Targeted Knockdown of G Protein Subunits Selectively Prevents Receptor-mediated Modulation of Effectors and Reveals Complex Changes in Non-targeted Signaling Proteins*

Heterotrimeric G protein signaling specificity has been attributed to select combinations of Gα, β, and γ subunits, their interactions with other signaling proteins, and their localization in the cell. With few exceptions, the G protein subunit combinations that exist in vivo and the significance of these specific combinations are largely unknown. We have begun to approach these problems in HeLa cells by: 1) determining the concentrations of Gα and Gβ subunits; 2) examining receptor-dependent activities of two effector systems (adenylyl cyclase and phospholipase Cβ); and 3) systematically silencing each of the Gα and Gβ subunits by using small interfering RNA while quantifying resultant changes in effector function and the concentrations of other relevant proteins in the network. HeLa cells express equimolar amounts of total Gα and Gβ subunits. The most prevalent Gα proteins were one member of each Gα subfamily (Gαs, Gαi3, Gα11, and Gα13). We substantially abrogated expression of most of the Gα and Gβ proteins expressed in these cells, singly and some in combinations. As expected, agonist-dependent activation of adenylyl cyclase or phospholipase Cβ was specifically eliminated following the silencing of Gαs or Gαq/11, respectively. We also confirmed that Gβ subunits are necessary for stable accumulation of Gα proteins in vivo. Gβ subunits demonstrated little isoform specificity for receptor-dependent modulation of effector activity. We observed compensatory changes in G protein accumulation following silencing of individual genes, as well as an apparent reciprocal relationship between the expression of certain Gαq and Gαi subfamily members. These findings provide a foundation for understanding the mechanisms that regulate the adaptability and remarkable resilience of G protein signaling networks.

Signal-transducing heterotrimeric G proteins are associated with the inner face of the plasma membrane, positioned as middlemen for activation by membrane-spanning, heptahelical receptors, and regulation of a variety of intracellular effectors. Interactions among these proteins are controlled by agonist-induced changes of receptor conformation and nucleotide-driven conformational changes of the ␣ subunits of the G proteins (G␣). A ligand-bound receptor catalyzes the exchange of GDP for GTP on a cognate G␣, and as a result, the (at least partial) dissociation of G␣ from a complex of G␤ and G␥ subunits. These acti-vated subunits are then capable of modulating the functional properties of effector proteins (e.g. adenylyl cyclases and phospholipases). The intrinsic GTPase activity of G␣ serves as a molecular timer, returning the protein to the GDP-bound state and allowing reformation of the inactive heterotrimer.
Much remains to be learned about the specificity of G protein signaling in vivo, the relative importance of isoforms of G protein subunits with apparently redundant functions, and the qualitative and quantitative significance of the fact that many hundreds of G protein heterotrimers can be assembled from the collection of G protein ␣, ␤, and ␥ subunits that are expressed in single cells. There is evidence for exquisite specificity of signaling through certain pathways. For example, intranuclear injection of antisense oligonucleotides against specific G protein subunits revealed that the M4-muscarinic receptor-mediated inhibition of L-type Ca 2ϩ channels requires G␣ oa , G␤ 3 , and G␥ 4 , whereas similar inhibition initiated by somatostatin receptors requires G␣ ob , G␤ 1 , and G␥ 3 subunits (1-3). The generality of these and related studies has not been examined, and there has been little effort to monitor and understand the compensatory mechanisms that such perturbations may set in motion. Furthermore, studies performed in vitro do not reveal such demanding specificity, and mechanisms of such phenomena are not known.
Herein we describe a more comprehensive attempt to examine these issues. We have sought information about the expression of most G protein subunits in a clonal human cell line (HeLa), and we have examined the functional and compensatory effects of siRNA 2 -mediated silencing of the expression of genes encoding members of the G␣ s , G␣ i , G␣ q , and G␤ subfamilies of G protein subunits.

EXPERIMENTAL PROCEDURES
Reagents-All reagents were purchased from Sigma unless noted otherwise.
Mammalian Cell Culture-HeLa cells (from ATCC) were cultured at 37°C in Dulbecco's modified Eagle's medium with high glucose (Invitrogen) supplemented with 10% fetal bovine serum under an atmosphere of 95% air, 5% CO 2 . Cells were cultured in the same lot of serum and passaged twice weekly by trypsinization. Fresh cultures were established from the same frozen stock after 10 passages.
Transient RNA Interference (RNAi) Transfection-Single-stranded 21-mer oligonucleotides (containing 19 ribonucleotides and 2 3Ј-deoxythymidine residues) targeting the open reading frames of selected proteins were designed using the Dharmacon siRNA Design Center. Candidate sequences were subjected to further BLAST analysis against the human genome data base (NCBI) and selected for further study if no more than 14 contiguous bases were identical to coding regions of known human gene sequences. Single-stranded oligonucleotides were annealed as described (4). The sequences of the RNAi sense strands and the targeted sites in the open reading frames are shown in Table 1. Transient transfections of RNAi duplexes were accomplished using Oligofectamine transfection reagent (Invitrogen) according to manufacturer's instructions. Briefly, HeLa cells were seeded at 44 ϫ 10 4 /100-mm dish 24 h before transfection. The cells were transfected using 19 l of Oligofectamine and a total of 190 pmol of RNAi duplex(es) in a final volume of 3.8 ml. Two days later the cells were trypsinized, counted, plated at 55 ϫ 10 4 /100-mm dish, and subjected to a second transfection. One day after the second transfection the cells were trypsinized, pooled, and distributed to 12-well plates. Three days after the second transfection, the cells were used for experiments or harvested for immunoblot analyses and determinations of protein concentration.
Total Inositol Phosphate Accumulation-Total inositol phosphate (IP) accumulation was determined by incubating HeLa cells for 48 h with 10 Ci/ml of myo- [2-3 H]inositol (1 mCi/ml, PerkinElmer Life Sciences) in inositol-free Dulbecco's modified Eagle's medium (Specialty Media, Phillipsburg, NJ) supplemented with 5% fetal bovine serum. Cells were incubated for 20 min at 37°C with bicarbonate-free Dulbecco's modified Eagle's medium, 20 mM Na-HEPES (pH 7.4) containing 20 mM LiCl prior to the initiation of ligand-dependent IP synthesis. Reactions were terminated by addition of ice-cold 10 mM formic acid to precipitate protein. The soluble lysate was applied to a 1-ml Dowex AG (1-X8) column, which was washed successively with 15 ml of 10 mM formic acid containing 10 mM myoinositol and 15 ml of 5 mM Na-tetraborate, 60 mM ammonium formate, prior to elution with 2.0 ml of 1.0 M ammonium formate, 0.1 M formic acid. Total [ 3 H]IP was determined by scintillation counting and normalized to the amount of acid-precipitated protein. Protein was quantified as described by Bradford (Bio-Rad) (6).
