Skip to main content
Molecular & Cellular Proteomics : MCP logoLink to Molecular & Cellular Proteomics : MCP
. 2012 Feb 8;11(6):M111.013292. doi: 10.1074/mcp.M111.013292

Cross-talk between Receptor Tyrosine Kinase and Tumor Necrosis Factor-α Signaling Networks Regulates Apoptosis but not Proliferation*

Elsa M Beyer , Gavin MacBeath §,
PMCID: PMC3433887  PMID: 22323825

Abstract

Although many of the signaling networks activated by receptor tyrosine kinases (RTKs) and cytokine receptors are well understood, how these networks interconnect is much less clear. We set out to determine how cells respond to simultaneous exposure to opposing signals and how their downstream networks process this information. Using six isogenic cell lines, each stably transfected with a different RTK, we found that, in each case, the cognate growth factor induced proliferation, whereas TNFα induced apoptosis. Surprisingly, when the cells were treated simultaneously with growth factor and TNFα, the growth factor enhanced, rather than antagonized, TNFα-induced cell death. In contrast, TNFα had no effect on growth factor-induced proliferation, suggesting that cross-talk between these networks is unidirectional. A quantitative, system-wide study of signaling at early and late time points corroborated this observation: proteins in the RTK networks were not affected by TNFα treatment, but proteins in the TNFα network were affected by growth factors. These studies also highlighted the stress mitogen-activated protein kinase proteins p38 and c-Jun N-terminal kinase as the key nodes of signal integration, and their activation states at an early time point correlated well with subsequent measurements of apoptosis. Knocking down cRaf reduced the growth factor enhancement of TNFα-induced apoptosis, highlighting its role as a regulator of network cross-talk upstream of p38 and c-Jun N-terminal kinase. Overall, we found that when cells encounter conflicting stimuli, their phenotypic response is determined not by the sum of isolated processes, but by how their signaling networks interconnect. This underscores the need to build mechanistic models of network integration as a first step in predicting cellular behavior in complex settings and in rationally designing combination therapies.


Receptor tyrosine kinases (RTKs)1 are single spanning cell surface receptors that transmit signals from the extracellular environment to the interior of a cell. Upon binding their cognate growth factors, they initiate signaling cascades that lead to diverse phenotypic responses ranging from adhesion to migration, proliferation to differentiation, and survival to apoptosis. The 58 RTKs encoded in the human genome activate largely overlapping signaling pathways, yet are able to induce a wide range of cell fates in response to diverse stimuli. We have previously shown that, when expressed in the same cellular background, different RTKs activate many of the same cytoplasmic proteins, but in ways that differ both qualitatively and quantitatively (1).

Programmed cell death, or apoptosis, is important in Metazoan development and in the maintenance of normal tissue homeostasis in adults. Insufficient apoptosis can manifest as cancer or autoimmunity, whereas excessive cell death contributes to degenerative diseases, immunodeficiency, and infertility (2). TNFα is an inflammatory cytokine that can either promote survival or induce apoptosis, depending on the cellular context. Receptors in the TNFR superfamily can activate the NFκB, JNK, and p38 pathways, as well as caspase signaling. The balance between pro-apoptotic and anti-apoptotic signals, which depends on the identity of the receptor and on the state of the cell, determines whether a cell exposed to TNFα survives, proliferates, or undergoes apoptosis (3). Although originally identified as a factor that promotes tumor cell death (4), it has subsequently been found that TNFα produced by tumor cells or in the tumor microenvironment often promotes tumor cell survival (510). As a result, TNFα antagonists are currently in clinical trials in patients with advanced cancer (1114).

In both physiological and pathological conditions, cells are regularly exposed to two or more external cues at the same time. These signals may be conflicting when, for example, a tumor cell secretes a pro-survival growth factor in an autocrine fashion and is also exposed to a pro-apoptotic cytokine, such as TNFα, that is secreted in a paracrine fashion from neighboring cells. The individual intracellular signaling networks downstream of each input are understood in great detail, but how these networks intersect when simultaneously activated has not been explored extensively. Identifying how these networks interconnect and identifying the key nodes at which cross-talk occurs constitutes the first step in building predictive models that capture cell decision processes. Some cross-talk between growth factor networks and the TNFα network has previously been reported (1521), and in several cases, this cross-talk has been shown to affect cell fate. For example, both epidermal growth factor (EGF) and insulin can antagonize TNFα-induced apoptosis in HT29 colon carcinoma cells, and they do so by different mechanisms (22).

We have previously shown that RTKs differentially activate the same signaling networks when placed in the same cellular background (1), and we proposed that the combination of these differences may allow different RTKs to generate diverse cellular outcomes in the same cell type. Cross-talk between RTK and TNFR signaling networks may therefore differ depending on the identity of the RTK, leading to different signaling and phenotypic responses. To further understand where cross-talk occurs between RTK and TNFR signaling networks and how signal integration affects cellular outcome, we treated six isogenic RTK-transfected cell lines individually or simultaneously with TNFα and the relevant growth factors. Quantifying proliferation and apoptosis under these conditions revealed that, irrespective of the identity of the RTK, growth factor treatment enhanced TNFα-induced apoptosis, but TNFα did not affect growth factor-induced proliferation. By carefully quantifying the activation states of a wide variety of intracellular signaling proteins, we identified several proteins that are stimulated by both growth factors and TNFα. Interestingly, cross-talk between these networks was unidirectional: whereas proteins in the TNFα network were affected by growth factor signaling, the converse was not observed. These studies also highlighted two key nodes of signal integration, JNK and p38. Interestingly, their activation states at an early time point correlated very well with the extent of subsequent apoptosis, suggesting that internetwork communication occurs at this level and plays a pivotal role in determining cell fate. Finally, cRaf was found to be a key node of cross-talk between the RTK and TNFR signaling networks, because knocking down cRaf reduced the ability of growth factors to enhance TNFα-dependent apoptosis. Overall, our studies highlight the need to develop mechanistic models of pathway integration as a first step in understanding how cells interpret conflicting signals to elicit the appropriate phenotypic response.

EXPERIMENTAL PROCEDURES

Cell Culture

The generation of isogenic RTK-transfected HEK293 cells was described previously (1). The cells were maintained in Dulbecco's modified Eagle's medium (DMEM; Mediatech; Herndon, VA) supplemented with 10% (v/v) fetal bovine serum (Hyclone; Logan, UT), 2 mm glutamine, 100 IU/ml Penicillin, 100 μg/ml streptomycin (all from Mediatech), as well as 150 μg/ml hygromycin B (Invitrogen). The cells were grown in the absence of hygromycin B for 48 h prior to all experiments.

