Presenilin regulates retinotectal synapse formation through EphB2 receptor processing

Abstract

As the catalytic component of γ-secretase, presenilin (PS) has long been studied in the context of Alzheimer’s disease through cleaving the amyloid precursor protein. PS/γ-secretase, however, also cleaves a multitude of single-pass transmembrane proteins that are important during development, including Notch, the netrin receptor DCC, cadherins, drebrin-A, and the EphB2 receptor. Because transgenic PS-KO mice do not survive to birth, studies of this molecule during later embryonic or early postnatal stages of development have been carried out using cell cultures or conditional knock-out mice, respectively. As a result, the function of PS in synapse formation had not been well-addressed. Here we study the role of PS in the developing Xenopus tadpole retinotectal circuit, an in-vivo model that allows for protein expression to be manipulated specifically during the peak of synapse formation between retinal ganglion cells and tectal neurons. We found that inhibiting PS in the postsynaptic tectal neurons impaired tadpole visual avoidance behavior. Whole cell recordings indicated weaker retinotectal synaptic transmission which was characterized by significant reductions in both NMDA receptor (NMDAR)- and AMPA receptor (AMPAR)-mediated currents. We also found that expression of the C-tail fragment of the EphB2 receptor, which is normally cleaved by PS/γ-secretase and which has been shown to upregulate NMDARs at the synapse, rescued the reduced NMDAR-mediated responses. Our data determine that normal PS function is important for proper formation and strengthening of retinotectal synapses through cleaving the EphB2 receptor.

1. Introduction

Presenilin (PS) is a family of multi-pass transmembrane proteins that are highly conserved across species (Otto et al., 2016). Vertebrates are known to express two forms of this protein, presenilin 1 and 2 in mammals (Tanzi and Bertram, 2001), and α and β in amphibians (Tsujimura et al., 1997). PS has been studied, and named, mainly in the context of Alzheimer’s disease (AD). As the catalytic component of γ-secretase, PS is well known for its participation in the cleavage of the amyloid precursor protein (APP) (Scheuner et al. 1996). Successive cleavages by β and γ-secretase of APP generate several different lengths of amyloid-β peptides (Aβ). Aβ is the component of neuritic plaques, a hallmark for Alzheimer’s disease (Iwatsubo et al. 1994).

In addition to cleaving APP, PS cleaves or interacts with molecules such as Notch (Song et al. 1999; Zhang et al. 2000; De Strooper et al. 1999), N-cadherin (Georgakopoulos et al. 1999), drebrin A (Lee and Aoki, 2012), β-catenin (Zhang et al. 1998), and Eph receptors (Xu et al. 2009) that are important across key phases of neural development including neurogenesis (Shen et al. 1997), differentiation (De la Pompa et al. 1997; Handler et al. 2000; Nye et al. 1994), neuronal migration (Wines-Samuelson et al. 2005, Patten et al. 2003, Imayoshi et al. 2010), and synapse adhesion and formation (Georgakopoulos et al. 1999; Zhang et al. 1998; Takahashi et al., 2006).

In addition, PS has been found to be involved in synaptic plasticity. Conditional PS knockout mice lacking both PS1 and PS2 in the postnatal forebrain display impairments in hippocampal memory and long-term potentiation (LTP). These mutant mice display a significant reduction in NMDA receptor-mediated responses and αCaMKII activity (Saura et al., 2004). Besides LTP, PS is known to be involved in homeostatic synaptic scaling, a compensatory form of synaptic plasticity that maintains action potential firing rate within an optimal range (Turrigiano et al., 1998). Hippocampal neurons derived from Psen1−/− mouse embryos display impaired homeostatic scaling up of excitatory synapses in response to chronic blockade of action potential firing (Pratt et al., 2011).

Because transgenic PS knockout (PS-KO) mice do not survive to birth (Donoviel et al., 1999; Saura et al., 2004), studies to elucidate the role of this molecule during later embryonic or postnatal stages of neural development have been carried out either using a conditional PS-KO mouse in which PS function is turned off after the circuit has formed, or by using cultures derived from E16 (embryonic day 16) mouse embryos. As a result, the functional role of PS specifically in synapse formation has not been addressed. Here, to determine the functional role of PS in synapse formation, we assessed the role of this protein in the developing retinotectal circuit of the Xenopus tadpole.

The retinotectal circuit is the main component of the amphibian visual system and consists of the presynaptic retinal ganglion cells (RGCs) in the eye, which project their axons directly to the contralateral optic tectum where they form synaptic connections with tectal neurons. The developmental progression of this circuit is stereotypical and has been described in detail (Wu et al., 1996; Pratt and Aizenman, 2007; Liu et al., 2016b), enabling PS function to be experimentally altered specifically during the peak of synaptogenesis (developmental stages 44–46, approximately 7–10 days post fertilization). Once the tadpoles have reached stage (st) 48/49 (approximately 2 weeks post fertilization), the stage when the circuit is processing visual stimuli, retinotectal circuit function is measured at the behavioral, circuit, and single cell level.

Using this approach we found that knocking down PS expression in the postsynaptic tectal neurons disrupts the visual avoidance behavior and weakens synaptic transmission between the RGCs and the tectal neurons. Significant decreases in both AMPA receptor (AMPAR)- and NMDA receptor (NMDAR)- mediated currents were observed. The reduction in NMDAR-mediated currents was rescued by the expression of the C-terminal fragment of the EphB2 receptor which is normally cleaved by PS/γ-secretase, indicating insufficient PS-dependent processing of EphB2 receptors is the likely mechanism underlying the compromised synaptic transmission. Together, our data suggest that normal PS function is important for proper formation and stabilization of retinotectal synapses through cleaving the EphB2 receptor.

2. Materials and Methods

All experiments have been approved by the University of Wyoming’s Institutional Animal Care and Use Committee. Xenopus laevis tadpoles were reared in 10% Steinberg’s solution at 20~23 °C with a 12:12-h light-dark cycle. Tadpole developmental stages were characterized according to Nieuwkoop and Faber (1994). To block γ-secretase chronically, tadpoles were reared in Steinberg’s solution containing 5μM L685,458 from stage 42 to 49. The drug and rearing solution were renewed every 3rd day.

