Splice Isoforms of Phosducin-like Protein Control the Expression of Heterotrimeric G Proteins*

Background: Heterotrimeric G proteins are essential for biological signaling; however, the mechanism of their biosynthesis remains poorly understood. Results: Long and short splice isoforms of phosducin-like protein stimulate and inhibit production of G proteins in the cell. Conclusion: Both G protein α and βγ functional units are subject to the regulation. Significance: We describe a potential mechanism for regulating the cellular levels of G proteins. Heterotrimeric G proteins play an essential role in cellular signaling; however, the mechanism regulating their synthesis and assembly remains poorly understood. A line of evidence indicates that the posttranslational processing of G protein β subunits begins inside the protein-folding chamber of the chaperonin containing t-complex protein 1. This process is facilitated by the ubiquitously expressed phosducin-like protein (PhLP), which is thought to act as a CCT co-factor. Here we demonstrate that alternative splicing of the PhLP gene gives rise to a transcript encoding a truncated, short protein (PhLPs) that is broadly expressed in human tissues but absent in mice. Seeking to elucidate the function of PhLPs, we expressed this protein in the rod photoreceptors of mice and found that this manipulation caused a dramatic translational and posttranslational suppression of rod heterotrimeric G proteins. The investigation of the underlying mechanism revealed that PhLPs disrupts the folding of Gβ and the assembly of Gβ and Gγ subunits, events normally assisted by PhLP, by forming a stable and apparently inactive tertiary complex with CCT preloaded with nascent Gβ. As a result, the cellular levels of Gβ and Gγ, which depends on Gβ for stability, decline. In addition, PhLPs evokes a profound and rather specific down-regulation of the Gα transcript, leading to a complete disappearance of the protein. This study provides the first evidence of a generic mechanism, whereby the splicing of the PhLP gene could potentially and efficiently regulate the cellular levels of heterotrimeric G proteins.

Heterotrimeric G proteins are a conserved group of molecules involved in a great number of signaling processes in eukaryotes (1). Although much is known about G protein-me-diated signaling, far less is known about G protein biogenesis. All heterotrimeric G proteins are composed of ␣, ␤, and ␥ subunits encoded by distinct genes. In humans, 16 genes code for G␣, 12 genes code for G␥, and five genes code for G␤. The mechanisms underlying the process of selection and assembly of these subunits into mature G proteins remain unknown.
Evidently, the folding of nascent G protein ␤ polypeptides requires the direct assistance of chaperonin containing t-complex protein 1 (CCT 2 ; also known as TRiC) (2,3) and is finalized by a permanent association with G␥ (4); however, the order of these events has not been well defined. Chaperonin CCT is a large, cylindrically shaped ATPase complex composed of two stacked rings. Each ring is made of eight t-complex protein 1 (TCP-1) subunits, designated ␣in mammals (5,6). In its open conformation, the substrate proteins are allowed to enter the central cavity. Nucleotide binding occurs once the substrate is bound, and ATP hydrolysis triggers the conversion from the open to the closed conformation (5). The exact mechanism by which CCT folds the substrate proteins has not been clearly defined, but it is thought to involve a set of sequential, hierarchical steps. The CCT substrates identified thus far are all soluble proteins, diverse in functions and structures (7). Many of these substrates, exemplified by the G protein ␤ subunits (8), possess WD-repeat domains that fold into a ␤-propeller structure (9 -11). Nevertheless, due to the diversity of CCT substrates, it is impossible to predict which proteins are folded by CCT based on sequence alone (12). Plausibly, the CCT substrates are selected and directed to CCT by specific co-factors. The proteins from the phosducin family emerge as one such set of CCT co-factors that are involved in substrate guidance.
The phosducin gene family is divided into subgroups I-III (13). Subgroup I consists of phosducin and PhLP1, which both bind G protein ␤ subunits with high affinity (2,14). PhLP1 (often called PhLP) is ubiquitously expressed (15) and is shown to act as a CCT co-factor during the folding of G protein ␤ (16 -20). In contrast, phosducin is expressed predominantly in rod and cone photoreceptors, where it helps maintain high lev-els of and assists in the trafficking of the visual G protein transducin without any known connection to CCT (21)(22)(23). Subgroup II consists of PhLP2A and PhLP2B, which were both discovered in humans (13). The PhLP2 ortholog in Saccharomyces cerevisiae interacts with CCT and is essential for cell growth (24) due to its function in actin biogenesis (25,26). In contrast to PhLP1, PhLP2 is not involved in G␤ folding. Subgroup III is composed of only PhLP3 (13), which is thought to participate in ␤-tubulin and possibly actin folding (27) but which binds G␤ poorly (24). Thus, one commonality among PhLP1 to -3 is that they all participate in protein folding by acting as co-chaperones of CCT (2).
