Deep learning reveals endogenous sterols as allosteric modulators of the GPCR–Gα interface
- Department of Computational Biology, Indraprastha Institute of Information Technology-Delhi (IIIT-Delhi), India
- Academy of Scientific and Innovative Research (AcSIR), India
- CSIR-Institute of Microbial Technology, Sector-39A, India
- Pharmacology Division, CSIR-Central Drug Research Institute, India
- Infosys Centre for AI, Indraprastha Institute of Information Technology-Delhi (IIIT-Delhi), India
Figures
Figure 1 with 3 supplements
Development, benchmarking, and validation of Gcoupler computational framework.
(a) Schematic workflow depicting different modules of the Gcoupler package. Of note, Gcoupler possesses four major modules, that is, Synthesizer, Authenticator, Generator, and BioRanker. (b) AUC–ROC curves of the finally selected model for each of the indicated GPCRs. Note: Experimentally validated active ligands and decoys were used in the testing dataset. (c) Bar graphs depicting the sensitivities and specificities of the indicated GPCRs with experimentally validated active ligands and reported decoys. (d) The AUC–ROC curve indicates the model’s performance in the indicated conditions. (e) Bar graphs indicating the prediction probabilities for each experimentally validated ligand. (f) Schematic workflow illustrates the steps in measuring and comparing the structural conservation of the GPCR–Gα-protein interfaces across human GPCRs. (g) Snake plot depicting the standard human GPCR two-dimensional sequence level information. Conserved motifs of the GPCR–Gα-protein interfaces are depicted as WebLogo. Asterisks represent residues of conserved motifs present in the GPCRs–Gα-protein interfaces. Of note, the location of the motifs indicated in the exemplary GPCR snake plot is approximated. (h) Schematic workflow illustrates the steps in measuring and comparing the structural conservation of the GPCR–Gα-protein interfaces across human GPCRs. (i) Representative structures of the proteins depicting highly conserved (low root mean square deviation [RMSD]) and highly divergent (high RMSD) GPCR–Gα-protein interfaces. PDB accession numbers are indicated at the bottom. (j) Heatmap depicting the RMSD values obtained by comparing all the GPCR–Gα-protein interfaces of the available human GPCRs from the protein databank. Of note, the RMSD of the Gα–protein cavity was normalized with the RMSDs of the respective whole proteins across all pairwise comparisons. (k) Heatmap depicting the pairwise cosine similarities between the in silico synthesized ligands of the GPCR–Gα-protein interfaces of the available human GPCRs using Gcoupler. (l) Schematic diagram depicting the hypothesis that the intracellular metabolites could allosterically modulate the GPCR–Gα interaction.
Figure 1—figure supplement 1
Gcoupler benchmarking using experimentally validated orthosteric ligands of GPCRs.
(a) AUC–ROC curves depicting the optimal probability cutoff recommended by G-means and Youden J index algorithms, respectively. Notably, the optimal threshold is indicated by the black dot on the ROC curve. (b) Molecular representation depicting the indicated experimentally validated orthosteric sites of the AA2AR, ADRB1, ADRB2, CXCR4, and DRD3 receptors, with the zoom-in inlet on the right highlighting the ligand and the cavity topologies. (c) Overlapping density plots comparing the distributions of synthetic compounds predicted to target the indicated receptors into High-Affinity Binders (HABs) and Low-Affinity Binders (LABs) using the Authenticator module of the Gcoupler package. (d) Bar graphs indicating the performance of the indicated models made for the experimentally elucidated orthosteric sites of the indicated GPCRs, generated using the Gcoupler workflow. (e) Confusion matrices indicating the relative proportions of experimentally determined ligands and their respective decoys (randomly selected to match # of ligands for each receptor). Note that the number of active ligands and decoys is mentioned at the bottom.
Figure 1—figure supplement 2
Gcoupler benchmarking using experimentally validated allosteric ligands of GPCRs.
(a) Molecular representation of the β2AR, CCR2, and CCR9 receptors with highlighted experimentally validated intracellular allosteric cavity and ligand. (b) Overlapping density plots comparing the distributions of synthetic High-Affinity Binders (HABs) and Low-Affinity Binders (LABs) predicted to target the allosteric cavities of the indicated GPCRs using the Gcoupler package. (c) Bar graphs depicting the performance metrics of the indicated classification model generated using Gcoupler for the highlighted allosteric cavities of the indicated proteins.
Figure 1—figure supplement 3
Run time comparison of Gcoupler and AutoDock.
(a) Flowchart depicting the entire workflow used to compute and compare the runtime using AutoDock and Gcoupler. Of note, the Synthesizer, Authenticator, and Generator module of Gcoupler are indicated alongside their key processes. (b) Ribbon diagram depicting the ADRA1A predicted structure using AlphaFold (ID: AF-P35348). (c) A waterfall chart depicts the timestamp information about the key steps involved in AutoDock. Time consumed at the indicated steps is mentioned in seconds and hours along the y-axis and above bars, respectively. (d) Waterfall chart depicting the timestamp information about the key steps involved in Gcoupler. Time consumed at the indicated steps is mentioned in seconds and hours along the y-axis and above bars, respectively. (e) Bar plot depicting the comparison between the total time consumed by the AutoDock and Gcoupler to predict the binding of experimentally validated ligands for ADRA1A (AF-P35348). (f) Box plot depicting the performance metrics of the 10-fold cross-validation of the training data obtained using Gcoupler. (g) Box plot depicting the performance metrics of the 10-fold cross-validation of the testing data obtained using Gcoupler. (h) Scatterplot depicting the relationship between AutoDock computed binding energies and Gcoupler computed prediction probabilities of the experimentally validated ligands of ADRA1A protein. (i) Density plot depicting the distribution of root mean square deviation (RMSD) between GPCR–Gα cavity across selected GPCRs–G-protein pairs. Of note, the RMSD of the Gα-protein cavity was normalized with the RMSDs of the respective whole proteins across all pairwise comparisons. (j) Heatmap depicting the RMSD values obtained by comparing all the GPCR–Gα-protein interfaces of the available human GPCRs from the protein data bank. (k) Heatmap depicting the RMSD values obtained by comparing the selected human GPCRs available in the protein data bank. (l) Density plot depicting the distribution of cosine similarities among the in silico synthesized ligands of the GPCR–Gα-protein interfaces of the available human GPCRs using Gcoupler.
