α6 nicotinic ACh receptor subunits are expressed in several catecholaminergic nuclei in the central nervous system, including retinal ganglion cells (Gotti et al., 2005b), locus coeruleus (Léna et al., 1999), and dopaminergic neurons located in the substantia nigra and ventral tegmental area (Whiteaker et al., 2000; Zoli et al., 2002; Champtiaux et al., 2003). Ligand-binding studies using the α6-specific probe α-conotoxin MII suggest that many α6^* (^* indicates that other subunits may be present in the receptor) receptors are located on presynaptic terminals in the superior colliculus and striatum (Whiteaker et al., 2000). Indeed, this binding activity disappears in the brains of α6 knockout mice (Champtiaux et al., 2002). This strikingly specific expression pattern could indicate a unique function for α6^* receptors, and α6^* receptors are candidate drug targets for diseases or disorders such as Parkinson's disease or nicotine addiction (Quik and McIntosh, 2006).

Functional, voltage-clamped α6-dependent responses are elusive in heterologous expression systems such as Xenopus laevis oocytes (Kuryatov et al., 2000; Broadbent et al., 2006), but native α6^* receptors are readily studied using synaptosome preparations from brain tissue (Whiteaker et al., 2000; Grady et al., 2002; Champtiaux et al., 2003; Gotti et al., 2005a). Indeed, α-conotoxin MII-sensitive receptors are pharmacologically and stoichiometrically distinct from α-conotoxin MII-resistant receptors in mediating [³H]dopamine release from striatal synaptosomes (Grady et al., 2002; Salminen et al., 2007). Recent studies using α4 and β3 knockout mice demonstrate the existence of functional α6β2, α6β2β3, α6α4β2, and α6α4β2β3 receptors (Salminen et al., 2007). It is noteworthy that native α6α4β2β3 receptors have the highest affinity (EC₅₀ = 0.23 ± 0.08 μM) for nicotine of any nicotinic receptor reported to date. Because nicotine is likely to be present at concentrations ≤0.5 μM in the cerebrospinal fluid of smokers (Rowell, 2002), only those receptors with the highest affinity for nicotine, including some α4^* and α6^* receptors, are likely to be important in nicotine addiction. Although previous studies offer major conceptual advances in our understanding of α6^* receptors in the brain, there is a lack of information regarding the subcellular localization and biophysical properties of α6 subunits.

β3 subunits are expressed in most of the same locations as α6, including midbrain dopaminergic neurons projecting to the striatum (Zoli et al., 2002). β3 knockout mice demonstrate that β3 subunits are important for the biogenesis of α6^* receptors in the brain (Cui et al., 2003; Gotti et al., 2005a). This is corroborated by studies in X. laevis oocytes and tissue culture cells (Kuryatov et al., 2000). β3 also increases α6-specific binding activity in HEK293 cells (Tumkosit et al., 2006). Uncertainty exists, however, because others have reported that β3 incorporation into nAChRs acts as a dominant negative (Boorman et al., 2003; Broadbent et al., 2006), suppressing ACh-evoked responses by an incompletely understood gating mechanism. This effect occurred apparently without significantly altering the surface expression of nAChRs. What is clear is that β3 acts more like a muscle β subunit than a “typical” neuronal β subunit; it does not participate in forming the α:non-α interface that comprises the neuronal ligand binding site, and other β subunits, either β2 or β4, must be present to form functional nicotinic receptors (Broadbent et al., 2006). This presents a problem both for basic and therapeutic-oriented research on β3^* receptors, because there are no pharmacological ligands that can visually or functionally isolate β3-specific actions in cell culture systems or intact brain tissue. Given the precise localization and unique functional properties of β3^* receptors, potential for therapeutic intervention that would be afforded by β3-specific probes, and involvement in nicotine addiction (Bierut et al., 2007), it is important to develop and exploit tools to study β3.

We have sought to compare characteristics of α6 and β3 with the better understood α4 and β2 subunits. We previously generated fluorescently labeled α4 and β2 subunits, and we used these subunits to study assembly, trafficking, and nicotine-dependent up-regulation of α4β2 receptors (Nashmi et al., 2003). We have now fluorescently labeled α6 and β3 subunits by inserting a yellow fluorescent protein (YFP) or cyan fluorescent protein (CFP) in the M3-M4 intracellular loop. With this approach, one can optically monitor functional nicotinic ACh receptors containing these subunits in live cells and in real time. We measured 1) functional responses using two-electrode voltage-clamp and patch-clamp electrophysiology, 2) subcellular distribution and colocalization in neurons using confocal microscopy and spectral imaging, 3) receptor assembly and subunit stoichiometry using Förster resonance energy transfer (FRET), and 4) plasma membrane localization and distribution patterns using total internal reflection fluorescence (TIRF) microscopy.

Materials and Methods

Reagents. Unless otherwise noted, all chemicals were from Sigma-Aldrich (St. Louis, MO). DNA oligonucleotides for PCR and site-directed mutagenesis were synthesized by Integrated DNA Technologies, Inc. (Coralville, IA). Restriction enzymes for molecular biology were purchased from Roche Diagnostics (Indianapolis, IN) or New England Biolabs (Ipswich, MA). Glass-bottomed dishes (35 mm) coated with l-polylysine were purchased from MatTek (Ashland, MA).

Cell Culture and Transfection. N2a cells (American Type Culture Collection, Manassas, VA) were maintained in Dulbecco's modified Eagle's medium (high glucose with 4 mM l-glutamine; Invitrogen, Carlsbad, CA)/Opti-MEM (Invitrogen) mixed at a ratio of 1:1 and supplemented with 10% fetal bovine serum (Invitrogen), penicillin (Mediatech, Herndon, VA), and streptomycin (Invitrogen). N2a cells were transfected in DMEM without serum or antibiotics. Transfection was carried out using Lipofectamine/PLUS (Invitrogen) according to the manufacturer's instructions and with the following modifications. For a 35-mm dish, 1 to 2 μg of total plasmid DNA was mixed with 100 μl of DMEM and 6 μl of PLUS reagent. DMEM/DNA was combined with a mixture of 100 μl of DMEM and 4 μl of Lipofectamine reagent. Rat hippocampal neurons were dissociated and plated on glass-bottomed imaging dishes as described previously (Slimko et al., 2002). For primary neuron transfection, Lipofectamine 2000 (Invitrogen) was used in conjunction with Nupherin (BIOMOL Research Laboratories, Plymouth Meeting, PA) as described below. In brief, in total 1 μg of DNA was incubated with 20 μg of Nupherin in 400 μl of Neurobasal medium without phenol red (Invitrogen), whereas 10 μl of Lipofectamine 2000 was mixed in 400 μl of Neurobasal medium (Invitrogen). After 15 min, the two solutions were combined and incubated for 45 min. Neuronal cultures in 35-mm glass-bottomed culture dishes were incubated in the resulting 800-μl mixture for 120 min, followed by removal of transfection media and refeeding of the original, pretransfection culture media.

Plasmids and Molecular Biology. Mouse α4 and β2 nAChR cDNAs in pCI-neo, both untagged and modified with YFP or CFP fluorescent tags, have been described previously (Nashmi et al., 2003). Mouse α3 and β3 nAChR cDNAs in pCDNA3.1 were a generous gift of Jerry Stitzel (Institute for Behavioral Genetics, University of Colorado, Boulder, CO). A full-length mouse α6 I.M.A.G.E. cDNA (ID no. 4501558) was obtained from Open Biosystems (Huntsville, AL). A modified α6 cDNA was constructed that 1) lacked the 5′ and 3′ untranslated regions and 2) contained a Kozak sequence (GCC ACC) before the ATG start codon to facilitate efficient translation initiation. Rat β4 cloned into pAMV was provided by Cesar Labarca (California Institute of Technology, Pasadena, CA). pEYFP-N1 and pECFP-N1 (Clontech, Mountain View, CA) were used to construct fluorescent nAChR cDNAs. mGAT1 labeled with CFP was provided by Fraser Moss (California Institute of Technology). YFP-Syntaxin was provided by Wolfhard Almers (Vollum Institute, Oregon Health and Science University, Portland, OR). CFP-tau was provided by George Bloom (University of Virginia, Charlottesville, VA). A QuikChange (Stratagene, La Jolla, CA) kit was used to construct β3 (WT or XFP-modified) cDNAs containing a V13'S point mutation.