Annexin V and Propidium Iodide Staining

To determine a saturating concentration of TNFα, untransfected parental HEK293 cells were treated for 48 h with five concentrations of TNFα (PeproTech, Rocky Hill, NJ) ranging from 12.5 to 200 ng/ml. Adherent cells were trypsinized and combined with floating cells, washed in PBS, and then resuspended in 200 μl of annexin V binding buffer containing annexin V-FITC and propidium iodide (Invitrogen). The samples were incubated for 15 min at room temperature in the dark, resuspended in 800 μl of annexin V binding buffer, and passed through 0.7-μm cell strainers. annexin V and propidium iodide staining was assessed by flow cytometry using an LSRII flow cytometer (BD Biosciences, Franklin Lakes, NJ).

For experiments with all cell lines, RTK-transfected HEK293 cells were treated with 100 ng/ml TNFα and/or saturating growth factor for 24 h. Growth factors were obtained from Peprotech, and concentrations used were as described previously (1). Untransfected parental HEK293 cells were treated with 100 ng/ml TNFα, and all of the cell lines were also mock treated with DMEM alone. The cells were collected, stained, and analyzed as above.

For inhibitor studies, EGFR-transfected cells were pretreated for 1 h with 10 μm SB202190 and/or 20 μm SP600125 (Sigma-Aldrich) or mock inhibited with DMSO alone. The total volume of DMSO was kept constant. The cells were then stimulated with EGF and/or TNFα or mock treated with DMEM for 24 h. The cells were collected, stained, and analyzed as above.

BrdU Incorporation Assays

The cells were treated with TNFα and/or growth factor or mock treated for 24 h. One hour prior to the end of the incubation, BrdU (BD Biosciences) was added to the culture medium to a final concentration of 20 μg/ml. After a 1-h incubation, trypsinized and floating cells were collected, washed twice with PBS, resuspended in cold 70% ethanol, and stored at 4 °C in the dark for 2–5 days. After ethanol fixation, the cells were resuspended in 1 ml of cold 0.1% Triton X-100, 0.1 m HCl and incubated on ice for 1 min. The cells were washed in 5 ml of room temperature denaturation buffer (150 μm sodium chloride, 15 μm sodium citrate), resuspended in 1 ml of denaturation buffer, and heated at 95 °C for 5 min. After cooling on ice for 5 min, the cells were resuspended and added to 5 ml of antibody dilution buffer (0.1% Triton X-100, 1% bovine serum albumin in PBS). After centrifugation, the cells were stained with FITC-conjugated anti-BrdU antibody (BD Biosciences) in antibody dilution buffer and incubated at room temperature for 30 min. The cells were washed with antibody dilution buffer, resuspended in 0.5% BSA in PBS, and passed through 0.7-μm cell strainers. BrdU incorporation was assessed by flow cytometry using an LSRII flow cytometer (BD Biosciences).

Cell Lysates

For the initial time course experiment, platelet-derived growth factor receptor-beta (PDGFRβ)-expressing cells were treated with 100 ng/ml TNFα and 4 nm platelet-derived growth factor-BB (PDGF-BB) and lysed at various times up to 48 h. For experiments with all cell lines, each RTK-transfected cell line was treated with growth factor and/or TNFα or mock treated with DMEM for 10 min or 16 h. For inhibitor studies, EGFR-expressing cells were pretreated for 1 h with 10 μm SB202190 and/or 20 μm SP600125 or mock inhibited with DMSO alone. The total volume of DMSO was kept constant. The cells were then stimulated for 10 min or 16 h with EGF and/or TNFα or mock treated with DMEM. In all of the experiments, for time points up to 4 h, the cells were washed twice with cold PBS, followed by the addition of 0.5 ml of lysis buffer (50 mm Tris-HCl, 1% Nonidet P-40 (v/v), 5 mm EDTA, 1 mm NaF, pH 8.0, supplemented with 10 mm β-glycerol phosphate, 1 mm phenylmethanesulfonyl fluoride, 1 mm sodium orthovanadate, 1% phosphatase inhibitor cocktail II; (Sigma), and 1 Complete-Mini Protease Inhibitor cocktail tablet (Roche Applied Science) per 10 ml). For time points longer than 4 h, floating cells were also collected and lysed in 200 μl of lysis buffer. All of the lysates were cleared by centrifugation at 20,000 × g for 15 min at 4 °C. Total protein concentration was determined using the MicroBCA protein assay (Pierce).

Immunoblotting

Prior to immunoblotting, lysates were boiled in standard SDS gel loading buffer and loaded onto 10 or 15% polyacrylamide gels for the initial time course experiment or E-PAGETM 48 8% gels (Invitrogen) for subsequent experiments. After separation by electrophoresis, the proteins were transferred to nitrocellulose, and the membranes were blocked with 5% nonfat dry milk (w/v) in Tris-buffered saline (20 mm Tris, 150 mm NaCl, pH 7.6) containing 0.1% Tween 20 (v/v). The membranes were probed using rabbit-derived primary antibodies from Cell Signaling Technologies (Beverly, MA). The bands were detected with IRDye 680-labeled goat anti-rabbit IgG (LI-COR Biosciences, Lincoln, NE) and imaged using an Odyssey infrared imaging system (LI-COR Biosciences). The intensity of each band was quantified and then normalized based on the protein concentration of the lysate.

Relative phosphorylation or total levels were calculated for each protein by dividing each value by the maximum observed value across the six cell lines and four conditions. The reported relative levels are the average of two biological replicates.

siRNA Transfections

PDGFRβ-expressing cells were reverse transfected with Dharmacon siRNAs (Thermo Scientific, Lafayette, CO) according to the manufacturer's protocol. Briefly, the cells were seeded in 6-well plates containing 50 nm ON-TARGETplus SMARTpool RAF1 or ON-TARGETplus nontargeting pool. After 48 h, the cells were treated with PDGF-BB and/or TNFα for 24 h. The cells were then stained with annexin V and propidium iodide or lysed for immunoblotting as described above.

