 |
The Journal of Neuroscience, July 20, 2005, 25(29):6911-6920; doi:10.1523/JNEUROSCI.0561-05.2005
 Previous Article | Next Article 
Cellular/Molecular
An Angstrom Scale Interaction between Plasma Membrane ATP-Gated P2X2 and  4 2 Nicotinic Channels Measured with Fluorescence Resonance Energy Transfer and Total Internal Reflection Fluorescence Microscopy
Baljit S. Khakh,1
James A. Fisher,1
Raad Nashmi,2
David N. Bowser,1 and
Henry A. Lester2
1Medical Research Council Laboratory of Molecular Biology, Cambridge CB2 2QH, United Kingdom, and 2Division of Biology, California Institute of Technology, Pasadena, California 91125
 |
Abstract
|
|---|
Structurally distinct nicotinic and P2X channels interact functionally, such that coactivation results in cross-inhibition of one or both channel types. It is hypothesized, but not yet proven, that nicotinic and P2X channels interact at the plasma membrane. Here, we show that plasma membrane  4 2 nicotinic and P2X2 channels form a molecular scale partnership and also influence each other when coactivated, resulting in nonadditive cross-inhibitory responses. Total internal reflection fluorescence and fluorescence resonance energy transfer microscopy between fluorescently labeled P2X2 and 42 nicotinic channels demonstrated close spatial arrangement of the channels in human embryonic kidney cells and in hippocampal neuron membranes. The data suggest that P2X2 and 42 channels may form a dimer, with the channels  80 Å apart. The measurements also show that P2X2 subunits interact specifically and robustly with the 2 subunits in 42 channels. The data provide direct evidence for the close spatial apposition of full-length P2X2 and 42 channels within 100 nm of the plasma membrane of living cells.
Key words: channel; cholinergic; purinergic; acetylcholine receptor; ACh; fluorescence microscopy; P2X
|
Introduction
|
|---|
Transmitter-gated ion channels are found in neurons and synapses, where they bind neurotransmitter and convert this presynaptic chemical signal into a postsynaptic signaling event (Hille, 2001 ). Mammalian transmitter-gated ion channels consist of three families: Cys-loop channels for ACh, GABA, glycine, and serotonin; glutamate-gated channels (kainate, AMPA, and NMDA); and the ATP-gated P2X channels (Green et al., 1998; Khakh, 2001; Lester et al., 2004), of which seven subunits are known (P2X1-P2X7). There is little sequence, topological or structural similarity between members of the Cys-loop, glutamate-gated, and P2X channel family members (North, 1996a,b), leading to the assumption that they act independently in neurons and synapses.
Motivated by pioneering work on neurons showing that P2X and nicotinic channels may not act independently (Nakazawa, 1994; Barajas-Lopez et al., 1998; Searl et al., 1998; Zhou and Galligan, 1998), we suggested that neuronal nicotinic and P2X channels undergo cross-inhibitory functional interactions because of conformational coupling akin to conformational spread (Khakh et al., 2000). Conformational spread has been proposed as a general mechanism for signal integration by cells (Bray and Duke, 2004): conformational changes in one protein are relayed to its neighbors because of close spatial interactions. Cross-inhibition also occurs for P2X, nicotinic, GABA, and 5-HT3 subunits in heterologous expression studies as well as for natively expressed channels in neurons and synapses (Barajas-Lopez et al., 1998; Searl and Silinsky, 1998; Searl et al., 1998; Zhou and Galligan, 1998; Khakh et al., 2000; Sokolova et al., 2001b; Barajas-Lopez et al., 2002; Boue-Grabot et al., 2003, 2004a,b). One functional measure of cross-inhibition between ATP-gated P2X and ACh-gated nicotinic channels is that coapplication of ATP and ACh to cells expressing these channels evokes responses smaller than the sum of the individual ATP and ACh-evoked currents (Searl and Silinsky, 1998). This lack of summation implies that, when P2X and nicotinic channels are simultaneously activated, fewer channels open than expected if one assumes that the two channel populations function independently. The underlying mechanisms responsible for cross-inhibition between P2X2 and nicotinic channels remain incompletely understood.
If conformational spread is the cause of cross-inhibition between P2X2 and nicotinic channels, a close spatial arrangement between the channels in the plasma membrane is expected (Bray and Duke, 2004) but has remained unproven. We took advantage of recent data demonstrating that both P2X2 and 42 channels can be innocuously labeled with GFP-based fluorophores in their cytosolic domains (Khakh et al., 2001; Nashmi et al., 2003; Fisher et al., 2004). We first asked whether channels formed by these brain-expressed subunits (P2X2, 4, 2) (Kanjhan et al., 1999; Labarca et al., 2001; Fonck et al., 2003) show cross-inhibition similar to that demonstrated for channels found in the peripheral nervous system, and we then used fluorescence resonance energy transfer (FRET) microscopy to probe interactions between P2X2 and 42 channels in the plasma membrane of single cells.
|
Materials and Methods
|
|---|
Molecular biology. cDNAs for wild-type (wt) P2X2, 4, and 2 subunits, as well as cyan fluorescent protein (CFP)- and yellow fluorescent protein (YFP)-labeled P2X2 (P2X2-C, P2X2-Y), 4 subunits (4-Y, 4-C), and 2 subunits (2-C, 2-Y) were used (where C is CFP, and Y is YFP). These were all available from previous work (Nashmi et al., 2003; Fisher et al., 2004). For the generation of P2X6-C and P2X6-Y, the starting vectors were P2X2-C and P2X2-Y. In these vectors, HindIII (5') and XhoI (3') digestion yielded P2X2. By PCR (PCR Core kit; Qiagen, Crawley, UK) the P2X6 coding region was amplified (no stop codon) with noncoding 5' HindIII and in frame 3' XhoI sites, using plasmids encoding P2X6 as template (Khakh et al., 1999). These PCR products were cloned into HindIII and XhoI sites of P2X2-C and P2X2-Y to generate P2X6-C and P2X6-Y, respectively. All constructs were sequenced (Lark Technologies, Saffron Walden, UK or MRC Gene Service, Cambridge, UK).