Alternative splicing of the PhLP1 gene in rats gives rise to two distinct transcripts, 297 and 218 amino acids long. A product of a short transcript was originally designated as phosducin-like protein short isoform (PhLPs) (28). Essentially the same transcript was later detected in the human retina and named phosducin-like orphan protein 1 (PHLOP1) (29). The precise biological function of PhLPs has remained elusive; however, some evidence has suggested that it acts as a negative regulator of heterotrimeric G proteins (17,30,31). To gain a better understanding of this important function, we ectopically expressed PhLPs in mouse rods, photoreceptor neurons of the retina that maintain very high levels of the heterotrimeric G protein, transducin (G␣ t1 ␤ 1 ␥ 1 ). We found that PhLPs disrupted several steps in the biosynthesis of rod transducin, which could also be observed in cell culture models. PhLPs specifically targeted the chaperonin CCT preloaded with nascent G␤ 1 . In contrast to the full-length PhLP, which facilitated the assembly of the G␤ 1 ␥ 1 heterodimer, PhLPs disrupted this process by trapping G␤ 1 inside the chaperonin. In addition, PhLPs evoked a strong and specific transcriptional suppression of G␣ t1 . This study, thus, provides the first evidence that the short isoform of PhLP can act as a negative regulator of the synthesis and assembly of heterotrimeric G proteins in vivo.

EXPERIMENTAL PROCEDURES
Animal Models-Transgenic mice expressing ⌬1-83 PhLP-FLAG have been previously described (32). All analyses were performed using ⌬1-83 PhLP-FLAG ϩ/Ϫ and ⌬1-83 PhLP-FLAG Ϫ/Ϫ littermates at postnatal day 10, according to the procedures approved by the animal care and use committees of West Virginia University.
Quantification of mRNA-The alternatively spliced PhLP transcripts were detected in the Human Total RNA Master Panel II (Clontech). One microgram from each RNA sample was reverse transcribed using random hexamers and RNase H(Ϫ) reverse transcriptase. The region of the phosducin-like protein gene (PDCL) containing exon 2 was amplified with primers located in exons 1 (PDCL_1-F, 6-FAM-TGAC-CCACTGCTCTTCCTCT) and 3 (PDCL_1-R, GATGGGAC-CTGCAAGTCATT). The forward primer carries a 6-carboxyfluorescein (6-FAM) fluorescent tag attached to its 5Ј-end. The resulting fluorescently labeled products were resolved using denaturing polyacrylamide electrophoresis and imaged on a Typhoon imager (GE Healthcare) in the fluorescence mode.
To measure the levels of mRNA in the retina, RNA was isolated from single retinas using the Absolutely RNA Miniprep kit (Agilent Technologies), according to the manufacturer's protocol for small samples and including two DNase treatments. The final RNA concentration was determined using a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies). Using a two-step quantitative RT-PCR process, cDNA from each retina was generated from 5 ng of total RNA using the AffinityScript qPCR cDNA synthesis kit (Agilent Technologies) and oligo(dT) primers. Negative controls included no reverse transcriptase and no RNA template. Quantitative realtime PCR experiments were performed in triplicate with 0.5 ng of cDNA using the Brilliant II SYBR Green qPCR master mix (Agilent Technologies), reference dye, and a 200 nM concentration of each primer. Reactions were incubated at 95°C for 10 min and then cycled 27 times at 95°C for 30 s, 55°C for 60 s, and 72°C for 30 s using a Stratagene Mx3000P TM real-time PCR system. A melting curve analysis was added at the end to verify a single product from each reaction, and the fluorescence was recorded during every qPCR cycle at both the annealing step (55°C) and the extension step (72°C). Fluorescence values were analyzed using MxPro qPCR software version 4.10, and amplification thresholds were normalized to those of GAPDH.
