Figures and data
Establishing a robust and efficient workflow for assessing changes in protein thermal stability in living cells.
A. Schematic of the cell-based PISA workflow utilized in the large-scale chemical perturbation screen. B. Histogram plotting the log2 fold change values for all solubility (thermal stability) measurements (main panel). Box plot depicting the %CV of curated targets vs. all other proteins (inset). C. Plots highlighting the significant changes in protein solubility following treatment with OTS167 (top) or GSK-LSD1 (bottom). The mean log2 fold change of duplicate measurements is plotted on the x-axis and the mean nSD is plotted on the y-axis. Blue boxes contain proteins that exhibit an increase in solubility (increase in melting temperature) and orange boxes contain proteins that exhibit a decrease in solubility (decrease in melting temperature). Orange points represent the known target of each compound. D. The total number of significant solubility changes quantified upon treatment with each compound. The number of stabilizing events (increase in solubility) is plotted on the y-axis and destabilizing events (decrease in solubility) are plotted on the x-axis. E. Plot comparing the log2 fold changes measurements for each protein in BI 2536 (y-axis)-and volasertib (x-axis)-treated K562 cells.
Using cell-based PISA data to assess compound target engagement.
A. A plot of the mean log2 fold change (x-axis) and mean nSD (y-axis) of each known compound-target pair that were quantified in the screen. Blue boxes contain proteins that exhibit an increase in solubility (increase in melting temperature) and orange boxes contain proteins that exhibit a decrease in solubility (decrease in melting temperature). B and C. Plots depicting the mean log2 fold change of p38α (B) and PIK3CB (C) in response to each of the compound and DMSO treatments. Orange points indicate compounds known to target each protein. Blue points indicate other compounds that induce a significant change in thermal stability for each protein. Green dashed lines mark a SD of 3.5 for each protein. D. A protein-centric view of PLK1 solubility (thermal stability) in response to all treatments. Log2 fold change is plotted on the y-axis. The points represent each of the 256 treatments that were performed. Orange points indicate compounds known to target PLK1. Blue points indicate other compounds that result in a significant decrease in solubility of PLK1. Green dashed lines mark a SD of 3.5 cutoff for PLK1. E and F. K562 cells were treated with the indicated concentrations of palbociclib or BI 2536 for 15 minutes. Changes in protein thermal stability are represented as a log2 fold change in soluble protein abundance for each treatment in reference to a DMSO-treated control. (E). PLK1 activity was determined using a western blot for p-TCTP (S46) and total TCTP (F). G. Chemical structures of BI 2536, palbociclib and NVP-TAE-226. Common structural features with BI 2536 are highlighted in blue for Palbociclib and NVP-TAE-226. H. Co-crystal structure of CDK6 bound to palbociclib (blue) with BI 2536 (gray) modeled into the active site. I. Co-crystal structure of PLK1 bound to BI 2536 (gray) with palbociclib (blue) modeled into the active site.
A chemical perturbation screen of 70 compounds in K562 native extracts.
A. Schematic of the lysate-based PISA workflow utilized in the large-scale chemical perturbation screen. B. Histogram plotting the log2 fold change values for all solubility measurements in lysate-based screen. C. Plots highlighting the significant changes in protein solubility (thermal stability) following treatment with MK-2206 (top) or GSK-LSD1 (bottom). The mean log2 fold change of duplicate measurements is plotted on the x-axis and the mean nSD is plotted on the y-axis. Blue boxes contain proteins that exhibit an increase in solubility (increase in melting temperature) and orange boxes contain proteins that exhibit a decrease in solubility (decrease in melting temperature). Orange points represent the known target of each compound. D. Plot depicting the total solubility changes quantified upon treatment with each compound. The number of stabilizing events (increase in solubility) is plotted on the x-axis and destabilizing events (decrease in solubility) are plotted on the y-axis. E. Plot depicting the number of compounds for which a target was quantified in both cell-and lysate-based experiments (light gray), the number of compounds that caused a significant change in a known target in cell-based experiments (orange), the number of compounds that caused a significant change in a known target in lysate-based experiments (blue), and the number of compounds that caused a significant change in a known target in cell-or lysate-based experiments.
Cell-and lysate-based PISA are complimentary approaches for assessing mechanism of action.
A. A plot depicting the mean log2 fold change for all compound-target pairs quantified in cell-(x-axis) and lysate-based (y-axis) PISA. Dashed lines indicate a log2 fold change of +/-0.2. B. The log2 fold change of a selected number of compound-target pair in cell-and lysate-based PISA. C-E. Plots depicting the log2 fold change values for BRD4 (C), BRAF (D), and RIPK1 (E) in response to treatment with all compounds assayed using cell-and lysate-based PISA. Dashed lines indicate a log2 fold change of +/-0.2. F. HCT116 cell lysates (N=3) were treated with each compound or DMSO for 15 minutes and assay using PISA. The plot indicates the log2 fold change of several proteins in response to treatment with each compound. G. HCT116 cells were treated with the indicated concentration of each compound for 1 hour following the initiation of necroptosis using TSZ treatment. RIPK1 autophosphorylation was assayed by western blot using antibodies targeting p-RIPK1 (S116) or total RIPK1.
