Serotonin receptor 4 regulates hippocampal astrocyte morphology and function

Astrocytes are an important component of the multipartite synapse and crucial for proper neuronal network function. Although small GTPases of the Rho family are powerful regulators of cellular morphology, the signaling modules of Rho‐mediated pathways in astrocytes remain enigmatic. Here we demonstrated that the serotonin receptor 4 (5‐HT4R) is expressed in hippocampal astrocytes, both in vitro and in vivo. Through fluorescence microscopy, we established that 5‐HT4R activation triggered RhoA activity via Gα13‐mediated signaling, which boosted filamentous actin assembly, leading to morphological changes in hippocampal astrocytes. We investigated the effects of these 5‐HT4R‐mediated changes in mixed cultures and in acute slices, in which 5‐HT4R was expressed exclusively in astrocytes. In both systems, 5‐HT4R‐RhoA signaling changed glutamatergic synaptic transmission: It increased the frequency of miniature excitatory postsynaptic currents (mEPSCs) in mixed cultures and reduced the paired‐pulse‐ratio (PPR) of field excitatory postsynaptic potentials (fEPSPs) in acute slices. Overall, our present findings demonstrate that astrocytic 5‐HT4R‐Gα13‐RhoA signaling is a previously unrecognized molecular pathway involved in the functional regulation of excitatory synaptic circuits.

In the present study, by combining quantitative molecular microscopy, time-lapse Förster resonance energy transfer (FRET) imaging, and biochemical approaches, we demonstrated that 5-HT 4 R was expressed in astrocytes of the mouse hippocampus, and that 5-HT 4 R-Gα 13 signaling in astrocytes increased RhoA activity leading to accumulation of filamentous actin structures. We also identified the role of 5-HT 4 R-Gα 13 -RhoA signaling in the regulation of astrocytic morphology. Moreover, electrophysiological experiments in mixed cultures and in acute slices revealed that astrocytic 5-HT 4 R modifies the function of excitatory hippocampal synapses.
On DIV3 the entire plating medium was replaced with 1 ml maintenance medium (49 ml Neurobasal-A medium, 1 ml B-27, 500 μl 200 mM glutamine, 5 μ/ml penicillin, 5 mg/ml streptomycin). With exception of the shRNA experiments, 1/2 of the medium was exchanged on DIV11 with maintenance medium prior to infection of the cells. Astrocytes were used for experiments on DIV14-17. Mixed hippocampal cultures were obtained from dissociated hippocampi of neonatal mice at P0-1 using an optimized protocol (Kobe et al., 2012).
At DIV7 cells were infected with AAVs. Cell cultures were maintained at 37 C in a humidified incubator in a 5% CO 2 atmosphere until they were used for experiments at DIV12. During microscopy, cells were kept in a balanced salt solution containing 115 mM NaCl, 5.4 mM KCl, 1 mM MgCl 2 , 2 mM CaCl 2 , and 20 mM HEPES, adjusted to pH 7.4 and 290 mOsm with glucose.

| Adeno-associated-viruses
The FRET-based biosensor RaichuEV-RhoA was a gift from Michiyuki Matsuda (Kyoto University, Japan). This biosensor contains a YPettagged RhoA covalently linked to a mTurquoise-tagged GTPasebinding domain of RBD-Rhotekin. Upon activation, conformational changes within the biosensor lead to changes in the FRET efficiency between acceptor (YPet) and donor (mTurquoise). Because the donor/ acceptor stoichiometry is 1:1, activation of RhoA can be simply quantified by calculation of the acceptor/donor emission ratio. This improved version of the RhoA biosensor developed by the group of Michiyuki Matsuda (Yoshizaki et al., 2003) was cloned into an adenoassociated-virus (AAV) vector under the control of the murine GFAP promoter and AAVs were produced using the AAV-DJ system. Primary astrocytes were infected with 1 × 10 4 viral genomes per well on To test for an increase in activated RhoA after 5-HT 4 R stimulation, a RhoA G-LISA Activation Assay Kit (#027BK124, Cytoskeleton Inc., Denver, CO) was used according to the manufacturer's protocol.

| Microscopy
Microscopic investigation was performed using Zeiss LSM780 with a LD C-Apochromat ×40/1.2 W objective and Zen2013 imaging software in online-fingerprinting mode with previously defined spectra for each fluorescent protein and dye obtained from single staining. Live cell imaging of RhoA activity in astrocytes with the FRET-based biosensor RaichuEV-RhoA was performed at 37 C in a continuous time series. Z-stacks of the same cell were acquired in both channels every 20 s using Zeiss Definite Focus to maintain focus during long-term imaging. Cells were recorded for 10 min with application of 10 μM BIMU8 or H 2 O after 5 min of imaging.
In experiments with F-and G-actin, astrocytes were labeled with a GFAP antibody, while in the shRNA experiments TurboFP650 was expressed under control of the GFAP promoter. Stimulated emission depletion (STED) imaging was conducted on an Abberior STEDYCON with an Olympus UPlanSApo ×100/1.40 oil objective. Excitation wavelengths were 594 and 640 nm for STAR 580 and STAR RED, respectively, while depletion wavelength was 775 nm. Images were acquired with a pixel size of 25 nm (1,232 × 1,116 pixels, 31 μm × 28 μm) and a pixel dwell time of 20 μs. A total of 17 z-planes were acquired with 1 μm distance. Image analysis was done using Matlab (Mathworks). 125, KCl 10, Na 2 Phosphocreatine 5, EGTA 0.5, MgATP 4, Na 2 GTP 0.3, HEPES 10, was adjusted to pH 7.3 and 290 mOsm. Patch electrodes were pulled to reach a resistance of 3-6 MΩ. Postsynaptic currents were low-pass filtered (2.9 kHz) and digitized at 20 kHz.