Western Blotting-Retinas were collected at postnatal day 10 and frozen on dry ice. Each retina was homogenized in 0.2 ml of buffer containing 125 mM Tris/HCl, pH 6.8, 4% SDS, and 6 M urea by short ultrasonic pulses, and the resulting extract was cleared by centrifugation. The total protein concentration was determined on a Nanodrop ND-1000 spectrophotometer. To prepare SDS-PAGE samples, the protein concentration in each of the compared samples was adjusted to the same value, and bromphenol blue tracking dye and 5% ␤-mercaptoethanol were added. Typically, 10-l samples from Tg(ϩ) and Tg(Ϫ) littermates were separated side by side on a 10 -20% Tris/HCl acrylamide gel (Bio-Rad). Quantification of the specific bands was performed using an Odyssey infrared imaging system (LI-COR Biosciences) according to the manufacturer's protocols. Data from three independent measurements utilizing at least three mice of each genotype each time were averaged.
Microarray Analysis-Total RNA were isolated from the retina using the Absolutely RNA Miniprep kit, and its quality was assessed with Bioanalyzer (both from Agilent Technologies). Samples were processed using an Affymetrix GeneChip whole transcript sense target labeling assay and hybridized to Affymetrix Mouse Exon 1.0 ST microarrays according to the manufacturer's instructions. Microarrays were analyzed using GeneBase software as described previously (33,34). GeneBase estimates the gene expression levels based on probes with highly correlated signals across multiple samples, thus reducing the bias from cross-hybridizing probes or probes located in alternative exons. To take advantage of the probe selection algorithm of GeneBase, we combined our mouse retina arrays with the Affymetrix mouse exon 1.0 ST mouse tissue panel, which contains 33 arrays hybridized to samples from 11 tissues. p values were adjusted using Benjamini-Hochberg's procedure.
Mass Spectrometry-⌬1-83 PhLP-FLAG was captured from the retinas with anti-FLAG-agarose and vacuum-dried, as described previously (32). A pull-down from 70 retinas was sent to Applied Biomics, Inc. (Hayward, CA), which conducted twodimensional PAGE separation, protein identification in the spots by LC/MS/MS, and data analysis. In another experiment, a pull-down from 150 retinas was dissolved in buffer containing 125 mM Tris/HCl, pH 6.8, 4% SDS, 6 M urea, 5% ␤-mercaptoethanol, and bromphenol blue tracking dye and subjected to a standard SDS-PAGE separation on a 12.5% Tris/HCl gel (Bio-Rad). Following a short run, the gel was stained with Colloidal Coomassie (LC 6025, Invitrogen), and the entire lane was excised from the gel and cut into identical fragments. The proteins in each gel fragment were digested with trypsin and analyzed by microcapillary reverse-phase HPLC nanoelectrospray tandem mass spectrometry (LC/MS/MS) on the Thermo LTQ-Orbitrap mass spectrometer at the Harvard Mass Spectrometry and Proteomic Resource Laboratory (Cambridge, MA). The acquired MS/MS spectra were correlated with known sequences using programs developed in this laboratory (35) and the Sequest algorithm (40). The data were analyzed to reach an estimated false discovery rate of ϳ1% using a reverse database strategy. When a peptide sequence could be assigned to more than one protein, the assignment was made to the protein that had accumulated the most peptide spectra; thus, no peptide spectrum was ever assigned to more than one protein.
