Sonic hedgehog enhances calcium oscillations in hippocampal astrocytes

Sonic hedgehog (SHH) is important for organogenesis during development. Recent studies have indicated that SHH is also involved in the proliferation and transformation of astrocytes to the reactive phenotype. However, the mechanisms underlying these are unknown. Involvement of SHH signaling in calcium (Ca) signaling has not been extensively studied. Here, we report that SHH and Smoothened agonist (SAG), an activator of the signaling receptor Smoothened (SMO) in the SHH pathway, activate Ca oscillations in cultured murine hippocampal astrocytes. The response was rapid, on a minute time scale, indicating a noncanonical pathway activity. Pertussis toxin blocked the SAG effect, indicating an involvement of a Gi coupled to SMO. Depletion of extracellular ATP by apyrase, an ATP-degrading enzyme, inhibited the SAG-mediated activation of Ca oscillations. These results indicate that SAG increases extracellular ATP levels by activating ATP release from astrocytes, resulting in Ca oscillation activation. We hypothesize that SHH activates SMO-coupled Gi in astrocytes, causing ATP release and activation of Gq/11-coupled P2 receptors on the same cell or surrounding astrocytes. Transcription factor activities are often modulated by Ca patterns; therefore, SHH signaling may trigger changes in astrocytes by activating Ca oscillations. This enhancement of Ca oscillations by SHH signaling may occur in astrocytes in the brain in vivo because we also observed it in hippocampal brain slices. In summary, SHH and SAG enhance Ca oscillations in hippocampal astrocytes, Gi mediates SAG-induced Ca oscillations downstream of SMO, and ATP-permeable channels may promote the ATP release that activates Ca oscillations in astrocytes.


Sonic hedgehog (SHH) is important for organogenesis during development. Recent studies have indicated that SHH is also
involved in the proliferation and transformation of astrocytes to the reactive phenotype. However, the mechanisms underlying these are unknown. Involvement of SHH signaling in calcium (Ca) signaling has not been extensively studied. Here, we report that SHH and Smoothened agonist (SAG), an activator of the signaling receptor Smoothened (SMO) in the SHH pathway, activate Ca oscillations in cultured murine hippocampal astrocytes. The response was rapid, on a minute time scale, indicating a noncanonical pathway activity. Pertussis toxin blocked the SAG effect, indicating an involvement of a G i coupled to SMO. Depletion of extracellular ATP by apyrase, an ATP-degrading enzyme, inhibited the SAG-mediated activation of Ca oscillations. These results indicate that SAG increases extracellular ATP levels by activating ATP release from astrocytes, resulting in Ca oscillation activation. We hypothesize that SHH activates SMO-coupled Gi in astrocytes, causing ATP release and activation of G q/11 -coupled P2 receptors on the same cell or surrounding astrocytes. Transcription factor activities are often modulated by Ca patterns; therefore, SHH signaling may trigger changes in astrocytes by activating Ca oscillations. This enhancement of Ca oscillations by SHH signaling may occur in astrocytes in the brain in vivo because we also observed it in hippocampal brain slices. In summary, SHH and SAG enhance Ca oscillations in hippocampal astrocytes, G i mediates SAG-induced Ca oscillations downstream of SMO, and ATP-permeable channels may promote the ATP release that activates Ca oscillations in astrocytes.
Astrocytes are a major glial cell population in the central nervous system (CNS) 2 with important roles in brain homeostasis, such as clearance of glutamate and GABA from the extracellular space, provision of nutrients from blood vessels to neurons, and control of extracellular pH (1,2). Astrocytes are transformed into reactive astrocytes in response to brain injury and inflammation. Reactive astrocytes have altered gene expression patterns and morphology and play roles in scar formation and in preventing the spread of inflammation. Astrocytes also modulate neural excitability and synaptic connectivity by releasing so-called gliotransmitters, among which glutamate and ATP are major components (3,4).

Sonic hedgehog
The Hedgehog gene was identified in the 1970s as a gene involved in Drosophila larval segmentation (5). There are three Hedgehog homologs in vertebrates, Sonic hedgehog (Shh), Desert hedgehog (Dhh), and Indian hedgehog (Ihh) (6). Shh is involved in organogenesis and development of the CNS and is expressed throughout the body. In the absence of SHH, an SHH receptor, Patched, keeps a 7-transmembrane receptor, Smoothened (SMO), from activating a transcription factor, GLI. Binding of SHH to Patched releases SMO to activate GLI, which translocates into the nucleus and activates transcription, thereby promoting cell proliferation and differentiation (7)(8)(9). Aside from this well-known canonical pathway, noncanonical pathways triggered by Patched activation have also been reported (10, 11). These pathways are not linked to GLI activation but regulate cell death (12), axon guidance (13), and cytoskeleton (14) with or without SMO activation.