Radioligand Binding-␤2-Adrenergic receptor concentrations were estimated for cells seeded in 6-well plates using 2 nM [ 125 I]CYP (PerkinElmer Life Sciences) in the presence or absence of the ␤2-adrenergic receptor-selective antagonist ICI-118,551 (10 M) or the nonselective ␤-adrenergic receptor antagonist timolol (10 M). Each measurement was conducted in triplicate. Ligand binding reactions were quenched in ice-cold 50 mM Tris-HCl (pH 7.4), 10 mM MgCl 2 , and 100 mM NaCl and immediately filtered through Whatman GF/C filters. Radioactivity was quantified by scintillation counting. Duplicate 6-well plates were cultured to permit determinations of cell number and total protein content.
Recombinant G␣ and G␤ Proteins-Recombinant bovine G␣ s,short , G␣ i1 (rat), G␣ i2 (rat), and G␣ i3 (rat) were purified after expression in Escherichia coli as described (7); the G␣ i proteins were coexpressed with protein N-myristoyl transferase and were thus myristoylated at their amino termini. Sf9 cells were coinfected with baculoviruses encoding human G␣ s,long , mouse G␣ 11 , or mouse G␣ 13 and baculoviruses encoding G␤1 and His 6 -G␥2, and the G proteins were isolated as described (8). Purified mouse G␣ q and G␣ 12 were gifts from G. Tall and P. Sternweis, respectively (University of Texas Southwestern Medical Center). Recombinant bovine G␤ 1 and human G␤ 2 , G␤ 3 , and G␤ 4 proteins were provided by S. Gibson (University of Texas Southwestern Medical Center). Protein concentrations were determined using an Amido Black staining assay (9). Purity was Ͼ90% for all proteins as assessed by Coomassie Blue-stained SDS-PAGE gels.
Plasma membrane fractions from HeLa cells were prepared essentially as described (10). Briefly, cells were disrupted by nitrogen cavitation (600 p.s.i. for 30 min). Nuclei and unbroken cells were removed by centrifugation for 5 min (600 ϫ g, 4°C). The resulting supernatant fraction was layered onto a 23 and 43% sucrose step gradient and centrifuged (100,000 ϫ g, 1 h, 4°C, Beckman SW 41 rotor). The enriched plasma membranes were collected at the gradient interfaces and recovered by centrifugation (200,000 ϫ g, 30 min, 4°C, Beckman TLA 100.3 rotor). The membranes were suspended in 20 mM Na-HEPES (pH 7.4), 150 mM NaCl, flash frozen in liquid N 2 , and stored at Ϫ80°C. Protein concentrations for total lysates and membranes were determined with the Amido Black protein assay (9) using bovine serum albumin as standard.
Antibody specificities, dilutions, and sources are shown in Table 1. Following incubation with primary antibody, immunoblots were washed with phosphate-buffered saline containing 0.1% Tween 20. Secondary anti-rabbit or anti-mouse antibodies conjugated to horseradish peroxidase were diluted in wash buffer and incubated with the blots for 1 h. The blots were subsequently washed and immunoreactive bands were detected by enhanced chemiluminescence (PerkinElmer Life Sciences).
Densitometry-Immunoblots were scanned into Adobe Photoshop 5.0, and the pixel intensities (arbitrary units) of the immunoreactive bands were quantified. For quantitative immunoblots, endogenous HeLa cell protein levels were interpolated from the pixel intensities of the known amounts of purified G␣ or G␤ subunit standards loaded on the same gel. For relative protein determinations, the pixel intensity of immunoreactive bands in samples prepared from RNAi-transfected cells was normalized to the corresponding immunoreactive band intensity identified in an equal amount of lysate prepared from mock-transfected cells.
Quantitative Real-time PCR-Relative quantification of mRNA levels was achieved using quantitative RT-PCR. Single component amplification reactions were conducted using the relative standard curve or comparative C t methods and TaqMan Gene Expression Assays (Applied Biosystems) according to manufacturer's instructions. Briefly, a TaqMan gene expression probe specific for amplifying a particular G␣ or G␤ subunit cDNA sequence was mixed together with a One-Step TaqMan RT-PCR master mixture and 100 ng of total RNA/reaction. Total RNA was isolated from HeLa cells using the RNAqueous-4PCR kit (Ambion) and quantified spectrophotometrically (A 260 ). Fluorescence data were collected following each amplification cycle using the ABI PRISM 7500 Sequence Detection System (Applied Biosystems). C t values were calculated from threshold values set to manual.
Data Analysis-Differences between measured values (mean Ϯ S.D.) were determined using a one-tailed t test or single analysis of variance (p Ͻ 0.05, InStat, Prism).

Endogenous G Protein Subunits in HeLa
Cells-Ten proteins were identified as products of the 16 mammalian G␣ genes by both RT-PCR (data not shown) and immunoblotting (see control lanes in Fig. 1A): G␣ s (long and short splice variants), G␣ q , G␣ 11 , G␣ i1 , G␣ i2 , G␣ i3 , G␣ o , G␣ 12 , and G␣ 13 . All positive RT-PCR signals were confirmed by cDNA sequencing, and antibody specificity was verified by target-specific RNAi-mediated silencing of antigens (also shown in Fig. 1A). Transcripts of two additional G␣ genes, G␣ z and G␣ 16 , were identified by RT-PCR, but the corresponding proteins were not detected despite multiple attempts to blot 100 g of enriched plasma membrane fractions with sensitive antibodies (detection limit of Ͻ1 ng of purified protein). No mRNA or protein was detected for G␣ 14 . We did not address possible expression of G␣ subunits that are primarily found in specialized sensory tissues (G␣ olf , G␣ t (1 and 2), and G␣ g ).
Of the five G␤ genes, we detected expression of four in HeLa cells: G␤ 1 , G␤ 2 , G␤ 4 , and G␤ 5 ; we found no evidence for expression of G␤ 3 . Because G␤ 5 subunits appear to form selective heterodimers with RGS proteins of the R7 family (which contain G␥-like domains) (12) and there is little credible evidence for the existence in vivo of heterodimers containing G␤ 5 and G␥ subunits, we focused our efforts on G␤ 1 , G␤ 2 , and G␤ 4 subunits (Fig. 1B).