Figure 2 with 4 supplements
Identification of endogenous, intracellular allosteric modulators of Ste2p using Gcoupler.
(a) Schematic diagram depicting the topology of all the cavities identified using the Synthesizer module of the Gcoupler Python package. Of note, the cavity nomenclature includes the cavity location, that is, EC (extracellular), IC (intracellular), and TM (transmembrane), succeeded by a numerical number. (b) Diagram depicting the three-dimensional view of the Ste2 protein, with highlighted Gα-protein-binding site (Gpa1) and the Gcoupler intracellular cavities (IC4 and IC5). The Venn diagram at the bottom depicts the percentage overlap at the amino acid levels between the Gα-binding site and predicted IC4 and IC5. (c) Schematic representation of the overall workflow used to predict the endogenous intracellular allosteric modulators of Ste2 receptor using Gcoupler and molecular docking technique. Of note, Yeast Metabolome Database (YMDB) metabolites were used as query compounds. (d) Overlapping density plots depicting and comparing the distributions of synthetic compounds predicted to target the IC4 and IC5 of the Ste2 receptor using the Gcoupler package. Of note, the Authenticator module of Gcoupler segregated the synthesized compound for each cavity (IC4 or IC5) into High-Affinity Binders (HABs) and Low-Affinity Binders (LABs). (e) AUC (area under the curve) plots representing the performance of the indicated models. Notably, the models were trained using the cavity-specific synthetic compounds generated using the Gcoupler package. (f) Scatter plots depicting the relationship (correlation) between the binding prediction probabilities using Gcoupler and binding free energies computed using molecular docking (AutoDock). (g) Scatterplot depicting the Pathway Over Representation Analysis (ORA) results of the endogenous metabolites that were predicted to bind to the GPCR–Gα-protein (Ste2p–Gpa1p) interface using both Gcoupler and molecular docking. (h) Alluvial plot showing five-level sub-activity spaces screening of the selected metabolites for IC4. (i) Schematic diagram depicting the workflow opted to narrow down on the single metabolic gene mutants. (j) Schematic diagram depicting the experimental design used to screen single metabolic gene mutants for α-factor-induced programmed cell death (PCD). Cell viability was assessed using a propidium iodide-based fluorometric assay. (k) Scatter plot depicting the impact of α-factor stimuli on cellular viability, assessed using propidium iodide-based fluorometric assay. The y-axis represents −log10(p-value) of the one-sample Student’s t-test between the normalized PI fluorescence of untreated and treated conditions. The x-axis represents the percentage inhibition or increase in cellular viability, estimated using a propidium iodide-based assay. The mutants reported to be involved in mating, PCD, or both are indicated in orange, green, and blue, respectively. The statistically non-significant mutants are indicated below the dashed line in black. (l) Heatmap depicting the relative enrichment/de-enrichment of differentially enriched metabolites (DEMs) in the indicated conditions. Of note, four biological replicates per condition were used in the untargeted metabolomics. (m) Venn diagram depicting the overlap between the predicted endogenous intracellular allosteric modulators of Ste2p and DEMs identified using untargeted metabolomics. (n) Mean-whisker plot depicting the relative abundance of ubiquinone 6 (CoQ6) and zymosterol (ZST) in the indicated conditions. Student’s t-test was used to compute statistical significance. Asterisks indicate statistical significance, whereas ns represents non-significance.
Figure 2—figure supplement 1
Ste2 protein cavities prediction using Gcoupler.
(a) Schematic diagram depicting the hypothesis that the intracellular yeast metabolites could allosterically modulate the GPCR–Gα-protein (Ste2p–Gpa1p) interface and, therefore, the Ste2p signaling pathway in yeast. (b) Overlapping Ste2p (GPCR) ribbon diagram depicting the structural similarity between the experimentally elucidated Ste2p structure before and after molecular dynamics simulation using GROMACS software. Root mean square deviation (RMSD) value is depicted at the bottom. (c) Flowchart depicting the key steps used for the molecular dynamics simulation of Ste2p using CHARMM-GUI and GROMACS software. (d) Line plot depicting the changes in the density (kg/m3) along the z-axis of the three-dimensional Ste2p structure. (e) Line plot depicting the decline in the potential (kJ/mol) over molecular dynamics simulation time in picoseconds (ps). (f) Line plot depicting the changes in the density (kg/m3) of the overall system over molecular dynamics simulation time in picoseconds (ps). (g) Line plots depicting the changes in the pressure (in bars) and temperature (in Kelvin) of the overall system during molecular dynamics simulation of Ste2 protein. (h) Line plot depicting the RMSD changes of the Ste2 protein during molecular dynamics simulation. Note that the RMSD values are indicated in nanometers (nm). (i) Ramachandran plot depicting the location of simulated Ste2 protein residues in the favorable and unfavorable regions. (j) Bar plot depicting the predicted cavity volumes (Å). (k) Scatterplot depicting the relationship between the cavity volume and surface area in the indicated cavities of the Ste2 protein. (l) Heatmap depicting the percentage of amino acids of the indicated cavities residing in the EC (extracellular), IC (intracellular), and TM (transmembrane) region of the Ste2 protein. (m) Scatterplot depicting the predicted drug score of the indicated cavities of the Ste2 protein. (n) Snake plot depicting the key amino acids of the Ste2 protein alongside their location within the indicated functional cavities. (o) Snake plot depicting the conservation of Ste2 protein at the amino acid level. Of note, Ste2 proteins from 15 related yeast species were used for computing the conservation score.