Western blot

Twenty tecta from stage 45–48 Xenopus tadpoles were directly dissected into RIPA buffer (50 mM Tris, 150 mM NaCl, 0.1% SDS, 0.5% sodium deoxycholate, 1% tritionX-100, 1 mM PMSF and 1× protease inhibitor-Thermo Scientific, Waltham, MA). Then the tissues were homogenized using dounce homogenizer and sonicated (Qsonica, Newtown, CT, USA) at amplitude of 2 having 4s pulse followed by 20 pause on ice for 10 min. Protein concentration in the lysates was measured by BCA. Protein samples were then mixed with DTT containing Laemmli’s buffer and incubated at 99 °C for 10 min and were separated via Polyacrylamide gel electrophoresis (PAGE) using 4–15 gradient gel. After the separation, protein samples were transferred to PVDF membrane and incubated overnight in blocking buffer (5% nonfat dry milk in TBS with 0.25% Tween 20) with rabbit anti-PS antibody at 4°C. Rabbit anti-PS (LS-C26137) was used at 1μg/ml concentration. After the incubation membrane was rinsed three times in TBTS for 10 min each. The membrane was then incubated with Abcam Sheep anti- rabbit HRP (ab6795) secondary antibody at a dilution of 1:40,000 for 1 hr at room temperature. Following incubation with secondary antibody, the membrane was rinsed three times with TBST for 10 min each and the protein bands were visualized using freshly prepared Super-Signal West Femto Kit (Pierce). Gel Doc XRS digital imaging system (Bio-Rad) was used to capture images. Once the images were captured, Restore PLUS western blot stripping buffer (Thermo Scientific) was used to strip the membrane. Complete stripping was confirmed by incubating the membrane with secondary antibody as described earlier. After confirming complete stripping, the same membrane was used to validate the PS antibody via preabsorption with PS blocking peptide 9LS-E39746/213. Briefly, PS antibody (1ug/ml) was incubated with PS blocking peptide (25 μg/ml) in blocking buffer for 4 hr at room temperature. After the incubation it was used to detect the proteins on membrane following the same protocol described above and images were captured using Gel Doc XRS.

Brain Preparations

Whole cell recordings were either performed by whole brain preparation as described by Wu et al. (1996) or in vivo. For whole brain preparation, briefly, tadpoles were anesthetized with 0.01% tricaine methane sulphonate (MS-222) and transferred to a recording dish that contains HEPES-buffered extracellular saline (in mM: 115 NaCl, 2 KCl, 3 CaCl2, 3 MgCl2, 5 HEPES, and 10 glucose, PH 7.25; osmolarity 255 mosM). The entire brain and brain stem were filleted along the dorsal midline and dissected out, then pinned to a submerged block of Sylgard. For in vivo preparation, following anesthesia, tadpoles were transferred to HEPES-buffered extracellular saline contain 0.1mM tubocurarine and stabilized by pins. Then the skin covering the brain was removed and the brain was fileted along the dorsal midline. For both preparations, the ventricular membrane covering the tectum was removed by using a broken glass pipette to access tectal cells (Liu et al., 2018).

Whole cell electrophysiology

Neurons were visualized using a Zeiss light microscope with a ×60 water-immersion objective that is connected to a Hamamatsu CCD camera. Morpholino-containing neurons were visualized using a fluorescence RFP filter and ZsGreen transfected neurons were visualized using a fluorescence GFP filter. Whole-cell recordings were recorded with glass micropipettes (7–12 MΩ) containing potassium gluconate internal saline (in mM: 100 K-gluconate, 8 KCl, 5 NaCl, 1.5 MgCl2, 20 HEPES, 10 EGTA, 2 ATP, and 0.3 GTP, pH 7.2; osmolarity, 255 mOsm). For light-evoked responses, one end of an optic fiber (300μm diameter) was aligned with the tadpole’s right eye and the other side was connected with an LED. The recordings were performed on the contralateral tectum. The light intensity measured from the tip of the optic fiber was 300nW. Light evoked currents of tectal neurons were recorded in a completely dark environment. Light was projected onto the retina via an optic fiber. The protocol consisted of first recording 1 second of baseline activity in darkness followed by the light turning on, remaining on for 15 seconds, then turning off.

For RGC-evoked recordings, a bipolar stimulating electrode was placed on the optic chiasm connected to an ISO-Flex stimulator (A.M.P.I.) to stimulate retinal ganglion cell (RGC) axons. The total amount of RGC input onto a given neuron was quantified by recording maximum RGC-evoked synaptic currents with the neurons voltage clamped at −40 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 minimal RGC-evoked responses, the strength of the stimulation electrode was adjusted such that failures were observed for approximately 50% of the trials. AMPAR- and NMDAR-mediated responses were measured by recording evoked responses with the neuron held at −60 mV to record AMPAR-mediated responses, and at +55 mV to allow recording of NMDAR-mediated responses. Inhibitory (GABAergic) transmission was blocked by adding picrotoxin (100μM) to the extracellular saline. Minimal RGC-evoked responses in the presence of Sr²⁺ were carried out as described by Pratt et al. (2008). The Ca²⁺ concentration in the extracellular saline was adjusted to 0.1nM and 3mM Sr²⁺ was added.

For paired pulse recordings, RGC axons were stimulated with a intermediate stimulation strength with a inter-stimulus interval of 50 ms. GABAergic transmission was inhibited by including picrotoxin (100 μM) in the bath. The amount of facilitation was measured as EPSC2/EPSC1. For coefficient of variation (CV) analysis, the CV of EPSC amplitudes was calculated by dividing the standard deviation by the mean EPSC amplitude and multiplying by 100 to express as a percentage.

Miniature excitatory postsynaptic currents (mEPSCs) were recorded in voltage-clamp at −60 mV in the in the presence of TTX (1 μM) to block action potential firing and picrotoxin (100μM) to block Inhibitory (GABAergic) events. Miniature inhibitory postsynaptic currents (mIPSCs) were recorded in voltage-clamp at −80mV in the presence of TTX (1 μM), NBQX (20 μM) and APV(50 μM) to block AMPAR- and NMDAR-mediated synaptic transmission.

All recordings were restricted to the middle third of the tectum to reduce potential variability. Signals were measured with Axon instruments MultiClamp 700B microelectrode Amplifier (Molecular Devices, Sunyvale, CA), and digitized using Digidata 1322A digitizer. PClamp software was used for recording and the traces were analyzed off-line with AxoGraph. Leak current was subtracted by pCLAMP software in real time. Cells with >50MΩ access (series) resistances were not included in the dataset. Statistical significance was calculated using unpaired two-tailed t-test for datasets that fit a Gaussian distribution and that had equal variance between groups. For datasets that did not meet these criteria, a nonparametric Mann-Whitney test was performed. Error bars represent SEM.

Visual avoidance behavior assay

The visual avoidance behavior assay used in this study was modified according to previous reports (Dong et al., 2009; Shen et al., 2014). Four tadpoles were put in a clear bottom tank with 1 cm high Steinberg’s rearing solution. Computer generated moving dots (0.4cm diameter) were projected to the tank bottom by a monitor (HP ZR22W). The behavior of each tadpole was assessed during its first 5 encounters with the moving dot. A dodging behavior was defined as the tadpole changing its trajectory and accelerating away. A failure was defined as the tadpole not changing its swimming direction or speed when encountering the dot. For each tadpole, a score was generated for the fraction of successful dodges achieved. For example, a score of 1.0 indicates that a particular tadpole successfully dodged an approaching dot 5/5 times.

