Abstract
It is well established that, during neural circuit development, glutamatergic synapses become strengthened via NMDA receptor (NMDAR)-dependent upregulation of AMPA receptor (AMPAR)-mediated currents. In addition, however, it is known that the neuromodulator serotonin is present throughout most regions of the vertebrate brain while synapses are forming and being shaped by activity-dependent processes. This suggests that serotonin may modulate or contribute to these processes. Here, we investigate the role of serotonin in the developing retinotectal projection of the Xenopus tadpole. We altered endogenous serotonin transmission in stage 48/49 (∼10–21 days postfertilization) Xenopus tadpoles and then carried out a set of whole-cell electrophysiological recordings from tectal neurons to assess retinotectal synaptic transmission. Because tadpole sex is indeterminate at these early stages of development, experimental groups were composed of randomly chosen tadpoles. We found that pharmacologically enhancing and reducing serotonin transmission for 24 h up- and downregulates, respectively, AMPAR-mediated currents at individual retinotectal synapses. Inhibiting 5-HT2 receptors also significantly weakened AMPAR-mediated currents and abolished the synapse strengthening effect seen with enhanced serotonin transmission, indicating a 5-HT2 receptor–dependent effect. We also determine that the serotonin-dependent upregulation of synaptic AMPAR currents was mediated via an NMDAR-independent, PI3K-dependent mechanism. Altogether, these findings indicate that serotonin regulates AMPAR currents at developing synapses independent of NMDA transmission, which may explain its role as an enabler of activity-dependent plasticity.
Significance Statement
Glutamate is the main excitatory transmitter in the central nervous system, and so the formation of synapses to support robust glutamatergic transmission is important for healthy brain function. It is well established that, during development, activity-dependent, Hebbian-like plasticity strengthens and refines glutamatergic synapses. Here, we report that the neuromodulator serotonin also strengthens developing glutamatergic synapses. This strengthening occurs through an activity-independent mechanism that involves the activation of phosphoinositide signaling. Overall, this study suggests that a combination of activity-dependent and activity-independent mechanisms strengthens glutamatergic synapses during development.
Introduction
In the vertebrate central nervous system, glutamate mediates most excitatory synaptic transmission mainly through two types of ionotropic receptors, α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid receptors (AMPARs) and N-methyl-D-aspartate receptors (NMDARs). During development, newly formed glutamatergic synapses are composed mostly of NMDARs. As development proceeds, synapses are strengthened through an activity-dependent, Hebbian-like mechanism that adds or upregulates AMPARs at the synapse (Durand et al., 1996; Liao et al., 1995; Wu et al., 1996; Xu et al., 2020).
While the current model of glutamate synapse formation and strengthening does not include a role for neuromodulators, it is well known that many neuromodulators are present in the developing nervous system and that they function as trophic factors during this time (Nguyen et al., 2001). The neuromodulator serotonin is one such neuromodulator. While a universal and well-known modulator of mature neural circuit function, serotonin is also known to play roles in many aspects of development including morphogenesis, neurogenesis, neural differentiation, and developmental neural plasticity (Lauder and Krebs, 1978; Sodhi and Sanders-Bush, 2004).
Here, we examine the role of endogenous serotonin in the development of the retinotectal projection of the developing Xenopus laevis tadpole. The Xenopus tadpole retinotectal projection is composed of the retinal ganglion cells (RGCs) in the eye, which project their axons to the optic tectum where they form synapses onto the dendrites of tectal neurons. The retinotectal synapse is glutamatergic: RGC axons release glutamate and postsynaptic tectal neurons express AMPA and NMDA receptors. It is well established that the maturation/strengthening of this projection involves an NMDAR-dependent mechanism that upregulates synaptic AMPAR-mediated currents (Wu et al., 1996; reviewed in Pratt et al., 2016). While there are no previous reports of a role for serotonin in the development of the retinotectal projection, our work showing that the innate preference for light displayed by Xenopus tadpoles is modulated by serotonin (Hunt et al., 2020; Bruno et al., 2022), the presence of serotonin receptors on tectal neurons (De Luccini et al., 2003), and our observation of dense innervation of serotonin immuno-positive axons throughout the neuropil of the tadpole optic tectum (Fig. 1A) led us to hypothesize that serotonin may contribute to the development of this circuit. To investigate this possibility, we pharmacologically increased or decreased endogenous serotonin transmission for ∼24 h in developing stage 48/49 Xenopus tadpoles (∼10–21 d post fertilization) followed by whole-cell electrophysiological recordings of RGC-evoked responses in tectal neurons using an isolated brain preparation. By developmental stages 48/49, the retinotectal circuit is known to be relatively stable and tadpoles display robust visually guided behaviors (Pratt and Aizenman, 2007; Dong et al., 2009). We found that increasing or decreasing serotonin signaling strengthened or weakened, respectively, retinotectal synapses. The changes in the strength of transmission were determined to be manifested mainly through the up- or downregulation of AMPAR-mediated currents at individual synapses. We also found that the strengthening was mediated via an NMDAR-independent, PI3K-dependent mechanism.
