Several related small molecules (Figure 1) were identified as active in a screen for inhibition of carbachol-evoked tonicity of isolated rat bladder strips. The effects of BMS-195270 in an ex vivo rat whole bladder model are shown in Figure 2. Typical cystometric curves (bladder pressure plotted versus infused volume) for BMS-195270–treated or vehicle-treated bladders are shown in Figure 2A and Figure 2B. Incubation of bladders in BMS-195270 produced a dramatic reduction in developed pressure at infusion volumes of 0.2–1.3 mls (Figure 2A; n = 5, p Figure 2B; n = 12 bladders). BMS-195270 treatment was also found to inhibit “spontaneous” contractions (Figure 2C and 2D). The finding that BMS-195270 was able to increase the filling capacity and reduce spontaneous contractions in this ex vivo model underscored the potential utility of this class of small molecules in treating UI resulting from overactive bladder contractions. Assays of structurally related small molecules revealed two small-molecule pairs that each contain one active and one inactive small molecule (BMS-192364/BMS-192365 and BMS-195270/BMS-195243; Figure 1A–1D, [16]). The mechanism of this bladder-specific activity, however, was not known.
A Chemical Genetic Approach in C. elegans to Identify Candidate Targets
We next utilized C. elegans to elucidate the molecular pathway affected by these small molecules. First, the active small molecules BMS-192364 and BMS-195270 were applied to wild-type adult C. elegans, and the resulting phenotypes observed (see Materials and Methods). Note that treatment of C. elegans with small molecules typically requires a higher concentration than cell-based assays, due to the worms' relatively impermeable cuticle.
Treatment of adult worms with 0.3 mM BMS-192364 resulted in a bloated egg-laying defective (Egl-d) phenotype that included retention of fertilized late-stage eggs (Figure 3A). This Egl-d phenotype indicates defective neuromuscular function in the egg-laying process. Treatment with BMS-192364 is dose-responsive, with a cumulative total of >90% of animals ultimately displaying an Egl-d phenotype (Figure 3B). Treatment of adult worms with BMS-195270 at 2.8 mM resulted in a similar Egl-d phenotype, as well as slowed or arrested pharyngeal pumping (Eat), and uncoordinated motion (Unc). These phenotypes also indicate a defect in normal neuromuscular signals. The pumping and movement defects were apparent within two hours of treatment, and the Egl-d phenotype was noted within 12 hours of treatment and maximal after 24 hours, consistent with an acute effect on the neuromuscular system. Structurally related small molecules that lacked the bladder-relaxing activity, such as BMS-192365, did not cause the Egl-d, Eat, or Unc phenotypes under any treatment conditions (unpublished data). This observation suggested that the structure–activity relationship as determined in the mammalian systems held true in C. elegans, and that the worm phenotypes corresponded to the therapeutic activity of the compounds. BMS-192364 was the more potent of the two active molecules in the egg-laying assays, and subsequent analyses in C. elegans used this compound.
To identify potential molecular targets or target pathway components for the small molecules, we first took a candidate-gene approach. The G-protein mediated pathways involved in egg laying have been extensively characterized in C. elegans. We tested BMS-192364 on 27 different worm strains, each carrying a mutation in pathways implicated in egg laying (Table 1). Of the 27 strains tested, three exhibited at least partial resistance to the Egl-d phenotype induced by BMS-192364. (Resistance was defined as no more than 15% of animals exhibiting an Egl-d phenotype in the presence of a small-molecule dose which renders >90% of wild-type animals Egl-d). The three resistant strains carried either the eat-16(ad702), the egl-19(n582ad952), or the egl-19(n2361) mutation.
Interestingly, the mutation giving rise to the strongest resistance phenotype, eat-16(ad702), disrupts an mRNA splice acceptor site and leads to truncation of an RGS protein. The EAT-16 RGS protein would normally downregulate G-protein signaling, an activity consistent with the phenotype of the small molecule [17]. The nIS51(egl-10+) allele phenotypically resembles the eat-16(ad702) allele (it overexpresses the EGL-10 RGS protein, which acts in an antagonistic pathway). However while the eat-16(ad702) strain was resistant to BMS-192364, a strain carrying the egl-10(nIS51) allele showed no resistance. Thus the resistance of the eat-16(ad702) strain seems closely linked to EAT-16 signaling status. In addition, the lack of resistance or hypersensitivity exhibited by strains carrying mutations in egl-8, dgk-1, tpa-1, or unc-68 (encoding phospholipase C, diacylglycerol kinase, PKC, and the Ryanodine receptor; Table 1) indicated that BMS-192364 was most likely acting upstream of or parallel to these genes' products. Taken together, the candidate-gene results suggested a target function in the area of the neurotransmitter/GPCR/G-protein complex. The resistance exhibited by two strains carrying gain-of-function mutations in egl-19 [18] also indicated that calcium channels could be the target of BMS-192364.
