Astrocytes restore connectivity and synchronization in dysfunctional cerebellar networks

Significance

Within the long-standing debate of whether glial cells contribute to neuronal circuits, we provide evidence of the impact of astrocytes on the physiopathology and the structural–functional organization of cerebellar neuronal circuits derived from Atm-deficient mice. In vitro and adult mouse characterizations show how disrupted astrocyte morphology is associated with an increased presence of synaptic markers in Atm^−/− compared with wild-type circuits, also related to reduce autophagy. Our study is corroborated by an in vitro demonstration of how replenishment of neuronal circuits with healthy astrocytes can restore Atm^−/− neuronal circuits’ dynamics.

Abstract

Evidence suggests that astrocytes play key roles in structural and functional organization of neuronal circuits. To understand how astrocytes influence the physiopathology of cerebellar circuits, we cultured cells from cerebella of mice that lack the ATM gene. Mutations in ATM are causative of the human cerebellar degenerative disease ataxia-telangiectasia. Cerebellar cultures grown from Atm^−/− mice had disrupted network synchronization, atrophied astrocytic arborizations, reduced autophagy levels, and higher numbers of synapses per neuron than wild-type cultures. Chimeric circuitries composed of wild-type astrocytes and Atm^−/− neurons were indistinguishable from wild-type cultures. Adult cerebellar characterizations confirmed disrupted astrocyte morphology, increased GABAergic synaptic markers, and reduced autophagy in Atm^−/− compared with wild-type mice. These results indicate that astrocytes can impact neuronal circuits at levels ranging from synaptic expression to global dynamics.

Brain degenerative diseases (BDDs) disrupt brain function by impairing the ability of specific neuronal circuitries to locally process (segregate) and distantly communicate (integrate) information across brain-spanning networks (1, 2). At the neuronal circuit level, BDDs induce the loss of different cell types and cellular functionality (2, 3) and alter circuit topology and dynamics. Currently, it is unclear how particular diseases explicitly alter specific brain circuits and neuro-glial populations.

In the last two decades, it has become appreciated that glial cells play a critical role in BDDs (4). The symptoms of BDDs arise from pathological changes to neuro-glia interactions (5), leading to neuronal cell death, disrupted neuro-glia communication, and impaired cell function (3), all of which affect global dynamics of brain circuitry. Astrocytes, a particular glial cell type, play key roles in regulating the pathophysiology of neuronal functions (6). Specifically, perisynaptic sheaths of astrocytes cover the majority of synapses in the central nervous system and are essential for synaptogenesis, maturation, and maintenance of synapses (7). This structural–functional unit, also called the “tripartite synapse,” enables astrocytes to influence synaptic communication via gliotransmission (5).

In this work, we tested the hypothesis that neuronal circuit dynamics were impacted as a consequence of disrupted neuron–astrocyte physiology in a mouse model of the BDD that results from a deficiency in Ataxia Telangiectasia Mutated (ATM) protein. The gene encoding ATM is mutated in the human genetic disease ataxia-telangiectasia (A-T) (8, 9). One of the most devastating symptoms of A-T is the cerebellar ataxia, with significant loss of Purkinje and granule neurons in the cerebellum, that leads progressively to general motor dysfunction (10).

We used primary cerebellar cultures grown from postnatal wild-type (WT) and Atm^−/− mice to study how Atm deficiency influences the structure and dynamics of cerebellar neuronal–astrocyte circuits. We hypothesized that Atm deficiency impairs the neuronal–astrocytic interactions underlying spontaneous neuronal synchronizations (NSs), a hallmark activity pattern of the developing nervous system that could be observed in many different circuits in various species and in in vitro preparations including primary neuronal cultures (11–13).

We report that the absence of Atm in neurons and astrocytes severely alters astrocyte morphology and the number of presynaptic and postsynaptic puncta, disrupting NSs within cerebellar networks. Higher numbers of synaptic puncta in Atm^−/− networks relative to numbers in WT cultures were associated with lower levels of autophagy. These reported structural and functional anomalies were all rescued in chimeric neuronal networks composed of Atm^−/− neurons and WT astrocytes. In contrast, cultures of WT neurons with Atm^−/− astrocytes led to significant neuronal cell death. Characterizations of adult Atm^−/− cerebella similarly showed disrupted astrocyte morphology, up-regulated GABAergic markers, and dysregulated mTOR-mediated signaling and autophagy. These results, obtained in an in vitro model system and confirmed in adult mice, demonstrate a possible role played by astrocytes in the physiopathological, structural, and functional organization of the cerebellar circuits.