Enhancing serotonin transmission strengthens, and reducing serotonin transmission weakens, retinotectal synaptic transmission. Ai, Bright-field image of a developmental stage 48 tadpole brain. The dashed box surrounds the left optic tectum. Scale bar, 400 μm. ii, Horizontal section, bright-field image, focused on the left optic tectum. The optic tectum is composed of layers of tectal neuron somata which reside adjacent to the midline and a more lateral neuropil where RGC axons synapse onto distal regions of tectal neuron dendrites. The dashed line shows the border between the more medial somata layers and the lateral neuropil. iii, Image of the same section as shown in ii, showing immuno-positive serotonergic axons innervating the neuropil. iv, Overlay of images ii and iii. Scale bar shown in iii represents 50 μm and applies to panels ii–iv. C, caudal; R, rostral; OB, olfactory bulb; OT, optic tectum. B, Timeline schematic. Tadpoles are exposed to the SSRI fluoxetine (to enhance serotonin) or pCPA (to reduce endogenous serotonin transmission) at approximately 2:00 P.M. in the afternoon (T = 0). Whole-cell electrophysiological recordings from tectal neurons were carried out at ∼24 h later, between 12:00 P.M. and 6:00 P.P. the following day. C, Schematic of the filleted isolated brain preparation used for electrophysiological recordings. RGC axons are controlled by placing a bipolar stimulation electrode on the optic chiasm. RGC-evoked responses are recorded from tectal neurons in whole-cell, voltage-clamp mode to record synaptic currents. OC, optic chiasm; HB, hindbrain; D, dorsal; M, medial. D, Examples of a range of maximum RGC-evoked responses recorded from tectal neurons of (left) control, (middle) fluoxetine, and (right) pCPA-exposed tadpoles. E, Dot plots showing the average maximum RGC-evoked current amplitudes for tectal neurons recorded from batch-matched control and fluoxetine-exposed tadpoles (control, n = 25; fluoxetine, n = 32; p = 0.0002; unpaired t test), and (F) batch-matched control and pCPA-exposed tadpoles (control, n = 21; pCPA, n = 27; p < 0.0001; Mann–Whitney test). Each dot represents the average of the five biggest RGC-evoked response amplitudes recorded from an individual neuron. Black bars represent the overall average amplitude for each group. ***p < 0.001. G, Dot plots showing the average maximum RGC-evoked current amplitudes recorded from batch-matched control and acute (bath applied) fluoxetine- and pCPA-exposed tectal neurons (control, n = 24; fluoxetine, n = 20; pCPA, n = 11; control vs fluoxetine, p = 0.2975; control vs pCPA, p = 0.7662; Mann–Whitney test). NS, not significant.
Materials and Methods
All experimental protocols involving animals was approved by the Institutional Animal Care and Use Committee (IACUC) of the University of Wyoming. Wild-type X. laevis embryos obtained by in-house mating were reared in Steinberg's solution at 22°C on a 12 h light/dark cycle. Tadpoles were staged according to the method of Nieuwkoop and Faber (1994). Developmental stage 48/49 tadpoles (∼10–21 d postfertilization) were used for all the experiments in this study.
Pharmacological manipulations
All pharmacological agents used in this study were prepared as stock solutions, aliquoted, and stored at −20°C until use. Tadpoles were exposed to fluoxetine (1 μM; Sigma-Aldrich, catalog #F132, dissolved in DMSO), p-chlorophenylalanine (pCPA; 15 μM; Sigma-Aldrich, catalog# C3635), 2-amino-5-phosphonovaleric acid (APV; 100 μM; Sigma, catalog #A8054), LY294002 (5 μM; Tocris, catalog #4A/147350, dissolved in DMSO), and ketanserin (20 μM; Sigma-Aldrich, catalog #S006, dissolved in DMSO) by adding the drug to the rearing solution for 24 h followed by whole-cell electrophysiology or immunohistochemistry. The concentrations used were the highest concentrations that did not elicit any noticeable effects on normal swimming patterns and behavior.
For drugs dissolved in DMSO, appropriate vehicle-only (DMSO alone) controls were used. We found no significant differences in control responses with or without DMSO, so both controls were combined when appropriate. For acute exposure experiments, fluoxetine or pCPA was added to the external recording solution and each recording session was completed within 3 h.
Brain preparation
For whole-cell recordings, the isolated brain preparation was carried out as described by Pratt and Aizenman (2007) and Wu et al. (1996): The tadpoles were anesthetized in 0.01% tricaine methanesulfonate (MS222) dissolved in Steinberg's solution and then pinned to a block of Sylgard silicone elastomer submerged in HEPES-buffered extracellular saline (in mM: 115 NaCl, 4 KCl, 3 CaCl2, 3 MgCl2, 5 HEPES, 0.01 glycine, and 10 glucose, pH 7.25). The skin overlying the brain was removed and the brain filleted along the midline to expose the ventral surface of the midbrain. The whole brain was then excised from the tadpole and attached to a submerged Sylgard block in the recording chamber via two insect pins placed through the olfactory bulb and the hindbrain. To access neurons for recording, we removed the ventricular membrane overlying the tectum using a broken glass pipette.
Whole-cell electrophysiology
Tectal neurons were visualized using a Zeiss light microscope with a ×60 water immersion objective. The microscope was connected to a Hamamatsu charge-coupled device (CCD) camera. Whole-cell recordings were carried out at room temperature (20–22°C) using a borosilicate glass micropipette with resistance in the range of 8–12 MΩ containing K+ gluconate internal saline (in mM: 100 potassium gluconate, 8 KCl, 5 NaCl, 1.5 MgCl2, 20 HEPES, 10 EGTA, 2 ATP, and 0.3 GTP, pH 7.2, osmolarity 255 mOsm). Electrophysiological recordings were carried out using an Axon Instruments 700B Multipatch Amplifier and digitized at 10 kHz using a Digidata 1322A Digitizer (Molecular Devices). pClamp software was used for data acquisition, and the traces were analyzed off-line using AxoGraph (version 1.8.0) and graphed using IGOR Pro (version 9.01). For all recordings, GABAergic transmission was inhibited by including picrotoxin (100 μM) in the bath.