| In vitro electrophysiological recordings
The access resistance was monitored throughout the recordings (5 mV steps every 2 min). Recordings with an access resistance of >50 MΩ or a leak current >200 pA were discarded. Miniature excitatory post synaptic currents (mEPSCs) were detected using Matlab (Mathworks) and reviewed manually to check for detection/analysis errors.

| Statistical analysis
Statistical analysis was done using GraphPad Prism7 software.
Two-tailed t-test (paired or unpaired) was applied to determine the statistical difference between two experimental groups. Analysis of variance (ANOVA, one-way or two-way where appropriate) followed by a post hoc test was used for multiple comparisons. A pvalue <.05 was considered statistically significant within each test.

| Hippocampal astrocytes express 5-HT 4 R in vivo
To understand the functional role of 5-HT 4 R in astrocytes, we first investigated whether this receptor was expressed on hippocampal astrocytes. Immunohistochemical staining revealed 5-HT 4 R localization on cells positive for the astrocytic marker S100β within the hippocampal formation of the adult mouse brain (Figure 1a-c). We confirmed antibody specificity using a corresponding blocking peptide, and by staining of hippocampal slices prepared from 5-HT 4 R-ko mice (Figure S1a-c). Since it appeared not all astrocytes were expressing 5-HT 4 R, we quantified the percentage of astrocytes positive for the 5-HT 4 R in the hippocampus and in the other brain regions. We found that 33 ± 12% of astrocytes in the hippocampus express the 5-HT 4 R, while in other brain regions percentage of 5-HT 4 R positive astrocytes was lower (cortex 9 ± 3%, midbrain 14 ± 9%, thalamus 11 ± 3%, and hypothalamus 12 ± 2%. Figure S1d). Of note, receptor expression within the hippocampus was similar throughout different hippocampal structures ( Figure S1e). We also observed differences in astrocytic and neuronal receptor expression, with higher fluorescence intensity detected from the somata of hippocampal neurons (Figure 1a,b). In astrocytes, 5-HT 4 R was expressed on both somata and astrocytic protrusions, although the receptor distribution was heterogeneous, with several protrusions lacking receptor expression (Figure 1d).
To acquire more detailed information regarding 5-HT 4 R distribution, we also performed STED imaging of S100β-positive astrocytes in fixed hippocampal slices. STED microscopy revealed that 5-HT 4 R appeared to form separated clusters with a mean size of 136 ± 21 nm (Figure 1e,f).

| Cultured hippocampal astrocytes are a suitable model for investigating 5-HT 4 R-signaling
After demonstrating 5-HT 4 R expression in hippocampal astrocytes in vivo, we next investigated the role of 5-HT 4 R-mediated signaling in the regulation of astrocytic morphology and function. As a model system, we utilized mouse primary hippocampal astrocyte cultures, raising stellate astrocytes (Wu et al., 2014).
Astrocytes can show an altered protein expression profile in vitro depending on the culture model used (Hertz, Chen, & Song, 2017); therefore, we first confirmed the presence of 5-HT 4 R on the cultured cells. The cultured astrocytes showed pronounced 5-HT 4 R expression ( Figure 2a,b). Additionally, antibody specificity was verified using F I G U R E 1 Astrocytes express the 5-HT 4 R in vivo. (a) Expression of the 5-HT 4 R in S100β-positive cells throughout the mouse brain shown by immunohistochemical staining. (b) Magnification of hippocampal structures shown in (a) revealed expression of the 5-HT 4 R in hippocampal astrocytes. Scale bar 50 μm. (c) Split channels for indicated area in (b). White arrows exemplarily point to 5-HT 4 R expressing astrocytes. Scale bar 50 μm. (d) 3D representation of a hippocampal astrocyte visualized by S100β and 5-HT 4 R staining. 5-HT 4 R was heterogeneously distributed. The grey arrow points towards a branch with less 5-HT 4 R staining than the one marked by the white arrow. Scale bar 50 μm. (e) STED microscopy images showed clustering of 5-HT 4 R on S100β-positive cells. The dotted white line represents the astrocyte outline identified by S100β labeling. Scale bar 2 μm. (f) Magnification of an astrocyte process shown in (e). Scale bar 1 μm [Color figure can be viewed at wileyonlinelibrary.com] several approaches, including pre-treatment with the corresponding blocking peptide, and the staining of astrocytes isolated from 5-HT 4 R-ko mice (Figures 2a,b and S2a,d). We also assessed the intracellular distribution of 5-HT 4 R in cultured astrocytes and found a high degree of co-localization with markers for the plasma membrane (i.e., PMCA) and the endoplasmatic reticulum (i.e., Calreticulin), while there was only partial co-localization with the Golgi marker (i.e., TGN38) and no evidence for co-localization with the lysosomal marker Lamp-1 ( Figure S2b,c).
F I G U R E 2 Cultured astrocytes as a model to investigate 5-HT 4 R signaling. (a) Immunocytochemical labeling of 5-HT 4 R astrocyte marker GFAP-positive cells in primary hippocampal cultures from WT mice. No signal of the 5-HT 4 R was detected in astrocytes from 5-HT 4 R-ko animals. (b) Visualization of 5-HT 4 R and S100β protein in cultured astrocytes from WT and 5-HT 4 R-ko mice. Scale bars in (a) and (b) 50 μm (left overview) and 20 μm (right magnification). (c) Schematic illustration of established 5-HT 4 R signaling pathways. Upon activation, the 5-HT 4 R induced signaling via Gα S or Gα 13 heterotrimeric G proteins, leading to cAMP upregulation via adenylyl cyclase (AC) or RhoA activation and subsequent actin cytoskeleton reorganization, respectively. (d) Relative mRNA expression levels of 5-HT 4 R and its down-stream effectors in primary hippocampal astrocyte cultures from WT and 5-HT 4 R-ko mice. Statistical significance was evaluated using unpaired two-tailed t test, N = 4 independent cultures [Color figure can be viewed at wileyonlinelibrary.com] As described above, 5-HT 4 R couples with the Gα S and Gα 13 heterotrimeric G proteins, thereby regulating cAMP levels and modulating the conformation of the actin cytoskeleton, respectively ( Figure 2c). To ensure that the proteins involved in 5-HT 4 R-mediated signaling were also available in the cultured hippocampal astrocytes, we performed RT-qPCR and western blot analysis. Both methods F I G U R E 3 Legend on next page. confirmed the expressions of 5-HT 4 R, Gα S , Gα 13 , and the small GTPase RhoA in the cultured astrocytes. Importantly, astrocytic cultures isolated from 5-HT 4 R-ko mice exhibited no differences in the expression levels of Gα S , Gα 13 , or RhoA (Figure 2d).