Results

Absence of Atm Impairs in Vitro Cerebellar Neuronal Network Synchronizations.

All Atm^−/− or WT astrocyte and neuronal cultures/networks referred to in the text were derived from single animals. A total of 80% of the cells in the primary cerebellar cultures were granule neurons (NeuN-positive; n = 19) (14, 15). The majority of the remaining cells were astrocytes (GFAP- and Vimentin-positive), with a small portion of only Vimentin-positive cells (oligodendrocytes and glia precursors) (refs. 16–18 and SI Appendix, Fig. S1 A and B).

The spontaneous activities of cultured cerebellar networks derived from Atm^−/− and WT mice were monitored by using calcium-based imaging (Fig. 1). The calcium signals that originated from the neuronal activity were characterized by stereotypical sharp calcium transients (on the order of a few dozens of milliseconds, abolished by tetrodotoxin; n = 3) and were generated and recorded from granule cell bodies (isolated or organized in multicellular clusters, as shown in SI Appendix, Fig. S1 C–H). Extended wider segmentation of calcium images including astrocytic processes (SI Appendix, Figs. S1H and S2) confirmed the presence of fast calcium signals only in neuronal cells. Stereotypical slow calcium signals of astrocytes were rarely observed (and not included in the network analysis; Methods). The neuronal calcium signals were mediated by glutamatergic synaptic connectivity, since they were blocked by an AMPA receptor antagonist (2,3-dioxo-6-nitro-1,2,3,4-tetrahydrobenzo[f]quinoxaline-7-sulfonamide; n = 7) (SI Appendix, Fig. S1E, Right). Spontaneous NSs (Methods), a general common feature of primary cultures independent of the originating circuit (11–13), were defined as neuronal coactivations of at least two neurons and were observed in both Atm^−/− and WT cultures at similar frequencies of ∼0.07 Hz (Fig. 1E). Global synchronizations (GSs) that recruited nearly the entire neuronal population were significantly more frequent in the WT compared with the Atm^−/− cultures [P < 0.0001; Mann–Whitney U (MWU) test; Fig. 1E]. Conversely, sparse synchronizations (SSs), recruiting < 3% of the active neurons, were significantly more frequent in Atm^−/− compared with WT cultures; in WT cultures, SSs were very infrequent (Fig. 1E; P < 0.0001, MWU test).

Fig. 1.

NS impairment in Atm^−/− cerebellar networks. (A and B) Raster plots of neuronal activity for WT network (A) and Atm^−/− network (B). (C) Calcium signals from an Atm^−/− network displaying a SS (marked in green arrow) and followed by two GS events (marked in black arrows). The relative variation in fluorescence to the baseline level (ΔF/F) is reported. (D) Calcium dye loading of the Atm^−/− network. Cells marked in green are recruited in the SS reported in B. (E) Synchronization frequencies in WT and Atm^−/− cultures. Significant differences are marked by asterisks. ***P < 0.001 (MWU test; n = 43 networks from 19 WT animals, n = 43 networks from 21 Atm^−/− animals).

When analyzing the GS build-up, we observed that early activated cells repeatedly preceding the GS in WT and Atm^−/− networks represented 9 ± 1% (n = 28) and 13 ± 3% (n = 22) of the active neuronal population, respectively, with no significant difference between the two groups (SI Appendix, Fig. S3; P > 0.05, two-sample Kolmogorov–Smirnov test). Cells recruited in the SS (n = 2,615 from 22 Atm^−/− networks) were spatially closely located, and this localization was significant compared with randomly computationally generated SSs (SI Appendix, Fig. S4).

No cell showed higher significant participation in SS since recruitment frequencies were similar in the Atm^−/− networks and in randomly computationally generated sequences of SSs that preserved the recruited cellular populations (Kolmogorov–Smirnov test, P > 0.05; Methods).

Absence of Atm Reduces Morphological Complexity of Astrocytes.

Previous studies showed that astrocyte coverage of blood vessels is dramatically reduced in the retinas of Atm^−/− mice (10). We performed immunostaining to determine whether astrocyte complexity was also altered in Atm^−/− cerebellar cultures. Staining of primary astrocytes with GFAP revealed a less complex cell arborization (Fig. 2 A and B) and fewer processes originating from the cell bodies in Atm^−/− cultures than in WT cultures (P < 0.0001, MWU test). In addition, the mean total length of astrocyte processes in the entire astrocytic network (marked by GFAP) normalized to the number of the nonneuronal cells (i.e., DAPI–NeuN-negative cells) was significantly shorter in Atm^−/− compared with WT networks (Fig. 2C; P < 0.0001, t test).

