Previous reports have established the critical role of astrocytes in synapse formation [16], [17], [19]–[21] and regulation of dendritic spines [10]–[13]. However the role of astrocytes in DS spine pathology has not been investigated. Our previous results indicate the presence of mitochondrial dysfunction in DS astrocytes leading to alterations in protein secretion and intracellular Aβ accumulation [22]. To establish whether deficits in DS astrocyte function could be involved in spine pathology, rat hippocampal neurons were cultured on top of NL or DS astrocyte monolayers (Figure 1A). Under these conditions, rat hippocampal neurons survived well for extended periods of time and developed fast-growing axons and dendrites (Figure 1A). No differences in neuronal survival were observed between regular rat hippocampal cultures and rat hippocampal/human astrocyte cocultures (Figure S1). After 21 DIV, neurons developed numerous spines (Figure 1B) exhibiting characteristic shapes including stubby-, mushroom-, thin- and filopodium-like spines [23]. Stubby spines were especially abundant (Figure 1B). In contrast, human cortical neurons exhibited poor spine development in culture (data not shown), precluding their use for this study. Thus, to assess the capacity of DS astrocytes to sustain spine and synapse formation we utilized human cortical astrocyte/rat hippocampal neuron cocultures.

We used immunofluorescence to analyze spine formation in the cocultures at 21 DIV. For simplicity, we assessed both stubby- and mushroom-like spines as “stubby”, and both thin- and filopodium-like spines as “filopodium” spines. We found a significant reduction in the total number of spines in neurons growing on top of DS astrocytes compared to neurons growing on top of NL astrocytes (NL: 9.1±0.5 per 50 µm of dendrite; DS: 7.1±0.6 per 50 µm of dendrite)(Figure 2A and B), indicating that DS astrocytes are less efficient than normal astrocytes in supporting dendritic spine formation. The reduction in spine density affected specifically stubby spines (NL: 4.6±0.5 stubby spines per 50 µm dendrite; DS: 2.1±0.3 stubby spines per 50 µm dendrite) (Figure 2C). Conversely, in DS astrocyte/hippocampal neuron cocultures we observed a significant increase in the number and length of filopodium spines (Figure 2A, D and E), which previous studies have characterized as immature spines in hippocampal neurons [24]. Similar alterations in spine number and structure are present in DS brains [2], [3], [25], suggesting that astrocyte dysfunction contributes to spine pathology in DS.

Soluble factors released by astrocytes such as cholesterol or thrombospondins have been shown to modulate spine formation and the establishment and maintenance of synaptic contacts [19], [20]. We assessed the effect of astrocyte-released factors on spine formation. Pure hippocampal cultures grown on polylysine and maintained with astrocyte conditioned medium (CM) exhibited somewhat lower spine density than their counterparts grown on top of astrocyte monolayers, indicating that physical contact with astrocytes is also relevant for spine formation (Figure 2B). Hippocampal neurons maintained in DS astrocyte CM showed a marked reduction in the number of spines compared to neurons maintained in NL astrocyte CM (NL CM: 6.2±0.5 spines per 50 µm of dendrite; DS CM: 3.6±0.3 spines per 50 µm of dendrite)(Figure 2B). Thus, soluble factors are critically involved in DS astrocyte-mediated spine alterations.

