![[Feedback]](/icons/banner/feedbackACT.gif){kind=icon}
![[For Subscribers]](/icons/banner/subscriberACT.gif){kind=icon}
![[Archive]](/icons/banner/archiveACT.gif){kind=icon}
![[Search]](/icons/banner/searchACT.gif){kind=icon}
![[Contents]](/icons/banner/tocACT.gif){kind=icon}
|
|
|
|
|
||
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Departments of Chemical Engineering (N.K., D.A.L.) and Biological Engineering (R.A., H.-D.K., D.A.L.), and Center for Cancer Research (D.A.L.), Massachusetts Institute of Technology, Cambridge, Massachusetts
Received November 25, 2007; accepted March 17, 2008
 |
Abstract |
|---|
 Top Abstract Materials and MethodsResultsDiscussionReferences |
|---|
; Sevecka and Macbeath, 2006; Lazo et al., 2007). These unintended effects can arise from direct "off-target" interactions of the drug with additional proteins (Davies et al., 2000; Fabian et al., 2005) as well as by indirect "cross-talk" between the targeted pathway and other pathways in the signaling network regulating the cell behavioral response of interest (Natarajan et al., 2006). Although substantial progress is being made in improving kinase inhibitor specificities to reduce off-target effects, pathway cross-talk effects will—because of the fundamental interconnectedness of signaling networks—almost always be significant. An important implication is that the effects of kinase inhibitors on downstream cell phenotypic behaviors are not easily predictable based on the degree of their direct target activity reduction. Accordingly, predictive understanding of drug efficacy requires a more holistic, multipathway perspective on consequent network perturbation and its ensuing influence on cell behavioral responses. Moreover, the notion of multitarget drugs, such as sorafenib and sunitinib (Hiles and Kolesar, 2008; Zhu, 2008), is arising, in which multiple pathways will be concomitantly modulated by design; this exciting approach will obviously require a network-wide view.
|
; Kemp et al., 2007; Miller-Jensen et al., 2007). On this foundation, we test here whether unintended effects resulting from treatment of human mammary epithelial cells with one of two commonly studied compounds—LY294002 [a phosphatidylinositol 3-kinase (PI3K) inhibitor] and PD98059 [a mitogen-activated protein kinase kinase (MEK) inhibitor]—can be predicted in terms of an analogous multipathway computational model. In particular, we examine the effects of these compounds in an in vitro cell assay characterizing the role of overexpression of human epidermal growth factor receptor 2 (HER2) in migration of mammary epithelial cells.
Increased migration, and particularly increased directional persistence of migration, is a hallmark of invasive breast cancer cells (Friedl et al., 1995; Deisboeck et al., 2005). HER2, a member of the ErbB or HER family of receptors, is overexpressed in 
30% of breast cancers and correlates with increased metastasis (Moasser, 2007), and enhanced motility and invasiveness of HER2-overexpressing breast carcinoma cells has been demonstrated (Spencer et al., 2000). Our laboratory has recently shown that effects of HER2 overexpression on cell migration transpire via enhancement of directional persistence (Kumar et al., 2006b). Induction of signaling in HER2 systems occurs through homodimerization or via ligand-driven heterodimerization with other HER family receptors. Epidermal growth factor (EGF) and heregulin (HRG), two HER family ligands implicated in tumor progression, bind HER1 (also called EGFR) and HER3, respectively, to induce the activation of HER2 through heterodimerization (Harari and Yarden, 2000). EGFR and HER3 can also form homodimers as well as heterodimers with each other upon ligand binding, although HER3 homo-dimers are expected to be inactive (Engelman and Cantley, 2006).
In this study, we measured inhibitor efficacy against persistent migration in the context of serum-free, EGF, or HRG treatment for normal (parental line) and HER2-overexpressing (24H subline) human mammary epithelial cells (hMECs). To quantify pathway cross-talk effects of the inhibitors and gain a further understanding of signaling giving rise to directional persistence, we measured phosphorylation of central actors in three key kinase pathways downstream of ErbB receptor activation: Akt (serine 473), Erk (threonine 202/tyrosine 204), and p38 (threonine 180/tyrosine 182). We also measured EGFR phosphorylation (tyrosine 1173 and tyrosine 1068) as a partial indicator of initiating receptor tyrosine kinase activity induced by ligand stimulation (Harari and Yarden, 2000). These kinases have been shown to regulate cell migration in breast cancer epithelial cells (Ho et al., 2001; Arboleda et al., 2003; Huang et al., 2004; Shin et al., 2005; Hirsch et al., 2006), have been implicated in the control of directional migration (Merlot and Firtel, 2003; Sawyer et al., 2003), and have been found for many cell behavioral functions to be highly informative integrators of multiple signaling pathways (Oda et al., 2005). Figure 1 offers a schematic illustration of these pathways, showing at least some appreciated loci of cross-talk along with downstream effectors of migration behavior. LY294002 and PD98059 each target one of the measured kinase pathways, do not directly interact with any of the other pathway signals measured here, but do have other off-target binding activities (Davies et al., 2000; Fabian et al., 2005; Bain et al., 2007). However, the presence of cross-talk leading to network modulation across their respective pathways has been indicated (Zimmermann and Moelling, 1999; Chaudhary et al., 2000; Mograbi et al., 2001; Moelling et al., 2002; Hausenloy et al., 2004) and might be further exacerbated by off-target activities of these fairly nonselective inhibitors as well as by convoluting feedback loops within the pathways (Asthagiri and Lauffenburger, 2001; Birtwistle et al., 2007).
