Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

Activation of G protein-coupled receptors (GPCRs) has been shown to produce inhibition of voltage-dependent Ca²⁺ channels (VDCCs) (Hille, 1994). This inhibition is important in modulating the strength of synaptic communication (Miller, 1998). N-type and P/Q-type Ca²⁺ channels are the major VDCCs located in presynaptic nerve terminals that seem to be involved in the regulation of neurotransmission (Dunlap et al., 1995). The major mechanism underlying their inhibition is rapid, voltage-dependent, and mediated by the direct binding of G subunits to the 1 subunits of Ca²⁺ channels. However, there are many examples of GPCR activation that do not produce Ca²⁺ channel inhibition, although G subunits are released. In many cases, these GPCRs are linked to the G_q/11 family of G proteins. The reasons for this selectivity are not clear (Simen et al., 2001). However, G52 has been shown to frequently interact with Gq (Fletcher et al., 1998). Interestingly, G subunits may not be the only dimerization partners for G5. A subgroup of RGS proteins, including mammalian RGS6, 7, 9, and 11, contain a G protein  subunit-like (GGL) domain that interacts specifically with G5 subunits (Sondek and Siderovski, 2001). We have shown that these GGL-containing RGS proteins can block the inhibitory effect of G52 on N-type Ca²⁺ channels, providing a possible explanation for the specificity of VDCC modulation by G protein subunits (Zhou et al., 2000). However, the molecular mechanism underlying these effects remains to be elucidated. Do these GGL-containing RGS proteins exert their effects as GTPase-activating proteins (GAPs) of Gq subunits, indirectly modulating G52-mediated signaling; or do they interact directly with G5 subunits thereby acting as de facto 5/2 antagonists?

RGS proteins have been shown to participate in channel modulation (Wickman and Clapham, 1995; Chen and Lambert, 2000). For example, it has been demonstrated that RGS proteins accelerate deactivation of G protein-coupled inwardly rectifying potassium channel channels, which are directly gated by G subunits (Doupnik et al., 1997; Saitoh et al., 1997). Furthermore, both heterologously (Jeong and Ikeda, 1998; Melliti et al., 1999) and intrinsically (Jeong and Ikeda, 2000) expressed RGS proteins reduce the magnitude and modulate the kinetics of G-mediated N-type Ca²⁺ channel inhibition. Such effects could be explained by the GAP activity of RGS proteins on G subunits. By accelerating GTP hydrolysis, RGS proteins enhance the formation of inactive GDP-G, and thereby attenuate signals transmitted by both G and G subunits (Berman and Gilman, 1998). On the other hand, biochemical studies have demonstrated that GGL-containing RGS proteins can form stable complexes with G5 (Snow et al., 1998, 1999). Additionally, the existence of native G5/RGS complexes in the brain and retina has been reported in immunoprecipitation studies (Cabrera et al., 1998; Liang et al., 2000; Witherow et al., 2000; Zhang and Simonds, 2000). Thus, GGL-containing RGS proteins might directly bind to G5 and selectively modulate G52-mediated signaling. Evidence from our former electrophysiological studies was consistent with the direct-binding model, because the blocking effect of RGS proteins was specific for G5-containing G heterodimers. Furthermore, the results from studies of mutated and truncated constructs indicated that the RGS GGL-domain played an important role in the antagonistic effect (Zhou et al., 2000). Overall, the functional and in vitro biochemical studies suggest that competitive binding is the molecular mechanism through which GGL-containing RGS proteins prevent the formation of functional G52 dimers.

In the present studies, we have used a fluorescence resonance energy transfer (FRET) imaging paradigm to examine whether GFP-tagged G protein subunits and RGS proteins can associate in vivo. Our results complement the findings of biochemical and functional studies, confirming the interaction between G5 and its potential cellular partners. This supports our previous hypothesis that these interactions may be involved in the specificity of GPCR inhibition of Ca²⁺ channels.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