Partial Least Squares Regression (PLSR)

The PLSR models were generated using SIMCA-P (Umetrics AB, Umea, Sweden) and XLSTAT-PLS (Addinsoft) software packages as described previously (2325). Briefly, the signaling and phenotypic response data were mean-centered and scaled to unit variance before performing PLSR. When performing PLSR, the mean values of signaling measurements were used as the X data set, and the mean values of phenotypic response data were used as the Y data set. The modeling was performed according to the following iterative formulas,

graphic file with name zjw00612-4142-m01.jpg
graphic file with name zjw00612-4142-m02.jpg

where Ei represents the residual of the ith principal component, with score vector ti, weight vector wi, and loading vector pi, and T represents transpose. Fi represents the residuals of the ith dependent principal component, with score vector ti and loading vector qi, and bi represents the coefficient characterizing the inner relation between the independent and dependent principal components.

The PLSR models were evaluated for goodness of fit (R2Y), indicating how well the variation of the phenotypic response (proliferation or apoptosis) is explained by the model and Q2 derived using leave-one-out cross-validation and indicating how well the phenotypic response can be predicted. The algorithm of PLSR and calculation of R2 and Q2 have been explained previously (1). A model with R2 > 0.7 and Q2 > 0.4 is considered as a good model with biological data (26). The contribution of variables on Y was evaluated using the variable importance in the projection (VIP) value. Generally, the variables with a VIP larger than 1 were considered as the most relevant for explaining Y, and those with a VIP smaller than 0.5 were insignificant and eliminated in model construction.

RESULTS

Cross-talk Regulates Apoptosis but Not Proliferation

We previously generated six isogenic cell lines based on HEK293-FlpIn cells, each stably expressing a different RTK: EGFR, fibroblast growth factor receptor 1 (FGFR1), insulin-like growth factor 1 receptor (IGF1R), hepatocyte growth factor receptor (MET), neurotrophic tyrosine kinase receptor type 2 (NTRK2), and PDGFRβ (1). To compare how different RTK-expressing cells respond to simultaneous stimulation with growth factor and TNFα, we used saturating concentrations of ligand. To determine a saturating concentration of TNFα that induced maximal apoptosis in HEK293 cells, the untransfected parental cell line was treated for 24, 48, or 72 h with five concentrations of TNFα ranging from 12.5 to 200 ng/ml. The cells were then stained with annexin V-FITC and propidium iodide to measure phosphatidylserine exposure and membrane permeability, respectively. The percentage of positive cells was determined by flow cytometry. Maximal apoptosis was observed with 100 ng/ml TNFα after 24 h (data not shown). Saturating concentrations of growth factor were determined previously (1): 12 nm EGF, 16 nm acidic fibroblast growth factor (aFGF), 12 nm insulin-like growth factor 1 (IGF1), 12 nm hepatocyte growth factor (HGF), 4 nm bone-derived neurotrophic factor, and 4 nm PDGF-BB.

To determine how cells respond to conflicting growth factor and TNFα signals and whether response depends on the identity of the RTK/growth factor, the six RTK-transfected cell lines were treated for 24 h with the appropriate growth factor and TNFα, either individually or simultaneously (Fig. 1). The untransfected parental cell line was also treated with TNFα, and all of the cell lines were mock treated with media as a control. Cells were stained with annexin V-FITC and propidium iodide, and the percentage of positive cells was determined by flow cytometry (Fig. 2).

Fig. 1.

Fig. 1.

Experimental overview. Six RTK-transfected cell lines were treated with growth factor (GF) and TNFα, individually or simultaneously. The untransfected parental cell line was also treated with TNFα, and all of the cell lines were mock treated (−). Response measurements were made by determining the percentages of proliferative and apoptotic cells after stimulation for 24 h. Signaling measurements were made by immunoblotting lysates of cells stimulated for 10 min or 16 h.

Fig. 2.

Fig. 2.

Growth factor- and TNFα-induced apoptosis. The cells were stimulated for 24 h as indicated and stained with annexin V-FITC and propidium iodide. The percentages of early apoptotic cells (annexin V-positive) and late apoptotic cells (annexin V- and propidium iodide-positive) were determined by flow cytometry. The error bars represent the S.E. of three biological replicates. GF, growth factor.

Several interesting observations arose from this experiment. First, in all six cell lines, the addition of growth factor to TNFα treatment had either no effect or led to increased levels of apoptosis. This contrasts with previously published studies in which both EGF and insulin were shown to antagonize TNFα-induced apoptosis in HT29 colon carcinoma cells (22), underscoring the importance of cellular context in determining how mixtures of opposing signals are interpreted. Second, TNFα induced apoptosis to different extents in the different cell lines, even without the addition of exogenous growth factor. This suggests that the six receptors exhibit some low level of basal activity, even in the absence of exogenously added ligand. Third, the combined action of growth factor and TNFα had quantitatively different effects depending on the identity of the growth factor. For example, in the EGFR and FGFR1 cell lines, EGF and aFGF alone induced similar amounts of apoptosis, and TNFα alone induced more apoptosis in the FGFR1 cell line than in the EGFR cell line. When cells were cotreated with growth factor and TNFα, however, EGF substantially increased TNFα-induced apoptosis in the EGFR-expressing cell line, but aFGF had little effect in the FGFR1-expressing cell line. Thus, at least in this cellular background, EGF and TNFα act synergistically to induce apoptosis, whereas aFGF and TNFα do not. Overall, the effects of growth factor plus TNFα on apoptosis were subadditive in some cell lines and superadditive in others, indicating that cross-talk occurs between these networks.

To measure proliferation in response to these treatments, bromodeoxyuridine (BrdU) was added to the tissue culture media for the last hour of the 24-h incubation. The cells were fixed and stained with an anti-BrdU antibody, and the percentage of proliferating cells was determined by flow cytometry (Fig. 3). Whereas different growth factors induced proliferation to different extents, TNFα did not affect proliferation, either alone or in combination with growth factor. The only exception was in the FGFR1-expressing cell line, where aFGF and TNFα each slightly decreased proliferation alone, but the combination treatment did not. In general, however, growth factors enhanced TNFα-induced apoptosis, but TNFα did not affect growth factor-induced proliferation, suggesting that cross-talk between these networks is unidirectional.

Fig. 3.

Fig. 3.

Growth factor- and TNFα-induced proliferation. The cells were stimulated for 24 h as indicated. BrdU was added to the culture media for the last hour of treatment. The percentage of BrdU-positive proliferating cells was determined by flow cytometry. The error bars represent the range of two biological replicates, except for MET samples, where the error bars represent the S.E. of four biological replicates. GF, growth factor.

