Discovery of small-molecule modulator of heterotrimeric G_i-protein by integrated phenotypic profiling and chemical proteomics

ABSTRACT

Discovery of small-molecule inducers of unique phenotypic changes combined with subsequent target identification often provides new insights into cellular functions. Here, we applied integrated profiling based on cellular morphological and proteomic changes to compound screening. We identified an indane derivative, NPD9055, which is mechanistically distinct from reference compounds with known modes of action. Employing a chemical proteomics approach, we then showed that NPD9055 binds subunits of heterotrimeric G-protein G_i. An in vitro [³⁵S]GTPγS-binding assay revealed that NPD9055 inhibited GDP/GTP exchange on a Gα_i subunit induced by a G-protein-coupled receptor agonist, but not on another G-protein from the Gα_s family. In intact HeLa cells, NPD9055 induced an increase in intracellular Ca²⁺ levels and ERK/MAPK phosphorylation, both of which are regulated by Gβγ, following its dissociation from Gα_i. Our observations suggest that NPD9055 targets Gα_i and thus regulates Gβγ-dependent cellular processes, most likely by causing the dissociation of Gβγ from Gα_i.

GRAPHICAL ABSTRACT

By combining integrated phenotypic profiling and chemical proteomics, a small-molecule modulator of heterotrimeric G_i-protein was discovered.

KEYWORDS:

• Phenotypic screening
• profiling
• chemical proteomics
• target identification
• heterotrimeric G-protein

Discovery of small-molecule modulators of proteins has accelerated the understanding of cellular processes at the molecular level and is also important for drug discovery. However, available small-molecule probes for many proteins are still lacking[1]. Therefore, exploring uncharacterized chemotypes with unique biological activity and connecting them with the target molecules is of great importance. Target-agnostic phenotypic screening, when compared to target-based screening, is a less biased approach that sometimes enables identification of unexpected targets necessary for achieving desired phenotypes[2]. Indeed, phenotypic screening combined with subsequent target identification has led to the identification of unique bioactive compounds and their target molecules, which has contributed to uncovering cellular functions and developing new drugs for the treatment of diseases [3,4].

Generally, target identification is a time-consuming, but critical process of elucidating mechanisms of action of bioactive compounds discovered through phenotypic screening. Techniques for target identification are divided into direct and indirect approaches [5,6]. Chemical proteomics, i.e., affinity-based enrichment of proteins, is a direct approach that is widely used in the identification of cellular targets of bioactive compounds. Indirect approaches, such as profiling for mechanism elucidation, have also been developed. We have previously established two profiling techniques, MorphoBase and ChemProteoBase, to characterize bioactive compounds based on the induced cellular morphological and proteomic changes, respectively [7,8]. Based on comparisons with 207 and 120 reference compounds with known modes of action, respectively (see Table S1 in Ref. 7; and Table S1), both methods have been utilized for target prediction of uncharacterized bioactive compounds[9]. More recently, a highly multiplexed morphology-based profiling method, termed Cell Painting assay, was established as a complementary approach based on similar logic[10].

G-protein-coupled receptors (GPCRs), the largest family of cell surface receptors, regulate a wide range of physiological and pathological processes, and heterotrimeric G-proteins are key mediators of GPCR-regulated signaling [11,12]. Heterotrimeric G-proteins consist of Gα, Gβ, and Gγ subunits, and are regulated via an activation/inactivation cycle, as follows: (i) Gα usually exists in a GDP-bound state and forms a trimer with Gβ and Gγ; (ii) an agonist-activated GPCR stimulates the exchange of bound GDP for GTP on the Gα subunit; (iii) GTP binding to Gα leads to Gβγ dissociation, and the subsequent binding of Gα and Gβγ to downstream effectors; (iv) the signal is terminated by GTP hydrolysis, resulting in the reformation of an inactive heterotrimer. Based on sequence similarity of the α subunit, G-proteins are divided into four families, G_s, G_i/G_o, G_q/G₁₁, and G₁₂/G₁₃. Since GPCRs are major targets for therapeutic drugs, G-proteins have also attracted attention [13,14].

In the present study, we integrated MorphoBase and ChemProteoBase profiling systems for compound screening, and thereby identified NPD9055, an indane derivative that was mechanistically distinct from the reference compounds. In addition, our chemical proteomics approach and subsequent biochemical assays suggested that NPD9055 directly targets Gα_i and modulates Gβγ-regulated cellular processes.

