⁺S100β⁺ and GLAST⁺ATP1B2⁺ at 8 days in vitro (DIV), in all independent assays performed (Figure S2). Consistently, expression of the intermediate filament Nestin, a NSC and progenitor cell marker, was reduced upon differentiation (Figure S2). We also verified the acquisition of relevant astrocyte functions such as glutamate uptake, following incubation of the cells with 100 μM glutamate in HBSS. For both strains, the presence of free glutamate in the media was reduced in differentiated astrocytes compared to NSCs, indicating an increased reuptake upon differentiation (Figure S2). Thus, we concluded that astrocytes differentiated in vitro from hippocampal NSCs (Diff‐Astrocytes hereafter) acquire mature astrocytic properties.

We next compared the SA‐β‐gal senescence mark in SAMR1 and SAMP8 NSCs and Diff‐Astrocytes and quantified the percentage of positive cells. As expected, SAMP8 NSC and SAMP8 Diff‐Astrocyte cultures showed a higher proportion of SA‐β‐gal⁺ cells compared to control SAMR1 cultures (p < 0.0001; Figure 2B,C). Interestingly, SAMP8 NSCs already displayed this senescence mark compared to SAMR1 NSCs, but the difference was greatly exacerbated upon NSC differentiation into astrocytes. Reduced Lamin B1 expression (p < 0.05; Figure 2D) and increased p16^INK4a protein levels (p < 0.05; Figure 2E,F), two other hallmarks of cellular senescence (Matias et al. 2021), were also detected in SAMP8 Diff‐Astrocytes compared to SAMR1 Diff‐Astrocytes by RT‐qPCR and immunofluorescence, respectively. Expression of Cdkn1a/p21^CIP1 and SASP genes further supported the acquisition of senescence (Figure S2).

Next, we evaluated if senescence influenced the synaptogenic capacity of the Diff‐Astrocyte cultures. To this end, we established primary hippocampal neuronal cultures from wild‐type embryos (E18.5) and tested the effect of astrocyte conditioned media (ACM) collected from SAMR1 and SAMP8 Diff‐Astrocytes (Figure 2A). Excitatory synapse formation was evaluated based on the immunofluorescence and co‐localization image analysis of the pre‐synaptic vesicle marker VGLUT1 (Vesicular Glutamate Transporter 1) and the post‐synaptic marker PSD95 (Post‐Synaptic Density protein 95), as previously described (Matias et al. 2021). ACM collected from SAMR1 Diff‐Astrocytes promoted co‐localization of the pre‐ and post‐synaptic markers, demonstrating a positive impact on the generation of new synapses compared to control medium (p < 0.05; Figure 2G,H). Most importantly, ACM from SAMP8 Diff‐Astrocyte cultures had no effect, demonstrating a loss of synaptogenic function in senescent astrocytes. Potential neurotoxic effects did not appear to drive the synaptic density changes (Figure S3A). Altogether, these results indicate that astrocytes differentiated in vitro from hippocampal SAMP8 NSCs enter senescence (as shown by the rise in SA‐β‐gal, p16^INK4a and p21^CIP1, and the reduction in Lamin B1) and are deficient in promoting synaptogenesis compared to control (SAMR1) astrocytes, providing a suitable model to study the impact of astrocyte senescence in astrocyte‐neuron communication.