SHH in the CNS
In early CNS development, SHH is secreted from the notochord and floor plate as a morphogen to direct dorso-ventral patterning of the CNS. During late CNS development, SHH is found in the cerebral cortex, optic tectum, and cerebellar cortex (15). SHH is also expressed in the adult CNS (16); SHH and Patched are expressed in the forebrain, cerebellar Purkinje cells, and spinal cord motor neurons. SMO is expressed in circumventricular organs, granular cells in the hippocampal dentate gyrus, and neurons in the reticular thalamic nuclei (17). The expression of SHH is particularly strong in the hippocampal dentate gyrus and the subventricular zone where adult neurogenesis takes place and retention, proliferation, and differentiation of neural stem cells occurs (16). In hippocampal neurons, SHH is present presynaptically and postsynaptically (18), and Patched and SMO are localized not only in cell bodies but also in dendrites and postsynapses (19). Involvement of SHH in synaptic plasticity was also reported (20). SHH expression is activated upon traumatic injury in the brain (21,22) including in astrocytes (21,23). During injury, released SHH increases the expression of glial fibrillary acidic protein in astrocytes and induces transformation of astrocytes to reactive astrocytes. SHH administration also induces transformation to reactive astrocytes (24,25) and gliotransmitter release from astrocytes. All of these observations indicate important roles of SHH in the regulation of astrocytes. However, detailed mechanisms for these actions have not been elucidated.

Ca oscillation in glia
Although neurons communicate with each other by electrical activity, astrocytes transmit information by changing intracellular Ca 2ϩ . Ca oscillation is frequently observed in astrocytes, in which Ca transients repeatedly occur in individual cells. These changing Ca patterns sometimes display wavelike propagation among astrocytes, which is called the Ca wave (26,27). Various extracellular stimuli evoke Ca oscillations in astrocytes through various plasma membrane receptors, whereas intracellular Ca release occurs from the endoplasmic reticulum (ER) through inositol trisphosphate receptors (IP3Rs) downstream of G q/11 -coupled receptors.
ATP is released from astrocytes as a gliotransmitter and influences neuronal excitability (3,4) and regulates Ca dynamics in astrocytes (29). Okuda et al. (30) reported that SHHstimulated astrocytes release ATP. Two mechanisms for releasing ATP from astrocytes are known: vesicular release and release through channels. Supporting vesicular release, a vesicular nucleotide transporter is expressed in astrocytes. For release through channels, ATP-permeable channels, maxianion channels, connexin hemichannels, pannexin hemichannels, and the P2X7 receptor are postulated to be involved. Several studies have shown that deprivation of oxygen and glucose, osmotic stimulation, and stretch stimulation induce ATP release from astrocytes through channels on the plasma membrane.

Aims of this study
The high levels of SHH and related proteins in the adult hippocampus together with the inferred roles of SHH signaling in adult neurogenesis and brain injury led us to characterize cellular responses to SHH in cultured hippocampal cells. We found the enhancement of Ca oscillations in astrocytes within several minutes after application of SHH pathway agonists, namely SHH and Smoothened agonist (SAG). This enhancement was blocked by inhibition of G i and removal of extracellular ATP. Together with other lines of evidence, we propose that the enhancement of Ca oscillations in astrocytes is initiated by the activation of SMO-coupled G i , which leads to ATP release through ATP-permeable channels. This released ATP then enhances Ca oscillations in nearby astrocytes. We also observed enhanced Ca oscillations in astrocytes in brain slices; therefore, this mechanism may be functional in the in vivo brain as well.

SHH and SAG evoke Ca oscillations in mouse hippocampal cultures
Calcium imaging was performed by loading cultured mouse hippocampal cells with Fura-2 (Fig. 1A). Primary cultures of mouse hippocampal cells were exposed to SHH (500 pM, 10 ng/ml) in the presence of 1 M tetrodotoxin (TTX), and some of the cells exhibited spontaneous Ca oscillations before the agonist application. Addition of SHH evoked Ca oscillations in quiescent cells and enhanced the frequency of Ca oscillations in cells that had already shown spontaneous Ca oscillations before the agonist application (Fig. 1B): a cumulative histogram during the agonist application periods was right-shifted from that during the baseline period ( Fig. 1B, b, p Ͻ 0.001), indicating an increase in Ca oscillation frequency by SHH. A SMO agonist, SAG (5 M), induced a similar increase in Ca oscillation frequency (Fig. 1C, p Ͻ 0.001). A series of different SHH and SAG concentrations resulted in increased or decreased Ca oscillation frequencies, but did not follow simple dose-response relationships ( Fig. 1, B and C, b panels, and Fig. S1). Agonist-induced Ca frequency increase was evaluated by subtracting the baseline Ca frequency from that after the drug application in each cell (⌬Frequency). ⌬Frequencies of cells applied with concentrations of SHH or SAG were compared with that of vehicleapplied control cells (0.1% DMSO; Fig. 1, D and E). SHH increased Ca oscillation frequency significantly at 500 pM (p Ͻ 0.001) but less at 50 pM (p Ͻ 0.01) and not at 5 nM (p ϭ 0.82). SAG increased Ca event frequency significantly at 5 M (p Ͻ 0.01) and 50 nM (p Ͻ 0.01), and decreased the Ca event frequency at 500 nM (p Ͻ 0.05) and 100 nM (p Ͻ 0.001). SAG at 1 M did not show a difference from the control (p ϭ 0.31). We used SHH at 500 pM hereafter. For SAG, we considered that 5 M would produce the most reliable results because 100 nM SAG showed a large decrease in Ca oscillation frequency (Fig.  S1F). The SAG concentration used hereafter was, therefore, 5 M unless otherwise indicated. Some cells that did not show Ca transients during the initial 10-min baseline period started showing Ca transients after SHH or SAG stimulation; therefore, the stimuli increased the proportion of cells showing Ca transients as well as the frequency of Ca oscillations in each cell. About half of the cells did not show Ca transients throughout the recording period.

