Cell Type–Specific Importance of Ras–c-Raf Complex Association Rate Constants for MAPK Signaling

See allHide authors and affiliations

Science Signaling  28 Jul 2009:
Vol. 2, Issue 81, pp. ra38
DOI: 10.1126/scisignal.2000397


We generated 17 c-Raf (RAF proto-oncogene serine-threonine protein kinase) mutants with altered Ras-Raf association and dissociation rates to investigate the role of electrostatically driven Ras-Raf association rates on epidermal growth factor (EGF)–activated mitogen-activated protein kinase (MAPK) signal transduction. Some of these mutants had compensating changes in association and dissociation rates, enabling the effects of changes in association rate to be distinguished from those of changes in affinity. In rabbit kidney (RK13) cells, these mutants affected downstream signaling, with changes in Ras–c-Raf association rates having a greater effect on MAPK signaling than did similar changes in dissociation rates. Mutants with compensating decreases in both association and dissociation rates stimulated less extracellular signal–regulated kinase (ERK)–dependent reporter activity than did wild-type c-Raf, whereas the converse was true for mutants with increased association and dissociation rates. In marked contrast, the mutants had little or no effect on signaling in human embryonic kidney (HEK) 293 cells. These two cell lines also showed distinct patterns of EGF-dependent ERK phosphorylation and signaling: ERK activation and signaling were transient in HEK293 cells and sustained in RK13 cells, with the difference resulting from the lack of negative feedback from ERK to Sos (Son of Sevenless) in the latter. Computer simulation revealed that, in the presence of negative feedback, changes in the rate of Ras–c-Raf binding have little effect on ERK activation. Thus, EGF-MAPK activation kinetics and feedback regulation is cell type specific and depends on the network topology.


Members of the epidermal growth factor (EGF) receptor (EGFR) family (also known as the ErbB receptor tyrosine kinases) transmit extracellular signals to the inside of the cell, eliciting various cellular responses, including cell proliferation, differentiation, and apoptosis (1). Signaling pathways downstream of the EGFR form a large, highly complex network (fig. S1), and EGFR, like other members of the ErbB family, has been implicated in human disease. EGFR signaling is highly robust because of redundancy, feedback loops, and the possibility of diverting signals into alternative pathways to achieve nearly identical outcomes. This robustness presents a problem for the design of drugs to target disease (2).

EGF binding and activation of the EGFR lead to the recruitment of adaptor proteins, such as Shc (Src homology 2 domain–containing transforming protein C) and Grb-2 (growth factor receptor–bound protein 2) to the phosphotyrosine residues of the EGFR (1, 3, 4). Binding of the proline-rich region of Sos1 (Son of Sevenless 1) to Grb2 translocates Sos1 to membrane, where it acts as a guanine nucleotide exchange factor for the small guanosine triphosphatase (GTPase) Ras (Fig. 1) (5). Ras in its active guanosine triphosphate (GTP)–bound form (Ras-GTP) can interact with effector molecules, such as Raf proteins, which leads to activation of the protein kinases MEK [MAPK (mitogen-activated protein kinase) kinase] and extracellular signal–regulated kinase (ERK) (68). Active ERK translocates to the nucleus, where it stimulates the activation of target genes. Active ERK also participates in a negative feedback loop, preventing Ras activation by phosphorylating Sos1 (9). In addition, the lifetime of active Ras depends on GTPase activating proteins (GAPs, here RasGAP) (10), which can be recruited to the plasma membrane through docking proteins [here, docking protein 1 (Dok1)] (11). Further down-regulation is achieved by phosphatases (12, 13) and by internalization of the EGFR (1, 14).

Fig. 1

Flow chart representation of the EGF-Ras-MAP kinase pathway as used for model generation. Activating processes are shown in black, inhibitory processes, in blue. EGFRi, internalized EGFR; RasGAP, here, p120GAP; TF, transcription factors phosphorylated by active ERK; PPN, nuclear phosphatase; SHP, protein-tyrosine phosphatase, phosphatase for EGFR; PP, Ser-Thr phosphatase PP, phosphatase for c-Raf; XPP, Ser-Thr phosphatase for MEK; MKP, MAP kinase phosphatase.