| 5-HT 4 R activation increases RhoA activity
To investigate the impact of 5-HT 4 R on astrocyte morphology, we focused on the 5-HT 4 R-Gα 13 -RhoA-actin signaling axis, which is a known regulator of cellular morphology (Kvachnina, 2005). We investigated whether this signaling pathway was preserved in hippocampal astrocytes using an ELISA-based RhoA activation assay. As shown in Figure 3a, cell treatment with the 5-HT 4 R-selective agonist BIMU8 resulted in significantly increased RhoA activity compared with the H 2 O-treated control.
This increase was not detected in astrocytes isolated from 5-HT 4 R-ko mice. Interestingly, the basal RhoA activity in 5-HT 4 R-deficient astrocytes was 1.6-fold higher than in astrocytes isolated from the hippocampus of WT animals (Figure 3a).
We then used the FRET-based biosensor RaichuEV-RhoA to moni-

| 5-HT 4 R-mediated signaling regulates actin cytoskeleton reorganization in astrocytes
RhoA activity plays a key role in actin cytoskeleton reorganization. To examine whether 5-HT 4 R-induced RhoA activation modulated actin filament dynamics, we compared the ratio of filamentous actin (F-actin) to globular actin (G-actin) in astrocytes isolated from WT and 5-HT 4 R-ko mice. F-actin and G-actin were visualized by staining astrocytes with phalloidin-TRITC and DNase I-Alexa Fluor 488, respectively. Based on the ratiometric overlay of F-actin and G-actin emissions, we visualized the F-actin fraction, which ranged from 0 (low F-actin, high G-actin) to 1 (high F-actin, low G-actin) (Figure 4a). WT astrocytes treated with BIMU8 for 30 min exhibited a significant increase of the F-to G-actin ratio, from 1.6 ± 0.7 in control cells to 3.1 ± 1.3 in BIMU8-treated astrocytes (Figure 4b). This effect was 5-HT 4 R-specific, as it did not occur in astrocytes from 5-HT 4 R-ko mice. Furthermore, astrocytes from 5-HT 4 Rko mice exhibited a higher basal F-actin to G-actin ratio (2.5 ± 0.9 in control cells, and 2.2 ± 0.9 after BIMU8 treatment) (Figure 4a,b).
Border-distance plots provided further information about the cellular distribution of F-and G-actin. Figure 4c shows the intensity distributions of F-and G-actin structures, as well as the F-actin to G-actin fraction, in WT and 5-HT 4 R-ko astrocytes as a function of distance from the cell border. Astrocyte treatment with BIMU8 increased the F-actin fraction in WT astrocytes, with the highest values observed near the plasma membrane.
Comparison of F-actin and G-actin intensities revealed that the increased F-actin fraction was due to an increased F-actin level rather than a decrease of G-actin signal. In non-stimulated cells, we observed more F-actin than G-actin structures near the plasma membrane, and more G-actin than F-actin structures toward the cell center. In contrast, 5-HT 4 R-deficient astrocytes exhibited a higher F-actin fraction throughout the whole cell at baseline, and no change in the F-actin fraction upon BIMU8 treatment (Figure 4c). Examination of astrocytic morphology by Sholl analysis revealed that BIMU8 treatment reduced arborization complexity (Figure 4d), indicating a general retraction of astrocytic protrusions upon 5-HT 4 R activation (see also greycolor contour image inset in Figure 4a).
In summary, our results indicated that 5-HT 4 R-mediated signaling in astrocytes resulted in an increased relative abundance of filamentous actin, and thereby influenced cell morphology.