Disparities in cell-and lysate-based PISA can point toward secondary changes in protein thermal stability.
A and B. Plots depicting the mean log2 fold change values for RB1 (A) and CRKL (B) in response to treatment with all compounds assayed using cell (x)-and lysate (y)-based PISA. Dashed lines indicate a log2 fold change of +/-0.2. C. K562 cells were treated with each indicated compound at 10 µM for 15 minutes. Western blots were used to assess levels of p-CRKL (Y207) and total CRKL. D. A plot of the log2 fold change of CRKL in response to each indicated compound. E. Plots highlighting the significant changes in protein solubility (thermal stability) following treatment with AZD-7762. The mean log2 fold change of duplicate measurements is plotted on the x-axis and the mean nSD is plotted on the y-axis. Blue boxes contain proteins that exhibit an increase in solubility (increase in melting temperature) and orange boxes contain proteins that exhibit a decrease in solubility (decrease in melting temperature). Orange points represent the known target of AZD-7762 and the blue points represents tyrosine kinases that experience a significant change in soluble protein abundance (thermal stability). F. K562 lysates were treated with the indicated concentrations of AZD-7762 for 15 minutes and assayed using PISA. Changes in protein thermal stability are represented as a log2 fold change in soluble protein abundance for each treatment in reference to a DMSO-treated control. G. K562 cells were treated with the indicated concentration of bafetinib or AZD-7762 for 15 minutes and assayed using phosphoproteomic profiling. Plots depict the relative abundance of several phosphorylation sites following treatment with the indicated compounds. * indicate significant changes, which were determined using a permutation-based FDR (FDR – 0.05, S0 – 0.1).
Concordance of protein engagement for protein complex members and structurally similar kinases.
A. Correlation network graph for all observed kinases in the cell-based PISA assays. Human kinases (nodes) are connected based on correlation coefficients (RSpearman > 0.35). Nodes are colored based on kinase families. B. Example of significant correlation between RPS6KA1, RPS6KA2, and RPS6KA3 with consistent but sometimes small changes in solubility (thermal stability). C. Sub-graph of the network in A showing CMGC-family p38α/MAPK14 (isoforms 1 and 2) correlation with CAMK-family member kinases MAPKAPK2 and MAPKAPK3. Edges are colored based on RSpearman. D. Correlation plots for the subnetwork proteins in C. E. Protein interaction network derived from the BioPlex interactome highlighting that MAPKAPK proteins directly interact with p38α/MAPK14 in multiple cell lines. F. Sub-graph of the network in A highlighting the AGC protein kinases. Edges are colored based on RSpearman. G. AGC kinases AKT1 and AKT2 have a significant correlation driven in part by thermal stability effects of MK-2206. H. Structural comparison of the CMGC kinases GSK3A (Alphafold) and CDK16 complexed with the TKI rebastinib (PDB: 5g6v). Alignment generated an RMSD of 0.89 Å. I. Plot of the significant correlation between CDK16 and GSK3A derived from the network in A.
Establishing a robust and efficient workflow for assessing changes in protein thermal stability in living cells.
A. Schematic depicting how thermal shifts are measured in a PISA experiment. An increase in melting temperature (thermal stabilization) will result in an increase in the AUC and a positive log2 fold change in soluble protein abundance. A decrease in melting temperature (thermal destabilization) will result in a decrease in the AUC and a negative log2 fold change in soluble protein abundance. B. PISA schematic depicting a thermal window vs. a full melting curve. C-G. K562 cells (N=4) were treated with 10 µM palbociclib (D), ribociclib (E), or abemaciclib (F) for 15 minutes and assayed using PISA. (C) The log2 fold changes for specific proteins are plotted for each treatment using a full melting curve (left) or thermal window (48°C-58°C; right). (D-F). Data is presented as a volcano plot to highlight significant changes in abundance utilizing the thermal window (48°C-58°C). Significant changes were determined using a permutation-based FDR (FDR – 0.05, S0 – 0.1). (G). Log2 fold change values for selected proteins following treatment with each compound in D-F.
Establishing a robust and efficient workflow for assessing changes in protein thermal stability in living cells.
