Astrocyte Specific Remodeling of Plasmalemmal Cholesterol Composition by Ketamine Indicates a New Mechanism of Antidepressant Action

Ketamine is an antidepressant with rapid therapeutic onset and long-lasting effect, although the underlying mechanism(s) remain unknown. Using FRET-based nanosensors we found that ketamine increases [cAMP]i in astrocytes. Membrane capacitance recordings, however, reveal fundamentally distinct mechanisms of effects of ketamine and [cAMP]i on vesicular secretion: a rise in [cAMP]i facilitated, whereas ketamine inhibited exocytosis. By directly monitoring cholesterol-rich membrane domains with a fluorescently tagged cholesterol-specific membrane binding domain (D4) of toxin perfringolysin O, we demonstrated that ketamine induced cholesterol redistribution in the plasmalemma in astrocytes, but neither in fibroblasts nor in PC 12 cells. This novel mechanism posits that ketamine affects density and distribution of cholesterol in the astrocytic plasmalemma, consequently modulating a host of processes that may contribute to ketamine’s rapid antidepressant action.

cAMP did not affect vesicles that underwent full fission from the plasmalemma. The mean diameter and the relative proportion of larger endocytotic vesicles were comparable in untreated and cAMP-treated cells ( Fig. 2c). In contrast, immunolabelling of early endosomes with the antibody against early endosomal marker EEA1 and vesicle loading with fluorescent dextrans revealed a cAMP-mediated effect on early endosomes. EEA1-positive endosomes were significantly larger after cAMP treatment ( Supplementary Fig. 2a-c). Similarly, significantly larger dextran-loaded vesicles were observed after cAMP treatment when compared to controls; after 15 min ( Supplementary Fig. 2d-f) and 3 h incubation, respectively (Supplementary Fig. 2g-i). These results suggest that cAMP stimulates intracellular fusion between early endosomes, similar to the cAMP-mediated intracellular vesicle-to-vesicle fusion reported in endocrine cells 26 . www.nature.com/scientificreports www.nature.com/scientificreports/ cAMP Favors an open fusion pore state of astroglial vesicles. In vesicles undergoing transient fusion/fission, the dwell time of an open or closed fusion pore can be determined (Fig. 3a); the open pore dwell time defines the time during which exocytotic cargo can be released from the vesicle lumen, whereas the closed pore dwell time defines the time of no release. Figure 3a depicts a representative recording of transient fusion and transient fission that can be interpreted as follows: at first, the baseline (dotted line) represents a state where vesicle "1" has a closed pore and vesicle "2" has an open pore. During the transient fusion, vesicle "1" opens and then closes its fusion pore to return to "baseline". Later on in the recording, vesicle "2" closes its pore during transient fission and then reopens it to return to the "baseline". The dwell times measured in this study ranged from 0.016 to 12.3 s. The open pore dwell time of 0.404 ± 0.087 s (n = 95) in controls was (P < 0.001) shorter than in cAMP-treated cells (1.03 ± 0.167 s, n = 106; Fig. 3b); this change may facilitate vesicle cargo discharge. The closed pore dwell time of 0.673 ± 0.161 s (n = 126) in controls was longer (P < 0.05) than in cAMP-treated cells (0.341 ± 0.079 s, n = 96; Fig. 3c). cAMP prolongs the open fusion pore configuration and shortens the closed fission pore configuration, which most likely facilitates vesicle cargo discharge. Although prolonged fusion dwell times could be related to an increase in full fusion events, our analysis indicated that cAMP had no effect on the frequency of full (or transient) fusions.
In contrast to ketamine 18 , cAMP increases vesicle fusion pore conductance that can be measured when discrete events in the imaginary admittance signal (I m ) are projected to the real admittance signal (R e ) 38 . The projection is the result of the formation of a narrow fusion pore that acts as an additional resistive element in the equivalent electrical circuit. Due to the detection limit 39 , the electrophysiological method enabled us to record fusion pores with conductance between 16.6 and 4201 pS, i.e. fusion pores with estimated diameter between 0.5 and 8.7 nm (Fig. 4a,b). Fusion pores with diameter >8.7 nm did not cause projections to R e and were thus undetectable. In controls, the proportion of transient projected exocytotic events was 84% (n = 185/221), whereas in cAMP-treated cells, this proportion was reduced to 66% (n = 133/202) (Fig. 4c). Thus, the proportion of transient fusion pores with diameter >8.7 nm increased from 16% in controls to 34% in cAMP-treated cells. Correspondingly, the relative conductance of fusion pores (normalized to vesicle capacitance) was significantly www.nature.com/scientificreports www.nature.com/scientificreports/ larger (P < 0.01) in cAMP-treated cells (81.8 ± 3.7 pS/fF, n = 290) than in controls (72.6 ± 3.4 pS/fF, n = 285; Fig. 4d). These results demonstrate that cAMP widens the fusion pore, opposite to the action of ketamine 18 , which indicates that ketamine-induced vesicle fusion pore modulation is not mediated by ketamine-induced increases in [cAMP] i ( Fig. 1).