To design fluorescently labeled α6 and β3 subunits, we chose to insert the XFP moiety in the M3-M4 loop each subunit. We have previously found that this region is appropriate for insertion in nAChR α4 and β2 subunits (Nashmi et al., 2003), the nAChR γ subunit (data not shown), and GluCl α and β subunits (Slimko et al., 2002). Similar to our previous studies, we inserted the XFP moiety in the M3-M4 loop at positions that avoided the conserved amphipathic α-helix and putative cell sorting motifs and phosphorylation sites (Fig. 1, A and B). To construct nAChRs with XFP inserted into the M3-M4 loop, a two-step PCR protocol was used. First, YFP or CFP was amplified with PCR using oligonucleotides (sequences available upon request) designed to engineer 5′ and 3′ overhangs of 15 base pairs that were identical to the site where XFP was to be inserted, in frame, into the nAChR M3-M4 loop. A Gly-Ala-Gly flexible linker was engineered between the nAChR sequence and the sequence for YFP/CFP at both the 5′ and 3′ ends. In the second PCR step, 100 ng of the first PCR reaction was used as a primer pair in a modified QuikChange reaction using Pfu Ultra II (Stratagene, Cedar Creek, TX) polymerase and the appropriate nAChR cDNA as a template. All DNA constructs were confirmed with sequencing and, in some cases, restriction mapping.

cRNA for injection and expression in X. laevis oocytes was prepared using a T7 or SP6 in vitro transcription kit (mMessage mMachine; Ambion, Foster City, CA) according to the manufacturer's instructions. RNA yield was quantified with absorbance at 260 nm. RNA quality was assessed by observing absorbance profiles across a range of wavelengths between 220 and 320 nm. Spectrophotometric analysis was performed using a ND-1000 spectrophotometer (Nano-Drop, Wilmington, DE).

Confocal Microscopy. N2a cells were plated on 35-mm glass-bottomed dishes, transfected with nAChR cDNAs, and they were imaged live 24-48 h after transfection. X. laevis oocytes were imaged 3 days after RNA injection. Oocytes were placed in an imaging chamber and allowed to settle for 20 min before imaging. To eliminate autofluorescence, growth medium was replaced with an extracellular solution containing the following components: 150 mM NaCl, 4 mM KCl, 2 mM CaCl₂, 2 mM MgCl₂, 10 mM HEPES, and 10 mM d-glucose, pH 7.4. Cells were imaged with a Nikon (Nikon Instruments, Melville, NY) C1 laser-scanning confocal microscope system equipped with spectral imaging capabilities and a Prior (Rockland, ME) remote-focus device. For oocytes, a Nikon Plan Apo 20 × 0.75 numerical aperture (NA) air objective was used, whereas a Nikon Plan Apo 60 × 1.40 NA oil objective was used for mammalian tissue culture cells. Pinhole diameter was 30-60 μm, and cells were imaged at 12-bit intensity resolution over 512 × 512 pixels at a pixel dwell time of 4 to 6 μs. CFP was excited using a 439.5-nm modulated diode laser, and YFP was excited with an argon laser at 514.5-nm. In most cases, imaging was carried out using the Nikon C1si DEES grating and spectral detector with 32 parallel photomultiplier tubes. This allowed us to collect spectral images (λ stacks). In such images, each pixel of the X-Y image contains a list of emission intensity values across a range of wavelengths. We collected light between 450 and 600 nm at a bandwidth of 5 nm. The 515-nm channel was intentionally blocked while we used the 514.5-nm laser for YFP bleaching. Because the emission profile of YFP and CFP significantly overlap, we used the Nikon EZC1 linear unmixing algorithm to reconstruct YFP and CFP images. Experimental spectral images with both YFP- and CFP-labeled nAChR subunits were unmixed using reference spectra from images with only YFP- or CFP-labeled nAChR subunits. For each pixel of a spectral image, intensity of YFP and CFP was determined from fluorescence intensity values at the peak emission wavelength derived from the reference spectra.

Spectral FRET Analysis. To examine FRET between various nAChR subunits, the acceptor photobleaching method (Nashmi et al., 2003) was used with a modified fluorescence recovery after photobleaching macro built into the Nikon EZC1 imaging software. In this method, FRET was detected by recording CFP dequenching during incremental photodestruction of YFP. A spectral image was acquired once before YFP bleaching and at six time points every 10 s during YFP bleaching at 514.5 nm. Laser power during bleaching varied from cell to cell, but was between 25 and 50%. One bleach scan per cycle was used. This bleaching protocol was optimized to achieve 70 to 80% photodestruction of YFP while still enabling us to record incremental increases in CFP emission at each time point. In the confocal microscope, nAChRs labeled with XFP usually exhibit a uniform, intracellular distribution, regardless of the subunit being examined. This is consistent with our previous observations (Nashmi et al., 2003). To measure FRET, spectral images were unmixed into their CFP and YFP components as described above. We found little or no difference in FRET for various cellular structures or organelles in N2a cells, and we measured CFP and YFP mean intensity throughout the entire cell by selecting the cell perimeter as the boundary of a region of interest in Nikon's EZC1 software. CFP and YFP components were saved in Excel format, and fluorescence intensities were normalized to the prebleach time point (100%). FRET efficiency (E) was calculated as E = 1 - (I_DA/I_D), where I_DA represents the normalized fluorescence intensity of CFP (100%) in the presence of both donor (CFP) and acceptor (YFP), and I_D represents the normalized fluorescence intensity of CFP in the presence of donor only (complete photodestruction of YFP). The I_D value was extrapolated from a scatter plot of the fractional increase of CFP versus the fractional decrease of YFP. The E values were averaged from several cells per condition (see Table 1 for n values). Data are reported as mean ± S.E.M.

TIRF Microscopy. N2a cells cultured in glass-bottomed, polyethylenimine-coated imaging dishes were transfected with cDNA mixtures as described above. Cells, superfused with the same imaging solution used for confocal microscopy, were imaged 18 to 24 h after transfection to minimize overexpression artifacts. TIRF images were obtained with an inverted microscope (Olympus IX71; Olympus America, Inc., Center Valley, PA) equipped with a 488-nm air-cooled argon laser (P/N IMA101040ALS; Melles Griot, Carlsbad, CA). Laser output was controlled with a UNIBLITZ shutter system and drive unit (P/N VMM-D1; Vincent Associates, Rochester, NY) equipped with a Mitutoyo (Mitutoyo America, City of Industry, CA) micrometer to control TIRF evanescent field illumination. TIRF imaging was carried out with an Olympus PlanApo 100 × 1.45 NA oil objective, and images were captured with a 16-bit resolution Photometrics Cascade charge-coupled device camera (Photometrics, Tucson, AZ) controlled by SlideBook 4.0 imaging software (Intelligent Imaging Innovations, Santa Monica, CA).

Two-Electrode Voltage-Clamp Electrophysiology. Stage V to VI X. laevis oocytes were isolated as described previously (Quick and Lester, 1994). Stock RNAs were diluted into diethyl pyrocarbonate-treated water and injected 1 day after isolation. RNA was injected in a final volume of 50 nl per oocyte using a digital microdispenser (Drummond Scientific, Broomall, PA). After injection, oocytes were incubated in ND-96 solution (96 mM NaCl, 2 mM KCl, 1 mM CaCl₂, 1 mM MgCl₂, and 5 mM HEPES/NaOH, pH 7.6) supplemented with 50 μg/ml gentamicin and 2.5 mM sodium pyruvate. After 1 to 4 days for nAChR expression, oocytes were used for recording or confocal microscopy.