Figure 2—figure supplement 2
Comparison of the chemical composition of cavity-specific ligands of Ste2 protein.
(a) Venn diagram depicting the number of overlapping amino acids constituting the extracellular cavity 1 (EC1), intracellular cavity 4 (IC4), and intracellular cavity 5 (IC5) of the Ste2p. (b) Two-dimensional and (c) three-dimensional principal component analysis plots depicting the segregation of in silico synthesized ligands for the indicated cavities via Gcoupler. Of note, atom pair fingerprints were used as features for this analysis. (d) Scatter plot depicting the relationship between the depth of the pharmacophore features and the neighboring grids. (e) Donut charts illustrating the properties of amino acids constituting the pharmacophore for each of the indicated cavities. (f) Two-dimensional and (g) three-dimensional principal component analysis plots depicting the chemical properties of in silico synthesized ligands from five independent runs on the IC4 of the Ste2p using the Gcoupler Synthesizer module. Of note, atom pair fingerprints were used as features for this analysis. (h) Heatmap depicting the Tanimoto Similarities scores between the synthesized ligands for IC4 of Ste2p across five independent runs using Gcoupler.
Figure 2—figure supplement 3
Sanity check of Gcoupler predictions.
(a) Principal component analysis depicting chemical heterogeneity of synthetic compounds synthesized using Gcoupler across intracellular cavities (IC4 and IC5). Notably, the compounds were segregated into High-Affinity Binders (HABs) and Low-Affinity Binders (LABs) by the Authenticator module of the Gcoupler. (b) Heatmap depicting the chemical similarity of the de novo synthesized HAB (dark green) and LAB (light green) across intracellular cavities (IC4 and IC5) of the Ste2 protein. Tanimoto similarities were computed using Atomic fingerprints. (c) Heatmaps illustrating the relative enrichment of the indicated functional groups (RNH2: primary amine, R2NH: secondary amine, R3N: tertiary amine, ROPO3: monophosphate, ROH: alcohol, RCHO: aldehyde, RCOR: ketone, RCOOH: carboxylic acid, RCOOR: ester, ROR: ether, RCCH: terminal alkyne, RCN: nitrile) among the Gcoupler-based de novo synthesized binders of Ste2p receptor across IC4 and IC5. (d) Box plots depicting the model’s performances across the 10-fold cross-validation. The upper and lower box plots depict the performance metrics of models trained using the de novo ligands for intracellular cavities (IC4 and IC5), respectively, of the Ste2 protein. (e) Venn diagrams depicting the number of yeast metabolites from YMDB predicted to bind to the Ste2p–Gpa1p interface, predicted using Gcoupler and AutoDock. Of note, IC4 and IC5 represent intracellular cavities 4 and 5 of the Ste2 protein, respectively. (f) Principal component analysis depicting chemical heterogeneity of endogenous yeast metabolites predicted to bind to the intracellular cavities (IC4 and IC5) of the Ste2 protein. Notably, the compounds were segregated into High-Affinity Metabolites (HAMs) and Low-Affinity Metabolites (LAMs) by the Generator module of the Gcoupler. (g) Heatmap illustrating the relative enrichment of the indicated functional groups among the endogenous intracellular allosteric modulators (metabolites) of Ste2 receptor across IC4 and IC5. (h) Percentage stacked bar plot depicting the relative enrichment of the indicated functional groups of the HAMs and LAMs across intracellular cavities (IC4 and IC5) of the Ste2 protein. (i) Heatmap depicting PageRank score of the selected metabolites for IC4 and (j) IC5 with respect to different biological properties. (k) Box plots depicting the performance parameters of the indicated models generated using Gcoupler against the IC4 cavity of Ste2p. Note that the training and testing datasets were generated randomly (five iterations) from the in silico synthesized ligands from Gcoupler. (l) Scatter plots depicting the segregation of HAMs/LAMs (indicated in green and red) identified using the Gcoupler workflow with 100% training data. Of note, models trained on lesser training data size (25%, 50%, and 75% of HAB/LAB) severely failed to segregate HAMs and LAMs (along the y-axis). The x-axis represents the binding affinity calculated using IC4-specific docking using AutoDock. (m) Bar plots depicting the correlation values obtained between the Gcoupler prediction probabilities and AutoDock computed binding energies in the indicated training data size. (n) Box plots depicting the distributions for binding energies of the HAMs and LAMs computed using cavity-specific docking and full docking via Autodock for IC4 and IC5 of the Ste2p.