Discussion
Understanding the mechanisms of the antidepressant action of ketamine is crucial for improving therapeutic strategies for MDD, a worldwide burden 1 . The clinical effects of ketamine consist of a rapid (hours) and a sustained (1-2 weeks) phases; the latter substantially outlasts the metabolic half-life of ketamine (~6 h) 3 . Although the rapid molecular pharmacology of ketamine activity is generally ascribed to NMDAR antagonism 5 , alternative pathways (such as ketamine modulation of cAMP-dependent cascades in the human brain) have been considered 12 . These pathways likely modulate brain function by primarily affecting the activity of astrocytes 42 .
Astrocytes are the primary homeostatic cells in the central nervous system that control homeostasis of major neurotransmitters, including glutamate, GABA, adenosine, and noradrenaline 43 . Major psychiatric disorders are associated with substantial decreases in astroglial density and in loss of astroglial functions, which arguably results in an imbalance in neurotransmitter homeostasis and hence in aberrant information processing in neuronal networks [44][45][46][47][48][49][50] . At the molecular level, mood disorders are associated with astroglia-specific changes in serotonin receptors and intracellular signaling pathways 51 . Pharmacological approaches that could specifically target astrocytes in the context of neuropsychiatric disorders have not yet been developed. Nonetheless, we present results that indicate that ketamine exerts astroglial-specific effects, which can arguably be linked to its antidepressant potential.
Ketamine, a drug with multiple targets, affects several cellular signaling cascades, including exo-and endocytosis 18,52 . Therefore, we first tested the hypothesis that ketamine regulates exo-/endocytosis through intracellular cAMP. We based this supposition on the recent finding that ketamine amplifies adrenergic receptor-mediated cAMP signaling in C6 glioma cells 11 . This in turn instigates the translocation of G αs -proteins from lipid rafts, allowing them to interact with and activate adenylyl cyclase 11 . It is conceivable that G αs -protein translocation could lead to an increase in [cAMP] i even in the absence of G-protein receptor stimulation. If ketamine-induced changes in the plasmalemmal structure are mirrored as lipid raft restructuring 11 , these may activate adenylyl cyclase and increase [cAMP] i . Indeed, we found that ketamine increased astroglial [cAMP] i (Fig. 1) in the absence of G-protein-activating neurotransmitters.
The ketamine-induced increase in [cAMP] i prompted us to compare the effects of both agents on vesicle fusion. It is generally accepted that cAMP potentiates exocytotic secretion 22 by regulating fusion of vesicles with the plasmalemma through the cAMP sensor cAMP-GEFII (Epac2) [53][54][55] . Protein kinases can also regulate exocytosis by increasing the population of vesicles sensitive to Ca 2+ 56 with cAMP-dependent potentiation being associated with protein kinase A 22 . It was also proposed that fusion pore flickering depends on protein kinases 57 as well as cAMP 19 . As ketamine was reported to induce fusion pore flickering 18 , this led us to question whether these effects may be mediated via cAMP.  (Figs 2-4, Supplementary Fig. 1), ketamine inhibits secretion of astroglial BDNF 52 . We found no effect of cAMP on the frequency of full vesicle and 25 µM ketamine (0.52 ± 0.03 D4/ µm 2 ) compared with controls (0.37 ± 0.03 D4/µm 2 ) (***P < 0.001, Holm-Sidak one-way ANOVA). Ketamine did not affect the density of the cholesterol-rich domains in the PC12 cell line and fibroblasts. The density of the D4-positive domains was significantly higher in astrocytes (0.37 ± 0.03 D4/µm 2 ) compared with PC12 cells (0.14 ± 0.01 D4/µm 2 ) and fibroblasts (0.10 ± 0.01 D4/µm 2 ) (***P < 0.001, Kruskal-Wallis test). (d) The average area of cholesterol-rich domains, measured in controls (treated with vehicle) and in cells treated with ketamine (2.5 µM and 25 µM) did not significantly differ in the PC12 cell line, fibroblasts and astrocytes, but it differed between different cell types. The average area of D4-positive domains was significantly higher in the PC12 cell line (0.080 ± 0.009 µm 2 ) and astrocytes (0.056 ± 0.001 µm 2 ) than in fibroblasts (0.037 ± 0.001 µm 2 ) (***P < 0.001, Kruskal-Wallis test). The data in the graphs are reported as mean ± SEM. Numbers above the bars represent the number of cells analyzed.  (Figs 3 and  4), as was previously described in pituitary cells 19 . The ratio between fusion pore conductance and vesicle size (G p /C v ) represents a measure of exocytotic secretion 58 , and cAMP increases this ratio in astrocytes. An increase in astrocyte [cAMP] i promotes transient fusion of small synaptic-like vesicles (Fig. 2), as observed in pituitary cells 19 in which Ca 2+ -dependent regulated exocytosis mediates the release of prolactin 59 . This suggests that cAMP preferentially stimulates full fusion of large non-synaptic-like vesicles 60 , whereas small vesicles remain attached to the plasmalemma with a narrow fusion pore (transient events). Consistent with this, immunocytochemical studies revealed that vesicles labelled with an antibody against VAMP2, a target of cAMP signaling and a marker of secretory vesicles 21 , are less abundant and of smaller diameter after cAMP treatment, indicating that cAMP prompts full fusion of larger secretory vesicles with the plasmalemma (Supplementary Fig. 1).