Human embryonic kidney cell culture and transfection. Human embryonic kidney 293 (HEK 293) cells (obtained from American Type Culture Collection via LGC Promochem, Teddington, UK) were maintained in 75 cm2 cell culture flasks (Corning; Fisher Scientific, Loughborough, UK) in DMEM/Ham's F-12 media with Glutamax (Invitrogen, Paisley, UK) supplemented with 10% fetal bovine serum and penicillin/streptomycin. Cells were grown in a humidified atmosphere of 95% air/5% CO2 at 37°C in a cell culture incubator. The cells were split 1 in 10 when confluence reached 60-90%, which was generally every 3-4 d. Cells were prepared for transfection by plating onto six-well plates at the time of splitting, 3-4 d before transfection. They were transfected at 60-90% confluence. For transient expression in HEK 293 cells, we used 0.5-1 µg of plasmid cDNA and the Effectene transfection reagent (Qiagen) for each well of a six-well plate. The manufacturer's instructions were followed, with 4 µl of enhancer and 10 µl of Effectene used for each transfection. The transfection efficiency was 40-60%. For the functional measurements of cross-inhibition between P2X2 and 42 channels, we used a previously available stable cell line expressing 42 channels (Egan and Khakh, 2004), which has been described in detail previously (Chavez-Noriega et al., 2000), and transfected subunit cDNAs using procedures described above. All recordings were made 1-2 d after transfection, as described below. For FRET experiments, we used wild-type HEK cells and transfected cDNAs for wt 4, 4-Y, 4-C, wt 2, 2-Y, 2-C, P2X2-Y, and P2X2-C subunits in all combinations, as appropriate (see Results and Table 1). Using this approach, we tested all possible partnerships between 42 and P2X2 channels with FRET (see Results) in a cell type otherwise devoid of 42 or P2X2 channels, thus avoiding the complication of nontagged subunits to FRET.
View this table:
[in this window]
[in a new window]
|
Table 1. Average data for FRET efficiency and a measure of channels within 100 Å of each other when expressed in HEK cells and hippocampal neurons
| |
Hippocampal neuron culture and transfection. Typically five rat pups at embryonic day 19 (E19) were used for hippocampal cultures. The mother was killed following Home Office (London, UK) and local procedures, and the embryos were removed to ice-cold dissection medium comprising Earle's buffered salt solution (with no Ca2+ or Mg2+ and no phenol red; Invitrogen), HEPES (10 mM, pH 7.4; Sigma, Poole, UK), and penicillin/streptomycin (1:100 dilution; Invitrogen). Hippocampi were dissected in Petri dishes filled with ice-cold medium. The dissected hippocampi (in medium, on ice) were cut in half, transferred to a 15 ml tube with 10 U/ml papain PAP2 powder [Worthington Biochemical (Lakewood, NJ) PAPAIN-022], and incubated for 15 min at 37°C. When the pieces of tissue had settled, digestion solution was removed and 10 ml of cold culture medium was added. This step was repeated, and the pieces were triturated five to six times in a 2 ml volume of ice-cold culture medium with paired flame-polished pipettes of progressively smaller bores. A hemacytometer was used to quantify yield (typically 2.5 million/ml), and 200,000 neurons were used for plating onto each coverslip. The coverslips were precoated with poly-D-lysine (50 µg/ml; Sigma) and then overnight with 150 µl of 20 µg/ml laminin (Sigma) in sterile dissection medium. Three hours after plating, the cells were fed with 2 ml of prewarmed culture medium and fed again 1 d after plating by substituting 1 ml of fresh prewarmed medium for 1 ml of existing medium. Cells were then fed every 4-5 d and used for experiments after 14 d in vitro. The culture medium comprised MEM with no phenol red (Invitrogen), glucose (20 mM), penicillin/streptomycin (Invitrogen), Na pyruvate (2 mM; Sigma), HEPES (25 mM; Sigma), N2 supplement (1:100 dilution; Invitrogen), and heat-inactivated horse serum (10%; Invitrogen). Effectene transfection reagent was used to transfect 0.5-1.0 µg of the appropriate cDNA per well of a six-well plate, which contained 3-4 coverslips. The instructions of the manufacturer were followed, except that 8 µl of enhancer and 24 µl of Effectene were used per 1 µg of cDNA. Experiments were performed 1-2 d after transfection, when transfection was successful typically in 1-2% of cells.