We found that no individual pathway signal is sufficient to predict drug efficacy on cell migration behavior from its target reduction—but we find that a partial least-squares regression model (Janes et al., 2005; Kemp et al., 2007; Miller-Jensen et al., 2007) that quantitatively integrates these three kinase pathway signals is successful in predicting the LY29004 and PD98059 effects despite their complex network modulations. We conclude that our multivariate computational modeling offers a novel and useful method for understanding how kinase inhibitors exhibiting direct and/or indirect effects on nontargeted signaling pathways influence downstream cell phenotypic behaviors.
|
Materials and Methods |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
; Wolf-Yadlin et al., 2006). Cells were stimulated with either 100 ng/ml EGF or 80 ng/ml HRG-β1 (Sigma, Inc., St. Louis, MO). For inhibitor experiments, cells were pretreated with either 20 µM LY294002 (Calbiochem, Inc., San Diego, CA) for 1 h, 25 µM PD98059 (Calbiochem, Inc.) for 1 h, 20 µM gefitinib (WuXi Pharmatech Co., Ltd., Shanghai, China) for 20 min, or 200 nM wortmannin (Calbiochem, Inc.) for 20 min.
Immunocytochemistry for Signal Profiling. Cells were seeded at 50,000 cells/well in a 96-well plate (Nalge Nunc International, Rochester, NY) in full DFCI-1 media. After 4 to 6 h, full media were replaced with serum-free media. After 12 to 16 h of serum starvation, cells were preincubated with inhibitor or treated with EGF, HRG, or fresh serum-free media. A wound of width 650 µm was then scraped in each well. Cells were stimulated for 0 (serum-free), 5, 15, 30, 60, and 90 min, with three or four wells used as biological replicates for each condition. After desired time had elapsed, cells were fixed, permeabilized, and stained as described in Supplemental Methods. The phospho-p44/42 MAPK (Erk, Thr202/Tyr204) and phospho-Akt (Ser473) rabbit monoclonal antibodies were purchased from Cell Signaling, Inc. (Danvers, MA) The phospho-p38 MAPK (pT180/pY182), phospho-EGFR (pY1068), and EGFR (pY1173) rabbit monoclonal antibodies were purchased from Epitomics, Inc (Burlingame, CA). Plates were imaged using an Odyssey instrument (Li-Cor, Lincoln, NE), and the integrated intensity was normalized to the loading control and then to parental serum-free intensity run on the same plate (see Supplemental Methods). All signaling data are plotted ± S.E.M., with n = 3 or 4.
Migration Assay. The migration assay was performed as described previously (Kumar et al., 2006b). In brief, cells labeled with 5-chloromethylfluorescein diacetate (Invitrogen, Carlsbad, CA) were diluted 1:20 with unlabeled cells and plated as described above, treated with ligand and inhibitor as described above, and wounded; cell movement was tracked for 12 h at 37°C and 5% CO2 using an automated epifluorescent microscope. Individual cell trajectories in a monolayer were analyzed using Imaris (Bitplane, Inc., Zurich, Switzerland), and cell trajectories were then fit to the persistent random walk equation to derive cell persistence. All data were normalized by the persistence of 24H cells stimulated with 100 ng/ml EGF on the plate, to account for plate-to-plate variability. For each condition, n > 300 and all data are reported as ± S.E.M.
Multilinear Modeling Using Partial Least-Squares Regression (PLSR). We regressed the signaling matrix X onto the persistence matrix P using partial least-squares regression (PLSR) in SIMCA-P 11.0 (Umetrics) as described in Supplemental Methods and elsewhere (Kumar et al., 2006a; Wolf-Yadlin et al., 2006). P is a column vector (Mx1) of persistence values corresponding to M cellular conditions, and X is the signaling matrix (MxN) in which each column represents a signaling metric. Both the parental and 24H cell data were normalized to the parental serum-free condition data obtained from the same plate after averaging biological replicates. The quantitative metrics used in the model for each phosphorylation site corresponded to these normalized levels of phosphorylation at each time point as well as the integral of phosphorylation to characterize "net" phosphorylation over 90 min. Coefficients for each metric were extracted from SIMCA-P as previously (Kumar et al., 2006a). All data were mean-centered and scaled to unit variance before analysis.