Interaction of the activated Ras-GTP with the downstream kinase c-Raf (RAF proto-oncogene serine-threonine protein kinase) constitutes a crucial element of the EGFR signal transduction pathway (6). The high affinity between Ras-GTP and the Ras-binding domain (RBD) of c-Raf (RafRBD) is mainly driven by the fast association rate constant (kon), which is the result of electrostatic steering due to the highly complementary charged surfaces of the two proteins (15). In a previous computational analysis, we showed that the electrostatic interactions and association rate constants of the Ras-RafRBD and of other Ras-effector complexes are evolutionarily conserved, suggesting that the magnitude of kon could be important for cellular signal transduction (16).

Here, we aimed to design and analyze the effect of introducing mutations into the Ras-RafRBD complex that have compensating effects on association and dissociation rate constants, so that the various mutants have similar equilibrium affinities (dissociation constants, Kd) (Fig. 2). This structure-based design, and the calculation of kon, dissociation rate constants (koff), and Kd values, was done with the protein design algorithm FoldX 2.8 ( (17, 18), which considers electrostatic steering in complex formation (15, 19, 20). We analyzed the effects of these mostly “mild” mutations (with changes in kon and koff of around 200 to 400%) in two cell lines that show transient or sustained ERK activation in response to EGF [human embryonic kidney (HEK) 293 and rabbit kidney (RK13) cells, respectively]. Computer simulation of the network was consistent with weak, or absent, negative feedback from ERK to Sos1 in RK13 cells and strong negative feedback in HEK293 cells. We found that, with strong negative feedback, moderate changes in kinetics and equilibrium binding constants had minor effects on ERK phosphorylation; with weak negative feedback, however, small changes in association rates had large effects on ERK activation.

Fig. 2

Amino acid residues in the Raf-RBD that were mutated to change affinities and kinetic rate constants regulating Ras and Raf-RBD interactions. (A) Mutations introduced to increase koff (R89L) and increase kon (N71K, D80K, and E125K). D38 of Ras is shown to depict its short-range interactions with R89 of c-Raf. (B) Mutations introduced to decrease koff (A85K) and decrease kon (T57D, K87D, and R73D). D38 of Ras is shown to depict its short-range interactions with R89 and the mutated K85 (blue) of c-Raf. Abbreviations for the amino acid residues are as follows: A, Ala; D, Asp; E, Glu; K, Lys; N, Asn; R, Arg; and T, Thr.


ERK phosphorylation kinetics after EGF stimulation in different cell lines

EGF is generally thought to elicit transient activation of ERK (2125); consistent with this, Western analysis of ERK phosphorylation kinetics in three cell lines [HEK293 cells (Fig. 3A), PC12 cells (fig. S4), and mouse embryonic fibroblasts (fig. S4)] showed fast activation [with the maximal amount of phosphorylated ERK (ERK-P) apparent at around 5 min], followed by fast ERK dephosphorylation. RK13 cells, however, showed an unusually sustained signal (Fig. 3B), with ERK phosphorylation taking 3 hours to decline to 40% of maximum. Sustained ERK activation after stimulation with EGF has been reported for carcinoma cell lines, such as A431 (epidermoid) and H3255 (lung adenocarcinoma) (14). However, in these cell lines, sustained ERK phosphorylation is the consequence of a constitutively active EGFR. Here, the difference in HEK293 and RK13 ERK-P kinetics cannot be the consequence of different receptor activation kinetics, because the time courses of the appearance and disappearance of phosphorylated EGFR (EGFR-P) are similar, even though RK13 cells have more EGF receptors (Fig. 3, C and D). Moreover, the difference in ERK phosphorylation kinetics cannot be explained on the basis of different affinities of human and rabbit EGFR for human EGF, because we used saturating concentrations of EGF in both cases (fig. S5).

Fig. 3

ERK and EGFR phosphorylation kinetics after EGF stimulation in HEK293 and RK13 cells. HEK293 (A and C) or RK13 (B and D) cells were serum-starved for 24 hours and then stimulated with EGF (50 ng/ml), lysed at indicated times, and analyzed by immunoblot. (A and B) Time course of ERK phosphorylation. (C and D) Time course of EGFR phosphorylation. Quantification of Western blot bands was done with ImageJ software.

