Proteome-wide solubility and thermal stability profiling reveals distinct regulatory roles for ATP

Abstract

Adenosine triphosphate (ATP) plays fundamental roles in cellular biochemistry and was recently discovered to function as a biological hydrotrope. Here, we use mass spectrometry to interrogate ATP-mediated regulation of protein thermal stability and protein solubility on a proteome-wide scale. Thermal proteome profiling reveals high affinity interactions of ATP as a substrate and as an allosteric modulator that has widespread influence on protein complexes and their stability. Further, we develop a strategy for proteome-wide solubility profiling, and discover ATP-dependent solubilization of at least 25% of the insoluble proteome. ATP increases the solubility of positively charged, intrinsically disordered proteins, and their susceptibility for solubilization varies depending on their localization to different membrane-less organelles. Moreover, a few proteins, exhibit an ATP-dependent decrease in solubility, likely reflecting polymer formation. Our data provides a proteome-wide, quantitative insight into how ATP influences protein structure and solubility across the spectrum of physiologically relevant concentrations.

Fig. 1

Effect of ATP and GTP on proteome thermal stability. a Experimental setup of 2D thermal proteome profiling (2D-TPP) using crude lysate system. Dotted rectangular box corresponds to one TMT10 experiment. Blue circles indicate untreated lysate and circle in increasing intensities of purple represent increasing concentration of ligand of interest added to the lysate. Data from three independent experiments have been analyzed. b Heat maps showing relative fold changes (FC) of protein abundance upon treatment with ATP (0.005, 0.05, 0.5, and 2 mM) compared to untreated crude lysate (first column on each plot) with increasing temperature (y-axis: 42, 44.1, 46.2, 48.1, 50.4, 51.9, 54.0, 56.1, 58.2, and 60.1 °C). c Heat maps showing relative FC of protein abundances upon treatment with GTP (0.001, 0.01, 0.1, and 0.5 mM) compared to untreated crude lysate (first column on each plot) with increasing temperature (y-axis: 42, 44.1, 46.2, 48.1, 50.4, 51.9, 54.0, 56.1, 58.2, 60.1, 62.4, and 63.9 °C). d, e Distribution of −log10 half-maximal effective concentration (pEC₅₀) values of different classes of proteins as annotated in UniProt stabilized by ATP (d), and GTP (e). Black line represents the median pEC₅₀ of annotated ATP- and GTP-binding proteins upon addition of ATP and GTP, respectively. Numbers above violin plots represent number of proteins. f Network diagram showing effect of ATP on proteasome stability. Nodes with green (filled) circles indicate ATP-stabilized subunits, and a red outline for nodes show known ATP-binding proteins of the complex. Edge thickness represents the Euclidean distance between the melting profiles of the different subunits. Thick lines are indicative of Euclidean distance less than 0.02. g Comparison of Euclidean distances between melting profiles of stabilized ATP-binding complex subunits and non-stabilized non-ATP-binding subunits, (left), and between stabilized ATP-binding complex subunits and stabilized non-ATP-binding subunits (right) within different complexes. Significance levels obtained from a Wilcoxon signed-rank test were encoded as ***p < 0.001. Numbers above violin plots represent numbers of non-ATP-binding subunits in protein complexes with at least one ATP-binding subunit. Violin plots represent relative densities. Center line in all box plots is the median, the bounds of the boxes are the 75 and 25% percentiles i.e., the interquartile range (IQR) and the whiskers correspond to the highest or lowest respective value or if the lowest or highest value is an outlier (greater than 1.5 * IQR from the bounds of the boxes) it is exactly 1.5 * IQR. Source data are provided as a Source Data file for panels B–G

Fig. 2

Effect of ATP depletion of proteome thermal stability. a Experimental setup of TPP on untreated cell (indicated in blue) and cells depleted of ATP by inhibiting glycolysis and oxidative phosphorylation with a combination of 10 mM 2-deoxyglucose and 1 nM Antimycin-A (indicated in grey). Dotted rectangular box corresponds to one TMT10 experiment. b Changes in median melting points of non-ATP-binding proteins, ATP-binding proteins, and ATP-binding proteins that were stabilized by ATP in the crude lysate experiment upon ATP depletion in cells from three independent experiments. Significance levels obtained from a Wilcoxon signed-rank test were encoded as *p < 0.05, **p < 0.01, and ***p < 0.001. c Changes in melting points of protein complex subunits stabilized by ATP in the crude lysate experiment (right), and all other proteins (left) upon ATP depletion in cells. Significance levels obtained from a Wilcoxon signed-rank test were encoded as *p < 0.05, **p < 0.01, and ***p < 0.001. d Changes in melting points of 19S and 20S proteasome subunits, and all other proteins upon ATP depletion in cells. Significance levels obtained from a Wilcoxon signed-rank test were encoded as *p < 0.05, **p < 0.01, and ***p < 0.001. Violin plots represent relative densities. For all box plots, the center line is the median, bounds of the boxes are the 75 and 25% percentiles (i.e., the IQR), and whiskers correspond to the highest or lowest respective value. If the lowest or highest value is an outlier (greater than 1.5*IQR from the bounds of the boxes) whiskers are exactly 1.5*IQR. Numbers above violin plots represent number of proteins. Source data are provided as a Source Data file for panels B–D