TSP-1 is an extracellular matrix protein synthesized and released by astrocytes [26], which is highly expressed during development of the nervous system [27]. It promotes neurite outgrowth and survival [28]–[30], neuronal migration [31], [32] and synaptogenesis [20]. Recently, it has been demonstrated that TSP-1 is necessary for synaptic and motor recovery after stroke [33], suggesting that TSP-1 also participates in neuronal plasticity. Given that TSP-1 is one of the most abundant proteins secreted by human astrocytes (data not shown), and that there are marked deficits in protein secretion in DS astrocytes [22], we next analyzed TSP-1 expression and secretion in DS astrocyte cultures and brain homogenates. We found high TSP-1 expression in human astrocytes and prominent surface localization (Figure 3A-D). TSP-1 subcellular localization and staining pattern was similar in normal and DS astrocytes. Quantification of TSP-1 protein levels by ELISA demonstrated significant reductions in both DS astrocyte cell lysates and CM (Figure 3E), suggesting that reduced TSP-1 expression leads to decreased TSP-1 in DS CM. TSP-1 expression was considerably higher in fetal than in adult brain homogenates (data not shown), consistent with a significant role of TSP-1 during development [27]. A comparison of TSP-1 protein expression in fetal brain lysates showed an average reduction of 57.9% in TSP-1 levels in DS brains despite the average age of DS brain samples being almost 4 weeks younger than the average age of NL brain samples (Figure 3F). qPCR analysis of mRNA expression indicated similar mRNA levels in DS brains than in NL brains for all thrombospondin isoforms (TSP-1, -2, -3 and -4)(Figure S2), suggesting that a post transcriptional mechanism underlies the changes observed in TSP-1 in DS brains and astrocyte cultures.

We assessed the effect of TSP-1 on spine development in pure hippocampal cultures. Recombinant TSP-1 was added to the culture medium at day 7, and fresh TSP-1 was replenished every 3 days for 14 days. We used a TSP-1 dose of 250 ng/ml because it is in the range of the difference in the level of TSP-1 in CM between NL and DS astrocyte cultures (Figure 3E). TSP-1 increased significantly the number of spines compared to vehicle-treated cultures (control: 6.5±0.7 spines per 50 µm dendrite; TSP-1: 9.0±0.7 spines per 50 µm dendrite)(Figure 4A and B). The effect of TSP-1 on spine number was similar to the effect of BDNF, a well known inducer of spine development in hippocampal neurons [34], [35]. Combined treatment of BDNF+TSP-1 had an additive effect on spine development, suggesting that the two factors act through non-overlapping, complementary signaling pathways (Figure 4A and B). A dose-response assay indicated that concentrations of TSP-1 between 250 ng/ml to 1250 ng/ml produced a gradual and marked increase in the number of dendritic spines in pure neuronal cultures (Figure 4C). Thus, TSP-1 is a strong promoter of spine development.

To determine whether depletion of TSP-1 is sufficient to alter spine number and morphology, NL astrocyte/hippocampal neuron cocultures were treated with anti-TSP-1 antibody to neutralize the endogenous TSP-1 secreted by NL human astrocytes. At day 7, anti-TSP-1 was added to the cultures and replenished every 3 days during 14 days. Hippocampal neurons treated with anti-TSP-1 exhibited dramatic changes in spine morphology, including a marked reduction in the density of stubby spines, and a major increase in the frequency and length of filopodium spines (Figure 5 and Figure S3). Anti-TSP-1 previously neutralized with excess TSP-1 and non-immune IgG had no effect on spine morphology (data not shown). Conversely, addition of recombinant TSP-1 to hippocampal neurons growing on top of DS astrocytes exhibited a significant reduction of filopodium spines (Figure 6A), and a marked increase in the number of stubby spines (Figure 6B). A similar effect on spines, although of slightly smaller magnitude, was observed after the addition of BDNF to the cocultures (data not shown). Heat-inactivated TSP-1 did not have any effect on spine morphology or number (Figure 6A and B). These results demonstrate that soluble TSP-1 modulates spine morphology and that reduced expression of TSP-1 in DS astrocytes leads to abnormal spine development.