The transient nature of the signal that follows EGF stimulation has been attributed mainly to the action of GAPs and to negative feedback from ERK-P to Grb-2 and Sos1 (9). In this negative feedback loop, activated ERK phosphorylates Sos1 at four serine residues in its Grb-2 binding domain, decreasing its binding affinity to Grb-2 (9). To determine if a decrease in negative feedback could account for the sustained ERK phosphorylation seen in RK13 cells, we followed the time course of Sos1 phosphorylation after EGF stimulation.

Sos1 phosphorylation results in a small shift of the main Sos1 band so that an upper band is detected (upshift) (24). Western analysis with an antibody directed against Sos1 that recognizes both phosphorylated and unphosphorylated forms revealed two Sos1 bands in HEK293 cells and showed that the upper band increased in intensity after EGF stimulation and then returned to basal intensity with a time course that lagged slightly behind that of ERK phosphorylation (Fig. 4A). We validated that this shift corresponds to Sos1 phosphorylation with ERK inhibitors, with which no upper band of Sos1 was detected (Fig. 4A). However, no noticeable upshift was observed in RK13 cells, and ERK inhibitor did not produce any effect on the Sos1 band in this cell line (Fig. 4B). Moreover, addition of moderate amounts of ERK inhibitor that do not fully eliminate ERK activation (Fig. 4, C and D) prolonged ERK phosphorylation in HEK293 cells (Fig. 4, E and G), but did not affect it in RK13 cells (Fig. 4, F and H).

Fig. 4

Effect of ERK inhibition on Sos1 phosphorylation kinetics and luciferase activity. HEK293 (A) or RK13 (B) cells were serum-starved and pretreated with ERK inhibitor II (see Materials and Methods) for 10 min and then stimulated with EGF (50 ng/ml). Cells were lysed at indicated time points and Sos1 was detected by immunoblot. The arrows in (A) indicate the upshift due to Sos1 phosphorylation. HEK293 (C) and RK13 (D) were transfected with luciferase-encoding reporter plasmids and the indicated amounts of ERK inhibitor II were added. The luciferase assay was performed as described. ERK phosphorylation kinetics in HEK293 cells (E and G) was assayed by preincubating cells with 25 μM ERK inhibitor II and then stimulating with EGF (50 ng/ml). Cells were lysed at indicated time points, and ERK 1/2-P immunoblots were performed. (E) ERK-P and total ERK kinetics on immunoblots; (G) ERK-P quantification with the ImageJ software. ERK phosphorylation kinetics in RK13 cells (F and H) was assayed by preincubating cells with 25 μM ERK inhibitor II and then stimulating with EGF (50 ng/ml). Cells were lysed at indicated time points, and ERK 1/2-P immunoblots were performed. (F) ERK-P kinetics on immunoblots; (H) total ERK.

Thus, these data suggest that the existence of little or no negative feedback from ERK-P to Sos1 contributes to sustained ERK activation in RK13 cells.

Effect of c-Raf kinetic mutants on EGF-dependent ERK phosphorylation

We used a luciferase reporter assay designed to measure ERK-dependent gene activation subsequent to Ras activation to analyze the long-term effects of c-Raf mutations on signal transduction in living cells (26). This assay has been used to show that, in RK13 cells, the magnitude of ERK target-dependent gene activation is linearly correlated with the in vitro Ras-RafRBD affinity (26); thus, it can be used to analyze the effects of the c-Raf mutants on MAPK signaling. We found good correlation between long-term ERK phosphorylation (measured by Western blot, see Materials and Methods) and the luciferase signal (fig. S6) and between the log of ERK inhibitor concentration and the decrease in luciferase signal (fig. S7).

One caveat for analyzing the mutants is the persistence of endogenous wild-type (WT) c-Raf. However, we found that 24 to 32 hours after transfection, more than 90% of c-Raf detectable by Western analysis was exogenous (fig. S8), consistent with the observation that c-Raf cannot be overexpressed in high amounts (27) and suggesting that its abundance is subject to autoregulatory negative feedback.