Preparation of cDNA Constructs-Mouse phosducin-like protein with a C-terminal FLAG tag (PhLP-FLAG) was prepared by reverse transcription of total retinal RNA from a 129/SV mouse using the gene-specific primer 5Ј-ACT AAA TGA GAC TAC AA, followed by PCR with the following primers: forward primer, 5Ј-GAG ACC ATG GCA ATG ACA ACC CTG GAT GAT AAG TTA CTG; reverse primers, 5Ј-CTT GTC ATC GTC GTC CTT GTA ATC ATC TAT TTC CAG ATC GCT GTC TTC and 5Ј-GCC TGG ATC CCT ACT ACT TGT CAT CGT CGT CCT TGT AAT C. The product was cut with NcoI and BamHI restriction endonucleases and inserted into the pTriEx4 expression vector (Novagen) at these sites. To utilize the potential for a Kozak sequence at the NcoI site and 5Ј-end of PhLP-FLAG, a few additional bases needed to be added, thereby expressing PhLP-FLAG with an additional methionine and alanine at the N terminus. PhLPs ( ⌬1-83 PhLP-FLAG) was subcloned from pRhop4.4k-⌬1-83 PhLP (32) using PCR and the following primers to add a C-terminal FLAG tag: forward primer, 5Ј-GAG ACC ATG GAG CGG CTG ATC AAA AAG CTG TC T ATG AG; reverse primer, 5Ј-GCC TGG ATC CCT ACT ACT TGT CAT CGT CGT CCT TGT AAT C. The product was cut with NcoI and BamHI restriction endonucleases and inserted into the pTriEx4 vector at these sites. The coding sequence of mouse Pdc was subcloned from another vector in our laboratory, pSC-A Kozak/Pdc/FLAG, using PCR and the following primers: forward primer, 5Ј-GAG ACC ATG GAA GAA GCC GCC AGC CAA AGC; reverse primer, 5Ј-GGG CCT CTC GAG TTC AAT GTC CTC GTC TTC CAT GTT G. The product and pTriEx4 vector were digested with NcoI and XhoI restriction endonucleases and ligated at these sites to generate a Pdc construct that would be expressed with a C-terminal His 6 epitope tag. Human rod transducin ␤ (G␤ 1 ) cDNA was amplified from a human retinal cDNA library (a kind gift from Dr. Jeremy Nathans, Johns Hopkins University) and subcloned into the EcoRI and NotI restriction sites in the pCMV-myc vector to create G␤ 1 with an N-terminal myc epitope tag (Clontech). Similar to the transducin ␤ subunit, human rod transducin G␥ 1 with an N-terminal HA epitope tag was generated using the NcoI (partial digestion) and XhoI sites in the pCMV-HA vector (Clontech). The integrity of all constructs was confirmed by sequence analysis.
Cell Culture-HEK 293 cells (ATCC) were cultured in Dulbecco's modified Eagle's medium (DMEM)/Ham's F-12 (50:50 mixture) medium (Mediatech, Inc.), supplemented with 10% defined fetal bovine serum (HyClone) and 1ϫ penicillin-streptomycin-L-glutamine (Invitrogen) or 100 units/ml penicillin, 100 g/ml streptomycin, and 0.292 mg/ml L-glutamine. For the transfection, cells were plated in 10-cm plates, so they would be at least 50% confluent the following day. 9.6 g of total plasmid DNA was transfected per plate using FuGENE 6 transfection reagent (Roche Applied Science and Promega) at a 6:1 FuGENE 6 reagent/plasmid DNA ratio. The total amount of cDNA was kept constant by adding empty pTriEx-4 vector to make up any differences in the co-transfections, and the transfections were performed in triplicate. Cells transfected with empty pTriEx 4 vector only were used as a negative control. All cells were collected in Hanks' balanced salt solution (HBSS), supplemented with 1 mM EDTA, 48 h after transfection, and pull-down assays were performed in either radioimmune precipitation assay buffer (R0278, Sigma) or buffer containing 20 mM MES, pH 6.5, 5 mM EDTA, and 1.5% Nonidet P-40 using anti-FLAG M2 and anti-c-Myc-agarose affinity gels (A2220 and A7470, Sigma) and anti-HA-agarose (26181, Pierce). Captured proteins were eluted with FLAG, HA, or Myc peptides in 1 mM Tris/HCl (pH 7.4), vacuum-dried, and reconstituted in USB buffer for Western blot analyses.