To further corroborate these findings, we next isolated astrocytes directly from the adult hippocampus of 6‐m SAMR1 and SAMP8 mice employing magnetic cell separation (MACS), taking advantage of the astrocyte cell surface marker ACSA‐2 (encoded by Atp1b2) (Figure 3A). We analyzed the acutely isolated ACSA‐2⁺ cells by RT‐qPCR for senescence and SASP markers. Lamin B1 expression was significantly decreased (p < 0.05; Figure 3B), while IL‐1β expression was upregulated (p < 0.05; Figure 3B) in ACSA‐2⁺ SAMP8 cells compared to ACSA‐2⁺ SAMR1 cells. Hippocampal ACSA‐2⁺ cells were then cultured in vitro in defined astrocytic growth medium (without serum) and were analyzed by immunocytochemistry, with specific astrocyte markers and SA‐β‐gal staining (Figure 3C). The vast majority of the hippocampal ACSA‐2⁺ isolated cells expressed astrocyte markers upon culture and were double‐positive for GLAST and ATP1B2 (average GLAST⁺ATP1B2⁺ cells ± S.E.M: 96.7% ± 0.7% in SAMR1 cultures and 93.3% ± 2.0% in SAMP8 cultures, n = 3). On average, 9.3% ± 4.5% of the GLAST⁺ATP1B2⁺ astrocytes were SA‐β‐gal⁺ in SAMR1 primary astrocyte cultures, and this percentage raised 3‐fold in SAMP8 primary astrocyte cultures (33.2% ± 4.1%, p < 0.05; Figure 3C,D). Similar results were obtained when SAMR1 and SAMP8 astrocytes were isolated and cultured from 2 and 10‐m animals (Figure S1).

Next, we evaluated the synaptogenic activity of ACM collected from primary ACSA‐2⁺ astrocytes, based on the co‐localization of VGLUT1 and PSD95 in ACM‐treated hippocampal neuronal cultures (Figure 3E,F). SAMR1 ACSA‐2⁺ ACM significantly promoted synapse formation compared to non‐conditioned control medium (p < 0.05; Figure 3F). In contrast, no synaptogenic effect was found when neurons were treated with SAMP8 ACSA‐2⁺ ACM (Figure 3F). One‐way ANOVA revealed no significant differences in neuronal density across conditions, suggesting that neurotoxic effects were not the primary driver of the observed reduction in synaptic density in the senescent SAMP8 medium (Figure S3B).

In summary, these results show that ACSA‐2⁺ hippocampal astrocytes isolated from SAMP8 animals show senescence marks and are deficient in synaptogenic activity. While primary SAMR1 hippocampal astrocytes release factors into the ACM that promote the formation of new excitatory glutamatergic synapses in hippocampal neuronal cultures, this capacity is lost in primary SAMP8 hippocampal astrocytes, in line with the results obtained with senescent astrocytes differentiated from SAMP8 NSCs.

We then followed a candidate approach to uncover the identity of putative factors repressed in hippocampal senescent astrocytes that could account for the loss of synaptogenic function. Healthy brain astrocytes release molecules to the extracellular space that contribute to the formation of synapses (Ullian et al. 2004). Thrombospondin‐1 (TSP‐1, encoded by the Thbs1 gene) has emerged as an astrocyte‐secreted protein with prominent effects on synaptogenesis and neuritogenesis during development (Cheng et al. 2016; Ikeda et al. 2010; Yu et al. 2008). We explored available bulk RNAseq databases in silico and confirmed the enriched expression of Thbs1 in astrocytes among all other major cell types of the murine brain (Zhang et al. 2014) (Figure 4A). Inspection of an RNAseq dataset from hippocampal astrocytes of postnatal, young, middle‐age, and old wild‐type animals highlighted the downregulation of Thbs1 expression during aging (Clarke et al. 2018) (Figure 4B). Given Thbs1 gene expression is also reduced in astrocytes isolated from other brain areas during aging (Boisvert et al. 2018), we decided to delve deeper into Thbs1/TSP‐1 expression in senescent astrocytes.

We evaluated Thbs1 mRNA and TSP‐1 protein levels in senescent SAMP8 hippocampal astrocytes by RT‐qPCR and immunocytochemistry. Our results demonstrate that primary SAMP8 ACSA‐2⁺ astrocytes isolated from the hippocampus show significantly lower TSP‐1 protein levels compared to SAMR1 ACSA‐2⁺ astrocytes (p < 0.05; Figure 4C,D). On the other hand, SAMP8 Diff‐Astrocyte cultures also show decreased TSP‐1 protein levels, as quantified by immunofluorescence (p < 0.05; Figure 4C,E) and Western blot (Figure S3), and reduced Thbs1 gene expression compared to SAMR1 Diff‐Astrocytes (p < 0.01; Figure 4F).

