Neuropharmacology 95 (2015) 130e143

Contents lists available at ScienceDirect

Neuropharmacology journal homepage: www.elsevier.com/locate/neuropharm

Time-dependent modulation of glutamate synapses onto 5-HT neurons by antidepressant treatment Sean D. Geddes a, b, c, Saleha Assadzada a, b, c, Alexandra Sokolovski a, e, ïque a, b, c, d, * Richard Bergeron a, e, Samir Haj-Dahmane f, Jean-Claude Be a

Department of Cellular and Molecular Medicine, Faculty of Medicine, University of Ottawa, K1H 8M5, Canada Neuroscience Graduate Program, Faculty of Medicine, University of Ottawa, Ottawa, Ontario K1H 8M5, Canada Canadian Partnership for Stroke Recovery, Faculty of Medicine, University of Ottawa, Ottawa, Ontario K1H 8M5, Canada d Centre for Neural Dynamics, Faculty of Medicine, University of Ottawa, Ottawa, Ontario K1H 8M5, Canada e Ottawa Hospital Research Institute, Ottawa, Ontario, K1Y 4E9, Canada f Research Institute on Addictions, University at Buffalo, State University of New York, Buffalo, NY 14203, USA b c

a r t i c l e i n f o

a b s t r a c t

Article history: Received 10 November 2014 Received in revised form 18 February 2015 Accepted 19 February 2015 Available online 5 March 2015

Antidepressants, including the selective serotonin reuptake inhibitors (SSRIs), are thought to exert their clinical effects by enhancing serotonin (5-HT) transmission. However, animal studies show that the full magnitude of this enhancement is reached only following prolonged treatments with SSRIs, consistent with the well-described therapeutic delay of this class of medications. Thus, the clinical efficacy of SSRIs most likely does not emerge from their acute pharmacological actions, but rather indirectly from cellular alterations that develop over the course of a sustained treatment. Here, we show that sustained administration of the SSRI citalopram leads to a homeostatic-like increase in the strength of excitatory glutamate synapses onto 5-HT neurons of the dorsal raphe nucleus that was apparent following one week of treatment. A shorter treatment with citalopram rather induced a paradoxical decrease in the strength of these synapses, which manifested itself by both pre- and postsynaptic mechanisms. As such, these results show that an SSRI treatment induced a concerted and time-dependent modulation of the synaptic drive of 5-HT neurons, which are known to be critically involved in mood regulation. This regulation, and its time course, provide a mechanistic framework that may be relevant not only for explaining the therapeutic delay of antidepressants, but also for the perplexing increases in suicide risks reportedly occurring early in the course of antidepressant treatments. © 2015 Elsevier Ltd. All rights reserved.

Keywords: Serotonin (5-HT) Antidepressants Glutamate Synapse Major depression Dorsal raphe nucleus (DRN) AMPAR NMDAR

1. Introduction Major depression is one of the most prevalent and lifethreatening forms of mental illness and is associated with significant disabilities and mortality (Chopra et al., 2011). Although modern antidepressants, including the widely prescribed Selective Serotonin Reuptake Inhibitors (SSRIs), provide much welcomed therapeutic benefits, they are unfortunately plagued by a lack of efficacy in a high proportion of patients and by a delay in their clinical actions (Slattery et al., 2004). These distressing limitations

* Corresponding author. University of Ottawa, Faculty of Medicine, Office: 3501N, 451 Smyth Road, Ottawa, ON K1H 8M5, Canada. Tel.: þ1 613 562 5800 4968; fax: þ1 613 562 5434. ïque). E-mail address: [email protected] (J.-C. Be http://dx.doi.org/10.1016/j.neuropharm.2015.02.027 0028-3908/© 2015 Elsevier Ltd. All rights reserved.

resolutely advocates for the need to gain a clearer understanding of the neural substrates of the clinical effects of this class of drugs. The mere presence of a therapeutic delay suggests that the clinical efficacy of antidepressants stems from the development of a neuroadaptive cellular mechanism rather than from their acute pharmacological actions. In part reinforced by the clinical findings that the serotonin (5-HT) system appears to be required for the antidepressant effect (Delgado et al., 1990; Neumeister, 2003; Young et al., 1985), the behavior of 5-HT neurotransmission during the time course of administration of several types of antidepressants has been under scrutiny for close to 40 years (e.g., (De Montigny and Aghajanian, 1978). Whereas the acute administration of SSRIs rapidly blocks 5-HT transporters and increases net 5HT output in target regions (Bel and Artigas, 1993; Artigas, 1993), the magnitude of this increase is constrained by activation of inhibitory somatodendritic 5-HT1A autoreceptors that suppresses

S.D. Geddes et al. / Neuropharmacology 95 (2015) 130e143

the firing activity of 5-HT neurons. Remarkably, the firing activity of these neurons recovers over the time course of sustained SSRI treatments, likely as a result of the desensitization of these 5-HT1A autoreceptors (De Montigny and Blier, 1984; Blier and De Montigny, 1994; Albert and Lemonde, 2004). Because the time course of these cellular events broadly matches the delayed therapeutic efficacy of SSRIs in humans, it may explain the delay in the clinical efficacy of these drugs. Unfortunately, despite some progress, capitalizing on this knowledge for therapeutic benefits has been difficult, likely because of the challenge in pharmacologically discriminating between pre- and postsynaptic 5-HT1A receptors (Blier et al., 1998; Segrave and Nathan, 2005). Moreover, the framework elaborated from these studies has mainly conceptualized 5-HT neurons as semi-autonomous units, largely ignoring the dynamic synaptic network in which these cells are embedded. Indeed, the raphe receives strong afferent projections from a remarkably vast amount of cortical and subcortical regions of the brain, forming an extensive network that synapses onto both 5-HT neurons and local GABAergic neurons (Weissbourd et al., 2014; Pollak Dorocic et al., 2014; Pan and Williams, 1989; Haj-Dahmane and Shen, 2005; Li and Bayliss, 1998). Yet, the basic characteristics of these glutamatergic synapses have been relatively little investigated, let alone how they are altered by chronic antidepressant treatments. Here, using ex vivo whole-cell electrophysiological recordings, we found, unexpectedly, that the strength of glutamate synapses onto 5-HT neurons was markedly reduced early in the time course of a treatment with an SSRI. Remarkably, a homeostatic-like synaptic plasticity mechanism emerged later during the treatment such that, by about one week of treatment, synaptic strength was not only no longer suppressed, but was rather potentiated. Because 5-HT neurons play a key role in mood regulation, this delayed, homeostatic-like, increase in synaptic drive onto these neurons by SSRIs may contribute to their clinical effectiveness. 2. Materials and methods 2.1. Animals Sprague Dawley rats (50e90 g; Charles River, St. Constant, Quebec, Canada) were received at least 7 days prior to implantation of mini-pumps for drug delivery and housed two to three per cage. They were kept on a 12:12h light/dark cycle, with access to food and water ad libitum. All experiments and procedures were performed in accordance with approved procedures and guidelines set forth by the University of Ottawa Animal Care and Veterinary Services.

2.2. In vivo treatments Citalopram (Toronto Research Chemicals; North York, Canada) was delivered to animals via subcutaneously implanted osmotic minipumps (Alzet; Cupertino, California) at a dose of 20 mg/kg/day. Animals (26e29 days old) were anesthetized with isofluorane at the time of minipump implantation. Citalopram was dissolved in 45% w/v 2-Hydroxypropyl-b-cyclodextrin (2-HbC) to enhance drug solubility in water. Citalopram or 45% w/v 2-HbC (as vehicle) was administered for 2 or 7 days before the electrophysiological experiments.

2.3. Slice preparation Brainstem slices containing the dorsal raphe nucleus were prepared from 28 to 37-day old Sprague Dawley rats following 2 or 7 days of treatments with citalopram or vehicle. Rats were anesthetized by inhalation of isofluorane (Baxter Corporation, Canada) and sacrificed by decapitation. The brain was rapidly removed and placed in ice-cold choline chloride-based cutting solution of the following composition (in mM): 119 choline-Cl, 2.5 KCl, 1 CaCl2, 4.3 MgSO4-7H2O, 1 NaH2PO4, 1.30 sodium Lascorbate, 26.20 NaHCO3, and 11 glucose, and equilibrated with 95% O2, 5% CO2. Two or three coronal slices (300 mm thick) containing the DRN were sectioned from a block of brainstem tissue in ice-cold choline chloride-based cutting solution using a Vibratome slicer Series 1000 Plus (Shepreth, England) or a Leica VT1000s vibrating blade microtome (Nussloch, Germany). Slices were then transferred into a recovery chamber containing standard Ringer's solution of the following composition (in mM): 119 NaCl, 2.5 CaCl2, 1.3 MgSO4-7H2O, 1 NaH2PO4, 26.2 NaHCO3, and 11 glucose, at a temperature of 37 C, continuously bubbled with a mixture of 95% O2, 5% CO2.

