Neuronal Nicotinic Synapse Assembly Requires the Adenomatous Polyposis Coli Tumor Suppressor Protein

Introduction

Nicotinic acetylcholine receptors (nAChRs) function at key interneuronal synapses. Their activation mediates excitatory transmission, reinforces nicotine addiction, and increases memory formation. Malfunction of cholinergic synapses has been implicated in Alzheimer's disease, schizophrenia, nocturnal frontal lobe epilepsy, and autoimmune autonomic neuropathies.

Despite the physiological importance of these synapses, little is known about molecular mechanisms that direct their assembly. Only two proteins have been identified that organize neuronal nicotinic synapses. First, one nAChR subunit, α3, is essential for targeting heteromeric α3-containing nAChRs to the postsynaptic membrane in autonomic neurons (Williams et al., 1998). Second, PSD (postsynaptic density)-93 regulates synaptic stability; α3-nAChR clusters disassemble faster after denervation in PSD-93-deficient versus wild-type autonomic ganglia (Parker et al., 2004). In addition, PSD-93 and PSD-95 couple receptor activation to downstream signaling cascades that regulate plasticity at nicotinic and glutamatergic synapses (Migaud et al., 1998; Conroy et al., 2003).

Because PSD-93 and PSD-95 organize both nicotinic and glutamatergic synapses, it is possible that a common core of proteins may organize diverse synaptic complexes. We propose that adenomatous polyposis coli (APC) may be one of these key organizers. APC is known best as a tumor-suppressor protein: loss-of-function mutations cause colorectal cancer (Livingston, 2001). Although the function of APC in neurons has yet to be defined, several lines of evidence suggest that it may play a role in synaptic differentiation. Muscle-type nAChR clustering induced by agrin requires APC interactions in cultured myotubes (Wang et al., 2003). APC localizes to glutamatergic synapses in hippocampal neurons and binds PSD-95 (Matsumine et al., 1996; Yanai et al., 2000). Moreover, APC is expressed in multiple neuron types and binds β-catenin, a widely expressed adhesion complex component of cadherin-mediated synapses, suggesting APC may have a broader synaptic distribution (Brakeman et al., 1999; Fearnhead et al., 2001). An essential role of APC in neurons is indicated by the correlation of APC gene deletions with human mental retardation (Raedle et al., 2001).

We tested for a role of APC in neuronal synaptic assembly in vivo using the chick parasympathetic ciliary ganglion (CG). In these experimentally tractable neurons, excitatory cholinergic synapses lie adjacent to inhibitory glycinergic synapses (Tsen et al., 2000). Single CG neurons express two nAChR types, α3-nAChRs and α7-nAChRs, and glycine receptors (GlyRs) that predominantly segregate to distinct synapse-associated surface membrane microregions. The α3-nAChRs are concentrated in postsynaptic membrane regions that oppose presynaptic terminal active zones (Jacob et al., 1986; Loring and Zigmond, 1987). Inhibitory GlyR clusters are present in separate but proximal postsynaptic membrane microregions, all under one presynaptic terminal (Tsen et al., 2000). In contrast, α7-nAChRs are excluded from the synapse and localize perisynaptically on somatic spines (Jacob and Berg, 1983; Shoop et al., 1999). The discrete localization of the different receptor types provides a valuable comparison for establishing the specificity of protein interactions responsible for orchestrating cholinergic synapse assembly.

We show here that APC interactions are required for high-density accumulations of α3-nAChRs, but not GlyRs, at postsynaptic sites in CG neurons in vivo. Our data identify APC as a key nicotinic cholinergic synapse-organizer and define a neural function for APC.

Materials and Methods

Antibodies. Primary antibodies used for immunolabeling (1:100 dilution) were: rabbit C-20 to APC C terminus (Santa Cruz Biotechnology, Santa Cruz, CA), mouse ab58 to APC N terminus (Abcam, Cambridge, UK), anti-Chapsyn-110/PSD-93 antiserum (Alomone Labs, Jerusalem, Israel), anti-end binding protein 1 (EB1) monoclonal antibody (mAb) (Transduction Laboratories, Lexington, KY), rabbit anti-EB1 (H-70; Santa Cruz Biotechnology), anti-β-catenin mAb and anti-N-cadherin mAb (Zymed Laboratories, San Francisco, CA), rat mAb35 to α3-nAChRs (Developmental Studies Hybridoma Bank, Iowa City, IA), biotinylated α-bungarotoxin (Molecular Probes, Eugene, OR) to detect α7-nAChRs, mAb2b to GlyRs and mAb7a to gephyrin (Alexis Biochemicals, San Diego, CA), anti-synaptic vesicle 2 (SV2) mAb (Developmental Studies Hybridoma Bank), and pan-PSD-95 family mAb (clone K28/86.2; Upstate Biotechnology, Lake Placid, NY). Rat anti-hemagglutinin (HA) mAb (clone 3F10; Roche Diagnostics, Indianapolis, IN) and anti-HA Y-11 antisera (Santa Cruz Biotechnology) were used for detecting HA-tagged APC dominant-negative (APC-dn) protein. Secondary reagents used at 1:1000-2000 dilution were as follows: Alexa 488- and 594-conjugated secondary antibodies raised in rabbit, rat, and mouse (affinity purified) (Molecular Probes), cyanine 3-conjugated streptavidin (Jackson ImmunoResearch, West Grove, PA), and FITC-conjugated avidin distinct cell sorter grade (Vector Laboratories, Burlingame, CA).