Expression Constructs. To make YFP-tagged G subunits, the cDNA of G5 was cut with BamHI/XbaI, and G1 was cut with EcoRI. The enzyme digestion products were inserted in frame into the polylinker of pEYFP-C1 vector (BD Biosciences Clontech, Palo Alto, CA). The cDNA of G3 was amplified by polymerase chain reaction with a 5' primer containing an XhoI site and a 3' primer containing an EcoRI site. The enzyme-cut fragment was inserted in-frame into the polylinker of pECFP-C1 vector (BD Biosciences Clontech). The cDNAs of G2 and G13 were cut with BamHI/XbaI; RGS11D was cut with EcoRI; and RGS11, RGS11n, and RGS11c were cut with EcoRI/XbaI. All were inserted in frame into the polylinker of pECFP-C1 vector. All fusion constructs were YFP- or CFP-tagged at their N termini. Linkers were derived from the polylinker region of pEGFP-C1 and/or pcDNA3.1 and are not expected to affect the protein functions. Linkers ranged from 13 to 25 amino acids, depending on the convenience of enzyme digestion sites (Fig. 1). They were confirmed by restriction enzyme digestion and sequencing as follows: Y-G1, TCC GGA CTC AGA TCT CGA GCT CAA GCT TCG AAT TCA CTA GTG ATT GCC GCC ACC; Y-G5, TCC GGA CTC AGA TCT CGA GCT CAA GCT TCG AAT TCT GCA GTC GAC GGT ACC GCG GGC CCG GGA TCC GCC GCC ACC; C-G2, TCC GGA CTC AGA TCT CGA GCT CAA GCT TCG AAT TCT GCA GTC GAC GGT ACC GCG GGC CCG GGA TCC CCG; C-Gg3, TCC GGA CTC AGA TCT CGA GTG ATC AGA TTC CCC AGG ACT GTC GCT GCC TGT GGC CTC AGG; C-Gg13: TCC GGA CTC AGA TCT CGA GCT CAA GCT TCG AAT TCT GCA GTC GAC GGT ACC GCG GGC CCG GGA TCC GAC GCC; C-RGS11, TCC GGA CTC AGA TCT CGA GCT CAA GCT TCG AAT TCG ACC ATG CCG CAT CTG AGG AAG; C-RGS11c, TCC GGA CTC AGA TCT CGA GCT CAA GCT TCG AAT TCG ACC; C-RGS11n: TCC GGA CTC AGA TCT CGA GCT CAA GCT TCG AAT TCG ACC; and C-GS11D, TCC GGA CTC AGA TCT CGA GCT CAA GCT TCG AAT TCC GAC.

View larger version (20K): [in this window] [in a new window]   Fig. 1.   Schematic structures of N-terminal GFP-tagged G, G, and different RGS11 constructs. Artificial linkers constructed from the polylinker region of pEGFP-C1 and pcDNA3.1, are drawn in double lines, and ranged from 13 to 25 amino acids. Linker sequences have been detailed under Materials and Methods. Artificial sequences retained from the original constructs and reported in Zhou et al. (2000) are drawn in thick gray lines. All fusion constructs were YFP- or CFP-tagged at their N termini. With the exception of G3 linker sequences and/or fluorophore tagging did not modify the protein function (see text and Fig. 2 for further details).

Cell Culture and Transient Transfection. The C2D7 cell line, derived from human embryonic kidney 293 cells and stably expressing N-type Ca²⁺ channel _1B, ₂, and _1-3 subunits (SIBIA Neurosciences, San Diego, CA), was used. Cell culture was maintained as described previously (Zhou et al., 2000). Cells were transfected using polyethylenimine-mediated method as described by Boussif et al. (1995). For electrophysiological recordings, cells were transfected as described previously (Zhou et al., 2000). For FRET experiments, cells were thinly plated on polylysine-coated 25-mm coverslips at 20 to 30% confluence the day before transfection. Different amount of fluorescent constructs were used in transfection, ranging from 1.5 to 4 µg. The purpose was to achieve a similar expression level, judged by intensity of fluorescence, for either YFP-tagged or CFP-tagged constructs across different transfection pairs.