Figure 2—figure supplement 4
Untargeted metabolomics and genetic screening suggest an interlink between metabolites and Ste2p-mediated programmed cell death (PCD).
(a) Screenshot of the central yeast metabolism from the Kyoto Encyclopedia of Genes and Genomes (KEGG) indicating the key metabolic pathways (highlighted in red) harboring metabolic intermediates predicted as High-Affinity Binders of GPCR–Gα-protein (Ste2p–Gpa1p) interface by Gcoupler. (b) Venn diagram depicting the percentage overlaps of the mating-related genes reported in the literature and those used in the screening. The STE2 was the only common gene. (c) Heatmap depicting the growth curve profiles of the wild-type, single metabolic mutants, and ste2Δ under optimal growth conditions. (d) Schematic diagram depicting the experimental workflow of the untargeted metabolomics experiment. Notably, the experiment involves the usage of increasing doses of α-factor to induce PCD and to enrich the surviving cells selectively. (e) Box plot depicting the increase in the propidium iodide fluorescence upon the increasing concentration of α-factor treatment. Heat-killed (HK), untreated, and vehicle (DMSO-treated) were used as controls, whereas the increasing concentration of α-factor surviving wild-type cells was used as test conditions. The Mann–Whitney U test was used to calculate statistical significance. Asterisks indicate statistical significance, whereas ns represents non-significance. (f) Heatmap depicting the relative (de)enrichment of differentially enriched metabolites in the indicated conditions. Of note, four biological replicates per condition were used in the untargeted metabolomics. (g) Principal component analysis depicting the segregation of indicated samples based on their metabolomics profiles along the PC1, PC2, and PC3. (h) Correlation plot depicting Pearson’s correlation of the metabolome of the indicated samples. (i) Heatmap depicting the relative enrichment of all the detected metabolites in the indicated condition. HK, untreated, and vehicle (DMSO-treated) were used as controls, whereas the increasing concentration of α-factor surviving wild-type cells was used as test conditions. (j) Venn diagram indicating that 38 predicted intracellular allosteric modulators of Ste2p were identified in the untargeted metabolomics profiling. (k) Scatterplot depicting the Pathway Over Representation Analysis (ORA) results of the differentially enriched metabolites (treated vs. vehicle control) identified using untargeted metabolomics.
Figure 3 with 3 supplements
Elevated endogenous metabolite levels stabilize Ste2p–Gpa1p interactions and selectively inhibit GPCR signaling.
(a) Scheme representing the key steps opted for preparing Ste2p structure for downstream computational analysis. (b) Bar plots depicting the binding energies obtained by the docking of Ste2p and indicated metabolites across IC4 and IC5. (c) Line plots depicting the root mean square deviation (RMSD) changes over simulation timeframes from the three independent replicates of the indicated conditions in the indicated conditions. The spread of the data is indicated as standard deviation (SD). Notably, RMSD is provided in Angstroms (Å), whereas the simulation time is in nanoseconds (ns). (d) Workflow depicting the steps involved in Ste2p-miniG-protein docking using HADDOCK and PRODIGY web servers. (e) Bar plots depicting the fold change of the dissociation constant (Kd) in the indicated conditions. Notably, fold change was computed with respect to the wild-type condition (Ste2p-miniG-protein). Inlets represent molecular representations of Ste2p-miniG-protein and the highlighted interface residues. (f) The schematic diagram depicts the experimental workflow used to quantify α-factor-induced programmed cell death (PCD) using a propidium iodide-based cell viability fluorometric assay. Box plot on the right depicting the rescue from the α-factor-induced PCD in the indicated conditions as inferred using propidium iodide-based cell viability fluorometric assay (n = 9 or 10 biological replicates; heat-killed = 2). The y-axis represents the fold change of the propidium iodide fluorescence values with respect to their respective controls. Mann–Whitney U test was used to calculate statistical significance. Asterisks indicate statistical significance, whereas ns represents non-significance. (g) Schematic representation (left) of the experimental approach used to measure cell vitality and viability using microscopy-based FUN1 staining. Representative micrographs (right) depicting the FUN1 staining results in the indicated conditions, Scale 10 µm. Mean-whisker plot depicting the relative proportion of the vital and viable yeast cells observed using FUN1 staining in the indicated conditions (n = 3 biological replicates). A Student’s t-test was used to compute statistical significance. Asterisks indicate statistical significance, whereas ns represents non-significance. Error bars represent the standard error of the mean (SEM). (h) Schematic representation (left) of the experimental design for the mating assay (n = 3 biological replicates, each with three technical replicates). MATa yeast cells were pre-loaded with the metabolites and then mated with MATα cells to evaluate the mating efficiency. Representative micrographs in the middle qualitatively depict the mating efficiency in the indicated conditions. The bar plots on the right depict the mating efficiency (mean ± SEM) in the indicated conditions. Student’s t-test was used to compute statistical significance. Asterisks indicate statistical significance, whereas ns represents non-significance. (i) Schematic representation depicting the experimental design of phospho-MAPK activity-based Western blot. Bar plots depicting the p-Fus3 levels (mean ± SEM; n = 5–6 biological replicates after IQR-based outlier removal) in the indicated conditions. Error bars represent the standard error of the mean (SEM). A Student’s t-test was used to compute statistical significance. Asterisks indicate statistical significance, whereas ns represents non-significance. (j) Schematic representation (left) of the experimental approach used to measure the fluorescence in PFUS1-eGFP-CYC1 yeast cells. Representative micrographs (right) depicting the eGFP expression in the yeast cells in the indicated conditions. Scale 20 µm. Bar plot depicting the Corrected Total Cell Fluorescence (CTCF) value (mean ± SEM; n = 3 biological replicates) in the indicated conditions. A Student’s t-test was used to compute statistical significance. Asterisks indicate statistical significance, whereas ns represents non-significance.