|
Results |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
; Sawyer et al., 2003). Thus, we set out to test whether this individual kinase signaling pathway might offer sufficient capacity for prediction of cell persistence. Our initial focus in this study was the quantitative relationship between Akt phosphorylation and HER2-mediated enhancement of mammary epithelial cell migration. As with our previous findings (Kumar et al., 2006b), HER2-overexpressing hMECs (24H cells) exhibited higher levels of directional persistence than those with low HER2 expression (parental cells) in the presence or absence of EGF or HRG (Figs. 2, A and B). Measurement of Akt phosphorylation under these same conditions revealed increased phosphorylation in response to HER2 overexpression, consistent with a putative role for Akt in persistent movement (Fig. 2C). Inhibition of Akt by treatment with LY294002 decreased persistence in 24H cells treated with HRG, further corroborating the role of the PI3K/Akt pathway in directional persistence (Figs. 2D,E). Inhibition with LY294002 in 24H cells treated with EGF, however, had no effect on directional persistence (Fig. 2F), despite the equivalent inhibition of Akt phosphorylation under both HRG and EGF treatments (Fig. 2D). Thus, the effect of LY294002 treatment on persistence could not be predicted solely by its effect on Akt phosphorylation, suggesting that assessing its effects on other pathways might be required to understand efficacy results.
|
|
Finally, p38 phosphorylation was higher in 24H cells but remained relatively constant in response to both HRG and EGF, although we observed a slight peak at 15 min under all conditions (only the 15-min HRG-stimulated peaks are significantly greater than the 0-min phosphorylation levels at p 
0.05, Fig. 3C).
Thus, the time-courses of measured phosphorylation levels indicate that upon ligand stimulation multiple signaling pathways in the network downstream of ErbB receptors respond to various levels and with diverse dynamics distinctly dependent on ligand and HER2 level. We hypothesize that migration persistence measured in Fig. 2, A and B, are integrated effects of these underlying signals.
Cell Migration Behavior Exhibited Ligand- and HER2-Dependent Inhibition Effects. To elucidate relationships between kinase phosphorylation and migration persistence, we measured cell movement in response to inhibition by the various compounds. As illustrated clearly in Fig. 2, E and F, treatment with LY294002 resulted in reduction of the migration of HER2-overexpressing 24H cells in the context of HRG stimulation but not EGF stimulation, even though Akt phosphorylation was diminished to comparable degrees for both ligand stimuli. This inhibitor similarly reduced migration of parental cells under HRG but not EGF induction (Fig. 4). Moreover, abrogation of Akt phosphorylation was found to be associated with diminished migration under serum-free conditions (Fig. 4), for which this functional behavior arises mainly from activation of EGFR by autocrine EGF family ligands (Wolf-Yadlin et al., 2006).
|
In analogous but not identical fashion, treatment with the MEK-Erk pathway inhibitor PD98059 reduced migration in a ligand- and HER2-dependent manner (Fig. 4). From our signaling data in Fig. 3, we hypothesized that PD98059 treatment would inhibit migration in response to EGF but not HRG. We were surprised to find that the inhibitor decreased levels of persistence in parental cells treated with EGF as well as HRG (Fig. 4). PD98059 reduced persistence in EGF-treated parental cells with greater efficacy than in 24H cells (50% versus 25% inhibition for parental and 24H cells, respectively), suggesting that HER2-overexpressing cells are less reliant on Erk for directional persistence. PD98059 treatment of HRG-stimulated 24H cells did not inhibit persistence (Fig. 4), further corroborating a diminished importance of Erk for 24H directional persistence. Thus, the varying efficacy of PD98059 treatment across different HER2 expression levels and ligand treatments indicates an important role for off-target and/or cross-talk effects in the governance of persistence.
Treatment with the EGFR kinase inhibitor gefitinib led to the greatest decrease in persistence across cell type and ligand treatment, indicating potential merit of further study into the potential use of gefitinib as an antimetastatic treatment in HER2-overexpressing tumors. Inhibition with gefitinib under EGF-stimulating conditions reduced persistence by 70 to 90%. It is noteworthy that gefitinib decreased persistence significantly more in 24H cells compared with parental cells (Fig. 4). To summarize, increasing HER2 expression in the presence of EGF and gefitinib lowered persistence, and HRG-treated 24H cells treated with gefitinib had persistence values less than those observed in parental cells without inhibitor.
Quantification of Phosphorylation in Response to Treatment with LY294002, PD98059, and Gefitinib Reveals Pathway Cross-Talk Effects. To better understand the above patterns of inhibitor efficacy, we measured multipathway effects by quantifying the phosphorylation of the previously mentioned kinases under all inhibitor conditions. Treatment with LY294002 greatly reduced Akt phosphorylation, although small differences between 24H and parental levels remained (Fig. 5). It is noteworthy that this PI3K inhibitor also abrogated Erk phosphorylation in response to EGF treatment; both negative and positive cross-talk from the PI3K/Akt pathway to the Erk pathway have been reported in previous literature, depending on cell type and ligand conditions (see Fig. 1) (Zimmermann and Moelling, 1999; Chaudhary et al., 2000; Mograbi et al., 2001; Moelling et al., 2002; Hausenloy et al., 2004). Replication of this result using wortmannin further corroborated a PI3K-related inhibition of Erk (Supplemental Fig. 2). Treatment with LY294002 also increased EGF-stimulated phosphorylation at both EGFR Tyr1068 and Tyr1173 (Fig. 5). This effect was not due to decreased serum-free phosphorylation levels, because serum-free levels of phosphorylated EGFR were the same in the presence or absence of inhibitor (Supplemental Fig. 3). This observation could suggest potential regulation of EGFR by the PI3K pathway; this kind of relationship has been previously reported for wortmannin, with alterations of endocytic trafficking indicated as an underlying mechanism (Bolander, 1998), although off-target effects may be involved (Chen and Wang, 2001).