Our results also demonstrate that cAMP treatment results in the formation of larger early endosomes ( Supplementary Fig. 2), suggesting homotypic vesicle-to-vesicle fusion between endosomal organelles, consistent with other studies that have suggested cAMP-evoked vesicle-to-vesicle fusion 26 . Homotypic fusion between vesicles also explains the amperometry results, which demonstrated that cAMP increased the quantal size of secretory vesicles [60][61][62] . We have observed a similar effect with capacitance measurements in this study, which revealed more full fusions of larger vesicles in cAMP-treated astrocytes. Our results have now demonstrated that cAMP stimulates full fusion of larger (non-synaptic-like) vesicles, widens the fusion pore and prolongs the open fusion pore dwell time, thus increasing secretory activity. Therefore, the action of cAMP contrasts with the previously reported ketamine-mediated fusion pore stabilization in a narrow state, which inhibits vesicle cargo discharge 18,52 .
This study has demonstrated that, although ketamine increases [cAMP] i , ketamine and cAMP have a distinct effect on vesicle fusion. Hence, we tested an alternative hypothesis whether ketamine directly alters the structure of the plasmalemma.
We visualized the cholesterol-rich plasmalemmal domains with a fluorescent cholesterol-specific peptide D4 from the toxin perfringolysin O 40,41 that labels the outer leaflet of the plasmalemma. D4-positive punctate structures thus represent cholesterol-rich domains in the astrocyte plasmalemma (Fig. 5). The domain density, relative to the imaged cell area, increased significantly after ketamine treatment, yet the size of individual D4-positive domains remained the same (Fig. 5). Therefore, a relatively short exposure of cells to ketamine (30 min) results in a visible change in membrane structure. These changes are specific to astrocytes, because they were not observed in neural-like PC12 cells or in non-excitable fibroblasts (Fig. 5). Although it is unclear how ketamine specifically affects the density of cholesterol-rich domains in the plasmalemma (Fig. 6), the overall increase in cholesterol production is unlikely, since no increase in serum level of cholesterol was observed even after application of a high ketamine dose (120 and 140 mg/kg) to male Wistar rats that were sacrificed 20 min after the administration of ketamine 63 . Moreover, intraperitoneal administration of ketamine to male Wistar rats (1 mg/kg) for 6 days did not affect the cholesterol synthesis 64 . Supporting the notion that vesicle-based mechanism may be affected by ketamine through affecting the distribution of cholesterol in the plasma membrane, are experiments where the fusion of single vesicles with the plasmalemma was studied in the presence and absence of ketamine; the results revealed a robust stabilization of the fusion-pore in a narrow state, suggesting that endocytotic vesicles may not transfer from the transient fission to full fission state 18 . This may contribute to an increased density of cholesterol at the plasmalemma, since cholesterol-rich membrane domains may not be internalized via endocytosis. The availability of cholesterol appears to impair upon synapse development 65 , and our results suggest an important role of astrocytes in cholesterol homeostasis in the central nervous system. Astrocytes provide cholesterol to neurons, where it is needed for shaping neuronal structures which is particularly critical for synaptogenesis. The ketamine-induced increase in the density of cholesterol-rich domains in the astroglial plasmalemma may thus enable more intense flux of cholesterol molecules from astrocytes to neurons. We may therefore speculate that ketamine boosts plasticity in neural networks, although such a proposal requires further experimentation.