Morpholino knockdown of PS in vivo

To knock down PS-α expression, we electroporated a 3’ lissamine-tagged, translation-blocking antisense morpholino oligonucleotide (“PS-MO”) into the middle ventricle. This same morpholino has been used previously by another group to knock down presenilin in Xenopus tadpole embryos (Paganelli et al., 2001). The electroporation process was modified from Haas et al., 2002. Briefly, tadpoles were anesthetized with 0.01% MS-222, and presenilin morpholino (PS-MO; GTCAGCGGAATCTTCAGACACTTGG; 0.1mM) or standard control morpholino (Ctrl-MO; CCTCTTACCTCAGTTACAATTTATA; 0.1mM) from Gene Tools (Philomath, OR) was injected into the middle ventricle of the optic tectum of stage 44/45 tadpoles by a glass micropipette. Next, two platinum electrodes are placed on the tadpole skin that flank the tectum and 3 × 60ms unipolar current pulses were delivered. The polarity of the current pulses was then reversed and the process repeated. After electroporation, tadpoles were transferred into a bowl that contains normal 10% Steinberg’s solution to recover. Neurons containing the lissamine-tagged morpholinos were identified through a RFP filter.

Ectopic expression of EphB2/CTF2 in vivo

DNA encoding the EphB2/CTF2 fragment was PCR-amplified from the pMX-EphB2/CTF2 construct (Litterst et al, 2007, a generous gift from Dr. Anastasios Georgakopoulos) using the primers 5’-tcgtgacgctagcgccgccaccatgattgccatcgtatgtaacagac-3’ and 5’-tcgtgacggatcctcaaacctctacagactggatctg-3’, which contain the restriction sites for NheI and BamHI, respectively. Following restriction digestion, the amplified EphB2/CTF2 fragment was subcloned into the NheI and BamHI sites within the MCS of the pEF1α-IRES-ZsGreen1 Vector (Clontech, Cat. No. 631976). The resulting construct was verified by sequencing the insert using the primers 5’-tcaagcctcagacagtggttc-3’ and 5’-aagcggcttcggccagtaacgtta-3’. Electroporation was performed as described for morpholinos. Neurons expressing ZS-green protein were identified through a GFP filter.

3. Results

To determine the role of PS in retinotectal synapse formation, we inhibited PS function–using both genetic and pharmacological manipulations – during the time in development when synapse formation was at its peak. Once the circuit had reached a more mature stage, the retinotectal circuit function was assessed at the behavioral, circuit and, single cell levels.

Presenilin is expressed in the optic tectum during retinotectal circuit formation

Although PS mRNA has been reported to be present in the tadpole embryo brain (Tsujimura et al., 1997), whether PS protein is expressed specifically in the optic tectum had not been determined. Therefore, western blot analysis of protein homogenate derived from stage 45–48 optic tecta was carried out using a PS-α antibody, resulting in a single band of approximately 60kDA (Figure 1A). Pre-absorbing the PS antibody with a 25-fold excess of PS blocking peptide greatly diminished the appearance of this band, indicating that the PS antibody was specifically detecting PS protein (Figure 1A). Thus, western blot data indicate PS is expressed in the optic tectum between developmental stages 45 and 48, the period when the retinotectal circuit is forming. This suggests that this protein may play a role in the development of the circuit. To test this possibility, PS function was inhibited in the optic tectum by transferring a PS morpholino (PS-MO) into the tectal neurons. This specific morpholino has been previously used to knock down PS function in early Xenopus embryos (Paganelli et al., 2001). For transferring morpholinos into tectal neurons, the morpholino solution was injected into the large middle ventricle in the brain, and morpholinos were transferred into the tectal neurons via electroporation (Truszkowski et al., 2016; Shen et al., 2014). Electroporation resulted in the uptake of morpholino into a large portion of tectal neurons (Figure 1B). As a control for both the electroporation procedure and the physical presence of a morpholino, batch-matched tadpoles were electroporated with a standard morpholino control (Ctrl-MO; Shen et al., 2014; Truszkowski et al., 2016). Electroporation of morpholinos was carried out at developmental stage 44/45, an early larval stage when the retinotectal circuit is in the midst of forming (van Rheede et al., 2015; Pratt and Aizenman, 2007). Experimental analyses of the effects of PS knockdown were carried out once tadpoles had reached stage 48/49, approximately 5–8 days after the morpholinos had been electroporated. By this time the normal retinotectal circuit has grown more mature and refined and visually guided behaviors are robust.

Figure 1. PS is expressed in the optic tectum during retinotectal circuit formation.

(A) Top, western blot using different amounts of protein from rat brain homogenate and stage 45–48 optic tecta samples (two separate samples). Bottom, preabsorping the PS antibody with a 25-fold excess of PS blocking peptide, greatly diminished the appearance of this band indicating PS antibody specificity. (B) Electroporation results in uptake of morpholino into many neurons of the optic tectum, Left: PS-MO, Right: Ctrl-MO. Red-dashed box indicates the left optic tectum.

Knocking down PS levels in tectal neurons compromises visually-guided avoidance behavior.

Visually guided behavior was quantified using an established visual avoidance test (Dong et al., 2009; Khakhalin et al., 2014; Shen et al. 2011; Shen et al., 2014; Truszkowski et al., 2016.) which is based on the finding that stage 47–49 tadpoles often display a robust escape behavior in response to a close encounter with a moving dot projected onto the floor of their glass container (Dong et al., 2009). For each tadpole, a total of 5 encounters with a dot were scored as either a “successful dodge” or “failure to dodge”. Figure 2A shows examples of a successful dodge and failure to dodge. A score was generated for each tadpole based on the success/failure for each of the 5 encounters and was expressed as a probability of a successful dodge. We found that the tadpoles that had been electroporated with the PS-MO performed significantly worse (i.e. displayed fewer successful dodges) than Ctrl-MO tadpoles. (Figure 2B; visual avoidance probability for Control: 0.65 ± 0.08, n =15; Ctrl-MO: 0.66 ± 0.06, n = 25 tadpoles; for PS-MO: 0.33 ± 0.05, n = 30; p = <<0.001, unpaired two-tailed t-test for comparing Ctrl-MO and PS-MO.) As the catalytic component of γ-secretase (Kimberly et al., 2003), a promoter of the Akt-PI3 signaling pathway (Baki et al., 2004), and a regulator of intracellular calcium stores (Tu et al., 2006; Rybalchenko et al., 2008), PS plays roles in a wide range of important cellular functions. To determine if this defect in visual avoidance behavior caused by the PS-MO is due specifically to deficits in secretase function, normal tadpoles were chronically exposed to the specific γ-secretase inhibitor L685,458 (5μM) during retinotectal circuit formation (stages 42 to 48/49). L685,458 is commonly used to distinguish between presenilin’s role as γ-secretase and other non-secretase functions (Pratt et al., 2011; Bai et al., 2011). To avoid possible acute effects, L685,458 was not present during testing. We found that, similar to inhibiting PS levels using the morpholino, blocking γ-secretase function by exposing tadpoles to L685,458 also reduced visual avoidance probability. (Figure 2C; visual avoidance probability for control tadpoles: 0.58 ± 0.09, n = 11; for L685-exposed tadpoles: 0.31 ± 0.04, n = 13; p = 0.0072, unpaired two-tailed t-test.) Thus, inhibiting specifically PS-dependent γ-secretase activity compromised visual avoidance behavior to a similar degree as inhibiting PS protein expression via the PS-MO (Figure 2B,C). These results suggest that the deficit in visual avoidance behavior as a result of knocking down PS protein levels in the tectum is due to decreased PS/γ-secretase function. Next, we characterized the underlying pathology by measuring light-evoked responses of control neurons and neurons containing the PS-MO.