Agonist-activated nicotinic receptor responses were measured by two-electrode voltage-clamp recording using a GeneClamp 500 (Molecular Devices, Sunnyvale, CA) voltage clamp. Electrodes were constructed from Kwik-Fil borosilicate glass capillary tubes (1B150F-4; WPI, Sarasota, FL) using a programmable microelectrode puller (P-87; Sutter Instrument Company, Novato, CA). The electrodes had tip resistances of 0.8 to 2.0 MΩ after filling with 3 M KCl. During recording, oocytes were superfused with Ca²⁺-free ND-96 via bath-application and laminar-flow microperfusion using a computer-controlled application and washout system (SF-77B; Warner Instruments, Hamden, CT) (Drenan et al., 2005). The holding potential was -50 mV, and ACh was diluted in Ca²⁺-free ND-96 and applied to the oocyte for 2 to 10 s followed by rapid washout. Data were sampled at 200 Hz and low-pass filtered at 10 Hz using the GeneClamp 500 internal low-pass filter. Membrane currents from voltage-clamped oocytes were digitized (Digidata 1200 acquisition system; Molecular Devices) and stored on a PC running pCLAMP 9.2 software (Molecular Devices). Concentration-response curves were constructed by recording nicotinic responses to a range of agonist concentrations (six to nine doses) and for a minimum of six oocytes. EC₅₀ and Hill coefficient values were obtained by fitting the concentration-response data to the Hill equation. All data are reported as mean ± S.E.M.

Whole-Cell Patch-Clamp Electrophysiology. N2a cells expressing YFP-labeled nicotinic receptors were visualized with an inverted microscope (Olympus IMT-2, DPlan 10 × 0.25 NA and MPlan 60 × 0.70 NA) under fluorescence illumination (mercury lamp). Patch electrodes (3-6 MΩ) were filled with pipette solutions containing 88 mM KH₂PO₄, 4.5 mM MgCl₂, 0.9 mM EGTA, 9 mM HEPES, 0.4 mM CaCl₂, 14 mM creatine phosphate (Tris salt), 4 mM Mg-ATP, and 0.3 mM GTP (Tris salt), pH 7.4 with KOH. The extracellular solution was 150 mM NaCl, 4 mM KCl, 2 mM CaCl₂, 2 mM MgCl₂, 10 mM HEPES, and 10 mM d-glucose, pH 7.4. Standard whole-cell recordings were made using an Axopatch 1-D amplifier (Molecular Devices), low-pass filtered at 2 to 5 kHz, and digitized online at 20 kHz (pClamp 9.2; Molecular Devices). Series resistance was compensated 80%, and the membrane potential was held at -70 mV. Recorded potentials were corrected for junction potential.

ACh was delivered using a two-barrel glass θ-shaped tube (outer diameter ∼200 μm; pulled from 1.5-mm-diameter θ-shaped borosilicate tubing) connected to a piezoelectric translator (LSS-3100, Burleigh Instruments, Fishers, NY) as described previously (Nashmi et al., 2003). ACh was applied for 500 ms (triggered by pCLAMP 9.2), and solution exchange rates measured from open tip junction potential changes during application with 10% extracellular solution were typically ∼300 μs (10-90% peak time). Data are reported as mean ± S.E.M. for the peak current response to 1 μM ACh, and statistical significance was determined using a Wilcoxon signed rank test.

Results

Design and Construction of α6 and β3 XFP Fusions. Based on previous work (Slimko et al., 2002; Nashmi et al., 2003), we chose to insert XFP fusions in the M3-M4 loop of mouse α6 and β3 nAChR subunits. Like all members of the Cys-loop family, α6 and β3 have predicted α-helices at the N- and C-terminal ends of their M3-M4 loop (Fig. 1, A and B) that may be important in ion permeation (Miyazawa et al., 1999). In addition to avoiding these regions, we also avoided potential phosphorylation sites and trafficking motifs (Fig. 1, A and B). Our XFP fusion cassette also consisted of a Gly-Ala-Gly flexible linker flanking the XFP open reading frame on both the N- and C-terminal side. We built three independent XFP fusions for α6 and two XFP fusions for β3 (Fig. 1C). These were designated according to the residue immediately N-terminal to the beginning of the Gly-Ala-Gly linker (e.g., α6-YFP^G366 denotes that the GAG-YFP-GAG cassette was inserted between G366 and V367). Unless otherwise noted, all experiments were conducted with α6-XFP^A405 and β3-XFP^P379.

Functional Expression of α6 and β3 Subunits. Despite exhaustive attempts to functionally reconstitute α6^* nAChRs in X. laevis oocytes and mammalian tissue culture cells, we recorded no robust, reproducible responses from cells expressing α6, either with untagged subunits or the fluorescent subunits (Supplemental Data; Table 1). β3-YFP, however, was well expressed on the plasma membrane of X. laevis oocytes when coexpressed with α3 and β4 subunits to support functional expression (Fig. 2A). As a control for oocyte autofluorescence, we imaged oocytes expressing untagged β3 subunits (Fig. 2A). No fluorescence was detected in this case, indicating that our β3-YFP signal was specific.

β3 subunits do not drastically alter the EC₅₀ for ACh or nicotine when incorporated into nAChRs (Boorman et al., 2003), but they do profoundly alter single-channel kinetics (Boorman et al., 2003). Channel burst duration was significantly shortened for nAChRs containing β3 versus those without it (Boorman et al., 2003), suggesting that β3 reduces the probability of channel opening, P_open. Consistent with this, macroscopic voltage-clamped responses from oocytes and mammalian cells expressing β3^* receptors were significantly smaller than for non-β3^* receptors (Broadbent et al., 2006). To assess the functionality of our β3-YFP construct, we compared the ability of untagged and YFP-labeled β3 subunits to attenuate nicotinic responses. β3 must be coexpressed with other α and β subunits, so we chose to use α3β4 receptors for this purpose. We did so because β3 has been well characterized with this receptor combination (Boorman et al., 2003; Broadbent et al., 2006). When WT, untagged β3 was coexpressed with α3β4 receptors, we found a significant attenuation of the peak response to 200 μM ACh (Fig. 2B), consistent with previous findings (Broadbent et al., 2006). When β3-YFP was tested in this assay, it was also able to attenuate the maximal response in a manner identical to untagged β3 (Fig. 2B). It is possible that, although β3-WT attenuates responses via a gating mechanism on the plasma membrane, YFP-labeled β3 might do so via a different mechanism such as sequestering α3 or β4 subunits inside the cell. To further test the functionality of YFP-labeled β3, we took advantage of the fact that a gain-of-function TM2 mutation in β3 is able to reverse the attenuation of peak responses seen for β3-WT (Broadbent et al., 2006). We reasoned that if the YFP label in the M3-M4 loop is not disturbing the function of β3, we should detect the same gain-of-function response for unlabeled and YFP-labeled β3 when they are engineered to express a mutation of this sort. When a Val13′ to Ser mutation (V13S) was introduced into unlabeled β3, we observed not only a reversal of this attenuation behavior, but a significant increase in the peak response to 200 μM ACh with α3β4 receptors (Fig. 2C). When β3-YFP^V13S was tested in this assay, we observed an identical behavior. Taken together, these data suggest that β3-YFP is fully functional and incorporates into nAChRs in X. laevis oocytes.