Results and discussion

The screening strategy for mechanistically unique bioactive compounds was to identify compounds that induced phenotypic changes distinct from those induced by well-characterized reference compounds in both MorphoBase and ChemProteoBase profiling systems. Since the throughput of both profiling methods is not high, we needed to narrow down test compounds prior to the profiling analyses. To this end, we first performed cytotoxicity assays using a panel of cancer cell lines, because morphology and the proteome are, in many cases, both affected during the process of cell death or cell-cycle arrest. In parallel, compound-induced morphological changes were observed under a microscope during the cytotoxicity assays. After testing 5472 compounds from the RIKEN Natural Products Depository (NPDepo) library, we found that an indane derivative, termed NPD9055 (1) (Figure 1a), was toxic against various cancer cell lines (IC₅₀ values, 1.8–18 μM; Table S2), and notably induced morphological changes in v-src-transformed normal rat kidney (src^ts-NRK) and human cervical cancer HeLa cells (Figure S1). Therefore, we next performed MorphoBase profiling for NPD9055. Using 71 parameters extracted from the nuclei, whole cells, and granules of NPD9055-treated src^ts-NRK and HeLa cells, we quantified the morphological changes. The mechanistic uniqueness of NPD9055 was suggested by the probability scores for 15 well-validated target classes and Euclidean distances to the 207 reference compounds (Table S3). Next, we performed ChemProteoBase profiling as a complementary and more informative analysis based on changes at the molecular level. After the treatment of HeLa cells with NPD9055, changes in 296 protein levels were analyzed. We found increases in the levels of multiple proteins localized to the mitochondria or the endoplasmic reticulum (ER) (Figure S2; Table S4). When the patterns of changes in protein expression level are compared, two compounds with a similar mode of action usually have a cosine similarity score >0.7. To identify compounds that have a mode of action similar to that of NPD9055, we analyzed the pattern of changes in protein expression level between NPD9055 and our 120 reference compounds. However, we did not identify any compounds that have a cosine similarity score with NPD9055 > 0.7, indicating that the mode of action of NPD9055 is unique among our reference compounds (Figure 1b and S3). This mechanistic uniqueness of NPD9055 suggested by the integrated phenotypic profiling prompted us to elucidate its mechanism of action.

Figure 1. Indane derivative NPD9055 has a unique mode of action

(a) Chemical structure of NPD9055 (1). (b) Analysis by ChemProteoBase profiling data. HeLa cells were treated with 10 µM NPD9055 for 18 h, and the cell lysate was analyzed by proteomics using the 2-D difference gel electrophoresis system. Quantitative data of the common 296 spots (x-axis) derived from NPD9055 and 120 reference compounds were analyzed by hierarchical clustering. In the heat map, log-fold (natural base) of the normalized volume is shown on the colored scale.

To identify the target molecule of NPD9055, we analyzed proteins binding to NPD9055 by using a chemical proteomics approach that employs affinity purification and subsequent mass spectrometric protein identification. To design a suitable affinity probe, we first synthesized and tested a small number of NPD9055 analogs. We found that the methyl group at the tertiary amino group can be extended to obtain compound 2 without the loss of cytotoxicity (Table S5). After testing compound 3, we synthesized a biotin-tagged pull-down probe 4; the probe exhibited a slight loss of cytotoxicity, most likely because of reduced cell membrane permeability (Figure 2a; Table S5). Using affinity probe 4 together with streptavidin beads, we performed pull-down assays using HeLa cell lysates in the presence or absence of unmodified NPD9055 as a competitor. Subsequent identification of the co-precipitated proteins by mass spectrometry and label-free quantification yielded six candidates whose intensity was significantly reduced in the presence of the competitor (Figure 2b): Gα_i2 and Gβ₂, which are components of a heterotrimeric G-protein[11]; nucleoporins Nup160, Nup37, and Nup85, which form the nuclear pore Nup107-160 complex[15]; and dynamin-like 120 kDa protein. Therefore, the six identified proteins were divided into three groups: (i) components of a heterotrimeric G-protein, (ii) components of the nuclear pore, and (iii) dynamin-like 120 kDa protein (Figure S4a). We have previously established a method for compound immobilization onto a matrix by a photoaffinity reaction that is functional group-independent[16]. Using this method, we prepared NPD9055-immobilized photoaffinity beads. To confirm the results of the experiment with the biotin-tagged pull-down probe (Figure 2b), we performed a competition pull-down assay with the photoaffinity beads. From the six potential binding proteins, only Gα_i2 and Gβ₂ (encoded by GNAI2 and GNB2, respectively) were identified in the photoaffinity bead experiment. Other G-protein subunits, Gα_i1, Gα_i3, Gβ₁, Gβ₄, and Gγ₁₂ (encoded by GNAI1, GNAI3, GNB1, GNB4, and GNG12, respectively), were also identified as potential NPD9055-binding proteins in the same experiment (Figures S4b and S4c). To confirm the physical engagement of Gα_i2 and Gβ₂ by NPD9055, we then performed a cellular thermal shift assay, which is based on the changes in thermal stability of small-molecule protein targets[17]. NPD9055 slightly disrupted the thermal stability of Gα_i2 in a HeLa cell lysate with a shift in the melting temperature of – 1.1°C (Figure 2c,d). Therefore, we prioritized G-proteins, especially the ones from the G_i subfamily, for further in-depth investigation.

