Limits...
Modeling HER2 effects on cell behavior from mass spectrometry phosphotyrosine data.

Kumar N, Wolf-Yadlin A, White FM, Lauffenburger DA - PLoS Comput. Biol. (2006)

Bottom Line: An understanding of how phenotype is determined requires the distributed characterization of signaling processes (e.g., phosphorylation states and kinase activities) in parallel with measures of resulting cell function.Finally, model analysis identified nine especially informative phosphorylation sites on six proteins that recapitulated the predictive capability of the full model.Thus, a PLSR modeling approach reveals critical signaling processes regulating HER2-mediated cell behavior.

View Article: PubMed Central - PubMed

Affiliation: Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts, United States of America.

ABSTRACT
Cellular behavior in response to stimulatory cues is governed by information encoded within a complex intracellular signaling network. An understanding of how phenotype is determined requires the distributed characterization of signaling processes (e.g., phosphorylation states and kinase activities) in parallel with measures of resulting cell function. We previously applied quantitative mass spectrometry methods to characterize the dynamics of tyrosine phosphorylation in human mammary epithelial cells with varying human epidermal growth factor receptor 2 (HER2) expression levels after treatment with epidermal growth factor (EGF) or heregulin (HRG). We sought to identify potential mechanisms by which changes in tyrosine phosphorylation govern changes in cell migration or proliferation, two behaviors that we measured in the same cell system. Here, we describe the use of a computational linear mapping technique, partial least squares regression (PLSR), to detail and characterize signaling mechanisms responsible for HER2-mediated effects on migration and proliferation. PLSR model analysis via principal component inner products identified phosphotyrosine signals most strongly associated with control of migration and proliferation, as HER2 expression or ligand treatment were individually varied. Inspection of these signals revealed both previously identified and novel pathways that correlate with cell behavior. Furthermore, we isolated elements of the signaling network that differentially give rise to migration and proliferation. Finally, model analysis identified nine especially informative phosphorylation sites on six proteins that recapitulated the predictive capability of the full model. A model based on these nine sites and trained solely on data from a low HER2-expressing cell line a priori predicted migration and proliferation in a HER2-overexpressing cell line. We identify the nine signals as a "network gauge," meaning that when interrogated together and integrated according to the quantitative rules of the model, these signals capture information content in the network sufficiently to predict cell migration and proliferation under diverse ligand treatments and receptor expression levels. Examination of the network gauge in the context of previous literature indicates that endocytosis and activation of phosphoinositide 3-kinase (PI3K)-mediated pathways together represent particularly strong loci for the integration of the multiple pathways mediating HER2's control of mammary epithelial cell proliferation and migration. Thus, a PLSR modeling approach reveals critical signaling processes regulating HER2-mediated cell behavior.

Show MeSH

Related in: MedlinePlus

PLSR-Generated Scores Plot Reveals Different Signaling Strategies for EGF and HRG(A) A scores plot identifies separation in signaling strategies associated with receptor overexpression or differential ligand treatment along two signaling axes.(B) HRG and EGF treatment give rise to different sets of signals, and the difference is exaggerated in 24H cells.(C) The linear superposition of the difference vector between 24H and parental serum-free conditions and each ligand's vector explains 24H + HRG signaling but not 24H + EGF signaling.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC1761044&req=5

pcbi-0030004-g003: PLSR-Generated Scores Plot Reveals Different Signaling Strategies for EGF and HRG(A) A scores plot identifies separation in signaling strategies associated with receptor overexpression or differential ligand treatment along two signaling axes.(B) HRG and EGF treatment give rise to different sets of signals, and the difference is exaggerated in 24H cells.(C) The linear superposition of the difference vector between 24H and parental serum-free conditions and each ligand's vector explains 24H + HRG signaling but not 24H + EGF signaling.