To quantify peak Na+ and K+ currents, neurons were held at −60 mV in a whole-cell voltage-clamp configuration, stepped to increasingly more depolarized potentials for a duration of 250 ms, and the resulting voltage-dependent Na+ and K+ currents were recorded. Resting membrane potential (RMP) was determined by switching out of voltage-clamp to “I = 0” mode and waiting an average of 5 s to get a stable RMP. Neurons with series resistances >50 MΩ were not included in the dataset (average series resistance was 23.5, 23.6, 25.2, and 19.37 MΩ for neurons recorded in control, fluoxetine, pCPA, and ketanserin groups, respectively).
For RGC-evoked responses, a bipolar stimulating electrode connected to an ISO-Flex stimulator (A.M.P.I.) was placed on the optic chiasm to activate RGC axons. The total amount of RGC input onto a given neuron was quantified by recording maximum RGC-evoked synaptic currents with the neuron voltage clamped at −60 mV. The maximum synaptic current was defined as the peak current amplitude that no longer increases in response to increasing the stimulation strength of the RGC axons. For every given tectal neuron, five peak currents were averaged to get the maximum evoked response.
For paired pulse recordings, RGC axons were paired-stimulated with a stimulus of intermediate strength, with an inter-stimulus interval of 50 ms. The paired pulse ratios were calculated by dividing the second excitatory postsynaptic current by the first (EPSC2 / EPSC1).
For asynchronous evoked responses, 3 mM Sr2+ was substituted for Ca2+ in the external recording solution. Evoked asEPSCs were analyzed in a 100 ms window starting 15 ms post stimulation.
Immunohistochemistry
Stage 48/49 tadpoles were fixed with 4% paraformaldehyde in 0.1 M phosphate-buffered saline (PBS) solution. They were then cryoprotected in 30% sucrose in PBS overnight at 4°C. Tadpoles were embedded in Optimal Cutting Temperature Compound (OCT, Tissue-Tek), and the brain and eye were serially sectioned (40 μm) in the horizontal plane using a cryostat (Microm HM 505 E).
For the serotonin immunohistochemistry, the slices were permeabilized in a blocking solution containing 5% normal goat serum (NGS) and 1% Triton X-100 (TX) in 0.1 M PBS followed by overnight incubation with primary rat anti-5-HT (1:250; Novus, #NB100-65037) diluted in 0.1 M PBS containing 1% NGS/0.1% TX at 4°C. Primary antibody was detected using Alexa Fluor 488, goat-anti-rat, 1:1,000. For PIP3 immunohistochemistry, the blocking solution contained 5% NGS, 1% TX, and 2% bovine serum albumin (BSA) in 0.1 M PBS. Slices were incubated overnight with mouse anti-PIP3 monoclonal antibody (1:100; Echelon Biosciences, #Z-P345b) diluted in 0.1 M PBS containing 5% NGS/0.1% TX/2%BSA at 4°C and detected using Alexa Fluor 488, rabbit-anti-mouse, 1:1,000. All secondary antibodies were incubated for 2 h. Samples were coverslipped in Vectashield Mounting Medium (Vector Laboratories) for imaging. The images were obtained using an upright Zeiss laser scanning confocal microscope (LSM 700). Images for each experiment were obtained using identical acquisition parameters. Anti-PIP3 immunofluorescence images were quantified by determining the mean fluorescence intensity (MFI) of horizontal eye and brain slices using ImageJ (v1.53k, NIH). Three slices each from two different experiments were analyzed. For each slice analyzed, the MFI of a nonfluorescent area of the same image (the background) was subtracted from the MFI of the tissue region of interest to arrive at the final MFI (Shihan et al., 2021).
Experimental design and statistical analysis
All experiments were carried out in stage 48/49 X. laevis tadpoles. Tadpole sex is indeterminate at these early stages of development. Unless indicated otherwise, summary data are presented using a dot plot that shows all data points. For the maximum RGC-evoked responses, each data point represents the average of 5 of the biggest evoked responses at maximum stimulation. The detailed sample size of each group can be found in the figure legends. All descriptive statistics are reported as mean ± standard error of the mean (SEM). Normality test was determined using the Shapiro–Wilk test. Datasets that passed the normality test were compared using either unpaired t test or ordinary one-way ANOVA with Tukey's multiple comparisons test. Datasets that failed the normality test were compared using nonparametric Mann–Whitney test or Kruskal–Wallis test with Dunn's multiple comparisons. Differences with a p value <0.05 were considered significant. All statistical analysis was done using GraphPad Prism version 9.5.1.