Figure 2. G_i subunits are NPD9055-binding proteins

(a) Chemical structure of the biotin-tagged pull-down probe, compound 4. (b) List of potential NPD9055-binding proteins derived from HeLa cells, whose label-free quantification (LFQ) intensities were significantly reduced (according to a t-test) in the presence of the competitor in pull-down assays (number of technical replicates, N = 3; number of biological replicates, n = 2). Just proteins with significantly reduced levels in both biological replicates are listed. Statistical significances were checked by means of a t-test comparing the LFQ intensity values of the proteins with and without competition for each biological replicate separately (see Materials and Methods for more detailed information). The numbers indicate nonlogarithmic enrichment ratios, i.e., mean of LFQ intensities of the respective protein in the absence of the competitor divided by the mean of LFQ intensities of the same protein in the presence of the competitor. “Specific” indicates that the protein was detected only in the absence of the competitor. Proteomics results after MaxQuant search of all proteins identified are listed for both biological replicates separately in Tables S6 and S7. c) Effect of NPD9055 on the thermal stability of Gα_i2 and Gβ₂. Cellular thermal shift assay using HeLa cell lysate treated with NPD9055 or DMSO with immunoblotting detection. Representative data for two independent experiments are shown. d) Quantification of the band intensities from c). The values were normalized to the corresponding band intensity at 37.0ºC.

During G-protein activation, GPCR agonist-stimulated GDP/GTP exchange occurs on Gα[11]. To monitor this process, we previously established an in vitro [³⁵S]GTPγS-binding assay, using a membrane fraction from transgenic silkworm expressing a human GPCR-Gα fusion protein to measure GPCR agonist-induced exchange of GDP to [³⁵S]GTPγS on Gα [18,19]. Although this system lacks the Gβγ subunits, it enables examination of the direct effect of small molecules on Gα. Using the silkworm experimental system, we first investigated whether NPD9055 induced the GDP/GTP exchange on Gα_i2 coupled to μ-opioid receptor in the absence of GPCR agonist. However, up to 100 µM NPD9055 did not induce GDP/GTP exchange on Gα_i2 unlike morphine, an agonist of μ-opioid receptor (Figure 3a). Therefore, we next examined the influence of NPD9055 on GPCR agonist-stimulated GDP/GTP exchange on Gα_i2. NPD9055 inhibited morphine-stimulated GDP/GTP exchange on Gα_i2 coupled to μ-opioid receptor (Figure 3b), suggesting a direct interaction of NPD9055 with Gα_i2. To assess the selectivity of inhibition, we also examined the effect of NPD9055 on Gα_s, a protein from another G-protein family, using a membrane fraction from silkworms expressing human Gα_s coupled to G-protein-coupled bile acid receptor (GPBA). NPD9055 did not affect taurolithocholic acid (TLC)-stimulated GDP/GTP exchange on Gα_s (Figure 3c). Collectively, these observations suggest that NPD9055 directly and selectively targets Gα_i, and inhibits GPCR agonist-stimulated GDP/GTP exchange on Gα_i.

Figure 3. NPD9055 inhibits GPCR agonist-induced GDP/GTP exchange on Gα_i in vitro.