Midbrain mesencephalic dopaminergic neuron cultures and transfection. Ventral mesencephalic neuronal cultures were prepared as described previously (Nashmi et al., 2003). Brain tissue containing primary mesencephalic neurons of the ventral tegmental area from E14 BDF1 mouse embryos (Shimoda et al., 1992) was incubated in plating medium (Neurobasal, 2% B27, 0.5 mM Glutamax, and 5% horse serum) with 1 mg/ml papain at 37°C for 20 min. The digested tissue was mechanically dispersed, laid on 4% bovine serum albumin (Fraction V; Sigma, St. Louis, MO) in PBS, centrifuged for 6 min at 300 x g and resuspended in plating medium. Cells (150,000) were plated onto 35 mm culture dishes with 10-mm-diameter glass coverslip bottoms (MatTek, Ashland, MA) that were coated with poly-DL-lysine (catalog #P9011; 1 mg/ml; Sigma, St. Louis, MO) dissolved in sterile water. To transfect one dish of neurons, 1 µg of each cDNA was mixed with 60 µg of Nupherin-neuron (Biomol Research Laboratories, Plymouth Meeting, PA) in 400 µl of Neurobasal medium without phenol red. Lipofectamine 2000 (3 µl; Invitrogen, Carlsbad, CA) was mixed in 400 µl of Neurobasal media. The two solutions were incubated for 15 min at room temperature, combined, and incubated for another 45 min at room temperature. The medium of the mesencephalic cultures was replaced with plating medium without added serum, and the transfection medium was added. The culture dishes were centrifuged at 100 x g for 5 min on a swinging bucket rotor and incubated for 30 min at 37°C. Then, the transfection mixture was removed and replaced with the original plating medium. Experiments were performed 2 d after transfection, when transfection was typically successful in 1-2% of cells.
Electrophysiology. HEK 293 cells were used for recordings 24-48 h after transfection, as described previously (Fisher et al., 2004). The cells were gently mechanically dispersed and plated onto glass coverslips 2-12 h before use. We included this step to ensure that the cells became spherical and therefore adequately space clamped during recordings (Egan et al., 1998). The extracellular recording solution comprised (in mM) 147 NaCl, 2 KCl, 1 MgCl2, 1 CaCl2, 10 HEPES, and 10 glucose, pH 7.4, and the pipette solution comprised (in mM) 154 KCl (or CsCl), 11 EGTA, and 10 HEPES. Whole-cell voltage-clamp recordings were made with 5 M borosilicate glass electrodes (World Precision Instruments, Stevenage, UK), using an Axopatch 1D amplifier controlled by a computer running pClamp 8.1 software via a Digidata 1322 interface (Molecular Devices, Union City, CA). Data were filtered at 0.5-2 kHz and digitized at 1-5 kHz. Agonists were applied for 0.2-4 s at 2-3 min intervals using triple-barreled theta-glass (Egan and Khakh, 2004) and a rapid solution changer system (Perfusion Fast-Step System SF-77; Warner Instruments, Hamden, CT). Voltage ramps (0.5 s duration; -120 to +60 mV) were applied at the peak response to each agonist.
Confocal FRET microscopy. Confocal FRET imaging of midbrain neurons has been described previously (Nashmi et al., 2003). Cultured and transfected mesencephalic neurons were visualized at room temperature in extracellular buffer comprising (in mM) 150 NaCl, 4 KCl, 2 CaCl2, 2 MgCl2, 10 HEPES, 10 D-glucose, and 1 ascorbic acid, pH 7.4. Ascorbic acid was included to minimize damage to dendrites during photodestruction. A series of  -stack x-y images were collected from an upright LSM 510 Meta laser-scanning confocal microscope using an Achroplan IR 63x 0.9 numerical aperture (NA) water-immersion objective (Zeiss, Oberkochen, Germany). Images were collected over a spectral range between 462.9 and 602 nm with bandwidths of 10.7 nm after excitation of CFP with the 458 nm laser line of an argon laser. Pinhole was 1.32 Airy units, and the Z resolution was 2.0 µm. Images were collected at a 12-bit intensity resolution over 512 x 512 pixels at a pixel dwell time of 12.8 µs. FRET was recorded by examining the dequenching of CFP after incremental photodestruction of YFP by the 514 nm argon laser line. Stacks were acquired at various time points before and after photodestruction of YFP. These time points included 0, 0.13, 0.26, 0.52, 1.05, 1.57, 2.09, and 6.81 min. We separated the fluorescence contribution of each fluorescent protein (CFP or YFP) using a linear unmixing algorithm based on the spectral signatures of YFP and CFP created from reference -stack images of cells expressing either soluble YFP or soluble CFP, respectively. These unmixed YFP and CFP images (Dickinson et al., 2001; Lansford et al., 2001) were analyzed for FRET efficiency. Background fluorescence for both CFP and YFP, determined as the mean fluorescence intensity from an area containing cells that did not express the constructs, was subtracted from the overall intensity. Mean fluorescence intensities of CFP and YFP were determined by tracing the outer perimeter of the cell, the neuronal soma, or dendrite with a region of interest. However, in the case of FRET calculated for cell membrane surface receptors, mean fluorescence intensity was calculated for only the outer 2 µm shell of fluorescence and did not include the cell interior.
Epifluorescence FRET microscopy. We used an Olympus (Southall, UK) BX50 microscope equipped with a Peltier-cooled (-15°C) half-inch interline transfer chip Imago CCD camera (640 x 480 pixels; each pixel, 9.9 x 9.9 µm), epifluorescence condenser (Olympus IX two-port flash/imaging condenser), control unit (containing ISC and PDC boards), and the Polychrome IV monochromator (all from TILL Photonics, Gräfelting, Germany). The light from the monochromator was led to the microscope through a quartz fiber light guide. The hardware was controlled by a personal computer, an appropriate frame grabber (TILL Photonics), and macros driven by TILLvisION version 3.3 software. The cells were viewed with a 40x water-immersion objective lens with a numerical aperture of 0.8 (Olympus). We used the following filters for acquiring CFP or YFP images (in nm; all from Glen Spectra, Stanmore, UK): CFP (dichroic, 455DRLP; emitter, 480AF30) and YFP (dichroic, 525DRLP; emitter, 545AF35). To photodestroy the YFP, we used 525 nm light from the monochromator, and a 525DRLP dichroic. FRET was determined as described below.