K562 cells were treated with each of the 96 compounds at 10 µM for 30 minutes. Three compounds and a DMSO control were assayed in duplicate (batch). All 96 compounds and 64 DMSO controls were assayed using thirty-two treatment batches. Following treatment, an equal number of cells were transferred to 10 PCR tubes. The cells were placed in a thermal cycler and heated across a thermal gradient from 48°C-58°C for 3 minutes. The cells were allowed to cool to room temperature for 5 minutes. An equal volume from each PCR tube was pooled and spun at 300 x g for 3 minutes to pellet cells. Cells were washed one with PBS and lysed in a buffer containing 0.5% NP-40, which will dissolve membranes without disrupting heat-induced protein aggregates. The lysates were centrifuged for 90 minutes at 21,000 x g to separate soluble protein from aggregates. An equal volume of each soluble fraction (∼20 µg) was prepared for LC-MS/MS analysis. Two treatment batches were combined for each TMTpro 16-plex. Each soluble fraction was reduced and alkylated. Each sample was precipitated onto SP3 carboxylate-coated beads to facilitate a buffer exchange. Proteomes were eluted off the SP3 beads into digestion buffer and digested with a combination of Lys-C and trypsin. Peptides from each sample were labeled with a unique TMTpro reagent. TMT-labelled peptides were pooled into a single sample, which was desalted using a sep-pak. Dried peptides were resuspended in HPLC buffer A and fractionated by basic reverse-phase HPLC. Twelve to twenty-four fractions were stage-tipped and analyzed on an Orbitrap Eclipse with a FAIMS device enabled (Thermo Fisher). Changes in thermal stability were determined by comparing soluble protein abundance in a compound-treated samples to vehicle-treated controls.
Establishing a robust and efficient workflow for assessing changes in protein thermal stability in living cells.
A. Minimal instrument time (assuming 12 fractions per plex) required to assay 96 compounds and 32 DMSO controls in duplicate using CETSA (left) or PISA (right). B. Total proteins quantified per TMTpro 16-plex. C. Schematic depiction of the filtering scheme used to define significant changes. D. Histogram depicting the log2 fold change measurements that were greater than 3.5 standard deviations from the mean. E,F. Total number of proteins passing each filter. G. Pie charts depicting the fraction of all measurements that pass each filter. H. Table highlighting proteins that pass the log2 fold change filter, but not the nSD filter and vice versa.
Establishing a robust and efficient workflow for assessing changes in protein thermal stability in living cells.
A-B. K562 cells (N=2) were treated with MK-2206 (top) or CCT128930 (bottom) at 10 µM for 30 minutes and assayed using PISA. The mean log2 fold change of duplicate measurements is plotted on the x-axis and the mean nSD is plotted on the y-axis. Blue boxes contain proteins that exhibit an increase in solubility (increase in melting temperature) and orange boxes contain proteins that exhibit a decrease in solubility (decrease in melting temperature). Orange points represent the known target of each compound. B. The plot displays the log2 fold change values for selected proteins following treatment with each compound in A. C. Plot comparing the log2 fold changes measurements for each protein in JQ-1 (x-axis)-and volasertib (y-axis)-treated K562 cells. Orange points represent the known targets of the compounds.
Establishing a robust and efficient workflow for assessing changes in protein thermal stability in living cells.
A. Plot highlighting the significant changes in protein thermal stability following treatment with PF-3758309. The mean log2 fold change of duplicate measurements is plotted on the x-axis and the mean nSD is plotted on the y-axis. Blue boxes contain proteins that exhibit an increase in solubility (increase in melting temperature) and orange boxes contain proteins that exhibit a decrease in solubility (decrease in melting temperature). The blue point represents the known target.
Using cell-based PISA data to assess compound target engagement.
A-B. A protein-centric view of p38α (A) and PIK3CB (B) solubility (thermal stability) in response to all treatments. Log2 fold change is plotted on the y-axis. The points represent each of the 256 treatments that were performed. Orange points indicate compounds known to target each respective protein. Blue points indicate other compounds that result in a significant thermal stabilization. Green dashed lines mark a SD of 3.5 cutoff for each treatment. C-D. K562 cells were treated with the indicated concentrations of NVP-TAE-226 or BI 2536 for 15 minutes. Changes in protein thermal stability are represented as a log2 fold change in soluble protein abundance for each treatment in reference to a DMSO-treated control (C). PLK1 activity was determined using a western blot for p-TCTP (S46) and total TCTP (D).
A chemical perturbation screen of 70 compounds in K562 native extracts.
A. Minimal instrument time required to assay 70 compounds and 10 DMSO controls in duplicate using TPP (left) or PISA (right). B. Total number of proteins quantified in each plex in the lysate-based screen. C. A plot of the mean log2 fold change (x-axis) and mean nSD (y-axis) of each known compound-target pair that were quantified in the lysate-based screen. Blue boxes contain proteins that exhibit an increase in solubility (increase in melting temperature) and orange boxes contain proteins that exhibit a decrease in solubility (decrease in melting temperature). D-G. Protein-centric view of HDAC1 (D), p38α (E), CDK2 (F), and CDK7 (G) solubility (thermal stability) in response to all treatments. Orange points indicate compounds known to target each respective protein. Blue points indicate other compounds that result in a significant thermal destabilization of each respective protein. Green dashed lines mark a SD of 3.5 cutoff for each compound.