Results
Immunohistochemistry reveals dense serotonergic input targeting the neuropil of the optic tectum
To visualize serotonergic projections, we carried out immunohistochemical studies on developmental stage 48 tadpole brain slices (Fig. 1A). Extensive serotonin immunoreactive axonal inputs were observed to innervate specifically the neuropil of the optic tectum, where the RGC axon terminals form synapses onto tectal neuron dendrites. We also observed that these serotonin immunoreactive axons contained brightly fluorescing varicosities, or boutons. Previous electron microscopy studies in rats established these boutons to be serotonin release sites (Belmer et al. 2017; reviewed in Descarries et al., 2010). Altogether, these observations suggest the presence of functional serotonergic input to the optic tectum, giving rise to the intriguing notion that retinotectal synaptic transmission in the larval tadpole may be modulated by serotonin.
Enhancing and reducing serotonin signaling increases and decreases, respectively, the strength of retinotectal synaptic transmission
To determine the effect of enhanced serotonin signaling on retinotectal synaptic transmission, we pretreated tadpoles for 24 h with fluoxetine, a selective serotonin reuptake inhibitor (SSRI). SSRIs inhibit serotonin uptake from the synapse and so are only effective in the presence of endogenously released serotonin. After tadpoles had been exposed to the SSRI for ∼24 h (Fig. 1B), the strength of retinotectal synaptic transmission was measured by carrying out electrophysiological recordings from tectal neurons in an isolated brain preparation. This preparation allows for the presynaptic RGC axons to be activated via a bipolar stimulating electrode placed on the optic chiasm, and for the resulting synaptic responses to be recorded from tectal neurons in whole-cell configuration (Fig. 1C; Wu et al., 1996; Pratt and Aizenman, 2007; Hamodi and Pratt, 2014). To study only the long-term effects of enhanced serotonin action, we did not add fluoxetine in the external recording solution. Tectal neurons were recorded in voltage-clamp mode, voltage clamped at −60 mV, to measure RGC-evoked synaptic currents. First, the maximum or total amount of RGC input onto a given tectal neuron was measured. For this, the strength of the stimulation electrode was increased—thus recruiting more axons—until the amplitude of the evoked synaptic current no longer increased with increased stimulation strength. This response amplitude reflects the maximal amount of direct monosynaptic RGC input (Liu et al., 2018). The average maximum evoked synaptic current for each neuron was derived by averaging its five biggest responses. We found that exposure to fluoxetine significantly increased the average maximum RGC-evoked EPSC amplitudes (Fig. 1D,E). Conversely, we found that inhibiting endogenous serotonin via exposing tadpoles for 24 h to pCPA, a tryptophan hydroxylase inhibitor commonly used to inhibit serotonin production (Cheng et al., 2016; Bruno et al., 2022), decreased the average maximum RGC-evoked response amplitude, indicating that inhibiting endogenous serotonin signaling weakened retinotectal synaptic transmission (Fig. 1D,F). In summary, these results show that experimentally increasing and decreasing serotonin transmission increased and decreased, respectively, the strength of retinotectal transmission. Neither drug appeared to affect the intrinsic electrical properties expressed by tectal neurons (Table 1). To determine whether acute exposure to fluoxetine or pCPA may alter the function of the retinotectal projection, we carried out recordings with the drug present only in the external recording solution (i.e., no pretreatment). Neither drug appeared to elicit acute effects on retinotectal synaptic transmission (Fig. 1G) nor intrinsic electrical properties of tectal neurons (Table 2).
Active and passive electrical properties of control, 24 h fluoxetine-treated, and 24-h pCPA-treated neurons
Active and passive electrical properties of control, acute fluoxetine-treated, and acute pCPA-treated neurons
The main way in which serotonin modulates retinotectal synaptic transmission is by altering AMPAR-mediated currents at the single synapse level
Next, we addressed how the serotonin effect on retinotectal synaptic transmission was manifested. Quantal analysis dictates that a change in the strength of evoked transmission could be due to a change in the strength of individual synapses, a change in the probability of transmitter release, a change in the number of synapses, or a combination of any of these (Korn and Faber, 1991). To determine whether the serotonin effect involved a change in presynaptic transmitter release, we used an established paired-pulse protocol to generate paired-pulse ratios (PPRs) for control, fluoxetine-, and pCPA-exposed tadpoles (Fig. 2A). Compared with controls, the fluoxetine-exposed group displayed a modest yet statistically significant decrease in PPR, indicating an increase in the probability of presynaptic transmitter release (Fig. 2B,C). The average PPR displayed by the pCPA-exposed group, however, was not different from controls (Fig. 2B,C), suggesting that the weaker maximum RGC-evoked synaptic current observed in the pCPA group is not due to a decrease in probability of release and that endogenous serotonin does not regulate this aspect of synaptic transmission.
Enhancing serotonin transmission increases probability of transmitter release. A, Examples of paired pulse recordings from tectal neurons from a control (top), fluoxetine-exposed (middle), and pCPA-exposed (bottom) tadpole. The pulses are 50 ms apart. B, Plots showing the amplitude of the first evoked EPSC (EPSC1; y-axis) and the second evoked EPSC (EPSC2; x-axis) for individual neurons recorded from control and fluoxetine-exposed (top) and control and pCPA-exposed (bottom) tadpoles. C, Bar graphs showing average paired pulse ratios (EPSC2/EPSC1) for control and fluoxetine groups (top; control, n = 27; fluoxetine, n = 28; p = 0.0169; Mann–Whitney test) and control and pCPA groups (bottom; control, n = 16; pCPA, n = 14; p = 0.8999; not significant; unpaired t test). *p < 0.05.