EGF stimulated a wide range of reporter activity in RK13 cells carrying the different mutants (fig. S9 and Fig. 5, A to C). In marked contrast, the different mutants had little effect on the response in HEK293 cells, with the sole exception of a slightly greater activity seen with the A85K mutant [a mutant with extremely high affinity for Ras relative to the other mutants that also binds slightly to Ras–guanosine diphosphate (GDP) (28); fig. S9]. The effects of the c-Raf mutants on reporter activity were confirmed by Western analysis of ERK phosphorylation in cells transfected with the extreme mutants (A85K and R89L) or WT c-Raf (fig. S6 and fig. S9). We also ruled out differences in expression of the c-Raf mutants as the cause of the changes in luciferase activity in RK13 cells (fig. S10).

Fig. 5

Comparison between predicted FoldX affinities (ΔGint) and FoldX kon values (ΔGkon) with experimental and simulated luciferase activity. (A) Correlation of luciferase activity and FoldX affinity (R = 0.34). (B) Correlation of luciferase activity and FoldX kon values (R = 0.74). (C) Luciferase activities from three to four repetitions measured in RK13 cells expressing the 17 different c-Raf mutants. The mutants on the left side all contain A85K, which decreases koff, plus compensating mutations that decrease kon. The mutants on the right side all contain R89L, which increases koff, plus compensating mutations that increase kon. The two mutants marked with an arrow are those with affinities closest to the WT, but also the greatest changes in both kon and koff. The error bars for luciferase activities and FoldX ΔG and ΔGkon values represent the SD. (D) Simulated luciferase activities of mutants with different association and dissociation rate constants for the Ras–c-Raf interaction (reaction 19 of the model). The color codes for (A), (B), and (C) are the same so that mutants can be identified in the correlation plots. The color code in (D) follows the same trend as that in (A), (B), and (C) in terms of changes in kon and koff: decreasing kon on the left side and increasing kon on the right side. The heavy black arrows point to mutants in which the Kd is predicted to be equivalent to that of the wild type, but with compensating changes in kon and koff.

There was a good correlation in RK13 cells between predicted changes in kon and luciferase activity (Fig. 5, B and C), whereas correlation of luciferase activity with the predicted ΔG (the Gibbs free energy, ΔG = −RT ln Kd) values was poorer (Fig. 5A). This does not contradict the perfect correlation between complex affinity and luciferase activity observed previously (26), because the mutations in the earlier study mainly influenced koff, but not kon.

The above results reflect the long-term activation of transcription factors regulated by ERK-P, but do not provide information on the kinetics of EGF-dependent ERK activation. Thus, we tested the effect of our mutants on ERK activation shortly after addition of EGF. Here, we monitored ERK activation by Western analysis. After addition of EGF to HEK293 cells harboring the c-Raf mutants, there was no significant change in ERK phosphorylation relative to that seen with WT c-Raf (fig. S9). With the RK13 cells, there was a significant decrease in EGF-dependent ERK phosphorylation with the R89L Raf mutant, which has a low binding affinity for Ras (fig. S9).

Thus, in HEK293 cells, c-Raf mutants do not affect either short-term or long-term activation of ERK by EGF. In RK13 cells, however, there were small differences in the short-term effects of the different c-Raf mutants on EGF-dependent ERK activation and substantial differences in the long-term effects.

To try to understand the differences in behavior between the two cell lines, we modeled the MAPK signaling pathway and simulated the effect of the mutations on ERK signaling.

Computer model of EGF signaling

Several computational models of different complexity have been published that enable the prediction of various experimentally observed aspects of EGF signal transduction [some recent models are found in (14), (2125), and (29)]. However, published models are often difficult to use or modify. Thus, we generated our own model on the basis of previous published models (14, 2125, 29), using when possible rate constants (see Supplementary Materials and Methods) and concentrations (25) that have been experimentally measured and incorporating, when needed, the latest published information on the EGF signaling network.

The computational model includes EGF-stimulated activation of the EGFR, subsequent activation of Ras by recruitment of adaptor proteins and Sos1, activation of the MAPK cascade (c-Raf, MEK, ERK), transport of active ERK to the nucleus, and transcriptional activation of target genes (Fig. 1). It also incorporates various factors that decrease EGF signaling: Negative feedback from ERK-P to Sos1, GAP recruitment, phosphatases, and receptor internalization can all decrease EGF signaling.