Fig. 3

Solubility proteome profiling characterizes broad solubilizing effects of ATP on the proteome. a Experimental setup of solubility proteome profiling (SPP). Cells were lysed by mechanical disruption and the resulting crude lysate was divided into ten aliquots. Two crude lysate aliquots were treated with vehicle and eight aliquots with different concentrations of a molecule (e.g., ATP—indicated in purple). One vehicle-treated aliquot and eight molecule treated samples were solubilized with a mild detergent (NP40—indicated in blue) and the other vehicle control was solubilized with strong detergent (SDS—indicated in orange). Dotted rectangular box corresponds to one TMT10 experiment. Proteins that showed at least 50% increase in abundance in NP40-processed molecule treatment (at least in one concentration) compared to NP40-processed vehicle, were fitted with sigmoidal dose-response curves and pEC_50,s values were calculated. b Heat map representation of proteome solubility from three independent experiments. The grey scale represents the log₂ ratio between NP40 and SDS-processed vehicle condition. The color scale represents the log₂ ratio between NP40-processed ATP treated and vehicle-treated conditions. c Solubility profiles of BANF1, DDX50, and FBL, following ATP (upper panel) and non-hydrolyzable ATP (AMP-PNP) (lower panel) treatment from three independent experiments, y-axis represents the log₂ ratio of NP40-processed ATP or AMP-PNP treated and vehicle-treated conditions. d Distribution of −log₁₀ half-maximal effective concentration (pEC_50,s) values of ATP-solubilized proteins allocated to different membrane-less organelles. Significance levels obtained from a Wilcoxon signed-rank test were encoded as *p < 0.05, **p < 0.01, and ***p < 0.001. e 2D density contours of log₂ fraction disorder vs. average isoelectric points of soluble and insoluble proteins, and of those solubilized by ATP. f Melting point differences of proteins solubilized by ATP (from SPP data) compared to all other proteins measured by TPP between 10 mM ATP and vehicle-treated crude lysate from two independent experiments. Significance levels obtained from a Wilcoxon signed-rank test were encoded as ***p < 0.001. Violin plots represent relative densities. For all box plots, the center line is the median, bounds of the boxes are the 75 and 25% percentiles (i.e., the IQR), and whiskers correspond to the highest or lowest respective value. If the lowest or highest value is an outlier (greater than 1.5*IQR from the bounds of the boxes) whiskers are exactly 1.5*IQR. Numbers above violin plots represent number of proteins. Source data are provided as a Source Data file for panels B–F

Fig. 4

SPP of ATP-depleted cells recapitulates effects observed in crude lysate. a Experimental setup of SPP in ATP-depleted cells. Cells were depleted of ATP by inhibiting glycolysis and oxidative phosphorylation with two combinations of 2-deoxyglucose (2DG) and Antimycin-A (AA) (D1: 0.1 nM AA and 1 mM 2DG, D2: 1 nM AA and 10 mM 2DG). The untreated cells and the cells from the two ATP-depleted conditions were divided into two aliquots each, of which one was solubilized with NP40 while the other using 1% SDS. All samples were digested with trypsin, labeled with different TMT10 isotope tags and analyzed by LC-MS/MS. b Change in solubility calculated for proteins binned according to their propensity for solubilization with ATP in crude lysate (pEC₅₀ x-axis), following ATP depletion in cells (from three independt experiments). Error bars represent standard error of the mean. Significance levels obtained from a Wilcoxon signed-rank test were encoded as *p < 0.05, **p < 0.01, and ***p < 0.001. c Change in solubility of FBL, DDX50, and BANF1 following ATP depletion in cells, measured by calculating the log₂ ratios of NP40-processed untreated and ATP-depleted conditions. d DNA-binding propensity of BANF1 in the presence of ATP. Recombinantly expressed and purified BANF1 was incubated with biotinylated double stranded DNA in the presence of increasing concentration of ATP. The DNA-bound fraction of BANF1 was pulled down using streptavidin beads and measured using quantitative mass spectrometry. Rel. DNA-bound fraction of BANF1 was calculated as the ratio protein bound to DNA in the presence of ATP compared to control sample without ATP. Data from four independent trials have been shown. Violin plots represent relative densities. For all box plots, the center line is the median, bounds of the boxes are the 75 and 25% percentiles (i.e., the IQR), and whiskers correspond to the highest or lowest respective value. If the lowest or highest value is an outlier (greater than 1.5*IQR from the bounds of the boxes) whiskers are exactly 1.5*IQR. Numbers above violin plots represent number of proteins. Source data are provided as a Source Data file for panels B–D

Fig. 5

SPP identifies proteins decreasing in solubility upon ATP addition. a Solubility profile for IMPDH1 and NUCKS1, fold change (FC) in y-axis represents the ratio of NP40-processed ATP treated and vehicle-treated conditions from three independent experiments. b Melting curves for IMPDH1 and NUCKS1 from two independent TPP experiments in crude lysates treated with vehicle (black) or 10 mM ATP (red) and corrected for solubility changes at 10 mM ATP using SPP data. c Maximal log2 fold changes observed for melting curves of insoluble versus soluble proteins in untreated crude lysate. Significance levels obtained from a Wilcoxon signed-rank test were encoded as *p < 0.05, **p < 0.01, and ***p < 0.001. Violin plots represent relative densities. For all box plots, the center line is the median, bounds of the boxes are the 75 and 25% percentiles (i.e., the IQR), and whiskers correspond to the highest or lowest respective value. If the lowest or highest value is an outlier (greater than 1.5*IQR from the bounds of the boxes) whiskers are exactly 1.5*IQR. Numbers above violin plots represent number of proteins. Source data are provided as a Source Data file for all panels

Fig. 6

Concentration-dependent proteome-wide effect of ATP. Density plots of pEC₅₀s of ATP-stabilized ATP-binding proteins, complexes, as well as of proteins solubilized by ATP as measured by 2D-TPP and SPP technologies. Despite a clear difference in the distributions, there is a substantial overlap, showing that changes in ATP concentrations will have simultaneous effects on the stability of ATP-binding proteins and complexes as well as on the solubility of disordered proteins