131

Slices were then allowed to recover for 1h in the recovery chamber and equilibrate to a temperature of approximately 25 C until recordings were performed. Hippocampal neuron recordings were performed from ex vivo organotypic slices cultures as previously described (Soares et al., 2013). In brief, individual hippocampi were removed from sprague dawley rats between the ages of 6e8 days. Coronal slices from hippocampi were gathered using a MX-TS Tissue Slicer (Siskiyou). As previously described, individual hippocampal slices were placed in membrane inserts and kept in six-well plates at 34 C in 95% O2 and 5% CO2 in culture media (Soares et al., 2013). Whole-cell recordings were performed following 8e10 DIV. 2.4. Whole-cell electrophysiology DRN neurons were visualized using an upright microscope (Examiner D1; Zeiss, Oberkochen, Germany) equipped with Dodt-contrast or differential-interference contrast (DIC) (40/0.75NA objective). 5-HT neurons were identified by morphological and biophysical characteristics using previously established criteria (HajDahmane, 2001; Aghajanian and Vandermaelen, 1982; Calizo et al., 2011). Using the aqueduct as a landmark, we surmise that our recordings were primarily from the rostral half of the DRN. Whole-cell recordings were carried out using an Axon Multiclamp 700B amplifier, sampled at 10 kHz, digitized with an Axon Digidata 1440A digitizer and filtered at 2 kHz. Whole-cell recordings were performed using borosilicate glass patch electrodes (3e6 MU; World Precision Instruments, Florida or Sutter, California) pulled on a Narishige PC-10 pipette puller (Narishige, Japan). All experiments were performed at room temperature in Ringer containing (in mM): 119 NaCl, 2.5 KCl, 1.3 MgSO4, 2.5 CaCl2, 1.0 NaH2PO4, 11 glucose, and 26.2 NaHCO3 (or low Mg2þ Ringer containing 0.1 MgSO4, 3.0 CaCl2) saturated with 95% O2 and 5% CO2 (pH ¼ 7.3; 295e310 mOsm/L). Excitatory post-synaptic currents (EPSCs) were recorded using an intracellular solution of either of the following compositions (in mM): (1) 115 cesium methane-sulfonate, 5 tetraethylammonium-Cl, 10 sodium phosphocreatine, 20 HEPES, 2.8 NaCl, 5 QX-314, 0.4 EGTA, 3 ATP(Mg2þ), and 0.5 GTP, pH 7.25 (adjusted with CsOH; osmolarity, 280e290 mOsmol/L; or (2) 77 cesium methane-sulfonate, 10 tetracesium BAPTA, 5 tetraethylammonium-Cl, 3 calciumchloride, 10 sodium phosphocreatine, 20 HEPES, 5 QX-314, 4 ATP(Mg2þ), and 0.5 GTP, pH 7.25 (adjusted with CsOH; osmolarity, 280e290 mOsmol/L. In some recordings, we bath applied the non-selective 5-HT1A receptor agonist 5Carboxamidotryptamine (5-CT) to elicit a 5-HT1A-medated outward current (or hyperpolarization). These recordings were carried out using an internal solution of the following composition (in mM): 115 potassium gluconate, 20 KCl, 10 sodium phosphocreatine, 10 HEPES, 4 ATP(Mg2þ), and 0.5 GTP, pH 7.25 (adjusted with KOH; osmolarity, 280e290 mOsmol/L). For post-hoc immunohistochemical identification of 5-HT neurons, Alexa-594 hydrazide (Na-salt; 30 mM; Invitrogen) was included in the recording pipette. For all voltage clamp recordings, access resistance was continuously monitored by applying a 125 ms, 2 mV hyperpolarizing pulse, 245 ms prior to stimulation. Recordings were discarded when the access resistance changed by >30%. Liquid junction potential was not compensated for. Excitatory post-synaptic currents were elicited using patch electrodes filled with standard Ringer solution and placed 50e100 mm dorsolateral to the recorded neurons. EPSCs were evoked with a single square pulse (100 ms) delivered at either 0.067 or 0.1 Hz. Unless stated otherwise, all recordings were performed in the presence of the GABAA receptor antagonist bicuculline (20 mM) or picrotoxin (100 mM). For determining NMDAR/AMPAR ratios (N/A ratios) of evoked EPSCs (eEPSCs), mixed NMDAR and AMPAR eEPSCs were recorded while voltage clamping the cell at þ40 mV. NBQX (20 mM) was then bath applied in order to isolate the NMDA receptor-mediated component. The AMPAR component was computed off-line by subtracting the NMDAR component from the mixed current traces. Paired-pulse ratios (PPRs) were determined by eliciting eEPSCs with an interstimulus interval (ISI) of 50 ms while holding the cell at 70 mV. These paired pulses were elicited at a stimulation frequency of 0.1 Hz. Miniature EPSCs (mEPSCs) were acquired at 70 mV in the presence of Tetrodotoxin (TTX, 1 mM). Currentevoltage relationships (IeV curves) of evoked AMPAR-mediated EPSCs were performed as previously described (Soares et al., 2013) with 0.1 mM spermine added to the internal solution and with D,L-APV (100 mM) in the extracellular solution. The sensitivity of AMPAR-mediated currents to NASPM (3 mM) was established by obtaining a 4-min baseline of AMPAR-EPSC and bath applying NASPM for 25 min. Similar to NASPM applications, the sensitivity of NMDAR-mediated currents to ifenprodil was established by obtaining a 4-min baseline NMDAR-mediated EPSC at þ40 mV in the presence of NBQX (20 mM) and bath applying ifenprodil (3 mM) for 25 min. NMDAR sensitivity to ifenprodil was calculated based on the 4-min baseline and the last 4 min from the 29 min recording. The decay of NMDAR-mediated eEPSCs was determined by a weighted tau value calculated using a bi-exponential fit as previously described (Beique et al., 2006). 2.5. Immunohistochemistry Brain slices used for immunohistochemistry were prepared from 28 to 37 day old rats with 1X phosphate buffered saline (PBS) used as the solvent for all solutions. Rats were anaesthetized through an intraperitoneal injection of 100 mg/kg pentobarbital. They were then intracardially perfused with 10 U/mL ice-cold heparin, followed by 4% PFA. Post fixation in 4% PFA was carried out for 2 h prior to sequential

132

S.D. Geddes et al. / Neuropharmacology 95 (2015) 130e143

cryoprotection steps in 15% and 30% sucrose over the course of two days. Brains were frozen in 30 C isopentane, mounted in cryomatrix, and cut to 40 mm slices on a cryostat (Leica CM 3050S, Houston, TX, USA). Free-floating sections were blocked in 1% BSA þ 0.03% Tx-100 for 45 min, and incubated in primary antibody against tryptophan hydroxylase enzyme (TPH, Sigma, 1:1000) in blocking solution at 4 C. The following morning slices were washed three times, for 5 min each, with 0.1% Tx100 prior to an incubation in the secondary antibody Alexa-488-conjugated goat anti-mouse IgG (Jackson ImmunoResearch) for 2 h at room temperature. Slices were washed with 0.1% Tx-100 three times, for 5 min each before mounting onto slides with Vectashield (Vector). To confirm that cells recorded on were in fact serotonergic, we performed posthoc immunolabeling of cells filled with Alexa 594 hydrazide (Na-salt; 30 mM; Invitrogen). Slices were fixed in 4% PFA for 24 h, and blocked in 1% BSA þ 0.1% Tx-100 for 45 min prior to an incubation in primary antibody against TPH for 36 h in blocking solution at 4 C. Slices were washed, incubated in secondary antibody, and mounted as above. Images were acquired using an Olympus BX61WI upright confocal microscope with 10-water and 63-oil immersion objectives, and laser emissions of 488 and 594 nm exciting Alexa-488 and Alexa-594, respectively. All post-processing was performed using ImageJ software. 2.6. Data analysis All electrophysiological recordings (with the exception of NSFA; see below) were analyzed offline using Clampfit 10.3 (Molecular Devices) and Origin 8.5 software (OriginLab). mEPSCs were analyzed using a mEPSC current-template search through Clampfit 10.3 software (Molecular Devices). The current template was created based on mEPSCs recorded in basal conditions. The same template was used for all mEPSC recordings. Non-stationary fluctuation analysis was performed using Minianalysis software (Synaptosoft) as previously described (Hartveit and Veruki, 2007; He et al., 2012; Soares et al., 2013). In brief, 50-100 mEPSCs were manually selected and peak scaled to match their average. Once aligned, the variance around the decay phase of average mEPSCs was plotted against the amplitude. The relationship was plotted using the parabolic equation: s2 ¼ iI  I2/N þ b, where s2 ¼ variance, i ¼ single channel current, N ¼ number of open channels at peak current, I ¼ mean current, b ¼ background variance. AMPAR channel conductance (y) was calculated by dividing single channel current by the driving force (70 mV). Cells were only included if they had a parabolic fit with r2 > 0.5. To test for multiplicative scaling of synaptic strength, we iteratively tested 1000 scaling factors ranging from 0.5 to 1.5 on the amplitude distributions of all mEPSC events using a recently reported test validated to appropriately deal with the potential distortion induced by uneven representation of subthreshold events in the amplitude distribution in the different treatment groups (Kim et al., 2012). All data are presented as means ± SEM. N refers to the number of cells recorded. Unless otherwise stated, differences between treatment and vehicle groups were analyzed using a standard Student's t-test (p < 0.05). P values of less than 0.05 were considered statistically significant and are indicated with an asterisk (*). 2.7. Drugs and chemicals Bicuculline, Picrotoxin, 2,3-Dioxo-6-nitro-1,2,3,4-tetrahydrobenzo[f]quinoxaline-7-sulfonamide di-sodium salt (NBQX) and DL-2-Amino-5-phosphonopentanoic acid (D,L-APV) were purchased from Abcam (Cambridge, MA). Tetrodotoxin (TTX), N-[3-[[4-[(3-Aminopropyl)amino]butyl]amino]propyl]-1-naphthaleneacetamide trihydrochloride (NASPM), ifenprodil, spermine, 2-Hydroxypropyl-b-cyclodextrin (2-HbC), and 5-Carboxyaminotryptamine (5-CT) were purchased from Tocris Bioscience (Ellisville, MO). Citalopram was purchased from Toronto Research Chemicals (Toronto, Canada).