Growth Factor and TNFα Signaling Networks Interconnect

Because extracellular cues are transmitted into phenotypic responses by intracellular signals, we reasoned that comparing intracellular signaling across the six cell lines and four stimulation conditions should help identify key nodes of pathway integration and the origin of the phenotypic differences induced by combinations of growth factors and TNFα. To determine the appropriate time points for measuring activation of intracellular signaling proteins, the PDGFRβ-expressing cell line was treated with saturating concentrations of TNFα and PDGF-BB simultaneously, and the cells were lysed at 11 time points over 48 h. This cell line was chosen because the combined treatment with growth factor and TNFα resulted in high levels of both apoptosis and proliferation (Figs. 2 and 3, respectively). Activation of a diverse set of signaling proteins was measured by quantitative immunoblotting, using pan-specific, phospho-specific, and cleavage-specific antibodies. Activation of most proteins was found to peak at either 10 min or 16 h, categorizing them as either early or late responders (Fig. 4).

Fig. 4.

Fig. 4.

Signaling time course in PDGFRβ-expressing cells. The cells were stimulated with saturating TNFα and PDGF-BB simultaneously and lysed at the indicated times. The lysates were immunoblotted with the indicated antibodies. Signal intensity is normalized to the maximum observed intensity for each antibody. For visualization purposes, the proteins were grouped as early responders (a) or late responders (b). The time points are equally spaced for visualization purposes.

Based on these results, all six RTK-expressing cell lines were mock treated or treated with growth factor and TNFα, individually or together, for 10 min or 16 h. Activation of a wide range of intracellular signaling proteins was then measured by quantitative immunoblotting (supplemental Tables 1 and 2). The proteins that we measured covered the canonical signaling pathways induced by growth factors and TNFα, as determined from our previous studies and from the literature. In total, 32 signals were measured after 10 min of stimulation, of which 19 responded to growth factors and/or TNFα. 33 signals were measured after 16 h of stimulation, of which 17 responded to growth factors and/or TNFα.

Some protein activation events, such as CrkL phosphorylation at 10 min, occurred only in response to growth factor treatment (Fig. 5a). Other events, such as NFκB p105 phosphorylation at 10 min, occurred only in response to TNFα treatment (Fig. 5b). Some proteins were activated by both treatments.

Fig. 5.

Fig. 5.

Growth factor- and TNFα-induced signaling. A subset of the entire signaling data set is shown. The cells were treated as indicated for 10 min or 16 h, and lysates were immunoblotted with the indicated antibodies. Signal intensity is normalized to the maximum observed intensity for each antibody. The error bars represent the range of two biological replicates, and representative immunoblots are shown below the bar graphs. a, CrkL phosphorylation at 10 min. b, NFκB phosphorylation at 10 min. c, RIP2 phosphorylation at 10 min. d, c-Jun phosphorylation at 16 h. e, p38 phosphorylation at 10 min. f, JNK phosphorylation at 10 min. g, total levels of p21 at 16 h. h, total levels of p53 at 16 h. GF, growth factor.

In several instances, growth factors and TNFα exhibited synergistic effects. For example, combined treatment induced more phosphorylation of RIP2 after 10 min than the sum of the individual treatments (Fig. 5c). The same was true for c-Jun phosphorylation after 16 h (Fig. 5d) and p38 and JNK phosphorylation at 10 min (Fig. 5, e and f). For some signals, such as the overall levels of p21 at 16 h, cross-talk occurred with some growth factors but not with others (Fig. 5g). In one case, growth factors and TNFα had opposing effects: growth factors decreased p53 levels at 16 h, whereas TNFα increased its levels, and the combination most often produced levels resembling growth factor treatment alone (Fig. 5h). Overall, however, in almost every case in which a protein was activated by both TNFα and growth factor individually, the combined treatment had a greater than additive effect.

Although it has previously been shown that TNFα can activate PI3K and Akt (27), this response was not observed in the current study, probably because this pathway already exhibits fairly high basal activity in these cells (supplemental Table 2). Additionally, it has been shown that Grb2, an upstream activator of the MAPK pathway, can directly interact with TNFR1 (28), but TNFα did not affect Ras/MAPK signaling in these cells.

The signaling and phenotypic outcomes observed in response to TNFα cannot be attributed to transactivation of the RTKs: TNFα treatment did not affect the levels of the RTKs or induce their phosphorylation (Fig. 6). Although no change in abundance of the six receptors was seen across any of the four treatments after 10 min, EGF and bone-derived neurotrophic factor (BDNF) did induce down-regulation of EGFR and NTRK2, respectively, by 16 h, as expected. In addition, the overall levels of insulin-like growth factor-1 receptor and PDGFRβ were consistently lower across all treatments at 16 h, possibly resulting from increased cell density at this time point.

Fig. 6.

Fig. 6.

Receptor expression and phosphorylation levels. Expression levels of the six RTKs after the indicated treatments were determined by immunoblotting with pan antibodies. Phosphorylation levels were determined by immunoblotting with phospho-specific antibodies. Signal intensity is normalized to the maximum observed intensity for each antibody, and the error bars represent the range of two biological replicates. The epitope for the MET pan antibody includes Y1234, which has been shown to be phosphorylated. Signal intensity was therefore affected by treatment conditions, and the antibody was determined not to be reliable for quantifying receptor expression level. GF, growth factor; N.D., not determined.

Activation of Network Integrators Correlates with Phenotypic Response

By placing the intracellular signaling proteins into a network diagram based on prior knowledge, system level properties become apparent (Fig. 7). As anticipated from the phenotypic results, cross-talk between the growth factor networks and the TNFα network does, indeed, appear to be unidirectional. Proteins that are commonly associated with RTK networks, such as Erk1/2 and Stat3, were not affected by TNFα treatment, whereas proteins considered part of the TNFR network, such as p38 and NFκB, were affected by growth factor treatment.

Fig. 7.

Fig. 7.

Signaling networks induced by growth factor and TNFα. Proteins were arranged into a network diagram based on known connectivity. a, all proteins investigated in this study. b, proteins investigated after 10 min of stimulation. c, proteins investigated after 16 h of stimulation. Faded proteins were not investigated at the indicated time point.