Mentions: Decomposition via PLSR of the signal (X) and response (Y) datasets provided a reduced-dimension map (called the scores vector t, see Methods) on which network signaling changes in response to ligand or receptor perturbation could be interpreted. The plot axes (referred to as principal components) are linear combinations of the signaling metrics described above. Figure 3 plots the changes corresponding to ligand or receptor perturbation on a two-dimensional graph whose axes are the first two principal components (the third component, which captures only 4% of the Y-block variance described by the full model, is omitted here for ease of visualization). The plot shows, as expected, that HRG treatment stimulates the signaling network distinctly from the EGF treatment case. If HRG treatment activated the same set of signals as EGF with different magnitudes, we would observe the +EGF and +HRG vectors overlapping, with one being longer than the other. Since HRG and EGF drive the formation of different HER dimers (i.e., HER3–HER2 versus EGFR–HER2 or EGFR–EGFR), we expect differential activation of the signaling network. Interestingly, the difference between EGF and HRG signaling is larger in 24H cells relative to parental cells as indicated by the offset of the two vectors (77 degrees versus 37 degrees, respectively), due in large part to a drastic shift in EGF-induced signaling with HER2 overexpression (Figure 3B). If we linearly superpose changes in signaling due to HER2 overexpression in the absence of ligand (i.e., the signaling changes under serum-free conditions between P and 24H cells) and the changes in signals as we add a ligand (i.e., signaling changes as either HRG or EGF are added to P cells), we can approximate the signals generated by 24H cells under HRG treatment (Figure 3C). We cannot do the same, however, for 24H cells under EGF treatment, emphasizing the nonadditive interplay between changes in cellular ligand–receptor conditions and the signals they generate (Figure 3C). In the case of our HMEC system, previous quantitative measurement and modeling of dimer kinetics has shown that HER2 overexpression in the presence of HRG leads to a relatively small shift in the number of HER2–HER3 dimers (∼10,000). Alternatively, HER2 overexpression in the presence of EGF leads to a large increase in EGFR–HER2 dimers (∼150,000) and a decrease in EGFR homodimers (∼65,000) [10]. Changes in signaling with HER2 overexpression under serum-free conditions could be due to either spontaneous homodimerization or autocrine signaling. Thus, given our analysis of the scores plot, we hypothesize that increases in HER2 under HRG treatment principally add to the signaling network through the independent addition of signals generated through HER2 homodimerization or autocrine signaling. HER2 overexpression under EGF treatment, however, results in the addition of these signals plus a novel suite of signals generated primarily through an increase in EGFR–HER2 dimers and loss of EGFR homodimers. These data support a previous finding that increasing levels of phosphorylated EGFR and HER2 induce novel signaling elements associated with lower affinity intracellular receptor–protein binding [12].


Modeling HER2 effects on cell behavior from mass spectrometry phosphotyrosine data.

Kumar N, Wolf-Yadlin A, White FM, Lauffenburger DA - PLoS Comput. Biol. (2006)

PLSR-Generated Scores Plot Reveals Different Signaling Strategies for EGF and HRG(A) A scores plot identifies separation in signaling strategies associated with receptor overexpression or differential ligand treatment along two signaling axes.(B) HRG and EGF treatment give rise to different sets of signals, and the difference is exaggerated in 24H cells.(C) The linear superposition of the difference vector between 24H and parental serum-free conditions and each ligand's vector explains 24H + HRG signaling but not 24H + EGF signaling.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC1761044&req=5