Genetic screens designed to identify both dominant and recessive mutations have been used successfully to identify components of pathways affected by small molecules [19,20]. In parallel to the candidate gene analysis, we also decided to carry out an unbiased genetic screen based on the robust Egl-d phenotype, looking for resistance to the small molecule BMS-192364. The candidate gene analysis indicated that such a screen should yield resistant mutants. 150,000 EMS-mutagenized genomes were generated and screened, resulting in the identification of four resistant mutants (Figure 3 C). In all cases these mutants were egg-laying constitutive (Egl-c) in the absence of treatment, and both the Egl-c phenotype and the drug resistance were dominant or semidominant. All BMS-192364–resistant mutants were also found to be cross-resistant to the phenotypes induced by BMS-195270 (unpublished data).
Dominant C. elegans–Resistant Mutations Affect G-Protein Pathway Components
The mutations present in the three resistant strains ep271, ep273, and ep275 were mapped to defined chromosomal regions using recombinant single nucleotide polymorphism (SNP) techniques [21,22]. Genes in these regions that were known to confer an Egl-c phenotype were obvious candidates. Linkage tests to known alleles, and direct sequence analysis, showed that three of the four mutated genes were allelic to the known genes egl-30, eat-16, and goa-1 (Figure 3D). The identity of the gene affected in the fourth resistant strain, ep272, remains unknown.
Strain ep271 carried a mutation in egl-30, which encodes the worm ortholog of the G-αq subunit and is required for egg laying [17,23]. Based on the Egl-c mutant phenotype observed, the mutation found in the egl-30(ep271) allele, M244I, is predicted to constitutively activate G-αq signaling. Methionine-244 of EGL-30 is conserved in many G-proteins that are regulated by an RGS domain–containing protein (Figure 4A).
Strain ep273 carried a mutation in eat-16, a member of the conserved RGS protein family [17]. EAT-16 has been shown to directly interact with and negatively regulate signaling via the G-αq protein EGL-30 [17,24,25]. The eat-16(ep273) allele was found to generate a nonconservative change, E158K, at a location between the DEP and GGL regions of the protein (Figure 4B). The Egl-c phenotype of the eat-16(ep273) mutant is similar to known loss-of-function eat-16 mutants [17,24].
Strain ep275 carried a mutation in goa-1 [17,26], which encodes the worm G-αo protein and acts to negatively regulate egg laying, most likely acting in parallel to the egl-30/eat-16 pathway [23,27]. The goa-1(ep275) allele was found to contain a stop codon that truncates the GOA-1 protein at amino acid 205, and is predicted to produce an inactive protein (Figure 3D).
The results of the genetic screen were striking. All three of the cloned resistant alleles (two dominant and one semidominant) affect proteins that are members of G-α/RGS complexes. This result, combined with the candidate gene analysis (Table 1), indicated that the small molecules were most likely acting on the signaling pathway at the level of the G-protein complexes. To summarize the genetic results, we observed two strongly compound resistant loss-of-function (lf) alleles of eat-16, one resistant and one sensitive gain-of-function (gf) allele in egl-30, and one resistant and one sensitive (lf) allele in goa-1, and finally two resistant (gf) alleles of egl-19, which encodes an L-type calcium channel [18]. The location of these gene products in a canonical signaling pathway is shown in Figure 5. In both the candidate gene analysis and the genetic screen, the strongest resistance was displayed by eat-16 (lf) alleles. The differential resistance of the egl-30(ep271) and egl-30(pk931) strains (Figure 4C) was quite surprising, as without compound treatment they display an identical Egl-c (gf) phenotype. This might indicate that these alleles have obtained their gain-of-function properties through different mechanisms: the egl-30(pk931) allele was originally identified as a suppressor of gpb-2 overexpression phenotypes (RP, personal communication).