Total internal reflection fluorescence microscopy and FRET. Briefly, we used an Olympus IX70 microscope equipped with a Princeton Instruments (Trenton, NJ) cooled I-PentaMAX camera with High Blue Gen III Intensifier (Roper Scientific, Trenton, NJ). The control of excitation and image acquisition was achieved using MetaMorph software and drivers (Universal Imaging, Marlow, UK) and shutters, filter wheels, and Proscan II control box (Prior Scientific, Cambridge, UK). The beams of 488/515 nm Ar (150 mW) and 442 nm He-Cd (12.5 mW) lasers (Melles Griot, Carlsbad, CA) at <5% power (constant for all experiments reported) were combined and controlled with an IX2-COMB (Olympus), Uniblitz shutters (Prior Scientific) and acousto-optical tunable filter and controller (AA Optoelectronics, Les Chevreuse, France) and fed into a fiber optic (FV5-FUR; Olympus) for entry into the total internal reflection fluorescence microscopy (TIRF) condenser (IX2-RFAEVA-2; Olympus). Cells were plated onto glass-bottom Petri dishes (170 µm thick; Willco Wells, Amsterdam, The Netherlands) 24-48 h before imaging and were viewed with a 60x oil-immersion objective lens, NA 1.45 (Olympus). FRET was determined with donor dequenching. During donor dequenching experiments on the TIRF rig, we noted that significant photodestruction of YFP occurred within 10 s, versus many minutes for epifluorescence imaging (Fisher et al., 2004). This difference, likely reflecting the use of different microscopes (IX70 vs BX50), objective lenses (60x, NA 1.45 vs 40x, NA 0.8), and light sources (laser vs xenon lamp), implied that, for TIRF experiments, we would obtain only a single time point for donor dequenching (see below, Data analysis and Results). The camera gain was adjusted for maximum signal-to-noise for each cell and kept constant for all image acquisitions of a particular cell.
Data analysis. For donor dequenching experiments, the FRET efficiency (e) was calculated as follows:
 | (1) |
where IC-before is the donor fluorescence intensity before photodestruction, and IC-after is the intensity after photodestruction. The photodestruction of the YFP proceeds with a rate equivalent to the dequenching of the donor, and plotting the photorecovery versus photodestruction yields a straight line (see Fig. 1). We used such linear plots and extrapolated to 100% acceptor photodestruction to calculate the maximum donor dequenching for epifluorescence microscopy (see Fig. 1): e is given by the y-axis intercept. For TIRF microscopy, we estimated FRET e from a single time point when the acceptor was 85% destroyed and extrapolated to 100% destruction by constraining the linear regression to pass through the origin. This was the most appropriate tactic, because it was not possible to track time-dependent photodestruction with TIRF: on one hand, as noted, the bleaching occurred too rapidly with laser illumination (<0.5-1 min). In contrast, with epifluorescence illumination, bleaching typically took 30 min, allowing multiple evenly spaced time points to be collected accurately (see Results) (Fisher et al., 2004); however, because during TIRF only the plasma membrane within 100 nm of the glass-cell interface, the footprint, is illuminated (Steyer and Almers, 2001), it is necessary to destroy the acceptor fluorophore rapidly within this region and to measure donor dequenching before channels that are not within the evanescent field of illumination (and therefore neither destroyed or dequenched) can either diffuse or be inserted into the foot-print. Thus, the fact that for TIRF microscopy maximal photodestruction occurred rapidly was a fortuitous feature that enabled us to determine FRET from a single time point. In Results, we present a detailed comparison of FRET values determined for a single time point versus those calculated from multiple time points.
To approximate distances, we used the Förster equation:
 | (2) |
where
 | (3) |
e is the experimentally determined FRET efficiency, and R is the distance between donor and acceptor fluorophores (Michalet et al., 2003; Bunt and Wouters, 2004). R0 includes terms for the donor quantum efficiency ( D), the solvent refractive index (n), overlap of the donor emission and acceptor absorption spectra (JDA), and the orientation factor ( 2). R0 is the distance at which e is 50% for any given fluorophore pair. In the case of CFP and YFP, this is 49.2 Å (Patterson et al., 2000) or 50.4 Å (Tsien, 1998). We have no good reason to choose between these values and have taken the average and rounded to the nearest whole number to give a value of 50 Å (5 nm). Importantly, R0 contains a term for the fluorophore dipole orientation factor (2), which, at 2/3, assumes random tumbling of the fluorophores (Bunt and Wouters, 2004). This means that R0 and the distance measurements reported here assume that the fluorophores adopt all spherically random orientations. Given that we do not know the relative orientation of the fluorophores on P2X2 and nicotinic channels, assuming a 2 value of 2/3 seems parsimonious and appropriate (Michalet et al., 2003; Bunt and Wouters, 2004) and is supported by empirical data for CFP and YFP fused to the cytosolic C termini of inwardly rectifying K+ channels (Riven et al., 2003), which have a membrane topology similar to P2X channels (North, 1996a). When interpreting FRET efficiency measurements in terms of distance, it is important to consider that, in the case of CFP and YFP, the fluorophore is buried in the center of a -can 30 Å in width and 45 Å in length (Tsien, 1998). An additional consideration is that we do not know the absolute ratio of donors to acceptors during FRET at the single molecule level (see Discussion).
For the FRET experiments, we chose regions of interest (ROIs) post hoc (Fisher et al., 2004). Our criteria were that ROIs (1) were bright in both CFP and YFP images, (2) were located at the edge, presumably plasma membrane of the cell, (3) did not change or drift in intensity during the control period, and (4) remained in focus during the entire experiment.