SHH enhances Ca oscillations in astrocytes
We tested cyclopamine (CPN), which is widely used as a SMO antagonist (31), on the assumption that it would block the effect of SAG. However, administration of CPN per se caused an increase in Ca oscillation frequency to a similar extent as SAG, and the addition of SAG to CPN further enhanced Ca oscillation frequency ( Fig. 2A, a and b, p Ͻ 0.001 for base versus CPN and base versus CPN ϩ SAG, respectively), which concealed the effect of subsequently applied SAG. ⌬Frequencies of Ca oscillation both in the CPN and CPN ϩ SAG periods were significantly larger than that of a control cell group ( Fig. 2A, c, p Ͻ 0.001, respectively). CPN seemed to act as an agonist on the Ca oscillation enhancement mechanism originating from SMO. Such an agonistic action of CPN on a noncanonical SHH pathway has been previously reported (32,33); therefore, we consider that the enhancement of Ca oscillations observed in this study was derived from the activation of a noncanonical SHH signaling pathway.

The SAG induced Ca oscillations are mediated by extracellular ATP
We next investigated whether extracellular ATP is involved in the enhancement of Ca oscillations by SHH signaling because ATP is a well-known activator of Ca oscillations in astrocytes (34,35). Administration of apyrase, an ATP-degrading enzyme, to the extracellular space abolished the enhanced Ca oscillations evoked by SAG (Fig. 2B, b, p Ͻ 0.001 for base versus SAG and p Ͻ 0.05 for SAG versus apyrase), which was also shown in the ⌬Frequency analysis (Fig. 2B, c, p Ͻ 0.001). This result indicates that SMO activation induced the increase

SHH enhances Ca oscillations in astrocytes
in extracellular ATP, which then enhanced Ca oscillations in the hippocampal cells. In a control experiment, a vehicle solution, HEPES-buffered saline (HBS) containing TTX, was added instead of apyrase, which produced an enhancement of Ca oscillation frequency (Fig. 2C, a, p Ͻ 0.001). ⌬Frequency[(SAG ϩ vehicle) Ϫ base] was greater than ⌬Frequency(SAG Ϫ base) ( Fig. 2C, b, p Ͻ 0.001). We have no clear explanation for this, but the dilution of SAG could be a cause because a relatively low concentration of SAG (50 nM) was as effective as 5 M (Fig. 1E).

Astrocytes are responsible for the enhancement of Ca oscillations
Primary hippocampal cell cultures are heterogeneous, containing neurons, astrocytes, microglia, oligodendrocytes, and other cell types. We, therefore, identified the cell types relevant to SAG-enhanced Ca oscillation. We assessed cell types in the hippocampal culture by immunohistochemistry with anti-MAP2, anti-S100␤, anti-Iba1, and anti-Olig2 antibodies, which are markers for neurons, astrocytes, microglia, and oligoden-drocytes, respectively (Fig. S2). MAP2-and S100␤-positive cells were observed throughout cultures (Fig. S2, A and B). Iba1positive cells were only found in one batch (10 -20 cells in each coverslip culture) of three culture batches (Fig. S2C). Olig2positive cells were only found where neuron density was very high, but not in fields of view where neuron density was modest; we used such fields of view with modest neuron density for Ca imaging (Fig. S2D). Next, we performed Ca imaging followed by immunohistochemistry with anti-MAP2 and S100␤ antibodies (Fig. 3A). All the Fura-2-loaded cells were stained with either MAP2 or S100␤ in three fields of view from different cultures. S100␤-positive cells were 58.9 Ϯ 9.8% of all cells and the remainder were all MAP2-positive (n ϭ 3 cultures). Ca transients were observed in 69.6 Ϯ 14.1% of the S100␤-positive cells and 32.5 Ϯ 16.0% of the MAP2-positive cells. In our cultures, neurons showed frequent Ca transients in a synchronized fashion unless TTX was included in the bath solution (data not shown), but neurons showed far fewer spontaneous Ca transients than astrocytes when action potential generation was . c, ⌬Frequency of the drug effects are displayed as cumulative histograms, together with that of a control cell group to which 0.1% DMSO was added in place of CPN, which is the same one shown in Fig. 1, D and E. B, a, apyrase (5 units/ml) suppressed the SAG-induced enhancement of Ca oscillation frequency. b, cumulative histograms of Ca oscillation frequency during the baseline period, after SAG application and after apyrase addition (n ϭ 444 cells from 3 cultures). c, cumulative histograms of ⌬Frequency. C, as a control, vehicle solution (HBS-TTX) was added in place of apyrase in B. Cumulative histograms of Ca event frequency (C, a) and ⌬Frequency (C, b) are shown (n ϭ 729 cells from 7 cultures).