|
Gefitinib differentially reduced EGFR Tyr1068 phosphorylation depending on HER2 expression levels; EGF-stimulated 24H phosphorylation was similar in the presence or absence of gefitinib, whereas parental phosphorylation was significantly decreased in response to gefitinib treatment (Fig. 5). Tyr1173 phosphorylation was substantially reduced in both cell lines (Fig. 5). Erk and Akt phosphorylation were eliminated by gefitinib treatment, including HRG-stimulated Akt phosphorylation in both parental and 24H cells (Fig. 5). The initially surprising ablation of HRG-stimulated Akt phosphorylation by gefitinib is consistent with its documented ability to trap HER2 and HER3 receptors in an inactive form (Anido et al., 2003). p38 phosphorylation levels equalized between the two cell lines over 90 min under gefitinib treatment, as observed under PD98059 treatment as well, suggesting HER2-dependent differences in off-target and cross-talk effects for both inhibitors.
Taken together, these results show that kinase inhibitor treatment leads to modulation of multiple pathways and that the dependence of the downstream cell migration phenotypic response on inhibition of the targeted pathway is not straightforward. Our goal in this work is not to elucidate the probably numerous molecular mechanisms by which the pathway cross-talk occurs, but instead to provide a computational modeling approach by which the combined effects of multiple pathways on downstream phenotypic behavior can be comprehended.
Multivariate Computational Model Enables Prediction of How Multiple Pathways Are Integrated to Influence Cell Migration. To quantify the contribution of each kinase to directional persistence, we constructed a PLSR model that linearly related our signal dataset to our migration data-set, following established methods (see Janes and Yaffe, 2006). In essence, weighted linear combinations of the most informative subsets of signal measurements are determined from a matrix algebraic equation relating the signal data across the full landscape of treatment conditions to the corresponding migration data across the same conditions; thus, this technique belongs to a class termed "data-driven modeling" approaches (Janes and Lauffenburger, 2006). We found that the resulting multivariate model (the "Full Model"; see Materials and Methods for computational details) accurately recapitulated measured migration persistence values (R2 = 0.89, r = 0.94, Fig. 6A). The high goodness-of-fit and correlation values helped validate the application of the multilinear modeling approach, consistent with previous successes in other cell systems (Janes et al., 2005; Kemp et al., 2007; Miller-Jensen et al., 2007).
|
A PLSR model constructed only from data obtained in the absence of inhibitor predicted a priori values of persistence in the presence of LY294002 and PD98059 that positively correlated with measured values of persistence (reduced model, r = 0.93; Fig. 6B). The reduced model accurately predicted the previously confounding effects of LY294002 on persistence under all ligand treatment conditions (Fig. 6D). Increased EGFR signaling in the case of EGF treatment under LY294002 compensated for the decrease in Akt signaling, explaining the multivariate model's accurate prediction of inhibitor nonefficacy and underscoring the role of indirect pathway cross-talk. In contrast, the model failed to accurately predict persistence in the presence of gefitinib (r = -0.36, Fig. 6C). Plausible explanations for this contrary result might lie in the fact that this inhibitor operates centrally on a kinase at the receptor level instead of downstream at the highly integrative "bow tie" of the signaling network (Oda et al., 2005) (see Fig. 1) and/or in substantial off-target effects on pathways besides those we measure here (Brehmer et al., 2005).
Inhibitor efficacy in 24H cells was accurately captured by the full multilinear model under both EGF and HRG stimulation (Figs. 6, E and F). In the case of HRG stimulation, the model accurately captured the ineffectiveness of treatment with PD98059, and the increased efficacy of gefitinib over LY294002 (Fig. 6E). It is noteworthy that the model not only captured trends but also accurately recapitulated absolute levels of persistence. In the case of EGF stimulation, the model captures the trend of increasing efficacy from LY294002 to PD98059 to gefitinib (Fig. 6F). In the case of gefitinib, the model failed to capture the absolute magnitude of migration persistence, presumably for the same reasons noted above. Nevertheless, the model, by accurately recapitulating measured persistence levels, defines a set of signal-integration rules by means of which the multipathway data effectively predict modulation of migration behavior.