In order to further explore Thbs1/TSP‐1 expression in the SAMP8 hippocampus in vivo, we measured Thbs1 mRNA levels by RT‐qPCR (Figure 5A) and total TSP‐1 protein levels by ELISA (Figure 5B) in hippocampal tissue extracts from 10‐m animals. Results confirmed the reduction in total Thbs1 mRNA and TSP‐1 protein content in SAMP8 hippocampus relative to SAMR1. This was further validated by immunofluorescence in hippocampal sections from SAMP8 relative to SAMR1 animals (Figure 5C), where TSP‐1 signal was found both co‐localizing with the astrocytic marker GFAP and in the extracellular matrix surrounding the astrocytic processes. Confocal image analysis showed decreased TSP‐1 levels in most astrocyte‐enriched layers of the 10‐m SAMP8 hippocampus compared to SAMR1 (Figure 5D). Finally, we also analyzed VGLUT1 and PSD95 co‐localization in 10‐m SAMP8 and SAMR1 hippocampal sections. As shown in Figure S4, we detected a significant decrease in co‐localized VGLUT1 and PSD95 excitatory synaptic puncta in most analyzed SAMP8 hippocampal layers, and a significant correlation between TSP‐1 signal intensity and puncta co‐localization (Figure S4). These data support the notion that loss of TSP‐1 expression in hippocampal senescent astrocytes may contribute to the reduction in excitatory synapses in SAMP8 animals.

We next designed a functional assay to explore whether the synaptogenic effect induced by SAMR1 astrocytes is related to TSP‐1 secretion. For this purpose, we employed Gabapentin (GBP), a pharmacological antagonist that blocks TSP‐1 signaling through its neuronal receptor, the voltage‐gated calcium channel (VGCC) subunit α2δ‐1 (Cheng et al. 2016; Eroglu et al. 2009). The α2δ‐1 receptor is encoded by the Cacna2d1 gene, which is highly expressed in neurons throughout the CNS. The α2δ‐1 receptor is enriched in hippocampal pyramidal neurons (Cole et al. 2005), and as shown by RT‐qPCR, is expressed in the hippocampal neurons used in our cell culture assays (Figure S5). As shown in Figure 6, GBP completely blocked the pro‐synaptogenic effect of SAMR1 Diff‐Astrocyte ACM (p < 0.01; Figure 6A,C). Similarly, GBP inhibited the positive effect of the SAMR1 ACSA‐2⁺ ACM on the formation of new synapses (p < 0.05; Figure 6B,D). No significant differences were found in GBP‐treated cultures exposed to SAMP8 Diff‐Astrocyte ACM or SAMP8 ACSA‐2⁺ ACM.

In summary, these results allow us to establish that the positive effect on the formation of new excitatory synapses triggered by healthy SAMR1 Diff‐Astrocytes and SAMR1 ACSA‐2⁺ primary hippocampal astrocytes is related to the presence of high TSP‐1 levels in the ACM, which signals through the neuronal receptor α2δ‐1.

Finally, we explored whether in vitro treatment with purified recombinant TSP‐1 restored the loss of synaptogenic activity of senescent SAMP8 astrocytes. Supplementing SAMP8 Diff‐Astrocyte ACM with TSP‐1 was sufficient to recover glutamatergic excitatory synapse formation in primary hippocampal neuronal cultures (p < 0.05; Figure 7A,B). As a control, we further demonstrated that the pharmacological antagonist GBP blocked the TSP‐1 effect on the co‐localization of VGLUT1 and PSD95 pre‐ and post‐synaptic markers (Figure S5). Overexpression of the mouse Thbs1 gene in SAMP8 differentiated astrocytes (Figure S5), followed by treatment of neuronal cultures with the resulting conditioned medium (Figure 7C,D) showed that restoring TSP‐1 expression suffices to rescue synaptogenic activity in senescent SAMP8 astrocytes.