3. Results To investigate the function of glutamate synapses onto 5-HT neurons during the course of a treatment with an SSRI, we treated rats with the SSRI citalopram (20 mg/kg/day) or vehicle for 2 or 7 days using osmotic minipumps and carried out ex vivo wholecell electrophysiological recordings from putative 5-HT neurons in midbrain slices. We sought to compare and contrast the effects on synapse function of a short-term (2-day) with those induced by a longer-term (7-day) treatment with citalopram in order to identify adaptive cellular mechanisms that might develop during a sustained antidepressant treatment regimen. SSRIs, including citalopram, typically requires 1e4 weeks of treatments to show antidepressant-like effectiveness depending on the depression model (eg., (Takamori et al., 2001; Telner and Singhal, 1981; Sherman et al., 1982; Telner et al., 1981; Papp et al., 2002; Robison et al., 2014; Venzala et al., 2012). Recordings were

performed on medium to large fusiform shaped 5-HT neurons located in the rostral half of the dorsal medial portion of the dorsal raphe nucleus (dmDRN). Putative 5-HT neurons were identified by a combination of morphological features (DIC imaging), electrophysiological properties (action potential discharge behavior) and finally by post-hoc immunohistochemical identification (see Materials and Methods and Fig. 1). In pilot experiments (n ¼ 25; using a Kþ-based solution, see Materials and Methods), we found that >85% of the cells recorded were hyperpolarized by bath administration of the 5-HT1A receptor agonist 5-CT, indicating that most neurons we sampled in the dmDRN are 5-HT containing neurons (Calizo et al., 2011). 3.1. A citalopram treatment transiently reduced the strength of glutamate synapses onto dmDRN 5-HT neurons As a first test to determine whether the SSRI citalopram modulates glutamate synaptic drive onto dmDRN 5-HT neurons, we determined the ratio of NMDA-to AMPA receptor-mediated components of evoked excitatory postsynaptic currents (eEPSCs) onto 5-HT neurons from rats treated with citalopram for either 2 or 7 days. We recorded eEPSCs in the presence of bicuculline (20 mM), while holding the cell at þ40 mV. In these conditions, mixed AMPAR and NMDAR-mediated outward currents are observed and the respective contributions from the two receptor subtypes can be determined pharmacologically by bath administration of the AMPAR antagonist NBQX (20 mM) (Fig. 1C; see Materials and Methods) (Beique and Andrade, 2003; Beique et al., 2006). Following 2 days of citalopram treatments, the NMDAR/AMPAR ratio was robustly reduced compared to age-matched vehicletreated littermates (Fig. 1D, VEH, 0.76 ± 0.2, n ¼ 7; CIT, 0.36 ± 0.1, n ¼ 9; p < 0.05, Student's t-test). However, this reduction in NMDAR/AMPAR ratio of eEPSCs did not persist over the 7-day regimen of citalopram treatment as no significant difference was observed between the 7-day treated animals and their agematched vehicle littermates (Fig. 1D, VEH, 0.84 ± 0.2, n ¼ 10; CIT, 0.96 ± 0.2, n ¼ 15; p ¼ 0.7). Thus, the NMDAR/AMPAR ratio of excitatory synapses onto 5-HT neurons is robustly, but only transiently, reduced by a treatment with citalopram. The reduction in the NMDAR/AMPAR ratio of eEPSCs following a 2-day citalopram treatment can reflect a reduction in the number and/or function of NMDARs, an increase in the number and/or function of AMPARs, or a combined, but disproportional, alteration in the contribution of these receptor subtypes to the eEPSCs. To distinguish between these possibilities, we recorded AMPARmediated miniature EPSCs (mEPSCs) from 5-HT neurons in citalopram- and vehicle-treated littermates at both the 2- and 7-day time points (Figs. 2 and 3) as a means to provide a quantifiable metric of solely AMPAR function at glutamate synapses. In slices prepared from 2-day citalopram-treated animals, we observed a robust reduction in the amplitude of mEPSCs compared to agematched vehicle-treated littermates (Fig. 2C, VEH, 14.3 ± 1.0 pA, n ¼ 12; CIT, 10.1 ± 1.1 pA, n ¼ 9; p < 0.05). Remarkably, not only was this reduction in the amplitude of mEPSCs no longer observed in the 7-day citalopram treatment regimen, these synaptic events were of larger amplitudes than those recorded from the vehicletreated littermates at this time point (Fig. 3C, VEH, 12.9 ± 0.4 pA, n ¼ 20; CIT, 15.8 ± 0.9 pA, n ¼ 21; p < 0.05). Thus, citalopram induced a reduction in the strength of AMPAR-mediated synaptic transmission early during the time course of the treatment, which was followed by an increase in synaptic strength that was apparent later in the treatment. Such results suggest that the function of glutamate receptors of both the AMPA and NMDA subtypes are decreased during the early phases of a citalopram treatment, albeit disproportionally. This reduction in overall glutamate receptor

S.D. Geddes et al. / Neuropharmacology 95 (2015) 130e143

133

Fig. 1. Two days of citalopram treatment reduced the NMDAR to AMPAR ratio of excitatory synapses onto 5-HT neurons. A1, Schematic of brain slice prepared from the brainstem containing the dorsal raphe nucleus (DRN). Neurons were recorded from the dorsal medial DRN (dmDRN) where putative serotonin neurons are located and distinctly identifiable by a 5-HT1AR-mediated current. A2, 5-HT1AR-mediated outward current in response to bath application of the 5-HT1AR agonist 5-carboxamidotryptamine (5-CT, 100 nM; Vm, 55 mV). A3, Application of 5-CT induced a 5-HT1AR-mediated hyperpolarization that blocked action potential firing. B, Confocal image of TPH2-expressing DRN serotonin neurons (top left). Post-hoc confocal images of an identified 5-HT neuron filled with Alexa-594 (30 mM; top right) and immuno-fluorescence of TPH2 (bottom left; see Materials and Methods). Overlay of Alexa-594 filled neuron and TPH2-flourescence (bottom right). C, Schematic of experimental outline. Ex-vivo recordings were performed following either 2 or 7 days of sustained citalopram (or vehicle) administration via subcutaneously implanted osmotic minipumps. D1, Representative current traces from recordings showing the NMDAR/AMPAR ratios of eEPSCs on 5-HT neurons following 2 days of treatment with vehicle (VEH) or citalopram (CIT). Neurons were voltage clamped at þ40 mV. Mixed AMPARand NMDAR-mediated currents were first recorded (Black traces; Vm, þ40 mV). NBQX (20 mM) was bath-applied to isolate the NMDAR-mediated component of the mixed current (blue traces). The isolated NMDAR-mediated component was then subtracted from the mixed current trace to acquire an isolated (subtracted) AMPAR-mediated component (red trace). D2, The NMDAR/AMPAR ratio was significantly reduced in 2 day CIT-treated animals compared to age-matched vehicle-treated littermates. E1, Representative traces from recordings measuring the NMDAR/AMPAR ratios on 5-HT neurons following 7 days of treatment with VEH and CIT (obtained by the method described in D1). E2, The NMDAR/ AMPAR ratio was not significantly altered in 7 day citalopram-treated animals compared to age-matched vehicle-treated littermates. Calibration: A3, 10 mV, 1 min; D1, E1, 10 pA, 25 ms. Scale: B, top left, 300 mm; top right, 50 mm. (*) Asterisk indicates statistical significance of p < 0.05. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