The regression coefficients associated with each phosphorylation metric reveal the metric's contribution to persistence. Coefficient values, plotted in Fig. 6G, indicated that Akt most positively correlated with persistence. However, EGFR Tyr1173 and early-phase Tyr1068 data also strongly correlate with persistence, indicating that an increase in phosphorylation at those sites helps to keep persistence high even in the absence of significant Akt phosphorylation. Erk plays a positive but limited role in stimulating persistence, consistent with the current view that Akt is more important for directional migration. p38 negatively correlates with migration in our context, which contrasts to results obtained in MCF10A breast epithelial cells (Shin et al., 2005). We do not observe acute p38 stimulation under any conditions, however, so our conclusions may not extend accurately to systems where p38 is transiently activated to a more dramatic degree.
Overall, we have found that a partial least-squares regression modeling framework based on multipathway integration yields a substantial advance in successfully accounting for kinase inhibitor effects on downstream cell phenotypic response behavior.
|
Discussion |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
; Kemp et al., 2007; Miller-Jensen et al., 2007), we reasoned that a similar approach could be usefully applied to this kinase inhibitor efficacy problem. As a particular manifestation of this general problem, in this new study, we tested whether we could predict efficacy of two fairly nonspecific kinase inhibitors, one directed against PI3K (LY294002) and the other directed against MEK (PD98059), in decreasing persistent migration of normal and HER2-overexpressing hMECs across a landscape of serum-free, EGF, or HRG stimulatory treatments. Our computational model was based on measurements of phosphoprotein levels in three key migration-related pathways downstream of EGF receptor family activation—Akt (serine 473), Erk (threonine 202/tyrosine 204), and p38 (threonine 180/tyrosine 182)—along with phosphorylation of EGF receptor (tyrosine 1173 and tyrosine 1068). For both of these inhibitors, we demonstrated that the decreases in migration they produced could not be predicted simply from their corresponding reduction of their targeted kinase, but that a PLSR model constructed on the five phosphopeptide levels could successfully predict the migration responses in a priori manner (see Fig. 6). We conclude that our multivariate computational modeling provides a framework for comprehending how downstream cell phenotypic responses are integratively influenced by the modulation of the signaling network caused by the combination of direct and/or indirect effects of kinase inhibitors on nontargeted pathways.
It is noteworthy that the PLSR model predictions for the EGFR kinase inhibitor gefitinib substantially under-predicted the diminution of cell migration by this compound (Fig. 6C); i.e., gefitinib treatment produced a much stronger reduction of migration persistence than was expected from the associated changes measured in the Akt, Erk, and p38 kinase signaling pathways. Our best explanation for this disparity relative to the highly successful results for the Akt and Erk pathway inhibitor model predictions are that inhibiting the upstream receptor kinase activity leads to significant modulation of additional pathways outside the realm of what is reflected in the Akt, Erk, and p38 signals. We have argued elsewhere (Kumar et al., 2006a; Miller-Jensen et al., 2007) that kinase pathway signals in the "bow-tie" of intracellular regulatory networks possess what can be viewed as "surrogate information," because they can integratively represent activities distributed more broadly across other pathways as a result of the the substantial degree of interpathway cross-talk (Oda et al., 2005). In contrast, more distal upstream measurements, such as at a receptor level, are much less likely to incorporate this kind of "surrogate information" arising from multipathway cross-talk integration.
It might be surprising that effects of kinase inhibitory compounds on downstream cell phenotypic functions can be effectively predicted using measurements from only 3 to 5 phosphoprotein signals [Akt, Erk, p38, EGFR (2 sites)] as we have found here. An immediate basis may reside in appreciating that our predictions were tested across a fairly limited experimental landscape: two related cell types (normal and HER2-overexpressing hMECs) and three environmental cue conditions [serum-free media (in which ErbB system activation arises mainly from autocrine ligands), EGF treatment (binding to EGFR, leading to EGFR-EGFR homodimers, EGFR-HER2 heterodimers, and possibly EGFR-HER3 heterodimers), and HRG treatment (binding to HER3, leading to HER3-HER3 homodimers, HER3-HER2 heterodimers, and perhaps HER3-EGFR heterodimers)]. Nevertheless, this landscape represents a more stringent test than is generally the case, for signal inhibition studies most typically consider only one environmental cue condition (whether in high-throughput assay format or not). Contemplating this situation more broadly, it must be recalled that the standard approach for attempting to predict drug efficacy relies on single biomarkers, which for the most part are inadequate (e.g., HER2 expression levels for trastuzumab (Herceptin; Kroese et al., 2007). Therefore, our PLSR model that quantitatively combines five measurements (in what might be considered a "model-based biomarker set") can be seen to offer increased power by taking advantage of an increased number of variables. Moreover, we have previously demonstrated that a relatively small fraction of available network variables can adequately represent the information content of the network more broadly, as a result of the surrogate information about other pathway activities that is reflected in the measured set because of the high degree of pathway cross-talk (Kumar et al., 2006a). Finally, we note that although genome-wide transcriptomics permits inclusion of hundreds to thousands of prospective variables for analogous models, recent study has found that a two-gene model is successfully diagnostic for distinctive diagnosis between two challengingly similar mesenchymal tumors (Price et al., 2007). Hence, a phosphoprotein-based model incorporating three to five signal measurements should not be considered an unreasonable success.