We also examined the expression of G␥ subunits in HeLa cells. Perhaps surprisingly, RT-PCR analysis positively identified transcripts of all 12 G␥ genes, including G␥ 1 , whose expression is largely restricted to the Positions of the long (52 kDa) and short (46 kDa) G␣ s splice variants, as well as immunodetectable bands corresponding to other G␣ subunits, are indicated by lines. G␣ q (42 kDa) and G␣ i1 (40 kDa) are each shown migrating below a nonspecific immunoreactive band. Equal amounts (30 g) of total lysates were loaded in each lane of gels used for Western immunoblotting of G␣ subunits. G␣ 12 antiserum did not cross-react with purified G␣ 13 standard (not shown). B, Western blots of endogenous G␤ subunits in total lysates of unperturbed HeLa cells. G␤ subunits targeted for RNAi silencing are shown across the top of each panel, and proteins identified by Western blotting are indicated at the left. G␤ 1 migrated as a 36-kDa protein, whereas G␤ 2 and G␤ 4 migrated as 35 kDa. Thirty g of each total lysate was loaded per lane for individual G␤ isoform examination by Western blotting and 5 g/lane of each total lysate was loaded for Western blotting examination of total G␤ 1-4 subunit expression. C, results of quantitative Western blotting for G␣ and G␤ subunits in unperturbed HeLa total cell lysates. Five to 30 g/lane of each lysate preparation was separated by SDS-PAGE along with purified G␣ or G␤ subunit standards (1-10 ng range). Western blots and densitometry analysis were conducted as described under "Experimental Procedures." Each bar represents the mean Ϯ S.D. of 10 different total cell lysates. G␣ protein amounts are represented by open bars and G␤ proteins by closed bars.
retina (13). Because of the lack of specific and sensitive antisera to most G␥ proteins, we have limited our studies to date to G␣ and G␤.
Prior to examining the effects of silencing individual G protein subunits, we quantified the amounts of G␣ and G␤ proteins in total lysates from control cells (Fig. 1C). Quantitative immunoblotting revealed that 80% of the entire G␣ subunit protein pool was accounted for by one member of each of the four subfamilies of G␣ proteins: G␣ i3 , G␣ 11 , G␣ 13 , and the sum of long and short splice variants of G␣ s . G␤ 1 and G␤ 2 were expressed at similar levels and together accounted for 80% of the total G␤ subunit pool; G␤ 4 comprised the remaining 20%. There was no significant difference between the total amount of G␤ protein detected with a pan G␤ antibody (G␤ [1][2][3][4] ) and the sum of G␤ subunits calculated from data obtained using isoform-specific G␤ antibodies. Transformation of the data in Fig. 1C to molar quantities reveals that the sum of the G␣ subunits (22 Ϯ 2 pmol/mg lysate) closely approximates that of the G␤ proteins (21 Ϯ 3 pmol/mg of lysate), consistent with the notion that most or all of these subunits exist as oligomers.
Knockdown of G Protein Subunits-Immunoblots showed remarkable (more than 90%) knockdown of endogenous protein expression for each G␣ and G␤ subunit (Fig. 1, A and B) after silencing by siRNAs, either individually or in combination. Greater than 90% knockdown of G␣ subunits was deemed necessary to test changes in receptor-modu-lated effector responses, because the reported stoichiometry of G protein to effector can, in some instances, exceed 10:1 (14). Thus, multiple RNAi duplexes (range: 2 to 10) were designed and generated toward each target; in general, most duplexes silenced target protein expression more than 75%.
In several instances (particularly with G␤ subunits) treatment with RNAi oligonucleotides not only silenced the intended target but also caused cell death. Interestingly, our most lethal RNAi duplex targeted green fluorescent protein using a sequence with no significant (Ͼ12 nucleotide) overlap with the open reading frame of any known human gene, indicating that cell death may be caused by mechanisms unrelated to specific mRNA sequence recognition by RISC complexes. We rigorously avoided using siRNAs that had adverse effects on cell growth or viability.
Knockdown of G␣ s : Anticipated Effect on Ligand-stimulated Activation of Adenylyl Cyclases-G␣ s proteins mediate activation of adenylyl cyclases by so called G s -coupled G protein-coupled receptors. As anticipated, knockdown of G␣ s completely abrogated ligand-dependent increases in cAMP synthesis (Fig. 2, A and B) stimulated by either INE, a ␤-adrenergic receptor agonist, or PGE1, an EP prostanoid receptor agonist. A combination of RT-PCR, radioactive ligand-binding assays, and RNAi-mediated silencing of cell surface receptors (see Table 1) demonstrated that INE stimulation of cAMP synthesis occurred exclu- sively via the ␤2-adrenergic receptor, whereas PGE1 signaling was mediated via two EP receptor isoforms, EP2 and EP4 (data not shown). We found no evidence of expression of ␤1or ␤3-adrenergic receptors or EP1 or EP3 prostanoid receptors in HeLa cells. The functional effects of loss of G␣ s were specific. Thus, silencing of G␣ s did not prevent UK 14,304-dependent inhibition of cAMP synthesis or stimulation of phospholipase C␤ activity by histamine (data not shown). These data con-firm the functionality and specificity of G␣ s -mediated activation of adenylyl cyclases using RNAi-mediated gene suppression and illustrate the utility of this approach to study G protein signaling in a population of intact cells.
Silencing of G␣ q and G␣ 11 : Anticipated Effects on Activation of Phospholipase C␤-G␣ q and G␣ 11 (along with G␣ 14 and G␣ 16 ) mediate cell surface receptor activation of PLC␤, which catalyzes the synthesis of inositol (1,4,5)-trisphosphate from phosphatidylinositol (4,5)-bisphosphate. Inositol phosphate (IP) accumulation in HeLa cells was increased following activation of the H1-histamine receptor by agonist. The simultaneous knockdown of G␣ q and G␣ 11 (denoted as G␣ q/11 ) eliminated histamine-stimulated IP accumulation (Fig. 2C). Interestingly, the silencing of either G␣ q or G␣ 11 individually decreased histamine-dependent IP accumulation by ϳ50%, despite the fact that concentrations of G␣ 11 in HeLa cells exceed those of G␣ q by 10-fold (Fig. 1C). The amount of [ 3 H] incorporated into acid-precipitable material was unchanged in cells lacking G␣ q/11 , suggesting that phosphatidylinositol kinase and phosphatase activities were unaffected. Silencing G␣ q/11 modestly (but significantly) increased ligand-dependent cAMP responses (ϳ30%) without affecting UK 14,304-mediated inhibition of adenylyl cyclase (data not shown).