Ciliary ganglion dissociation. White Leghorn embryonated chick eggs were obtained from Charles River Spafas (North Franklin, CT) and maintained at 37°C in a forced air-draft humidified incubator until use. Embryos were staged according to the classification scheme by Hamburger and Hamilton (1951), and the days of embryonic development refer to the stage (st) rather than the actual days of incubation. The developmental stages used in this study include the following: embryonic day 7 (E7) (st 31-32), E11-E13 (st 37-39), and E18 (st 44). CGs at E11-E13 and E18 were acutely dissociated in Ca²⁺-free avian basal salt solution using collagenase (1 mg/ml; Type A; Roche) (Williams et al., 1998). Cells were plated onto poly-d-lysine and laminin-coated chamber slides (BD Labware, Bedford, MA) and allowed to attach to the substratum (30 min at 37°C) before immunolabeling.

Quantitative confocal microscopy. Double-labeling immunofluorescence and confocal microscopy were done as described with minor modifications (Tsen et al., 2000). Surface nAChRs on freshly dissociated intact CG neurons were immunolabeled at 4°C for 1 hr to avoid staining the large internal biosynthetic pool. Next, cells were rinsed with cold basal salt solution, fixed (1% paraformaldehyde for 5 min), and permeabilized (0.1% saponin for 2 min) before labeling with other antibodies (to APC, PSD-93, β-catenin, EB1, N-cadherin, gephyrin, or HA) followed by fluorescent-conjugated secondary reagents. Controls for specific binding in double-labeling studies included omitting the first or second primary antibody in separate tests; only background labeling was detected (data not shown). Cells were examined by confocal microscopy with a Leica (Heidelberg, Germany) TCS SP2 microscope equipped with HeNe (633 nm), Kr (568 nm), and Ar (488 nm) laser sources, using a 63× 1.32 numerical aperture lens. Optical sections were acquired from the top to the bottom of each neuron at ∼0.5 μm step intervals. Laser intensity and photomultiplier tube gain were kept constant across experiments. Settings were chosen such that pixel intensities fell below saturation levels. In addition, the wavelengths of light collected in each detection channel were set such that no detectable bleed-through occurred between the different channels. Pixel intensity profiles along randomly chosen ∼3 μm segments of labeled neuronal surface membrane were assessed using Leica imaging software. Pearson's correlation coefficients (r) were used to quantify the extent of overlap between two immunolabels and were calculated using Leica, Scion Image (Scion, Frederick, MD), and Microsoft Excel (Microsoft, Seattle, WA) software. Pearson's correlation coefficient values represent the mean ± SEM of calculations from two to three different labeled surface membrane areas per cell for four to six neurons for each immunolabel pair.

Light microscopy. For some experiments, APC-dn infected and uninfected age-matched control CGs (E7, E11-E13, and E18) were processed in parallel for frozen sectioning and immunolabeling as described previously (Tsen et al., 2000). By E8, all CG neurons were functionally innervated. To detect surface α3-nAChRs, freshly dissected CGs were treated with collagenase (1 mg/ml; 10 min) to increase reagent access and incubated with mAb35 at 4°C. CGs were rinsed in cold PBS, fixed in 1% paraformaldehyde for 30 min, and processed for frozen sectioning (8-10 μm thick cryosections). Sections were labeled with other antibodies (to HA, gephyrin, APC, PSD-93, or N-cadherin), examined with a Zeiss Axioscope (Zeiss, Thornwood, NY) microscope, and photographed with a SPOT color CCD camera and software (Diagnostic Instruments, Sterling Heights, MI). For all experiments, images were acquired using identical settings for gain and exposure time such that pixel intensities were below saturation levels. To quantify changes in surface membrane-associated labeling in APC-dn versus uninfected control neurons, we assessed pixel intensity profiles along ∼3 μm length segments of the surface membrane using Scion Image. For each immunolabel, pixel intensities for three or more segments per neuron and a total of four to six neurons were evaluated for control and APC-dn expressing CGs. The pixel intensities of the sampled membrane segments were then binned into incremental groups of 10 pixel intensity steps (e.g., 0-9, 10-19..., up to 255 saturation). We then calculated the percentage of pixels that belonged to each pixel intensity category and plotted the relative frequency distributions of pixel intensities (data shown as graphs of the relative frequency polygons) (see Figs. 5i, 6). For HA with EB1, β-catenin, and PSD-93, relative frequency distributions of pixel intensities were acquired using immunolabeled freshly dissociated neurons and confocal microscopy. In addition, we obtained similar results using either frozen sections or freshly dissociated neurons at both E11-E13 and E18.