To directly address the role of TSP-1 on the regulation of spine formation, rat hippocampal neurons were grown on top of astrocyte monolayers derived from wild type (WT) or TSP-1 KO mice [36]. There were no significant differences in general morphology, viability, between both cocultures (Figure 7A). However, neurons growing on top of TSP-1 KO astrocytes exhibited a striking increase in the number of filopodium spines and a marked reduction in the number of stubby spines (Figure 7B and C). Treatment of TSP-1 KO cocultures with recombinant TSP-1 for 4 days reduced significantly the number of filopodium spines and increased the number of stubby to similar levels than WT cocultures (Figure 7C). The effect of TSP-1 was more pronounced after 7 days of treatment (Figure 7C). The spine phenotype in TSP-1 KO cocultures is reminiscent to the alterations observed in DS cocultures, providing direct evidence for a role of TSP-1 on spine formation.

We investigated the effect of spine alterations and TSP-1 deficits on synapse formation and activity in DS cocultures. Synaptic density was assessed as the frequency of colocalization of the pre- and post-synaptic markers synaptophysin and PSD95 at 21 DIV [37]–[39]. Consistent with the alterations in spine morphology and density, we found a significant reduction in the number of synaptic contacts (NL: 12.1±1.1 synaptophysin/PSD95 colocalized puncta per 50 µm of dendrite; DS: 5.0±0.5% synaptophysin/PSD95 colocalized puncta per 50 µm of dendrite)(Figure 8A and B). To determine the number of synapses specifically localized in spines (Figure 8C), we measured the frequency of colocalization of synaptophysin, PSD95 and drebrin, (Figure 8D and Figure S4). Of the total number of synapses in the culture, represented by synaptophysin/PSD95 colocalization (Figure 8A and B), more than 50% were localized at dendritic spines, represented by synaptophysin/PSD95/drebrin colocalization (Figure 8C and D), indicating that spines are principal sites of synaptic formation in the cocultures. The number of synapses localized at spines was also markedly reduced in DS cocultures (Figure 7D). To study the presence of functional synapses, we assessed vesicular uptake using the fluorescent cationic styryl fixable dye AM4-64. Depolarization induced by 20 mM KCl causes rapid vesicular uptake of A4-64, allowing the visualization of recycled vesicles [37], which in this case was used as an indicator of active synapses. Depolarization induced massive vesicle recycling in both NL and DS cocultures, indicating the presence of a high number of functional synapses in the cocultures (Figure 8E). However, there was a significantly lower density of A4-64-positive vesicles in DS cocultures after stimulation with KCl (Figure 8F). These findings demonstrate that the total number of synapses, the number of synapses at spines, and the number of active synapses are all reduced in DS cocultures.

The ability of TSP-1 to modulate spine formation (Figure 4), to rescue spine defects associated with DS and TSP-1 KO cocultures (Figure 6 and 7), and its capacity to stimulate synaptogenesis [20], prompted us to examine the effect of TSP-1 treatment on synapse formation in neurons growing on top of DS astrocytes. Addition of 250 ng/ml TSP-1 resulted in a major increase in both total synaptic density (Figure 8B) and in the number of synapses localized at spines (Figure 8D). Restoration of TSP-1 levels also increased synaptic activity in DS cocultures as measured by A4-64 uptake (Figure 8E and F).

Protein synthesis and secretory function are highly dependent on mitochondrial energy production and are particularly vulnerable to the effects of intracellular oxidative damage. For example, APP secretory deficits in DS astrocytes are directly linked to impaired mitochondrial metabolism and abnormal secretory function [22], and increased apoptosis of DS neurons can be prevented by antioxidants [40]. To establish whether reduced expression of TSP-1 in DS astrocytes was associated with oxidative stress and/or energy deficits in DS cells, we analyzed the effect of free radical scavengers, antioxidants, and energy substrates on TSP-1 expression. DS astrocytes were treated with sPBN (100 µM), trolox (100 µM), resveratrol (100 µM), nicotinamide (15 mM), nicotinamide adenine dinucleotide (β–NAD, 15 mM), creatine (5 mM) or glucose (5 mM), and TSP-1 levels were quantified 24 hr later. None of the compounds tested were able to modify TSP-1 levels inside the cells or in the CM, suggesting that neither mitochondrial dysfunction nor oxidative stress are primary factors mediating TSP-1 deficits in DS astrocytes (Figure S5).