Knockdown of G␣ i/o : Incomplete Effects on Inhibition of Adenylyl
Cyclase-G␣ i/o subunits transduce signals from cell surface receptors to a variety of intracellular pathways (see Ref. 15 for review), including inhibition of adenylyl cyclase activity. cAMP synthesis in HeLa cells can be inhibited by ␣2aand ␣2b-adrenergic receptors. 3 Despite silencing each of the G␣ i1-3 subunits, individually or in combinations of two, we observed no significant change in the ability of the synthetic ␣2-adrenergic receptor agonist UK-14,304 to inhibit INE-(data not shown) or forskolin-stimulated cAMP synthesis (Fig. 2D). We observed a 60% reduction in the magnitude of such inhibition only when all three G␣ i subunits were silenced (denoted G␣ i1-3 ) (Fig. 2D). Thus, ␣2-adrenergic receptors displayed no absolute specificity for G␣ i1-3 isoforms for the inhibition of INE or forskolin-stimulated cAMP synthesis.
Interestingly, knockdown of G␣ i1 did specifically interfere with UK 14,304-mediated inhibition of PGE1-stimulated cAMP synthesis (Fig.  2E), suggesting that a PGE1-stimulated adenylyl cyclase may preferentially localize with or couple to this G protein. It is most puzzling, therefore, that simultaneous silencing of G␣ i1-3 subunits did not abrogate inhibition of PGE1-activated cAMP synthesis by UK 14,304. Residual inhibition of adenylyl cyclase activity in all experiments was eliminated by treatment of cells with pertussis toxin, suggesting that the G␣ i subunit expression may have been insufficiently silenced and/or that G␣ o subunits, in the absence of G␣ i1-3 subunits, may have contributed to the inhibition of cAMP synthesis (16). Individual silencing of G␣ o (both a and b splice variants) had no effect on UK-14,304-mediated inhibition of cAMP synthesis. Recall that G␣ z was not detected in these cells.
Silencing of G␣ Subunits: Alterations in Accumulation of Other G␣ Proteins-Silencing of individual G protein ␣ subunits routinely resulted in loss of more than 90% of that protein, detected by immunoblotting ( Table 2, highlighted in yellow). Quantification of non-targeted G␣ proteins revealed that the concentrations of G␣ i3 and G␣ 11 (Table 2) and G␣ 13 (data not shown) were not significantly altered by such perturbations. However, we observed modest (ϳ2-fold) increases in concentrations of G␣ i1 protein following knockdown of G␣ i2 or G␣ i3 ( Table  2, highlighted in green) and increased G␣ i2 protein after silencing of G␣ i3 . Thus, loss of the major isoform of G␣ i resulted in increased accumulation of the other two isoforms of G␣ i . Increased accumulation of G␣ i1 protein was associated with a dramatic increase in detectable mRNA for this protein (Fig. 3A). The level of G␣ o protein also increased modestly, but significantly, after combined knockdown of all three G␣ i subunits ( Table 2). These data indicate that reductions in the concen-3 S. K. Gibson, unpublished observations.

TABLE 2
Relative G␣ and G␤ protein expression following select G protein subunit silencing G protein subunits targeted for siRNA silencing are shown across the top of the table and proteins identified by Western blotting are indicated at the left. Western blots and densitometry analyses were conducted as described under "Experimental Procedures." Each value corresponds to the pixel intensity of a particular antigen identified in siRNA-transfected total cell lysates normalized against the pixel intensity obtained for the same antigen in unperturbed control cell lysates loaded on the same gel. Data are represented as the mean Ϯ S.D. for a minimum of three separate total lysate preparations. Expression data were analyzed using single-sided analysis of variance (p ⌬ 0.05) comparison of log (base 2) transformation of relative antigen expression for a particular perturbation against log transformation of the variance of protein expression for the same antigen determined from multiple control lysate preparations (range 5-12). Effects of siRNA transfection directed at the antigen are highlighted in yellow. Significant increases in non-targeted protein expression following a knockdown of G protein subunits are highlighted in green, and significant decreases are highlighted in red.
trations of G␣ i isoforms are associated with compensatory changes in the levels of other proteins within the G␣ i subfamily.
Unexpectedly, we detected substantially increased concentrations of G␣ i1 and G␣ o protein following the knockdown of either G␣ q , G␣ 11 , or both of these proteins (denoted as G␣ q/11 ) (Fig. 3B and Table 2, highlighted in green). These changes in protein concentrations were accompanied by increases in mRNA for both G␣ i1 and G␣ o (Fig. 3A). Taken together with the fact that the absolute amounts of G␣ 11 and G␣ q are so different (and observed effects are thus less likely to be caused by alterations of ratios of G␣ to G␤␥), these data suggest that G␣ q/11 proteins regulate (directly or indirectly) the rate of synthesis of G␣ i1 and G␣ o . Although knockdown of G␣ q and G␣ 11 increased accumulation of G␣ i1 and G␣ o , the reciprocal was not generally true. However, a modest increase in G␣ q protein level was observed following knockdown of G␣ i3 (Fig. 3B and Table 2, highlighted in green). Amounts of G␣ 12 mRNA were also increased about 4-fold in cells lacking G ␣q/11 protein (data not shown), but the effects on protein concentrations were not examined because of lot variability of the anti-G␣ 12 antibody.
G␣ i1 stands out in examination of all of the effects of G␣ protein knockdowns on G protein expression. Although apparently a minor player in HeLa cells in terms of protein concentration (Fig. 1C), levels of G␣ i1 message and protein increased significantly when expression of G␣ s , G␣ q , G␣ 11 , G␣ i2 , or G␣ i3 was compromised (Fig. 3, A and B, and Table 2).
We also examined the effects of silencing G␣ protein expression on G␤ subunits ( Fig. 3C and Table 2). Interesting and in some cases specific effects were observed, although they were not predictable by consideration of the relative concentrations of the proteins involved or other facts known to us. Thus silencing of G␣ s , G␣ 11 , or individual G␣ i isoforms (G␣ i2 and G␣ i3 ) significantly decreased accumulation of G␤ 4 , the least prevalent G␤ subunit. Knockdown of G␣ s also decreased levels of G␤ 2 , whereas silencing G␣ i2 decreased levels of G␤ 1 substantially, while having only modest effects on other G␤ proteins. The simultaneous knockdown of the three G␣ i proteins (denoted as G␣ i1-3 ) decreased the total immunodetectable G␤ subunit pool by more than 50%. Decreased accumulation of G␤ protein was associated with decreased G␤ mRNA levels following knockdown of G␣ s and G␣ i3 . However, the effect of G␣ i2 knockdown was not correlated with changes in levels of G␤ mRNA.