To further characterize our β3-YFP construct, we constructed concentration-response curves for α3β4 receptors containing either β3WTor β3-YFP. Consistent with previous reports (Boorman et al., 2000), we measured an EC₅₀ for ACh of 230 ± 22 μM for α3β4β3 receptors, which is slightly higher than for α3β4 (165 ± 9 μM) (Fig. 3A). When β3-YFP was substituted for WT β3, the EC₅₀ was shifted slightly, but acceptably, to 109 ± 8 μM (Fig. 3A). We also noticed that the addition of β3 to α3β4 receptors increased the Hill coefficient from 1.5 ± 0.1 to 2.0 ± 0.3, and this effect was retained when β3-YFP was coexpressed with α3β4 receptors. Likewise, we also constructed concentration-response relationship curves for oocytes expressing β3^V13S and β3-YFP^V13S. Compared with the EC₅₀ for α3β4β3 (230 ± 22 μM), we measured an EC₅₀ for α3β4β3^V13S of 28 ± 3 μM (Fig. 3B). This is consistent with others who have reported an approximate 6-fold reduction in EC₅₀ for the inclusion of β3 with a similar hypersensitive mutation, Val9'Ser (Boorman et al., 2000). We reasoned that if β3-YFP retained the WT function of β3, then there should be a similar gain-of-function phenotype when it is coexpressed with α3β4. We measured an EC₅₀ for α3β4β3-YFP^V13S of 34 ± 3 μM (Fig. 3B), confirming that this construct behaves identically to β3-WT. Collectively, our work in X. laevis oocytes with YFP-labeled β3 subunits suggests that insertion of YFP into the M3-M4 loop does not significantly alter the assembly, subcellular trafficking, or function of this subunit.

Subcellular Localization and Trafficking of α6 and β3 Subunits. To probe the subcellular localization and trafficking of α6^* and β3^* receptors, we chose a mouse neuroblastoma cell line, N2a, to transiently express our fluorescent nicotinic receptor subunits. We prefer these cells over, for example, HEK293 cells, because they 1) are of mouse origin, the same species as our fluorescent constructs; 2) are of neuronal origin, suggesting that they will be a permissive environment for correct expression, subcellular localization, and assembly of our ectopic nAChR subunits; and 3) express only moderate quantities of transfected membrane protein. To study the subcellular localization of β3^* receptors, we coexpressed β3-YFP with the previously described fluorescently labeled α4 and β2 subunits (Nashmi et al., 2003). β3 is able to assemble and function when coexpressed with α4β2 receptors (Broadbent et al., 2006). When coexpressed with fluorescent α4 or β2 receptors, β3-YFP was localized primarily in the endoplasmic reticulum of live N2a cells (Fig. 4A, i and B, i). We used CFP-labeled α4 (Fig. 4B, ii) or β2 (Fig. 4, A and C, ii) subunits along with a confocal microscope with spectral imaging capabilities to unambiguously assign YFP and CFP signals to each pixel for the spectral images of the cells. In these experiments, YFP was assigned green, CFP was assigned red, and yellow indicated pixels where β3-YFP was colocalized with either β2-CFP (Fig. 4A, iii) or α4-CFP (Fig. 4B, iii). We noted that β3-YFP was completely colocalized with either α4 or β2 in this experiment, suggesting that these subunits are assembled in the same pentameric receptors. To further define the extent of this colocalization, we plotted the β3-YFP and α4-CFP or β2-CFP pixel intensity across a two-dimensional region of interest transecting the cell (Fig. 4, A and B, iv). We noted that the YFP and CFP intensity profiles strongly resembled each other, suggesting that these subunits were indeed colocalized and coassembled in intracellular compartments of the cell. With respect to α4β2^* receptors, this localization pattern is not an artifact of overexpression, because this is the same pattern we observed previously (Nashmi et al., 2003). This is also the expression pattern of endogenous, YFP-labeled α4^* receptors in α4-YFP knockin mice (Nashmi et al., 2007). This indicates that 1) a large pool of intracellular receptors exists in neurons, and 2) YFP tag does not interfere with the delivery of receptors to the plasma membrane. Thus, the localization pattern we observe here for β3 subunits is the expected result if it is assembling with α4β2 receptors.

We expressed α6-YFP along with β2-CFP in N2a cells, and we analyzed its localization pattern as described above for β3. We also found that α6 was localized in intracellular compartments in the cell (Fig. 4C, i), and that it was completely colocalized with β2 subunits (Fig. 4C, ii-iv). Although this is the first fluorescence imaging reported for α6^* receptors, there is other evidence to corroborate our findings. Studies with [³H]epibatidine demonstrate that a significant portion of α6β2 and α6β2β3 receptors are intracellular (∼50 and ∼20%, respectively), although some are delivered to the surface (Kuryatov et al., 2000; Tumkosit et al., 2006).

To further investigate the subcellular localization and trafficking of α6 and β3 subunits, we imaged live, differentiated N2a cells and primary neurons. N2a cells can be induced to differentiate and undergo neurite outgrowth if serum is withdrawn and an activator of protein kinase A, dibutyryl-cAMP, is added (Fowler et al., 2001). In our previous work, α4β2 receptors were localized to dendrites, but not axons, when expressed in primary midbrain neurons (Nashmi et al., 2003). We were interested in whether our fluorescent nicotinic receptor subunits were localized to N2a cell processes in a manner analogous to dendrites in primary neurons. Furthermore, we wanted to address the question of whether β3 is localized with other subunits at distal sites such as dendrites. This is an unsolved question, as there is no high-affinity probe (pharmacological or immunological) that can reliably and unambiguously isolate β3^* receptors. N2a cells were plated on glass-bottomed dishes, and they were then differentiated for 2 days (see Materials and Methods) followed by transfection with various combinations of YFP-labeled and unlabeled nAChR subunits. Cells were also co-transfected with an expression plasmid for soluble CFP to mark total cell morphology. We found that α4β2 receptors were indeed localized to neuronal processes in differentiated N2a cells (Fig. 4D, arrow), along with abundant expression in the cell soma. When β3-YFP was coexpressed with α4β2, we observed a very similar pattern. We found that β3 was present even at the most distant elements of neuronal processes (Fig. 4D, arrow). Because this pattern is identical to that of α4β2 in differentiated N2a cells, we conclude that β3 is likely assembling with α4β2 receptors and that the YFP label in the M3-M4 loop is not disrupting the normal cellular trafficking of α4β2β3 pentamers. To further characterize the localization of β3^* receptors, we coexpressed β3-YFP with α4β2 receptors in primary rat hippocampal neurons (Nashmi et al., 2003). To minimize overexpression artifacts, cells were imaged live only 18 to 24 h after transfection. We found that β3^* receptors were localized very similarly to α4β2 receptors in our previous studies with primary neurons; we noted uniform localization in the soma, suggestive of endoplasmic reticulum, and dendritic localization (Fig. 5A, arrow) and an absence of localization in axons. A high-magnification micrograph demonstrates the dendritic localization of these putative α4β2β3 receptors (Fig. 5A, right). In cells coexpressing α4/β2/β3Y with soluble CFP [to mark total cell morphology, similar to Nashmi et al. (2003)], β3 subunits did not traffic to a subregion of the cell interior likely to be axons (data not shown). To more directly determine whether β3^* receptors could be localized to axons in these neurons, we coexpressed α4β2β3Y receptors with a CFP-labeled axonal marker, tau. The tau-CFP decorated axons in hippocampal neurons, with proximal (relative to the cell body) portions of the axon being labeled more strongly than distal portions (Fig. 5B). In all cells examined, we noted the presence of YFP-labeled β3 subunits in these proximal axons but not distal axons (Fig. 5B, arrow). These data in differentiated N2a cells and primary neurons suggest that β3 assembles efficiently with α4β2 receptors, and it is thus cotrafficked and targeted to distal sites in neurons.