The signaling changes resulting from inhibitor treatments revealed modulations in the pathway activities beyond the targeted kinase pathway, consistent with either direct off-target or indirect cross-talk effects. For instance, treatment with LY294002 and Wortmannin diminished phosphorylation of Erk (Fig. 5, Supplemental Fig. 2). Previous studies documenting inhibition of Raf by Akt have proffered the hypothesis that PI3K inhibition should increase Erk phosphorylation (Wennström and Downward, 1999; Zimmermann and Moelling, 1999). Our data here show that PI3K inhibition can result in strong ablation of Erk phosphorylation, possibly through a MEK-independent pathway, because treatment with LY294002 or PD98059 reduces Erk phosphorylation in a similar manner. As demonstrated in a recent study of signaling network responses across different cell types, it is quite likely that multipathway cross-talk is cell-type dependent and accordingly difficult to anticipate in absence of dedicated experimental measurements (Miller-Jensen et al., 2007).
The full PLSR model suggested that Akt phosphorylation correlates positively with persistent migration (Fig. 6G). An explicit role for Akt, other than as a surrogate for upstream PI3K activity in breast cancer cell migration is not clearly established. Recent reports indicate that Akt may act as either an inhibitor or promoter of metastasis (Arboleda et al., 2003; Yoeli-Lerner et al., 2005), but our results indicate that Akt inhibition via LY294002 treatment can lead to an effective decrease of HER2-mediated cell migration in the absence of off-target effects. Because both HRG and EGF-family ligands have been found to circulate in breast cancer tumors (Normanno et al., 2006), we speculate that inhibition of Akt with a PI3K inhibitor similar to LY294002 might be effective against cell migration in tumors with high levels of HRG but not necessarily in those with high EGF-family ligand expression, as a result of the LY294002 treatment consequence of increasing EGFR phosphorylation. Our data also contrast with an earlier report suggesting that PI3K does not play a role in directed migration in the absence of chemical attractant gradients (Pankov et al., 2005) but cell-type or substratum differences could underlie this disparity.
The likelihood of indirect, network cross-talk-mediated off-target effects raises the consideration that this kind of off-target effect cannot be engineered away through the creation of more target-specific drugs, and thus must motivate broader strategies including computational modeling approaches such as those we have described here. In fact, prospects are now being raised regarding the potential advantage for promiscuous, multitarget compounds in explicitly acting across multiple pathways intentionally (Zhang and Meier, 2006; Cascone et al., 2007). If this promising avenue is followed, our quantitative, multipathway integration framework will be absolutely required to effectively predict resulting effects on cell phenotypic behaviors.
|
Acknowledgements |
|---|
|
Footnotes |
|---|
ABBREVIATIONS: LY294002, 2-(4-morpholinyl)-8-phenyl-4H-1-benzopyran-4-one; PI3K, phosphatidylinositol 3-kinase; PD98059, 2'-amino-3'-methoxyflavone; MEK, mitogen-activated protein kinase kinase; EGF, epidermal growth factor; HRG, heregulin; EGFR, epidermal growth factor receptor; HER, human epidermal growth factor receptor; hMEC, human mammary epithelial cell; Akt, protein kinase B; Erk, extracellular signal-regulated kinase; MAPK, mitogen-activated protein kinase; PLSR, partial least-squares regression.
The online version of this article (available at http://molpharm.aspetjournals.org) contains supplemental material. 
Address correspondence to: Douglas A. Lauffenburger, Building 56, Room 341, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge MA 02139. E-mail: lauffen{at}mit.edu.
|
References |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
Arboleda MJ, Lyons JF, Kabbinavar FF, Bray MR, Snow BE, Ayala R, Danino M, Karlan BY, and Slamon DJ (2003) Overexpression of AKT2/protein kinase B-beta leads to up-regulation of beta1 integrins, increased invasion, and metastasis of human breast and ovarian cancer cells. Cancer Res 63: 196-206.
Asthagiri AR and Lauffenburger DA (2001) A computational study of feedback effects on signal dynamics in a mitogen-activated protein kinase (MAPK) pathway model. Biotechnol Progr 17: 227-239.[CrossRef][Medline]
Bain J, Plater L, Elliott M, Shpiro N, Hastie CJ, McLauchlan H, Klevernic I, Arthur JSC, Alessi DR, and Cohen P (2007) The selective of protein kinase inhibitors: a further update. Biochem J 408: 297-315.[Medline]
Birtwistle MR, Hatakeyama M, Yumoto N, Ogunnaike BA, Hoek JB, and Kholodenko BN (2007) Ligand-dependent responses of the ErbB signaling network: experimental and modeling analyses. Mol Syst Biol 3: 144.[Medline]
Bolander FF Jr (1998) Transduction pathways involved in rapid hormone regulation in the mammary epithelium. Am J Physiol 275: E553-E557.[Medline]
Brehmer D, Greff Z, Godl K, Blencke S, Kurtenbach A, Weber M, Muller B, Cotton M, Keri G, Wissing J, et al. (2005) Cellular targets of gefitinib. Cancer Res 65: 379-382.