Cloning. An E15 chick CG cDNA library was constructed in Hybrizap vector (Stratagene, La Jolla, CA). Full-length human APC cDNA (gift from Dr. Bert Vogelstein, Johns Hopkins University, Baltimore, MD) was used as a probe to screen 5 × 10⁸ clones by standard methods. Six positive clones were isolated, sequenced, and identified as overlapping cDNAs of APC.

Yeast two-hybrid assay. Yeast two-hybrid assays were performed as described previously (Tsen et al., 2000). For the bait construct, chick APC cDNA encoding amino acids 2498-2844 (numbering on the basis of homology to human and mouse sequences) was cloned into pBTM116 vector and expressed as a LexA DNA binding domain fusion in yeast L40 reporter strain. The target was an E15 chick CG cDNA library constructed in pAD-Gal4 vector (excised from Hybrizap).

Coprecipitation assays. In vitro binding assays with recombinant fusion proteins were used to test whether HA-tagged APC dominant-negative peptide prevented the coprecipitation of APC with its targeted binding partners EB1 and PSD-93a. We generated glutathione S-transferase (GST) and maltose binding protein (MBP) fusions of chick APC, EB1, and PSD-93a proteins by cloning into pGEX4T-1 and pMalC2 vectors. The fusion proteins were expressed in Escherichia coli BL21-DE3 cells and purified using GST Sepharose (Sigma, St. Louis, MO) or amylose resin (New England Biolabs, Beverly, MA). The affinity-isolated recombinant fusion peptide was used to coprecipitate its binding partner from bacterial lysates. The coprecipitation assays were performed in the presence or absence of purified HA-tagged APC dominant-negative peptide. Bound protein complexes were eluted, separated by SDS-PAGE, transferred to nitrocellulose, and probed with antisera to the epitope tag. Specific binding was established by using GST alone (negative control) to precipitate the MBP-tagged protein, and using the MBP tag alone for binding to the recombinant GST protein.

In vivo coimmunoprecipitation of APC with its binding partners β-catenin and PSD-93 was performed on isolated postsynaptic densities (psds). The psds were isolated by subcellular fractionation using previously described methods (Ehlers, 2003; Parker et al., 2004). Briefly, two E18 chick brains (without the brain stem and cerebellum) were homogenized in homogenization buffer (HB) containing protease inhibitors (4 mm HEPES-NaOH, pH 7.4, 0.32 m sucrose, EDTA free protease inhibitor mixture Complete; Roche) and centrifuged at low speed to remove nuclei and blood cells. The supernatant was diluted with 2 ml of 10% Percoll-HB, layered on top of a discontinuous 10>20% Percoll-HB gradient, and centrifuged 33,000 × g in a Beckman L8 - 80 Ultracentrifuge (Beckman Instruments, Fullerton, CA) using SW-28 rotor. The interface between the two Percoll layers was collected, diluted to 12 ml with HB, and further centrifuged at 20,000 × g for 20 min to pellet the psds. The pellet was resuspended in psd buffer (40 mm HEPES-NaOH, pH 8.1) containing protease inhibitors plus 1 mm EDTA (Complete tablets; Roche), solubilized with 1% Triton X-100, recentrifuged, and the final pellet resuspended in 200-400 μl of psd buffer. The isolated psds were stored at -80°C and used in immunoprecipitation assays. The psds were incubated with 2 μg of antibody to APC or β-catenin, and the immune complexes were affinity precipitated with protein-G or protein-A Sepharose beads (for ms and rb antibodies, respectively). The bound protein complexes were eluted, separated by SDS-PAGE, and analyzed by immunoblotting with the antibody directed against the other protein of the candidate interacting pair (Ikonomov et al., 1998; Conroy et al., 2003). Controls for antibody binding specificity were included.