Identification of the PhLPs Transcript in Human Tissues-
The skipping of exon 2 in the phosducin-like protein gene gives rise to two alternative transcripts encoding full-length PhLP and N-terminally truncated PhLPs (Fig. 1A). To determine the expression profile of these transcripts, we quantified their relative abundance in 20 human tissues by RT-PCR (Fig. 1B). Both transcripts could be detected in all tested tissues, with PhLP being the predominant splice isoform. The highest levels of PhLPs were found in neural tissue, where it comprised ϳ10% of the total PhLP mRNA. Interestingly, expressed sequence tag evidence suggests that this PhLP splicing does not take place in mice. As such, there are 82 spliced expressed sequence tags and six full-length mRNAs that align to the genomic region containing exon 2 of the mouse PhLP (Pdcl) gene. All of these sequences include exon 2. The alignment can be viewed at the University of California Santa Cruz genome browser. Subsequently, we were unable to detect the PhLPs transcript by RT-PCR in mouse retina and brain (not shown). Thus, the alternative splicing of the PhLP gene, which produces the truncated PhLPs protein, appeared to be widespread in human tissues although not conserved evolutionarily in mice. Due to the low expression levels of PhLP and lack of adequate antibodies, our attempts to detect endogenous PhLP isoforms by Western blotting were unsuccessful.
PhLPs Disrupts the Expression of Heterotrimeric G Proteins in Mouse Photoreceptors-Having found that mice lack endogenous PhLPs, we used this model to explore the function of this protein. Toward this goal, a transgene encoding epitope-tagged PhLPs was expressed in mice. This transgene, designated here as ⌬1-83 PhLP-FLAG, encodes truncated PhLP that starts from Met-84, and is identical to PhLPs. Driven by a 4.4-kb rhodopsin promoter, the transgene was turned on at around postnatal day 5, and its expression was primarily restricted to the rod photoreceptors of the retina. We found that although the majority of rods survived until postnatal day 10 ( Fig. 2A), these cells did not elaborate their light-sensing compartments, the outer segments, and failed to express heterotrimeric G proteins (Fig. 2, B  and C). As such, the G␣ t1 subunit of the abundant rod G protein, transducin, was undetectable on the protein level, and two other transducin subunits, G␤ 1 and G␥ 1, were reduced by 63 Ϯ 4 and 95 Ϯ 1%, respectively. The photoreceptor-specific long isoform, G␤ 5 L, was also significantly reduced by 84 Ϯ 3%, whereas the short isoform, G␤ 5 S, expressed in the other retinal cell types, remained unaffected. The protein levels of photoreceptor-specific visual arrestin and ubiquitously expressed ␤-tubulin were also affected insignificantly. With the exception of G␣ t1 , whose transcript, Gnat1, was down-regulated 10-fold, the mRNA levels of G␤ 1 , G␥ 1 , and G␤ 5 L were reduced by less than 50%, which is insufficient to account for the observed reduction in these proteins. Hence, it appeared that all G␤ and G␥ affected by ⌬1-83 PhLP-FLAG were destabilized primarily on a posttranslational level.
The observed significant down-regulation of the Gnat1 gene prompted us to examine the global impact of ⌬1-83 PhLP-FLAG on gene expression. We used an Affymetrix GeneChip whole transcript exon array to identify the changes in mRNA expression in the retinas of ⌬1-83 PhLP-FLAG-positive mice, as compared with their littermate controls at postnatal day 10. The change in gene expression was considered significant if the -fold change was greater than 2.00 at a false discovery rate below 0.05. We identified nine significantly down-regulated genes, including Gnat1, which was down-regulated the most (Table 1). Besides phosducin-like protein, which was overexpressed from the transgene, the only significantly up-regulated gene was activating transcription factor 3 (Atf3). Forty-three additional transcripts showed a statistically significant (false discovery rate Ͻ0.05), albeit small in magnitude, change in expression levels (supplemental Table 1). Of these, the majority (38 transcripts) were down-regulated, which is consistent with the general repression of transcription by Atf3. Thus, the early transcriptional responses of rods evoked by ⌬1-83 PhLP-FLAG appeared to be specifically focused on the suppression of the Gnat1 gene, encoding the G␣ t1 subunit of transducin, and on the activation of the stress response via Atf3.