SHH enhances Ca oscillations in astrocytes
blocked by TTX. From these results, we consider that the major cell population that showed Ca oscillations was astrocytes. We tested whether the enhancement of Ca oscillations by SHH or SAG could be evoked in astrocyte cultures as well as in the primary heterogeneous culture, and this was indeed the case: the enhancing effects of SHH or SAG on Ca oscillation frequency were observed in the astrocyte culture (Fig. 3B, c, p Ͻ 0.001; and Fig. 3C, b, p Ͻ 0.001, respectively). ⌬Frequency of SHH and SAG compared with a DMSO (0.1%) applied to the control cell group showed increased Ca oscillation frequency with SHH or SAG (Fig. 3D, p Ͻ 0.01 and p Ͻ 0.001, respectively). We performed immunostaining of the astrocyte culture with the same antibodies as above. The anti-S100␤ antibody labeled cells over entire coverslips from two culture batches. However, no signals for anti-MAP2, anti-Iba1, and anti-Olig2 antibodies were observed (data not shown). Thus, it was clear that astrocytes were activated by SHH or SAG and showed enhanced Ca oscillations, although there is a possibility that neurons could also receive SHH or SAG and activate astrocyte Ca oscillations. To eliminate the possibility of astrocyte activation by neurons, we used astrocyte cultures in the following experiments.

Ca release from ER is necessary for Ca oscillations
To characterize the mechanism of SAG-enhanced Ca oscillations, the source of Ca 2ϩ was investigated. Removal of Ca 2ϩ from the extracellular solution did not alter the SAG-or SHHinduced enhancement of Ca oscillations in astrocytes (Figs. 4A and Fig. S3A, respectively): SAG enhanced Ca oscillation frequency with Ca-free medium (Fig. 4A, b, p Ͻ 0.05), and ⌬Frequency(SAG Ϫ base) in Ca-free medium showed apparent difference from that of a control cell group (Fig. 4A, c, p Ͻ

SHH enhances Ca oscillations in astrocytes
0.001). In contrast, disruption of intracellular Ca release mechanisms resulted in drastic changes: 2-aminoethyl diphenylborinate (2-APB, 50 M), an IP3R inhibitor, and thapsigargin (Tg, 100 nM), an inhibitor of the ER Ca-ATPase (36), blocked the SAGinduced enhancement of Ca oscillations (Fig. 4, B and C). SAG did not take effect under 2-APB (Fig. 4B, b, base versus 2-APB:  (Fig. S3B). Although 2-APB also blocks transient receptor potential (TRP) channels on the plasma membrane (37), we consider that the result with 2-APB was not due to TRP blockade because removal of extracellular Ca 2ϩ did not have an effect. Tg induced a slow increase in intracellular Ca 2ϩ concentration ([Ca 2ϩ ] i ) over 15 min. The level then decreased to the original baseline level. This is usually seen because of inhibited uptake of cytosolic Ca 2ϩ into ER. Ca oscillations occurred riding on the slow Tg-induced [Ca 2ϩ ] i increase, but the frequency was not increased by SAG (Fig. 4C,  b), and ⌬Frequency compared with the control cell group shows an apparent block of the SAG-induced Ca oscillation enhancement by Tg (⌬Frequency [(Tg ϩ SAG) Ϫ base] versus ⌬Frequency[(DMSO ϩ SAG) Ϫ base]: p Ͻ 0.001). Application of dantrolene (10 M), which inhibits ryanodine receptormediated Ca release from the ER, did not result in a clear inhibition of the SAG-enhanced Ca oscillations: SAG further enhanced the Ca oscillation frequency, which had been enhanced by dantrolene (Fig. 4D, b, (Fig. 4D, c). These results suggest that Ca release not though ryanodine receptor but through IP3Rs is relevant to the Ca oscillations enhanced by SAG.