Electrophysiological Recording and Data Analysis. Whole-cell patch-clamp recordings from C2D7 cells were acquired as described previously (Zhou et al., 2000) 40 to 72 h after transfection. Prepulse experiments were carried out using a prepulse protocol consisting of a 50-ms +10-mV depolarization test pulse from 80-mV holding potential with (test pulse 2) or without (test pulses 1) a 50-ms + 80-mV prepulse (Fig. 2A). Facilitation was indicated by calculating the facilitation ratio (P2/P1), which was defined as the peak current of test pulse 2 divided by the current of test pulse 1 at the same time point (Fig. 2A) as described by Simen and Miller (1998). Electrophysiological and FRET data were plotted as mean ± S.E.M. Electrophysiological data were statistical analyzed using StatMost program (StatMost3.5; Dataxiom Software, Inc., Los Angeles, CA). One-way ANOVA followed by nonparametric Kolmogorov-Smirnov test was used for multiple comparisons. Unpaired t test was used for two-group comparisons.

View larger version (27K): [in this window] [in a new window]   Fig. 2.   Modulation of N-type Ca2+ channels by GFP-tagged G, G, and RGS11 constructs. A, prepulse protocol and calculation of the facilitation ratio. The facilitation ratio (P2/P1) is defined as the peak current of test pulse 2 divided by the current of test pulse 1 at the same time point. B, N-terminal GFP-tagged G and G had similar abilities to inhibit N-type Ca2+ channels, compared with untagged proteins when coexpressed, except for CFP-tagged G3. Data are plotted as mean ± S.E.M.; *, p < 0.01, one-way ANOVA analysis followed by nonparametric Kolmogorov-Smirnov test (n = 5-7). C, N-terminal CFP-tagged RGS11 constructs had similar abilities compared with their untagged counterparts to antagonize Ca2+ channel inhibition by G52. Data are plotted as mean ± S.E.M.; *, p < 0.01, unpaired t test between G52 and G52/RGS11-constructs (n = 5-15).

FRET Data Acquisition and Analysis. Forty-eight to 72 h after transfection, C2D7 cells were washed with phosphate-buffered saline and fixed with 4% paraformaldehyde at 4°C for 20 min. Then cells were washed three times with PBS and mounted with mounting solution (60% glycerol, 3% n-propyl gallate, 120 mM Tris base, sonicated to dissolve, pH 7.4, and kept in dark at 4°C). Images were captured with a fluorescence microscope (BX50WI; Olympus, Tokyo, Japan), equipped with an intensified 12-bit Penta-MAX charge-coupled device camera (Princeton Instruments, Trenton, NJ) controlled by MetaMorph software (Universal Imaging Corporation, Downingtown, PA) using a 100× numerical aperture 1.3 objective. Filter sets for the fluorescence channels used were as follows: YFP (excitation filter, 500 ± 20 nm; beam splitter, 515 nm long pass; and emission filter, 535 ± 15 nm), CFP (440 ± 20 nm, 455 nm, 485 ± 20 nm), and FRET (440 ± 20 nm, 455 nm, 535 ± 15 nm). No neutral density filter was used (neutral density = 0). Briefly, three consecutive images were collected from the same field, using the filter settings for YFP, FRET, and CFP. After that, a fourth image was also collected using a 340-nm excitation filter to obtain the background image. Occasionally, autofluorescence from cells was clearly visible on the background images, but the pixel values never exceeded twice the values of the background where no cells were found. Emissions from CFP or YFP constructs were not detectable using the 340-nm excitation filter. The background image obtained this way was used in the background subtraction subroutine of the MetaMorph software to obtain the FRET fluorescence image (Ff) from the raw FRET image, the donor fluorescence image (Df) from the CFP image, and the acceptor fluorescence image (Af) from the YFP image. The two-letter symbols were used as defined by Gordon et al. (1998). Shading correction was also applied at the time of background subtraction.

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

GFP-Variants of G and RGS11 Proteins Preserve Their Ability to Inhibit N-type Ca²⁺ Channels. To use FRET as an indicator of protein-protein interactions, two variants of GFP, CFP and YFP, were tagged to proteins of interest. Based on the structures of G and G subunits (Wall et al., 1995; Sondek et al., 1996) and molecular modeling of GGL-containing RGS proteins (Snow et al., 1998), fluorescent tags were fused in-frame to the amino termini of these proteins (Fig. 1). Sequences around the linker area were confirmed by sequencing. YFP-G1 contains an 18 amino acid linker, YFP-G5 25 aa, CFP-G2 23 aa, CFP-G3 20 aa, CFP-G13 24 aa, CFP-RGS11 19 aa, and CFP-RGS11c, CFP-RGS11n, and CFP-RGS11D all contain a 13 aa linker. These linkers were derived from the polylinker regions of pEGFP-C1 and pcDNA3.1 vectors and were chosen based on availability of restriction enzyme sites.