Confirmation of G-Protein Involvement Using a Mammalian System
The functional roles of the C. elegans modifier genes pointed us toward a small-molecule mechanism involving calcium signaling via G-proteins (Figure 5). To validate and extend this model in mammalian cells, the small molecules were evaluated for their effects on muscarinic GPCR-mediated calcium release in mammalian cells. Both BMS-192364 and BMS-195270 inhibited the response of HEK293 cells to the muscarinic agonist carbachol (Figure 6A and 6B; the EC50 for BMS-192364 was 9 μM, and for BMS-195270 was 2 μM). Experiments with muscarinic receptor subtype-specific inhibitors indicated that the majority of the carbachol-evoked signal was via muscarinic receptor type 3 (M3), with a small amount via M2 (unpublished data). The degree of inhibition observed with BMS-192364 and BMS-195270 was similar to that seen with known pan-muscarinic receptor inhibitors. The ability of the small molecules to inhibit calcium fluxes downstream of muscarinic receptor activation was entirely consistent with our observations in mammalian bladder tissue and C. elegans.
Utilizing a primary bladder smooth muscle cell line with Histamine as a receptor agonist, we also observed inhibition of calcium flux by BMS-192364 (Figure 6C) and BMS-195270 (unpublished data). Our genetic results had suggested that the compounds acted downstream of muscarinic receptors. To further investigate this, we carried out competitive binding assays with BMS-192364 and the radio-labeled muscarinic ligand N-methylscopolamine. We observed no effect on radio-ligand binding to the muscarinic receptor types M1–5 [28,29] using BMS-192364 at concentrations of 10μM. Similar radio-ligand binding experiments with the histamine receptors H1–4 were also negative (unpublished data) [30–32]. While we can not rule out the existence of a previously undescribed allosteric site shared by the two GPCR types, from these results it is unlikely that BMS-192364 and related small molecules act directly on muscarinic or histamine receptors. Additional BMS compound/radio-ligand binding assays run on adrenergic, angiotensin, bradykinin, cannabinoid, dopamine, endothelin, neuropeptide Y, nicotinic, serotonin, and other GPCRs were also negative.
The mammalian N- and L-type calcium channels were also of interest as potential targets, due to the partial resistance exhibited by two egl-19(gf) alleles in C. elegans (Table 1). To address the possibility that BMS-192364 and BMS-195270 were acting directly on the channels, we repeated the carbachol-stimulated Ca-flux assays, but pretreated cells with a high concentration of the calcium channel blocker niguldipine [33]. We reasoned that if BMS-195270 were acting directly on calcium channels, by preblocking those channels with niguldipine, we should observe no further effect of BMS-195270 on calcium flux. We observed that while pretreatment with niguldipine slightly dampened carbachol-induced calcium flux, BMS-195270 treatment was clearly additive to this effect (Figure 7A). This additive response indicated that BMS-195270 retained inhibitory activity even when endogenous calcium channels were inactivated. In addition to the cellular assays, competitive binding assays were performed using radio-labeled diltiazem (L-type ligand) [30] or ω-conotoxin (N-type ligand) [34]. BMS-192364 at 10μM had no effect on binding of either radio-ligand, suggesting that the L- and N-type channels are not the direct target of BMS-192364.
Our genetic analysis for resistance to BMS-192364 had identified multiple alleles affecting the RGS protein EAT-16 and single alleles for G-αq (egl-30) and G-α0 (goa-1). The genetic data also indicated that the targeted process was likely upstream or parallel to PKC (tpa-1), PLC (egl-8), DGK (dgk-1), and the RyR (unc-68), as mutations in these genes did not affect sensitivity to BMS-195270 (Table 1). Clearly a comprehensive analysis of potential branched effecter pathways needs to be carried out. Consistent with these genetic observations, pretreatment of HEK293 cells with inhibitors of the ryanodine receptor (ryanodine), PI3-kinase (wortmanin), or PLC (U73122) did not affect inhibition of carbachol-induced calcium flux by BMS-195270 (unpublished data). Therefore, we next examined the G-protein/RGS complex more closely.