Figure 2. Knocking down PS expression in tectum disrupts visual avoidance behavior.

(A) Examples of tadpole visual avoidance behavior in response to moving visual stimuli. Top, example of a successful dodge displayed by a tadpole. Bottom, example of a failure to dodge displayed by a tadpole. Black arrow indicates the direction of movement of the dots. Red arrow indicates the swimming direction of the tadpole. Right panel shows swimming trajectories of the tadpole. Bar graphs showing the visual avoidance probability (successful dodge rate) for (B) Ctrl, Ctrl-MO and PS-MO (Ctrl-MO versus PS-MO: *** p < 0.001,); (C) Ctrl and L685 (** p < 0.01).

Inhibiting PS function in tectal neurons impairs light-evoked responses

Compromised visual avoidance behavior suggests faulty synaptic transmission between RGCs and tectal neurons. To test this possibility, in vivo light-evoked responses from Ctrl-MO and PS-MO tectal neurons were recorded. For this, light was projected onto the retina via an optic fiber and the responses were recorded from individual tectal neurons in voltage clamp mode. Tectal neurons at these developmental stages tend to respond to changes in luminance, responding to both light turning on (ON response) and light turning off (OFF response). Therefore, we used a protocol that tested both ON and OFF responses. The protocol consisted of first recording 1 second of baseline activity in darkness followed by the light turning on, remaining on for 15 seconds, then turning off (Figure 3A). A typical response exhibited by a control neuron is shown in Figure 3A. This is a compound response consisting of a monosynaptic component which is due to direct synaptic activation of the tectal neuron by RGC axons, and the polysynaptic portion, due to RGC-evoked activation of the microcircuitry within the tectum. Figure 3B and C show examples of ON and OFF responses recorded from a Ctrl-MO and PS-MO neuron respectively. We found that, compared to Ctrl-MO neurons, PS-MO neurons displayed dramatically reduced ON (Figure 3D; evoked compound synaptic currents (eCSCs) charge: measured from a 1.5s window, 0.5s after light turns on, for Ctrl-MO: 12.42±1.66 pA*S, n=15; for PS-MO: 5.64 ± 0.85 pA*S, n = 16; p = 0.0012, Mann-Whitney test) and OFF synaptic responses (Figure 3E; eCSCs charge for Ctrl-MO: 11.83±1.92 pA*S, n=15; for PS-MO: 5.07 ± 0.87 pA*S, n = 16; p = 0.0011, Mann-Whitney test). To determine whether this effect, like the effect on visual avoidance behavior, was also due to inhibiting specifically the role of PS as the catalytic component of γ-secretase, light-evoked responses were recorded from L685,458-exposed neurons. A significant decrease in light-evoked responses was again observed (Figure 3F, G; On response: eCSCs charge for Ctrl: 15.84 ± 4.29 pA*S, n=9; for L685,458: 5.41 ± 1.14 pA, n = 13; p = 0.012, unpaired two-tailed t-test. Off response: eCSCs charge for Ctrl: 19.89 ± 4.84 pA, n=9; for L685,458: 7.03 ± 2.60 pA, n = 13; p = 0.017, Mann-Whitney test). The magnitude of inhibition observed in L685,458 neurons was similar to that of PS-MO neurons, suggesting a pathology associated with impaired secretase activity. The significantly weaker light-evoked responses observed in both PS-MO and L685,458-exposed neurons likely accounts for the compromised visual avoidance behavior.

Figure 3. Knocking down PS expression in tectal neurons impairs light-evoked responses.

(A) Left, diagram showing the in vivo recording configuration. Right, an example of a light-evoked response. The protocol consisted of first recording 1 second of baseline activity in darkness followed by the light turning on, remaining on for 15 seconds, then turning off. (B) Sample traces of ON response from Ctrl-MO and PS-MO neurons. (C) Sample traces of OFF response of Ctrl-MO and PS-MO neurons. (D) Plots showing the ON response - amount of charge (pA*S; a 1.5s window after light turns on) of the evoked compound synaptic currents of Ctrl-MO and PS-MO neurons (** p < 0.01). For all dot plots, each dot represents a single neuron. (E) Plots showing the OFF response - amount of charge (1.5s window after light turns off) of the evoked compound synaptic currents of Ctrl-MO and PS-MO neurons (** p < 0.01). (F) Plots showing the ON response - amount of charge (1.5s window after light turns on) of the evoked compound synaptic currents from Ctrl and L685-exposed neurons (* p < 0.05). (G) Plots showing the OFF response - amount of charge (1.5s window after light turns off) of the evoked compound synaptic currents of Ctrl and L685-exposed neurons (* p < 0.05).

Neither method of inhibiting PS affected passive electrical properties or voltage-dependent intrinsic currents

Changes in passive or active electrical properties expressed by tectal neurons could cause the observed compromises in light-evoked responses. For example, a decrease in membrane resistance would manifest as dampened synaptic responses. Several passive and active properties expressed by PS-MO, Ctrl-MO (Table 1), control and L685,458-exposed (Table 2) neurons were quantified. None of the properties differed between groups (unpaired two-tailed t-test). The deficits in light-evoked responses associated with PS-MO neurons, therefore, cannot be explained by alterations in passive or active electrical properties.

Table 1.

Active and passive electrical properties of Ctrl-MO and PS-MO-transfected neurons.

	Ctrl-MO	n	PS-MO	n	p
Resting membrane potential, mV	−38.8±1.38	53	−38.4±0.85	60	0.82
Input resistance, GΩ	1.24 ±0.17	53	1.51±0.13	60	0.13
Membrane capacitance, pF	11.81±0.36	53	11.21±0.31	60	0.21
Maximum Na+, pA	376.2±60.4	17	331.4±38.3	22	0.51
Maximum K+, pA	626.9±58.2	17	487.8±73.0	22	0.16

Values are means ±SE; n, no. of neurons. No properties were significantly altered.

Table 2.

Active and passive electrical properties of Control and L685-rear neurons.

	Control	n	L685	n	p
Resting membrane potential, mV	−37.5±2.81	9	−36.4±2.40	13	0.62
Input resistance, GΩ	1.24 ±0.17	9	1.74 ±0.21	13	0.10
Membrane capacitance, pF	10.39±0.73	9	10.53 ±0.82	13	0.90
Maximum Na+, pA	399.7±55.5	9	327.8±45.5	13	0.33
Maximum K+, pA	526.5±50.0	9	445.2±67.9	13	0.39

Values are means ±SE; n, no. of neurons.