Because α6-YFP^* receptors do not function in our hands, we wanted to determine whether this is due to a subtle trafficking defect that could prevent the correct delivery of α6-YFP to the plasma membrane. Although we could readily detect α6 fluorescence in the cell body of undifferentiated N2a cells, we wanted to further probe the cellular trafficking of α6^* receptors by expressing them in differentiated N2a cells that contain processes. To evaluate the subcellular localization of α6^* receptors, we expressed α6-YFP with β2 in differentiated N2a cells. To our surprise, we found that α6β2 receptors were trafficked to neuronal processes in a manner analogous to α4β2 and α4β2β3 receptors (Fig. 4D, arrow). To further address this question, we expressed α6-YFPβ2 receptors in rat hippocampal neurons as described for α4β2β3-YFP. We observed a localization pattern for α6-YFP that was very similar to α4β2β3-YFP. These receptors were well expressed in the cell soma, but they were readily detectable in dendrites as well (Fig. 5A, arrow). In experiments with coexpressed soluble CFP and α6-YFPβ2 receptors, α6 subunits were not detected in putative axons (data not shown). In tau-CFP/α6-YFPβ2 coexpression experiments, α6 subunits (similar to α4β2 but not α4β2β3 receptors) were not detected in tau-labeled axons (Fig. 5B, arrow). These data indicate that, although α6^* receptors produce little or no agonist-induced conductance in mammalian tissue culture cells, they are expressed well and trafficked similarly compared with α4β2 and β3^* receptors.

FRET Revealed Assembly of α6 and β3 Subunits into nAChR Pentamers. The fact that α4/β2/β3 and α6/β2 subunits are colocalized in the cell body and cotargeted to processes and dendrites in neurons suggests that they are assembled into pentameric receptors. The question of receptor assembly is often answered by simply measuring agonist-induced conductance increases in cells expressing a subunit combination of interest, or by applying a selective agonist or inhibitor to a pure receptor population of known pharmacological properties. This approach is not applicable, however, for α6^* and β3^* receptors. α6^* receptors do not function well in heterologous expression systems, so it is not straightforward to determine the extent to which free α6 subunits assemble into pentameric receptors. Similarly for β3, although it is functional in oocytes (Fig. 2), there are no pharmacological probes that can be applied to β3^* receptors to study their assembly or subunit composition. Others have indirectly measured receptor assembly of nicotinic subunits by using biochemical techniques such as immunoprecipitation (Zoli et al., 2002; Champtiaux et al., 2003) and centrifugation (Kuryatov et al., 2000) or by forcing subunits to assemble by using molecular concatamers (Tapia et al., 2007). To directly determine whether two nicotinic receptor subunits interact and, possibly, assemble to form pentameric receptors, we have used FRET coupled with our CFP- and YFP-tagged receptors. In the context of our nicotinic receptor subunits labeled with YFP or CFP in the M3-M4 loop, only subunits that interact will undergo FRET, because FRET occurs only when donors and acceptors are within 100 Å. Furthermore, we previously demonstrated that the efficiency of FRET directly correlates with the number of functional, plasma membrane-localized pentameric receptors (Nashmi et al., 2003). To measure FRET between subunits, we used the acceptor photobleaching method (Nashmi et al., 2003). In this method, we measure CFP dequenching during incremental photodestruction of YFP. CFP was excited at 439 nm, whereas YFP was bleached at 514 nm (Fig. 6A). Because the emission spectra for CFP and YFP overlap significantly, we imaged using a confocal microscope with spectral imaging capabilities along with a linear unmixing algorithm (described under Materials and Methods).

Fluorescent α4 and β2 subunits are functional and undergo robust FRET in mammalian cells (Nashmi et al., 2003), so we used these subunits in our acceptor photobleaching assay with XFP-tagged β3 and α6. We expressed β3-YFP with untagged α4 and β2-CFP in N2a cells, followed by live cell FRET imaging. We recorded the whole-cell fluorescence intensity for β3-YFP and β2-CFP before and after photobleaching of YFP with the 514-nm laser, and we expressed with pseudocolor intensity scaling (Fig. 6B). In this experiment, β2-CFP was clearly dequenched after β3-YFP photodestruction (Fig. 6B), indicating that the two subunits had been undergoing FRET. In a similar experiment, we recorded multiple spectral images at several time points during YFP photodestruction. This revealed a corresponding increase in CFP intensity (Fig. 6C). A reciprocal experiment was also done, where β3-YFP was coexpressed with α4-CFP and untagged β2. We recorded a similar dequenching for α4-CFP after YFP photobleaching (Fig. 6, D and E), indicating FRET between these subunits as well. Both for β2/β3 and α4/β3 FRET, we found no difference between FRET inside the cell versus FRET at the cell periphery at or near the plasma membrane. These results directly demonstrate that β3 is able to assemble with α4β2 receptors in neuronal cells. This assembly likely occurs in the endoplasmic reticulum, which is consistent with previous findings (Nashmi et al., 2003).

There are many different putative α6^* receptor subtypes in brain, including α6β2, α6β2β3, α6α4β2, and α6α4β2β3 (Salminen et al., 2007). To begin to study α6^* receptor assembly, we measured FRET between α6-YFP and β2-CFP. In response to YFP bleaching, we recorded a robust dequenching of β2-CFP throughout the cell, indicating FRET between these subunits (Fig. 6, F and G). The pattern of localization and FRET pattern was identical to α4β2β3 receptors.

To further quantify FRET between α4/β2 subunits and β3 or α6, we measured FRET E values for various receptor subtypes. α4-CFPβ2-YFP, α4β2-CFPβ3-YFP, and α4-CFPβ2β3-YFP receptors were expressed in N2a cells followed by acceptor photobleaching FRET (Fig. 7A). We acquired spectral images with 439-nm laser excitation before and during incremental photobleaching of YFP-labeled subunits, followed by extraction of true CFP and YFP image data using linear unmixing (see Materials and Methods). A scatterplot of CFP intensity in response to YFP photobleaching reveals FRET between the subunits in question (Fig. 7B) when the slope of the linear regression line is >0. This slope was used to calculate FRET efficiency values, which were expressed as bar graphs (Fig. 7C) or listed (Table 1). As shown qualitatively in Fig. 6, significant FRET occurred in all nAChR pentamer conditions. We noted a higher FRET E for α4C/β2Y than for β3Y with either β2C or α4C (Y, YFP; C, CFP). To assess the specificity of this measurement, we also measured FRET between β3 and a non-nAChR, CFP-labeled protein, mGAT1. GAT1 is also a multipass transmembrane protein with a CFP-tag at its C terminus, which faces the cytoplasm. This protein is mainly localized to the endoplasmic reticulum (data not shown). These two points are important, because it was critical for a specificity probe to have 1) the same membrane topology as our labeled nicotinic receptors, with respect to the attached fluorophore; and 2) the same subcellular localization such that they are capable of interacting with each other. In N2a cells expressing β3-YFP and mGAT1-CFP, we could not detect any FRET between these proteins (Table 1; Fig. 7C). In an even more rigorous test, we assessed FRET between β3-YFP and another Cys-loop receptor labeled in the M3-M4 loop, the CFP-labeled GluCl β subunit (Slimko et al., 2002; Nashmi et al., 2003). FRET between β3 and the GluCl β subunit was significantly smaller (FRET E = 6 ± 4%) than for α4 or β2 nAChR subunits (Table 1; Fig. 7C). Thus, our FRET results between β3 and other labeled nAChR subunits cannot be explained by random collision or interaction with unassembled subunits. Finally, we were interested in whether subtle changes in the location of the fluorophore within the β3 M3-M4 loop could influence its ability to undergo FRET with another subunit. FRET E decreases strongly with the distance between fluorophores. We reasoned that changes in the insertion point of YFP in β3, while holding the position of CFP in β2 constant, might alter FRET between these two subunits. To address this, we compared the FRET E between β2-CFP and two different β3-YFP constructs, β3-YFP^P379 and β3-YFP^G367, which have different insertion points for YFP within the M3-M4 loop. To our surprise, there was no change in the FRET E for these two subunits (Table 1; Fig. 7D).