Knockdown of Single and Paired G␤ Isoforms: Alterations in Expression of Other G Protein Subunits-G␤ subunits were targeted for silencing in attempts to observe the coordinated disappearance of combinations of G␣ and G␤ subunits that might be prevalent in vivo, isoformspecific functions of G␤ subunits, and regulation of G␤ subunit expression. siRNA-mediated targeting of G␤ 1 , G␤ 2 , or G␤ 4 resulted in loss of more than 90% of immunodetectable protein (Fig. 1B and Table  2, highlighted in yellow). No dramatic changes were observed in nontargeted G␤ proteins or mRNA (not shown) following the silencing of . Relative mRNA and protein expression of G protein subunits following select G␣ silencing. A, quantitative PCR were conducted using total RNA isolated from control cells and cells lacking G␣ s , G␣ q , G␣ 11 , G␣ i1 , G␣ i2 , G␣ i3 , or G␣ o . Relative amounts of G␣ subunit mRNA expression (G␣ s (black), G␣ q (pale blue), G␣ 11 (orange), G␣ 13 (purple), G␣ i1 (pink), G␣ i2 (pale green), G␣ i3 (yellow), and G␣ o (blue)) are shown for cells lacking selected G␣ protein expression (shown across the bottom). -Fold differences were calculated using the D⌬C t method. C t values for each probe were corrected for the amount of endogenous 18 S rRNA in each sample and normalized using C t values determined for each probe in unperturbed cell total RNA. Each bar represents the mean Ϯ S.D. from three separate total RNA preparations. B, Western blots of G␣ o , G␣ i1 , and G␣ q following knockdown of select G␣ subunits. G␣ proteins targeted for RNAi silencing are shown across the top of each panel and proteins identified by Western blot are indicated to the left of each panel. Migration of G␣ q , indicated by the line, is below a nonspecific band. The data are representative of at least three separate experiments. C, relative G␤ mRNA levels following knockdown of G␣ subunits. The relative G␤ subunit mRNA expression (G␤ 1 (pale red), G␤2 (green), and G␤ 4 (brown)) is shown using the same total RNA preparations described in A. Each bar represents the average Ϯ S.D. of three separate preparations.
individual G␤ proteins, although G␤ 1 and G␤ 2 levels appeared to rise modestly following silencing of G␤ 4 . No G␤ 3 protein was detected after G␤ 1 , G␤ 2 , or G␤ 4 knockdown. Regulation of expression of individual G␤ isoforms thus appears to occur largely independently of other individual isoforms.
Because G␣ proteins are stabilized by association with G␤␥ dimers (17), we hypothesized that loss of a single G␤ subunit might reduce the level of cognate G␣ proteins. Because three G␤ subunits must "service" 10 G␣ proteins, overlapping relationships are mandatory. Silencing of G␤ 1 resulted in reduced accumulation of G␣ s and G␣ i3 ; knockdown of G␤ 2 decreased G␣ s , G␣ i1 , and G␣ i3 ; and silencing G␤ 4 diminished G␣ q and G␣ o while increasing G␣ i1 and G␣ i2 (Fig. 4A and Table 2). Reduced accumulation of G␣ proteins was not well correlated with changes in mRNA levels ( Fig. 4B and not shown), consistent with the notion that G␤␥ complexes promote stability of G␣. In examining these data, G␣ i1 again stands out. Silencing of any G␤ provoked an increase in G␣ i1 mRNA; this is especially true of G␤ 4 and correlates with the increase in G␣ i1 protein seen with this perturbation. These G␤ subunit-specific effects on accumulation of G␣ protein are consistent with the existence of isoform-specific combinations of G␣ and G␤ subunits in unperturbed HeLa cells.
Silencing of individual G␤ isoforms did not decrease concentrations of G␣ i2 , G␣ 11 , and G␣ 13 , suggesting that G␣ s , G␣ i1 , G␣ i3 , and G␣ o are less stable than G␣ i2 , G␣ 11 , and G␣ 13 in the absence of G␤ or that the latter group more readily redistributes among other G␤ subunits. To shed further light on these questions, we evaluated the effects of simultaneous silencing of multiple G␤ subunits.
Simultaneous knockdown of G␤ 1 and G␤ 2 was accomplished with either of two distinct RNAi duplexes, targeting site 865 (denoted as G␤ 1/2 (a)) or 866 (denoted as G␤ 1/2 (b)), both of which contain 19-nucleotide identities with the G␤ 1 and G␤ 2 open reading frames. Concomitant silencing of G␤ 1 and G␤ 2 eliminated nearly all G␣ i family protein, in contrast with a recent study (18) in which the G␣ i subfamily protein was maintained in a cultured mouse J774A.1 macrophage cell line following the knockdown of G␤ 1 and G␤ 2 .
Surprisingly, G␣ s protein expression was maintained at 50% of control values despite the loss of G␤ 1/2 ( Fig. 4C and Table 2), as were G␣ q , G␣ 11 , G␣ 13 , and G␣ o . Interestingly, each G␣ subfamily was represented despite the knockdown of 80% of the initial G␤ pool. Sustained expression of some G␣ proteins in our G␤ 1/2 ablated cells could not be attributed to greater stability of these free G␣ subunits because expression of nearly all G␣ subunits was lost following the silencing of all G␤ isoforms ( Fig. 4A and Table 2). Decreased G␣ protein concentrations following knockdown of multiple G␤ subunits occurred independently of changes in levels of G␣ subunit mRNA (Fig. 4B), likely indicating that G␤ subunits are requisite for the stability of virtually all G␣ proteins.  13 , G␣ i1 , G␣ i2 , and G␣ i3 using total lysates prepared from unperturbed cells and cells lacking G␤ 1 , G␤ 2 , G␤ 4 , G␤ 1/2 , or G␤ 1/2/4 . The data are representative of four independent preparations for each condition. B, relative G␣ mRNA expression in cells lacking G␤ protein expression. Relative G␣ mRNA amounts (G␣ s (black), G␣ 11 (orange), G␣ 13 (purple), G␣ i1 (pink), G␣ i2 (pale green), and G␣ i3 (yellow)) were determined as described previously using total RNA prepared from cells lacking G␤ 1 , G␤ 2 , G␤ 4 , G␤ 1/2 , and G␤ 1/2/4 . Each bar represents the average Ϯ range of relative mRNA levels calculated from two independent preparations. C, amount of G␣ and G␤ subunit protein expression in total lysates prepared from cells lacking G␤ 1/2 expression. Total cell lysates were prepared from G␤ 1/2 RNAi-transfected cells and quantitative Western blots were obtained for each G␣ and G␤ subunit as described previously. Each bar represents the average Ϯ S.D. from six independent preparations.