-
Figure 3—source data 1
PDF file containing original western blots of p-Fus3, Fus3, and Pgk1 displayed in Figure 3i.
- https://cdn.elifesciences.org/articles/106397/elife-106397-fig3-data1-v1.pdf
-
Figure 3—source data 2
Original files for western blot analysis of p-Fus3, Fus3, and Pgk1 displayed in Figure 3i.
- https://cdn.elifesciences.org/articles/106397/elife-106397-fig3-data2-v1.zip
Figure 3—figure supplement 1
Docking and molecular dynamics (MD) simulations suggest the stability of the Ste2–metabolite complex.
(a) Ligplots depicting the interacting amino acid residues and atoms of the Ste2 protein and indicated metabolites for the intracellular cavities (IC4 and IC5). (b) Table indicating the Van der Waals (ΔEvdw), electrostatic (ΔEelec), polar solvation (ΔEpsolv), and non-polar solvation (ΔEnpsolv) energies along with net binding free energy (ΔGbind) in kcal/mol of the three independent replicates of the indicated metabolites across intracellular cavities IC4 and IC5 of the Ste2p. Note the values in the table are provided as mean ± standard errors. (c) Line plot depicting the variation in the root mean square deviation (RMSD) of the indicated metabolites bound to the intracellular cavity 4 (IC4) at the GPCR-Gα interface of the Ste2p, obtained using longer MD simulation runs. Notably, RMSD is provided in Angstroms (Å), whereas the simulation time is in nanoseconds (ns).
Figure 3—figure supplement 2
Divergent characteristics of site-directed missense mutants of STE2 gene.
(a) Line plots depicting the total energy decomposition of the individual amino acids of the Ste2p (three independent replicates) computed using molecular dynamics (MD) simulations in IC4 (left) and IC5 (right). (b) Molecular representation of the Ste2 protein structures (blue) and the miniGpa1–proteins (cyan), with the highlighted interface residues (red and green) in the indicated conditions. (c) Scatterplot depicting the net binding free energy ΔG (kcal/mol) of the Ste2p (with and without indicated metabolites) with miniG-protein (y-axis) and α-factor (x-axis). (d) Bar plot depicting root mean square deviation (RMSD) values of the Ste2 protein with indicated metabolite across intracellular cavities IC4 and IC5.
Figure 3—figure supplement 3
Exogenous supplementation of selective metabolites rescues α-factor-mediated programmed cell death (PCD).
(a) Growth curve profiles of the wild-type (control), ste2Δ, and wild-type yeast pre-treated with the indicated metabolites. (b) Schematic diagram (left) depicting the workflow opted to monitor α-factor-induced PCD in the wild-type yeast cells pre-loaded for 24 hr with ubiquinone 6 (CoQ6), zymosterol (ZST), and lanosterol (LST) at three different concentrations. Bean plots on the right depict the changes in the α-factor-mediated PCD in the indicated conditions. Notably, cell viability was assessed by computing the area under the curve (AUC) of the growth kinetics. The Mann–Whitney U test was used to calculate statistical significance. Asterisks indicate statistical significance, whereas ns represents non-significance. (c) Bean plot depicting the fold change in the α-factor-mediated programmed cell death, assessed using the propidium iodide-mediated programmed cell death in the indicated conditions. The Mann–Whitney U test was used to calculate statistical significance. Asterisks indicate statistical significance, whereas ns represents non-significance. (d) Representative micrographs (left) depicting the growth of the cells (mating response) pretreated with indicated metabolites in the indicated single dropout medium (SC-Lys and SC-Met). Quantifications of the mating response with respect to the SC-Met control are indicated as a bar graph (mean ± SEM) on the right (n = 3 biological replicates, each with three technical replicates). The Student’s t-test was used to compute the statistical significance. Asterisks indicate statistical significance, whereas ns represents non-significance.
Figure 4 with 3 supplements
Site-directed Ste2p mutants disrupt metabolite-mediated rescue.