We quantitatively measured FRET between α6 and α4/β2 subunits as well. We expressed either α6Yβ2C or α6Yα4Cβ2 in N2a cells to measure FRET (Fig. 8A). The latter receptor was studied because recent work indicates that nAChR receptors containing both α6 and α4 1) exist and are functional in mouse brain tissue (Salminen et al., 2007), and 2) are both necessary to form the nAChR subtype with the highest affinity for nicotine yet reported in a functional assay (Salminen et al., 2007). Acceptor photobleaching FRET experiments reveal robust CFP dequenching in response to YFP photobleach for both of these receptor subtypes, indicating FRET (Fig. 8B). Similar to β3-YFP^* receptors, we measured FRET E values for these two subtypes, and we found a FRET E of 36.0 ± 2.4% for α6Yβ2C and 21.9 ± 1.1% for α6Yα4Cβ2 (Table 1; Fig. 8C). We also assessed the specificity of our FRET measurements for α6 by measuring FRET between α6-YFP and mGAT1-CFP as described above for Fig. 7. Similar to β3 and mGAT1, we could record no significant FRET between α6 and mGAT1 (Table 1; Fig. 8C). FRET experiments between α6 and the GluCl β subunit, the most rigorous test conducted, yielded a small FRET signal (FRET E = 14 ± 2%) (Table 1; Fig. 8C). Because these subunits presumably do not form functional channels, there may be a small distortion of our α6 FRET signals that is due to partially assembled receptors. Because this signal is significantly smaller than for all other α6 combinations, FRET between subunits in pentameric receptors remains the most plausible explanation for the energy transfer we observe for α6. Finally, we studied FRET between β2-CFP and three α6-YFP constructs (α6-YFP^A405, α6-YFP^G387, and α6-YFP^G366) that differed only in their insertion point for YFP within the M3-M4 loop. Again, we were surprised to find no significant difference in FRET E between these three α6 constructs (Table 1; Fig. 8D).

Several results described above gave us confidence that our XFP-labeled β3 and α6 constructs were performing as expected. After confirming that these subunits assemble and traffic normally when expressed independently of each other, we used these constructs together to study α6β2β3 nAChRs. This receptor represents a modest population of the total striatal nAChR pool, and it contributes to nicotine-stimulated dopamine release (Salminen et al., 2007). α6β2β3 receptors, where one subunit is untagged and the remaining subunits are either YFP- or CFP-tagged (α6Yβ2Cβ3, α6Yβ2β3C, and α6β2Yβ3C), were expressed in N2a cells (Fig. 9A). We measured robust donor dequenching for all receptor subtypes (Fig. 9B), which was confirmed with FRET E measurements (Table 1; Fig. 9C). Thus, aside from α6 functional measurements, we conclude that XFP-labeled α6 and β3 subunits exhibit normal subcellular trafficking and assembly compared with our well characterized fluorescent α4 and β2 subunits.

α6 and β3 Subunit Stoichiometry Probed with FRET. Having established that fluorescently labeled α6 and β3 are functional (β3 only), have a reasonable subcellular localization pattern, and assemble into nicotinic receptor pentamers with other subunits, we used these tools to probe an important question facing the nicotinic receptor field: subunit stoichiometry. A variety of creative approaches have been taken to understand subunit stoichiometry, including immunoprecipitation (Zoli et al., 2002; Champtiaux et al., 2003), density gradient centrifugation (Kuryatov et al., 2000), molecular concatamers/linked subunits (Tapia et al., 2007), reporter mutations (Boorman et al., 2000), and mouse genetic approaches (Gotti et al., 2005a; Salminen et al., 2007). We now use FRET to address the problem of subunit stoichiometry because FRET occurs only when subunits are directly interacting, and often assembled, with one another.

We have previously shown that FRET efficiency correlates directly with functional receptor pentamers (Nashmi et al., 2003). To examine the number of α6 and β3 subunits in a nicotinic receptor pentamer, we first used FRET to examine the stoichiometry of a well studied receptor, namely, α4β2 receptors. It is widely accepted that α4 and β2 subunits assemble to form both high-sensitivity (HS) and low-sensitivity (LS) receptors. Cells often produce a mixture of these two receptors (Buisson and Bertrand, 2001; Nashmi et al., 2003), although they can be induced to express a pure population of one or the other (Nelson et al., 2003; Briggs et al., 2006). The subunit stoichiometry of HS receptors is postulated to be (α4)₂(β2)₃, whereas the LS receptors is thought to be (α4)₃(β2)₂ (Nelson et al., 2003). Regardless of the fraction of HS and LS receptors, we took advantage of the fact that all α4β2 receptors presumably contain two or more α4 and two or more β2 subunits. We reasoned that when cells express α4-YFP and α4-CFP along with β2 (Fig. 10A), a fraction of the receptors will contain both YFP- and CFP-labeled α4 subunits, and they will therefore be detectable by FRET. Confirming this hypothesis, we did detect modest dequenching of α4-CFP upon incremental α4-YFP photobleaching (Fig. 10B). FRET E for α4Yα4Cβ2 receptors was 22.2 ± 2.3% (Table 1; Fig. 10C). We also conducted a similar experiment with β2, and we found a modest FRET signal (FRET E = 16.3 ± 1.7%) between β2-YFP and β2-CFP within the same pentamer (Table 1; Fig. 10, B and C). We next used this assay to determine whether α6^* and β3^* receptors have one or more than one α6 or β3 subunit per pentamer. N2a cells expressing either α6Yα6Cβ2 or α4β2β3Yβ3C receptors were analyzed for FRET (Fig. 10A). We measured a strong FRET signal between α6-YFP and α6-CFP in donor dequenching (Fig. 10B), corresponding to a robust FRET E of 27.8 ± 1.7% (Table 1; Fig. 10C). Thus, these data are the first to directly demonstrate that α6^* receptors are capable of containing at least two α6 subunits, similar to other α subunits such as α3 and α4. In contrast to α6, β3 is thought to be an “ancillary subunit”, only able to incorporate into nAChRs with other α and β subunits (Groot-Kormelink et al., 1998). We could detect little or no FRET between β3-YFP and β3-CFP (FRET E = 2.6 ± 1.3%) (Table 1; Fig. 10, B and C). This was a specific result, because β3-YFP and β3-CFP were able to FRET with other subunits (Figs. 7 and 9), thus ruling out the notion that one of these subunits is not able to undergo FRET. These data are the first to directly demonstrate that receptors containing β3 subunits are only able to incorporate a single copy of this subunit. We interpret this to mean that β3 incorporates into the “accessory” position in a nAChR pentamer (Tumkosit et al., 2006), and it likely does not contribute to either of the two α:non-α interfaces that form the ligand-binding sites.

After confirming via FRET that β3 incorporates into nAChRs at a frequency of one subunit per pentamer, we used β3 coexpression to further probe the subunit stoichiometry of α4^* and α6^* receptors. We coexpresssed β3-WT with α4-YFP, α4-CFP, and β2 such that β3 was in excess. In this experiment, β3 is incorporated into α4-XFPβ2 receptors and will displace either an α4 or β2 subunit. There was a significant decline in FRET for cells expressing α4Yα4Cβ2β3 receptors versus those expressing α4Yα4Cβ2 (Table 1; Fig. 10, D and G). We interpret this result to mean that β3 incorporation has fixed the subunit stoichiometry of FRET-competent receptors to (α4Y)₁(α4C)₁(β2)₂(β3)₁ versus the following mixture of FRET-competent receptors without β3: (α4Y)₂- (α4C)₁(β2)₂,(α4Y)₁(α4C)₂(β2)₂ and (α4Y)₁(α4C)₁(β2)₃. A reduction in FRET for two XFP-labeled α4 subunits (YFP and CFP) versus three is reasonable and expected based on the work of others (Corry et al., 2005), and on our calculations that predict the relative FRET efficiencies in pentamers with XFP-labeled subunits (data not shown). Thus, β3 incorporation into nAChR pentamers likely displaces one subunit, and results in a decrease in α4 to α4 FRET for pentamers with a mixed subunit stoichiometry.