| 5-HT 4 R signaling in astrocytes is G proteindependent
As mentioned above, 5-HT 4 R can activate both Gα 13 and Gα S proteins; thus, we next aimed to determine which G protein primarily F I G U R E 3 5-HT 4 R stimulation leads to RhoA activation. (a) RhoA activity determined by RhoA activation assay in primary astrocyte cultures from WT and 5-HT 4 R-ko mice after 5 min stimulation with 5-HT 4 R-agonist BIMU8 or H 2 O (control). WT astrocytes showed higher RhoA activity when treated with BIMU8 compared with H 2 O treated cells, while 5-HT 4 R-ko astrocytes displayed higher basal RhoA activity which was not changed upon BIMU8 application. Statistical significance was calculated using two-way ANOVA with Sidak's multiple comparisons post hoc test, N = 6 (WT) and N = 3 (5-HT 4 R-ko) independent cultures. The silencing of Gα 13 protein in astrocytes significantly increased the F-actin fraction without 5-HT 4 R stimulation, while silencing of Gα S did not influence the basal actin cytoskeleton composition (Figure 5a,b). Similar to control cells, Gα S -deficient astrocytes exhibited a significantly increased F-actin fraction following treatment with the 5-HT 4 R agonist BIMU8 (Figure 5a,b). In contrast, in Gα 13 -F I G U R E 4 5-HT 4 R-activation leads to actin reorganization. (a) Representative images showing F-actin fraction in fixed hippocampal astrocytes. In WT astrocytes the stimulation with BIMU8 (30 min, 10 μM) resulted in an increase in the F-actin fraction. In 5-HT 4 R-ko astrocytes the initial fraction was higher and addition of BIMU8 had no effect. The cells from the excerpts indicated by white boxes are additionally shown uniformly in grey color on the right, to emphasize the cell area and branching. Scale bar 20 μm. (b) The quantification reflected an increase in the F-to G-actin ratio upon BIMU8 treatment in WT but not in 5-HT 4 R-ko astrocytes. Statistical differences were calculated using one-way ANOVA with Dunnett's multiple comparisons post hoc test, n ≥ 25 cells, N = 3. (c) Border-distance plots indicate the gradient of F-and G-actin intensities from the outer cell border to the center with a proportional shift upon BIMU8 treatment in WT (upper plot) but not in 5-HT 4 R-ko astrocytes (lower plot), n ≥ 25 cells, N = 3. (d) Sholl analysis reflected a reduction in arborization complexity in BIMU8 treated astrocytes (Area under the curve [AUC] 224 ± 15) compared with control cells (AUC 270 ± 15). Unpaired t-test with Welch's correction, p = .038, n ≥ 31 cells, N = 3. Arrows point towards direction of change [Color figure can be viewed at wileyonlinelibrary.com] deficient astrocytes, which exhibited a high basal F-actin fraction, receptor stimulation with BIMU8 induced a significant reduction of filamentous actin (Figure 5a,b). These findings indicated a balance between Gα S and Gα 13 protein-mediated signaling under basal conditions, and upon 5-HT 4 R activation, which exerted a bidirectional influence on the actin cytoskeleton in astrocytes.
F I G U R E 5 Impact of 5-HT 4 R activation on the actin cytoskeleton is G protein dependent. (a) Representative images showing F-actin fraction in cultured hippocampal astrocytes infected with AAVs encoding shRNA scramble, shRNA against Gα 13 or shRNA against Gα S proteins. Stimulation with 5-HT 4 R agonist BIMU8 (30 min, 10 μM) increased the F-actin fraction under control conditions as well as in cells expressing less Gα S . If Gα 13 expression was downregulated, this effect was reversed and stimulation of the receptor led to restoration of levels similar to the control. The cells from the excerpts indicated by white boxes are additionally shown uniformly in grey color on the right, to emphasize the cell area and branching. Scale bar 20 μm. (b) Quantification of the F-actin fraction. Statistical differences were calculated using two-way ANOVA with Tukey's multiple comparisons post hoc test, n ≥ 101 cells, N = 3. (c) Sholl analysis revealed a reduction in astrocyte complexity upon BIMU8 stimulation. Knockdown of Gα S by shRNA enhanced this effect, while stimulation had the opposite effect in astrocytes with downregulated Gα 13 protein levels, leading to increased complexity. Two-way ANOVA with Tukey's multiple comparisons test. shRNA-scr and shRNA-scr + BIMU8: radius 18 μm p = .0195; shRNA-Gα 13 and shRNA-Gα 13 + BIMU8: radius 70 μm p = .0045; shRNA-Gα S and shRNA-Gα S + BIMU8: radius 46 μm p = .0007; n ≥ 25 cells, N = 3. Arrows point out direction of change upon BIMU8 treatment. (d) Quantification of astrocyte size (2D) showed the impact of G protein signaling in regulation of cell size. Astrocytes occupied more space when expressing shRNA against Gα 13 compared with astrocytes expressing shRNA against Gα S . This difference was more prominent upon stimulation with 5-HT 4 R agonist BIMU8. Statistical differences were calculated using two-way ANOVA with Tukey's multiple comparisons post hoc test, n ≥ 98 cells, N = 3 [Color figure can be viewed at wileyonlinelibrary.com] We further performed Sholl analysis to confirm the dual role of Gα S and Gα 13 in regulating astrocytic morphology. At 30 min after treatment with BIMU8, astrocytes expressing shRNA-scramble or shRNA against Gα S showed reduced arborization complexity ( Figure 5c). In contrast, in astrocytes with downregulated Gα 13 , BIMU8 treatment led to increased morphological complexity and an altered cell size. Under basal conditions, astrocytes expressing shRNA against Gα 13 were larger than astrocytes expressing shRNA against Gα S (0.206 ± 0.007 mm 2 vs. 0.172 ± 0.007 mm 2 ). These size differences were even more pronounced after 5-HT 4 R stimulation with BIMU8 (0.224 ± 0.009 mm 2 vs. 0.158 ± 0.005 mm 2 ; Figure 5d, see also grey-color contour image inset in Figure 5a). To further confirm these effects to be directly mediated by the Gα 13 -RhoA signaling axis, we repeated these experiments in presence of Y-27632, a cell-permeable, highly potent, and selective inhibitor of the RhoA downstream effector Rho associated kinase (ROCK) ( Figure S4). In the presence of this ROCK inhibitor, the 5-HT 4 R agonist BIMU8 failed to change the relative F-actin fraction ( Figure S4a,b) and the branching pattern of astrocytic processes ( Figure S4c,d). This indicates that ROCK mediates the overall effect of 5-HT 4 R activation on both.
Taken together, the emerging picture is that activation of astrocytic 5-HT 4 R increases RhoA activity (Figure 3), which increases the relative abundance of F-over G-actin (Figure 4) and decreases the arborization of astrocytes ( Figure 5) via ROCK signaling ( Figure S4).