(a) Effect of NPD9055 on the GDP/GTP exchange on Gα_i2 in the absence of a GPCR agonist. Membrane fractions were prepared from the fat body of silkworms expressing μ-opioid receptor-Gα_i2, and compound-stimulated exchange from GDP to [³⁵S]GTPγS on Gα_i2 was measured. Morphine was used as a positive control. (b) Morphine-stimulated GDP/GTP exchange on Gα_i2 coupled to μ-opioid receptor in the presence (100 μM) or absence of NPD9055. c) Membrane fractions were prepared from the fat body of silkworms expressing G-protein-coupled bile acid receptor (GPBA)-Gα_s, and taurolithocholic acid (TLC)-stimulated exchange from GDP to [³⁵S]GTPγS on Gα_s was measured in the presence (100 μM) or absence of NPD9055. Data are means ± SD (N = 3, n = 2).

Since NPD9055 inhibited the GPCR agonist-stimulated GDP/GTP exchange on Gα_i2 in vitro (Figure 3b), we next investigated whether NPD9055 affected G_i-regulated cellular processes. Following the GPCR agonist-stimulated GDP/GTP exchange on Gα_i and the subsequent dissociation of Gα_i from Gβγ, both components separately modulate their effector molecules[11]. One of the known effectors of Gα_i is Rap1GAP2, which inactivates Rap1, resulting in the phosphorylation of ERK/MAPK[20]. On the other hand, Gβγ regulates phospholipase Cβ, which hydrolyzes phosphatidylinositol 4,5-bisphosphate to the second messengers inositol 1,4,5-trisphosphate (IP₃) and diacylglycerol[21]. IP₃ stimulates Ca²⁺ release from the ER via the IP₃ receptor on the ER membrane[22]. Furthermore, both IP₃ and diacylglycerol activate protein kinases C, resulting in the phosphorylation of ERK/MAPK[23]. We first examined the influence of NPD9055 on Ca²⁺ homeostasis, and found that NPD9055 increased the intracellular Ca²⁺ levels directly after its addition to HeLa cells (Figure 4a). To investigate the source of the increased Ca²⁺ levels, we performed another Ca²⁺ assay in the presence or absence of extracellular Ca²⁺. NPD9055 increased the intracellular Ca²⁺ levels even in the absence of extracellular Ca²⁺ (Figure 4b), suggesting that the increased Ca²⁺ levels are at least partly derived from intracellular organelle(s), such as the ER. We next investigated whether NPD9055 affected the phosphorylation level of ERK/MAPK, and found that NPD9055 induced ERK/MAPK phosphorylation 10 min after its addition to HeLa cells (Figure 4c). These observations suggest that NPD9055 modulates Gβγ-regulated cellular processes, such as Ca²⁺ release and ERK/MAPK phosphorylation.

Figure 4. NPD9055 affects Gβγ-regulated cellular processes

(a) Effect of NPD9055 on the intracellularCa²⁺ levels in HeLa cells, in the presence of 1.8 mM CaCl₂. Data are means ± SD (N = 3, n = 3). (b) Intracellular Ca²⁺ levels in HeLa cells were measured after 3-min treatment with 10 μM NPD9055 (NPD) or DMSO in the presence (1.8 mM) or absence of CaCl₂ in the culture medium. Data are means ± SD (N = 3, n = 3). Statistical analysis was performed by using Student’s t-test. *p < 0.05, **p < 0.01. c) Effect of NPD9055 on the level of ERK/MAPK phosphorylation. HeLa cells were treated with NPD9055 or DMSO for 10 min, harvested, and the cell lysate analyzed via immunoblotting. Protein kinase C activator, 12-O-tetradecanoylphorbol-13-acetate (TPA), was used as a positive control.

To date, several G-protein modulators targeting Gα_i or Gβγ have been reported [13,14]. One of the most recognized Gα_i modulators is pertussis toxin produced by the bacterium Bordetella pertussis. Pertussis toxin catalyzes ADP-ribosylation of the key C-terminal cysteine residue in the Gα_i subfamily, and thereby interferes with the activation of G_i mediated by an agonist-bound GPCR[24]. Pertussis toxin inhibits Ca²⁺ level increase induced by a chemotactic peptide formyl-methionyl-leucyl-phenylalanine in differentiated HL-60 cells[25]. Since NPD9055 increased the intracellular Ca²⁺ levels (Figure 4a), NPD9055 seems to be mechanistically distinct from pertussis toxin. Another related G-protein modulator is Surfen/NSC12155. Gβγ binds to many effectors at a common “hot spot” that is obscured by GDP-bound Gα in the inactive state, and Surfen was identified as binding to the hot spot[26]. Surfen caused the dissociation of Gβγ from GDP-bound Gα without inducing a nucleotide exchange on Gα in vitro, and induced Ca²⁺ release and ERK/MAPK phosphorylation in differentiated HL-60 cells[26]. Based on these previous findings and on data presented in the current study (Figure 4a-4c), we hypothesized that the Gα_i-binder NPD9055 was similar to Surfen in terms of the mechanism of action. However, unlike NPD9055, Surfen did not increase the intracellular Ca²⁺ levels, induce ERK/MAPK phosphorylation, or was not cytotoxic up to 30 µM, at least in HeLa cells (Figure S5), suggesting a different mechanism of action. In addition, neither Gallein (a direct inhibitor of Gβγ) nor UK14304 (an agonist of G_i-coupled α₂-adrenoceptor expressed in HeLa cells) [27] showed any effect in the same assays (Figure S5).