Interaction of G and G Subunits in Cells. We used the FRET technique to study protein-protein interactions between G and G subunits. YFP-tagged G subunits and CFP-tagged G subunits were cotransfected into cells. Figure 3A demonstrates typical images observed from the YFP, FRET, and CFP channels. All colors are artificially assigned. Signals from the YFP channel are presented as green, signals from CFP channel as blue, and signals from the FRET channel are shown in pseudocolor. Numerical values for each point of each channel represent the intensity of the signal, but do not encode color information. However, the color images in Figs. 3A and 4A do illustrate some important points. The yellow (top) and cyan (bottom) images demonstrate that the cellular distribution of the YFP- and CFP-tagged constructs were fairly similar. Moreover, although the raw data from the FRET channel (middle) does not carry information about the FRET interaction by itself, it does demonstrate the importance of the proper normalization in the FRET methodology. Owing to different expression levels of the constructs within cells and inhomogeneities in the thickness of the cells, the raw FRET intensities vary from cell to cell. However, after normalization, the normalized data from different regions of interest fell into the same range regardless of the original intensities of the raw FRET values.

View larger version (24K): [in this window] [in a new window]   Fig. 3.   Interaction between G and G subunits in cells detected by FRET. Cells were transfected with indicated constructs. FRET signal acquisition was carried out 48 to 72 h after transfection using a two-cube setup (see Materials and Methods). FRETN was calculated as described; see Materials and Methods and Gordon et al. (1998). A, representative fluorescent signals observed from three channels. All colors are arbitrarily assigned to indicate signal strength. B, summary of FRETN between different protein pairs as indicated. cfp-yfp is a construct that CFP and YFP fluorophores are directly linked (see Materials and Methods). Data are plotted as mean ± S.E.M. (n = 20-40).

View larger version (24K): [in this window] [in a new window]   Fig. 4.   Interaction between G and RGS11 constructs in cells detected by FRET. A, representative fluorescent signals observed from three channels. All colors are arbitrarily assigned to indicate signal strength. B, summary of FRETN between different protein pairs as indicated. Data are plotted as mean ± S.E.M. (n = 20-40).

5

5

5

5

5

5

5

Interaction of G5 and RGS11 Constructs in Cells. To test whether GGL-containing RGS11 proteins associate directly with G5 in cells and which domain(s) is involved in the association, protein-protein interactions between G5 and different RGS11 constructs were studied using FRET. The results are summarized in Fig. 4. Again, the FRETN signal from expressed CFP-YFP concatemer was used as a positive control, and FRETN from coexpression of the YFP vector and CFP-G2 was used as a negative control. Coexpression of YFP-G5 and the CFP-tagged full-length RGS11 protein (c-R11) generated a strong FRETN signal of (92.6 ± 1.7) × 10⁵, similar to the positive control. On the other hand, coexpression of YFP-G1 and CFP-RGS11 had an FRETN value of (12.5 ± 1.5) × 10⁵, similar to the negative control. RGS11n contains the N-terminal, the DEP domain, the DEP-GGL linker, and the GGL domain of RGS11 (Fig. 1). Coexpression of CFP-tagged RGS11n (c-R11n) and YFP-G5 gave an FRETN of (63.3 ± 3.2) × 10⁵. RGS11c contains the GGL-RGS linker, the RGS domain, and the C-terminal of RGS11 (Fig. 1). FRETN for coexpression of CFP-R11c and YFP-G5 was (26.8 ± 1.3) × 10⁵. RGS11D contains the GGL domain, the GGL-RGS linker, the RGS domain, and a truncated C-terminal. Coexpression of CFP-R11 D and YFP-G5 had an FRETN of (70.0 ± 1.1) × 10⁵ (Fig. 4B).