For the electrophysiological experiments, we used analysis that has been described previously (Khakh et al., 2000). Briefly, we measured the currents activated by ATP (IATP), ACh (IACh), and ATP and ACh together (IATP/ACh). From these, we calculated the predicted current (IPredicted) and the occluded current (ILost) on a cell-by-cell basis, where
 | (4) |
 | (5) |
Thus, the percentage contribution of any one component such as IACh to the predicted current can be calculated as (IACh/IPredicted) x 100. Rectification index was calculated as the ratio of macroscopic slope conductances at voltages ±40 mV from the reversal potential. Reversal potentials were measured using two- or three-order polynomial fits to the current-voltage relationships and were taken as the values at which the fitted line first intersected the x-axis (see Fig. 1C, bottom). Electrophysiological analysis was performed with Clampfit (Molecular Devices) or Origin 6.1 (OriginLab, Northampton, MA), and all statistical tests were run in GraphPad Instat 3.0 (GraphPad Software, San Diego, CA). Image analysis was performed with ImageJ (http://rsb.info.nih.gov/ij/). Data in the text and graphs are shown as mean ± SEM from n determinations as indicated.
|
Results
|
|---|
Electrophysiological evidence for cross-inhibition between P2X2 and 42 channels
We applied ATP (100 µM) to activate P2X2 and ACh (100 µM) to activate nicotinic channels in HEK cells coexpressing wt P2X2 and 42 channels (Fig. 1A,B). We measured robust ATP- and ACh-evoked inward currents (at -60 mV: IATP at -1161 ± 311 pA and IACh at -522 ± 113 pA; n = 9) when either agonist was applied alone, but less than the predicted (IPredicted at -1635 ± 402 pA; n = 9) current when both ACh and ATP were coapplied (IATP/ACh) (Fig. 1B). The measured current (IATP/ACh at -1037 ± 265 pA; n = 9) on coapplication of ATP and ACh was consistently smaller than the predicted current by 39 ± 4% (ILost). This ILost percentage was thus approximately equal to the IACh percentage (35 ± 3%) and significantly smaller than the IATP percentage (70 ± 2%), where 100% equals IPredicted (see Materials and Methods, Data analysis). Interestingly, the rectification indices and reversal potentials of IATP/ACh (0.18 ± 0.03 and +6.6 ± 0.6 mV; n = 5) were similar to those of IATP (0.21 ± 0.02 and +7.5 ± 0.9 mV; n = 5) and significantly different from IACh (0.04 ± 0.02 and -5.4 ± 1.0 mV; n = 5) (Fig. 1C), as though few nicotinic channels opened after coapplication of ATP and ACh. Together, these data extend previous work on peripheral nervous system channels (see Introduction) to 42 nicotinic and P2X2 channels abundantly found in the brain (Kanjhan et al., 1999; Labarca et al., 2001; Fonck et al., 2003).
FRET between P2X2 and 42 channels in HEK 293 cells
We used the donor dequenching method (Michalet et al., 2003) to estimate FRET efficiency for P2X2 and 42 channels expressed in HEK 293 cells (Fig. 2A). If FRET occurs, one expects full removal (photodestruction) of the YFP acceptor to increase CFP fluorescence by a percentage amount that is equal to FRET e. We determined this value by plotting the photorecovery of CFP as a function of the photodestruction of YFP and estimated e when we extrapolated the data to 100% YFP photodestruction for a membrane ROI (Fig. 2B-E). Our values of e for P2X2-C/P2X2-Y and 4-Y/2-C channels (Table 1, 28% and 25%, respectively) agree with previous work (Nashmi et al., 2003; Fisher et al., 2004). The 10% difference between 4-Y/2-C (e = 25%) and 4-C/2-Y (e = 34%) (Table 1, Fig. 2) may reflect a preferential 4:2 stoichiometric ratio of 2:3 in the heteromeric nicotinic channels (Table 1).
View larger version (23K):
[in this window]
[in a new window]
|
Figure 1. Cross-inhibition between P2X2 and 42 channels expressed in HEK cells. A, Diagram of presently understood membrane topology and oligomeric state of nicotinic and P2X channels. N, N terminus; C, C terminus. B, Three examples of inward currents (-60 mV) evoked by sequential 5 s applications of ATP, ACh, and ATP/ACh from an HEK cell expressing P2X2 and 42 channels (for average data, see Results). The dashed line indicates the level of the peak predicted current if there was no cross-inhibition. C, Top, Representative current-voltage relationships of currents evoked by ATP, ACh, and ATP/ACh together (for average data, see Results). Bottom, Expanded plot near the reversal potential (Erev) for the current-voltage relationships shown in the top panel. The thin lines are the raw data, and the thick lines are two- or three-order polynomial fits to the data (see Materials and Methods, Data analysis). In all cases, agonists (100 µM) were applied at 2 min intervals.
|
|
Importantly, we also measured strong FRET between CFP- or YFP-labeled P2X2 and 42 channels under some conditions. FRET between P2X2 and 42 was strong when the cognate acceptor or donor fluorophore was on the 2 subunit (e = 23 and 26%) (Table 1, Fig. 2D) but was weak or absent when the acceptor (e = 10%) or donor (e = 0%) fluorophores were on the 4 subunit (Table 1, Fig. 2E). However, we know that the P2X2-2 interaction does require assembled 42 channels, because no FRET was measured from cells expressing CFP-labeled P2X2 channels and YFP-labeled 2 subunits (e =0%; n = 9) (Table 1). These data demonstrate that (1) P2X2 and 42 channels were spatially close and that (2) FRET is stronger for P2X2 and 2 than for P2X2 and 4.