As with any computer model, this one simplifies the in vivo situation. For example, databases such as MINT (30) predict that at least 40 different proteins can associate with the intracellular part of the EGFR. Because we do not have experimental rate constants for many of these reactions we decided to see if we could simplify our model, while still capturing the key aspects of ERK activation. To do so, we performed various experiments.

First, we expressed nonactive dominant-negative forms of various key proteins. Transient expression of a nonactivatable Ras S17N mutant attenuated the effects of EGF on ERK phosphorylation in both RK13 and HEK293 cells (Fig. 6A), as did expression of a catalytically inactive form of c-Raf (c-Raf K375M) (Fig. 6A). These data indicate that the signal that activates ERK predominantly involves the Sos1–Ras–c-Raf pathway.

Fig. 6

Experimental analysis in RK13 and HEK293 cells. Effect of dominant-negative Ras (S17N) and a c-Raf mutant with impaired kinase activity on ERK phosphorylation. HEK293 (A) or RK13 (B) cells were transiently transfected with HRas S17N (a mutant that cannot be activated by Sos1) or c-RafK375M (a mutant with impaired kinase activity), starved for 24 hours, and stimulated with EGF (50 ng/ml).

Inhibition of phosphatidylinositol 3-kinase (PI3K) affected the amplitude of peak ERK-P (figs. S11 and S12) but did not change the transient or sustained time course of ERK phosphorylation in HEK293 and RK13 cells, respectively, or the qualitative behavior of our mutants (see below and fig. S13). Thus, we did not include it in the model. RK13 cells have many more EGFRs than do HEK293 cells [Fig. 3, C and D; see also (31)], and this was incorporated in the model.

The in silico time course of ERK phosphorylation after EGF stimulation in HEK293 cells (Fig. 6B) is in agreement with the in vivo data (Fig. 3A). Moreover, the model correctly predicts the time course of phosphorylated Sos1, the long-term duration of ERK phosphorylation for a time course of 7 hours, and EGFR activation kinetics (fig. S14).

To generate a model that could explain the sustained activation kinetics observed in RK13 cells, we kept the same kinetic values for all reactions and analyzed the effect of deleting negative feedback from ERK-P to Sos1-Grb2, the inhibition of Ras by GAP, or both. We found that diminishing negative feedback with all other parameters kept constant reproduced well our experimental results (Fig. 6C). Other experimental findings, including long-term activation kinetics and the EGF concentration effects in RK13 cells were also predicted correctly (fig. S15).

Computer model analysis of the effects of kinetic mutants in RK13 and HEK293 cells

We modeled the effect of the mutants on ERK-P by changing the rate constants and affinities of the Ras-Raf complex. We find that with the HEK293-like model, the two to four times higher-affinity mutants slightly increased the maximum of ERK-P, whereas basal ERK-P was not increased (Fig. 6B). Similarly, mutants with two times decreased kon or increased koff reduced the maximum ~10%, and four times lower-affinity mutants decreased the maximum ERK response (to about 60% of maximal activity). In contrast, the same changes in kinetic properties and affinities simulated in the “RK13-like” model led to much larger effects (Fig. 6C), especially for the ERK-P values after 3 hours, whereas the maximum short-term values are similar to the values for the “HEK293-like” model (Fig. 7).

Fig. 7

Computer modeling results for ERK-P in HEK-like and RK13-like model. Simulation results of the effect of changing Ras-Raf kinetic rate constants (reaction R19) on simulated ERK phosphorylation changes after EGF stimulation in the HEK-like (A) and the RK13-like (B) model. The inset in (A) shows the end levels of the simulation (from 120 to 180 min) with a different y scale, to demonstrate that the relative end levels of ERK-P in the HEK293 and the RK13-like models are similar for all mutants, but the absolute values differ substantially.

Thus, the effects of mutations were diminished under conditions with strong negative feedback (HEK293 cells). In contrast, the sustained ERK activation kinetics in RK13 cells leads to predicted long-term effects on reporter activity that correlate with Ras-Raf association rate constants and, to a lesser extent, affinities. Furthermore, the model reproduced the results of the experimental analysis, in which we found that changes in kon have greater effects on luciferase expression than do corresponding changes in koff (Fig. 5, C and D).