Most notably, RIP2, p38, and JNK were phosphorylated after 10 min in response to both TNFα and several of the growth factors (Fig. 7b). Based on this observation, we hypothesized that these nodes serve as key points of integration between the two networks. Consistent with this hypothesis, when we regressed each of the signals that we measured against phenotypic response, we found that, of all the signals, the phosphorylation levels of JNK and p38 at 10 min correlated best with the extent of apoptosis at 24 h (r = 0.90, p < 0.0001 and r = 0.90, p < 0.0001, respectively; Fig. 8, a and b). These nodes of signal integration may therefore be mechanistically involved in determining the extent of apoptosis in response to treatment with these conflicting signals. Although the number of proteins affected by both TNFα and growth factor treatment increased at later times (Fig. 7c), these late signals do not correlate well with apoptosis, with the exception of the caspases and poly(ADP-ribose) polymerase, which are themselves considered read-outs of apoptosis.

Fig. 8.

Fig. 8.

Correlations between signals and phenotype. a, JNK phosphorylation at 10 min is plotted against the percentage of early apoptotic (annexin V-positive) cells. b, p38 phosphorylation at 10 min is plotted against the percentage of late apoptotic (annexin V- and propidium iodide-positive) cells. c, cRaf phosphorylation at 16 h is plotted against the percentage of proliferating (BrdU-positive) cells. d, Erk phosphorylation at 16 h is plotted against the percentage of proliferating cells.

In contrast to apoptosis, no early signal correlated significantly (p < 0.05) with proliferation. Phosphorylation of cRaf and Erk at 16 h, however, correlated very well (r = 0.92, p < 0.0001 and r = 0.87, p < 0.0001; Fig. 8, c and d). Because these proteins are known to be integrally involved in RTK-induced mitogenic responses, this correlation emphasizes the central role of Erk in proliferation and further explains why TNFα does not affect proliferation.

To determine whether combinations of signals can better explain the observed phenotypic responses, we used PLSR to regress our responses, Y, against our measurements of signaling, X. PLSR reduces the dimensionality of X by decomposing it into a small number of orthogonal components that capture most of the covariance between X and Y. Each component is a linear combination of signals, weighted by how much they contribute to predicting response. To guard against overfitting and to assess the predictive value of the signaling measurements, we built our models using leave-one-out cross-validation: each model was trained using data from five cell lines and then asked to predict the phenotypic data for the sixth cell line from its signaling measurements. We found that two components were sufficient to capture ∼70% of the covariance with both proliferation and apoptosis (Fig. 9, a and b). To determine which signals were most predictive of either proliferation or apoptosis, we calculated their VIP as previously described (1). Consistent with our earlier results, the levels of p-cRaf and p-Erk at 16 h were most predictive of proliferation (Fig. 9a, bottom panel), whereas the levels of p-p38 and p-JNK at 10 min were most predictive of apoptosis (Fig. 9b, bottom panel). In general, late signaling events were most closely connected with proliferation and early signaling events with apoptosis. Together, these observations led us to posit that the decision to undergo apoptosis is made early and depends largely on the stress MAP kinases p38 and JNK, whereas the decision to proliferate is made later and depends on sustained Raf/Erk signaling.

Fig. 9.

Fig. 9.

PLSR models of proliferation and apoptosis. Top panels, correlation loading plot (first and second partial least squares components) of PLSR showing all 34 protein measurements and the cellular outputs with cell proliferation (a) or apoptosis (b). Bottom panels, bar graph showing the VIP with 95% confidence intervals for each signal in the PLSR model. The horizontal dotted line indicates the threshold value of 1.0. The top two signals (p-cRaf and p-Erk) and top three signals (p-JNK, p-p38, and p-NFκb) important for proliferation and apoptosis, respectively, are indicated.

Network Redundancy Preserves Response despite Pharmacological Intervention

To investigate the causal role of p38 and JNK in defining the extent of apoptosis in response to TNFα and growth factor treatment, we used small molecule inhibitors of p38 and JNK (SB202190 and SP600125, respectively) to modulate their kinase activities. EGFR-expressing cells were pretreated with SB202190 or SP600125 for 1 h prior to stimulation with EGF and TNFα, either alone or in combination. p38 and JNK activity were determined by measuring phosphorylation of their substrates, Hsp27 and c-Jun, respectively. Interestingly, we found that inhibiting p38 activity increased signaling through JNK, and inhibiting JNK activity increased signaling through p38 (Fig. 10a). The system thus appears to have built-in functional redundancy at these key points of network integration. This is consistent with previous observations that knocking down proteins in highly interconnected networks often leads to up-regulation in the activity of proteins in parallel pathways (29). Combination treatment with both inhibitors resulted in activation levels that were similar to those in DMSO-treated cells, preserving the overall apoptotic response to the external cues despite pharmacological intervention (Fig. 10b). This result underscores the need to build quantitative, mechanistic models of the integrated network that could better highlight how information flux is redirected as the system is perturbed.

Fig. 10.

Fig. 10.

Inhibition of JNK and p38 signaling. The cells were pretreated for 1 h with p38 inhibitor (SB202190, SB) or JNK inhibitor (SP600125, SP) alone or in combination or with DMSO alone and then stimulated as indicated. a, p38 and JNK activity was determined by measuring phosphorylation of their substrates, Hsp27 and c-Jun, respectively. Signal intensity is normalized to the maximum observed intensity for each antibody, and the error bars represent the range of two biological replicates. b, cells were stained with annexin V-FITC and propidium iodide, and the percentage of apoptotic cells was determined by flow cytometry. The error bars represent the range of two biological replicates. GF, growth factor.

Growth Factor-induced Enhancement of Apoptosis Is Mediated by cRaf

RIP2 lies upstream of p38 and JNK and was phosphorylated by both growth factors and TNFα. RIP2 and cRaf have previously been shown to coimmunoprecipitate when overexpressed (30), and we propose that growth factors may activate RIP2 through cRaf. RIP2 and cRaf phosphorylation are reasonably well correlated at 10 min for all samples (r = 0.81, p < 0.0001). When the treatment conditions are considered separately, RIP2 and cRaf phosphorylation are correlated for samples treated with growth factor alone (r = 0.82, p = 0.02) or with growth factor and TNFα (r = 0.88, p = 0.01) but not for mock treated samples (r = −0.03) or samples treated with TNFα alone (r = −0.1). To determine whether cRaf is responsible for growth factor-induced enhancement of TNFα-dependent apoptosis, we knocked down cRaf in the PDGFRβ-expressing cells. Cells were transfected with siRNA 48 h prior to cytokine treatment. cRaf protein levels in knockdown cells were ∼35–60% of control levels (data not shown). In two independent experiments, combined PDGF-BB and TNFα treatment in control cells increased apoptosis by ∼7–8% compared with TNFα alone. In cRaf knockdown cells, however, this difference was only ∼3% (Fig. 11). These results show that cRaf mediates growth factor-induced enhancement of apoptosis and are consistent with the hypothesis that cRaf and RIP2 constitute a key node of cross-talk between the RTK and TNFR signaling networks.