⌬1-83
PhLP-FLAG is suppressing G proteins, we identified its protein interactions in rods using a pull-down assay with anti-FLAG-agarose. As a first approach, the proteins in pull-downs were separated by two-dimensional gel electrophoresis, and the resulting protein spots were excised from the gel and identified by MALDI/MS/MS (Fig. 3). We identified seven major spots, present only in the Tg(ϩ) preparation, as the TCP-1␣, TCP-1␤, TCP-1␥, TCP-1⑀, TCP-1, TCP-1, and TCP-1 subunits of the cytosolic chaperonin CCT. As a second approach, the pulldowns from Tg(ϩ) and Tg(Ϫ) preparations were separated side by side by a short run on a conventional one-dimensional SDS gel, and then the entire lane was excised from the gel and cut into progressive fragments. The protein content of each gel fragment was analyzed by LC/MS/MS, using spectral counts to quantify the relative abundance of each target. In order to reduce error, of 4048 proteins identified by this approach, we selected those with minimal spectral counts of 20 and a Tg(ϩ)/ Tg(Ϫ) ratio greater than 2 (Table 2). Again, we found that eight core subunits of CCT were the most abundant proteins in the pull-downs. Combined, these data demonstrate that the chaperonin CCT, implicated in the posttranslational processing of G proteins (2), is a primary binding target of ⌬1-83 PhLP-FLAG in rods.
Other less abundant protein partners of ⌬1-83 PhLP-FLAG were functionally categorized into several groups, including membrane biogenesis and dynamics (Palm, Osbp2, Dstn, Exoc1, and Fis1), protein biosynthesis and degradation (Pmm1, Psmd6, Psmd11, and Cog8), intracellular trafficking (Snx12, Snx1, and Cask), signal transduction (Itgb1, Klraq1, and Dvl2), nucleic acid biosynthesis and regulation (Polr1a, Wdr82, and Med27), metabolism (Aldoc), and cell death (Maged2). It is plausible that some of them may represent substrate proteins trapped inside the CCT folding chamber, whereas others interact with PhLP directly. Indeed, several bona fide clients of CCT from the actin and tubulin families were present in significant amounts in the pull-downs. However, all of them also appeared in the controls; therefore, the specificity of these interactions could not be reliably verified (Fig. 3 and Table 2).
PhLPs Forms a Complex with CCT Preloaded with Nascent G␤-The ⌬1-83 PhLP-FLAG pull-downs from the retina contain significant amounts of G␤ 1 , a subunit of rod transducin and a known client of CCT. This G␤ 1 was not associated with G␥ 1 , indicating that the G␤ 1 is nascent protein (Table 2). Also, ⌬1-83 PhLP-FLAG lacks the first 83 amino acids, which include several points of interaction with G protein ␤ subunit (14), and these proteins do not co-precipitate in the in vitro assay (31). Thus, this G␤ 1 appears to be client protein trapped inside the folding chamber of the chaperonin co-precipitated with  t test). Bottom graph, the same operation was performed to compare the levels of corresponding mRNA determined by quantitative RT-PCR. The levels of mRNA are 12 Ϯ 2% for G␣ t1 , 60 Ϯ 7% for G␥ 1 , 54 Ϯ 9% for G␤ 1 , and 63 Ϯ 8% for G␤ 5 (mean Ϯ S.E., n ϭ 3).

⌬1-83
PhLP-FLAG. The abundance of G␤ 1 in the pull-downs indicates that ⌬1-83 PhLP-FLAG forms a stable complex with those CCT that are preloaded with G␤. To test whether ⌬1-83 PhLP-FLAG binding to CCT is indeed stimulated by G␤, we utilized cell culture. In these experiments, ⌬1-83 PhLP-FLAG or PhLP-FLAG was transiently expressed in HEK 293 cells, with and without G␤ 1 -myc (Fig. 4). Both proteins were captured from cell extracts with an excess of anti-FLAG magnetic beads, and the resulting pull-downs were assayed by Western blotting. We found that the endogenous CCT of HEK 293 cells robustly co-precipitated with either PhLP-FLAG or ⌬1-83 PhLP-FLAG, and when G␤ 1 -myc was overexpressed in the cells, this protein was present in the pull-downs as well (Fig. 4B). Importantly, overexpressing G␤ 1 -myc markedly increased the amount of ⌬1-83 PhLP-FLAG and co-precipitated CCT while having no such effect on PhLP-FLAG (Fig. 4C). These data further support the notion that ⌬1-83 PhLP-FLAG forms a stable tertiary complex with the chaperonin CCT preloaded with nascent G␤.