In conclusion, we here revealed that ketamine induces visible structural changes in the outer leaflet of the astroglial plasmalemma, observed as redistribution of cholesterol-rich domains. This action appears astroglia specific and may affect diverse homeostatic responses that could modulate the functional activity and plasticity www.nature.com/scientificreports www.nature.com/scientificreports/ of neuronal networks. In particular, these changes may influence the flux of cholesterol from astrocytes to neurons that is critical for morphological plasticity of synaptic connections 65 . In addition, structural changes of the astroglial plasmalemma likely affect adenylyl cyclase, with consequent increases in [cAMP] i in the absence of G-protein-coupled receptor stimulation 11 . This new mechanism of ketamine action may explain its multiple effects on depressive behavior and highlights the role of astrocytes in the search for new antidepressants.
The acetoxymethyl ester of cAMP (cAMP-AM, A022, BIOLOG Life Science Institute) was applied to cells in serum-free medium. To optimize cAMP-AM loading, cells were pre-incubated for 15 min in DMEM without FBS; cAMP-AM (60 µM) or DMSO (0.2%; vehicle) were applied for 30 min.

FRET Measurements of Cytosolic cAMP and Data Analysis.
Astrocytes expressing the Epac1-camps 28 were examined 24-48 h after transfection with a Plan NeoFluoar 40×/1.3 NA oil differential interference contrast (DIC) immersion objective using a LSM510 META confocal microscope (Carl Zeiss, Jena, Germany). Real-time Epac1-camps FRET signal acquisition was performed as described 66 .
Unless stated otherwise, the FRET signal is reported as the ratio of the CFP/YFP fluorescence after subtracting the background fluorescence using Excel (Microsoft, Seattle, WA). The values of the FRET signals were normalized to 1.0. An increase in the FRET signal reflects an increase in the [cAMP] i .
Initially, astrocytes were kept in ECS and treated with 25 µM ketamine (Tocris Bioscience, Bristol, UK) for 900 s following a 100-s baseline. Control cells were treated with ECS (vehicle). The amplitude of the FRET signal (ΔFRET (%)) was determined for individual recordings by subtracting the mean FRET signal of the signal spanning the last 100 s before treatment from the first 100 s of the recording (baseline). The initial rate of the FRET signal change (ΔFRET/Δtime) was calculated for each recording as the slope of the linear regression function (ΔFRET (%) = slope (%/min) × Δtime (min)) fitting the initial FRET signal change.
Electrophysiology. Astrocyte-coated coverslips maintained in ECS were mounted on an inverted microscope (Zeiss Axio Observer.A1). Compensated cell-attached patch-clamp recordings were performed to measure discrete step-like changes in membrane capacitance (C m ) 68 and fusion pore conductance 31 . Full vesicle fusion/ fission was defined as a discrete upward/downward step in imaginary (I m ) part of the admittance signal 30 that was not followed by a step of similar amplitude (±15%) and opposite direction within 15 s, whereas transient vesicle fusion/fission was defined as a step in I m that was followed within 15 s, as reported previously 18 . As C m is proportional to the membrane area, the vesicle surface area and diameter (d) were determined assuming spherical vesicle geometry and a specific membrane capacitance of 10 fF/μm 2 31 . Immunocytochemistry. Immunocytochemical staining of astrocytes treated with 0.2% DMSO or 60 µM cAMP-AM (30 min) was performed as described 52  www.nature.com/scientificreports www.nature.com/scientificreports/ mCherry-D4-PFO and EGFP-D4-PFO expression and purification. For the construction of a plasmid encoding a His-Tag-eGFP-D4 or His-Tag-mCherry-D4 fusion proteins, a DNA fragment containing eGFP or mCherry coding region was first cloned into the pGA2.1 bacterial expression vector 69 after removing Equinatoxin II coding sequence using XhoI and MluI sites. Flexible Gly-Ser linker and additional AvrII restriction site were created at 3′ end of the gene encoding fluorescent protein as non-complementary ends of the amplification primers. In the next step, DNA fragment encoding for D4 domain of PFO was introduced into the prepared vector using AvrII and MluI restriction sites. The tagged D4-PFO variants were expressed in E. coli BL21(DE3) cells, which were grown in 1 l of LB medium supplemented with 100 µg/ml ampicillin (LBA) to an A 600 of 0.5-0.7, induced with 0.5 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) and incubated overnight at 20 °C, with shaking at 180 rmp. The bacteria were harvested by centrifugation for 15 min at 4 000 g and 4 °C, resuspended in 10 ml/g wet mass of lysis buffer (50 mM NaH 2 PO4, 300 mM NaCl, 10 mM imidazole, pH 8.0) and lysed by sonication. Cellular remnants were removed by centrifugation at 50 000 g for 1 hour at 4 °C. The supernatant was filtered through a 0.2-µm cellulose-acetate filter, added to the 0.5 ml 50% Ni-NTA slurry (Ni-NTA Superflow, Qiagen) and incubated at 4 °C on a rotary shaker for 60 minutes. Lysate-Ni-NTA mixture was loaded on polypropylene columns, washed twice with 10 ml wash buffer (50 mM NaH 2 PO4, 300 mM NaCl, 20 mM imidazole, pH 8.0) and the protein was eluted with 2 ml of elution buffer (50 mM NaH 2 PO4, 300 mM NaCl, 250 mM imidazole, pH 8.0). Purified protein was dialyzed (Slide-a-lyzer from Thermo Scientific, 10 000 MWCO) against 0.5 l of a buffer (50 mM Tris-HCl, 200 mM NaCl, 5% (v/v) glycerol, pH 7.4).