Cascone T, Gridelli C, and Ciariello F (2007) Combined targeted therapies in non-small lung cancer: a winner strategy? Curr Opin Oncol 19: 98-102.[Medline]
Chaudhary A, King WG, Mattaliano MD, Frost JA, Diaz B, Morrison DK, Cobb MH, Marshall MS, and Brugge JS (2000) PI3K regulates Raf1 through Pak phosphorylation of serine 338. Curr Biol 10: 551-554.[CrossRef][Medline]
Chen X and Wang Z (2001) Regulation of EGFR endocytosis by Wortmannin through activation of Rab5 rather than inhibition by PI3K. EMBO Rep 2: 842-849.[CrossRef][Medline]
Davies SP, Reddy H, Caivano M, and Cohen P (2000) Specificity and mechanism of action of some commonly used protein kinase inhibitors. Biochem J 351: 95-105.[CrossRef][Medline]
Deisboeck TS, Demuth T, and Mansury Y (2005) Correlating velocity patterns with spatial dynamics in glioma cell migration. Acta Biotheor 53: 181-190.[CrossRef][Medline]
Engelman JA and Cantley LC (2006) The role of ErbB family members in non-small cell lung cancers sensitive to epidermal growth factor receptor kinase inhibitors. Clin Cancer Res 12: 4372s-4376s.
Fabian MA, Biggs WH 3rd, Treiber DK, Atteridge CE, Azimioara MD, Benedetti MG, Carter TA, Ciceri P, Edeen PT, Floyd M, et al. (2005) A small molecule-kinase interaction map for clinical kinase inhibitors. Nat Biotechnol 23: 329-336.[CrossRef][Medline]
Friedl P, Noble PB, Walton PA, Laird DW, Chauvin PJ, Tabah RJ, Black M, and Zanker KS (1995) Migration of coordinated cell clusters in mesenchymal and epithelial cancer explants in vitro. Cancer Res 55: 4557-4560.
Harari D and Yarden Y (2000) Molecular mechanisms underlying ErbB2/HER2 action in breast cancer. Oncogene 19: 6102-6114.[CrossRef][Medline]
Hausenloy DJ, Mocanu MM, and Yellon DM (2004) Cross-talk between the survival kinases during early reperfusion: its contribution to ischemic preconditioning. Cardiovasc Res 63: 305-312.
Hiles JJ and Kolesar JM (2008) Role of sunitinib and sorafenib in the treatment of metastatic renal cell carcinoma. Am J Health Syst Pharm 65: 123-131.
Hirsch DS, Shen Y, and Wu WJ (2006) Growth and motility inhibition of breast cancer cells by epidermal growth factor receptor degradation is correlated with inactivation of Cdc42. Cancer Res 66: 3523-3530.
Ho W, Uniyal S, Meakin SO, Morris VL, and Chan BM (2001) A differential role of extracellular signal-regulated kinase in stimulated PC12 pheochromocytoma cell movement. Exp Cell Res 263: 254-264.[CrossRef][Medline]
Huang C, Jacobson K, and Schaller MD (2004) MAP kinases and cell migration. J Cell Sci 117: 4619-4628.
Janes KA, Albeck JG, Gaudet S, Sorger PK, Lauffenburger DA, and Yaffe MB (2005) A systems model of signaling identifies a molecular basis set for cytokine-induced apoptosis. Science 310: 1646-1653.
Janes KA and Lauffenburger DA (2006) A biological approach to computational models of proteomic networks. Curr Opin Chem Biol 10: 73-80.[CrossRef][Medline]
Janes KA & Yaffe MB (2006) Data-driven modeling of signal transduction networks. Nat Rev Mol Cell Biol 7: 820-828.[CrossRef][Medline]
Kemp ML, Wille L, Lewis CL, Nicholson LB, and Lauffenburger DA (2007) Quantitative network signal combinations downstream of TCR activation can predict IL2 production response. J Immunol 178: 4984-4992.
Kroese M, Zimmern RL, and Pinder SE (2007) HER2 status in breast cancer - an example of pharmacogenetic testing. J R Soc Med 100: 326-329.
Kumar N, Wolf-Yadlin A, White FM, and Lauffenburger DA (2006a) Modeling HER2 effects on cell behavior from mass spectrometry phosphotyrosine data. PLoS Comp Biol 3: e4.
Kumar N, Zaman MH, Kim HD, and Lauffenburger DA (2006b) A high-throughput migration assay reveals HER2-mediated cell migration arising from increased directional persistence. Biophys J 91: L32-L34.[CrossRef][Medline]
Kung C and Shokat KM (2005) Small molecule kinase inhibitor target assessment. Chembiochem 6: 523-526.[CrossRef][Medline]
Lazo JS, Brady LS, and Dingledine R (2007) Building a pharmacological lexicon: small molecule discovery in academia. Mol Pharmacol 72: 1-7.
Merlot S and Firtel RA (2003) Leading the way: directional sensing through phosphatidylinositol 3-kinase and other signaling pathways. J Cell Sci 116: 3471-3478.