Fig. 11.

Fig. 11.

Effect of cRaf knockdown on apoptosis. PDGFRβ-expressing cells were transfected with cRaf siRNA or control siRNA 48 h prior to treatment with TNFα and PDGF-BB for 24 h. The cells were stained with annexin V-FITC and propidium iodide, and the percentage of positive cells was determined by flow cytometry. Two independent experiments were performed, and one representative experiment is shown. The error bars represent the range of replicate wells.

DISCUSSION

The results of this study reveal that rather than signaling in parallel, RTK networks and the TNFR network intersect. Cellular outcome is determined by how signals are integrated, rather than simply by the balance of the two conflicting stimuli. A similar result was seen in a previous study, in which RAW 264.7 macrophages were treated with 22 different ligands, alone and in pairwise combinations (31). A large number of nonadditive interactions were observed, suggesting that diverse external stimuli converge on intracellular signals to provide for context-dependent signaling.

Similarly, previous studies have shown that different epithelial cell types use different transducer molecules to activate a common set of effector molecules across different cell types. The various combinations of these effector molecules then generate diverse outcomes (32). This principle of common processing also appears to operate in HEK293 cells, where the activation of different RTK networks produces different cellular contexts. Rather than each RTK-TNFR supernetwork using distinct effectors, a small set of common effectors, most notably p38 and JNK, appear to determine cell fate.

Our results suggest that early p38 and JNK signaling promote apoptosis, which is consistent with the previous observation that early activity of the p38 substrate MK2 correlates with apoptosis in HT29 cells (22). Other reports, however, have shown that inhibiting early p38 and JNK signaling increases apoptosis (33) or that JNK is not involved in TNFα-induced apoptosis (34). Together, the results of the current study and the above cited previous studies show that a single protein may have different functions depending on the cellular context and the timing of its activation. Regardless of context, however, it is likely that the nodes of pathway cross-talk are invariant, because they are hard-wired into the integrated TNFR/RTK signaling network. Thus, although p38 and JNK elicit different outcomes in different cell types, they probably always function as common effectors that determine cellular response to these stimuli. Additionally, signal integration between these two networks appears to be unidirectional: RTK-mediated signals converge on TNFR-mediated signals to enhance apoptosis, but TNFR-mediated signals do not affect growth factor-induced proliferation. It remains to be determined whether this one-way flow of information is cell type-specific or whether it is in fact a fundamental feature of the RTK and TNFR signaling networks.

Although p38 and JNK act as common effectors in this integrated supernetwork, we propose that the key point of network integration is cRaf and RIP2, which lie upstream of these two proteins. RIP2 was phosphorylated by both growth factors and TNFα. It contains a serine/threonine kinase domain and a caspase activation and recruitment domain (35). It can be recruited to TNFR1 via TRAF proteins and can activate caspases and NFκB signaling (36, 37). Although RIP2 phosphorylation at 10 min did not correlate well with apoptosis, it did correlate with both p38 phosphorylation (r = 0.90, p < 0.0001) and JNK phosphorylation (r = 0.87, p < 0.0001), and RIP2 has previously been shown to lie upstream of these proteins mechanistically (3840). RIP2 has primarily been studied in the context of innate and adaptive immune responses, but lipopolysaccharide-induced activation of NFκB, p38, and JNK was shown to be reduced in macrophages from RIP2-deficient mice (38). RIP2 has also been shown to activate TAK1, a MAPK kinase kinase that lies upstream of p38 and JNK (40), and RIP2 activates p38 and JNK when overexpressed in HEK293 cells (39, 40). With respect to growth factor signaling, activation of RIP2 may be mediated by cRaf, because these two proteins have been shown to coimmunoprecipitate when overexpressed (30). Here we show that knocking down cRaf reduces the ability of PDGF-BB to enhance TNFα-induced apoptosis but does not affect apoptosis induced by TNFα alone. These results suggest that cRaf, possibly through RIP2, regulates growth factor-induced apoptosis. Activation of RIP2 by TNFα alone probably occurs through direct recruitment of RIP2 to TNFR.

Future studies investigating growth factor- and TNFα-induced changes in gene expression and the role of extracellular autocrine feedback loops in determining cell fate will further elucidate how these stimuli induce cellular responses and how cellular context can profoundly affect the outcome. Additionally, systematically inhibiting or knocking down individual network components, combined with computational analyses, may elucidate the topology of each of these growth factor-TNFα networks and enable identification of additional upstream molecules that initiate cross-talk. Cells are exposed to multiple stimuli under physiological and pathological conditions, and integration of these and other signaling networks is almost certainly a universal phenomenon. Understanding how networks integrate external signals at the molecular level will be essential for predicting the effects of therapeutic intervention in these networks and for identifying mechanistic biomarkers that are predictive of response.

Acknowledgments

We thank Taran Gujral for help with PLSR analysis and Jordan Krall for helpful discussions and critical reading of the manuscript.

G. M. is a founder, employee, and shareholder of Merrimack Pharmaceuticals; a founder, consultant, and shareholder of Makoto Life Sciences; and a scientific advisory board member of Aushon Biosystems.

Footnotes

* This work was supported by National Institutes of Health Grants R21 CA126720 and P50 GM068762. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Inline graphic This article contains supplemental material.

1 The abbreviations used are:

RTK
receptor tyrosine kinase
aFGF
acidic fibroblast growth factor
EGF
epidermal growth factor
EGFR
EGF receptor
IGF1
insulin-like growth factor 1
IGF1R
insulin-like growth factor 1 receptor
MET
hepatocyte growth factor receptor
HGF
hepatocyte growth factor
NTRK2
neurotrophic tyrosine kinase receptor type 2
BDNF
bone-derived neurotrophic factor
FGFR1
fibroblast growth factor receptor 1
NFκB
nuclear factor κB
PDGF-BB
platelet-derived growth factor-BB
PDGFRβ
platelet-derived growth factor receptor-β
TNFR
TNF receptor
JNK
c-Jun N-terminal kinase
MAPK
mitogen-activated protein kinase
DMEM
Dulbecco's modified Eagle's medium
siRNA
small interfering RNA
PLSR
partial least squares regression
VIP
variable importance in the projection
BrdU
bromodeoxyuridine.