In C. elegans, the RGS protein EAT-16 has been shown to interact with both G-αq/EGL-30 and G-α0/GOA-1, in different complexes. Therefore, we investigated which G-α subunit was involved in the BMS-195270 activity seen in HEK293 cells. Given that the muscarinic receptor M3 (a G-αq–coupled receptor) was responsible for the majority of the carbachol-induced signal, G-αq was the most likely candidate. We tested the effects of BMS-195270 in combination with pertussis toxin, which is a known blocker of G-αo/i and G-αs signaling but does not affect G-αq [35]. Pertussis toxin has been utilized extensively to differentiate G-α signals. Pertussis toxin reduced the peak calcium flux evoked by carbachol stimulation of HEK293 cells, but the effect of pertussis toxin was nearly additive with that of BMS-195270 (Figure 7B). This result indicated that BMS-195270 inhibits signaling downstream of muscarinic receptors, at least in part, by a G-αq-dependent mechanism. This observation was entirely consistent with the genetic analysis in C. elegans, but does not exclude the possibility that some BMS-195270 activity in mammalian cells might also be due to inhibition of another G-α subtype activity.
Genetic Tests Differentiate between Models of Small-Molecule Action
The biochemical and genetic data collectively directed us toward a hypothesis that the BMS small molecules affect the RGS/G-protein complex. In simple, single-target models, the behavior of BMS-192364 and BMS-195270 is consistent with action as an RGS protein agonist, or as a G-α antagonist. In the former model, the small molecule would increase the rate at which RGS protein is able to stimulate GTP catalysis by G-α (RGS GAP activity) and thereby shorten the active period of G-protein signaling. In the latter model, the small molecule inhibits G-αq activity directly. The small molecule could affect the activation of G-αq or its ability to signal to downstream partners (Figure 8). A precedent for small-molecule effects on G-αq has been set by YM-254890 [36], which has been reported to inhibit the activity of G-αq. Genetic tests in a variety of systems were used to investigate these two single-target models.
A Genetic Test of the RGS Agonist Model in Yeast
If the compounds act via the RGS protein, G-αq proteins that are resistant to RGS GAP activity should also exhibit resistance to the compound. In C. elegans the egl-30 (ep271) allele of G-αq, which contains the amino acid substitution M244I, confers resistance to compound BMS-192364. Therefore, we wished to determine whether this mutation also confers resistance to RGS GAP activity, as predicted by the first model.
Ideally, resistance to compound activity and to RGS GAP activity should be studied in the same genetic system. However, there are no C. elegans strains that carry bichemically proven RGS GAP–insensitive alleles of G-α, and there is no system for evaluating response to RGS GAP activity in C. elegans. Instead, RGS GAP activity is studied in the yeast Saccharomyces cerevisiae, which offers both a robust assay system and also the ability to substitute the mutant alleles under analysis [11,37]. (Unfortunately, the insolubility of compound BMS-192364 in yeast growth media prohibits direct evaluation of its action in yeast). The yeast RGS/G-α interaction is highly conserved with respect to higher eukaryotes, and the resistance of a mammalian G-α mutant protein to RGS GAP activity has been determined in this system [11,37]. Therefore, observations in yeast will be informative as to the properties of the C. elegans egl-30(ep271) allele.
RGS/G-α interactions in yeast are evaluated by treatment with the peptide ligand α–factor, which acts on the GPCR Ste2 initiating a signaling cascade that blocks cell division. This can easily be assayed as a halo of growth arrest around a point source of the ligand (Figure 9A) [11,38–42]. In the absence of the GAP activity of yeast RGS protein Sst2, this growth arrest can be achieved at lower ligand concentrations, hence larger halos are observed (Figure 9A). To test whether the EGL-30(M244I) substitution produced a G-α protein that was sensitive to RGS GAP activity, we constructed the equivalent mutation, M362I, in the yeast G-α protein Gpa1. The gpa1-M362I or wild-type GPA1 alleles were then expressed from a plasmid [43] in the background of a yeast strain lacking both the RGS and wild-type G-α (sst2 gpa1). When tested in the halo-formation assay, the sst2 gpa1 (p-gpa1-M362I) and the sst2 gpa1 (p-GPA1) strains produced the same diameter of halo, demonstrating that a functional G-α protein is produced from the gpa1-M362I allele (Figure 9A and 9B). The halo diameter was large, as expected in the absence of the Sst2 RGS function. When Sst2 function was returned to the two yeast strains (by expression of the SST2 gene from a second plasmid), halo diameter was reduced to the same degree in the presence of the gpa1-M362I or the wild-type GPA1 allele (Figure 9C and 9D). Thus, the yeast gpa1-M362I allele produces a functional G-α protein that has a normal response to the GAP activity of the RGS protein. Since the equivalent G-α allele in C. elegans, egl-30(M244I), is resistant to BMS-192364, a model in which the small molecule acts as an RGS protein agonist is unlikely.