134

S.D. Geddes et al. / Neuropharmacology 95 (2015) 130e143

Fig. 2. A treatment with citalopram for 2 days reduced the number of AMPARs at glutamate synapses of dmDRN 5-HT neurons. A, Schematic of 2 day experimental regimen. B1eB2, Representative mEPSCs (Vm, 70 mV) from VEH- and CIT-treated animals respectively. These recording were conducted in the presence of TTX (1 mM) and Bicuculline (20 mM). C, Cumulative probability plot of all mEPSC amplitudes in VEH- and CIT-treated animals. Inset, cell-averaged mEPSC amplitude histogram. mEPSC amplitude was significantly reduced following 2 days of citalopram treatment. D, Cumulative probability plot of all mEPSC inter-event intervals in VEH- and CIT-treated animals. Inset, cell-averaged mEPSC frequency histogram. mEPSC frequency was significantly reduced following 2 days of citalopram treatment. E, Scatter plot of correlation between mEPSC frequency and amplitude following 2 day VEH and CIT treatments (VEH, r2 ¼ 0.901; CIT, r2 ¼ 0.539). Inset, correlation between frequency and amplitude of mEPSCs from hippocampal neurons (r2 ¼ 0.004). F, Left, Cellaveraged AMPAR-mediated mEPSCs from single voltage clamp recordings (Vm, 70 mV) in VEH- (black) and CIT- (green) treated animals. Peak-scaled individual mEPSC events are shown (gray) behind the cell-averaged mean mEPSC. Right, Current-variance plots from peak-scaled non-stationary fluctuation analysis (NSFA) of mEPSC recordings (see Materials and Methods). G, Number of channels (N) open at peak of cell-averaged AMPAR-mEPSCs was significantly reduced in CIT-treated animals compared to age-matched VEH-treated littermates (VEH, 8.39 ± 1.29, n ¼ 8; CIT, 4.12 ± 0.370 n ¼ 7; P < 0.05). H, AMPAR conductance (g; normalized) calculated by NSFA was not significantly different in animals following 2 days of CIT treatment. Calibration: B1, B2, 10 pA, 100 ms; F, 10 pA, 25 ms. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

function, however, was transient, since the function of both receptor subtypes appeared to be increased following a prolonged treatment with citalopram (i.e., unaltered NMDAR/AMPAR ratio along with an increase in the amplitude of AMPAR-mediated mEPSCs at 7 days of treatment). Interestingly, previous studies have repeatedly shown that a short-term treatment (2 days) with several SSRIs, including citalopram, leads to a pronounced reduction in the mean firing activity of 5-HT neurons (Blier et al., 1987, Blier et al., 1990; El Mansari et al., 2005). As such, the delayed upregulation of excitatory synapse function onto 5-HT neurons observed following 7 days of treatment of citalopram appears to develop as a means to increase excitability of 5-HT neurons

following a period of decreased neuronal activity. This behavior is highly reminiscent of the phenomenon of homeostatic synaptic plasticity, whereby neurons, in response to prolonged periods of inactivity, increases several excitability metrics (Turrigiano, 2011, Turrigiano, 2008; Turrigiano and Nelson, 2004; Lee et al., 2013), including the upregulation of both the AMPAR and NMDAR subtypes (Soares et al., 2013; Watt et al., 2000). In principle, the behavior of AMPAR-mediated mEPSC amplitude observed during the course of a citalopram treatment could reflect changes in the number and/or function of AMPARs (i.e., conductance). To distinguish between these possibilities, we carried out non-stationary fluctuation analysis (NSFA) of mEPSC events

S.D. Geddes et al. / Neuropharmacology 95 (2015) 130e143

135

Fig. 3. A treatment with citalopram for 7 days increased the function of AMPARs in the dmDRN. A, Schematic of 7 day experimental regimen. B1eB2, Representative mEPSCs (Vm, 70 mV) from VEH- and CIT-treated animals respectively (as described in Fig. 2). C, Cumulative probability plot of all mEPSC amplitudes in VEH- and CIT-treated animals. Inset, Cellaveraged mEPSC amplitude histogram. mEPSC amplitude was significantly increased following 7 days of CIT treatment compared to VEH treatment. D, Cumulative probability plot of all mEPSC inter-event intervals in VEH- and CIT-treated animals. Inset, cell-averaged mEPSC frequency histogram. mEPSC frequency was not significantly changed following 7 days of CIT treatment compared to VEH treated littermates. E, Scatter plot of correlation between mEPSC frequency and amplitude following 7 day VEH and CIT treatments (VEH, r2 ¼ 0.846; CIT, r2 ¼ 0.854). F, Left, Average AMPAR-mediated mEPSCs from single voltage clamp recordings (Vm, 70 mV) in VEH- (black) and CIT- (green) treated animals. Peakscaled individual mEPSC events are shown (gray) behind the cell-averaged mean mEPSC. Right, Current-variance plots from peak-scaled non-stationary fluctuation analysis (NSFA) of mEPSC recordings (see Materials and Methods). G, Number of channels (N) open at peak of cell-averaged AMPAR-mEPSCs was not significantly different in CIT-treated animals compared to age-matched VEH-treated littermates. H, AMPAR conductance (g; normalized) calculated by NSFA was significantly increased in animals following 7 days of CIT treatment. Calibration: B1, B2, 10 pA, 100 ms; F, 10 pA, 25 ms. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

(Figs. 2F and 3F; see Materials and Methods) recorded in the different treatment groups. This analysis revealed that a 2-day treatment with citalopram induced a reduction in the number (N) of postsynaptic AMPARs located on 5-HT neurons (Fig. 2G, VEH, 8.4 ± 1.3, n ¼ 8; CIT, 4.1 ± 0.4, n ¼ 7; p < 0.05) without leading to any measurable changes in AMPAR conductance (Fig. 2H, g, normalized, VEH, 1.0 ± 0.1; CIT, 1.2 ± 0.1; p ¼ 0.3). Interestingly, following 7 days of treatments, NSFA revealed no significant difference in the number of AMPARs (Fig. 2G, VEH, 5.9 ± 1.1, n ¼ 11; CIT, 6.5 ± 1.1, n ¼ 16; p ¼ 0.7) but rather an increase in AMPAR conductance (g; normalized) compared to vehicle-treated pairs (Fig. 3H, VEH, normalized, 0.9 ± 0.2; CIT, 1.4 ± 0.1; p < 0.05). Thus, in comparison

to the situation that prevails at 2 days of citalopram treatment, these results suggest that the homeostatic-like upregulation of AMPAR-function that develops over the first week of treatment is itself mediated by a compensatory increase in both the number and single-channel conductance of AMPARs (Figs. 2F and 3F). 3.2. The citalopram-induced alterations in glutamate synaptic strength onto dmDRN 5-HT neurons showed synapse specificity In order to gain further mechanistic insights into the synaptic changes occurring during the course of a citalopram treatment, we asked whether the changes in unitary synaptic strength outlined

136

S.D. Geddes et al. / Neuropharmacology 95 (2015) 130e143

above occurred over the entire population of synapses onto 5-HT neurons or, rather, whether they showed some level of synapse specificity (i.e., the changes would be restricted to a subset of synapses). To begin addressing this issue, we analyzed the amplitude distributions of mEPSCs in the different treatment groups and determined whether the respective differences could be accounted for by a multiplicative mathematical transformation (i.e., whether the treatments uniformly scaled the amplitude distributions). The traditional interpretation of this outcome would be that, not only the treatment affected all synapses, but that every synapse was affected by the treatment in a way that is proportional to its initial strength (Turrigiano and Nelson, 2004). Alternatively, a nonmultiplicative alteration in the distribution of mEPSCs induced by citalopram would rather indicate that a subpopulation of synapses are preferentially altered by the drug treatment, in effect outlining a synapse-specificity in its action (Goel and Lee, 2007; Kim et al., 2012; Lee et al., 2013). To test for multiplicativity, we iteratively tested 1000 scaling factors ranging from 0.5 to 1.5 on the amplitude distributions using a recently reported test validated to appropriately deal with the potential distortion induced by uneven representation of subthreshold events in the amplitude distribution in the different treatment groups (Kim et al., 2012). Using this approach, we found that no unique scaling factor could account for the shift in amplitude distribution induced by a 2-day treatment with citalopram i.e., it was impossible to scale the two distributions such that they would become overlapping (Fig. 4A; p ¼ 0.00087; KS test). We carried out a similar test on the upregulation of mEPSC amplitude obtained following a 7-day treatment with citalopram and found that it was non-multiplicative as well (Fig. 4B; p ¼ 2.64  107; KS Test). We last compared the distribution obtained in the 2-day citalopram treatment to that obtained following the 7-day citalopram treatment, in effect placing the time of treatment as the independent variable. In these conditions, again no unique scaling factor could coalesce satisfactorily the two distributions (Fig. 4C; p ¼ 8.71  1012; KS Test). Thus, these results suggest that the time-dependent regulation of AMPAR function induced by a citalopram treatment shows some level of synapsespecificity i.e., that a subpopulation of synapses onto 5-HT neurons are preferentially regulated by this treatment. Future work will be required to identify the nature of this synapse specificity and identify with networks are preferentially regulated by prolonged antidepressant treatment. 3.3. A treatment with citalopram decreased the release probability at glutamate synapses onto dmDRN 5-HT neurons Thus far, our results show that a prolonged treatment with citalopram induced a concerted alteration over time in the number and/or function of both the AMPA and NMDARs located on 5-HT neurons. In order to identify additional synaptic alterations induced by this treatment, we next analyzed the effects of citalopram treatments on the frequency of mEPSCs. In the course of carrying out these experiments, we first noticed in control conditions the presence of a strong correlation (r2 ¼ 0.7; all vehicle data pooled) between the amplitude and the frequency of mEPSCs on a per individual cell basis (Figs. 2E and 3E). This is peculiar inasmuch as other cell types typically do not show such a strong correlation. As a means of comparison, we plotted the amplitude and frequency of mEPSCs from random hippocampal CA1 neuronal recordings and found no such correlation between these metrics (Fig. 2E (inset), r2 ¼ 0.004). Although the mechanistic underpinnings of this tight correlation observed in 5-HT cells are unclear, these results nonetheless point to an intriguing non-canonical regulation of synaptic gain by 5-HT neurons. We next found that a 2-day treatment with citalopram induced a profound reduction in the frequency of