Next, we investigated whether the serotonin effect on retinotectal synaptic transmission was manifested via changes in the strength of individual retinotectal synapses. For this, RGC-evoked responses were recorded using an external recording solution in which Ca2+ was replaced with strontium (Sr2+). This desynchronizes evoked transmission, thereby allowing for the measurement of individual retinotectal synaptic currents, referred to as asynchronous excitatory postsynaptic currents, (asEPSCs; Xu-Friedman and Regehr, 2000; Deeg and Aizenman, 2011). Examples of asEPSCs recorded from control, fluoxetine, and pCPA neurons are shown in Figure 3A. We observed that enhancing serotonin transmission increased the amplitude of individual synaptic currents (Fig. 3B), while inhibiting serotonin decreased the amplitude (Fig. 3C). These differences are reflected in the averaged asEPSC traces (Fig. 3D,E). Given that the amplitude of individual glutamatergic synaptic currents recorded at −60 mV are mediated solely by AMPARs, these results indicate that serotonin alters postsynaptic AMPAR-mediated currents at the level of individual synapses. Given the finding reported by Shutoh et al. (2000) that long-term (1 week) reduction in serotonin levels led to alterations in the subunit composition of AMPARs expressed in the cortex of the rat (increase in GluR2, decrease in GluR1), averaged asEPSC traces were scaled to normalize amplitudes. The scaled average traces showed no differences in rise or decay of the currents (Fig. 3D,E), suggesting that the observed serotonin-dependent alterations in synaptic strength are not manifested through major changes in AMPAR subunit composition.
Serotonin signaling-dependent changes in retinotectal synaptic transmission are manifested through strengthening or weakening of individual synapses. A, Examples of RGC-evoked asynchronous excitatory postsynaptic currents (asEPSCs) recorded from tectal neurons of a control (left), fluoxetine-exposed (middle), and pCPA-exposed (right) tadpole. Vertical dashed lines represent the time window that synaptic events were collected. B, Cumulative probability plot showing the distribution of individual asEPSC amplitudes recorded from tectal neurons of control and fluoxetine-exposed tadpoles. Notice that the entire population of fluoxetine asEPSCs is shifted to the right of control asEPSCs. (Inset) Bar graph showing the average asEPSC amplitude for control and fluoxetine groups (control, n = 24; fluoxetine, n = 23; p < 0.0001; Mann–Whitney test). C, Cumulative probability plot showing the distribution of individual asEPSCs recorded from neurons of control and pCPA-exposed tadpoles. (Inset) Bar graph showing the average asEPSC amplitude for control and pCPA groups (control, n = 16; pCPA, n = 23; p = 0.0033; Mann–Whitney test). Error bars represent the standard error of the mean. D, Average (left) and scaled average (right) asEPSC traces for control and fluoxetine groups. E, Average (left) and scaled average (right) asEPSC traces for control and pCPA groups. F, Dot plots showing frequency of asEPSCs in controls and fluoxetine-exposed neurons (controls, n = 20; fluoxetine, n = 22; p = 0.8121; unpaired t test). G, Dot plots showing frequency of asEPSCs in controls and pCPA-exposed neurons (controls, n = 16; pCPA, n = 24; p = 0.0679; unpaired t test). **p < 0.01, ***p < 0.001, NS, not significant.
While we did not determine directly whether the serotonin-dependent effect on maximum RGC-evoked responses involved changes in synapse number, the following observations suggest that it does not: (1) Similar numbers of asEPSC events per sweep were obtained from control, fluoxetine, and pCPA groups (Fig. 3F,G), and (2) the control versus fluoxetine cumulative probability plot in Figure 3B shows the entire population of asEPSC amplitudes is shifted to the right. If fluoxetine was both strengthening existing synapses and inducing the formation of new ones, we would expect to see a fraction of smaller current amplitudes corresponding to newly formed synapses.
Overall, quantal analysis suggests that the main way in which serotonin modulates retinotectal synaptic transmission is via altering the strength of AMPAR-mediated currents at individual synapses. Therefore, for the remainder of this study, we focus on the mechanism through which serotonin affects the strength of individual retinotectal synapses.