Fig. 2.

Astrocytic processes atrophy in Atm^−/− cerebellar cultures. (A) Images of a single astrocyte stained with GFAP (green) in WT (Upper) and Atm^−/− (Lower) culture. The yellow dashed squares highlight the processes originating from the cell body. (B, Left) Immunocytochemical staining of neurons (NeuN; red), astrocytes (GFAP; green), and nuclei (DAPI; blue) in WT (Upper) and Atm^−/− (Lower) cultures. (B, Right) Automatic extraction of processes contours (red lines) from GFAP staining (gray scale). (C) Quantification of processes length (normalized to the pixel size) per astrocyte. Significant difference is indicated by asterisks. ***P < 0.001 (t test). n = 7 networks from four WT cultures, n = 7 networks from four Atm^−/− cultures.

WT Astrocytes Restore WT NSs in Atm^−/− Cerebellar Networks.

Next, we tested the hypothesis that impaired neuro–astrocyte interactions induced by Atm deficiency disrupted spontaneous NSs. We generated cerebellar chimeric cultures from two mice with half the cerebellum of each animal plated as a control and half used to create the chimeric network composed of neurons from one animal and astrocytes from another animal with a different genotype (SI Appendix, Fig. S5). The cellular type and origin in the chimeric cultures will be specified according to the following nomenclature: N_WT, A_WT, N_Atm^−/−, and A_Atm^−/−, where N and A refer to neurons and astrocytes, respectively, and subscripts WT and Atm^−/− refer to the genotype of the donor animal. To validate the procedure, we prepared chimeric cultures from two different animals with the same genotype. The chimeras showed no statistical differences in terms of synchronization compared with single-animal cultures (P > 0.05, MWU tests for GS, SS, and NS).

Immunostaining of neurons (NeuN) and astrocytes (GFAP) in [N_WT A_Atm^−/−] and [N_Atm^−/− A_WT] chimeras compared with single-animal WT and Atm^−/− cultures showed that the morphology of the astrocytes were not influenced by the presence of neurons from a different mouse and/or genotype [Fig. 3 A and B; P < 0.0001, Kruskal–Wallis (K-W) with Dunn’s multiple comparison (Dunn’s) test; SI Appendix, Fig. S6]. Notably, only a few neurons in [N_WT A_Atm^−/−] chimeric cultures survived (Fig. 3A).

Fig. 3.

WT astrocytes restore GSs in Atm^−/− cerebellar networks. (A) Immunocytochemical staining of the neurons (NeuN; red), astrocytes (GFAP; green), and nuclei (DAPI; blue). Half of the cerebellar cells from each mouse were plated as controls (Upper) and the other half as chimeric cultures (Lower). WT astrocytes are present in Left, whereas Atm^−/− astrocytes are present in Right, and in both cases they maintain their original morphology independently of the genotype of the cocultured neurons. (B) Number of astrocytic processes originating from the cell body in WT networks (n = 18; five cultures), [N_Atm^−/− A_WT] chimeric networks (n = 16; four cultures), Atm^−/− networks (n = 17; five cultures), and [N_WT A_Atm^−/−] chimeric networks (n = 8; three cultures) (P < 0.001). (C–E) Synchronization frequencies of NS (C; P > 0.05), GS (D; P < 0.001), and SS (E; P < 0.001) for the WT, Atm^−/−, and [N_WT A_Atm^−/−] chimeric networks. Note that the WT (n = 43; 19 animals) and the [N_Atm^−/− A_WT] chimeric (n = 11; six animals) networks are characterized by higher levels of GS and lower levels of SS events, whereas the Atm^−/− (n = 43; 21 animals) networks display low levels of GS and high levels of SS. Significant differences are marked by asterisks. **P < 0.0; ***P < 0.001 (K-W test followed by Dunn’s test).