(a) Workflow depicting the steps involved in Ste2p-miniG-protein docking of the wild-type and site-directed Ste2p mutants. Notably, docking was performed using HADDOCK and PRODIGY web servers. (b) Bar plots depicting the dissociation constant (Kd) fold change in Ste2p site-directed mutants and wild-type. Notably, fold change was computed with respect to the metabolite-influenced wild-type condition (Ste2p-miniG-protein). (c) The schematic diagram depicts the experimental workflow used to quantify α-factor-induced programmed cell death in generated site-directed missense mutants (T155D, L289K, and S75A), alongside reconstituted wild-type STE2 (rtWT), using a propidium iodide-based cell viability fluorometric assay. The box plot (left) depicts the increase in the relative proportion of dead cells upon α-factor exposure. Box plot (right) depicting the loss of rescue phenotype from the α-factor-induced programmed cell death in the indicated conditions when pre-loaded with metabolites as inferred using propidium iodide-based cell viability fluorometric assay. The y-axis represents the fold change of the propidium iodide fluorescence values with respect to their respective controls. The Mann–Whitney U test was used to calculate statistical significance. Asterisks indicate statistical significance, whereas ns represents non-significance. (d) Schematic representation (top) of the experimental approach used to measure cell vitality and viability using microscopy-based FUN1 staining. Representative micrographs (below) depicting the FUN1 staining results in the indicated conditions, Scale 10 µm. Mean-whisker plot depicting the relative proportion of the vital and viable yeast cells observed using FUN1 staining in the indicated conditions (n = 4 biological replicates). A Student’s t-test was used to compute statistical significance. Asterisks indicate statistical significance, whereas ns represents non-significance. Error bars represent the standard error of the mean (SEM). (e) Schematic representation (up) depicting the experimental design of phospho-MAPK activity-based Western blot. Bar plots (down) depicting the p-Fus3 levels (mean ± SEM) in the indicated conditions (n = 3 biological replicates). The y-axis represents the p-Fus3/Fus3 ratio for the stimulated condition normalized by its corresponding unstimulated sample. A Student’s t-test was used to compute statistical significance. Asterisks indicate statistical significance, whereas ns represents non-significance. (f) Schematic representation depicting the experimental design of RNA sequencing, featuring treatment duration and the sequencing parameters. (g) Heatmap depicting the expression of differentially expressed genes obtained from RNA sequencing in the indicated conditions. Notably, Control and LST represent yeast cells unloaded and pre-loaded with lanosterol, respectively. α-factor is represented as α, where plus and minus signs represent its presence and absence, respectively. (h) Schematic representation of the experimental workflow followed to deduce the impact of indicated metabolites treatment on isoproterenol (ISO)-induced, GPCR-mediated hypertrophy response in human (AC16) and neonatal rat cardiomyocytes. Notably, in the case of AC16 cells, wheat germ agglutinin (WGA) was used to stain the cardiomyocytes, whereas, for neonatal cardiomyocytes, alpha-sarcomeric actinin staining was used. (i) Micrographs depicting the human (above; green colored) and neonatal rat (below; red colored) cardiomyocytes in the indicated conditions. Scale 50 µm. (j) Box plots depicting the surface area of human (AC16) and neonatal rat cardiomyocytes in the indicated conditions. Statistical significance of indicated metabolites with untreated control and isoproterenol-treated conditions is indicated in green and gray text, respectively. Mann–Whitney U test with Bonferroni-corrected p-values was used to compute statistical significance.
-
Figure 4—source data 1
PDF file containing original western blots of p-Fus3, Fus3, and Pgk1 displayed in Figure 4e.
- https://cdn.elifesciences.org/articles/106397/elife-106397-fig4-data1-v1.pdf
-
Figure 4—source data 2
Original files for western blot analysis of p-Fus3, Fus3, and Pgk1 displayed in Figure 4e.
- https://cdn.elifesciences.org/articles/106397/elife-106397-fig4-data2-v1.zip
Figure 4—figure supplement 1
Effects of Ste2 mutants on shmoo formation in response to α-factor.
(a) Ligplots depicting the interacting amino acid residues of the mutant Ste2 proteins (mutation site indicated) and atoms of indicated metabolites at the intracellular cavities (IC4 and IC5). (b) Growth curve profiles of the reconstituted wild-type (rtWT) and indicated missense site-directed mutants. (c) Schematic representation depicting the overall experimental workflow designed for the quantitative assessment of the shmoos formation upon α-factor treatment. (d) Micrographs depicting the phase contrast images of the rtWT cells, pretreated with the indicated metabolites, with and without α-factor treatment, Scale 10 µm. Mean-whisker plot (right) indicating the relative proportions of shmoos forming cells in the indicated conditions. (e) Micrographs depicting the phase contrast images of the indicated site-directed missense mutants of the STE2 gene, pretreated with the indicated metabolites, with and without α-factor treatment, Scale 10 µm. Mean-whisker plot (right) indicating the relative proportions of shmoos forming cells in the indicated conditions. The Student’s t-test was used to compute the statistical significance of d and e. Asterisks indicate statistical significance, whereas ns represents non-significance. Error bars represent the standard error of the mean (SEM).
Figure 4—figure supplement 2
RNA-sequencing unveils genes involved in attenuating α-factor-induced cell death response.
(a) A table highlighting the key differentially expressed genes in the indicated conditions reported to be involved in regulating yeast mating response. Arrowheads and their direction represent the direction of differential expression. (b) A table highlighting the key differentially expressed genes in the indicated conditions reported to be involved in regulating yeast programmed cell death (PCD)-like signaling response. Arrowheads and their direction represent the direction of differential expression. (c) M–D plots (left) from the NOIseq analysis highlighting the differentially expressed genes (colored) in the indicated conditions. Bar plots (middle) depicting the prominent Gene Ontologies in the indicated conditions. Venn diagram (right) depicting the overlaps between the differentially expressed genes and the genes used in biochemical assays. (d) Venn diagram depicting the overlap between the differentially expressed genes obtained from the RNA-Seq analysis and the yeast knockout library. (e) Growth curve profiles of the wild-type and indicated knockouts from the yeast knockout library. (f) Box plot depicting the relative proportion of dead cells upon α-factor exposure in the indicated knockouts, from the yeast knockout library as inferred using propidium iodide-based cell viability fluorometric assay (n = 12 biological replicates from 2 experiments). The y-axis represents the fold change of the propidium iodide fluorescence values with respect to their respective controls. The Mann–Whitney U test was used to calculate statistical significance. Asterisks indicate statistical significance, whereas ns represents non-significance.