Fluoxetine-induced strengthening of retinotectal synapses is through an NMDAR-independent, PI3 kinase-dependent mechanism
The observed upregulation of AMPAR-mediated currents recorded from fluoxetine-exposed neurons suggests an LTP-like mechanism. If so, then the effect would be expected to require activation of NMDARs. To test this, we carried out the same 24 h exposure to fluoxetine in the presence of NMDAR blocker APV and then measured maximum RGC-evoked responses as previously described. We found that blocking NMDARs did not significantly interfere with the fluoxetine-induced strengthening of retinotectal synaptic transmission (Fig. 4A), indicating an NMDAR-independent mechanism. What then may be the mechanism? Tectal neurons are known to express serotonin type 2B and 2C receptors (De Lucchini et al., 2003). To test whether these type 2 receptors are involved in the serotonin-induced strengthening of retinotectal synapses, we exposed tadpoles to fluoxetine in the presence of the 5-HT2 antagonist ketanserin (20 µM). We found that ketanserin completely blocked the fluoxetine-induced strengthening of average maximum RGC-evoked responses displayed by tectal neurons (Fig. 4B), suggesting that serotonin strengthens retinotectal synapses through 5-HT2 receptors. Furthermore, we observed that exposing tadpoles to the 5-HT2 antagonist alone weakened retinotectal synaptic transmission to levels below that of control (Fig. 4B)—thus similar to the effect elicited by pharmacologically inhibiting serotonin levels via pCPA (Fig. 1F). These serotonin receptor subtypes are G-protein coupled (Conn and Sanders-Bush, 1984, McCorvy and Roth, 2015), which, when activated by serotonin, are known to activate PI3K (Liu and Fanburg, 2006) and the concomitant generation of PIP3 (Niswender et al., 2003). Based on the finding that PIP3 strengthens glutamate synapses via decreasing the rate at which AMPARs are cycled out of the synapse (Arendt et al., 2010), we reasoned that this could be the intracellular pathway through which serotonin strengthens retinotectal synapses. If so, PIP3 levels would be expected to be elevated in fluoxetine-exposed tectal neurons. Consistent with this hypothesis, immunohistochemical studies revealed higher levels of immuno-tagged PIP3 throughout the tectum of the fluoxetine-exposed tadpoles compared with those of controls (Fig. 5A,B). Levels of immuno-tagged PIP3 across the retina of fluoxetine-treated tadpoles, however, were not different from control levels (Fig. 5C,D), suggesting that the locus of fluoxetine-induced strengthening of retinotectal transmission is the optic tectum, not the retina. To further test whether the fluoxetine-induced strengthening of retinotectal synapses is due to serotonin-induced activation of PI3K/PIP3 and the concomitant increase in AMPARs at the synapse, we carried out the same 24 h exposure to fluoxetine in the presence of 5 μM PI3K inhibitor LY294002 (LY). This concentration of LY is known to specifically inhibit PI3K (and the concomitant generation of PIP3) without significantly affecting phosphoinositide kinases required for PIP2 synthesis (Vanhaesebroeck et al., 2001; Arendt et al., 2010). We observed that LY significantly attenuated fluoxetine-induced strengthening of the average maximum evoked EPSC (Fig. 5E) while having no effect on the control maximum evoked EPSC (Fig. 5F). LY also attenuated fluoxetine-induced increase in asEPSC amplitudes (Fig. 5G). Altogether, these results indicate that fluoxetine is upregulating AMPAR-mediated currents at individual synapses through a PI3K-PIP3–dependent signaling mechanism.
Enhanced serotonin transmission strengthens retinotectal synapses through an NMDA receptor-independent mechanism and requires 5-HT2 receptor activation. A, Dot plot showing the maximum RGC-evoked EPSC amplitudes recorded from tectal neurons of control tadpoles, tadpoles exposed to fluoxetine, and tadpoles exposed to fluoxetine + the NMDA receptor blocker APV. Each dot represents the average of the five biggest amplitude responses from an individual neuron. Blocking NMDA receptors did not significantly decrease the fluoxetine-induced strengthening of retinotectal synaptic transmission (control, n = 25; fluoxetine, n = 20; fluoxetine + APV, n = 20; Ctrl vs Fluox, p = 0.0007; Ctrl vs Fluox + APV, p = 0.0092; Fluox vs Fluox + APV, p > 0.9999; Kruskal–Wallis with Dunn's multiple-comparisons test). B, Dot plot showing the maximum RGC-evoked EPSC amplitudes recorded from tectal neurons of control tadpoles, tadpoles exposed to fluoxetine, tadpoles exposed to fluoxetine + the 5-HT2 receptor antagonist ketanserin, and tadpoles exposed to only ketanserin. Blocking 5-HT2 receptors in tectal neurons significantly reduced the maximum RGC-evoked EPSC amplitudes compared with controls and abolished the fluoxetine-induced strengthening of retinotectal synaptic transmission (control, n = 12; fluoxetine, n = 9; fluoxetine + ketanserin, n = 13; ketanserin, n = 11; Ctrl vs Fluox, p < 0.0001; Ctrl vs Ket, p = 0.0076; Ctrl vs Fluox + Ket, p = 0.0244; Fluox vs Ket, p < 0.0001; Fluox vs Fluox + Ket, p < 0.0001; Ket vs Fluox vs Ket, p = 0.9385; ordinary one-way ANOVA with Turkey's multiple-comparisons test). *p < 0.05, **p < 0.01, ***p < 0.001, NS, not significant.