SMO is coupled to G i
SMO is a 7-transmembrane receptor that couples with a trimeric G protein, G i (38,39). We tested if the SMO enhancement of Ca oscillations requires G i activation by incubating astrocytes with pertussis toxin (PTX), a G i inhibitor. SAG induced an enhancement of Ca oscillations in PTX-treated astrocytes (Fig. 5A, b, p Ͻ 0.001), but it was significantly smaller than the SAG-induced enhancement without PTX treatment. Furthermore, the addition of DMSO (0.1%) in place of SAG in PTX-treated astrocytes induced an even larger increase in Ca oscillation frequency than that produced by SAG in PTXtreated astrocytes (Fig. 5A, c, p Ͻ 0.001). ⌬Frequency analysis shows that the SAG-induced increase in Ca event frequency in the PTX-treated cells was smaller than that by DMSO in PTX-treated cells (Fig. 5A, d, ⌬Frequency(SAG Ϫ base in PTX) versus ⌬Frequency(DMSO Ϫ base in PTX): p Ͻ 0.001) and much smaller than that by SAG without PTX treatment (⌬Frequency(SAG Ϫ base in PTX) versus ⌬Frequency(SAG Ϫ base without PTX): p Ͻ 0.001). Thus, we concluded that the enhancement of Ca oscillations by SAG requires G i activity. G i suppresses cAMP production from ATP by inhibiting adenylate cyclase. We confirmed that G i was actually activated by SAG stimulation by intracellular cAMP imaging using a fluorescent protein-based cAMP indicator, Flamindo2 (40). SAG stimulation increased the fluorescence intensity of Flamindo2 in astrocytes (Fig. 5B, c) compared with a control cell group to which vehicle (0.1% DMSO) was applied in place of SAG (p Ͻ 0.001, two-way analysis of variance), indicating that cAMP concentration was decreased. These results showed that SAG stimulation activated G i , presumably via SMO, which is necessary for the downstream SAG-induced enhancement of Ca oscillations.

SHH enhances Ca oscillations in astrocytes Blockers of ATP-release channels inhibit the enhancement of Ca oscillations
We hypothesized that the SMO and G i activation by SAG induced ATP release, which led to the increased extracellular ATP concentration. To test this, carbenoxolone (CBX; 100 M) and 1-octanol (2 mM), blockers of connexin hemichannels (41,42), gadolinium (Gd 3ϩ , 50 M), a blocker of maxianion chan-nels (43), and Brilliant Blue G (BBG; 1 M), a blocker of P2X7 receptors (44), were applied to astrocytes. The SAG-induced enhancement of Ca oscillation frequency was completely suppressed by CBX, 1-octanol, and Gd 3ϩ (Fig. 6, A-C). CBX and 1-octanol induced slow Ca transients lasting 5-10 min. CBX induced Ca event frequency enhancement, which was then inhibited by the addition of SAG (Fig. 6A, b, base versus CBX: , a P2X7 receptor antagonist, was applied and addition of SAG followed. b, cumulative histograms of Ca oscillation frequency before application of the blockers, under the blockers, and after SAG addition. c, cumulative histograms of ⌬Frequency together with those from a control cell group in which 0.1% DMSO was applied in place of the blockers (n ϭ 368 cells from 5 cultures).

Ca oscillations in brain slice astrocytes are enhanced by SAG
Astrocytes in culture have different features to those in the brain, e.g. in morphology and proliferation state (45). To test if the enhancement of Ca oscillations by SHH pathway activation is an atypical phenomenon in cultured astrocytes or a general feature of astrocytes, we performed Ca imaging in astrocytes in hippocampal slices. [Ca 2ϩ ] i in astrocytes labeled with SR101 (46) was monitored with a Ca dye, Fluo-4 (Fig. 7). In dentate gyrus of hippocampus, intensively stained cells with SR101 were found in the molecular layer and weakly stained cells in the granule cell layer. Cell bodies of some interneuron species are known to exist in the inner molecular layer (47)(48)(49), and we indeed observed NeuN-positive, a general neuron marker, cell bodies in the inner molecular layer: some in proximity to the granule cell layer and much fewer distant from granule cell layer (Fig. S4). Therefore, ROIs were placed on SR101 and Fluo-4 double positive cells in the molecular layer excluding cells within 30 m from the granule cell layer, where most, not to say all, SR101-positive cells should be astrocytes. During the baseline period, some astrocytes showed Ca oscillations, and SAG (50 nM) enhanced Ca oscillation frequency (Fig. 7C, p Ͻ 0.01), whereas vehicle (0.1% DMSO) did not (Fig. 7B, p ϭ 0.45). ⌬Frequency analysis also showed a significant difference between SAG and vehicle (Fig. 7D, p Ͻ 0.01). This result indicates that the enhancement of Ca oscillations by SHH signaling operates not only in cultured astrocytes but also in brain slice astrocytes.