Figure 4—figure supplement 3
Evolutionary conservation of GPCR–Gα interface.
(a) Molecular representations depicting the topology of human and rat ADRB1 and ADRB2 receptors. GPCR–Gα interface cavities are color-coded at the intracellular sites. Of note, in the case of human ADRB1, the nomenclature of the two cavities detected at the interface includes ‘Cv’ (cavity) succeeded by the numerical number. (b) Bar plots depicting the binding energies obtained by the docking of Human ADRB1 (7BVQ), Human ADRB2 (5X7D), Rat ADRB1 (P18090), and Rat ADRB2 (P10608) with the indicated metabolites at the GPCR–Gα interfaces. (c) Table depicting the conservation of yeast Ste2p–metabolite (ZST, CoQ6, and LST) interacting residues in the adrenergic receptors from humans and rats. (d) The schematic workflow illustrates the steps involved in measuring and comparing the sequential and ligand-interaction conservation of GPCRs across six different species. (e) Donut charts illustrating the GPCR count across different species. (f) Phylogenetic tree representing 75 unique GPCRs from six species. (g) Empirical Cumulative Distribution Function (ECDF) plot depicting binding energy distribution of Gcoupler identified potential allosteric modulators along with five negative controls. (h) Heatmap depicting the pairwise d-values/p-values from the Kolmogorov–Smirnov statistical test under specified conditions. The lower triangle of the heatmap represents d-values, while the upper triangle shows p-values for the specified pair. The test provides two key outputs: the D-value, which represents the maximum difference between the cumulative distribution functions (CDFs) of the two groups, and the p-value, which indicates the statistical significance of the observed difference. Symbols *, **, ***, and **** refer to p-values <0.05,<0.01,<0.001, and <0.0001, respectively.
Author response image 1
Schematic representation of the key strategy used in the Synthesizer module of Gcoupler.
Author response image 2
Scatter plots depicting the segregation of High/Low-Affinity Metabolites (HAM/LAM) (indicated in green and red) identified using Gcoupler workflow with 100% training data.
Notably, models trained on lesser training data size (25%, 50%, and 75% of HAB/LAB) severely failed to segregate HAM and LAM (along Y-axis). X-axis represents the binding affinity calculated using IC4-specific docking using AutoDock.
Author response image 3
(a) Affinity purification of Ste2p from Saccharomyces cerevisiae. Western blot analysis using anti-His antibody showing the distribution of Ste2p in various fractions during the affinity purification process. The fractions include pellet, supernatant, wash buffer, and sequential elution fractions (1–4). Wild-type and ste2Δ strains served as positive and negative controls, respectively. (b) Optimization of Ste2p extraction protocol. Ponceau staining (left) and Western blot analysis using anti-His antibody (right) showing Ste2p extraction efficiency. The conditions tested include lysis buffers containing different concentrations of CHAPS detergent (0.5%, 1%) and glycerol (10%, 20%).
Author response image 4
Scatter plot comparing the predicted binding affinity calculated by Docking and MM/GBSA methods, against experimental ΔG (10.1007/s10822-023-00499-0).
Tables
Appendix 1—table 1
The list of primers used in PCR amplification of indicated genes and plasmids for Gibson Assembly.
| Primer name | Sequence (5′–3′) |
|---|---|
| pFUS1 Forward | GTAAAACGACGGCCAGTGAGCTCAATCCTTCAATTTTTCTGGCAACTTTTCTC |
| pFUS1 Reverse | GCGTGACATAACTAATTACATGACTCGAGTTACTTGTACAGCTCGTCCATGCCG |
| eGFP Forward | CCATCAAGTTTCTGAAAATCAAAGGATCCATGAGTAAGGGCGAGGAGCTGTTCACCG |
| eGFP Reverse | GCGTGACATAACTAATTACATGACTCGAGTTACTTGTACAGCTCGTCCATGCCG |
| pRS306 Forward | CTCGAGTCATGTAATTAGTTATGTCACG |
| pRS306 Reverse | GAGCTCACTGGCCGTCGTTTTAC |
Appendix 1—table 2
The list of primers used in cloning and site-directed mutagenesis.
| Primer name | Sequence (5′–3′) |
|---|---|
| STE2 FWD overlap | CTTTAACGTCAAGGAGGGATCCATGTCTGATGCGGCTCCTTCATTG |
| SDM1 STE2 FWD | GTCATGTGGATGACATCGAGAGCTAGAAAAACGCCGATTT |
| SDM1 STE2 REV | AAATCGGCGTTTTTCTAGCTCTCGATGTCATCCACATGAC |
| SDM4 STE2 FWD | TTTCAGATAAAAGTTATTTTCGACGGCGACAACTTCAAAAGGATA |
| SDM4 STE2 REV | TATCCTTTTGAAGTTGTCGCCGTCGAAAATAACTTTTATCTGAAA |
| SDM5 STE2 FWD | ACATTACTTGCTGTATTGTCTAAACCATTATCATCAATGTGGGCC |
| SDM5 STE2 REV | GGCCCACATTGATGATAATGGTTTAGACAATACAGCAAGTAATGT |
| STE2 REV overlap | CATAACTAATTACATGACTCGAGTCATAAATTATTATTATCTTCAGTCCAGAACTTTCTG |
| CYC.F | ATAATAATTTATGACTCGAGTCATGTAATTAGTTATGTCACGCTTAC |
| GAL1 REV BAMHI | GCCGCATCAGACATGGATCCCTCCTTGACGTTAAAGTATAGAGG |
Author response table 1
Comparison of de novo drug design tools.