PhLP but Not PhLPs Stimulates the Assembly of G␤␥ Subunit
Complexes-As described above, ⌬1-83 PhLP-FLAG specifically targets chaperonin complexes engaged in the folding of G protein ␤ subunits. Such a mode of action, together with the significant reduction in the level of G␥ 1 observed in rods (Fig. 2B), was consistent with ⌬1-83 PhLP-FLAG disrupting the assembly of G␤ 1 and G␥ 1 subunits in these cells. To test this possibility directly, we studied the assembly of G␤␥ dimers in HEK 293 cells from transiently expressed, epitope-tagged G␤ 1 -myc and G␥ 1 -HA (Fig. 5A). Specifically, we sought to determine how this process was affected by the overexpression of PhLP-FLAG or ⌬1-83 PhLP-FLAG in the same cells (Fig. 5B). Our assay was based on capturing the majority of expressed G␥ 1 -HA with an excessive amount of anti-HA-agarose and detecting the co-purified G␤ subunits by Western blotting (Fig. 5C). We found that exogenous G␥ 1 -HA was successfully processed by the endogenous G␤␥ folding pathway, as evident from its forming indiscriminately a dimer with both G␤ 1 -myc and other G␤ subunits of the HEK 293 cells (Fig. 5C, lane 2). This process was suppressed by ⌬1-83 PhLP-FLAG (Fig. 5C, compare lanes 2 and 3). Conversely, PhLP-FLAG strongly stimulated the association between G␥ 1 and G␤ (Fig. 5C, compare lanes 2 and 4). On average, overexpressing PhLP-FLAG in cells increased the  amounts of G␥ 1 -HA in the pull-downs by 4.8 Ϯ 2.6-fold and the co-precipitated G␤ 1 -myc and endogenous G␤ subunits by 7.1 Ϯ 2.7-and 2.5 Ϯ 0.9-fold, respectively (means Ϯ S.E.; n ϭ 5) (Fig. 5D). These data demonstrate that PhLP-FLAG markedly stimulates the assembly of G␤␥ subunit complexes in the cells, whereas ⌬1-83 PhLP-FLAG shows a tendency to suppress this process. Hence, the observed decline of G␥ 1 in rods was probably caused by a major disruption in G␤ 1 ␥ 1 dimer assembly by ⌬1-83 PhLP-FLAG preventing the folded G␤ 1 from being released from CCT.
Binding of the ␥ Subunit Triggers the Release of PhLP from G␤-Having observed the robust stimulatory effect of PhLP on G␤␥ dimer formation, we next tested whether PhLP and G␤␥ are engaged in direct interaction. In these experiments, HEK 293 cells were also co-transfected with phosducin, a close PhLP homolog abundantly expressed in rod photoreceptors (Fig. 6A). When G␥ 1 -HA was captured from the cells, we again observed that it was assembled with G␤ 1 -myc as well as endogenous G␤ subunits, and the amount of these complexes in the cells increased in the presence of PhLP-FLAG (Fig. 6B, compare  lanes 2 and 3). On average, the amounts of G␥ 1 -HA in the pull-downs increased by 3.6 Ϯ 0.8-fold (p Ͻ 0.05), and the amounts of co-precipitated G␤ 1 -myc and endogenous G␤ subunits increased by 3.3 Ϯ 0.5-fold (p Ͻ 0.05) and 1.4 Ϯ 0.6-fold, respectively (means Ϯ S.E.; n ϭ 3). Surprisingly, no detectable amounts of PhLP-FLAG were retained by these newly assembled G␤⅐␥ 1 -HA complexes, which instead contained substantial amounts of bound phosducin (Fig. 6B). Moreover, when PhLP-FLAG itself was pulled down from the same cells with anti-FLAG-agarose, it co-precipitated with G␤, which was essentially free from G␥ 1 -HA (Fig. 6C). These data support a model according to which PhLP interacts exclusively with nascent G␤ that has already been folded by CCT, perhaps to keep this protein in a conformation optimal for assembly with G␥.  The ensuing association of G␤ with G␥ triggers PhLP release, and from now on, the protection of the newly assembled G␤␥ dimer becomes a prerogative of a different molecular chaperone, phosducin. As our model predicts, the amount of G␤ coprecipitated with PhLP-FLAG decreased when the cells were co-transfected with G␥ 1 -HA (Fig. 6D). Among two types of G␤-containing complexes captured in this assay, G␤⅐CCT⅐ PhLP-FLAG and G␤⅐PhLP-FLAG, only the second type is expected to dissociate in the presence of increasing amounts of G␥ 1 -HA. The degree of the observed G␤ reduction indicates that G␤⅐CCT⅐PhLP-FLAG is significantly more abundant in the cell than a transient G␤⅐PhLP-FLAG.