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

Activation of Gq-coupled GPCRs often fails to produce fast, voltage-dependent, G/-mediated inhibition of Ca²⁺ channels (Simen et al., 2001). We have previously shown that Gq-selective G5 subunits can produce Ca²⁺ channel inhibition when paired with G2. However, the inhibitory effect of G52 can be specifically blocked by GGL domain-containing RGS proteins (Zhou et al., 2000). RGS proteins have been shown to be involved in channel modulation either as GAPs or as effector antagonists (Berman and Gilman, 1998). Our previous study suggested that GGL-containing RGS proteins might also selectively attenuate signaling mediated by G5-containing G heterodimers through direct interaction between the GGL-domain and G5. We have now tried to confirm this hypothesis by studying protein-protein interactions between G5 and GGL-containing RGS11 proteins in cells.

Interactions between G and G or RGS proteins have been studied using biochemical methods. The affinity of G52 association is relatively weak and sensitive to isolation conditions (Snow et al., 1999). On the other hand, GGL-containing RGS proteins interact strongly with G5 in in vitro binding studies, and endogenous complexes of G5/RGS have been identified from the brain and retina in immunoprecipitation studies. We addressed this question further by studying protein-protein interactions in cells using FRET techniques. The efficiency of FRET is inversely related to the 6th power of the distance between the two fluorophores. For the CFP/YFP pair we used the critical distance (R₀) is about 50 Å. Therefore, any significant FRET signal should indicate direct interactions between the tagged proteins.

Although the underlying mechanism that generates FRET may be relatively straightforward, the actual data analysis is complicated. Several methods have been developed (Wouters et al., 2001). Our present study used an intensity-based method termed "sensitized emission". Sensitized emission is defined as the increased acceptor emission due to energy transfer from the donor to the acceptor. Intensity-based methods for FRET data analysis suffer from a high level of false-positives when not properly corrected for direct fluorescence contributions from both donor and acceptor. We adopted the formula of Gordon et al. (1998) to correct for direct fluorescence contributions and concentration effects. There are several different versions of this formula (Gordon et al., 1998; Xia and Liu, 2001). However, based on our observations, the concentration effect was not completely corrected using either of the formulae. The calculated FRETN exhibited a large variation due to heterogeneity of the expression levels of donor and acceptor proteins. In the present studies, several measures were adopted to control for this effect. First, the amount of cDNA used in transfections was adjusted to obtain similar expression levels for YFP-tagged and CFP-tagged constructs, respectively, judged from their fluorescence intensity. Second, only cells with similar expression levels of YFP-tagged and CFP-tagged proteins were selected for data analysis. Finally, data were corrected for differences in expression levels using statistical methods. To check the sensitivity and selectivity of the protocol we used for FRET detection, controls were always run in parallel with our experimental samples. As expected, the positive control, a CFP-YFP concatemer, produced an FRETN value significantly larger than that produced by the negative controls. Thus, the protocol was used in our study of protein-protein interaction.

For the FRET analysis to be meaningful, the fusion proteins must be correctly assembled and functional. All YFP/CFP-tagged G, G, and RGS11 constructs produced similar effects compared with their untagged counterparts in modulating N-type Ca²⁺ channels, with only one exception. Interestingly, N-terminal CFP-tagged G3 "gained" the ability to form functional dimers with G5 and to inhibit N-type Ca²⁺ channels; untagged G5/3 was not effective in this functional test. This gain of function was due to the CFP-G3, rather than YFP-G5. Coexpression of untagged G5 with CFP-G3 inhibited channels, whereas coexpression of YFP-G5 with untagged G3 did not (data not shown). Because the lack of effect of untagged G5/3 is presumably due to the inability of these two subunits to form heterodimers (Watson et al., 1994), an N-terminal CFP tag on G3 may help to stabilize the interaction with G5. The N-terminal coiled-coil structure between G and G subunits is important for their interaction, and the "unusual" G5 subunit differs from other G subunits mainly at the N-terminal portion (Clapham and Neer, 1997). Therefore, an N-terminal CFP tag on G3 subunit may somehow stabilize the N-terminal structure of G5 and G3 and help to form a functional heterodimer. Alternatively, tagging G3 might greatly increase its level of expression, resulting in the observed interaction, although we have no evidence that this is the case. Furthermore, epifluorescent and confocal microscopy showed similar subcellular distributions of GFP-tagged constructs in transfected cells (data not shown) compared with that reported for endogenous proteins (Zhang et al., 2001), indicating that the N-terminal GFP-tag did not disrupt normal cellular distributions of the proteins.