Cholesterol and Ganglioside Monosialic Acid (GM1) Staining. Astrocytes, PC12 cells and fibroblasts were exposed to 2.5 or 25 µM ketamine for 30 min at 37 °C in culture medium; controls were exposed only to ECS. Cells were washed with ECS at room temperature (RT) and mCherry-D4 (0.25 µM) was applied for 30 min at RT. Then cells were washed with ECS 3× for 3 min at RT, fixed in 4% formaldehyde for 10 min at RT, and washed with ECS 3× for 3 min at RT. Samples were mounted onto glass slides with Slow Fade Gold Antifade agent. In a subset of experiments, astrocytes were first labelled by mCherry-D4 (0.25 µM) and then by the non-toxic B-subunit toxin from Vibrio cholerae (CT-B) conjugated to Alexa Fluor 488 (Thermo Fisher Scientific) that interacts with the penta-saccharide chain of ganglioside monosialic acid (GM1), as described 70 . Structured Illumination Microscopy and Image Analysis. Astrocytes labelled with dextrans, antibodies or mCherry-D4 were imaged with a Zeiss ELYRA PS.1 super-resolution microscope with an oil-immersion plan apochromatic DIC objective (63×/NA 1.4), an EMCCD camera (andor iXon 885, Andor Technology, Belfast, UK), and 5 different grating directions for SIM.
Alexa Fluor 488 and Alexa Fluor 546 /mCherry-D4 were excited by 488 nm argon and 561 nm DPSS laser lines, respectively, and emission fluorescence was filtered with 495-575 nm or 570-650 nm bandpass filters. The number and surface area of fluorescent vesicles were obtained by exporting tiff images to ImageJ (NIH, Bethesda, MD). To identify individual vesicles, the intensity threshold was set to 20% of the maximum fluorescence, and the minimum fluorescent spot size considered to be a vesicle was five adjacent pixels (5 × 0.04 × 0.04 µm); the minimum surface area covered by a vesicle was 0.008 µm 2 .
In mCherry-D4-stained cells, we acquired 500-nm thick z-stacked images that were analyzed in Fiji 71 . Individual cells were cropped and auto-thresholded with IsoData method to measure number and the area of D4-labelled entities in the range of 7-3000 pixels (0.01-4.80 µm 2 ).
Confocal Microscopy and Image Analysis. Z-stacked confocal images of astrocytes expressing ANP.emd were obtained with a Zeiss LSM 780 as described 17 . The number and surface area of ANP.emd-positive vesicles were obtained by ImageJ. The intensity threshold was set to 20% of the maximum fluorescence and the minimum spot size considered to be an individual vesicle was four pixels (4 × 0.088 × 0.088 μm); the minimum surface area covered by a vesicle was 0.031 µm 2 . Double-fluorescent (mCherry-D4-and CT-B-labelled) cells were observed by a plan-apochromatic oil-immersion objective 63×/NA 1.4. Z-stacked images were obtained with a 488-nm argon laser and 561-nm diode-pumped solid-state laser excitation; the emission fluorescence was bandpass filtered at 500-550 nm and 590-640 nm, respectively. Fluorescence co-localization between green-emitting Alexa Fluor 488 and red-emitting mCherry-D4 was quantified in 8-bit TIFF files exported to ColocAna software 72 as described 70 . Statistics. Data analysis was performed with SigmaPlot (Systat Software, San Jose, CA). The parameters are presented as mean ± SE. Unless stated otherwise, Student's t test and the Mann-Whitney U test were used to determine statistical significance; P < 0.05 was considered significant.

Data Availability
All data generated or analyzed during this study are included in this published article and its Supplementary Information files or are available from the corresponding author on reasonable request.