mEPSCs (Fig. 2D, VEH, 2.3 ± 0.6 Hz, n ¼ 12; CIT, 0.6 ± 0.9 Hz, n ¼ 9; p < 0.05). This robust reduction subsided during the time course of the treatment such that there were no longer significant differences in the frequency of mEPSCs by 7 days of sustained citalopram treatments compared to controls (Fig. 3D, VEH, 2.0 ± 0.4 Hz, n ¼ 20; CIT, 1.9 ± 0.3 Hz, n ¼ 21; p ¼ 0.8). In principle, these robust, timedependent alterations in the frequency of mEPSC induced by a citalopram treatment could be attributed to a decrease in presynaptic release probability or in the number of AMPAR-containing synapses. Likewise, the compensatory recovery in the frequency of mEPSCs apparent at 7 days of treatment can also be accounted for by changes in release probability and/or the number of synapses. To begin discriminating between these possibilities, we set out to determine the effects of citalopram on the release probability of glutamate synapses impinging onto 5-HT neurons. The paired-pulse ratio (PPR) is a reliable index of presynaptic neurotransmitter release probability (Dobrunz and Stevens, 1997). A 2-day treatment with citalopram induced an increase in the PPR compared to age-matched vehicle-treated littermates, indicative of a reduction in release probability (Fig. 5A, VEH, 0.8 ± 0.1, n ¼ 8; CIT, 1.3 ± 0.1, n ¼ 8; p < 0.05). As such, this increase in PPR suggests that the reduction in mEPSC frequency observed early in the citalopram treatment (Fig. 2D, VEH, 2.3 ± 0.6 Hz; CIT, 0.6 ± 0.9 Hz; p < 0.05) can be accounted for by a reduction in release probability of glutamate synapses onto 5-HT neurons. Thus, a short-term treatment with citalopram reduced the strength of excitatory drive onto 5-HT neurons by both pre- and postsynaptic mechanisms. Because we have observed a recovery in the frequency of mEPSCs during the course of a prolonged citalopram treatment (compare 2 and 7 days of treatment; Figs. 2D and 3D), we reasoned that this recovery would be accompanied by a parallel recovery in the release probability of glutamate synapses. However, we found that the PPR of eEPSCs at 7 days of citalopram treatment was still significantly larger compared to controls, indicating that release probability of glutamatergic synapses onto 5-HT neurons was still significantly reduced following 7 days of citalopram administration (Fig. 5B, VEH, 0.9 ± 0.1, n ¼ 16; CIT, 1.4 ± 0.1, n ¼ 6; p < 0.05). Thus, these results suggest that the compensatory increase in the frequency of mEPSCs following 7-days of citalopram treatment does not appear to be accounted for by a change in probability of release. Rather, these results suggest that the increase in mEPSC frequency reflects a compensatory upregulation in the number of AMPARcontaining synapses. 3.4. A treatment with citalopram did not induce detectable alterations in the subunit composition of glutamate receptors The results outlined above show that a prolonged treatment with citalopram induces a dynamic regulation of synaptic strength onto 5-HT neurons that occurs through both pre- and postsynaptic mechanisms. In principle, the differential regulation of postsynaptic glutamate receptors' function during this treatment could be accompanied by changes in the subunit makeup of these receptors. Because subunit compositions influence several key features of AMPA and NMDA receptor function, we next addressed this possibility. AMPARs are tetrameric ligand-gated ionotropic glutamate receptors composed of different combinations of four subunits (GluA1-4; (Shepherd and Huganir, 2007; Beique and Huganir, 2009). The presence or absence of the GluA2 subunit in the poreforming complex is the most consequential for receptor function as it modulates a number of biophysical features of the channel and dictates its calcium permeability (Isaac et al., 2007). GluA2-lacking AMPARs exhibit a signature inward rectification consequent of a polyamine block and are readily blocked by a subtype-specific

S.D. Geddes et al. / Neuropharmacology 95 (2015) 130e143

137

Fig. 4. The citalopram-induced alterations in glutamate synaptic strength onto dmDRN 5-HT neurons showed synapse specificity. A1, B1, and C1, Cumulative distribution plots of pooled mEPSC amplitudes in both vehicle (gray) and citalopram (green) treated conditions at 2- and 7-day time points. Implemented scaling factor on cumulative probability plot (blue) derived by testing 1000 scaling factors (insets). A1, Pooled mEPSCs cumulative probability plot of 2-day vehicle (gray), citalopram (green), and scaled citalopram (blue; scaling factor ¼ 0.95; see inset). A2, Distribution histogram plotting the percentage of mEPSCs against mEPSC amplitude at the 2-day vehicle time point (gray) and the scaled 2-day citalopram distribution (blue). The difference of the two distributions is superimposed on the x-axis of the plot (delta; red line). B1, Pooled mEPSC cumulative probability plot of 7-day vehicle (gray), citalopram (green), and scaled citalopram (blue; scaling factor ¼ 1.078; see inset). B2, Distribution histogram plotting the percentage of mEPSCs against mEPSC amplitude at the 7-day vehicle time point (gray) and the scaled 7-day citalopram distribution (blue). Subtraction of the two distributions is superimposed on the x-axis of the plot (delta; red line). C1, Pooled mEPSC cumulative probability plot of 2-day citalopram (gray), 7-day citalopram (green), and scaled 7-day citalopram (blue; scaling factor ¼ 1.13; see inset). C2, Distribution histogram plotting the percentage of mEPSCs against mEPSC amplitude at the 2-day citalopram time point (gray) and the scaled 7-day citalopram distribution (blue). The difference of the two distributions is superimposed on the plot (delta; red line). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

138

S.D. Geddes et al. / Neuropharmacology 95 (2015) 130e143

Fig. 5. A treatment with citalopram decreased the release probability at glutamatergic synapses onto dmDRN 5-HT neurons. A1, Representative traces of evoked AMPAR-mediated EPSCs (ISI: 50 ms) following 2 days of VEH (black) or CIT (green) treatment. A2, Summary histogram showing an increase in the PPR following 2 days of CIT treatment. B1, Representative traces of evoked AMPAR-mediated EPSCs (ISI: 50 ms) following 7 days VEH (black) and CIT (green) treatment. B2, Summary histogram showing an increase in the PPR following 7 days of CIT treatment. Calibration: A1, left, B1, right, 10 pA, 25 ms; A1, right, 20 pA, 25 ms; B1, left, 25 pA, 25 ms. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

antagonist. We thus determined the contributions of GluA2-lacking AMPARs to synaptic transmission onto 5-HT neurons by determining the current-voltage (IV) relationship of AMPAR-mediated eEPSCs as well as their sensitivity to the specific GluA2-lacking AMPAR antagonist NASPM. In recordings from control slices, eliciting eEPSCs while holding the 5-HT cells at different voltages yielded IV curves that were predominantly linear, with only slight deviation from an ohmic relationship (Fig. 6A, C, n ¼ 19, vehicle data from both time-point groups), suggesting at very small contribution of GluA2-lacking AMPARs to the AMPAR-EPSCs recorded at these synapses. In line with this, we found that bath administration of the selective GluA2lacking AMPAR antagonist NASPM (30 mM) induced a small, but highly consistent, reduction in the amplitude of AMPAR-mediated eEPSCs (Fig. 6B, D, 32.3 ± 6.0%, n ¼ 15, vehicle data from both time-point groups). Altogether, these results indicate that a small proportion of AMPARs located at glutamate synapses of 5-HT neurons are of the calcium-permeable GluA2-lacking subtype. We carried out similar experiments in parallel on slices from rats treated for either 2 or 7 days with citalopram. At both treatment time points, the IV curves (Fig. 6A, C; A, VEH, n ¼ 14, CIT, n ¼ 11; C,

VEH, n ¼ 5, CIT, n ¼ 8) and the sensitivity to NASPM of AMPARmediated eEPSCs were indistinguishable from those observed in interleaved control experiments (i.e., in slices from animals treated with vehicle; Fig. 6B, D; B, VEH, 31.1 ± 6.3%, n ¼ 10, CIT, 33.7 ± 13.8%, n ¼ 9, p ¼ 0.9; D, VEH, 34.2 ± 12.4%, n ¼ 6, CIT, 30.0 ± 11.9%, n ¼ 8, p ¼ 0.8). Thus, the citalopram-induced modulation in the number and/or function of postsynaptic AMPAR function outlined above occurs independently of detectable alterations in AMPAR subunit composition, at least as it pertains to the incorporation of the GluA2 subunit. As outlined above (see Fig. 1), the regulation of the AMPA/ NMDAR ratio of eEPSCs during the time course of a citalopram treatment suggests that NMDAR function is also modulated during citalopram treatment. Since a number of features of NMDAR gating and function are dependent on its subunit composition, we next asked whether the subunit composition of NMDARs onto 5-HT neurons is regulated during the time course of a citalopram treatment. Because of the well-characterized subunit dependence of NMDAR decay kinetics (Vicini et al., 1998), we probed the subunit composition of synaptic NMDARs by first analyzing the decay of pharmacologically isolated NMDAR-mediated eEPSCs. In control