| 5-HT 4 R activation in astrocytes changes excitatory synaptic transmission in vitro and in situ
Activation of 5-HT 4 R reportedly causes a long-lasting increase in the excitability of hippocampal neurons (Mlinar, Mascalchi, Mannaioni, Morini, & Corradetti, 2006), converts a weak synaptic potentiation into persistent LTP in the CA1 area (Matsumoto et al., 2001) and directly potentiate CA3-CA1 synapses (Teixeira et al., 2018). Since astrocytes and neurons are both involved in regulating multiple brain functions, including synaptic transmission, it is crucial to understand whether 5-HT 4 R activation in astrocytes could have functional consequences on the neuronal network. To investigate the possible role of astrocytic 5-HT 4 R on synaptic activity, we prepared primary hippocampal mixed cultures from 5-HT 4 R-ko mice. In these cultures, we selectively rescued 5-HT 4 R expression in either neurons or astrocytes, using AAV vectors encoding 5-HT 4 R under control of a synapsin or GFAP promoter, respectively. The control cell cultures were infected with AAVs encoding tdTomato under control of a synapsin or GFAP promoter (Figures 6a and S5a). We verified the selective 5-HT 4 R expression in neurons or astrocytes by immunofluorescence staining with antibodies against GFAP (astrocytic marker) or βIII-tubulin (neuronal marker), respectively ( Figure S5b). Whole-cell patch-clamp recordings were obtained from cultured neurons to compare the influence of 5-HT 4 R activation on mEPSCs between those conditions. In the control condition (5-HT 4 R-ko cultures expressing tdTomato), application of BIMU8 did not change the mEPSC frequency (Figure 6a,b). Receptor stimulation with BIMU8 also did not change the mEPSC frequency in cells with selective rescue of neuronal 5-HT 4 R expression (Figure 6a,b). In contrast, 5HT 4 R activation induced a significantly increased mEPSC frequency within minutes in cells with selective rescue of 5-HT 4 R expression in astrocytes (Figure 6a,b). To determine whether this effect was mediated via the 5-HT 4 R-Gα 13 -RhoA-ROCK signaling pathway, we performed electrophysiological recordings in the presence of the ROCK inhibitor Y-27632. Although Y-27632 affects RhoA signaling in both neurons and astrocytes, our observation, that the selective rescue of 5-HT 4 R expression in neurons did not influence BIMU8-mediated changes in mEPSC (Figure 6b), demonstrates that the 5-HT 4 R-RhoA-ROCK signaling in neurons is not implemented in BIMU8-mediated EPSC frequency response. Therefore, the effect of Y-27632 was analyzed only in cultures where 5-HT 4 R expression was selectively rescued in astrocytes. In these experiments, no increase of the mEPSC frequency was observed after BIMU8 stimulation in 5-HT 4 R expressing astrocyte cultures ( Figure 6b). We observed a nonspecific rundown of mEPSC amplitudes, which was present in all groups to the same extent. These results suggest that astrocytic 5-HT 4 R modulated an increase of spontaneous neurotransmitter release at glutamatergic synapses in cultured cells mediated by the 5-HT 4 R-Gα 13 -RhoA-ROCK signaling.
Finally, we examined the role of astrocytic 5-HT 4 R in glutamatergic synaptic transmission in the hippocampus in situ. To this end, we stereotactically injected 5-HT 4 R-ko mice in the hippocampal CA1 region of one hemisphere with a control AAV-GFAP-tdTomato construct. Each mouse was also injected in the CA1 hippocampus of the other hemisphere with AAV-GFAP-5-HT 4 R-eGFP to selectively rescue astrocytic 5-HT 4 R expression (Figure 7a and S6a-c). Three weeks after stereotactic injection, acute hippocampal slices were prepared. AAV-mediated rescue of 5-HT 4 R expression resulted in a heterogeneous expression pattern similar to that obtained for the endogenous receptor ( Figure S6d). We identified the slices containing transfected astrocytes using two-photon excitation fluorescence microscopy ( Figure 7b). Next, fEPSPs were evoked by electrical stimulation of CA3-CA1 Schaffer collateral axons and recorded in the CA1 stratum radiatum near astrocytes expressing either 5-HT 4 R-eGFP (rescue) or tdTomato (knockout). Basal synaptic transmission did not differ between the rescue and knockout slices (Figure 7c), and the fEPSP slope was not affected by bath application of 10 μM BIMU8 for 20 min (fEPSP slope normalized to pre-drug baseline: knockout, 99.7 ± 2.93%, n = 6, p = .919; rescue, 97.7 ± 4.88%, p = .658, n = 8; paired t-tests; not illustrated). However, the paired-pulse-ratio (PPR) at short inter-stimulus intervals of 25 ms was significantly reduced in slices with rescued astrocytic 5-HT 4 R expression compared with control slices (Figure 7d). Thus, the selective rescue of 5-HT 4 R expression in astrocytes affected glutamate release at these synapses. Again, the rescue of 5-HT 4 R had no effect when recordings were performed in the presence of the ROCK inhibitor Y-27632 (Figure 7e