A chemical perturbation screen of 70 compounds in K562 native extracts.
A. A venn diagram depicting the total number of compounds that cause a change in a known target in cell-and lysate-based PISA. B,C. Log2 fold change measurements for the indicated drug-target pair (GSK429286A_ROCK1 [A] and AZ 628_BRAF [B]) in cell-and lysate-based PISA.
Cell-and lysate-based PISA are complimentary approaches for assessing mechanism of action.
A. Changes in protein solubility (thermal stability) for the indicated proteins in cell-and lysate-based PISA are represented as a log2 fold change in protein abundance for each treatment in reference to a DMSO-treated control. Dashed lines indicate a log2 fold change of +/-0.2. B. A protein-centric view of RIPK1 solubility (thermal stability) in response to all treatments. Log2 fold change is plotted on the y-axis. The points represent each of the 256 treatments that were performed. Blue points highlight compounds that result in a significant increase in solubility (thermal stabilization). Green dashed lines mark a SD of 3.5 cutoff for RIPK1. C-F. HCT116 lysates were treated with necrostatin-2 (C), GSK2606414 (D), AZD-5438 (E), or tozasertib (F) (N=3) or DMSO (N=4) for 15 minutes and any changes in thermal stability were determined using PISA. Data is presented as a volcano plot to highlight significant changes in abundance. Significant changes were determined using a permutation-based FDR (FDR – 0.05, S0 – 0.1).
Disparities in cell-and lysate-based PISA can point toward secondary changes in protein thermal stability.
A. Plot highlighting the significant changes in protein solubility (thermal stability) following treatment with PF-3758309 in cells (no fill) and lysates (black points). The mean log2 fold change of duplicate measurements is plotted on the x-axis and the mean nSD is plotted on the y-axis. Blue boxes contain proteins that exhibit an increase in solubility (increase in melting temperature) and orange boxes contain proteins that exhibit a decrease in solubility (decrease in melting temperature).
Disparities in cell-and lysate-based PISA can point toward secondary changes in protein thermal stability.
A and B. A protein-centric view of CRKL solubility (thermal stability) in K562 cells (A) and lysates (B) in response to all treatments. Log2 fold change is plotted on the y-axis. The points represent each treatment that was performed in each screen. Green dashed lines mark a SD of 3.5 cutoff for each treatment. C. Plots depicting the log2 fold change values for CRK in response to treatment with all compounds assayed using cell-and lysate-based PISA. D. Changes in protein solubility (thermal stability) for the indicated proteins in cell-and lysate-based PISA are represented as a log2 fold change in soluble protein abundance for each treatment in reference to a DMSO-treated control.
Disparities in cell-and lysate-based PISA can point toward secondary changes in protein thermal stability.
A-C. K562 lysates were treated with the indicated concentrations of AZD-7762 for 15 minutes and assayed using PISA. Changes in protein thermal stability are represented as a log2 fold change in protein soluble abundance for each treatment in reference to a DMSO-treated control. Significant changes were determined using a permutation-based FDR (FDR – 0.05, S0 – 0.1).
Disparities in cell-and lysate-based PISA can point toward secondary changes in protein thermal stability.
A-F. K562 cells were treated with the indicated concentration of bafetinib or AZD-7762 for 15 minutes and assayed using phosphoproteomic profiling. A-D. Data is displayed as a volcano plot. Significant changes were determined using a permutation-based FDR (FDR – 0.05, S0 – 0.1). E. PCA plot of phosphoproteomic data. F. Total number of significant changes that result from each treatment (left) or significant changes with a log2 fold change > 1 or < -1 (right).
Concordance of protein engagement for protein complex members and structurally similar kinases.
A. Subgraph from Figure 7A highlighting the correlation between PRKC kinases. Edges are colored based on RSpearman. B. Correlation of PRKCB and either PRKCA or PRKCQ. These correlations are driven by consistent but relatively small changes in thermal stability. C. Non-protein kinases of the correlation network graph. Library compounds did not specifically target any non-protein kinases though these compounds elicited consistent solubility (thermal stability) changes in these kinases. Comparison of compound engagement for PIP4K2 non-protein kinases (D) PIP4K2A and PIP4K2B as well as (E) PIP4K2B and PIP4K2C. Compound engagement generates statistically significant correlations in both cells and lysates, with a larger solubility (thermal stability) effect size in cell-based assays. F-G. Significant correlation with small solubility (thermal stability) fold changes for the non-protein kinases SRPK1, SRPK2, and RIOK2.