pcbi-0030004-g003: PLSR-Generated Scores Plot Reveals Different Signaling Strategies for EGF and HRG(A) A scores plot identifies separation in signaling strategies associated with receptor overexpression or differential ligand treatment along two signaling axes.(B) HRG and EGF treatment give rise to different sets of signals, and the difference is exaggerated in 24H cells.(C) The linear superposition of the difference vector between 24H and parental serum-free conditions and each ligand's vector explains 24H + HRG signaling but not 24H + EGF signaling.
Mentions: Decomposition via PLSR of the signal (X) and response (Y) datasets provided a reduced-dimension map (called the scores vector t, see Methods) on which network signaling changes in response to ligand or receptor perturbation could be interpreted. The plot axes (referred to as principal components) are linear combinations of the signaling metrics described above. Figure 3 plots the changes corresponding to ligand or receptor perturbation on a two-dimensional graph whose axes are the first two principal components (the third component, which captures only 4% of the Y-block variance described by the full model, is omitted here for ease of visualization). The plot shows, as expected, that HRG treatment stimulates the signaling network distinctly from the EGF treatment case. If HRG treatment activated the same set of signals as EGF with different magnitudes, we would observe the +EGF and +HRG vectors overlapping, with one being longer than the other. Since HRG and EGF drive the formation of different HER dimers (i.e., HER3–HER2 versus EGFR–HER2 or EGFR–EGFR), we expect differential activation of the signaling network. Interestingly, the difference between EGF and HRG signaling is larger in 24H cells relative to parental cells as indicated by the offset of the two vectors (77 degrees versus 37 degrees, respectively), due in large part to a drastic shift in EGF-induced signaling with HER2 overexpression (Figure 3B). If we linearly superpose changes in signaling due to HER2 overexpression in the absence of ligand (i.e., the signaling changes under serum-free conditions between P and 24H cells) and the changes in signals as we add a ligand (i.e., signaling changes as either HRG or EGF are added to P cells), we can approximate the signals generated by 24H cells under HRG treatment (Figure 3C). We cannot do the same, however, for 24H cells under EGF treatment, emphasizing the nonadditive interplay between changes in cellular ligand–receptor conditions and the signals they generate (Figure 3C). In the case of our HMEC system, previous quantitative measurement and modeling of dimer kinetics has shown that HER2 overexpression in the presence of HRG leads to a relatively small shift in the number of HER2–HER3 dimers (∼10,000). Alternatively, HER2 overexpression in the presence of EGF leads to a large increase in EGFR–HER2 dimers (∼150,000) and a decrease in EGFR homodimers (∼65,000) [10]. Changes in signaling with HER2 overexpression under serum-free conditions could be due to either spontaneous homodimerization or autocrine signaling. Thus, given our analysis of the scores plot, we hypothesize that increases in HER2 under HRG treatment principally add to the signaling network through the independent addition of signals generated through HER2 homodimerization or autocrine signaling. HER2 overexpression under EGF treatment, however, results in the addition of these signals plus a novel suite of signals generated primarily through an increase in EGFR–HER2 dimers and loss of EGFR homodimers. These data support a previous finding that increasing levels of phosphorylated EGFR and HER2 induce novel signaling elements associated with lower affinity intracellular receptor–protein binding [12].

Bottom Line: An understanding of how phenotype is determined requires the distributed characterization of signaling processes (e.g., phosphorylation states and kinase activities) in parallel with measures of resulting cell function.Finally, model analysis identified nine especially informative phosphorylation sites on six proteins that recapitulated the predictive capability of the full model.Thus, a PLSR modeling approach reveals critical signaling processes regulating HER2-mediated cell behavior.

View Article: PubMed Central - PubMed

Affiliation: Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts, United States of America.

ABSTRACT
Cellular behavior in response to stimulatory cues is governed by information encoded within a complex intracellular signaling network. An understanding of how phenotype is determined requires the distributed characterization of signaling processes (e.g., phosphorylation states and kinase activities) in parallel with measures of resulting cell function. We previously applied quantitative mass spectrometry methods to characterize the dynamics of tyrosine phosphorylation in human mammary epithelial cells with varying human epidermal growth factor receptor 2 (HER2) expression levels after treatment with epidermal growth factor (EGF) or heregulin (HRG). We sought to identify potential mechanisms by which changes in tyrosine phosphorylation govern changes in cell migration or proliferation, two behaviors that we measured in the same cell system. Here, we describe the use of a computational linear mapping technique, partial least squares regression (PLSR), to detail and characterize signaling mechanisms responsible for HER2-mediated effects on migration and proliferation. PLSR model analysis via principal component inner products identified phosphotyrosine signals most strongly associated with control of migration and proliferation, as HER2 expression or ligand treatment were individually varied. Inspection of these signals revealed both previously identified and novel pathways that correlate with cell behavior. Furthermore, we isolated elements of the signaling network that differentially give rise to migration and proliferation. Finally, model analysis identified nine especially informative phosphorylation sites on six proteins that recapitulated the predictive capability of the full model. A model based on these nine sites and trained solely on data from a low HER2-expressing cell line a priori predicted migration and proliferation in a HER2-overexpressing cell line. We identify the nine signals as a "network gauge," meaning that when interrogated together and integrated according to the quantitative rules of the model, these signals capture information content in the network sufficiently to predict cell migration and proliferation under diverse ligand treatments and receptor expression levels. Examination of the network gauge in the context of previous literature indicates that endocytosis and activation of phosphoinositide 3-kinase (PI3K)-mediated pathways together represent particularly strong loci for the integration of the multiple pathways mediating HER2's control of mammary epithelial cell proliferation and migration. Thus, a PLSR modeling approach reveals critical signaling processes regulating HER2-mediated cell behavior.

Show MeSH
Related in: MedlinePlus