We next analyzed neuronal dynamics in the WT, Atm^−/−, and chimeric cultures. The frequencies of NS events were similar in WT, Atm^−/−, and [N_Atm^−/− A_WT] chimeric cultures (Fig. 3C; K-W with Dunn’s test). In contrast, synchronizations were absent in the 3 of 10 [N_WT A_Atm^−/−] chimeric cultures that survived 17 d in vitro (DIV). [N_Atm^−/− A_WT] chimeric cultures were statistically similar to WT cultures in terms of GSs and SSs, compared with Atm^−/− networks (Fig. 3 D and E; P < 0.001, K-W with Dunn’s test).

Astrocytic Genotype Impacts Synaptic and Autophagy Levels in Cerebellar Networks.

Based on the known roles of astrocytes in the development, maturation, and dynamics of synapses (5, 7, 19, 20), we hypothesized that the impairment of neuronal dynamics in the cerebellar networks containing Atm^−/− astrocytes was due to disruption of neuronal connectivity. To test this hypothesis, we analyzed synaptic protein levels in Atm^−/− and WT cultures. We found that levels of both presynaptic [Synapsin1 (Syn1)] and postsynaptic (PSD-95) markers were significantly higher (68% and 58% increases, respectively) in the Atm^−/− cultures in comparison with WT cultures (Fig. 4 B and C; P < 0.01, MWU test). To evaluate the contribution of astrocytes to the neuronal circuit, we compared the levels of Syn1 and PSD-95 protein in Atm^−/−, [N_Atm^−/− A_WT] chimeric, and WT cultures: Syn1 (MWU test) and PSD-95 (MWU test) were expressed at similar levels (Fig. 4 D and E). Using immunocytochemistry, we further compared the synaptic termini in WT, Atm^−/−, and [N_Atm^−/− A_WT] chimeric cultures. In Atm^−/− cultures, we observed a higher average numbers of presynaptic puncta (Syn1) (Fig. 4 F and G; P < 0.001, K-W with Dunn’s test) and postsynaptic puncta (PSD-95) (Fig. 4 H and I; P < 0.05, K-W with Dunn’s test) per neuron in comparison with the WT and [N_Atm^−/− A_WT] chimeric cultures.

Fig. 4.

Reduced autophagy leads to increased levels of Syn1 and PSD-95 in Atm^−/− cultures in comparison with WT and [N_Atm^−/− A_WT] chimeric cultures. (A) Representative immunoblot (nine independent cultures) of Syn1 and PSD-95 at 16 DIV of WT, Atm^−/−, and [N_Atm^−/− A_WT] chimeric cultures. (B) Fold difference in Syn1 levels in Atm^−/− (n = 8) compared with WT (n = 8) cultures. (C) Fold difference in PSD-95 levels in Atm^−/− (n = 6) relative to WT (n = 6) cultures. (D) Fold difference in Syn1 levels in [N_Atm^−/− A_WT] chimeric (n = 4) compared with WT (n = 8) cultures. (E) Fold difference in PSD-95 levels in [N_Atm^−/− A_WT] chimeric (n = 4) relative to those in WT (n = 6) cultures. (F) Representative images of cerebellar granule cultures of specified genotypes stained for presynaptic puncta (Syn1; green), neurons (MAP2 red + NeuN; magenta), and nuclei (DAPI; blue). (G) Average number of presynaptic puncta (Syn1) per neuron. Atm^−/− networks (n = 31; six cultures) have a higher number of presynaptic puncta than the WT (n = 27; five cultures) and [N_Atm^−/− A_WT] chimeric networks (n = 19; three cultures; P < 0.001, K-W test followed by Dunn’s test). (H) Representative images of cerebellar granule cultures stained for postsynaptic puncta (PSD95; green). (I) Average number of postsynaptic puncta (PSD95) per neuron. Atm^−/− networks (n = 4; two cultures) have a higher number of presynaptic puncta than the WT (n = 4; two cultures) and [N_Atm^−/− A_WT] chimera networks (n = 4; two cultures; P < 0.05, K-W test followed by Dunn’s test). Note that number of puncta per neuron refers only to the puncta and neurons present in the field of view. (J) Representative immunoblot (nine independent cultures) of p62 and p70S6K at 16 DIV of WT, Atm^−/−, and [N_Atm^−/− A_WT] cultures. (K) Fold difference in p70S6K levels in Atm^−/− (n = 6) and [N_Atm^−/− A_WT] chimeric (n = 4) cultures relative to WT (n = 4) cultures (P < 0.05, K-W test followed by Dunn’s test). (L) Fold difference in p62 levels in Atm^−/− (n = 6) and [N_Atm^−/− A_WT] chimeric cultures (n = 4) relative to WT (n = 4) levels (P < 0.01, K-W test followed by Dunn’s test). Significant differences are marked by asterisks. *P < 0.05; **P < 0.01; ***P < 0.001. MWU test was used in B–E.