The retention of some G␣ subunits in cells lacking the G␤ 1/2 protein was likely because of formation of new associations with increased concentrations of G␤ 4 (Fig. 1B, Table 2, and Fig. 5). No G␤ 3 protein was detected in cells where G␤ 1 and G␤ 2 were eliminated. Increased G␤ 4 protein was detected 48 -72 h after transfection with siRNA directed against G␤ 1/2 until a new steady-state level of G␤ 4 protein was established (compare the relative G␤ 4 level 72 h after transfection (Fig. 5C) with that after 144 h ( Table 2)). Interestingly, there was no detectable increase in G␤ 4 mRNA during the same time period.
Knockdown of Single and Multiple G␤ Subunits: Functional Effects-Silencing of G␤ 1 (by either of two RNAi duplexes) substantially reduced PGE1-dependent activation of cAMP synthesis by 70% (Fig. 6A). Targeted knockdown of G␤ 2 and/or G␤ 4 , on the other hand, had little effect on PGE1-dependent activity. The data indicate that at least one of the EP2 and EP4 prostanoid receptors primarily stimulates cAMP synthesis via G␣ s associated with G␤ 1 in HeLa cells. This interesting example of specificity is perhaps reminiscent of the observation (above) that G␣ i1 was specifically required for ␣2-adrenergic receptor-mediated inhibition of PGE1 (but not INE-or forskolin-)-stimulated adenylyl cyclase activity.
Knockdown of individual G␤ subunits did not affect histamine-dependent activation of PLC␤ (Fig. 6C), and little requirement for specific G␤ isoforms was observed for ␣2-adrenergic receptor-dependent inhibition of ligand-or forskolin-activated cAMP synthesis (Fig. 6, D and E, and data not shown).
We then examined the effects of simultaneous elimination of G␤ 1 and G␤ 2 on ligand-dependent effector activity; these proteins account for 80% of the total cellular G␤ in unperturbed cells. Histamine-dependent accumulation of IP was unaffected (Fig. 6C), whereas ␣2-adrenergic receptor-mediated inhibition of ligand-activated cAMP synthesis was reduced by at least 90% (Fig. 6, D and E). These data indicate that G␤ 4 could fully sustain both the stability and function of the G␣ q subfamily of G␣ s proteins but not the G␣ i subfamily, whose members largely disappear under this condition. Coincidental silencing of all three G␤ subunits largely abrogated ligand-dependent accumulation of cAMP and IP. No G␤ 3 protein was detected in cells lacking G␤ 1 , G␤ 2 , and G␤ 4 .
Very curiously, silencing of both G␤ 1 and G␤ 2 did not cause the loss of PGE1-dependent cAMP synthesis that occurred following the knockdown of only G␤ 1 (Fig. 6A). It also did not impair INE-stimulated adenylyl cyclase activity, which is not unanticipated because G␣ s concentrations are reasonably well maintained under this circumstance. Instead, PGE1-and INE-dependent activities were augmented nearly 2-fold in cells transfected with one siRNA duplex directed at both G␤ 1 and G␤ 2 (G␤ 1/2 (b)), but not the other (G␤ 1/2 (a)). Immunoblots revealed that G␣ s protein concentrations were 45% lower in cells transfected with G␤ 1/ 2(a) than in those transfected with G␤ 1/2 (b). The difference may be attributed to a modestly better G␤ subunit silencing efficiency of the G␤ 1/2 (a) RNAi duplex compared with the G␤ 1/2 (b) RNAi duplex (data not shown). No significant differences in relative protein expression were observed for any other G␣ subunit (data not shown).
We addressed a number of mechanistic possibilities that might account for the augmented cAMP response seen after silencing G␤ 1/2 with the G␤ 1/2 (b) siRNA. Of interest, forskolin, a direct activator of adenylyl cyclases, caused a 3-5-fold greater cAMP response in cells lacking G␤ 1/2 (regardless of which G␤ 1/2 RNAi duplex was transfected) compared with control cells (Fig. 7). No significant difference was noted in the concentration of forskolin required to elicit a half-maximal response (control EC 50 ϭ 110 Ϯ 60 M and G␤ 1/2 EC 50 ϭ 60 Ϯ 20 M). Cells lacking G␤ 1/2 that were exposed to a maximally effective concentration of forskolin demonstrated up to a 10-fold greater rate of cAMP accumulation than control cells (Fig. 7, inset).
This forskolin-sensitized cAMP response could not be attributed to increased ␤2-adrenergic receptor density or decreased phosphodiesterase activity (data not shown). Differences in basal and maximal cAMP responses (assayed with a mixture of INE, PGE1, and forskolin at saturating concentrations) in cells lacking G␤ 1 /G␤ 2 expression and cells transiently transfected with a constitutively active mutant of G␣ s (G␣ s Q227L) (19) were not consistent with augmented cAMP activity because of an increased pool of free (unassociated with G␤␥) G␣ s subunits.
We also explored whether decreased expression of G␣ i isoforms in cells lacking G␤ 1 and G␤ 2 contributed to the augmented cAMP response. Concomitant knockdown of all three G␣ i subunits (G␣ i1-3 ) did not increase the cAMP response; instead, agonist-dependent and forskolin-stimulated adenylyl cyclase was modestly reduced by 20% (not shown). Individually silencing G␣ i1 or G␣ i3 did not significantly affect INE-or PGE1-stimulated cAMP synthesis; however, forskolin-dependent adenylyl cyclase activity was increased 90% in cells lacking G␣ i3 expression. Interestingly, silencing of G␣ i2 augmented PGE1-, INE-, and forskolin-dependent cAMP synthesis by 50, 70, and 200%, respectively (data not shown). However, silencing G␣ i2 concomitantly with G␤ 1/2 caused additive effects on forskolin-dependent cAMP responses (data not shown), consistent with independence of the mechanisms involved. Thus it is unlikely that the loss of G␣ i contributed significantly to the augmented adenylyl cyclase activity in cells lacking G␤ 1 and G␤ 2 expression.