This study is part of an ongoing research protocol approved by the Health and Hospital Corporation of the City of New York, the Albert Einstein College of Medicine Committee on Clinical Investigation and the Internal Review Board of the University of California-Irvine. The protocols for obtaining post-mortem fetal brain tissue comply with all federal and institutional guidelines with special respect for the confidentiality of the donor's identity. Written informed consent was obtained from all tissue donors.

All procedures for the generation of rat hippocampal neuronal suspensions were reviewed and approved by the Institutional Care and Use Committee (IACUC) at the University of California Irvine, protocol number 2008-2779.

Astrocyte cultures were established from postmortem NL and DS human fetal brain tissue samples as described [22]. Normal and DS fetal human brain samples are procured at the Human Fetal Tissue Repository, Albert Einstein School of Medicine (AECOM), NY, and received and processed in the Busciglio laboratory at UCI. Cortical tissue samples were dissociated into a single-cell suspension by incubation with 0.25% trypsin/40 µg/ml DNAse (Sigma, St Louis, MO) in PBS at 37°C for 30 min, and mechanically dissociated with a fire-polished glass Pasteur pipette. Cells were plated on the bottom of tissue culture dishes or on glass coverslips, and maintained in DMEM (Invitrogen, Grand Island, NY) supplemented with 10% fetal bovine serum (HyClone, Road Logan, UT). After growing to confluence, cells were subjected to two passages to generate pure astrocytic cultures. The identity and purity of the astrocyte cultures was confirmed by immunocytochemistry with anti-GFAP and anti-S100 antibodies. To generate the cocultures, hippocampi were dissected from rat newborn pups, incubated with trypsin, and triturated through a glass Pasteur pipette as described by Kaech and Banker [112]. The neurons were plated on top of NL or DS astrocyte monolayers at 100,000 cells/mm². Two hr after neuronal plating, the medium was changed to Neurobasal plus N2 and B27 supplements (Invitrogen, Grand Island, NY). Partial medium changes were performed every 5 days. Cultures were maintained for 21 days to allow for the development and maturation of dendritic spines. Under these conditions, there is minimal growth of rat astrocytes in the coculture (Figure S6). Pure hippocampal cultures were maintained in NL or DS astrocyte conditioned medium (CM) obtained from astrocyte cultures. For pure hippocampal cultures, the neurons were plated on coverslips pretreated with 1 mg/ml poly-L-lysine immediately after dissociation (Sigma, St Louis, MO). Two hr after plating, the medium was switched to NL or DS CM. Partial medium changes with fresh CM were performed every 3 days. For some experiments, astrocyte monloayers were generated from the cortex of WT and TSP-1 KO [36] newborn pups following the procedures described above.

Cultures were fixed with 4% paraformaldehyde/0.12 M sucrose/PBS for 15 min at 37°C, permeabilized 15 min with 0.2% Triton X-100/PBS, and blocked for 30 min in 5% bovine serum albumin/PBS. Then, the cultures were incubated with one or more of the following primary antibodies: mouse anti-drebrin, a marker of dendritic spines (1∶250, Stressegen, Ann Arbor, MI), GFAP (1∶1000, Sigma, St Louis, MO), anti-β-tubulin isotype III (1∶1000, Sigma, St Louis, MO), mouse anti-thrombospondin-1 (1∶250 Calbiochem, La Jolla, CA), anti-synaptophysin (1∶250, Calbiochem, Gibbstown, NJ), anti-PSD95 (1∶250, Abcam, Cambridge, MA), and anti S-100β (1∶500, R&D systems, Minneapolis, MN), for 1 hr at RT, followed by a 30 min incubation in fluorescent-conjugated secondary antibody (1∶500, Alexa 350, Alexa 488 and Alexa 594, Invitrogen, Eugene, OR). An Axiovert 200 inverted microscope (Zeiss, Jena, Germany) was used for examination, imaging and quantification of various parameters of spine morphology. Images were captured with a digital camera and processed using AxioVision software (Zeiss). When required, Z-stacks were captured at 500 nm intervals, processed and rendered using the Apotome imaging system (Zeiss). Spine-like protrusions were classified according to Hering and Sheng [23]. For spine assessment, dendritic spines were defined as protrusions from the surface of dendritic processes exhibiting a high concentration of drebrin [113], [114]. Morphologically, we identified 2 populations of spines: stubby-/mushroom-like spines, short, small and rounded spines with or without a short neck, and filopodium-like spines, which are typically longer, thin and lack a head. For quantification, 10 neurons per culture were selected from 3–6 separate cultures. Density and morphology of spines were scored in dendritic segments approximately 50 µm away from the cell body. Spine length was measured using the measurement tool of AxioVision. Spine length was defined as the distance from the base of the neck to the tip of the spine. To assess synaptic density, the colocalization of pre- and post-synaptic markers was performed as described [37]–[39].