Moelling K. Schad K, Bosse M, Zimmermann S, and Schweneker M (2002) Regulation of Raf-Akt cross-talk. J Biol Chem 277: 31099-31106.
Miller-Jensen KE, Janes KA, Brugge JS, and Lauffenburger DA (2007) Common effector-processing mediates cell-specific responses to stimuli. Nature 448: 604-608.[CrossRef][Medline]
Moasser MM (2007) The oncogene HER2: its signaling and transforming functions and its role in human cancer pathogenesis. Oncogene 26: 6469-6487.[CrossRef][Medline]
Mograbi B, Bocciardi R, Bourgett I, Busca R, Rochet N, Farahi-Far D, Juhel T, and Rossi B (2001) Glial cell-derived neurotrophic factor-stimulated PI3K and Akt activities exert opposing effects on the ERK pathway. J Biol Chem 276: 45307-45319.
Natarajan M, Lin KM, Hsueh RC, Sternweis PC, and Ranganathan R (2006) A global analysis of crosstalk in a mammalian cellular signaling network. Nat Cell Biol 8: 571-580.[CrossRef][Medline]
Normanno N, DeLuca A, Bianco C, Strizzi L, Mancino M, Maiello MR, Carotenuto A, DeFeo G, Caponigro F, and Salomon DS (2006) Epidermal growth factor receptor (EGFR) signaling in cancer. Gene 366: 2-16.[CrossRef][Medline]
Oda K, Matsuoka Y, Funahashi A, and Kitano H (2005) A comprehensive pathway map of epidermal growth factor receptor signaling. Mol Syst Biol 1:2005.0010.
Pankov R, Endo Y, Even-Ram S, Araki M, Clark K, Cukierman E, Matsumoto K, and Yamada KM (2005) A Rac switch regulates random versus directionally persistent cell migration. J Cell Biol 170: 793-802.
Price ND, Trent J, El-Naggar AK, Cogdell D, Taylor E, Hunt KK, Pollock RE, Hood L, Shmulevich I, and Zhang W (2007) Highly accurate two-gene classifier for differentiating gastrointestinal stromal tumors and leiomyosarcomas. Proc Natl Acad Sci U S A 104: 3414-3419.
Sawyer C, Sturge J, Bennett DC, O'Hare MJ, Allen WE, Bain J, Jones GE, and Vanhaesebroeck B (2003) Regulation of breast cancer cell chemotaxis by the PI3K p110
. Cancer Res 63: 1667-1675.
Sevecka M and Macbeath G (2006) State-based discovery: a multidimensional screen for small-molecule modulators of EGF signaling. Nat Methods 3: 825-831.[CrossRef][Medline]
Shin I, Kim S, Song H, Kim HR, and Moon A (2005) H-Ras-specific activation of Rac-MKK3/6-p38 pathway: its critical role in invasion and migration of breast epithelial cells. J Biol Chem 280: 14675-14683.
Spencer KS, Graus-Porta D, Leng J, Hynes NE, and Klemke RL (2000) ErbB2 is necessary for induction of carcinoma cell invasion by ErbB family receptor tyrosine kinases. J Cell Biol 148: 385-397.
Wennström S and Downward J (1999) Role of phosphoinositide 3-kinase in activation of ras and mitogen-activated protein kinase by epidermal growth factor. Mol Cell Biol 19: 4279-4288.
Wolf-Yadlin A, Kumar N, Zhang Y, Hautaniemi S, Zaman M, Kim HD, Grantcharova V, Lauffenburger DA, and White FM (2006) Effects of HER2 overexpression on cell signaling networks governing proliferation and migration. Mol Syst Biol 2: 54.[Medline]
Yoeli-Lerner M, Yiu GK, Rabinovitz I, Erhardt P, Jauliac S, and Toker A (2005) Akt blocks breast cancer cell motility and invasion through the transcription factor NFAT. Mol Cell 20: 539-550.[CrossRef][Medline]
Zhang Z and Meier KE (2006) New assignments for multitasking signal transduction inhibitors. Mol Pharmacol 69: 1510-1512.
Zhu AX (2008) Development of sorafenib and other molecularly targeted agents in hepatocellular carcinoma. Cancer 112: 250-259.[CrossRef][Medline]
Zimmermann S and Moelling K (1999) Phosphorylation and regulation of Raf by Akt (protein kinase B). Science 286: 1741-1744.
This article has been cited by other articles:
|
|
 |
|
  H.-D. Kim, T. W. Guo, A. P. Wu, A. Wells, F. B. Gertler, and D. A. Lauffenburger Epidermal Growth Factor-induced Enhancement of Glioblastoma Cell Migration in 3D Arises from an Intrinsic Increase in Speed But an Extrinsic Matrix- and Proteolysis-dependent Increase in Persistence Mol. Biol. Cell, October 1, 2008; 19(10): 4249 - 4259. [Abstract] [Full Text] [PDF]
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
 |
 |
|
 |
 |
 |