To determine whether α6^* receptors have a fixed or a mixed subunit stoichiometry, we coexpressed β3 in excess with α6-YFP, α6-CFP, and β2. If α6^* receptors only incorporate two α6 subunits, we expect to observe little or no change in FRET between α6-YFP and α6-CFP because β3 will only displace one unlabeled β2 subunit. However, if α6^* receptors exist as a mixture of (α6)₂(β2)₃ and (α6)₃(β2)₂ subtypes similar to α4^* receptors, we expect to observe a similar decline in FRET when β3 is present to induce only the (α6)₂(β2)₂β3 stoichiometry. The latter was the case. We noted a significant decline in the slope of the donor dequenching profile for α6Yα6C^* receptors when β3 was present (Table 1; Fig. 10E) and a decline in the FRET E for α6Yα6Cβ2β3 (21.7 ± 1.4%) versus α6Yα6Cβ2 (27.8 ± 1.7%) (Table 1; Fig. 10G). Thus, fluorescent α6^* receptors behave identically to α4^* receptors, and these results suggest that α6^* receptors are capable of forming either of two subunit stoichiometries: (α6)₂(β2)₃ and (α6)₃(β2)₂.

Several groups have reported the existence of α4α6^* receptors in brain tissue (Zoli et al., 2002; Salminen et al., 2007), and α4α6β2β3^* receptors (presumably α4₁α6₁β2₂β3₁) have high affinity for nicotine (Salminen et al., 2007). To learn about the subunit stoichiometry of α4α6^* receptors, we expressed α6-YFP and α4-CFP along with β2 subunits in N2a cells. We noted a modest FRET signal, indicating that these subunits are present in some of the same nicotinic receptor pentamers (Fig. 8, B and C; Fig. 10F). In contrast to our results with α6β2β3 and α4β2β3 receptors, there was no difference between cells transfected with α6Y/α4C/β2 and α6Y/α4C/β2/β3 subunits (Fig. 10, F and G). This shows that addition of excess β3 subunits did not reduce FRET between α6Y and α4C. Thus, this experiment is not informative regarding α4α6β2β3 receptors in N2a cells. For instance, the α:β subunit stoichiometry as measured by FRET may not change in the presence of β3.

TIRF Revealed α4^*, α6^*, and β3^* Receptor Plasma Membrane Localization. Having confirmed that fluorescent α6 and β3 subunits assemble to form nicotinic receptor pentamers, we probed the plasma membrane localization of nAChRs containing these subunits using TIRF microscopy. For nicotinic receptor subunits fused to fluorescent proteins, TIRF illumination selectively excites only receptors at or very close to the plasma membrane. We imaged live N2a cells expressing α4β2β3-YFP, α6-YFPβ2, and α4-YFPβ2. In epifluorescence mode (Fig. 11, epi), these receptors exhibited an intracellular, endoplasmic reticulum-like localization identical to our confocal imaging data in Fig. 4. In TIRF mode, however, we noted robust plasma membrane fluorescence for all receptor combinations. We were surprised to find α4Yβ2 receptors to be localized to distinct, filamentous structures protruding from the cell body (Fig. 11A, arrow). This specific filamentous pattern was seen for >90% of the plasma membrane fluorescence. These structures were reminiscent of filopodia, which are actin-dependent plasma membrane protrusions. To test whether these structures contain actin, a hallmark of filopodia, we imaged cells stained with rhodamine-phalloidin, a marker of polymerized actin. We noted distinct, actin-containing protrusions (Fig. 11B, arrow). These structures were actin-dependent, because they were destroyed by treatment with latrunculin B, an actin-disrupting agent (Fig. 11B, right). These data indicate that, in N2a cells, α4Yβ2 nicotinic receptors are localized to membrane protrusions that strongly resemble filopodia.

Similar to α4Yβ2, we imaged live N2a cells, in TIRF mode, expressing either α4β2β3Y or α6Yβ2. To our surprise, we noted a very different localization pattern compared with α4Yβ2. β3^* and α6^* receptors were well expressed on the plasma membrane, but there was no evidence of membrane protrusion or filopodia localization for these receptors. Rather, these proteins exhibited a punctate, lattice-like localization pattern on the plasma membrane (Fig. 11, C and D). This pattern was consistently seen in other cells types such as HEK293 (data not shown), and it suggests that β3^* or α6^* receptors cluster in microdomains distinct from α4β2 receptors. Alternatively, some of these puncta could be clusters of assembled receptors adjacent to the plasma membrane within the 100-nm evanescent wave. We also recorded movies to monitor plasma membrane nAChRs, and we noted that although they exhibited localized, stochastic movements, most of these receptor clusters did not travel or translocate to any significant degree (data not shown). This localization pattern resembles that of the soluble N-ethylmaleimide-sensitive factor attachment protein receptor protein syntaxin1, which was localized to distinct granules or microdomains in the plasma membrane when observed in TIRF (Ohara-Imaizumi et al., 2004). Because soluble N-ethylmaleimide-sensitive factor attachment protein receptor proteins are important regulators of ion channel subcellular trafficking and function in neuronal soma and synaptic terminals (Bezprozvanny et al., 1995), we compared the plasma membrane localization pattern of YFP-syntaxin1A with β3-YFP^* and α6-YFP^* receptors in N2a cells. We observed a plasma membrane distribution pattern for syntaxin1A that was very similar to β3 and α6 subunits; syntaxin1A was also localized to distinct clusters adjacent to the plasma membrane or microdomains on the plasma membrane (Fig. 11E).

It is possible that the different plasma membrane localization pattern observed for α4Yβ2 versus α4β2β3Y reflects the localization pattern of functional versus nonfunctional nicotinic receptors, respectively. To address this, we used whole-cell patch-clamp electrophysiology to record voltage-clamped responses from functional, fluorescent nAChRs expressed in N2a cells. Because WT (Broadbent et al., 2006) or YFP-tagged β3 subunits (Fig. 2B) significantly attenuated nAChR responses, we used β3-YFP^V13S subunits to reverse this attenuation. We reasoned that, if coexpressed and coassembled with α4β2 receptors, β3-YFP^V13S subunits should 1) induce the high-sensitivity (α4)₂(β2)₂(β3)₁ subunit stoichiometry similar to previous work (Broadbent et al., 2006); and 2) lower the EC₅₀ for activation of this high-sensitivity form by approximately 1 order of magnitude (Fig. 3B). When voltage-clamped N2a cells expressing α4Yβ2 receptors were stimulated with 1 μM ACh, a dose that induces minimal (20-30 pA) responses in our previous work with HEK293 cells (Nashmi et al., 2003), we observed an identical phenotype (Fig. 11F). Responses to 300 μM ACh were robust (200-400 pA), indicating significant plasma membrane expression of these receptors (data not shown). However, cells expressing α4β2β3Y^V13S receptors exhibited robust responses to 1 μM ACh (Fig. 11F), which were significantly larger than the response size for α4β2 receptors (Fig. 11G). This is the expected result if β3-YFP^V13S subunits are incorporated into functional nAChRs in N2a cells, and it is consistent with our X. laevis oocyte experiments (Figs. 2 and 3), and with the work of others (Broadbent et al., 2006). These data confirm that both α4Yβ2 and α4β2β3Y receptors are functional in N2a cells and that the observed plasma membrane localization pattern for functional α4Yβ2 and α4β2β3Y receptors is significantly different.

Discussion

Subcellular Localization of Nicotinic ACh Receptors. α4β2 receptors were localized intracellularly in our previous studies in HEK293 cells and primary midbrain neurons (Nashmi et al., 2003), although enough receptors are expressed on the cell surface to record responses using electrophysiology. Furthermore, knockin mice with YFP-labeled α4 subunits show uniform intracellular and plasma membrane localization of α4^* receptors (Nashmi et al., 2007). Other investigators have found a similar localization pattern for α4β2 (Xu et al., 2006), α3β4 (Grailhe et al., 2004), α7 (Xu et al., 2006), and 5-hydroxytryptamine_3A (Grailhe et al., 2004) receptors. In light of these studies, it is not surprising that we found fully assembled β3^* and α6^* receptors in intracellular stores in N2a cells. This suggests that neurons produce more assembled nAChRs than they can use at any particular time and that they may require the ability to rapidly change either their total number or specific stoichiometry of receptor subtypes on the plasma membrane in response to different extracellular signals. The specific phenomenon of up- or down-regulation of nicotinic receptors occurs during chronic nicotine exposure (Marks et al., 1983; Nashmi et al., 2007) and in other neurological disorders including autism and Alzheimer's disease (for review, see Graham et al., 2002).