| DISCUSSION
The role of serotonergic signaling in the brain has been extensively studied over the last decades. However, such research has been largely focused on the impact of serotonin receptors on neurons. As increasing evidence highlights the importance of astrocytes in regulating physiological brain functions and in many neurological disorders, these cells must be considered important contributors to diseases involving serotonergic signaling changes (Lundgaard, Osório, Kress, Sanggaard, & Nedergaard, 2014;Miyazaki & Asanuma, 2016;Peng, Verkhratsky, Gu, & Li, 2015). Moreover, astrocytes represent an essential part of the multipartite synapse. Thus, the expression of serotonin receptors on astrocytes in brain regions with serotonergic innervation might represent an important, yet largely unexplored, signaling pathway.
Our present findings confirmed the astrocytic expression of 5-HT 4 R (Boisvert et al., 2018;Cahoy et al., 2008;Parga et al., 2007) and demonstrated that this receptor was expressed on a subset of astrocytes. These observations are in line with multiple studies revealing astrocyte heterogeneity in terms of morphology and function (Farmer & Murai, 2017;Matyash & Kettenmann, 2010;Oberheim, Goldman, & Nedergaard, 2012). The distinct expression pattern may also result from the defined structures of serotonergic projections within the hippocampus. Moreover, we observed irregular and clustered distributions of 5-HT 4 R within single astrocytes, which could be F I G U R E 6 5-HT 4 R activation in astrocytes impacts neuronal signaling. (a) Illustration of the experimental setup with three conditions. 5-HT 4 R-ko mice were used for preparation of mixed hippocampal cultures (HCC) and the receptor was either rescued in astrocytes or neurons using viral infection of AAV-mGFAP-5-HT 4 R-eGFP or AAV-syn-5-HT 4 R-eGFP, respectively. In a third (control) condition cells were infected with AAV-tdTomato. Representative traces of electrophysiological recordings of neurons in those cultures are shown on the right for all three conditions. Only rescued expression of the 5-HT 4 R in astrocytes, but not neurons, influenced mEPSC frequency after BIMU8 stimulation. (b) Quantification of mEPSC frequency in hippocampal neurons after application of 10 μM 5-HT 4 R agonist BIMU8. Increase of mEPSC frequency was only present if astrocytes expressed 5-HT 4 Rs, but not if the receptor was only present in neurons. No change was observed upon BIMU8 stimulation in a 5-HT 4 R-ko culture or when RhoA signaling was blocked by cell-permeable, highly potent ROCK inhibitor Y-27632 (50 μM). Statistical analysis was performed using two-way ANOVA with Sidak's multiple comparisons post hoc test, n ≥ 7 [Color figure can be viewed at wileyonlinelibrary.com] F I G U R E 7 Astrocytic 5-HT 4 R impact neuronal properties in vivo. (a) Schematic overview of acute slice recordings. 5-HT 4 R-ko mice were stereotactically injected with AAV-GFAP-5-HT 4 R-eGFP and AAV-GFAP-tdTomato into separate hemispheres to selectively restore astroglial 5-HT 4 Rs in the hippocampus. After 3 weeks acute slices were subjected to electrophysiological investigations. fEPSPs were evoked by electrical stimulation of CA3-CA1 Schaffer collaterals and recorded in the stratum radiatum (S.R.) of the CA1 region. (b) Astroglial expression of AAVs was visualized using two-photon excitation fluorescence microscopy. Transfected astrocytes were uniformly distributed across stratum oriens (S.O.) and S.R. Unlike cytosolic tdTomato the 5-HT 4 R-eGFP fusion protein was located to astrocytic membranes. Stratum pyramidale (S.P.) indicated for orientation. (c) Representative fEPSPs recorded in S.R. with gradually increasing stimulation intensities (left panel, examples for 50, 100, 200 μA). fEPSP slopes recorded near 5-HT 4 R-eGFP expressing astrocytes (green) or areas covered by tdTomato expressing astrocytes (red) were not statistically different (right panel, p = .87, two-way repeated measures ANOVA, eGFP n = 11, tdTomato n = 14). (d) Representative pairs of fEPSPs recorded with an inter-stimulus interval of 25 ms (left panel). The paired-pulse ratio (PPR) obtained at short inter-stimulus intervals (ISI) was lower when astroglial 5-HT 4 R expression was rescued compared with tdTomato-positive control slices (p = .0084, two-way repeated measures ANOVA, eGFP n = 10, tdTomato n = 14, Tukey post hoc test: 25 ms, p = .0065, **50 ms, p = .145). (e-f) The experiments were repeated in the presence of ROCK inhibitor Y-27632 (20 μM). The fEPSP slopes recorded near 5-HT4R-eGFP-expressing astrocytes were not significantly different from fEPSPs recorded close to tdTomato-expressing astrocytes (p = .225, two-way