REFERENCES

  • 1. Gordus A., Krall J. A., Beyer E. M., Kaushansky A., Wolf-Yadlin A., Sevecka M., Chang B. H., Rush J., MacBeath G. (2009) Linear combinations of docking affinities explain quantitative differences in RTK signaling. Mol. Syst. Biol. 5, 235. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2. Danial N. N., Korsmeyer S. J. (2004) Cell death: Critical control points. Cell 116, 205–219 [DOI] [PubMed] [Google Scholar]
  • 3. Aggarwal B. B. (2003) Signalling pathways of the TNF superfamily: A double-edged sword. Nat. Rev. Immunol. 3, 745–756 [DOI] [PubMed] [Google Scholar]
  • 4. Carswell E. A., Old L. J., Kassel R. L., Green S., Fiore N., Williamson B. (1975) An endotoxin-induced serum factor that causes necrosis of tumors. Proc. Natl. Acad. Sci. U.S.A. 72, 3666–3670 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5. Moore R. J., Owens D. M., Stamp G., Arnott C., Burke F., East N., Holdsworth H., Turner L., Rollins B., Pasparakis M., Kollias G., Balkwill F. (1999) Mice deficient in tumor necrosis factor-α are resistant to skin carcinogenesis. Nat. Med. 5, 828–831 [DOI] [PubMed] [Google Scholar]
  • 6. Spriggs D., Imamura K., Rodriguez C., Horiguchi J., Kufe D. W. (1987) Induction of tumor necrosis factor expression and resistance in a human breast tumor cell line. Proc. Natl. Acad. Sci. U.S.A. 84, 6563–6566 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7. Beissert S., Bergholz M., Waase I., Lepsien G., Schauer A., Pfizenmaier K., Krönke M. (1989) Regulation of tumor necrosis factor gene expression in colorectal adenocarcinoma: In vivo analysis by in situ hybridization. Proc. Natl. Acad. Sci. U.S.A. 86, 5064–5068 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8. Naylor M. S., Stamp G. W., Foulkes W. D., Eccles D., Balkwill F. R. (1993) Tumor necrosis factor and its receptors in human ovarian cancer. Potential role in disease progression. J. Clin. Invest. 91, 2194–2206 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9. Kulbe H., Thompson R., Wilson J. L., Robinson S., Hagemann T., Fatah R., Gould D., Ayhan A., Balkwill F. (2007) The inflammatory cytokine tumor necrosis factor-α generates an autocrine tumor-promoting network in epithelial ovarian cancer cells. Cancer Res. 67, 585–592 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10. Petersen S. L., Wang L., Yalcin-Chin A., Li L., Peyton M., Minna J., Harran P., Wang X. (2007) Autocrine TNFα signaling renders human cancer cells susceptible to Smac-mimetic-induced apoptosis. Cancer Cell 12, 445–456 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11. Brown E. R., Charles K. A., Hoare S. A., Rye R. L., Jodrell D. I., Aird R. E., Vora R., Prabhakar U., Nakada M., Corringham R. E., DeWitte M., Sturgeon C., Propper D., Balkwill F. R., Smyth J. F. (2008) A clinical study assessing the tolerability and biological effects of infliximab, a TNF-α inhibitor, in patients with advanced cancer. Ann. Oncol. 19, 1340–1346 [DOI] [PubMed] [Google Scholar]
  • 12. Harrison M. L., Obermueller E., Maisey N. R., Hoare S., Edmonds K., Li N. F., Chao D., Hall K., Lee C., Timotheadou E., Charles K., Ahern R., King D. M., Eisen T., Corringham R., DeWitte M., Balkwill F., Gore M. (2007) Tumor necrosis factor α as a new target for renal cell carcinoma: Two sequential phase II trials of infliximab at standard and high dose. J. Clin. Oncol. 25, 4542–4549 [DOI] [PubMed] [Google Scholar]
  • 13. Madhusudan S., Foster M., Muthuramalingam S. R., Braybrooke J. P., Wilner S., Kaur K., Han C., Hoare S., Balkwill F., Talbot D. C., Ganesan T. S., Harris A. L. (2004) A phase II study of etanercept (Enbrel), a tumor necrosis factor α inhibitor in patients with metastatic breast cancer. Clin. Cancer Res. 10, 6528–6534 [DOI] [PubMed] [Google Scholar]
  • 14. Madhusudan S., Muthuramalingam S. R., Braybrooke J. P., Wilner S., Kaur K., Han C., Hoare S., Balkwill F., Ganesan T. S. (2005) Study of etanercept, a tumor necrosis factor-α inhibitor, in recurrent ovarian cancer. J. Clin. Oncol. 23, 5950–5959 [DOI] [PubMed] [Google Scholar]
  • 15. Aguirre V., Uchida T., Yenush L., Davis R., White M. F. (2000) The c-Jun NH2-terminal kinase promotes insulin resistance during association with insulin receptor substrate-1 and phosphorylation of Ser307. J. Biol. Chem. 275, 9047–9054 [DOI] [PubMed] [Google Scholar]
  • 16. Cursi S., Rufini A., Stagni V., Condò I., Matafora V., Bachi A., Bonifazi A. P., Coppola L., Superti-Furga G., Testi R., Barilà D. (2006) Src kinase phosphorylates Caspase-8 on Tyr380: A novel mechanism of apoptosis suppression. EMBO J. 25, 1895–1905 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17. Daigle I., Yousefi S., Colonna M., Green D. R., Simon H. U. (2002) Death receptors bind SHP-1 and block cytokine-induced anti-apoptotic signaling in neutrophils. Nat. Med. 8, 61–67 [DOI] [PubMed] [Google Scholar]
  • 18. Habib A. A., Chatterjee S., Park S. K., Ratan R. R., Lefebvre S., Vartanian T. (2001) The epidermal growth factor receptor engages receptor interacting protein and nuclear factor-κB (NF-κB)-inducing kinase to activate NF-κB. J. Biol. Chem. 276, 8865–8874 [DOI] [PubMed] [Google Scholar]