Inhibiting PS levels in postsynaptic tectal neurons decreases the strength of synaptic transmission between presynaptic RGC axons and postsynaptic tectal neurons

To further define the pathology in synaptic transmission between RGC axons and PS-MO tectal neurons, we switched to a whole brain preparation. With this preparation a bipolar stimulation electrode can be placed directly onto the optic chiasm, allowing for the RGC axons to be directly activated in a controlled manner (Figure 4A; Wu et al., 1996; Aizenman et al., 2003; Pratt and Aizenman, 2007; Liu et al., 2016a). The total amount of RGC input onto a given neuron was quantified by recording maximum RGC-evoked synaptic currents with the neurons voltage clamped at −40 mV. The maximum synaptic current is defined as the peak current amplitude that no longer increases in response to increasing the stimulation strength of the RGC axons (Truszkowski et al., 2016; Liu et al., 2016a,). Example traces of maximum responses from Ctrl-MO and PS-MO neurons are shown in Figure 4B. Similar to the light-evoked responses, the average maximum monosynaptic response amplitude displayed by PS-MO neurons was significantly reduced compared to Ctrl-MO neurons (Figure 4C; Peak synaptic amplitude for Ctrl-MO: 88.36 ± 10.74 pA, n=15; for PS-MO: 38.12 ± 5.22 pA, n = 14; p = 0.0003, unpaired two-tailed t-test). In addition, we also measured the strength of individual RGC axons by recording minimal RGC-evoked responses. For this, the strength of the stimulation electrode was adjusted such that failures were observed for approximately 50% of the trials. In this configuration, any non-failure event is likely due to activation of a single RGC axon synapsing onto the neuron being recorded (Wu et al., 1996; Pratt and Aizenman, 2007). This protocol allows for both the AMPAR- and NMDAR-mediated responses to be readily measured by recording evoked responses with the neuron held at −60 mV to record AMPAR-mediated responses, and at +55 mV to record NMDAR-mediated responses (Wu et al., 1996). Examples of AMPAR- and NMDAR-mediated responses from a Ctrl-MO neuron and PS-MO neuron are shown in Figure 4D. Compared to the control, we observed a significant reduction in both AMPAR- and NMDAR-mediated minimal-stimulation response amplitudes exhibited by PS-MO neurons (Figure 4E and F; AMPA current for Ctrl-MO: 9.98 ± 0.59 pA, n=20; for PS-MO: 6.93 ± 0.42 pA, n = 16; p = 0.003, unpaired two-tailed t-test. NMDA current: measured from a 10 ms window, 20 ms after peak AMPA currents, for Ctrl-MO: 10.57 ± 1.47 pA, n=22; for PS-MO: 6.25 ± 0.85 pA, n=15; p= 0.0057, Mann-Whitney test). The AMPAR- and NMDAR-mediated responses decreased such that a normal AMPA:NMDA ratio – a measure of synapse maturity – was maintained (Figure 4G and H; AMPA:NMDA ratio for Ctrl-MO: 1.12 ± 0.15, n=20; for PS-MO: 1.28 ± 0.51, n=14; p=0.48, unpaired two-tailed t-test). Overall, these data suggest that inhibiting PS function in a postsynaptic tectal neuron significantly weakens the synaptic transmission at the single axon level, while the AMPA:NMDA ratio is that of a mature synapse, suggesting that even though overall synaptic transmission is weaker in PS-MO neurons, the functional relationship between NMDARs and AMPARs is not disrupted.

Figure 4. Inhibiting PS levels in postsynaptic tectal neurons decreases the strength of connectivity between presynaptic RGC axons and postsynaptic tectal neuron dendrites.

(A) Diagram showing the preparation and recording configuration. Grey shaded area represents the brain preparation. The bipolar stimulating electrode was placed on the optic chiasm where the two axons cross (indicated as green and red line). (B) Three example traces for maximum RGC-evoked synaptic currents from Ctrl-MO (blue) and PS-MO (magenta) neurons. (C) Plots showing amplitude of the maximum RGC-evoked synaptic currents of Ctrl-MO and PS-MO neurons (*** p < 0.001). (D) Example traces of the minimal RGC-evoked responses from an individual Ctrl-MO (blue) and PS-MO (magenta) neuron. The upward trace represents NMDAR-mediated response and downward trace represents AMPAR-mediated response. Dashed line represents the analysis window for NMDAR-mediated response. (E) Plots showing amplitude of the minimal RGC-evoked AMPAR-mediated response of Ctrl-MO and PS-MO neurons, (** p < 0.001). (F) Plots showing amplitude of the minimal RGC-evoked NMDAR-mediated response of Ctrl-MO and PS-MO neurons, (** p < 0.001). (G) AMPAR- (y-axis) and NMDAR- (x-axis) mediated response amplitudes within individual cells, blue: Ctrl-MO neurons; magenta: PS-MO neurons. (H) Plots showing AMPA: NMDA of Ctrl-MO and PS-MO neurons.

Probability of transmitter release from RGC axons onto PS-MO neurons is normal

The weaker synaptic transmission between RGC and tectal neurons could be due to a decrease in probability of transmitter release from the presynaptic RGC axons (p), a decrease in the number of synapses (n), a decrease in the strength of the individual retinotectal synapses (q), or changes in a combination of these factors (Korn and Faber, 1991). Therefore, to identify the locus of the diminished synaptic transmission, we carried out a set of established recording protocols designed to assess the individual factors (p, n, and q). To assess probability of presynaptic transmitter release, paired pulse recordings were carried out. For this, RGC axon stimulation strength was adjusted to induce a reliable response of intermediate amplitude. Paired stimulations, separated by 50 ms, were delivered to the axons and the resulting synaptic currents were recorded in postsynaptic tectal neurons. To minimize recurrent polysynaptic portion of the response, GABAergic transmission was inhibited by including picrotoxin (100 μM) in the bath. Sample recordings from PS-MO and Ctrl-MO tectal neurons are shown in Figure 5A. Paired pulse ratios were generated from this data by dividing the amplitude of the second response by the amplitude of the first response. Average paired pulse ratios for PS-MO neurons were found to be similar to Ctrl-MO tectal neurons (Figure 5B,C: Paired pulse ratios for Ctrl-MO: 1.51 ± 0.14, n=17; for PS-MO: 1.82 ± 0.13, n=22, p=0.11, unpaired two-tailed t-test). To test this further we carried out a coefficient of variability (CV) analysis, comparing the degree of variability between the first and second EPSCs (Waldeck et al., 2000) of Ctrl-MO and PS-MO neurons. The higher the probability of transmitter release (i.e. the greater the quantal content) the lower the CV, and visa versa (Lin and Faber, 1988). Therefore, if inhibiting PS in the postsynaptic tectal neurons decreases the probability of presynaptic transmitter release, then we would expect the amplitudes of first EPSC (EPSC1) to be more variable compared to control, and correspondingly, the second EPSC (EPSC2) to be less variable. These two effects, combined, would create a greater difference in CV values between EPSC1 and EPSC2 in the PS-MO neurons. Conversely, if inhibiting PS in the postsynaptic neuron does not alter the probability of transmitter release, then the difference in the CV between EPSC1 and EPSC2 amplitudes of PS-MO neurons should be similar to that of Ctrl-MO neurons. Our results indicate the latter: the differences in CVs between EPSC1 and EPSC2 of PS-MO (6.6% ± 4.2%, n = 10) and Ctrl-MO neurons (3.1% ± 2.4%, n = 10) were not significantly different (p = 0.47). Additional CV analysis data is shown in supplementary figure 1. The CV analysis indicates no change in probability of release onto PS-MO neurons. Thus, the compromised transmission appears not to be due to altered probability of transmitter release (p) from presynaptic RGC axons.