Retroviral vector-mediated gene transfer. APC-dn construct [amino acids 2500-2844; corresponding to the C-terminus tail of APC that contains the EB1 and PDZ (postsynaptic density-95/Discs large/zona occludens-1) binding domains] (Fearnhead et al., 2001) was generated and coupled to the sequence encoding the hemaglutinin tag (YPYDVPDYA) at its 5′ end by PCR. The APC-dn cDNA was subcloned into the avian-specific retroviral vectors RCASBP of the A and B envelope subgroup types (Homburger and Fekete, 1996). Viral stocks were prepared in DF1 chicken fibroblast cells (American Type Culture Collection, Manassas, VA). CGs were infected in ovo as described previously (Williams et al., 1998) and sampled 1-2 weeks later.

Electron microscopy. E18 CGs were processed for ultrastructural localization of APC relative to synapses using horseradish peroxidase for detection using previously described methods (Williams et al., 1998).

Surface α3-nAChR binding assay. Surface levels of α3-nAChRs were assayed by saturable specific binding of mAb35, followed by biotinylated anti-rat antibody and ¹²⁵I-labeled strepavidin to freshly dissociated, intact neurons from either two retroviral infected CGs or uninfected control CGs as we have done previously (Williams et al., 1998).

Quantitative reverse transcription-PCR. To compare α3 subunit mRNA levels in individual APC-dn infected CGs versus uninfected control CGs at matched ages, total RNA was isolated, and quantitative reverse transcription (RT)-PCR was performed using α3-specific primers and mutated internal standards as described previously (Levey et al., 1995).

Results

Discussion

The major finding reported here is that the tumor-suppressor protein APC is required for high-density accumulations of α3-nAChRs at postsynaptic sites in CG neurons in vivo. This study identifies APC as the first nonreceptor protein to function in localizing nAChRs at neuronal synapses. We identify EB1, β-catenin, and PSD-93 as APC binding partners in the CG. We show that APC, PSD-93, β-catenin, and N-cadherin (the cell adhesion molecule that binds β-catenin) are selectively enriched at cholinergic synapses; they colocalize with surface clusters of α3-nAChRs and one another and are juxtaposed to synaptic vesicle clusters on CG neurons. In contrast, little colocalization was observed for either the neighboring synaptic GlyRs or perisynaptic α7-nAChRs. In addition, the microtubule end-binding protein EB1 shows close proximity to and partial overlap with surface clusters of α3-nAChRs and APC. Importantly, in vivo overexpression of an APC dominant-negative peptide causes dramatic decreases in α3-nAChR surface levels and clusters but not of GlyRs or α7-nAChRs. In addition, the overexpressed APC-dn also leads to reductions in PSD-93 and EB1 surface membrane-associated clusters. These results demonstrate that APC plays a key role in the assembly of excitatory cholinergic, but not inhibitory glycinergic, postsynaptic specializations in CG neurons. This work provides new insights into molecular interactions that direct the formation of neuronal nicotinic synapses in vivo and defines a neural function for APC.

Blocking the interactions of APC with both EB1 and PSD-93 dramatically decreases α3-nAChR surface clusters. However, new studies indicate that PSD-93 is not required for α3-nAChR clustering at neuronal synapses (Conroy et al., 2003; Parker et al., 2004). Together, the data suggest that the interactions of APC with EB1, but not with PSD-93, target α3-nAChRs to postsynaptic sites in neurons.

We speculate that the interactions of APC with EB1 direct α3-nAChR surface delivery (microtubule-mediated transport) or stabilization at postsynaptic sites by regulating the microtubule cytoskeleton. In our model, EB1 tags the plus ends of a subset of microtubules, and APC accumulates at synapses marked by β-catenin/N-cadherin complexes. APC directs α3-nAChR delivery to the synapse by capturing EB1-tagged microtubule plus ends and thereby positioning a microtubule-based transport pathway in close proximity to β-catenin/N-cadherin-marked postsynaptic regions. In addition, APC and EB1 interactions may be required for stabilizing α3-nAChRs at postsynaptic sites by anchoring microtubules and thereby regulating microtubule-cortical actin cytoskeletal interactions. Moreover, these two APC functions may not be mutually exclusive. Our model is based on the roles of APC and EB1 in polarized epithelial cells (Dikovskaya et al., 2001; Lu et al., 2001; Bienz, 2002). In these cells, APC accumulates at specific surface sites and directs the capture of EB1-tagged microtubules (to the β-catenin/cadherin-rich adherens junctions in Drosophila ectodermal cells) (Lee et al., 2000; Lu et al., 2001; McCartney et al., 2001; Barth el al., 2002; Mimori-Kiyosue and Tsukita, 2003). Thus, APC and EB1 interactions anchor microtubule plus ends at precise positions of the cell surface. A key role for APC in tethering microtubules is also suggested by studies that show heterozygous APC mutant mice have substantially fewer microtubule arrays anchored to the specialized basal membrane of polarized epithelial supporting cells (Mogensen et al., 2002). All together, the data suggest that the interactions of APC with EB1 may direct high-density accumulations of α3-nAChRs at postsynaptic sites in vivo.