Given these results and the previous reports of disrupted number of synapses in rodent models of inherited developmental disorders (21, 22) and impaired pruning with dysregulated mTOR-autophagy signaling observed in autism (23), we investigated whether mTOR hyperactivation and reduction of autophagy occurs in Atm^−/− cerebellar cultures. We characterized the levels of the phosphorylated p70S6K protein, indicative of mTOR activation (24), and p62/SQSTM1 (abbreviated here as p62), a scaffold protein that binds LC3 and ubiquitinated substrates destined for degradation (24). The levels of p62 decrease when autophagy is induced (24). Both the p70S6K and the p62 protein were significantly increased in Atm^−/− cultures in comparison with WT cultures (Fig. 4 K and L; P < 0.05, P < 0.01, K-W with Dunn’s test), indicating that mTOR activation and inhibition of autophagy occurs in Atm^−/− cultures. Although intermediate values of p70S6K were observed in the [N_Atm^−/− A_WT] chimeric cultures with no significant difference from either WT or Atm^−/− networks, we found similar levels of p62 in WT and [N_Atm^−/− A_WT] chimeric cultures that were significantly lower than levels in the Atm^−/− culture (Fig. 4L; P < 0.01, K-W with Dunn’s test), implying that, to some extent, the autophagy process occurs even when mTOR is induced.

Astrocytic Atrophy and Impaired Autophagy Are Associated with Increased GABAergic Synaptic Levels in Adult Atm^−/− Mouse Cerebella.

To understand whether the impact of astrocytic atrophy on the neuronal circuitry demonstrated in vitro results in similar consequences in adult cerebellum, cerebellar circuits in adult WT and Atm^−/− mice were examined. First, similarly to the in vitro results, histological sections confirmed that there was a reduction in GFAP-stained area in both Bergmann glia (41% decrease, MWU test) and Velate astrocytes in the granule layer (47% decrease, MWU test) in the cerebella of adult (4 mo) Atm^−/− mice compared with WT mice (Fig. 5 A–C and SI Appendix, Fig. S7). Second, synaptic protein levels in adult cerebella were analyzed. In accordance with in vitro results, Syn1 levels were up-regulated in Atm^−/− relative to WT adult cerebella (25% increase, t test; Fig. 5G). Levels of PSD-95 (Fig. 5I) and Vglut, a glutamate transporter (Fig. 5L), were similar in both genotypes (t test and MWU test, respectively). Up-regulations of Gephyrin, an inhibitory postsynaptic protein (23% increase, t test; Fig. 5H), and Vgat, a vesicular inhibitory transporter (25% increase, t test; Fig. 5K), in Atm^−/− cerebella indicated that the increased synaptic levels in the adult cerebellum were confined to GABAergic synapses or pathways. Third, we examined the levels of the autophagy markers in the adult cerebellum. As we found in vitro, both p70S6K and p62 were significantly overexpressed in Atm^−/− compared with WT cerebella (Fig. 5 M–O; MWU test).

Fig. 5.

In adult Atm^−/− cerebella, GFAP immunoreactivity is reduced in astrocytic processes; levels of Syn1, Gephyrin, and Vgat are increased; and autophagy is reduced. (A and B) Z-stack histological images of cerebella from 4-mo-old WT and Atm^−/− mice at different magnifications. Astrocytes are stained with GFAP (green), granule neurons with NeuN (magenta), and nuclei with DAPI (blue). The images show Bergmann glia processes (green) at the molecular layer of cerebellar folia. Note that there are fewer GFAP-stained processes in the Atm^−/− cerebella. (C) Morphology of a single representative valet astrocyte (n = 4 for both genotypes). These astrocytes had fewer numbers of processes in Atm^−/− mice than in WT mice. (D–F) Representative immunoblot (four independent mice) of Syn1 (D), Gephyrin (E), and PSD-95 (F) from 4-mo-old WT and Atm^−/− mice. Panel D and F are from the same blot; thus, measured using the same loading control. (G) Fold difference in Syn1 levels in Atm^−/− (n = 5) relative to WT (n = 5) cerebella (t test). (H) Fold difference in Gephyrin in Atm^−/− (n = 5) relative to WT (n = 5) cerebella (t test). (I) Fold difference in PSD-95 in Atm^−/− (n = 5) relative to WT (n = 5) cerebella (t test). (J) Representative immunoblot (four independent mice) of Vgat and Vglut from 4-mo-old WT and Atm^−/− cerebella. (K) Fold difference in Vgat levels in Atm^−/− (n = 5) relative to WT (n = 5) cerebella (t test). (L) Fold difference in Vglut in Atm^−/− (n = 5) relative to WT (n = 5) cerebella (MWU test). (M) Representative immunoblot (four independent mice) of p62 and p70S6K from 4-mo-old WT and Atm^−/− cerebella. (N) Fold difference in p62 levels in Atm^−/− (n = 5) relative to WT (n = 5) cerebella (t test). (O) Fold difference in p70S6K in Atm^−/− (n = 4) relative to WT (n = 4) cerebella (MWU test). Significant differences are marked by asterisks. *P < 0.05; **P < 0.01; ***P < 0.001.