S.D. Geddes et al. / Neuropharmacology 95 (2015) 130e143

139

Fig. 6. A sustained citalopram treatment did not have detectable effects on the subunit composition of glutamate receptors onto dmDRN 5-HT neurons. A1, Representative traces of AMPAR-mediated EPSCs evoked while voltage clamping the neuron at 70, 40, 0, þ20 and þ40 mV. Recordings were performed in the presence of bicuculline (20 mM), D-L, APV (100 mM), and intracellular spermine (100 mM). A2, Average currentevoltage (IV) relationship curve following 2 days of VEH (black) and CIT (green) treatment. B, Left, Scatter plot of normalized average amplitudes of evoked AMPAR-mediated EPSCs (Vm, 70 mV) before (baseline) and during a 25 min post-baseline application of the GluA2-selective antagonist NASPM (30 mM) from 2 day treated animals. Right, representative traces from before and after application of NASPM. C1, Representative traces of IeV relationship of evoked AMPARmediated EPSCs following 7 days of sustained VEH and CIT treatment. C2, Average IeV relationship curve following 7 days of VEH (black) and CIT (green) treatment. D, Left, Scatter plot depicting normalized average amplitudes of evoked AMPAR-mediated EPSCs (Vm, 70 mV) before (baseline) and during a 25 min post-baseline application of the GluA2selective antagonist NASPM (30 mM) from 7 day treated animals. Right, representative traces from before and after application of NASPM. E1, Pharmacologically isolated NMDAR EPSCs (Vm, þ40 mV) following 2 days of VEH- (black) and CIT- (green) treatment. E2, NMDAR weighted tau was not significantly different between VEH and CIT treated animals at the 2-day time point. F, NMDAR-mediated EPSC did not show any significant difference in ifenprodil (3 mM) sensitivity following 2 days of VEH and CIT treatment. G1, Isolated NMDAR-mediated EPSCs (Vm, þ40 mV) following 7 days of VEH- and CIT-treatment. G2, NMDAR weighted tau was not significantly different between VEH- and CIT-treated animals at the 7-day time point. H, NMDAR-mediated EPSCs did not show a significant difference in ifenprodil sensitivity following 2 days of VEH and CIT treatment. Calibration: A1, left, 10 pA, 20 ms; right, 25 pA, 20 ms; B, top, 25 pA, 10 ms; bottom, 50 pA, 10 ms; C1, left, 20 pA, 25 ms; right, 10 pA, 10 ms; D, top, 50 pA, 10 ms; bottom, 25 pA, 10 ms; E1 & G1, 500 ms; F & H, 10 pA, 50 ms. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

140

S.D. Geddes et al. / Neuropharmacology 95 (2015) 130e143

animals, NMDAR-mediated eEPSCs decayed with a time constant of 207.3 ± 15.3 ms (Fig. 6E, G, n ¼ 41, vehicle data from both timepoint groups; weighted tau of a bi-exponential fit; see Materials and Methods) and were reduced by 53.5 ± 3.9% (Fig. 6F, H, n ¼ 15, vehicle data from both time-point groups) by bath administration of the preferential GluN2B-selective antagonist ifenprodil. As a means of comparison, these functional metrics put NMDAR-mediated eEPSCs onto 5-HT neurons close to those observed at the canonical CA1 synapse in the hippocampus (e.g., (Gray et al., 2011), that are composed of GluN1, GluN2A and GluN2B subunits, forming a mixture of di-heteromeric and tri-heteromeric NMDARs (Rauner and Kohr, 2011; Tovar et al., 2013; Soares and Lee, 2013). Recordings from slices from citalopram-treated animals and interleaved controls showed that neither the decay kinetics nor the ifenprodil sensitivity of NMDAR-mediated eEPSCs onto 5-HT neurons were altered by any of the prolonged treatment regimens (2or -7-day) with citalopram (Fig. 6EeH, E, VEH, 216.1 ± 23.4 ms,

n ¼ 18; CIT, 298.4 ± 45.3 ms, n ¼ 18; p ¼ 0.1; F, VEH, 52.4 ± 5.1%, n ¼ 10; CIT, 59.8 ± 7.8%, n ¼ 8; p ¼ 0.4; G, VEH, 200.4 ± 20.5 ms, n ¼ 23; CIT, 259.4 ± 28.4 ms, n ¼ 20; p ¼ 0.1; H, VEH, 56.3 ± 5.5%, n ¼ 5; CIT, 49.6 ± 16.5%, n ¼ 6; p ¼ 0.7) Thus, the profound reduction of NMDAR function observed early during the time course of citalopram treatment, as well as its recovery, occurred independently of significant alterations in the subunit makeup of the channel, at least as it relates to GluN2A/2B ratio. 4. Discussion Here, we report that a treatment with an SSRI induces a concerted and dynamic regulation of glutamatergic synaptic strength onto 5-HT neurons. A sustained treatment with citalopram initially leads to a reduction in synaptic strength that was expressed by both postsynaptic (reduction in the function and/or number of both the AMPA and the NMDA subtypes of glutamate

Fig. 7. Summary of citalopram-induced alterations of glutamate synapses onto 5-HT neurons of the dmDRN.

S.D. Geddes et al. / Neuropharmacology 95 (2015) 130e143

receptors) and presynaptic (reduction in glutamate release probability) mechanisms. By approximately one week of sustained treatment, this reduction in transmission had however waned and was superseded by an adaptive, homeostatic-like, increase in synaptic strength (Fig. 7). Because an increase in 5-HT neurotransmission is necessary for the clinical efficacy of antidepressants, this time-dependent modulation of the function of the synaptic network in which 5-HT neurons operate may be intimately involved not only in the clinical efficacy of these medications, but also in the distressful delay of the onset of their therapeutic effects. Several studies have consistently shown that in vivo treatments with several types of SSRIs (including citalopram) initially lead to a substantial reduction in the mean firing activity of 5-HT neurons that gradually recovers by 2e3 weeks of treatment. Borrowing from the well-described phenomenon of homeostatic plasticity, we hypothesized that this reduction in firing activity induced by SSRIs may trigger the emergence of homeostatic-like processes in these 5-HT neurons. Indeed, work in several experimental preparations, including following in vivo sensory deprivation paradigms (He et al., 2012; Lee, 2012; Goel and Lee, 2007; Gao et al., 2010; Deeg and Aizenman, 2011), show that neurons that are experimentally deprived of activity for a prolonged period of time regulate a number of cellular processes to restore their excitability levels, including upregulating excitatory synaptic function (Soares et al., 2013; Lee et al., 2013; Turrigiano, 2008). Although this form of plasticity has overwhelmingly been studied in cortical neurons, more recent evidence shows that other cell types are homeostatically regulated (eg., (Mendoza Schulz et al., 2014; Sun and Wolf, 2009; Beique, 2009), and it is likely that analogous mechanisms operate in monoaminergic neurons as well. The delayed increase in glutamatergic synapse strength occurring during an SSRI treatment outlined here is consistent with this idea. Moreover, the time course of this adaptive response suggests that it may be intimately involved in the slow onset of the antidepressant response observed in humans. Future studies will be required to not only identify the degree of mechanistic and molecular convergence between homeostatic mechanisms operant in 5-HT neurons and those in cortical neurons, but also to firmly establish whether the occurrence of homeostatic mechanisms in 5-HT neurons are necessary for the antidepressant effect. The adaptive increase in glutamate synapse function following a prolonged treatment with citalopram appeared to occur predominantly through postsynaptic mechanisms. This was primarily evidenced by the observation that the recovery (and overshoot) of both the amplitude and the frequency of mEPSCs following 7 days of citalopram was not accompanied by a parallel recovery in release probability of glutamate synapses onto 5-HT neurons. In combination with those from the NSFA analysis, these results suggest that the plasticity process engaged in 5-HT neurons during citalopram treatment effectively upregulates the number and function of glutamate receptors per synapse as well as the total number of synapses. Indeed, upregulation in the frequency of mEPSCs occurring independently of changes in release probability (irrespective of changes in amplitude) is traditionally interpreted as reflecting an increase in the number of AMPAR-containing synapses (e.g., (Beique et al., 2006; Beique and Andrade, 2003; El-Husseini et al., 2000; Adesnik et al., 2008). Whereas presynaptic mechanisms have been shown to be involved in homeostatic plasticity at some synapses (eg.,(Frank, 2014; Lindskog et al., 2010; Jakawich et al., 2010)), homeostatic adaption in hippocampus can be mediated solely by postsynaptic mechanisms (eg., (Arendt et al., 2013)). Altogether, the behavior of glutamate synapse function onto 5-HT neurons during the time course of an SSRI treatment (specifically, from 2 days onwards to 7 days) is fully consistent with that expected for a homeostatically-driven synaptic plasticity process.