Nevertheless, it is widely recognized that Gβγ can dissociate from Gα and regulate Gβγ effectors without an agonist-activated GPCR or nucleotide exchange. Some proteins, such as activator of G-protein signaling 3 (AGS3), share the G-protein regulatory (GPR)/GoLoco motif consisting of 19 amino acid residues, and interact with GDP-bound Gα_i/o via that motif. Proteins belonging to the GoLoco family act as GDP dissociation inhibitors that selectively interfere with the GDP/GTP exchange on Gα_i/o [28–30]. Importantly, despite the inhibitory activity of GDP dissociation, the GoLoco peptide derived from AGS3 causes the dissociation of Gβ₁γ₂ from Gα_i1[31]. A crystal structure analysis revealed that the GoLoco motif interacts with the nucleotide-binding pocket of Gα_i1, suggesting that Gβγ- and GoLoco-binding are mutually exclusive[32]. Based on the above reports on the GoLoco motif, we propose a mechanism of action for NPD9055, namely, that is (i) targeting GDP-bound Gα_i, (ii) dissociation of Gβγ from Gα_i, and (iii) activation of Gβγ-regulated signaling.

Further, to the best of our knowledge, neither the nuclear pore Nup107-160 complex nor dynamin-like 120 kDa protein indicated by chemical proteomics (Figure 2b) (neither of them was identified as NPD9055-binding in the pull-down assay using the photoaffinity beads) has been reported as a Ca²⁺ level modulator. Therefore, these cellular components do not appear to be the targets responsible for the NPD9055-associated phenotype. Taken together, our observations suggest that G_i is the primary target of NPD9055. Further experimental comparison with known G-protein modulators will provide additional insight into the mechanism of action of NPD9055.

Conclusion

By integrating morphology- and proteome-based profiling systems, we successfully identified the biologically uncharacterized indane derivative NPD9055. In addition, employing a chemical proteomics technique, we connected the indane chemotype with its cellular target Gα_i. Since the dysregulation of G-protein-mediated signaling is involved in various diseases including cancer, exploration of the available small-molecule modulators that directly target G-proteins is of great interest. NPD9055 may serve as a chemical tool to uncover the functions of Gα_i, Gβγ, and related molecules. Furthermore, by combining integrated phenotypic profiling and chemical proteomics, the developed strategy may open up new opportunities for drug discovery targeting unexplored molecules.

Author contributions

T.K., Y.F., M.M., P.J., M.U., J.W., and S.T. performed the biological experiments and analyzed the data. E.S. synthesized NPD9055 analogs. Y.K. prepared photoaffinity beads. T.K., Y.F., H.W., and H.O. conceived and designed the experiments. H.O., H.W., N.W., and S.Z. supervised the research project. T.K., Y.F., and H.O. wrote the manuscript with the support of other authors.

Supporting information

The Supporting Information is available free of charge on the journal website. Materials and methods; supporting tables and figures. (PDF and Excel files)

Acknowledgments

We thank M. Tanaka, H. Aono, K. Honda, and T. Nogawa for technical support, T. Shimizu and M. Kawatani for suggestions (Chem. Biol., RIKEN CSRS), and the RIKEN NPDepo for providing compounds. We also thank Y. Uehara (Iwate Medical University) for providing src^ts-NRK cells.

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplementary material

Supplemental data for this article can be accessed here.

Additional information

Funding

This research was supported in part by the Max Planck Society, the RIKEN-Max Planck Joint Research Center for Systems Chemical Biology, research grants from JSPS KAKENHI (Grant Numbers JP16H06276, JP17H06412, JP19K05746, JP18H05503, JP18K05366, JP19H05302, and JP20K05857), AMED (Grant Number JP20cm0106112), and the RIKEN Incentive Research Projects.