We used a combined experimental and computational approach to analyze the role of kinetic rate constants in cellular signal transduction. On the basis of x-ray structural information and using the principles of electrostatic steering, we designed 17 c-Raf mutants that affect the Ras-RafRBD complex, with compensating changes in association and dissociation rate constants. We analyzed the effects of these mutants on EGF-dependent ERK signaling in two cell lines; RK13, which shows sustained EGF-induced ERK activation, and HEK293, which shows transient ERK activation.

Negative feedback suppresses the effects of c-Raf mutants on MAPK signaling

Analysis of the mutants with a luciferase assay that reflects ERK activity showed that, in RK13 cells with weak negative feedback, the mutants had long-term effects on EGF-stimulated MAPK signaling. However, in cells with strong negative feedback, this was not the case [the small effect observed for the A85K c-Raf mutant could be explained by the fact that it also binds to Ras-GDP (28)]. Even though the relative effect of the mutations in the HEK293-like model are similar when compared with those in the RK13-like model, the quantitative magnitude of the effects on ERK phosphorylation are different. In the first case, the differences in the number of ERK-P molecules are less than 80 molecules per cell, whereas in the second case, differences of up to 2000 molecules of ERK-P are observed with the different mutations. Linear negative feedback loops will not change the relative magnitudes of active downstream kinases, but it is the absolute amount of proteins that determine signaling strength. This is illustrated well when comparing the effects produced by NGF and EGF in PC12 cells, where in both cases an ERK-P peak of similar magnitude is observed, but the phenotypic result depends on ERK-P concentration over time (32). Negative feedback loops are important for reducing noise and controlling the degree of expression or activity of the genes or proteins involved (33). In agreement with our results, it has been theoretically shown that a negative-feedback loop when compensating for changes in the amount of the drug target protein will neutralize the effect of this drug (34).

Association rate constants affect signaling

Kinetic properties and complex formation driven by electrostatic steering are important in signaling events, such as in the cell cycle–dependent regulation of MAPK signaling in yeast. For instance, Strickfaden et al. (35) showed that Cln2-CDK phosphorylates the scaffolding protein Ste5, which acts as a sensor for high G1 CDK activity, thereby disrupting its membrane localization and inhibiting Ste5-mediated signaling. The inhibition of Ste5-mediated signaling is proportional to the negative net charge on Ste5, which was modified by the addition of artificial phosphorylation sites. Such a kinetics-based regulatory mechanism has been suggested to provide evolutionary plasticity to proteins regulated by phosphorylation (36), because of the weak spatial requirements of the phosphorylated sites (the net charge counts, but not the position of the phosphorylation sites). A possible role for association rate constants in biological function was previously suggested when a correlation between kon and biological affinity was demonstrated in an antigen-antibody system (37). It has also been suggested that strong electrostatic steering allows for low Kd values, while at the same time permitting dynamic complex formation by allowing fast koff values (16).

Here, we show that Ras–c-Raf association rate constants are important for signal transduction in cells (RK13) with sustained ERK activation (due to reduced negative feedback). Mutations with similar affinity, but lower kon (less electrostatic steering) and koff, stimulated less ERK-dependent reporter activity than did WT c-Raf. Likewise, mutants with higher kon (more favorable electrostatic steering) and koff had higher activity. The same results were obtained when the two cell lines were modeled using the same parameters and obviating negative feedback from ERK-P to Sos1 in RK13 cells (Fig. 5, A to D). Thus, mutants with higher kon had greater biological activity than did those with similar Kd values but lower kon, suggesting that kon values are important for ERK activation.

It is difficult to understand why this should be the case, particularly because the main differences in ERK-P abundance are seen in the long term and not immediately after addition of EGF. To try to explain this, we have plotted the changes in concentration for all species in the model consequent to simultaneously increasing or decreasing kon and koff fourfold (see Appendix). The only differences we found were all downstream of Ras–c-Raf interaction and all species were affected. Careful analysis suggests that these downstream effects had to do with the relationship between the kinetics of binding and release of Ras and c-Raf and the activation kinetics of c-Raf. (Activation of c-Raf requires that the activated Ras stays bound to c-Raf for a certain amount of time.) When kon and koff were both four times higher than the wild-type rates, we found slightly less free c-Raf and concomitantly more c-Raf in complex with Ras-GTP and a slight increase in the abundance of all activated species downstream. When kon and koff were both four times lower than the wild-type rates, the changes are opposite in direction and larger in magnitude. There is an increase in the amount of Ras-GDP, concomitant with an increase in free c-Raf, which results in less activated ERK downstream (see Appendix). Thus, subtle effects at the competition level could result in substantial downstream effects in signaling. Our results illustrate how important computer modeling could be in understanding the intricacies of signal transduction.