Enhanced serotonin transmission strengthens retinotectal synapses through a PI3K-dependent mechanism. A, PIP3 immunohistochemistry shows that, compared with controls, a 24 h exposure to fluoxetine increases PIP3 levels in tectal neuron somas. Also notice a relatively higher level of PIP3 immunofluorescence in the outer neuropil region in the fluoxetine-exposed slice. Scale bar, 50 μm. B, Dot plots of anti-PIP3 MFI of the optic tectum of controls and of tadpoles exposed to fluoxetine (controls, n = 6; fluoxetine, n = 6; p = 0.0357; unpaired t test). C, PIP3 immunohistochemistry shows no difference in the expression of PIP3 in the retina of controls and 24 h fluoxetine-exposed tadpoles. Scale bar, 50 μm. PRL, photoreceptor layer; ONL, outer nuclear layer; INL, inner nuclear layer; IPL, inner plexiform layer; GCL, ganglion cell layer. D, Dot plots of anti-PIP3 MFI of the retina of controls and of tadpoles exposed to fluoxetine (controls, n = 6; fluoxetine, n = 6; p = 0.3631; unpaired t test). NS, not significant. E, Dot plot showing the maximum RGC-evoked EPSC amplitudes recorded from tectal neurons of control tadpoles, tadpoles exposed to fluoxetine, and tadpoles exposed to fluoxetine + the PI3K inhibitor LY294002 (LY; control, n = 13; fluoxetine, n = 10; fluoxetine + LY, n = 16; Ctrl vs Fluox, p = 0.0024; Ctrl vs Fluox + LY, p = 0.9486; Fluox vs Fluox + LY, p = 0.0034; ordinary one-way ANOVA with Turkey's multiple-comparisons test). F, Dot plot showing the maximum RGC-evoked EPSC amplitudes recorded from tectal neurons of control tadpoles and tadpoles exposed to LY alone (control, n = 9; LY, n = 12; p = 0.6511; Mann–Whitney test). G, Cumulative probability plot showing the distribution of individual asEPSCs recorded from neurons of control tadpoles, tadpoles exposed to fluoxetine, and tadpoles exposed to fluoxetine + LY. Inhibiting PI3K prevented the fluoxetine-induced rightward shift of asEPSCs compared with controls. (Insert) Bar graph showing the average asEPSC amplitude for control, fluoxetine, and fluoxetine + LY groups (control, n = 9; fluoxetine, n = 9; fluoxetine + LY, n = 11; Ctrl vs Fluox, p = 0.0002; Ctrl vs Fluox + LY, p = 0.8253; Fluox vs Fluox + LY, p = 0.0005; ordinary one-way ANOVA with Turkey's multiple-comparisons test). *p < 0.05, **p < 0.01, ***p < 0.001, NS, not significant.
Discussion
The overall findings of this study are that during development of the retinotectal projection, experimentally enhancing and reducing serotonin transmission for 24 h up- and downregulates, respectively, AMPAR-mediated currents at individual retinotectal synapses. It was also determined that the upregulation happens through an NMDAR-independent, PI3K-dependent mechanism. The identification of this mechanism by which serotonin signaling influences AMPAR-mediated currents reveals a link between two known signaling pathways: (1) serotonin-dependent activation of PI3K and concomitant generation of PIP3 and (2) PIP3-dependent upregulation of AMPAR-mediated currents.
Theoretical model of how serotonin strengthens glutamatergic synapses
Altogether, the results of this study suggest a scenario in which serotonin strengthens retinotectal synaptic transmission by binding to 5-HT2 receptors on dendrites of tectal neurons. Activation of 5-HT2 receptors would activate PI3K. Activated PI3K, in turn, would generate PIP3, which would upregulate nearby retinotectal synapses by slowing down the recycling of AMPARs out of the synapse, a mechanism described in detail by Arendt et al. (2010). It is important to note that the focus of this study was on the mechanism by which serotonin upregulates AMPAR-mediated currents at individual synapses, a postsynaptic effect. Thus, the model is specific for the effect of serotonin on the postsynaptic side of the synapse. A presynaptic effect cannot be ruled out, however, given our finding that exposure to fluoxetine significantly increased the probability of transmitter release from presynaptic RGC axon terminals (Fig. 2B,C). In addition, the same study that identified serotonin type 2B and 2C receptors in the optic tectum also reported these same receptors in the proliferative zones of the Xenopus tadpole retina (De Lucchini et al., 2003). Indeed, the effect of enhanced serotonin in the retina may explain the increased probability of transmitter release observed in tadpoles exposed to fluoxetine. Thus, while our data indicate a role for serotonin at the level of the optic tectum, enhanced serotonin may also be modulating the probability of transmitter release from RGC axon terminals.
A role for serotonin in a developing glutamatergic synapse
During development, glutamatergic synapses are known to be strengthened and refined through a Hebbian-like mechanism. This mechanism requires the activation of NMDARs so that calcium can flow through NMDAR channels. The calcium then activates local CaMKII which in turn upregulates AMPAR-mediated currents (Wu et al., 1996). A feature of Hebbian strengthening is that it requires the temporal coordination of the release of glutamate from the presynaptic side of the synapse followed immediately by depolarization of the postsynaptic dendrite to relieve the voltage-dependent magnesium block of NMDAR channels. Early in development, however, most glutamatergic synapses contain no or relatively few AMPARs, and so there exists a catch 22: with a paucity of AMPARs, how can there be sufficient postsynaptic depolarization to relieve the magnesium block? There are theories on how this may happen. For instance, the NMDARs of nascent synapses have been found to be less sensitive to magnesium blockade than those of more mature synapses (Bowe and Nadler, 1999). Another theory suggests that depolarizing GABA currents in immature neurons create sufficient postsynaptic depolarization to relieve the magnesium block (Leinekugel et al., 1997; Akerman and Cline, 2006). Still, the presence of silent synapses (nascent synapses that contain only NMDARs) which do not flux current at hyperpolarized potentials suggest that activation of NMDARs in immature synapses may not happen readily. Our data suggest that serotonin boosts the strength of glutamatergic synapses through the PI3K-PIP3 signaling pathway, thus an additional mechanism that boosts AMPARs at the synapse without the need for activation of NMDARs. The observed reduction in AMPAR-mediated currents caused by reducing endogenous serotonin transmission or by blocking serotonin receptors (Figs. 1F, 4B) suggests that this NMDAR-independent mechanism is important for the normal strengthening of retinotectal synapses (Wu et al., 1996).