Discussion
Astrocytes are known to respond to stimuli by evoking or altering their Ca oscillation patterns (29,50,51), in which G q/11 -coupled receptors, namely metabotropic glutamate receptors and P2Y receptors are often involved (52)(53)(54). In this study, the enhancement of Ca oscillations by SAG was inhibited by the degradation of extracellular ATP with apyrase or inhibition of G i with PTX. Therefore, the enhancement of Ca oscillations did not result from direct activation of G q/11 but from G i activation by SMO, which in turn evoked ATP release. ATP release from astrocytes by SHH has been previously reported (30). The increased extracellular ATP may have activated P2Y receptors in adjacent astrocytes, and possibly in the astrocytes that released ATP, thereby enhancing the Ca oscillations in these cells. P2Y1/2/4/6/12/13/14 receptors are expressed in astrocytes, and P2Y1/2/4/6 receptors are coupled with G q/11 (52,55).
The finding that CBX, 1-octanol, and Gd 3ϩ inhibited the enhancement of Ca oscillations by SAG raises the possibility that the ATP release downstream of activated SMO was not mediated by a vesicular release mechanism but through ATPpermeable channels, namely connexin hemichannels and maxianion channels, which are sensitive to these antagonists. SLCO2A1 is a core component of the maxianion channel (56), although the precise molecular composition of the maxianion channel is still unclear (57). G protein-coupled receptor activation (58) and dephosphorylation of maxianion channels (59) are involved in the opening mechanisms. The presumed suppression of cAMP-dependent protein kinase (PKA) downstream of G i activation by SAG could lead to dephosphorylation of maxianion channels. The opening mechanism of connexin hemichannels has not been clarified, and thus the mechanisms underlying ATP release through them are unknown (60). However, because both CBX and Gd 3ϩ block a wide range of channels (61), they could directly inhibit the Ca oscillation mechanism downstream of ATP release rather than affecting the ATP release channels. Furthermore, 2-APB inhibits connexin channels with varying potencies depending on the subunit (62). Thus, an alteration of connexin-mediated ATP release by 2-APB could also be involved in the 2-APB inhibition of the SAG-induced Ca oscillation enhancement (Fig. 4) in addition to the inhibition of IP3R by 2-APB. Characterization of the ATPrelease mechanisms downstream of SMO in astrocytes remains to be elucidated.
The lag in the onset of changes in the frequency of Ca transient on application of SHH or SAG varied considerably among cells. In some cells the increase started within 1 min and in others responses started after delays of 2-5 min. This variability indicates that some of the steps from SMO activation to the increase in extracellular ATP concentration are slow with variable speeds among cells. Extracellular ATP accumulation is one such variable step. The not-so-simple dose-response in Ca oscillation frequency in response to SHH or SAG (Fig. 1, D and  E) may reflect such a complicated mechanism, together with a possibility that SHH and SAG have multiple sites of action on their receptors.
Although we consider that the direct ATP release and subsequent Ca oscillations was via a noncanonical SHH pathway through G i activation, the canonical SHH pathway, in which GLI is activated and transported to the nucleus, may be influenced by the noncanonical series of events. When the SHH signal is turned off, GLI is ubiquitinated by ␤-TrPC, which is induced by the activities of PKA, glycogen synthase kinase 3␤,

SHH enhances Ca oscillations in astrocytes
and CK1. This removes the GLI active domain and keeps GLI inactive (9). When the SHH signal is turned on, GLI escapes from ubiquitination, and full-length GLI dissociates from the negative regulator, Suppressor of Fused (SuFu), and migrates into the nucleus (9,63,64). Although the complex interactions between SMO and GLI are not understood in detail, several lines of evidence indicate that PKA is a key negative regulator of the canonical SHH signal downstream of SMO (65-69). Thus, the presumed PKA deactivation following G i activation and the decrease in cAMP concentration, which were confirmed in this study, may reduce the negative effect of PKA and push the balance of inactive/active GLI molecules to the active side. Whether this deactivation of PKA by activation of G i downstream of SMO is peculiar to astrocytes or a more general feature needs to be clarified.
Adenylyl cyclase 5 and 6 (AC5/6) localize to primary cilia (70) and are sensitive to [Ca 2ϩ ] i (71). SHH and SAG raise [Ca 2ϩ ] i in primary cilia, possibly via Trp channels and Gd 3ϩ -sensitive plasma membrane channels (72)(73)(74). Moore et al. (74) reported that the cAMP concentration in primary cilia is 5-fold higher than that of whole cells, and SHH stimulation increased ciliary [Ca 2ϩ ] i and decreased ciliary cAMP concentration in mouse embryonic fibroblasts. They suggested that inhibition of AC5/6 by Ca 2ϩ is the mechanism for the reduction of cAMP in cilia by SHH. Primary cilia may behave as isolated compartments where concentration and dynamics of signaling molecules, including Ca 2ϩ , are separate from global cytoplasmic Ca 2ϩ (73,75). Therefore, the Ca increase and cAMP reduction observed in this study may have occurred in SAG-stimulated primary cilia of astrocytes in parallel with the global enhancement of cytoplasmic Ca oscillations and reduced concentrations of cAMP.
In the adult brain, SHH is secreted upon brain injury and makes astrocytes reactive (21,25). In severe cases, the gathered