Lastly, we examined whether the lack of G␤ 1/2 affected concentrations of adenylyl cyclase proteins. There are nine known genes that encode adenylyl cyclases; they are all activated by G␣ s but they are differentially regulated by G␣ i , G␤␥, Ca 2ϩ , Ca 2ϩ -binding proteins, and/or phosphorylation (for review, see Ref. 20). RT-PCR analyses revealed that HeLa cells express AC types I, III, VI, VII, and IX; there was no evidence of AC types II, IV, V, or VIII using a variety of RT-PCR primers and total RNA isolated from control cells or cells lacking G␤ 1/2 . However, immunoblots of enriched plasma membranes (Fig. 8) revealed that concentrations of AC type VI (glycosylated plus non-glycosylated) and AC type III were increased 250 and 60%, respectively, in fractions prepared from cells lacking G␤ 1/2 compared with controls. Lack of reliable antibodies precluded examination of AC types I, VII, and IX. No changes in mRNA levels were observed for any of the five AC isotypes expressed in HeLa cells (data not shown). Increased concentrations of AC types VI and III, in conjunction with the fact that increased accumulation of G␤ 4 permits retention of significant amounts of G␣, provide an explanation for the robust forskolin-and ligand-stimulated cAMP responses in cells lacking G␤ 1 and G␤ 2 .
Knockdown of G␥ Subunits-HeLa cells express all 12 G␥ isoforms. We sought to silence the expression of all 12 endogenous G␥ subunits using siRNA while monitoring the effects on ligand-dependent cAMP responses. The efficacy and specificity of RNAi duplexes targeting G␥ subunits were determined by co-transfection of siRNAs with expression plasmids encoding G␥-GFP fusion proteins ( Table 1). Silencing of individual G␥ subunits reduced PGE1-stimulated cAMP synthesis for 7 of the 9 G␥ subunits tested, and the sum of the individual percentage decreases was significantly greater than 100% (data not shown). Further examination of G␥ contribution to signaling specificity will require specific and sensitive antibodies to all 12 G␥ proteins.

DISCUSSION
This study brings appreciation of the fact that concentrations of G protein subunits are highly inter-related by a variety of mechanisms and that these effects can extend as well to cognate receptors and effectors. It also sounds a clear warning that those who undertake other than short-term perturbations of such networks must be willing to search broadly for the consequences of their actions, focusing on the proteins themselves and not just on more easily measured mRNA transcripts. Our best example of these two conclusions: simultaneous knockdown of G␤ 1 and G␤ 2 in HeLa cells results in a compensatory, non-transcriptionally mediated increase in G␤ 4 ; this is sufficient to stabilize an adequate (but reduced) amount of G␣ s , which can activate substantially increased levels of at least two isoforms of adenylyl cyclase. The result is an enhanced response to agonist when severe impairment was anticipated. This perturbation also resulted in loss of essentially all G␣ i protein, despite a large increase in mRNA for G␣ i1 . Substantial changes in G␥ subunits certainly occurred as well, but were not measured. How much further were these changes felt in these normal appearing viable cells?
Concentrations of G␣ and G␤ in HeLa Cells-We know of no other study where there has been reasonably comprehensive quantification of G protein ␣ and ␤ subunits in a homogeneous cell population. HeLa cells contain very similar total amounts of G␣ and G␤ proteins, suggesting that all subunits are present largely as heterotrimers. Stabilization of G␣ by ␤␥ is likely a major mechanism for the maintenance of the stoichiometric equivalence between G␣ and G␤␥, because silencing of the G␤ subunits present in these cells results in very substantial loss of all G␣ proteins.
Four G␣ subunits (G␣ s , G␣ 11 , G␣ i3 , and G␣ 13 ), each representing one of the four G␣ subfamilies, were the most prevalent G␣ proteins and together accounted for Ͼ80% of the total G␣ pool. Contrary to the prevalent notion that levels of G␣ i proteins usually exceed that of either G␣ s or G␣ q/11 , HeLa cells contain comparable amounts of G␣ s , G␣ 11 , and total G␣ i1-3 , with somewhat lower amounts of G␣ 12/13 . The ratios of the amounts of the family members are at least to some extent celltype or tissue-specific; the amounts of G␣ i proteins in brain are very high.
Functions of G␣ Proteins: Implications for Signaling Specificity-Certain tenets of G protein signaling, such as the specific requirement for G␣ s to activate adenylyl cyclase, G␣ q/11 to stimulate PLC␤, and G␤␥ to stabilize G␣ were verified herein. The effects of silencing G␣ s on agonist-dependent signaling were straight forward: one gene and one functional response.
By comparison, activation of PLC␤ through H1-histamine receptors is more complex. Two G␣ proteins, G␣ q and G␣ 11 , are coupled to the same type of receptor. Of interest, silencing of G␣ q or G␣ 11 caused indistinguishable phenotypes, loss of half of histamine-stimulated PLC␤ activity, despite the fact that concentrations of G␣ 11 exceed those of G␣ q by 10-fold. Because biochemical studies indicate that recombinant G␣ q and G␣ 11 are equally effective at stimulating PLC␤ (21), it is possible that specific localization or organization could be invoked to explain this observation. For example, perhaps only a fraction of the highly expressed G␣ 11 in HeLa cells is accessible to H1 receptors and/or PLC␤.
No specificity was detected among the G␣ i proteins when ␣2-adrenergic receptor-mediated inhibition of INE-or forskolin-stimulated adenylyl cyclase activity was examined, but silencing of G␣ i1 did specifically interfere with ␣2 receptor-mediated inhibition of PGE1-stimulated adenylyl cyclase. Speculatively related, silencing of G␤ 1 interfered with stimulation of cAMP accumulation by PGE1 but not INE, and we observed no effect of this perturbation on other signaling pathways. There appears to be preferential communication between at least one EP-receptor, G␣ s , G␣ i1 , and G␤ 1 , although the mechanism for these apparent relationships are obscure. Knockdown of G␤ Isoforms Affects G␣ Protein Stability-Concomitant knockdown of all three G␤ subunits expressed in HeLa cells resulted in near complete loss of each G␣ without affecting G␣ mRNA levels (except for an increase in G␣ i1 ). Thus, G␤␥ subunits stabilize endogenous G␣ proteins in intact cells. A nearly G protein-free and viable cell is produced, but little G protein-coupled receptor-mediated signaling remains.
Our data demonstrating the loss of G␣ protein in cells lacking G␤ subunits are in contrast to a recently published study demonstrating the apparent stability of free G␣ subunits in a mouse J774A.1 macrophage cell line (18). The authors hypothesized and subsequently demonstrated that G␣ i proteins modified with both myristate and palmitate were more stable than G␣ subunits carrying only one lipid modification. We speculate that the differences observed between the two cell types may be because of the existence of proteins that control the fate of free (i.e. unassociated with G␤␥) G␣ subunits or that the stoichiometry of palmitoylation of the G␣ i proteins may differ substantially in the two cells.