Figure 5. Paired pulse facilitation is normal in PS-MO neurons.

(A) Example traces of paired pulse facilitation recordings collected from Ctrl-MO and PS-MO neurons, a medium stimulation strength was used. (B) Graph showing EPSC1 amplitude (y-axis) and EPSC2 amplitude (x-axis) from an individual Ctrl-MO (Blue) or PS-MO (magenta) neurons. Data points above the dash lines indicate paired pulse depression and data points below the dashed line indicate facilitation. (C) Plots showing paired pulse ratio (EPSC2: EPSC1) of Ctrl-MO and PS-MO neurons.

PS-MO neurons display a significant decrease in strength and number of individual retinotectal synapses

The strength of individual retinotectal synapses was quantified by recording minimum RGC evoked responses in an external recording solution in which Ca²⁺ was replaced with Sr²⁺, which desynchronizes synaptic vesicle release and allows examination of individual quantal events (Xu-Friedman and Regehr 2000; Aizenman and Cline, 2007; Pratt et al., 2008; Wen and Barth, 2011). For this, RGC axons were activated using the minimal stimulation protocol in the presence of 3mM Sr²⁺ (Figure 6A). We found that both amplitude and frequency (number of events within the analysis time window) of the asynchronously evoked EPSCs (asEPSCs) were decreased in the PS-MO neurons (Figure 6B,C; asEPSCs amplitude for Ctrl-MO: 7.91 ± 0.35, n=29; for PS-MO: 6.83 ± 0.31, n=32; p=0.0167, Mann-Whitney test; asEPSCs frequency for Ctrl-MO: 1.62 ± 0.11, n=15; for PS-MO: 1.16 ± 0.11, n=18; p=0.0072, unpaired two-tailed t-test). These data suggest that inhibiting PS in a given postsynaptic tectal neuron decreases both the strength and number of retinotectal synapses received by that neuron. Besides receiving direct input from RGCs, tectal neurons receive strong local inhibitory and excitatory input from other tectal neurons, (Miraucourt et al., 2012; Pratt et al., 2008; Gambril et al., 2016). To determine whether inhibiting PS function alters the formation of local inhibitory synapses onto PS-MO neurons, individual inhibitory synapses were quantified by recording miniature inhibitory postsynaptic currents (mIPSCs) from Ctrl-MO and PS-MO neurons (Figure 6D). Neither the amplitude nor the frequency of PS-MO mIPSCs were found to be different from controls. (Figure 6E,F, mIPSCs amplitude for Ctrl-MO: 5.4 ± 0.23, n=17; for PS-MO: 5.8 ± 0.26, n=17; p=0.2, unpaired two-tailed t-test; mIPSCs frequency for Ctrl-MO: 0.96 ± 0.14, n=17; for PS-MO: 0.93 ± 0.14, n=17; p=0.88, unpaired two-tailed t-test). This result indicates that inhibiting PS function does not appear to affect the formation of inhibitory synapses between tectal neurons. To assess whether the strength of all the excitatory synapses, not only retinotectal synapses, onto a given PS-MO neuron were decreased, miniature excitatory postsynaptic currents (mEPSCs) were recorded. To isolate mEPSCs, neurons were recorded in the presence of the GABA receptor inhibitor picrotoxin (100 μM). Similar to what was found for the retinotectal synapses (i.e. the asEPSCs), the mEPSCs data also show decreased frequency and amplitude of synaptic events, however only the decrease in frequency was statistically significant (Supplementary figure 2). That the decrease in average mEPSC amplitude (which reflects the strength of all the different sources of excitatory synapses) recorded from PS-MO neurons was less pronounced than the decrease observed in average asEPSC amplitude (which reflects the strength of purely retinotectal synapses) is potentially interesting because it could be an indication that PS may be more important for the formation of retinotectal synapses and may not be involved in, for example, the formation of local excitatory tectal-tectal synapses. Further implications and limits of interpreting the mEPSC data appear in the Discussion section.

Figure 6. PS-MO neurons display a significant decrease in amplitude and frequency of asynchronously evoked EPSCs (asEPSCs) events.

(A) Sample traces showing asEPSCs of a Ctrl-MO and a PS-MO neuron with the dashed lines showing the analysis window (100ms). (B) Plots showing the amplitude of asEPSCs of Ctrl-MO and PS-MO neurons (* P < 0.05). (C) Plots showing the number of events (within the analysis window) of asEPSCs of Ctrl-MO and PS-MO neurons (** p < 0.01). (D) Example traces of mIPSCs of Ctrl-MO and PS-MO neurons. (E) Plots showing the amplitude of mIPSCs of Ctrl-MO and PS-MO neurons. (F) Plots showing the frequency of mIPSCs of Ctrl-MO and PS-MO neurons.

EphB2/CTF2 rescues the reduced NMDAR mediated currents in PS-MO neurons.

One of the substrates of presenilin/γ-secretase is the EphB2 receptor. It has been reported that the EphB2 receptor, through direct interactions with NMDARs, recruits NMDARs and thereby regulates excitatory synapse formation (Dalva et al. 2000). Furthermore, studies show that the EphB2/CTF2 peptide, generated by γ-secretase (Figure 7A), directly phosphorylates and promotes cell surface localization of NMDARs through its tyrosine kinase activity (Litterst et al., 2007; Xu et al., 2009). Finally, it is well established that the RGC axons express ephrin and the tectal neurons express EphB2 (Mann et al., 2002; Feldheim and O’Leary, 2010; Lim et al., 2010). Therefore, the decreased NMDAR-mediated currents we observed in the PS-MO neurons could be due to impaired PS-dependent processing of the EphB2 receptor. If this is true, then an already-cleaved EphB2/CTF2 should rescue the reduced NMDAR mediated currents in PS-MO neurons. To test this possibility, we generated a construct to allow the simultaneous ectopic expression of EphB2/CTF2 and ZsGreen in tectal neurons following electroporation in vivo (Figure 7B). Using the ZsGreen fluorescence to identify successfully electroporated neurons (Figure 7C), we used minimum stimulation (same method as shown in Figure 4) to quantify AMPAR- and NMDAR-mediated currents. We found that expressing EphB2/CTF2 in the PS-MO neurons significantly increased the NMDAR-mediated currents, rescuing the currents to a similar level as controls (Figure 7D, NMDAR-mediated currents for Ctrl-MO+ZsGreen: 9.53 ± 1.26, n=11; for Ctrl-MO+CTF2: 9.62±1.08, n=9; for PS-MO+ZsGreen: 6.01 ± 0.58, n=12; for PS-MO+CTF2: 8.21 ± 0.78, n=10. p=0.0323, one-way ANOVA test; p = 0.030, Mann-Whitney test for comparing PS-MO+CTF2 with PS-MO+ZsGreen). It is important to note that the expression of EphB2/CTF2 in Ctrl-MO neurons did not affect the NMDAR-mediated current (p= 0.95, unpaired two-tailed t-test), indicating a specific rescuing of NMDAR-mediated responses in the PS-MO neurons. For AMPAR-mediated responses, a similar trend of rescuing is observed following expression of EphB2/CTF2 in PS-MO neurons, however it is not statistically different from PS-MO neurons. (Figure 7E, AMPAR-mediated currents for Ctrl-MO+ZsGreen: 9.91 ± 0.93, n=14; for Ctrl-MO+CTF2: 9.63 ± 0.70, n=9; for PS-MO+ZsGreen: 6.99 ± 0.58, n=13; for PS-MO+CTF2: 9.06 ± 1.14, n=10. p=0.0515, one-way ANOVA test; p=0.101, unpaired two-tailed t-test for comparing PS-MO+CTF2 with PS-MO+ZsGreen). By showing that expressing EphB2/CTF2 rescues the reduced NMDAR-mediated currents in PS-MO neurons, our results indicate that the compromised synaptic transmission between RGC and PS-MO neurons is likely due to a lack of PS-dependent processing of EphB2 receptors and a concomitant reduction of functional NMDARs.