141

In contrast to the homeostatic-like upregulation of synapse function observed following prolonged SSRI treatment, the robust reduction in synaptic strength observed early in the treatment was unexpected. This effect was multifaceted in that it involved both pre- and postsynaptic mechanisms. First, our electrophysiological analyses suggest that both AMPA- and NMDA-receptor function were reduced early in the treatment with citalopram. Second, paired-pulse ratio analysis revealed that glutamate release probability was decreased in response to short-term treatment with citalopram. These alterations in synaptic transmission most likely reflect an LTD-like phenomenon induced by the in vivo treatment with citalopram rather than a pharmacological effect by residual drug in the slices at the time of recordings. Although the cellular mechanisms causing this LTD-like phenomenon at glutamate synapses is unclear at present, it is pertinent to speculate more broadly on the functional consequence of this initial shut-down of excitatory drive onto 5-HT neurons induced by an SSRI. More specifically, a number of reports suggest that antidepressant therapy is associated with increases in suicidal behavior early in the treatment, especially in adolescent populations (Licinio and Wong, 2005; Murray et al., 2005). Although it has not been unambiguously established that this paradoxical increase in suicidal ideation reflects a property of antidepressants themselves rather than the unabated progression of the disease, the transient nature of the reduction in glutamate drive of neurons involved in mood regulation described here during the initial phases of an SSRI treatment may have a contributory role to this distressing aspect of SSRIs. Future studies will be required to directly address this possibility. Although the role of the 5-HT and glutamate system in mood regulation and in the mechanism of action of antidepressants is well recognized, little mechanistic attention has been given to the synaptic network that modulates the excitability of 5-HT neurons and how they are modulated by SSRIs. By describing a dynamic regulation of excitatory glutamatergic function onto 5-HT neurons of the DRN by a prolonged treatment with an SSRI, these results expand the repertoire of cellular processes that are modulated by antidepressants. An outstanding challenge that remains lies in identifying with precision which of the numerous synaptic netwoks that innervates the DRN, including inhibitory networks (Tao and Auerbach, 2003; 2000), are regulated by antidepressants, including by NMDAR antagonists such as ketamine that have rapid antidepressant effects (Autry et al., 2011; Li et al., 2010; Diazgranados et al., 2010). A clearer mechanistic understanding of the synaptic network that regulates 5-HT neuron function may open up new avenues for rational therapeutic strategies for faster acting and higher-efficacy antidepressants. Acknowledgments This work was supported by an award from the Brain and Behavior Research Foundation (JCB) and by grants from the Canadian Institute for Health Research (JCB; RB) and by the Canadian Partnership for Stroke Recovery (JCB). We would like to thank ïque and Bergeron labs for helpful comments members of the Be during the completion of this work. We would also like to thank Dr. Tim Benke (UC Denver) for personal communications regarding NSFA. Thank you to Kevin F.H. Lee and Cary Soares for help with some analytical tools and thank you to Dr. Denise Cook for reading this manuscript. References Adesnik, H., Li, G., During, M.J., Pleasure, S.J., Nicoll, R.A., 2008. NMDA receptors inhibit synapse unsilencing during brain development. Proc. Natl. Acad. Sci. U. S. A. 105, 5597e5602.

142

S.D. Geddes et al. / Neuropharmacology 95 (2015) 130e143

Aghajanian, G.K., Vandermaelen, C.P., 1982. Intracellular recordings from serotonergic dorsal raphe neurons: pacemaker potentials and the effect of LSD. Brain Res. 238, 463e469. Albert, P.R., Lemonde, S., 2004. 5-HT1A receptors, gene repression, and depression: guilt by association. Neuroscientist 10, 575e593. Arendt, K.L., Sarti, F., Chen, L., 2013. Chronic inactivation of a neural circuit enhances LTP by inducing silent synapse formation. J. Neurosci. 33, 2087e2096. Artigas, F., 1993. 5-HT and antidepressants: new views from microdialysis studies. Trends Pharmacol. Sci. 14, 262. Autry, A.E., Adachi, M., Nosyreva, E., Na, E.S., Los, M.F., Cheng, P.F., Kavalali, E.T., Monteggia, L.M., 2011. NMDA receptor blockade at rest triggers rapid behavioural antidepressant responses. Nature 475, 91e95. Beique, J.C., 2009. Homeostatic plasticity in a reward processing region: accumbens neurons scale too! Eur. J. Neurosci. 30 (4), 537e538. Beique, J.C., Andrade, R., 2003. PSD-95 regulates synaptic transmission and plasticity in rat cerebral cortex. J. Physiol. 546, 859e867. Beique, J.C., Huganir, R.L., 2009. AMPA receptor subunits get their share of the pie. Neuron 62, 165e168. Beique, J.C., Lin, D.T., Kang, M.G., Aizawa, H., Takamiya, K., Huganir, R.L., 2006. Synapse-specific regulation of AMPA receptor function by PSD-95. Proc. Natl. Acad. Sci. U. S. A. 103, 19535e19540. Bel, N., Artigas, F., 1993. Chronic treatment with fluvoxamine increases extracellular serotonin in frontal cortex but not in raphe nuclei. Synapse 15, 243e245. Blier, P., De Montigny, C., 1994. Current advances and trends in the treatment of depression. Trends Pharmacol. Sci. 15, 220e226. Blier, P., De Montigny, C., Chaput, Y., 1987. Modifications of the serotonin system by antidepressant treatments: implications for the therapeutic response in major depression. J. Clin. Psychopharmacol. 7, 24Se35S. Blier, P., De Montigny, C., Chaput, Y., 1990. A role for the serotonin system in the mechanism of action of antidepressant treatments: preclinical evidence. J. Clin. Psychiatry 51 (Suppl. l), 14e20 discussion 21. Blier, P., Pineyro, G., El Mansari, M., Bergeron, R., De Montigny, C., 1998. Role of somatodendritic 5-HT autoreceptors in modulating 5-HT neurotransmission. Ann. N. Y. Acad. Sci. 861, 204e216. Calizo, L.H., Akanwa, A., Ma, X., Pan, Y.Z., Lemos, J.C., Craige, C., Heemstra, L.A., Beck, S.G., 2011. Raphe serotonin neurons are not homogenous: electrophysiological, morphological and neurochemical evidence. Neuropharmacology 61, 524e543. Chopra, K., Kumar, B., Kuhad, A., 2011. Pathobiological targets of depression. Expert Opin. Ther. Targets 15, 379e400. De Montigny, C., Aghajanian, G.K., 1978. Tricyclic antidepressants: long-term treatment increases responsivity of rat forebrain neurons to serotonin. Science 202, 1303e1306. De Montigny, C., Blier, P., 1984. Effects of antidepressant treatments on 5-HT neurotransmission: electrophysiological and clinical studies. Adv. Biochem. Psychopharmacol. 39, 223e239. Deeg, K.E., Aizenman, C.D., 2011. Sensory modality-specific homeostatic plasticity in the developing optic tectum. Nat. Neurosci. 14, 548e550. Delgado, P.L., Charney, D.S., Price, L.H., Aghajanian, G.K., Landis, H., Heninger, G.R., 1990. Serotonin function and the mechanism of antidepressant action. Reversal of antidepressant-induced remission by rapid depletion of plasma tryptophan. Arch. Gen. Psychiatry 47, 411e418. Diazgranados, N., Ibrahim, L., Brutsche, N.E., Newberg, A., Kronstein, P., Khalife, S., Kammerer, W.A., Quezado, Z., Luckenbaugh, D.A., Salvadore, G., MachadoVieira, R., Manji, H.K., Zarate Jr., C.A., 2010. A randomized add-on trial of an Nmethyl-D-aspartate antagonist in treatment-resistant bipolar depression. Arch. Gen. Psychiatry 67, 793e802. Dobrunz, L.E., Stevens, C.F., 1997. Heterogeneity of release probability, facilitation, and depletion at central synapses. Neuron 18, 995e1008. El Mansari, M., Sanchez, C., Chouvet, G., Renaud, B., Haddjeri, N., 2005. Effects of acute and long-term administration of escitalopram and citalopram on serotonin neurotransmission: an in vivo electrophysiological study in rat brain. Neuropsychopharmacology 30, 1269e1277. El-Husseini, A.E., Schnell, E., Chetkovich, D.M., Nicoll, R.A., Bredt, D.S., 2000. PSD95 involvement in maturation of excitatory synapses. Science 290, 1364e1368. Frank, C.A., 2014. Homeostatic plasticity at the Drosophila neuromuscular junction. Neuropharmacology 78, 63e74. Gao, M., Sossa, K., Song, L., Errington, L., Cummings, L., Hwang, H., Kuhl, D., Worley, P., Lee, H.K., 2010. A specific requirement of Arc/Arg3.1 for visual experience-induced homeostatic synaptic plasticity in mouse primary visual cortex. J. Neurosci. 30, 7168e7178. Goel, A., Lee, H.K., 2007. Persistence of experience-induced homeostatic synaptic plasticity through adulthood in superficial layers of mouse visual cortex. J. Neurosci. 27, 6692e6700. Gray, J.A., Shi, Y., Usui, H., During, M.J., Sakimura, K., Nicoll, R.A., 2011. Distinct modes of AMPA receptor suppression at developing synapses by GluN2A and GluN2B: single-cell NMDA receptor subunit deletion in vivo. Neuron 71, 1085e1101. Haj-Dahmane, S., 2001. D2-like dopamine receptor activation excites rat dorsal raphe 5-HT neurons in vitro. Eur. J. Neurosci. 14, 125e134. Haj-Dahmane, S., Shen, R.Y., 2005. The wake-promoting peptide orexin-B inhibits glutamatergic transmission to dorsal raphe nucleus serotonin neurons through retrograde endocannabinoid signaling. J. Neurosci. 25, 896e905.