SHH enhances Ca oscillations in astrocytes
astrocytes then form a characteristic astrocyte scar around the injured site (76,77). Although the intercellular and intracellular signaling mechanisms that make astrocytes reactive and form scars are not well-characterized, an increase in extracellular ATP has been implicated in the initial microglial activation (78), leading to transformation of astrocytes to the reactive state (79). Although it is proposed that ATP is released from injured cells as "find me" (80) or "eat me" (81) signals, intact astrocytes could be involved in the increase in extracellular ATP in the early phase of the response to injury by responding to SHH release because SHH levels are raised after traumatic brain injury (21,22). SHH is also expressed after brain ischemia (82), and administration of SHH to rats just after stroke partially relieves neurological damage with improved angiogenesis and neuron survival (83). The release of ATP from astrocytes and enhancement of Ca oscillations in astrocytes may also be involved in recovery from brain ischemia.
Ca oscillations are observed in a wide range of cell types (51), and controls various vital functions of cells, e.g. egg activation in fertilization (84) and differentiation of osteoclasts (85). However, outcomes of many Ca oscillation events are not known, including for the Ca oscillations in astrocytes. A possible scenario is that Ca oscillations activate Ca-dependent transcription factors and alters gene expression patterns, which may modulate cellular proliferation, differentiation, and programmed cell death. It was proposed that different sets of transcription factors are activated according to the frequency and amplitude of Ca oscillations following experiments in which intracellular Ca patterns were artificially controlled (86,87). NFAT activity is controlled by Ca oscillations (86 -88), and dephosphorylation of NFAT by calcineurin, a Ca-dependent phosphatase, is postulated as a mechanism for NFAT activation by Ca oscillations (89). It is conceivable that the activity of GLI via the canonical SHH pathway is regulated by Ca oscillations evoked through a noncanonical pathway, because the activity of GLI is controlled by dephosphorylation (68). The expression of SHH is increased during development and upon traumatic injury; therefore, the Ca oscillations in astrocytes observed in this study may play roles in differentiation and activation of astrocytes and may affect nearby neurons by releasing gliotransmitters. In future research, it will be of great interest to determine the outcome of Ca oscillations induced by SHH in slice preparations or in vivo.
In summary, we found that: 1) Ca oscillations in the hippocampus were enhanced in astrocytes in response to SHH or SAG; 2) the enhancement of Ca oscillations by SAG required IP3R-dependent Ca release; 3) G i plays a role downstream of SMO in the enhancement of Ca oscillations by SAG; 4) ATP-permeable channels may be responsible for the ATP release that activates Ca oscillations in surrounding astrocytes; and 5) the enhancement of Ca oscillations by SHH signaling was not peculiar to cultured astrocytes and was also observed in slice preparations.

Animal care
Animal care was in accordance with guidelines outlined by the Institutional Animal Care and Use Committee of Waseda University. The protocol was approved by the Committee on the Ethics of Animal Experiments of Waseda University. All efforts were made to minimize the number of animals used and their suffering during experiments.

