Identification of crosstalk between phosphoprotein signaling pathways in RAW 264.7 macrophage cells

Abstract

Signaling pathways mediate the effect of external stimuli on gene expression in cells. The signaling proteins in these pathways interact with each other and their phosphorylation levels often serve as indicators for the activity of signaling pathways. Several signaling pathways have been identified in mammalian cells but the crosstalk between them is not well understood. Alliance for Cellular Signaling (AfCS) has measured time-course data in RAW 264.7 macrophage cells on important phosphoproteins, such as the mitogen-activated protein kinases (MAPKs) and signal transducer and activator of transcription (STATs), in single- and double-ligand stimulation experiments for 22 ligands. In the present work, we have used a data-driven approach to analyze the AfCS data to decipher the interactions and crosstalk between signaling pathways in stimulated macrophage cells. We have used dynamic mapping to develop a predictive model using a partial least squares approach. Significant interactions were selected through statistical hypothesis testing and were used to reconstruct the phosphoprotein signaling network. The proposed data-driven approach is able to identify most of the known signaling interactions such as protein kinase B (Akt) --> glycogen synthase kinase 3alpha/beta (GSKalpha/beta) etc., and predicts potential novel interactions such as P38 --> RSK and GSK --> ezrin/radixin/moesin. We have also shown that the model has good predictive power for extrapolation. Our novel approach captures the temporal causality and directionality in intracellular signaling pathways. Further, case specific analysis of the phosphoproteins in the network has led us to propose hypothesis about inhibition (phosphorylation) of GSKalpha/beta via P38.

Figure 1. Heat-map of the correlation matrix between the input and the output variables.

The rows and columns correspond to inputs and outputs, respectively. Negative values of the correlation are small in magnitude (e.g., AKT to P40) compared to positive values of the correlation. Hence, to enhance the visualization, asymmetric color-scale is used.

Figure 2. Identification of statistically significant phosphoproteins (acting as inputs) in the regulation of the signaling pathways (PPs acting as outputs).

The labels on the X-axis are the names of the output PPs. For each output PP, one bar is drawn for each input PP. The Y-axis represents the ratio of the coefficient for the input in the actual model to the standard deviation of the corresponding coefficients in the random models. Thus, Y-axis is equivalent to z-score. The horizontal dashed-line denotes the threshold, 2.58 on the ratio, corresponding to 99% confidence level. The bars crossing the threshold line represent the statistically significant interaction. The names of the inputs with the absolute ratio-value greater than 90% of the threshold are also listed for each input.

Figure 3. Predictive power of the reduced models containing only significant predictors.

The X- and Y-coordinates represent the experimental and predicted values, respectively. The central diagonal line is the y = x line (i.e. perfect fit with no residual error). The dashed and dotted lines, denote the y = x±σ_f and y = x±2σ_f lines, respectively; where σ_f is the fit-error between the experimental data and the prediction made by the full model that included all the inputs and was obtained by linear-regression.

Figure 4. Reconstructed phosphoproteins signaling network in RAW 264.7 macrophages.

(A) Full network with 16 phosphoprotein nodes. Activation and inhibition are shown with black (arrow end) and red (blunt end) lines, respectively. Thickness is proportional to (square root of) the confidence in that interaction. (B) Short network after combining ERK1 and 2, GSKα and β, ST1A and B, and EZR and MOE in single node. To do so, the corresponding rows and columns in the matrix of significant ratios were averaged without any ambiguity in the signs of the incoming or outgoing edges.

Figure 5. Distribution of the consistency of pathways for nodes which are activated through two different pathways.

X-axis represents four possibilities: (1) neither path consistent (‘None’), (2) path 1 consistent, but not 2 (3) path 2 consistent but not 1, (4) both paths consistent (see text). Y-axis represents the number of experiments counted for each case.

Figure 6. Display of experimental data corresponding to the four cases of valid paths from P38 and/or AKT to GSKα/β in Figure 5B.

Red and green colors are for positive and negative values (log(x), where x is ratio of raw value at current time to the raw value at t = 0), respectively, with the darker color indicating larger magnitude. (A) None of the two edges are valid because the level of GSKα/β is changing in opposite direction to what the signals through the paths P38 → GSKα/β or AKT → GSKα/β would otherwise result in. (B) The effective signaling is through the path AKT → GSKα/β alone. (C) Only the path P38 → GSKα/β is valid. (D) Both paths are valid as indicated by the same color in each column for all rows. 1 min data was used for AKT and P38. 3 min data was used for GSKα/β.

Figure 7. Ligand distribution for all four cases of GSKα/β activation (discussed in Figure 5B).

X-axis and Y-axis represent the name of ligand and counts of the cases, respectively. For dual ligand experiments, the case is added to both of the ligands. The panels A–D also correspond to the heat-maps of Figure 6 A–D, respectively.

Figure 8. Prediction of data at 10 min using the experimental data at 3 min and the reduced models containing only significant predictors.

Numerical integration is used to eliminate the error due to discrete approximation of time-derivatives. The central diagonal line is the y = x line (i.e. perfect fit with no residual error). The dashed and dotted lines, denote the y = x±σ_f and y = x±2σ_f lines, respectively; where σ_f is the best (minimum) fit-error obtained by linear-regression between the experimental data at 3 min (as input) and the experimental data at 10 min (as output).