Cell type–specific responses

As previously noted, different cell types can respond differently to the same stimulus and to the same mutant proteins. Indeed, we found that the effect of inhibiting PI3K on EGF-induced ERK activation in RK13 cells was opposite to that reported in HEK293 cells (38) (fig. S11). Thus, we found that at low EGF concentrations, inhibition of PI3K increased ERK phosphorylation, whereas the opposite was true for high EGF concentration (fig. S11).

All of the above raises the question of whether computer models used for simulations can be applied directly from one cell type to another, or whether each cell type will require its own in-depth experimental analysis. It has been proposed that common effector processing mediates cell-specific responses to various stimuli and, therefore, that models can be broadly applied to many cell types (39). This hypothesis was validated on different epithelial cell types, although the authors noted that the model failed when used on a T cell line. Our results suggest that assuming that the connectivity in a network is the same for all cells, with only the specific transducers and responses varying, is not necessarily true for different cell lineages (rabbit versus human kidney cells). Thus, all applications of modeling, including drug validation, predicting the effects of mutations, and so forth, need to be carefully analyzed before a model developed in one cell type can be applied to another.

The question remains if what we have found here for RK13 cells is just an odd curiosity resulting from anomalies of this cell line, or whether similar sustained activation of ERK downstream of the Ras–c-Raf interaction could take place in other cells. The latter would explain why electrostatic steering seems to be well conserved throughout the animal kingdom for the Ras–c-Raf complex (16). Alternatively, this could be because Ras–c-Raf participates in other signaling pathways (for instance, HRG, heregulin) where sustained activation is a norm and, therefore, rapid association rate constants could be crucial.

Materials and Methods

Structure-based design of Ras–c-Raf mutants using FoldX

We designed the c-Raf mutants on the basis of the 1GUA pdb file of the Ras-Raf complex, using the build model option in FoldX 2.8 (17, 18). During this design procedure, FoldX tests different rotamers and allows neighboring side chains to move. Mutagenesis was repeated five times for all mutations. Interaction energies were calculated with FoldX 2.8 ( (17, 18) at physiological ionic strength (0.125 M).

Computer simulations of EGF signal transduction

The model was generated on the basis of the latest experimental results (40) and previous models (see Supporting Material). Simulations were performed in SmartCell (41) with ordinary differential equations (ODEs).

Plasmids and mutagenesis in pCDNA3 c-Raf and Ras constructs

Single lysine mutations were introduced into pCDNA3:Raf fl with the QuikChange site-directed mutagenesis kit (Stratagene) using pCDNA3:Raf fl.WT as a template. For double and higher-order mutants the corresponding pCDNA3:Raf fl.WT mutant was used as template. All mutations were checked by sequencing. The pCDNA3 (Flag-StrepII) construct was obtained from M. Ueffing. The c-Raf K375M mutant and the S17N HRas construct was obtained from W. Kolch.

Cell culture and Western blot

HEK293 and RK13 cells were grown in normal medium [Dulbecco’s minimum essential medium (DMEM) supplemented with 10% fetal calf serum and 2 mM glutamine] at 37°C and 5% CO2. When indicated, cells were transfected with 6 μl of Lipofectamine 2000 per 6-cm dish (Invitrogen). Transfection efficiency was around 90%, as determined by transiently expressing green fluorescent protein. For short-term kinetic analysis after EGF stimulation, cells were transfected and 6 hours after transfection cells were washed (to remove transfection reagent) and incubated in serum-free medium (DMEM plus 2 mM glutamine). Twenty-four hours after transfection, cells were stimulated with EGF and samples were taken at the indicated time points for Western analysis.