Cell culture imaging
Coverslips holding cultured cells were mounted in a stainless steel chamber containing HBS (in mM, 20 HEPES, 115 NaCl, 5.4 KCl, 1 MgCl 2 , 2 CaCl 2 , 10 glucose, pH 7.4). A Ca indicator, Fura-2 was loaded into cells by incubation with 2.5 M Fura-2-AM/HBS (Dojindo, Kumamoto, Japan) at 37°C for 10 min followed by three washes with HBS. TTX (1 M) was added to HBS throughout the imaging procedures (HBS-TTX) when primary culture preparations were used to avoid Ca events evoked by neuronal activities. Time-lapse imaging was performed with an inverted microscope (IX71, Olympus, Tokyo, Japan) with a ϫ20 objective (UApo/340, N.A., 0.75, Olympus). Fura-2 was excited by 340 and 380 nm wavelength light alternating every 3 s and fluorescence was detected with a cooled CCD camera (ORCA-ER, Hamamatsu Photonics, Hamamatsu, Japan) through a beam splitter (400 nm) and an emission filter (420-nm long pass). The fluorescence intensity ratio (F 340 /F 380 ) was calculated for each time point of each ROI by dividing fluorescence intensity excited by 340-nm light by that excited by 380-nm light after background subtraction. Fluorescence signals of each wavelength image were calculated by averaging pixel intensities in an oval ROI covering each cell body. ROIs were set on all cells identifiable by the Fura-2 staining in each field of view. Ca transients were detected by finding F 340 /F 380 changes exceeding a threshold from a baseline value along the time course of the F 340 /F 380 ratio. The baseline value for each time point of each ROI (R base ) was determined by calculating the average of the F 340 /F 380 values of the ROI for 20 frames just before the time point. To reject large deflecting values due to Ca transients, data points with values twice larger than the standard deviation of the values of the ROI throughout the recording period were omitted from the baseline calculation. In experiments where the Ca recording baseline suffered from artifactual slow drifts because of optical interference by applied reagents (Tg, 2-APB, CBX, and 1-octanol; see Figs. 4 and 6), linear regression instead of averaging was used to determine R base , which was effective to cancel slow baseline drift (Fig. S5). Cells showed various Ca changes in this study; from apparent large spikes of Ca tran-sients to much smaller Ca changes just above the noise level. Because we focused on the former clear Ca transients in this study, we determined the threshold for detection of Ca events as 0.08 for cultured cells, which effectively isolated the clear Ca transients. In most of the Ca measurement experiments the frequencies of Ca transients before drug application (baseline) and after drug application periods were calculated in each cell. These frequencies of all cells from repeated experiments were pooled as a group, and are displayed by cumulative histograms in the figures. To make the Ca frequency changes clearer, the baseline Ca event frequency was subtracted from the frequency after drug application in each cell (⌬Frequency). ⌬Frequencies of all cells from repeated experiments were pooled as a group, and are displayed as cumulative histograms in figures and compared with other groups of different experimental conditions by considering the null hypothesis that ⌬Frequency of cells is the same between different experimental conditions. When the cAMP concentration was measured, Fura-2 was loaded into cultured astrocytes that had been transfected 2-5 days previously with plasmid encoding Flamindo 2, and time-lapse imaging of three alternating wavelengths was performed every 5 s, by adding a blue excitation and a green emission acquisition for Flamindo 2 (490 -500 nm excitation and 515-560 nm emission filters with a 505-nm beam splitter) to the Fura-2 dual wavelength acquisition described above. There was no detectable cross-talk between the Fura-2 and Flamindo 2 optical paths. For time course analysis of Flamindo 2 results, the fluorescence signal of each frame was divided by that of the first image frame (F/F 0 ) after background subtraction. Image acquisition and data analysis were performed with custom-made TI Workbench software written by T. I. (92).

Brain slice
Hippocampal slices, 400 m thick, were prepared from 4 -8week-old male C57BL/6 mice using a standard method (93) and used within 1 day of preparation. A Ca indicator, Fluo-4 AM (Thermo Fisher Scientific), and an astrocyte marker, sulforhodamine 101 (SR101, sc-215929, Santa Cruz), were loaded onto the slices by incubating with 5 M Fluo-4 AM in artificial cerebrospinal fluid (ACSF; in mM, 1 NaCl 125; KCl 2.5; NaH 2 PO 4 1.25; NaHCO 3 26; MgCl 2 1; CaCl 2 2; and glucose 20, 310 -315 mosmol and bubbled with a 95% O 2 and 5% CO 2 gas mixture) at 34°C for 40 min. SR101 (1 M) was then added and the slices were stained for another 20 min. After staining, slices were transferred to ACSF and kept at room temperature for 30 min before the time-lapse imaging. Time-lapse imaging was performed with an in-house two-photon microscope (92) mounted on an upright microscope (BX51, Olympus) with a ϫ20 water immersion objective (XLUMPlanFl, N.A., 0.95, Olympus). Brain slices were continuously superfused with ACSF containing 1 M TTX at 34°C. An excitation laser of wavelength at 920 nm (titanium-sapphire pulse laser, Mai Tai DeepSee, Spectra-Physics, Santa Clara, CA) was used and the emission was divided with a 580-nm beam splitter and passed through a 495-540 or 575-630 -nm band-pass filter for the Fluo-4 and the SR101 signals, respectively. Data analysis was performed as described for the cell culture experiments except that the ROIs were placed on SR101 and Fluo-4 double positive SHH enhances Ca oscillations in astrocytes cells in the molecular layer and the Fluo-4 fluorescence intensity of each ROI was normalized by dividing by that of the first image frame (F/F 0 ). A baseline value for each data point was calculated as described above by averaging the preceding 20 frames, and fluorescence intensity of the data point of the ROI was subtracted with the baseline value. We used 0.3 as the threshold to detect Ca transients. Although SR101 is widely used to stain astrocytes, hyperexcitation of neurons is a known side effect (94,95). We consider that if this side effect occurred in this study it was minute because the concentration of SR101 used was much lower than the suggested threshold (between 50 and 250 M) (95).

Statistics
Mann-Whitney U test included in the ALGLIB library (www.alglib.net) 3 implemented in TI Workbench was used to compare two groups unless otherwise indicated. All indicated data are given as the average Ϯ S.D.