We next repeated the experiments with TIRF-FRET microscopy (Fisher et al., 2004). This method allows for the visualization of channels within 100 nm of the plasma membrane (Fig. 3A) in an area called the footprint (Steyer and Almers, 2001). We used donor dequenching to estimate FRET e. Figure 3B shows donor and acceptor images from cell footprints expressing P2X2-C/P2X2-Y, 4-C/2-Y, and P2X2-C/42-Y subunits before and after acceptor photodestruction. From images such as these, we calculated FRET e by extrapolating to 100% acceptor photodestruction (Fig. 4A) for equally sized footprints (Fig. 4B). In contrast to the experiments with epifluorescence microscopy (Fig. 2B,C), we could obtain only one time point, at 85% acceptor photodestruction (see Materials and Methods). FRET e was measured as 34 ± 3, 35 ± 4, 41 ± 5, and 22 ± 2% for P2X2-C/Y (n = 31), 4-Y/2-C (n = 17), 4-C/2-Y (n = 16), and P2X2-C/42-Y channels (n = 22), respectively (Fig. 4B).
As discussed above (see Materials and Methods) it is necessary to photodestroy the acceptor rapidly for TIRF experiments, so as to minimize distortion by channels outside the evanescent field diffusing into the footprint. We could achieve this photodestruction readily within 10-30 s (see Materials and Methods) (Figs. 3B, 4A). However, with this approach, we were restricted to estimating FRET from a single time point (Fig. 4A), and we next sought to address the issue of whether measurements of FRET e from a single point are reliable compared with measurements of FRET e from several points. This was not possible for TIRF, because of the necessity for rapid acceptor photodestruction (see Materials and Methods), but was possible for epifluorescence experiments when the whole cell, rather than the footprint alone, was illuminated. For these experiments, we estimated FRET e from linear plots derived from a single point when the acceptor was 85% destroyed (as in the TIRF experiments) (Fig. 4A) and compared these values with FRET e estimates derived from linear plots such as those illustrated in Figure 2B using multiple points. We performed this analysis for six sets of constructs. Figure 4C shows near equality between FRET e estimates determined from a single point and those determined for the same constructs with multiple points. This indicates that the measurement of FRET e from a single point, when the acceptor is 85% photodestroyed, provides a reliable measure of true FRET e. Moreover, for our most complete data set (P2X2-C/P2X2-Y channels), the FRET e estimates determined with TIRF at a single time point (34 ± 3%; n = 31), epifluorescence with five time points (28 ± 8%; n = 7), epifluorescence with one time point (28 ± 8%; n = 7), or linear unmixing [33 ± 5%; n = 18 from Fisher et al. (2004)] were not significantly different. These data indicate that the technical necessity to estimate FRET from a single time point for the TIRF experiments (Fig. 4A) does not underestimate or overestimate FRET efficiency. These data and analysis give us confidence in the TIRF data presented in Figure 4A.
FRET between P2X2 and 42 channels in neurons
P2X2 and 42 channels are endogenously expressed in brain neurons (Kanjhan et al., 1999; Labarca et al., 2001; Fonck et al., 2003), and we next determined whether P2X2 and 42 channels undergo FRET when transfected into hippocampal neurons. We measured marked colocalization between P2X2-C/P2X2-Y, 4-C/2-Y, and between P2X2-C and 42-Y channels in hippocampal neuron somata (supplemental Fig. 1, available at www.jneurosci.org as supplemental material), as well as strong FRET between P2X2 and 42 channels in hippocampal neuron membrane and in whole-cell ROIs (Table 1, Fig. 5A). This provides strong evidence that the FRET measured between P2X2 and 42 channels in HEK cells also occurs when the subunits are coexpressed in neurons that normally express these channel types (Kanjhan et al., 1999; Labarca et al., 2001; Fonck et al., 2003).
We repeated these experiments for ventral midbrain neurons, which also endogenously express P2X2 and 42 channels (Kanjhan et al., 1999; Labarca et al., 2001; Fonck et al., 2003). We measured strong colocalization between CFP- and YFP-labeled P2X2 channel subunits and between CFP- and YFP-labeled 4 and 2 subunits but weak colocalization between P2X2-C and 42-Y channels (supplemental Fig. 2, available at www.jneurosci.org as supplemental material). Moreover, within midbrain neuron somata, we measured robust FRET for P2X2-C/P2X2-Y and 4-Y/2-C channels (e = 38 ± 8%, n = 4; and 24 ± 6%, n = 3, respectively), but 0% FRET for P2X2-C and 4/2-Y channels (Fig. 5C) (n = 5). These data suggest that there is no interaction between P2X2 and 42 channels in midbrain neuron somata, consistent with their distinctly weak colocalization (supplemental Fig. 2, available at www.jneurosci.org as supplemental material). Data for dendrites followed a similar pattern: we measured strong FRET for dendritic P2X2-C/P2X2-Y and 4-Y/2-C channels at 36 ± 5% (n = 7) and 27 ± 4% (n = 4), but 0% FRET for dendritic P2X2-C and 4/2-Y channels (n = 7). The FRET differences between hippocampal and midbrain neurons extend previous functional studies suggesting that interactions between P2X and nicotinic channels may vary depending on the neuron type (Nakazawa, 1994; Rogers et al., 1997).