α4β2 receptors were localized to actin-dependent membrane protrusions akin to filopodia, whereas β3^* and α6^* receptors were not. Filopodia are critical sensory components of growth cones, influencing growth cone orientation and turning toward extracellular cues. Nicotinic receptors are required for growth cone orientation in some neuronal types (Zheng et al., 1994). Interestingly, α4 and β2 subunits are highly expressed during embryogenesis (Azam et al., 2007), suggesting a role for these subunits in neuronal development and in adult function. α6 and β3 subunits, in contrast, are not expressed at appreciable levels until after birth (Azam et al., 2007). We speculate that the differences we observe for α4β2 versus β3 or α6 receptors on the plasma membrane could reflect their involvement (or lack thereof) in neuronal development.

α6 Functional Expression. Although it remains unclear why it is very difficult to record functional responses from α6^* receptors, this study advances our knowledge of this problem. In 20 different attempts, using several expression/assay systems, we could record no α6 functional responses (Supplemental Data; Table 1). This could be a result of many different problems, such as α6 mRNA stability, α6 protein production, proper assembly of α6^* receptors, intracellular trafficking, or plasma membrane delivery. We (herein) and others have demonstrated that cells do not have an apparent problem synthesizing α6 subunits (Kuryatov et al., 2000; Tumkosit et al., 2006), and either partial or full assembly of α6^* receptors occurs in mammalian tissue culture cells and X. laevis oocytes (Kuryatov et al., 2000; Tumkosit et al., 2006). We found robust ectopic expression of α6-YFP subunits in tissue culture cells and in primary neurons. Furthermore, based on our FRET measurements, α6 subunits are fully capable of assembling with α4, β2, and β3 subunits in a manner indistinguishable from that of α4 and β2. These are the presumptive subunits necessary for expression of α6^* receptors in vivo (Salminen et al., 2007). Assembled α6^* receptors are also localized identically to α4^* receptors in our experiments. Finally, and most surprisingly, we ruled out the possibility that α6^* receptors are not delivered to the plasma membrane. Plasma membrane localization for α6^* receptors was identical to that of assembled, functional β3^* receptors. From these data, we conclude that, although fully assembled and partially localized on the plasma membrane, α6^* receptors do not yield responses in standard functional assays. This information should facilitate the design of new experimental approaches to develop robust, reproducible reconstitution of α6 function in vitro.

Because we found no difference in FRET between α4α6β2 receptors ± β3, the experiments probing the stoichiometry of α4α6β2β3 receptors are uninformative. It is possible that there are no assembled α4α6β2β3 receptors in N2a cells. We suggest that catecholaminergic or retinal ganglion cells, which are those cells in vivo that produce high levels of functional α6^* receptors (Léna et al., 1999; Whiteaker et al., 2000; Champtiaux et al., 2002, 2003; Zoli et al., 2002; Gotti et al., 2005b), express a unique protein or factor that is essential for proper function of these receptors. It is also possible that these cells are specially suited to traffic α6^* and/or β3^* receptors to distal axons/presynaptic terminals. Our results probing α6 and β3 axonal targeting (Fig. 5B) may be negative as a result of differences in the cell trafficking machinery in hippocampal (α6-negative) versus catecholaminergic (α6-positive) neurons, and they highlight the need for a more detailed study of α6/β3 axonal targeting in α6-positive neurons. Perhaps there is an α6-associated protein similar to other nicotinic receptor-associated proteins, such as lynx1 (Miwa et al., 1999), which remains to be characterized.

α6^* and β3^* Receptor Assembly and Subunit Stoichiometry. We previously demonstrated that FRET between XFP-labeled nicotinic receptor subunits not only reveals proximity between subunits but that increased FRET efficiency correlates with increased assembled, functional receptors (Nashmi et al., 2003). Because we cannot measure functional responses from α6^* receptors, we must draw conclusions about their behavior inferentially by comparing it to α4β2 receptors, which are functionally expressed. Using these criteria, we conclude that α6 subunits assemble with α4, β2, and β3 subunits. Furthermore, we conducted several specificity controls (FRET with GAT1C and GluCl β) that revealed that FRET between α6 or β3 and other nAChR subunits is robust and likely explained by pentameric assembly. These experiments also reveal that a subpopulation of α6 subunits may be contained in partially assembled receptors. It may be this feature that precludes routine measurement of functional responses, further supporting the notion of a special factor in vivo that promotes α6^* nAChR assembly and function.

We noted a slightly higher FRET efficiency for α6^* versus α4^* receptors in all assays reported herein. Because FRET E depends on distance, we speculate that this is largely due to the smaller M3-M4 intracellular loop of α6(∼136 residues) versus α4(∼270 residues). When the same XFP-labeled β2 construct is expressed with α4 versus α6, the relative distance between fluorophores, and therefore the efficiency of FRET, will be different. Although the lack of structural information about these M3-M4 loops precludes us from making any quantitative predictions or correlations with our observed FRET E values, we assert that, at a qualitative level, the differences in FRET E for α4 versus α6 are likely explained by the differences in M3-M4 loop length.

In this study, we use FRET to describe α6 and β3 subunit stoichiometry in assembled pentamers containing these subunits. Based on our data, only one β3 subunit is able to incorporate into a nicotinic receptor pentamer with other α and β subunits. It is not clear whether β3 does so because it lacks residues required for formation of an α:β ligand-binding interface, or whether it is due to specific residues in the transmembrane segments or intracellular loops. What is clear is that β3 is able to displace one subunit in a pentamer, which affords the ability to alter the subunit stoichiometry of receptors containing this subunit. Although this has been assumed based on indirect experiments (Broadbent et al., 2006), the present study is the first to directly elucidate the stoichiometry of β3^* receptors.

Based on our data, α6^* receptors, like α4^*, are stoichiometrically heterogeneous. Previous reports using immunoprecipitation or genetic techniques have identified the specific subunits coassembled with α6, but not their stoichiometry (Zoli et al., 2002; Salminen et al., 2007). For example, Salminen et al. (2007) demonstrated the existence of native α6β2, α6β2β3, and α4α6β2β3 subtypes, among others. But for α6β2 receptors, how many α6 and β2 subunits are present in a given pentamer? Our FRET results suggest that a mixture of stoichiometries exist for α6^* receptors. We interpret the β3-induced decline in FRET between α6Y and α6C in α6β2 receptors to mean, at least in part, that β3 is displacing a third α6 subunit and stabilizing a stoichiometry of (α6)₂(β2)₂(β3)₁. The alternative, that α6^* receptors adopt a strict (α6)₂(βX)₃ stoichiometry, is less likely. Such a scenario would require that β3 is able to significantly reduce FRET between α6Y and α6C without changing α6 stoichiometry. It is more reasonable to assume that α6 is behaving similar to α4, whose stoichiometry is varied and can be altered by β3 coexpression (Broadbent et al., 2006). α6 and β3 subunits are present in α4α6β2β3 receptors in striatum and nucleus accumbens, which have the highest affinity for nicotine of any nicotinic subtype yet reported (Salminen et al., 2007). Their localization on dopaminergic nerve terminals coupled with this high affinity for nicotine ensures that they are among the first nicotinic subtypes activated during a smoking-induced bolus of nicotine. These and other high-affinity receptors are important targets for smoking cessation and Parkinson's disease medications, so understanding their subunit stoichiometry is important for the rational design of small molecule modulators of nAChR function.