Implications of boosting glutamatergic synaptic strengths through PI3K/PIP3
There are many reports that serotonin enables activity-dependent plasticity expressed by developing glutamatergic circuits. This has been studied mostly in the developing visual system. For instance, it is well established that serotonin is required for the proper expression of ocular dominance plasticity in the visual cortex (Gu and Singer, 1995; Wang et al., 1997). Reducing serotonin levels in neonatal rats is reported to inhibit activity-dependent refinement of RGC inputs onto the superior colliculus, the homolog of the amphibian optic tectum (Gonzalez et al., 2008). Studies from the same group show that chronic fluoxetine treatment amplifies lesion-induced plasticity of the RGC inputs to the superior colliculus (Bastos et al., 1999), whereas neonatal reduction in serotonin levels attenuates it (Penedo et al., 2009). Blocking serotonin type 1A receptors in the Rana pipiens adult frog leads to loss of refinement of the topographic map, while blocking type 1B receptors enhances further the already refined circuit (Butt et al., 2002). These are all glutamatergic circuits. Thus, serotonin is thought to be an enhancer or enabler of glutamatergic circuit plasticity during critical periods of development when the circuit is being shaped by activity-dependent mechanisms.
While the intracellular mechanism by which serotonin enables activity-dependent plasticity during development remains unclear, it has been hypothesized that it may involve a phosphoinositide signaling pathway (Gu and Singer, 1995). While our study does not address the role of serotonin in activity-driven plasticity such as refinement or ocular dominance plasticity, it could be that boosting synaptic strengths through a PI3K/PIP3 pathway, as we report here, may be precisely how serotonin enables such activity dependent plasticity. This would also support the hypothesis posed by Gu and Singer (1995) that the way in which serotonin enables ocular dominance plasticity is through a phosphoinositide signaling pathway. In another study, it was shown that the acute application of serotonin facilitates synaptic plasticity on specifically 5-HT2C-expressing layer IV neurons of the visual cortex (Kojic et al., 2000). Because 5-HT2 receptors are linked to the phosphoinositide signaling system, it was proposed that perhaps acute application of serotonin facilitates synaptic plasticity in the developing visual cortex by activating IP3-dependent release of intracellular calcium (Kirkwood, 2000; Kojic et al., 2000). Our data do not suggest this pathway because the concentration of LY that was used in this study only interferes with the PI3K-dependent generation of PIP3. Furthermore, the focus of the Kojic et al. (2000) study is on the acute (phasic) effects of 5-HT on synaptic plasticity, while the focus of our study is on the tonic, or steady-state role of serotonin during development. The signaling pathways activated by tonic versus phasic patterns of serotonin activity may be different and work on different time scales.
The way in which PIP3 enhances AMPAR-mediated currents was previously shown to be through slowing down AMPAR recycling away from the synapse (Arendt et al., 2010). Slowing down the recycling of AMPAR's that are already in the synapse would theoretically boost all AMPAR-containing glutamatergic synapses in a way that preserves their relative synaptic strengths (i.e., the more AMPARs at the synapse, the more the synapse is enhanced by PIP3), perhaps amplifying both LTP and LTD types of plasticity which underlie circuit refinement. Also, because this mechanism is NMDAR independent, it would not interfere with NMDAR-dependent plasticity. These attributes theoretically render serotonin an ideal enabler of activity-driven, NMDAR-dependent plasticity.
In addition to enhancing activity-dependent plasticity during development, work by Lamberto Maffei's group shows that fluoxetine enhances activity-dependent plasticity expressed by the adult visual cortex by increasing BDNF levels and decreasing GABA levels, hallmarks of a younger, more plastic circuit (Maya Vetencourt et al., 2008). In another study, the same group establishes that, similar to fluoxetine treatment, enhanced visual experience increases visual cortical plasticity via elevating serotonin levels—which leads to, again, an increase in BDNF and decrease in GABA (Baroncelli et al., 2010), thus indicating that the way in which enhanced experience increases plasticity is through serotonin, that serotonin is the primum movens. While the intracellular mechanism through which serotonin elicits these effects on the adult visual cortex is not known, it is possible that the way in which serotonin modulates circuit plasticity may depend ultimately on the state of the circuit itself and the specific and perhaps changing set of (intracellular) downstream effectors that are modulated by it.
Endogenous levels of serotonin signaling are intermediate
Overall, the results of this study show that experimentally increasing levels of serotonin activity strengthens retinotectal synapses and that reducing serotonin signaling weakens them. This indicates that endogenous/natural levels of serotonin signaling are intermediate—neither fluoxetine high nor pCPA low. Could this intermediate level of signaling, and thus an intermediate degree of synapse boosting, be a goldilocks level that optimally supports activity-dependent plasticity? While it may seem intuitive that enhancing the strengthening effect of endogenous serotonin via fluoxetine exposure would enhance NMDAR-dependent plasticity, it could also be that strengthening synapses to an abnormally high degree may occlude the ability of these synapses to respond to NMDAR-dependent instruction. This is a remaining question which is generated by, but beyond the scope of, this study.
Footnotes
This work was supported by the National Science Foundation (2212591) to K.G.P. U.G.U. was supported by the National Institute of General Medical Sciences (P20GM103432). Research reported in this publication was supported in part by the Institutional Development Award (IDeA) from the National Institute of General Medical Sciences of the National Institutes of Health under grant number P20GM121310.
The authors declare no competing financial interests.
- Correspondence should be addressed to Kara G. Pratt at kpratt4{at}uwyo.edu.