Results from the FRET studies complement those of biochemical and functional studies. The FRETN data for YFP-G5/CFP-G2 and YFP-G5/CFP-G13 had intermediate values, suggesting that G5 forms heterodimers with G2 and G13. This was in agreement with the results of electrophysiological studies (Blake et al., 2001). Electrophysiological studies also indicated that G53 did not form functional dimers that inhibited N-type channels, whereas, as we show here, G5/CFP-G3 did. Thus, as discussed above, the N-terminal CFP tag on G3 may stabilize its interaction with G5. Coexpression of YFP-G5 and CFP-RGS11 generated a large FRETN value, suggesting that G5 and RGS11 did form stable heterodimers in cells. On the other hand, YFP-G1 did not form heterodimers with CFP-RGS11 and had an FRETN value similar to negative controls. These findings further support the results from biochemical binding studies (Snow et al., 1998) and immunoprecipitation studies (Witherow et al., 2000; Zhang and Simonds, 2000) and are in support of the notion that interaction between G5 and GGL-containing RGS proteins underlies the blocking effect of RGS11 on G52 mediated N-type channel inhibition. Furthermore, the interaction between G5 and GGL-containing RGS proteins, indicated by FRETN values, depends on the GGL domain. CFP-RGS11n and CFP-RGS11D, both containing the GGL domain, interacted with YFP-G5, whereas CFP-RGS11c had much weaker interaction with YFP-G5. The pattern of interactions between G5 and RGS11 constructs (Fig. 4B) coincides with the pattern of blocking effect of different RGS11 constructs on G52-mediated N-type channel inhibition (Fig. 2C). A reasonable explanation would be that interaction between G5 and GGL-containing RGS proteins underlies the blocking effect.

FRET techniques provide a way of studying protein-protein interaction in vivo and are potentially useful for quantifying the interaction and dynamics of protein interaction. However, great care must be taken to interpret the energy transfer data. Because FRET is sensitive to the distance between the two fluorophores and the dipole orientation, such effects may be so significant that different energy transfer readings could be due to either conformational changes or binding-unbinding, which cannot be distinguished from the FRET data. Furthermore, an apparent energy transfer efficiency for sensitized emission is equal to the true energy transfer efficiency multiplied by the fraction of tagged acceptor molecules in a complex and the ratio of donor and acceptor brightness (Wouters et al., 2001). The FRETN values for YFP-G1/CFP-G2 and YFP-G5/CFP-RGS11 were quite high, similar to the positive control of CFP-YFP concatemer. This is intriguing when one considers that all (100%) donor and acceptor pairs coexist in the CFP-YFP concatemers, whereas presumably only a fraction of donor/acceptor pairs would exist in complex resulting from G/ or G/RGS transfections. However, although the fraction in a complex may be lower for G/ or G/RGS transfections, their true FRET efficiency could be higher. Indeed, that is what might be expected. True FRET efficiency is determined by the strength of protein-protein interaction and falls off with the inverse 6th power of distance. The CFP and YFP fluorophores are simply linked together by a linker of nine amino acids in the concatemer; there is no active binding activity between the two molecules. On the other hand, biochemical studies have shown that there is active interaction in the complex of G1/2 and G5/RGS11, possibly resulting in a higher FRET efficiency. However, the relative contributions of energy transfer efficiency and the fraction ratio could not be distinguished from the observed FRET signals. At each point in an image, a mixture of states, both bound and unbound, coexists. To quantify the true FRET efficiency, the relative amounts of bound and unbound acceptor must be known at each point. This requires precise knowledge of all of the biochemical parameters governing the interaction between the two proteins under consideration, so that correct models can be used. Therefore, both biochemical studies and FRET studies offer important complementary information.

In summary, we show here that GGL-containing RGS proteins can interact directly with G5 in cells. The pattern of interactions between G5 and RGS11 constructs, indicated by FRET values, coincides with the pattern of blocking effects of different RGS11 constructs on G52-mediated N-type channel inhibition in electrophysiological studies. Therefore, we suggest that GGL-containing RGS proteins can act as G5 effector antagonists and specifically attenuate G5-mediated signaling, including inhibition of N-type Ca²⁺ channels.