DISCUSSION
These studies reveal a mechanism whereby the cellular levels of heterotrimeric G proteins are controlled by two splice isoforms of phosducin-like protein (Fig. 7). We demonstrate that a short transcript produced by the alternative splicing of the phosducin-like protein gene, PhLPs, previously found in rat brain (28), human retina (29), and chromaffin cells of bovine adrenal gland (36), is widely expressed in normal human tissues at levels typically lower than those of PhLP. Seeking to understand the biological significance of PhLPs, we expressed this protein in the rod photoreceptors of transgenic mice and found that this manipulation ablated the expression of the rod major heterotrimeric G protein, transducin (G␣ t1 ␤ 1 ␥ 1 ). All transducin subunits were affected on either transcriptional or posttranslational levels. The decrease of G␣ t1 was primarily caused by a dramatic and rather specific down-regulation of its transcript, although a mutual posttranslational destabilization due to the G␤ 1 ␥ 1 deficiency in these cells, similar to that observed in G␥ 1 -null mice (37,38), may also contribute to the reduction. The molecular mechanism of this negative transcriptional feedback remains unknown and will be the subject of future studies. The decline of two other transducin subunits, G␤ 1 and G␥ 1 , and a remarkable propensity of PhLPs toward the chaperonin CCT responsible for the posttranslational stabilization of these proteins, suggests that PhLPs acts as a CCT suppressor. This notion gains additional support from our cell culture studies, demonstrating that, in sharp contrast to PhLP, which markedly stimulates the dimerization of G␤ and G␥ subunits, PhLPs suppresses this process. This observation, thus, further corroborates the notion that PhLP acts as a co-factor of CCT, facilitating the assembly of the G␤␥ subunit complex (2), whereas PhLPs competitively inhibits this function. According to the currently accepted model, a permanent association of the G protein ␤ and ␥ subunits, via a high affinity coiled-coil interaction (39), is required for their posttranslational stabilization in cells (4) and occurs subsequent to the folding of G␤ by CCT (2). PhLP may potentially facilitate this interaction by rendering the newly folded G␤ competent to bind G␥. As evident from the lack of any significant co-precipitation of PhLP and G␥ 1 observed in all of our experiments, binding of G␥ 1 apparently triggers the release of PhLP from G␤ 1 and also serves as a signal precluding the assembled G␤ 1 ␥ 1 dimer from re-entering CCT. All of these latter steps in the posttranslational processing of G␤ are inhibited by PhLPs, which binds CCT preloaded with nascent G␤ with a high affinity and, being unable to productively assist with the release of G␤, traps this protein in the folding chamber of the chaperonin. Such a mode of action is consistent with a broad suppression of G␤, which was indeed observed in rods unable to maintain a normal level of G␤ 5 L.
The only physiological function of PhLPs demonstrated thus far was its ability to inhibit Ca 2ϩ -induced exocytosis in chromaffin cells (36). Our data provide the first evidence that PhLPs can also effectively regulate the expression of heterotrimeric G proteins in vivo. What could be the physiological role of PhLPs in photoreceptors? At this point, we can only speculate that the effects of PhLPs would become significant only when its protein level is up-regulated through alternative mRNA splicing. If such regulation does occur in the photoreceptors under a stress condition (e.g. excessive light stimulation), the ensuing temporal blockage of heterotrimeric G protein-mediated visual signaling and perhaps Ca 2ϩ -mediated membrane trafficking may play a protective role, increasing the chance of photoreceptors to survive under this condition. Determining whether such a stress response, encompassing global suppression of extracellular signaling, is widespread among eukaryotes will remain the subject of future studies. PhLP stimulates the subsequent steps of this process by facilitating the assembly of processed G␤ 1 with G␥ 1 (green arrow). PhLPs inhibits these steps by blocking the release of G␤ 1 from CCT and also suppresses the transcription of G␣ t1 subunit (red arrow).