Discussion

To dissect the impact of WT and Atm^−/− astrocytes on the structure and dynamics of cerebellar networks, we first used WT, Atm^−/−, and chimeric cultures of cerebellar cells. Using the chimeric culture approach, we found that the absence of Atm in both astrocytes and neurons led to (i) reductions in lengths and numbers of astrocyte processes; (ii) reduced spatial extent of NSs; (iii) increased numbers of synaptic markers; and (iv) a decreased level of autophagy. The increase in number of presynaptic and postsynaptic sites, corroborated both by an up-regulation of synaptic protein markers and decreased autophagy levels, implies a possible dysregulated pruning process. In [N_Atm^−/− A_WT] chimeric cultures, astrocyte morphology, NSs, synaptic levels, and p62 protein levels were no different from cultures with WT neurons and WT astrocytes. Analysis of adult WT and Atm^−/− cerebella confirmed both the altered astrocyte morphology and the most likely decreased level of autophagy in the mutant cerebella and confined the up-regulation of synaptic protein markers to GABAergic synapses. These results suggest that the impairments in network dynamics observed in Atm^−/− cerebellar networks originate in part from malfunctioning astrocytes that cause aberrant structural connections among neurons, culminating in impaired network dynamics.

Astrocytes’ morphological and pathological changes have been documented in Atm-deficiency studies (10, 25) as well as in numerous brain diseases (7) such as amyotrophic lateral sclerosis (26, 27) and Alzheimer’s disease (AD) (28, 29). The correlation between hyperactive mTOR and impaired autophagy with increased dendritic spine density has been reported in an autism mouse model (23). Interestingly, ATM is involved in a cellular homeostasis pathway that negatively regulates mTOR. Through mTOR repression, ATM can enhance autophagy (9). Therefore, the absence of ATM enhances mTOR levels and results in inhibition of autophagy. We assume that the insignificant reduction of mTOR in [N_Atm^−/− A_WT] chimeric cultures either was sufficient to induce autophagy or enabled induction of a noncanonical autophagy that bypasses some level of activated mTOR (30). The fact that p62 was enhanced in Atm^−/− whole cerebella and Atm^−/− cultures with the increased synaptic levels suggests that autophagy and synaptic connectivity regulation are closely interconnected.

Similar increases in synaptic markers, as shown in in vitro cultures, but limited to GABAergic synapses, were shown in cerebella of Atm^−/− adult mice. These data suggest that up-regulation of GABAergic markers is associated with the cerebellar cell death observed in A-T (31) due to impairment in the development of cerebellar networks [where GABAergic transmission plays a key role (11, 32)].

In this work, spontaneous granule synchronizations mediated by AMPA receptors have been described in primary cerebellar cultures. Although glutamatergic synapses between granule cells are not present in vivo or in situ, coherent oscillations have been recorded in the granular layer in vivo driven by mossy fiber (33, 34) and synchronized by Golgi cells (35), inputs absent in the in vitro model. The engineered in vitro chimeric networks, despite displaying artificial topologies, allowed a clear dissection of the role of astrocytes in cerebellar networks, which can be hardly obtained in in vivo experiments.