repeated measures ANOVA, eGFP n = 13, tdTomato n = 12). Also the PPR was not different between these conditions when ROCK was blocked (p = .873, two-way repeated measures ANOVA, eGFP n = 13, tdTomato n = 12) [Color figure can be viewed at wileyonlinelibrary.com] temporally or structurally correlated, as recently demonstrated for the GABA transporter GAT-3 (Boddum et al., 2016;Mederos, González-Arias, & Perea, 2018). Our results indicated that activation of the 5-HT 4 R-Gα 13 signaling pathway induced activation of the small GTPase RhoA in astrocytes, which led to an increased F-actin fraction and morphological changes. The observed morphological changes could be critically involved in synapse formation, maintenance, and plasticity (Chung et al., 2015;Dallérac, Zapata, & Rouach, 2018;Haber, Zhou, & Murai, 2006).
We found that treatment with the selective 5-HT 4 R agonist BIMU8 induced an increase in filamentous actin structures. Moreover, this effect was absent in 5-HT 4 R-ko astrocytes, demonstrating 5-HT 4 R specificity. Interestingly, cultured astrocytes from 5-HT 4 Rko mice exhibited altered properties under basal conditions: elevated RhoA activity and an increased basal F-actin fraction. The same phenomenon was observed following the downregulation of Gα 13 proteins in astrocytes from WT mice. We propose that under basal conditions, 5-HT 4 R is negatively coupled to Gα s and Gα 13 proteins, with the latter being the preferential partner ( Figure S7). Upon receptor knockout, Gα 13 becomes disinhibited leading to activation of the RhoA cascade and increased F-actin formation. Knockingdown the Gα 13 subunit shifts a negative coupling of 5-HT 4 R towards Gα s , which can lead to decreased RhoA phosphorylation via cAMPdependent kinases resulting in RhoA disinhibition (see below).
Receptor stimulation with an agonist activates the Gα 13 -RhoA signaling pathway and boosts F-actin formation ( Figure S7). In case of Gα 13 knockdown, receptor stimulation preferentially induces a Gαsmediated rise of cAMP levels, leading to increased RhoA phosphorylation by PKA (Jaffe & Hall, 2005;Kim et al., 2018). As consequence, RhoA-mediated activation of ROCK kinase is attenuated, leading to inhibition of F-actin formation. Supporting this view, a new concept, in which Gα S signaling negatively regulates guanine nucleotide exchange factors (GEFs) that are responsible for activating the small GTPases Cdc42 and Rac1 was recently presented by Sugiyama and co-workers (Sugiyama et al., 2017). Since low Cdc42 and Rac1 activity are reportedly accompanied by elevated RhoA activity (Bishop & Hall, 2000;Chauhan, Lou, Zheng, & Lang, 2011), chronic upregulation of active RhoA may account for the presently observed increase in F-actin.
As mentioned above, in astrocytes with diminished Gα 13 signaling, stimulation with BIMU8 drastically decreased the elevated basal F-actin to G-actin ratio, possibly via signaling through Gα S or G protein-independent pathways. Several possible links between Gα S signaling and the actin cytoskeleton have been proposed, including acute phosphorylation of RhoA at serine residue 188 by the cyclic nucleotide-dependent protein kinase A (PKA), which directly inhibits RhoA activity by enhancing its interaction with RhoGDI and its attendant withdrawal from the plasma membrane (Ellerbroek, Wennerberg, & Burridge, 2003;Forget, Desrosiers, Gingras, & Béliveau, 2002;Tkachenko et al., 2011). Additionally, RhoA phosphorylation by cAMP-dependent kinases causes decreased interaction with the effector protein Rho-associated coiled-coil-containing kinase (ROCK) (Nusser et al., 2006;Takemoto, Ishihara, Mizutani, Kawabata, & Haga, 2015), which regulates actin dynamics and cell morphology via numerous effector proteins, including ROCK, LIMkinases, Cofilin, F-actin, and the regulatory light chain of myosin (MLC) (Jaffe & Hall, 2005;Kim et al., 2018). Several reports describe antagonistic roles of RhoA and cAMP in regulating cellular morphology (Bandtlow, 2003;Dong, Leung, Manser, & Lim, 1998). Moreover, RhoA phosphorylation reduces ROCK interaction-without altering the interactions of other effector protein, including rhotekin, mDia-1, and protein kinase N (PKN)-thus, adding an additional switch and increasing the complexity of actin regulation (Nusser et al., 2006). Downregulation of Gα S protein levels in astrocytes did not affect 5-HT 4 R-mediated regulation of F-actin dynamics and cell morphology neither under basal conditions nor after receptor stimulation. Therefore, we conclude that under physiological conditions, 5-HT 4 R signaling influences actin cytoskeleton composition predominantly via the Gα 13 -RhoA signaling pathway. Furthermore, our findings indicated that actin dynamics are in a balanced state, and that 5-HT 4 R activity can bidirectionally regulate the turnover of F-actin and G-actin structures. Bidirectional control of actin structures is reportedly important for the regulation of spine morphology in hippocampal neurons, albeit through different mechanisms (Okamoto, Nagai, Miyawaki, & Hayashi, 2004). We also demonstrated that astrocyte morphology underlies the regulation of actin by serotonergic signaling, which adds a new player in the debate about regulation of astrocyte appearance (Matyash & Kettenmann, 2010;Oberheim et al., 2012;Zhou, Zuo, & Jiang, 2019).