Figure 7. EphB2/CTF2 rescues the NMDA receptor mediated currents in PS-MO neurons.

(A) Schematic of the EphB2 receptor showing the γ-secretase cleavage site and the CTF2 fragment. (B) Schematic of the EphB2/CTF2 expression construct. The bicistronic ZsGreen1 gene enables the identification, via fluorescence microcopy, of neurons that express the construct. (C) Images of PS-MO+CTF2 neuron being recorded. Left, GFP filter (expression of ZsGreen); Middle, RFP Red filter (presence of MO); Right, merged. Arrow indicates recoding pipette. (D) Plots showing the amplitude of NMDA receptor mediated currents of Ctrl-MO+ZsGreen, Ctrl-MO+CTF2, PS-MO+ZsGreen and PS-MO+CTF2 neurons (* Mann Whitney test comparing PS-MO+CTF2 with PS-MO+ZsGreen, P < 0.05). (B) Plots showing the amplitude of AMPA receptor mediated currents of Ctrl-MO+ZsGreen, Ctrl-MO+CTF2, PS-MO+ZsGreen and PS-MO+CTF2 neurons.

4. Discussion

The main findings reported here are: (1) knocking down PS expression in the postsynaptic tectal neurons when retinotectal synaptogenesis is at its peak impaired visual avoidance behavior and weakened synaptic drive between RGC axons and tectal neurons. (2) Both AMPAR- and NMDAR-mediated synaptic currents were markedly reduced in PS-MO tectal neurons, while inhibitory synapses appeared normal. (3) Expression of the EphB2/CTF2 rescued the reduced NMDAR-mediated currents, and, to a lesser extent, the AMPAR-mediated currents, in PS-MO neurons, indicating insufficient PS-dependent processing of EphB2 receptors is the likely mechanism underlying this compromised synaptic transmission. By inhibiting PS specifically during the peak of synaptogenesis we have found that this protein is critical for EphB2-dependent excitatory synapse formation and strengthening. These results give rise to new questions about whether PS is important for the maintenance and plasticity of synapses after they have formed. Once formed, does a synapse require perpetual EphB signaling to continue to exist and function? Might the processing of EphB2 by PS be regulated by activity? We are addressing these questions using the same tadpole retinotectal circuit as the model, knocking down PS well after the peak of synaptogenesis when the synapses are relatively mature, then carrying out a similar set of experiments to quantify synapse function and plasticity.

Normal PS/γ-secretase activity in postsynaptic tectal neurons is required for optimal visual avoidance behavior

Our data show that knocking down PS function during the development of the retinotectal circuit, by either transfer of a PS-MO into tectal neurons or globally by using the PS/γ-secretase inhibitor L685,458, impairs visual avoidance behavior and reduces light-evoked synaptic responses in tectal neurons (Figures 2 and 3). The observation that knocking down PS levels in postsynaptic tectal neurons compromised visual avoidance behavior and light-evoked responses to a similar degree as inhibiting PS function globally reveals that the locus of the impaired behavior resides on the postsynaptic side of the synapse. L685,458 inhibits only PS’s role as a secretase suggesting that the observed deficits in visual avoidance behavior and light-evoked tectal responses are due specifically to compromised PS/γ-secretase activity. These conclusions are confirmed by the EphB2/CTF2 rescue experiment.

Weaker and fewer retinotectal synapses recorded from PS-MO tectal neurons.

Minimal stimulation data indicate that PS-MO neurons receive relatively weaker synaptic input from individual RGC axons due to significant reductions is both AMPAR- and NMDAR-mediated currents (Figure 4). To determine if the weakening of individual RGC axons onto a given PS-MO neuron was due to a decrease in the strength of individual synapses or the number of synapses, or both, minimal stimulation responses were recorded in the presence of strontium to desynchronize the synaptic events. The observed decrease in amplitude and frequency of strontium-evoked synaptic events indicates a decrease in the strength and number of individual synapses (Figure 6). We do not distinguish whether the decrease in synapse number is due to fewer synapses being formed or fewer being maintained. Nevertheless, given that weaker synapses are known to be less stable and have a higher probability of being lost than stronger ones (Ruthazer and Cline, 2006), the observed decrease in synapse number would be expected.

In addition to visual input, tectal neurons receive input from several different non-visual sensory modalities including the auditory and lateral line systems (Hiramoto and Cline, 2009; Deeg et al., 2009; Hamodi et al., 2016), and local excitatory and inhibitory input from other tectal neurons. To determine if all excitatory synapses associated with PS-MO neurons were reduced, mEPSCs were recorded, and these data indicate a less pronounced decrease in current amplitude compared to that of the asEPSCs. The interpretation of this result is limited by the fact that mEPSCs reflect a sampling of all the different excitatory synapses received by individual tectal neurons and it is not possible to know which events are retintotectal synapses, which are input from other sensory modalities, and which are local tectal-tectal synapses. Nevertheless, that the average mEPSC amplitude was not significantly reduced while the asEPSC was, gives rise to the compelling notion of input-specific mechanisms for synapse formation. This could be how different inputs get targeted to specific regions of dendrite and it would also prevent competition between the different types of inputs (Hamodi et al., 2016). We also observed that, while PS-MO neurons displayed major impairments in the formation/function of excitatory retinotectal synapses, inhibitory GABAergic synapses appeared unaltered, suggesting that they do not involve PS-dependent mechanisms.

Reduced AMPAR-mediated currents in PS-MO neurons most likely stems from reduced NMDAR-mediated mechanisms

Normally, formation of retinotectal synapses, and glutamatergic synapses in general, begins with the recruitment of NMDARs to the postsynaptic site, followed by NMDAR activity-dependent mechanisms that recruit AMPARs (Wu et al., 1996; Barry and Ziff, 2002; Akaneya et al., 2010). Based on our results (the minimal stimulation data; figure 4), the decrease in NMDAR-mediated currents is most likely the direct consequence of knocking down PS expression in these tectal neurons during the time when retinotectal synapse formation is at a peak. The decrease in functional NMDARs, in turn, decreases the (NMDAR-dependent) recruitment of AMPARs – resulting in preservation of the normal AMPA:NMDA ratio, but weaker synapses. It is important to clarify that the AMPA:NMDA ratios are measured at later developmental stages 48/49, when the synapses are expressing a relatively high AMPA:NMDA ratio that is characteristic of mature synapses. The preserved ratio suggests that recruitment of AMPARs to the synapse is actually not impaired at PS-MO synapses, but is scaled down due to lower levels of NMDAR-dependent processes. The preserved ratio also underscores a strong functional connection/correlation between NMDARs and AMPARs at the individual synapse level.