On the percentage of P2X2 and 42 channels within 100 Å of each other
Table 1 shows that FRET e was lower for whole cells compared with membrane ROI, in agreement with previous data (Nashmi et al., 2003; Fisher et al., 2004). If one assumes that channels in the plasma membrane display a maximal FRET e equal to that determined with our highest resolution experiments (TIRF microscopy) and also that unassembled subunits do not FRET because their separation greatly exceeds the Förster distance for CFP and YFP (5 nm; see Materials and Methods), the percentage of fluorescent channels that are within 100 Å (Nashmi et al., 2003) can be calculated as follows:
 | (6) |
where eTIRF is the FRET efficiency determined with TIRF, eROI is the value for FRET from an ROI (membrane or whole cell) from epifluorescence microscopy and enoise is the contribution of noise to the FRET measurements. We determined enoise from experiments with P2X6-C/P2X6-Y channels, because these are unassembled (Torres et al., 1999a) and not functionally expressed in the plasma membrane (Khakh et al., 1999). We calculated a FRET value of 3.5% for P2X6-C/P2X6-Y channels for whole-cell ROIs from the average data, because it was not possible to calculate FRET e for all of the single cells, some of which showed negligible photorecovery (Fig. 6). The low FRET value implies that P2X6 channels are expressed mainly as unassembled monomers, a suggestion consistent with previous biochemical experiments (Torres et al., 1999a). From the perspective of the present study, these data thus provide a measure of noise that may arise because of "nonspecific" FRET between subunits that occasionally pass within 80 Å of each other when coexpressed in cells (Fig. 6, Table 1). This value, 3.5%, is significantly lower than the FRET measurements for P2X2, 42, or P2X2/42 channels (Table 1). Using equation 1, we calculated the percentage of fluorophore-labeled channels that were within 100 Å of one another for P2X2, 42, and P2X2/42 channels in HEK cells and hippocampal neurons (Table 1). This value was generally higher for P2X2, 42, and P2X2/42 channels in the membrane than for whole cells (Table 1), consistent with the suggestion that whole-cell ROIs contain contributions from the endoplasmic reticulum, which contain a proportion of unassembled channels (Nashmi et al., 2003; Fisher et al., 2004). We suggest that 30 and 100% of whole-cell and plasma-membrane channels formed by coexpressing P2X2 and 42 channels in HEK cells are dimers between these channel types, with an arrangement such that the P2X2 subunit is  80-100 Å from the 4 subunit, but within 60 Å of the 2 subunit (Fig. 7) (see Materials and Methods and Discussion).
|
Discussion
|
|---|
The main finding, summarized diagrammatically in Figure 7, is that P2X2 and 42 channels form a molecular scale partnership within 100 nm of the plasma membrane.
Functional and optical measures of an interaction between P2X2 and 42 channels
The electrophysiological data for P2X2 and 42 channels expressed in HEK cells demonstrate that these channels undergo functional interactions similar to those described previously for P2X2 and nicotinic, GABA, or 5-HT3 channels (Nakazawa, 1994; Barajas-Lopez et al., 1998; Searl and Silinsky, 1998; Searl et al., 1998; Zhou and Galligan, 1998; Khakh et al., 2000; Sokolova et al., 2001b; Barajas-Lopez et al., 2002; Boue-Grabot et al., 2003, 2004a,b). In future work, it will be important to test whether cross-inhibition also occurs for P2X and nicotinic channels in brain neuron nerve terminals, where they are often coexpressed (Kanjhan et al., 1999; MacDermott et al., 1999; Khakh and Henderson, 2000; Labarca et al., 2001; Fonck et al., 2003; Engelman and MacDermott, 2004). Indeed, there is evidence that some P2X-like channels colocalize and display cross-inhibition with nicotinic channels within cholinergic nerve terminals (Diaz-Hernandez et al., 2002, 2004).
View larger version (17K):
[in this window]
[in a new window]
|
Figure 4. FRET estimates from TIRF microscopy. A, Linear plots of donor (CFP) photorecovery versus acceptor (YFP) photodestruction for channels formed by coexpressing the indicated subunits. The y-axis intercept is a measure of FRET e (see Materials and Methods, Data analysis). B, Footprint areas for the cells shown in A. C, A comparison of FRET e values determined from a single point or for multiple points with epifluorescence microscopy (as in Fig. 2). The data for cytosolic CFP and YFP (CFP/YFP) and yellow cameleon 2.1 (YC2.1) determined with multiple points have been reported (Fisher et al., 2004). The data point for P2X6 has no error bars, because we could estimate FRET e only from the average data (see Results). Error bars represent SEM.
|
|
View larger version (44K):
[in this window]
[in a new window]
|
Figure 5. FRET for channels expressed in hippocampal and midbrain neurons. A, Representative donor and acceptor images of a hippocampal neuron expressing P2X2-C/P2X2-Y channels before and after photodestruction of the acceptor fluorophore. B, Linear plots of donor (CFP) photorecovery versus acceptor (YFP) photodestruction for channels formed by coexpressing P2X2-C and P2X2-Y channels in hippocampal neurons. The y-axis intercept is equal to FRET e. C, As in B but for subunits coexpressed in ventral midbrain neurons.
|
|
Interestingly, we found robust FRET between P2X2 and 42 channels in hippocampal neurons but not in midbrain neurons. Neuron-specific interactions also occur in the periphery: cross-inhibition occurred in sympathetic neurons from the celiac ganglion (Searl et al., 1998) but not in those from the superior cervical ganglion (Rogers et al., 1997). Perhaps there are neuron-specific cues that govern cross-inhibition, for example variable expression of interacting or targeting proteins.
We proposed previously that conformational spread may explain the ability of P2X2 and nicotinic channels to display cross-inhibition (Khakh et al., 2000). In the simplest cases, the proteins that display conformational spread are spatially close or linked (Bray and Duke, 2004) so that motions in a molecule are communicated to its neighbors. Conformational spread is well studied for bacterial chemotaxis and may also occur for receptors and channels (Bray and Duke, 2004). However, there have been no data on the spatial arrangement or possible physical association of P2X2 and nicotinic channels in the plasma membrane, a prerequisite for conformational spread. The present study provides FRET-based evidence for an angstrom scale interaction between full-length P2X2 and 42 nicotinic channels in the plasma membrane of living cells. Thus, we complement previous electrophysiological studies on cross-inhibition. The FRET measurements imply that P2X2 interacts more strongly with the 2 than with the 4 nicotinic receptor subunit and provide the percentage of channels that are within 100 Å of each other and also within 100 nm of the plasma membrane (Fig. 7B).