It has been suggested that 5-HT 4 R activity might affect overall network excitability (Schill et al., 2020). We investigated whether the manipulation of 5-HT 4 R signaling in astrocytes affected glutamatergic synaptic transmission. In dissociated cultures, pharmacological activation of 5-HT 4 Rs quickly increased the frequency of spontaneous synaptic glutamate release (mEPSCs). Because this happened on a timescale of 10 min, this effect was likely due to an acute increase in the probability of spontaneous release from existing presynaptic glutamatergic terminals, and not synaptogenesis. The rapid onset of the increase was in line with the fast increase of RhoA activity after 5-HT 4 R activation.
In situ, we recorded fEPSPs evoked by single stimulations of CA3-CA1 axons, and found that they were not changed by the rescue of astrocytic 5-HT 4 R expression or by the acute pharmacological activation of the restored 5-HT 4 Rs by BIMU8. There are at least two possible explanations for the absence of an acute effect. First, the modulation of synaptic transmission via astrocytic 5-HT 4 Rs in situ may occur on a longer timescale than in vitro, for example, over hours or days. Second, 5-HT 4 Rs might be fully activated by ambient serotonin in the slice preparation, which would prevent effects of direct receptor activation. The latter could indicate that astrocytic  Rs are already fully activated in acute slices. However, we also detected a reduced paired-pulse facilitation of synaptic transmission at short inter-stimulus intervals after rescuing astrocytic 5-HT 4 R expression. Since the paired-pulse facilitation at CA3-CA1 synapses is thought to reflect changes of presynaptic glutamate release (Debanne, Guérineau, Gähwiler, & Thompson, 1996;Dobrunz & Stevens, 1997 and references therein), this observation suggests that astrocytic 5-HT 4 Rs modify presynaptic release (i.e., synapses with a high initial release probability show low paired-pulse facilitation, as in our case). This reduced paired-pulse facilitation was not due to an increase of the overall basal presynaptic release probability, because we did not observe a corresponding increase of the slope of single fEPSPs. Rather, it is likely that astrocytic 5-HT 4 Rs modify rapid repetitive release. Both in situ and in vitro effects mediated by astrocytic 5-HT 4 Rs were not observed when ROCK signaling was inhibited. Therefore, astrocytic 5-HT 4 Rs can modify synaptic glutamate release via ROCK signaling both in dissociated cells and organized tissue, although the change of the mEPSC frequency in vitro and of the paired-pulse facilitation in situ do not point towards a specific presynaptic mechanism. Regarding the latter, it needs to be noted that viral delivery of 5HT 4 R to hippocampal astrocytes does not lead to transduction of all astrocytes ( Figure 7) and may not reproduce the relatively sparse endogenous 5HT 4 R expression pattern ( Figure S1), which could explain differences in the presynaptic effects between cultures and acute slices. Optical quantal analysis of release at single spines (Oertner, Sabatini, Nimchinsky, & Svoboda, 2002) near astrocytic processes positive and negative for 5-HT 4 R could help to dissect the involved mechanism in future. In combination with pharmacology, this approach could also be useful to identify the involved signaling molecules (Alfonso Araque et al., 2014;Rusakov et al., 2014) that link astrocytic 5-HT 4 R, RhoA, ROCK and presynaptic function.

ACKNOWLEDGMENT
The RaichuEV-RhoA biosensor was provided by Michiyuki Matsuda, which we greatly appreciate. This study was supported by the German Research Foundation (DFG) (PO732 to E. P., ZE994/2 to A. Z., SFB1089 B03, FOR2795, SPP1757 HE6949/1, HE6949/3 to C. H.), Lobachevsky University 5-100 academic excellence program to E.P., and SFB870 B05 to C. W. S. This manuscript is part of the PhD thesis of F. E. M. Open access funding enabled and organized by Projekt DEAL.

DATA AVAILABILITY STATEMENT
The data that supports the findings of this study are available in the