SBDD refers to Structure-Based Drug Design, and LBDD refers to Ligand-Based Drug Design.
| Tools | Drug design Type | Open Source | Code Availability | End-to-End Solution | Command Line support |
|---|---|---|---|---|---|
| Gcoupler | SBDD+LBDD | Yes | Yes | Yes | Yes |
| Pocket Crafter | SBDD | Partially | Partially | No | Yes |
| DeepLigBuilder | SBDD | -- | No | -- | -- |
| AutoDesigner | LBDD | No | No | -- | Yes |
Additional files
-
Supplementary file 1
The table contains information about the structure of the human GPCR–Gα-protein complexes downloaded from the Protein Data Bank (PDB).
It also contains information about the PDB IDs, protein name, species information, whether it is used or not, UniProt IDs, and other information.
- https://cdn.elifesciences.org/articles/106397/elife-106397-supp1-v1.xlsx
-
Supplementary file 2
The table contains information about the normalized root mean square deviation (RMSD) of the pairwise comparison of the GPCR–Gα-protein interface (cavities) for the indicated GPCRs.
- https://cdn.elifesciences.org/articles/106397/elife-106397-supp2-v1.xlsx
-
Supplementary file 3
The table contains information about the chemical similarities of the de novo synthesized synthetic compounds of the indicated GPCR–Gα-protein interface (cavities).
Of note, the chemical similarity was computed using the Tanimoto coefficient on the Atomic fingerprints.
- https://cdn.elifesciences.org/articles/106397/elife-106397-supp3-v1.xlsx
-
Supplementary file 4
The table contains information about the YMDB metabolites predicted to bind at the intracellular cavities IC4 and IC5 of the Ste2 protein.
Of note, predictions were made using molecular docking and Gcoupler. It contains information about the YMDB ID, metabolite name, SMILES (Simplified Molecular Input Line Entry System), and cavity information. Of note, BE refers to binding energies, whereas probabilities were the outcome of Gcoupler.
- https://cdn.elifesciences.org/articles/106397/elife-106397-supp4-v1.xlsx
-
Supplementary file 5
The table contains information about the list of metabolite-gene pairs from the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway.
- https://cdn.elifesciences.org/articles/106397/elife-106397-supp5-v1.xlsx
-
Supplementary file 6
The table contains the results of the untargeted metabolomics.
Of note, the values represent normalized and imputed peak intensities of the indicated metabolites.
- https://cdn.elifesciences.org/articles/106397/elife-106397-supp6-v1.xlsx
-
Supplementary file 7
The table contains the TMM normalized read counts after RNA sequencing.
- https://cdn.elifesciences.org/articles/106397/elife-106397-supp7-v1.xlsx
-
Supplementary file 8
The table contains the list of differentially expressed genes (DEGs) of the Control groups, alongside the other standard output from the NOISeq analysis, such as M and D values, mean values of the two conditions, probability ranking, and other gene-associated features.
- https://cdn.elifesciences.org/articles/106397/elife-106397-supp8-v1.xlsx
-
Supplementary file 9
The table contains the list of differentially expressed genes (DEGs) of the Control versus LST group, alongside the other standard output from the NOISeq analysis, such as M and D values, mean values of the two conditions, probability ranking, and other gene-associated features.
- https://cdn.elifesciences.org/articles/106397/elife-106397-supp9-v1.xlsx
-
Supplementary file 10
The table contains the list of differentially expressed genes (DEGs) of the LST group, alongside the other standard output from the NOISeq analysis, such as M and D values, mean values of the two conditions, probability ranking, and other gene-associated features.
- https://cdn.elifesciences.org/articles/106397/elife-106397-supp10-v1.xlsx
-
Supplementary file 11
The table contains information about the comparison of de novo drug design and cavity detection tools.
Please note, SBDD refers to Structure-Based Drug Design, and LBDD refers to Ligand-Based Drug Design.
- https://cdn.elifesciences.org/articles/106397/elife-106397-supp11-v1.xlsx
-
Supplementary file 12
The table contains information about the known mutants from the literature that could modulate Ste2 signaling.
Of note, the table also contains information about the site of mutation, the nature of the mutation, its position within the intracellular cavities IC4 and IC5, its impact on the Ste2 signaling pathway, and references.
- https://cdn.elifesciences.org/articles/106397/elife-106397-supp12-v1.xlsx
-
Supplementary file 13
Details of the nucleotide and amino acid Multiple Sequence Alignments (MSAs) of the wild-type STE2 and the site-directed missense mutants of STE2.
Each row starts with the sequence identifier followed by the aligned sequence (in chunks) with the ending position of the aligned sequence provided at the end, separated by tabs. The mutation sites are highlighted in red (S75A), yellow (T155D), and blue (L289K).
- https://cdn.elifesciences.org/articles/106397/elife-106397-supp13-v1.docx
-
MDAR checklist
- https://cdn.elifesciences.org/articles/106397/elife-106397-mdarchecklist1-v1.docx
Download links
A two-part list of links to download the article, or parts of the article, in various formats.
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Deep learning reveals endogenous sterols as allosteric modulators of the GPCR–Gα interface
eLife 14:RP106397.
https://doi.org/10.7554/eLife.106397.3