PSD-93 is a member of the PSD-95-related protein family and a well characterized component of the glutamatergic postsynaptic complex. PSD-93 interacts with and links cytoskeletal, signaling, and transmembrane proteins (Brenman et al., 1996; Kim et al., 1996; McGee et al., 2001; Sheng and Sala, 2001; Garner et al., 2002; Malinow and Malenka, 2002; McGee and Bredt, 2003). Surprisingly, we and others found that PSD-93 is also present at neuronal nicotinic cholinergic postsynaptic sites (Conroy et al., 2003; Parker et al., 2004; our study). However, studies of PSD-93 function using knock-out mice or in vitro models have not yet elucidated the role of PSD-93 in cholinergic synapse formation. The studies show that PSD-93 is not required for cholinergic synapses and α3-nAChR clusters to form or for normal levels of functional nAChRs on autonomic neurons (Conroy et al., 2003; Parker et al., 2004). However, loss of synaptic PSD-93 and PSD-95 in vitro reduces spontaneous EPSC frequency (Conroy et al., 2003). The decreased synaptic activity suggests that PSD-93 may be required to organize a complex that signals retrogradely to the presynaptic terminal and thereby regulates synaptic function (Dean et al., 2003; Prange and El-Husseini, 2003). We identified PSD-93 as an APC binding partner in the CG. Importantly, we show that the overexpressed APC-dn led to decreases in PSD-93 surface clusters in CG neurons, suggesting that APC interactions are required to localize or selectively retain PSD-93 at nicotinic postsynaptic sites in vivo.

Similar to our findings in neurons in vivo, recent work in muscle shows that APC localizes to the neuromuscular junction and is required for nAChR clustering induced by agrin on myotubes in vitro (Wang et al., 2003). Interestingly, APC binds directly to the β1-subunit of the muscle-type nAChRs. In contrast, we found that neither APC nor its binding partners EB1 and PSD-93a directly bind to α3 (yeast two-hybrid-directed interaction assays and in vitro coimmunoprecipitation assays; data not shown). We focused on α3 because our previous studies showed that this subunit is essential to target heteromeric nAChR channels to synapses in CG neurons in vivo (Williams et al., 1998). However, there is no significant sequence identity between the region of β1 that binds to APC and any of the nAChR subunits expressed in CG neurons (α3, α5, β2, β4, or α7).

Consistent with our model that APC functions with other proteins to localize α3-nAChRs to synapses, α3-nAChRs and PSD-93 coprecipitate in a complex in vivo from mammalian and avian autonomic ganglia (Conroy et al., 2003; Parker et al., 2004). However, the proteins do not interact directly in the in vitro binding assays, similar to our findings (Conroy et al., 2003). None of the nAChR subunits have an apparent consensus PDZ-binding motif. Thus, an adapter-like protein may be required to link α3-nAChRs to PSD-93 or APC. Precedence exists for an intermediate protein (Stargazin) linking neurotransmitter receptors (AMPA receptors) and PSD-95 (Chen et al., 2000; Schnell et al., 2002).

The importance of the role of APC in the vertebrate nervous system is highlighted by the correlation of APC gene deletions with various forms of human mental retardation (Raedle et al., 2001). However, the neural function of APC was previously undefined. We show here that APC interactions with its binding partners are required for high-density accumulations of α3-nAChRs at CG synapses in vivo and thereby define a neural function for APC. In addition, APC is the first overlapping component in neurons and muscle identified to function in nicotinic cholinergic synapse formation. Moreover, APC is also concentrated at glutamatergic synapses in hippocampal neurons (Matsumine et al., 1996). The APC binding partner PSD-93 has organizing functions at glutamatergic and nicotinic cholinergic synapses, whereas both proteins are excluded from inhibitory neuronal synapses (Sheng and Sala, 2001; Conroy et al., 2003; McGee and Bredt, 2003; Parker et al., 2004; our study). In summary, accumulating evidence indicates that APC and its interacting proteins may be core organizers of excitatory, but not inhibitory, synapses in the vertebrate nervous system.