  • 19. Kleber S., Sancho-Martinez I., Wiestler B., Beisel A., Gieffers C., Hill O., Thiemann M., Mueller W., Sykora J., Kuhn A., Schreglmann N., Letellier E., Zuliani C., Klussmann S., Teodorczyk M., Gröne H. J., Ganten T. M., Sültmann H., Tüttenberg J., von Deimling A., Regnier-Vigouroux A., Herold-Mende C., Martin-Villalba A. (2008) Yes and PI3K bind CD95 to signal invasion of glioblastoma. Cancer Cell 13, 235–248 [DOI] [PubMed] [Google Scholar]
  • 20. Lin Y., Yang Q., Wang X., Liu Z. G. (2006) The essential role of the death domain kinase receptor-interacting protein in insulin growth factor-I-induced c-Jun N-terminal kinase activation. J. Biol. Chem. 281, 23525–23532 [DOI] [PubMed] [Google Scholar]
  • 21. Senft J., Helfer B., Frisch S. M. (2007) Caspase-8 interacts with the p85 subunit of phosphatidylinositol 3-kinase to regulate cell adhesion and motility. Cancer Res. 67, 11505–11509 [DOI] [PubMed] [Google Scholar]
  • 22. Janes K. A., Albeck J. G., Gaudet S., Sorger P. K., Lauffenburger D. A., Yaffe M. B. (2005) A systems model of signaling identifies a molecular basis set for cytokine-induced apoptosis. Science 310, 1646–1653 [DOI] [PubMed] [Google Scholar]
  • 23. Gaudet S., Janes K. A., Albeck J. G., Pace E. A., Lauffenburger D. A., Sorger P. K. (2005) A compendium of signals and responses triggered by prodeath and prosurvival cytokines. Mol. Cell. Proteomics 4, 1569–1590 [DOI] [PubMed] [Google Scholar]
  • 24. Janes K. A., Kelly J. R., Gaudet S., Albeck J. G., Sorger P. K., Lauffenburger D. A. (2004) Cue-signal-response analysis of TNF-induced apoptosis by partial least squares regression of dynamic multivariate data. J. Comput. Biol. 11, 544–561 [DOI] [PubMed] [Google Scholar]
  • 25. Wolf-Yadlin A., Kumar N., Zhang Y., Hautaniemi S., Zaman M., Kim H. D., Grantcharova V., Lauffenburger D. A., White F. M. (2006) Effects of HER2 overexpression on cell signaling networks governing proliferation and migration. Mol. Syst. Biol. 2, 54. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26. (2005) User's Guide to SIMCA-P, SIMCA-P+, Umetrics AB, Inc. Umea, Sweden [Google Scholar]
  • 27. Ozes O. N., Mayo L. D., Gustin J. A., Pfeffer S. R., Pfeffer L. M., Donner D. B. (1999) NF-κB activation by tumour necrosis factor requires the Akt serine-threonine kinase. Nature 401, 82–85 [DOI] [PubMed] [Google Scholar]
  • 28. Hildt E., Oess S. (1999) Identification of Grb2 as a novel binding partner of tumor necrosis factor (TNF) receptor I. J. Exp. Med. 189, 1707–1714 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29. Sevecka M. (2008) Development of lysate microarray technology for quantitative analysis of EGFR signaling. Ph.D. dissertation, Harvard University [Google Scholar]
  • 30. Navas T. A., Baldwin D. T., Stewart T. A. (1999) RIP2 is a Raf1-activated mitogen-activated protein kinase kinase. J. Biol. Chem. 274, 33684–33690 [DOI] [PubMed] [Google Scholar]
  • 31. Natarajan M., Lin K. M., Hsueh R. C., Sternweis P. C., Ranganathan R. (2006) A global analysis of cross-talk in a mammalian cellular signalling network. Nat. Cell Biol. 8, 571–580 [DOI] [PubMed] [Google Scholar]
  • 32. Miller-Jensen K., Janes K. A., Brugge J. S., Lauffenburger D. A. (2007) Common effector processing mediates cell-specific responses to stimuli. Nature 448, 604–608 [DOI] [PubMed] [Google Scholar]
  • 33. Roulston A., Reinhard C., Amiri P., Williams L. T. (1998) Early activation of c-Jun N-terminal kinase and p38 kinase regulate cell survival in response to tumor necrosis factor α. J. Biol. Chem. 273, 10232–10239 [DOI] [PubMed] [Google Scholar]
  • 34. Liu Z. G., Hsu H., Goeddel D. V., Karin M. (1996) Dissection of TNF receptor 1 effector functions: JNK activation is not linked to apoptosis while NF-κB activation prevents cell death. Cell 87, 565–576 [DOI] [PubMed] [Google Scholar]
  • 35. Inohara N., del Peso L., Koseki T., Chen S., Núñez G. (1998) RICK, a novel protein kinase containing a caspase recruitment domain, interacts with CLARP and regulates CD95-mediated apoptosis. J. Biol. Chem. 273, 12296–12300 [DOI] [PubMed] [Google Scholar]
  • 36. McCarthy J. V., Ni J., Dixit V. M. (1998) RIP2 is a novel NF-κB-activating and cell death-inducing kinase. J. Biol. Chem. 273, 16968–16975 [DOI] [PubMed] [Google Scholar]
  • 37. Yin X., Krikorian P., Logan T., Csizmadia V. (2010) Induction of RIP-2 kinase by proinflammatory cytokines is mediated via NF-κB signaling pathways and involves a novel feed-forward regulatory mechanism. Mol. Cell. Biochem. 333, 251–259 [DOI] [PubMed] [Google Scholar]
  • 38. Kobayashi K., Inohara N., Hernandez L. D., Galán J. E., Núñez G., Janeway C. A., Medzhitov R., Flavell R. A. (2002) RICK/Rip2/CARDIAK mediates signalling for receptors of the innate and adaptive immune systems. Nature 416, 194–199 [DOI] [PubMed] [Google Scholar]
  • 39. Thome M., Hofmann K., Burns K., Martinon F., Bodmer J. L., Mattmann C., Tschopp J. (1998) Identification of CARDIAK, a RIP-like kinase that associates with caspase-1. Curr. Biol. 8, 885–888 [DOI] [PubMed] [Google Scholar]
  • 40. Windheim M., Lang C., Peggie M., Plater L. A., Cohen P. (2007) Molecular mechanisms involved in the regulation of cytokine production by muramyl dipeptide. Biochem. J. 404, 179–190 [DOI] [PMC free article] [PubMed] [Google Scholar]

Articles from Molecular & Cellular Proteomics : MCP are provided here courtesy of American Society for Biochemistry and Molecular Biology

RESOURCES