To monitor the long-term effects of the mutations on ERK-P (fig. S9), we transfected cells in serum-free medium; 6 hours after transfection, cells were washed (to remove transfection reagent), incubated in normal medium, and stimulated with EGF. We determined that EGF was not degraded after several hours of incubation by checking the capacity of the cell culture supernatant to induce ERK phosphorylation when added to serum-starved cells. Cells were lysed in standard lysis buffer, containing protease (Protease Inhibitor Cocktail Complete EDTA free from Roche) and phosphatase inhibitors (Phosphatase Inhibitor Cocktail 1 and 2 from Sigma). Antibodies used were obtained from Sigma [ERK 1/2, ERK 12-P, SOS1, c-Raf, EGFR-P (Tyr1068)] or Calbiochem (EGFR). The antibody for the phosphorylated form of the receptor is raised against the region around pTyr1068. The antibody for the nonphosphorylated form of the EGFR was derived with the use of a peptide corresponding to the area around Tyr1068 as well. This region is conserved in human, mouse, rat, and rabbit. ERK inhibitors (ERK inhibitor II and ERK inhibitor) were purchased from Calbiochem. PI3K inhibitor LY29002 was purchased from Cell Signaling. PI3K inhibitor wortmannin was purchased from Sigma.

Luciferase reporter assay

RK13 cells or HEK293 cells were grown to 50% confluence in 6-cm dishes. Luciferase assay conditions were modified from those described earlier (26). Cells were transfected with 2 μg of reporter constructs (E743-tk80-luc) and 1.5 μg of pCDNA3 c-Raf construct in serum-free medium. Six hours after transfection, cells were washed and incubated in normal medium and stimulated with EGF. Twenty-four hours after transfection, cells were harvested and lysed, and luciferase activities were determined as described (42).


We thank N. Kuemmerer for help with mutagenesis and cell culture, B. N. Kholodenko and W. Kolch for critical reading of the manuscript, and F. Wittinghofer, M. Ueffing, and W. Kolch for valuable discussions and for providing many plasmids used in this work. The work was supported by the European Union (INTERACTION proteome, grant LSHG-CT-2003–505520).

Supplementary Materials

Materials and Methods

Fig. S1. Overview of EGF signal transduction.

Fig. S2. Comparing experimental and FoldX predicted kon and ΔG values.

Fig. S3. Predicted ΔGint values for Ras-Raf WT and mutant complexes as predicted by FoldX.

Fig. S4. ERK phosphorylation kinetics and Sos1 upshift in PC12 cells and mouse embryo fibroblasts.

Fig. S5. Dose-response curve of ERK-P 5 min after EGF stimulation over a concentration range of 0.075 to 50 ng/ml.

Fig. S6. Correlation of ERK phosphorylation and luciferase activity in RK13 cells.

Fig. S7. Effect of ERK inhibitor on luciferase activity, as a measure of ERK target gene–dependent gene activation, in RK13 cells.

Fig. S8. Transient expression of TAP (Flag-StrepII)-c-Raf plasmids.

Fig. S9. Luciferase assay and ERK-P kinetics for c-Raf WT, A85K, and R89L in RK13 and HEK293 cells.

Fig. S10. Abundance of c-Raf WT and mutants in RK13 and HEK293 cells.

Fig. S11. PI3K contribution to ERK phosphorylation at different EGF concentrations in RK13 cells.

Fig. S12. PI3K contribution to ERK phosphorylation in HEK293 cells.

Fig. S13. Luciferase activities for c-Raf WT, A85K, and R89L in HEK293 and RK13 cells treated with the PI3K inhibitors LY294002 or wortmannin.

Fig. S14. Comparing experimental and simulated time courses for Sos1, EGFR-P, and ERK-P in HEK293 cells and HEK293-like model.

Fig. S15. Comparing experimental and simulated time courses for EGFR-P and ERK-P in RK13 cells and RK13-like model.

Table S1. Comparing FoldX mutagenesis results and ionic strength dependence of WT with in vitro experiments using stopped-flow and isothermal titration calorimetry.

Table S2. Design of c-Raf mutants using FoldX.

Table S3. Modeling parameters.

Appendix: Sensitivity Analysis

References and Notes

View Abstract

Navigate This Article