The finding that NMDAR-mediated currents are significantly impaired in PS-MO neurons is important because it means that all of the NMDAR-dependent processes, including refinement (Dong et al., 2009; Hamodi et al., 2016), LTP (Tao et al., 2001), growth and stabilization of RGC axons (Munz et al., 2014; Van Horn et al., 2017) and tectal cell dendrites (Rajan et al., 1999; Sin et al., 2002) could also be at risk.

Cleavage product of EphB2 receptor, peptide EphB2/CTF2 rescues the decreased NMDAR-mediated responses displayed by PS-MO neurons

The Eph receptors belong to a family of receptor tyrosine kinases that are activated by ephrins. There are two classes of Eph receptors, EphA and EphB receptors, which bind to glycosylphosphatidyl-inisotol (GPI)-anchored ephrinAs or transmembrane ephrinBs respectively (Lisabeth et al., 2013). It is well established that gradients of ephrins A and B and their corresponding Eph receptors across retina and the optic tectum instruct the formation of the topographic map formed by RGC axonal inputs (Feldheim and O’Leary, 2010; Mann et al., 2002). Besides the well-studied role of ephrin-Eph signaling in RGC axon pathfinding in the developing optic tectum (Klein et al., 2004; Feldheim and O’Leary, 2010), evidence suggests that EphB receptors through interactions with NMDARs lead to the recruitment of NMDARs and regulate excitatory synapse formation (Dalva et al. 2000). EphB2R knock-out mice display reduced NMDAR mediated synaptic current and decreased number of AMPARs and NMDARs at excitatory synapses (Henderson et al., 2001; Henkemyer et al., 2003). Important to this work, a study by Xu and colleagues (Xu et al. 2009) reveals that peptide EphB2/CTF2 released to the cytosol by γ-secretase processing of EphB2 receptor directly phosphorylates and promotes cell surface localization of NMDARs. Here, we found that expressing the EphB2 intracellular fragment (EphB2/CTF2), produced by the γ-secretase cleavage of EphB2, rescues the reduced synaptic input received by PS-MO neurons (Figure 7). This strongly suggests that the PS-MO phenotype is likely due to impaired EphB2/CTF2–dependent NMDAR upregulation/function. That EphB2/CTF2 was observed to rescue NMDAR-mediated responses better than it rescued AMPAR-mediated responses is consistent with NMDARs being the direct downstream target of EphB2/CTF2, and the rescue of AMPAR currents being due to the increased NMDAR function (Malinow and Malenka, 2002; Shi et al., 2001; Shi et al., 1999). Furthermore, we found that the expression of the EphB2/CTF2 did not alter NMDAR or AMPAR- mediated responses of control neurons. This shows that adding EphB2/CTF2 to PS-MO neurons corrects the specific deficit instead of simply increasing NMDAR-mediated responses via some other mechanism. The lack of effect of EphB2/CTF2 on control neurons also suggests that, in normal tectal neurons, the generation of the fragment is not the limiting factor in synapse formation.

A previous study by Saura and colleagues (Saura et al., 2004) addresses the role of PS in LTP using a PS conditional knockout (PS-cKO) mouse which lacks both PS1 and PS2 in the postnatal forebrain. These mice were found to display impairments in hippocampal memory and LTP, and, consistent with our findings, hippocampal neurons of the PS-cKO mice displayed a significant reduction in NMDAR-mediated responses. Unlike our findings, however, the basal AMPAR-mediated currents displayed by these neurons were not different than wild-type. The discrepancy in the two studies may reflect the differences in timing of PS inhibition: In our study, PS is inhibited during synapse formation – when NMDAR are in the midst of arriving at the synapse and begin to support activity driven recruitment of AMPARs – while in the study by Saura and colleagues, PS is not inhibited until after the circuit has formed (postnatal day 18). Interestingly, in both cases, the observed decrease in basal NMDAR activity displayed by PS-deficient neurons significantly impairs NMDAR-dependent recruitment of AMPARs – whether during activity-dependent synapse formation or activity-dependent plasticity (LTP). The common deficit identified in these two studies underscores the principle that similar mechanisms underlie synapse formation and LTP. Indeed, it would be interesting to determine if the EphB2/CTF2 could rescue the impaired LTP in the PS-cKO mice.

Together, our data suggest a role for PS in the initial formation of excitatory synapses by cleaving and thereby activating the EphB2/CTF2 which then upregulates NMDARs at the synapse. Our data provide functional evidence of the original biochemical studies carried out in cell culture (Xu et al., 2009) and support a molecular model for how the initial stages of excitatory synapse formation may proceed (Figure 8). During development, axons expressing ephrin interact with dendrites expressing EphB2. When ephrin on the axon membrane binds to an EphB2 receptor on the dendrite, PS/γ-secretase is then able to cleave the intracellular domain of the EphB2 receptor (Litterst et al., 2007) which subsequently upregulates NMDARs at the synapse (Xu et al., 2009). The requirement for ephrin to be bound to the EphB receptor for the cascade of events to ensue is a key feature of the model because it ensures that postsynaptic sites form adjacent to ephrin-expressing presynaptic axons to create a functional synapse.

Figure 8. Schematic of the involvement of PS in retinotectal synapse formation.

During early formation and development of the retinotectal synapses, EphrinB on the RGC axon interacts with EphB2 receptor on the tectal cell dendrite and triggers the PS-dependent cleavage of the EphB2 receptor to generate the CTF2 fragment. The EphB2/CTF2 phosphorylates NMDAR and promotes its synapse expression. Increased functional NMDAR at the synapse can further lead to NMDAR dependent recruitment of AMPAR that further contributes synapse strengthening and development.

Linking neurodevelopment with neurodegeneration

It is now well established that neurogenesis and the formation of new synapses occurs not only during development, but also in specific regions of the adult brain. In the adult hippocampus, birth of neurons in the underlying subventricular zone (SVZ) and their subsequent incorporation into the synaptic circuitry of the overlying dentate gyrus is regulated by the same mechanisms that regulate neurodevelopment during embryonic stages of life. It follows, therefore, that the PS/γ-secretase-dependent processes that are associated with neural circuit development are probably also taking place in the healthy adult hippocampus, but perhaps not taking place in PS-deficient neurons or neurons expressing an FAD-linked PS mutation. Prime evidence for this is the discovery of decreased neurogenesis in the SVZ of the adult AD hippocampus which is due to inhibition of Notch processing (Veeraraghavalu et al., 2010). Our work suggests an even more dire situation, for not only will there be fewer neurons born in the SVZ of the AD brain, the inability to form healthy new synapses could impact the proper integration of the new neurons into the circuitry, leading to suboptimal circuit function.