The cerebellar cultures’ spontaneous synchronizations occurred at a frequency of 0.07 Hz, similar to frequencies reported for primary cultures from cortical, retinal, and spinal circuits (11–13, 36). Since the frequency of NSs in WT and Atm^−/− cultures is similar, but presents a lower number of GSs and an increased number of SSs in Atm^−/− cultures, we hypothesize that SSs could represent aborted GSs (i.e., network synchronization failures). Similar results have been observed in the neocortex of a mouse model of AD, showing a greater number of false up-state transitions and a smaller number of true up-state transitions in mutant compared with WT animals (37). The presence of spurious or aberrant burst (i.e., SSs) in Atm^−/− cerebellar cultures that do not lead to GSs (consistent with the fact that cells recruited in the SSs do not play a role in the GS build-up) can be explained by the overexpression of synaptic contacts. This overexpression is possibly a consequence of an impaired synaptic pruning mechanism (38) as discussed above. The apparent contradiction between a larger number of synapses in the Atm^−/− circuits and lower occurrence of network synchronizations could result from the homeostatic downscaling of synaptic weights between neurons (aborted effective connectivity hypothesis) or from the presence of nonfunctional connections (aborted functional connectivity hypothesis) (39). Both scenarios are consistent with the reduced autophagy in Atm^−/− cerebellar networks and the hypothesis of lack of synaptic pruning.

Our data support the hypothesis that Atm-deficient astrocytes induce a multiscale accumulation of structural and functional defects into cerebellar circuits. These defects have the potential to scale up the impact from molecular pathways to the synaptic level, to cell assemblies, and neuronal circuits.

Methods

Generation of WT and Atm-Deficient Mice.

The construction of the Atm-deficient mice was described (40). All animal care protocols and all experimental protocols used for this study were conducted according to the animal research guidelines from the Institutional Animal Care and Use Committee, Tel Aviv University (ethical approval no. L-14-019).

Cerebellar Dissociated Cultures Enriched in Granule Neurons.

Primary cultures of mouse cerebellar granule neurons were prepared from 8-d-old Atm^−/− and WT mice as described (14). Cultures were prepared from a single mouse (SI Appendix, SI Methods).

Generation of Chimeric Networks.

Cerebellar cultures were prepared by using the cerebellar dissociated culture protocol (see above). For each pair of pups, a total of three 35-mm Petri dishes were prepared: a control Petri dish from each pup with neurons and astrocytes from the same animal and a single chimeric culture (neurons obtained from one pup and astrocytes from a second pup) (SI Appendix, SI Methods).

Neuronal Recording.

Recordings were performed on cultures at 17 DIV. The loading procedure and recording setup was described (41) (SI Appendix, SI Methods).

Immunocytochemistry.

At the end of calcium imaging experiments, cells on coverslips were washed with PBS and fixed with 4% paraformaldehyde for 10 min at room temperature. Details and antibodies list are provided in SI Appendix, SI Methods.

Immunohistochemistry.

Fresh-frozen cerebella were obtained from 4-mo-old mice, sectioned sagittally in a cryostat to 25-μm sections, and mounted on slides. Details and antibodies list are provided in SI Appendix, SI Methods.

Western Blot Analysis.

Western blots were done on extracts from both primary granular cultures and 4-mo-old WT and Atm^−/− mice cerebellar tissues as described (10) (SI Appendix, SI Methods).

Data Analysis of Network Dynamics Based on Calcium Imaging.

The neuronal activity monitored through calcium imaging was analyzed by using custom code written in Matlab (MathWorks), which implements a few upgrades from what was described (41, 42).

Contour Detection of Neuronal Cell Bodies and Segmentation of Calcium Images.

Segmentation of calcium images was aimed to identify neuronal cells. Cell soma boundaries were automatically identified by using the first 1,000 frames acquired at the beginning of the calcium imaging session (SI Appendix, SI Methods).

Automatic Detection of Astrocytic Arborization from GFAP Staining in Cultures.

The total length of astrocytic arborizations was calculated by using the same automatic identification of boundaries described for neuronal cell body identification (SI Appendix, SI Methods).

Statistical Analyses.

Following a Shapiro–Wilk normality test, data comparisons were carried out by using a two-tailed t test or two-tailed MWU test for two groups and one-way analysis of variance followed by Tukey’s multiple comparison test or K-W test followed by a Dunn’s post hoc test for multiple group comparisons. All statistical analyses were performed by using GraphPad Prism (Version 6 for Windows). P values < 0.05 were considered statistically significant. Values are expressed as means ± SEM. Significant data are denoted with asterisks: *P < 0.05; **P < 0.01; ***P < 0.001. Error bars represent mean + SEM. Statistical details of experiments can be found in the figure legends