Hartveit, E., Veruki, M.L., 2007. Studying properties of neurotransmitter receptors by non-stationary noise analysis of spontaneous postsynaptic currents and agonist-evoked responses in outside-out patches. Nat. Protoc. 2, 434e448. He, K., Petrus, E., Gammon, N., Lee, H.K., 2012. Distinct sensory requirements for unimodal and cross-modal homeostatic synaptic plasticity. J. Neurosci. 32, 8469e8474. Isaac, J.T., Ashby, M., Mcbain, C.J., 2007. The role of the GluR2 subunit in AMPA receptor function and synaptic plasticity. Neuron 54, 859e871. Jakawich, S.K., Nasser, H.B., Strong, M.J., Mccartney, A.J., Perez, A.S., Rakesh, N., Carruthers, C.J., Sutton, M.A., 2010. Local presynaptic activity gates homeostatic changes in presynaptic function driven by dendritic BDNF synthesis. Neuron 68, 1143e1158. Kim, J., Tsien, R.W., Alger, B.E., 2012. An improved test for detecting multiplicative homeostatic synaptic scaling. PLoS One 7, e37364. Lee, H.K., 2012. Ca-permeable AMPA receptors in homeostatic synaptic plasticity. Front. Mol. Neurosci. 5, 17. Lee, K.F., Soares, C., Beique, J.C., 2013. Tuning into diversity of homeostatic synaptic plasticity. Neuropharmacology 78, 31e37. Li, N., Lee, B., Liu, R.J., Banasr, M., Dwyer, J.M., Iwata, M., Li, X.Y., Aghajanian, G., Duman, R.S., 2010. mTOR-dependent synapse formation underlies the rapid antidepressant effects of NMDA antagonists. Science 329, 959e964. Li, Y.W., Bayliss, D.A., 1998. Presynaptic inhibition by 5-HT1B receptors of glutamatergic synaptic inputs onto serotonergic caudal raphe neurones in rat. J. Physiol. 510 (Pt. 1), 121e134. Licinio, J., Wong, M.L., 2005. Depression, antidepressants and suicidality: a critical appraisal. Nat. Rev. Drug Discov. 4, 165e171. Lindskog, M., Li, L., Groth, R.D., Poburko, D., Thiagarajan, T.C., Han, X., Tsien, R.W., 2010. Postsynaptic GluA1 enables acute retrograde enhancement of presynaptic function to coordinate adaptation to synaptic inactivity. Proc. Natl. Acad. Sci. U. S. A. 107, 21806e21811. Mendoza Schulz, A., Jing, Z., Sanchez Caro, J.M., Wetzel, F., Dresbach, T., Strenzke, N., Wichmann, C., MOSER, T., 2014. Bassoon-disruption slows vesicle replenishment and induces homeostatic plasticity at a CNS synapse. EMBO J. 33, 512e527. Murray, M.L., Wong, I.C., Thompson, M., 2005. Do selective serotonin reuptake inhibitors cause suicide? Antidepressant prescribing to children and adolescents by GPs has fallen since CSM advice. BMJ 330, 1151. Neumeister, A., 2003. Tryptophan depletion, serotonin, and depression: where do we stand? Psychopharmacol. Bull. 37, 99e115. Pan, Z.Z., Williams, J.T., 1989. GABA- and glutamate-mediated synaptic potentials in rat dorsal raphe neurons in vitro. J. Neurophysiol. 61, 719e726. Papp, M., Nalepa, I., Antkiewicz-Michaluk, L., Sanchez, C., 2002. Behavioural and biochemical studies of citalopram and WAY 100635 in rat chronic mild stress model. Pharmacol. Biochem. Behav. 72, 465e474. Pollak Dorocic, I., Furth, D., Xuan, Y., Johansson, Y., Pozzi, L., Silberberg, G., Carlen, M., Meletis, K., 2014. A whole-brain atlas of inputs to serotonergic neurons of the dorsal and median raphe nuclei. Neuron 83, 663e678. Rauner, C., Kohr, G., 2011. Triheteromeric NR1/NR2A/NR2B receptors constitute the major N-methyl-D-aspartate receptor population in adult hippocampal synapses. J. Biol. Chem. 286, 7558e7566. Robison, A.J., Vialou, V., Sun, H.S., Labonte, B., Golden, S.A., Dias, C., Turecki, G., Tamminga, C., Russo, S., Mazei-Robison, M., Nestler, E.J., 2014. Fluoxetine epigenetically alters the CaMKIIalpha promoter in nucleus accumbens to regulate DeltaFosB binding and antidepressant effects. Neuropsychopharmacology 39, 1178e1186. Segrave, R., Nathan, P.J., 2005. Pindolol augmentation of selective serotonin reuptake inhibitors: accounting for the variability of results of placebo-controlled double-blind studies in patients with major depression. Hum. Psychopharmacol. 20, 163e174. Shepherd, J.D., Huganir, R.L., 2007. The cell biology of synaptic plasticity: AMPA receptor trafficking. Annu Rev. Cell. Dev. Biol. 23, 613e643. Sherman, A.D., Sacquitne, J.L., Petty, F., 1982. Specificity of the learned helplessness model of depression. Pharmacol. Biochem. Behav. 16, 449e454. Slattery, D.A., Hudson, A.L., Nutt, D.J., 2004. Invited review: the evolution of antidepressant mechanisms. Fundam. Clin. Pharmacol. 18, 1e21. Soares, C., Lee, K.F., 2013. A prominent role for triheteromeric GluN1/GluN2A/ GluN2B NMDARs at central synapses. J. Neurosci. 33, 14975e14977. Soares, C., Lee, K.F., Nassrallah, W., Beique, J.C., 2013. Differential subcellular targeting of glutamate receptor subtypes during homeostatic synaptic plasticity. J. Neurosci. 33, 13547e13559. Sun, X., Wolf, M.E., 2009. Nucleus accumbens neurons exhibit synaptic scaling that is occluded by repeated dopamine pre-exposure. Eur. J. Neurosci. 30, 539e550. Takamori, K., Yoshida, S., Okuyama, S., 2001. Repeated treatment with imipramine, fluvoxamine and tranylcypromine decreases the number of escape failures by activating dopaminergic systems in a rat learned helplessness test. Life Sci. 69, 1919e1926. Tao, R., Auerbach, S.B., 2000. Regulation of serotonin release by GABA and excitatory amino acids. J. Psychopharmacol. 14, 100e113. Tao, R., Auerbach, S.B., 2003. Influence of inhibitory and excitatory inputs on serotonin efflux differs in the dorsal and median raphe nuclei. Brain Res. 961, 109e120. Telner, J.I., Singhal, R.L., 1981. Effects of nortriptyline treatment on learned helplessness in the rat. Pharmacol. Biochem. Behav. 14, 823e826. Telner, J.I., Singhal, R.L., Lapierre, Y.D., 1981. Reversal of learned helplessness by nortriptyline. Prog. Neuropsychopharmacol. 5, 587e590.

S.D. Geddes et al. / Neuropharmacology 95 (2015) 130e143 Tovar, K.R., Mcginley, M.J., Westbrook, G.L., 2013. Triheteromeric NMDA receptors at hippocampal synapses. J. Neurosci. 33, 9150e9160. Turrigiano, G., 2011. Too many cooks? Intrinsic and synaptic homeostatic mechanisms in cortical circuit refinement. Annu Rev. Neurosci. 34, 89e103. Turrigiano, G.G., 2008. The self-tuning neuron: synaptic scaling of excitatory synapses. Cell 135, 422e435. Turrigiano, G.G., Nelson, S.B., 2004. Homeostatic plasticity in the developing nervous system. Nat. Rev. Neurosci. 5, 97e107. Venzala, E., Garcia-Garcia, A.L., Elizalde, N., Delagrange, P., Tordera, R.M., 2012. Chronic social defeat stress model: behavioral features, antidepressant action, and interaction with biological risk factors. Psychopharmacol. Berl. 224, 313e325.

143

Vicini, S., Wang, J.F., Li, J.H., Zhu, W.J., Wang, Y.H., Luo, J.H., Wolfe, B.B., Grayson, D.R., 1998. Functional and pharmacological differences between recombinant Nmethyl-D-aspartate receptors. J. Neurophysiol. 79, 555e566. Watt, A.J., Van Rossum, M.C., Macleod, K.M., Nelson, S.B., Turrigiano, G.G., 2000. Activity coregulates quantal AMPA and NMDA currents at neocortical synapses. Neuron 26, 659e670. Weissbourd, B., Ren, J., Deloach, K.E., Guenthner, C.J., Miyamichi, K., Luo, L., 2014. Presynaptic partners of dorsal raphe serotonergic and GABAergic neurons. Neuron 83, 645e662. Young, S.N., Smith, S.E., Pihl, R.O., Ervin, F.R., 1985. Tryptophan depletion causes a rapid lowering of mood in normal males. Psychopharmacol. Berl. 87, 173e177.