TSP-1 levels were determined in astrocyte cultures and fetal brain homogenates. Astrocytes were plated in 24-well plates. Cultures were washed in ice-cold PBS and collected in Eppendorf tubes in 0.1 M NaOH. The lysates were centrifuged at 14,000 rpm in an Eppendorf microcentrifuge for 30 min at 4°C, and the supernatants were used to determine the intracellular concentration of TSP-1. Aliquots of CM from the same cultures were processed to determine the level of secreted TSP-1. Samples of fetal cortex were homogenized in RIPA buffer plus protease inhibitors (Complete, Roche, Mannheim, Germany) with a Turrax homogenizer (Ika Works, Wilmington, NC), and centrifuged at 14,000 rpm for 30 min at 4°C. TSP-1 levels in the supernatant fraction were determined using a commercial ELISA kit following the vendor's instructions (R&D systems, Minneapolis, MN). For some experiments, the following compounds (purchased from Sigma, St Louis, MO) were added to pure astrocyte cultures, and TSP-1 levels were determined after 24 hr: s-PBN (100 mM), Trolox (100 µM), resveratrol (100 mM), nicotinamide (15 mM), nicotinamide adenine dinucleotide (β–NAD, 15 mM), creatine (5 mM) and glucose (5 mM).

Cocultures were plated in 24-well plates and human recombinant TSP-1 (R&D system Minneapolis, MN) was added every 3 days at the indicated concentrations. Control cultures were incubated with heat-inactivated TSP-1 (100°C for 5 min). Cocultures were fixed at 21 DIV and analysis of spines was performed as described above. For some experiments, pure hippocampal cultures were treated with TSP-1 (250 ng/ml), BDNF (10 ng/ml, PeproTech Inc, Rocky Hill, NJ) or with TSP-1+BDNF.

Cocultures were plated in 24-well plates. Starting at day 7, anti-TSP-1 (1∶1000 Calbiochem, Gibbstown, NJ) was added to the cultures, and fresh anti-TSP-1 was replenished with partial medium changes every 3 days. Control cultures were treated with anti-TSP-1 previously neutralized with an excess of recombinant human TSP-1 or with non-immune rabbit serum. Cocultures were fixed at 21 DIV to assess spine density and morphology.

To assess activity-dependent vesicle recycling, the cultures were incubated with the fixable fluorescent probe AM4-64 (Biotium Inc, Hayward, CA) for 5 min with or without 20 mM KCl [37]. AM4-64 fluorescent density was quantified using Axiovision.

All experiments were repeated three to six times using cultures derived from different brain specimens. Each individual experiment was performed at least in in triplicate samples. Data were analyzed by one-way analysis of variance (ANOVA) followed by Fisher's test. Results were expressed as the mean ± SEM. Significance was assessed at p<0.05. All results shown correspond to individual representative experiments.