We found that modest amounts of G␣ 11 and G␣ 13 remain after silencing of G␤ 1 , G␤ 2 , and G␤ 4 . We speculate that this may be a reflection of the intrinsic rates of nucleotide exchange that characterize these G␣ proteins. Their relatively slower rates of nucleotide exchange minimizes time spent in the nucleotide-free state, which is unstable. Other explanations are possible, including high intrinsic affinity for residual G␤␥ or interactions with other proteins.
Unanticipated Functional Effects of Silencing G␤ Isoforms-No requirement for a specific G␤ isoform was revealed for ␤2-adrenergic, H1-histaminergic, or ␣2-adrenergic receptor-dependent modulation of effector activities. However, as noted above, some degree of functional specificity was exhibited for G␤ 1 coupling of prostanoid receptors to adenylyl cyclase. It was thus surprising that prostanoid-dependent activation of cAMP synthesis was re-established when G␤ 2 was silenced along with G␤ 1 . There are a number of possibilities that could explain the retention of PGE1-mediated signaling under this circumstance. The simplest of these involves the increase in adenylyl cyclase concentrations noted after silencing of G␤ 1 and G␤ 2 . Alternatively, severe perturbation of the system, such as the knockdown of G␤ 1 and G␤ 2 , may have disrupted and redistributed prostenoid-dependent adenylyl cyclase signaling components from specialized plasma membrane compartments (22).
Our study also revealed some interesting features about G␤ 4 , the least well studied of the conventional G␤ subunits. The amino acid sequence of G␤ 4 is more than 90% identical to those of G␤ 1 and G␤ 2 , it is ubiquitously expressed (23), and it behaves similarly to G␤ 1 and G␤ 2 in vitro (23,24). We show that endogenous G␤ 4 has the ability to maintain G protein-coupled receptor-dependent signaling by interacting functionally with and supporting the coupling of G␣ s , G␣ q , and G␣ 11 to effectors in cells that lack G␤ 1 and G␤ 2 .
There appear to be differences in the regulation of expression of G␤ 4 compared with that of G␤ 1 and G␤ 2 . G␤ 4 was the only G␤ subunit whose protein expression was significantly reduced (50% or more) following knockdown of more than one G␣ subfamily. With the exception of G␣ i2 , loss of G␤ 4 protein was correlated with decreased G␤ 4 mRNA, indicating that G␣ s , G␣ i3 , and G␣ 11 each, in some manner, regulated expression of G␤ 4 mRNA. Moreover, elimination of the functionally important (but not prevalent) G␤ 4 increased accumulation of G␤ 1 and G␤ 2 , whereas only the elimination of both G␤ 1 and G␤ 2 had a significant effect on G␤ 4 .
Alterations of Non-targeted Proteins-Knockdown of a specific G protein subunit often resulted in increased accumulation of a member of the same subfamily. Thus, loss of the prevalent G␣ i3 doubled the levels of G␣ i1 and G␣ i2 , whereas loss of G␣ i2 doubled G␣ i1 . Silencing of all G␣ i1-3 proteins impacted the accumulation of G␣ o . The relationships between expression of G␤ 1 , G␤ 2 , and G␤ 4 isoforms is noted above. Some of these effects appeared to be mediated transcriptionally; some did not. The exception is G␣ q and G␣ 11 , where compensatory increases were not observed following loss of either protein.
Less expected was observation of an apparently reciprocal relationship between expression of members of the G␣ q and G␣ i subfamilies. Loss of G␣ q or G␣ 11 caused increased accumulation of G␣ i1 and G␣ o ; loss of G␣ i3 resulted in increased accumulation of G␣ q . The significance is not clear.
It is also notable that levels of G␣ i1 and its mRNA increased following the knockdown of almost every G␣ and G␤ subunit, which is suggestive of some as yet unknown common mechanism. Little is known about the specific signaling pathways affected by G␣ i1 . It is widely expressed, particularly in brain, but there has been no obvious phenotype associated with G␣ i1 knock-out mice (25).
The impact of G␤ knockdowns went beyond G protein family members. We documented changes in the amounts of ␤-adrenergic receptors with the knockdown of G␤ 1 and increased levels of adenylyl cyclase when G␤ 1 and G␤ 2 were silenced simultaneously. Increased ␤2-adrenergic receptor expression correlated with a 2-3-fold increase in mRNA for the receptor in cells lacking G␤ 1 , whereas, AC type VI protein appeared to increase independently of changes in mRNA. We do not know if the changes in adenylyl cyclases are a direct result of loss of G␤ 1 and G␤ 2 , increase in G␤ 4 , partial loss of G␣ s , or other factors; however, knockdown of G␣ s increased adenylyl cyclase (types III and VI) expression 60 -75%, suggesting that G␣ s likely contributes to the regulation of adenylyl cyclase expression.
We cannot comment on the generality of the specific network interactions that are revealed by these knockdown experiments in HeLa cells. Our analyses revealed many changes in non-targeted G protein subunit expression and specific network interactions are likely to be dependent on the specific cell type, the presence of different regulatory mechanisms, and relative stoichiometries of G protein subunits. Future studies using different cells and tissues will reveal the generality of such relationships. FIGURE 8. Adenylyl cyclase protein expression in plasma membranes prepared from cells lacking G␤ 1/2 subunits. One hundred g/lane of enriched plasma membranes were blotted for: A, type VI; or B, type III adenylyl cyclase. Glycosylated adenylyl cyclase bands migrated as large smears with apparent M r values slightly larger than 150,000. Glycosylation of both type VI and III adenylyl cyclases was confirmed by observing increased mobility following enzymatic removal of glycosyl groups with peptide: N-glycosidase F (data not shown). Non-glycosylated type III adenylyl cyclase could not be detected, whereas non-glycosylated type VI adenylyl cyclase co-migrated with a 100-kDa standard (Precision Plus Protein Standards, Bio-Rad). Antibody specificity was verified following isotype-specific siRNA-mediated knockdown of select adenylyl cyclases. No cross-silencing effects were observed with either isotype-specific RNAi duplex. Data are representative of three independent preparations.
It is clear from this study that G protein signaling pathways in HeLa cells are elastic and resilient in their ability to maintain robust regulation of effectors in the face of change. The mechanisms used to achieve this are varied and complex. It will be a daunting challenge to understand how and to what extent the concentrations and activities of the many interacting components of such networks are regulated to resist and respond to the forces that impinge upon them.