Molecular interpretations of FRET efficiency data
It is unlikely that the FRET data reported here arise from a non-specific phenomenon such as "overcrowding in the membrane." First, FRET for P2X channels depends strongly on the subunit combination: P2X6 shows negligible FRET (0-3.5%), P2X1 displays intermediate FRET (19%), and P2X2 shows the highest FRET (30-35%) (this study and Fisher et al., 2004). Importantly, all these subunits express robustly in HEK cells (Torres et al., 1999a; North, 2002) and yet FRET varies by a factor of 10. This is inconsistent with simple overcrowding causing FRET in our experiments. Second, previous experiments show that FRET is specific to heteromeric 42 channels or to heteromeric GluCl channels and does not occur between nicotinic or GluCl subunits when they are expressed in the same cell (Nashmi et al., 2003), also arguing against overcrowding. Likewise, in this study, FRET for 42 channels varied depending on the : ratio, as expected of preferential donor/acceptor ratios (2:3 vs 3:2), arguing against overcrowding phenomena. Third, FRET between P2X2 and 42 channels was specific, occurring only between P2X2 and 2 subunits in 42 channels. If FRET occurred because of overcrowding, it would presumably not discriminate between 4 and 2 subunits in assembled 42 channels. Fourth, we measured no FRET in cells expressing P2X2 and 2 subunits alone. Fifth, direct imaging shows that P2X2 channels do not form clusters of over-crowded receptors but rather adopt a relatively uniform membrane distribution (Khakh and Egan, 2005). Sixth, both P2X2 and 42 channels are expressed at a density of <10 channels/µm2 (Khakh et al., 2001; Nashmi et al., 2003; Fisher et al., 2004; Fujiwara and Kubo, 2004; Khakh and Egan, 2005). In contrast, channels become crowded at plasma membrane densities >10,000 channels/µm2 (Hille, 2001). Seventh, FRET has been used for decades to measure protein-protein inter-actions; no substantive data or arguments question the validity of the approach because of nonspecific overcrowding in the membrane (Michalet et al., 2003; Bunt and Wouters, 2004). Future work may overcome the technical challenges of systematically varying the expression levels of P2X2 and 42 channels and determine the impact on FRET.
View larger version (40K):
[in this window]
[in a new window]
|
Figure 6. FRET after coexpression of CFP and YFP-labeled P2X6 subunits. A, Donor (CFP) and acceptor (YFP) images of two HEK cells expressing P2X6-C/P2X6-Y subunits before and after photodestruction of the acceptor fluorophore. Note the minimal dequenching of the donor. B, Linear plots of donor (CFP) photorecovery versus acceptor (YFP) photodestruction for channels formed by coexpressing P2X6 subunits. The y-axis intercept is equal to FRET e (see Materials and Methods, Data analysis).
|
|
The FRET efficiency values for P2X2 and 42 channels reported here agree with previous estimates (Nashmi et al., 2003; Fisher et al., 2004) and in the case of nicotinic channels with values expected from a 4.6 Å structural model (Miyazawa et al., 1999). Extrapolating from this model to 42 channels (Fig. 7A) suggests that the distance between the edges of any 4 subunit and its nearest or farthest 2 subunit would be 38 Å (a) or 45 Å (b), respectively. Note that the fluorophore is buried in a -can 30 Å in diameter and 45 Å in length (Tsien, 1998). Assuming that the fluorescent protein domains line up with their long axes parallel, this implies that the surfaces of the two fluorescent protein domains are separated in 42 channels by 60 minus 30 Å (from two fluorophore radii) or 30 Å. We consider this probable because the five fluorophores likely stick out from the helices forming the 42 cytosolic domain rather than being buried within them, which we presume would be detrimental given the relatively small dimensions of, and available space within, the entire domain (Miyazawa et al., 1999). Because of the 1/R6 dependence, the estimate of 30 Å measures primarily the distance a between an 4 subunit and its nearest 2 (Fig. 7A). Finally, if the 2:3 : stoichiometry and arrangement of Figure 7 is correct, JDA is effectively doubled, and the value of 30 Å is an underestimate by 12%. Thus, the experiments suggest a distance estimate for 42 channels at 33 Å, comparable with 39-41 Å expected from a structural model of a related channel. Similar comparisons are not possible for P2X channels, because no structural models exist, but the FRET efficiency measurements agree roughly with expectations from structural models of the cytosolic domains of several ion channels (Chang et al., 1998; Miyazawa et al., 1999; Sokolova et al., 2001a; Higgins et al., 2002; Jiang et al., 2002; da Fonseca et al., 2003; Kuo et al., 2003).
The finding that FRET is 30-34% for P2X2 channels and 3.5% for P2X6 channels provides strong evidence that FRET efficiency for P2X channels is subunit specific. In previous data, P2X1 channels display intermediate FRET efficiency at 19% (Fisher et al., 2004). The difference for FRET between probes at the extreme C terminus of P2X2 and P2X1 channels may arise because P2X2 channels have a longer C terminus (120 residues) compared with P2X1 (47 residues); perhaps the P2X2 cytosolic domain resembles a phonograph needle, with the C terminus at the cytoplasmic tip (Fisher et al., 2004). The value of 3.5% represents an upper limit and probably measures noise in our experiments; FRET for P2X6 channels may not exist, consistent with electrophysiological (Le et al., 1998; Khakh et al., 1999), biochemical (Torres et al., 1999a |
Top
Abstract
Introduction