In the yeast halo formation assay, one qualitative difference was apparent between the p-gpa1-M362I and the p-GPA1 strains, regardless of their RGS status. The halos produced by the strain expressing Gpa1-M362I protein were less turbid (Figure 9B and 9D), indicating that growth was arrested for longer by a given dose of pheromone. The phenomenon of growth resumption in the continued presence of the α–factor ligand is called adaptation [42]; the gpa1-M362I allele is thus hypo-adaptive. Interestingly, the M362I mutation affects a region of Gpa1 where a number of hyper-adaptation alleles have also been described [42,44,45] (Figure 10). These mutations all affect amino acid residues on the surface that is known to interact with downstream effectors of G-α [46]. Thus the C. elegans egl-30(M244I) mutation is likely to confer resistance to compound by strengthening interactions with effectors, rather than by insensitivity to RGS GAP activity. This M/I substitution creates the first hypo-adaptation allele of G-α to be described.
A Genetic Test of the G-α Antagonist Model in C. elegans
The evidence presented so far suggested that the compound could be acting on the G protein, the RGS protein or potentially both. We devised a genetic test to attempt to help distinguish between these possibilities. The experiment took advantage of a putative activated egl-30 mutant that is sensitive to the compound (unlike the egl-30(M244I) mutant identified in the resistance screen). The egl-30(pk931) strain has an Egl-c phenotype, indicative of constitutively activated G-αq signaling. The R210Q substitution in egl-30(pk931) affects a predicted key residue on the face of G-α that interacts with the RGS protein. While there is no direct proof that EGL-30(R210Q) is constitutively activated due to an insensitivity to RGS GAP activity, a mammalian G-αq mutation of the neighboring amino acid (Q209L) results in both constitutive activation and insensitivity to RGS GAP activity. The egl-30(pk931) strain was found to retain sensitivity to BMS-192364 (Figure 4C). The finding that different mutant alleles of the same gene have different sensitivities to the compound suggests that the G-α protein, EGL-30, may be directly involved in compound action. However, it cannot be ruled out that the difference in sensitivity is due simply to a difference in the degree of activation of the alleles.
What about the role of EAT-16? The assumption for the following test is that if the G-αq protein is the sole component interacting with the compound, then the presence or absence of the RGS protein should not affect resistance/sensitivity to the compound. Conversely, if presence of the RGS protein is critical for compound activity, removing it (by mutation) should confer resistance regardless of the status of the G-αq subunit. Therefore, a double mutant strain containing the compound-sensitive egl-30(pk931) allele and the RGS protein loss-of-function eat-16(ad702) allele was constructed and tested for small-molecule sensitivity. The eat-16(ad702); egl-30(pk931) double mutant strain was found to be completely resistant to BMS-192364 (Figure 4C), like the eat-16(ad702) single mutant, implicating EAT-16 in compound action. Overall, these results are consistent with a model in which both the G-αq and RGS proteins interact with the compound.
Chemical Genetic and Biochemical Tests of the G-α Antagonist Model in Mammalian Cells
To investigate the G-α antagonist model in mammalian cells utilizing a better characterized G-αq allele, we overexpressed the mammalian G-αq-G188S mutant protein in Hek-293 cells (construct courtesy of the Guthrie Research Institute). The G-αq-G188S mutant protein requires ligand stimulation, but is known to be resistant to subsequent deactivation by RGS proteins, thus behaving as a constitutive signaling molecule [12]. However, overexpression of G-αq-G188S did not suppress the inhibitory effect of BMS-192364 upon carbachol-stimulated calcium flux (Figure 7C). This result (similar to the G-αq experiment in C. elegans above), also fails to support a model where compound directly antagonizes G-αq.
Finally, we tested the compound's direct biochemical action on wild-type G-αq. We quantified binding of a radio-labeled nonhydrolysable substrate (GTP-γS) to the G-α protein downstream of the histamine H1 receptor (which is G-αq coupled) [47] and the muscarinic receptor M4 (which is G-αi/G-αo coupled) [48]. BMS-192364 displayed no significant activity in either assay, when performed in either agonist or antagonist mode (unpublished data). These results suggest that BMS-192364 does not directly affect GDP/GTP exchange at G-αq or G-αi/o.
Answers to all your Biology Questions
