Cancer Research AACR Membership Sign up for Cancer Research eTOC's
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
Cancer Research Clinical Cancer Research
Cancer Epidemiology Biomarkers & Prevention Molecular Cancer Therapeutics
Molecular Cancer Research Cancer Prevention Research
Cancer Prevention Journals Portal Cancer Reviews Online
Annual Meeting Education Book Meeting Abstracts Online
Cancer Research 69, 6713, August 15, 2009. Published Online First July 28, 2009;
doi: 10.1158/0008-5472.CAN-09-0777
© 2009 American Association for Cancer Research
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Supplementary Data
Right arrow All Versions of this Article:
0008-5472.CAN-09-0777v1
69/16/6713    most recent
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Google Scholar
Right arrow Articles by Faratian, D.
Right arrow Articles by Harrison, D. J.
PubMed
Right arrow PubMed Citation
Right arrow Articles by Faratian, D.
Right arrow Articles by Harrison, D. J.

Systems Biology and Emerging Technologies

Systems Biology Reveals New Strategies for Personalizing Cancer Medicine and Confirms the Role of PTEN in Resistance to Trastuzumab

Dana Faratian1, Alexey Goltsov2, Galina Lebedeva2, Anatoly Sorokin2, Stuart Moodie2, Peter Mullen1, Charlene Kay1, In Hwa Um1, Simon Langdon1, Igor Goryanin2,3 and David J. Harrison1

1 Edinburgh Breakthrough Research Unit and Division of Pathology, 2 Centre for Systems Biology at Edinburgh, and 3 School of Informatics, Informatics Forum, University of Edinburgh, Edinburgh, Scotland, United Kingdom

Requests for reprints: Dana Faratian, Edinburgh Breakthrough Research Unit and Division of Pathology, University of Edinburgh, Crewe Road South, Edinburgh EH4 2XR, Scotland, United Kingdom. Phone: 44-131-537-1763; Fax: 44-131-537-3159; E-mail: d.faratian{at}ed.ac.uk.


   Abstract
Top
 Abstract
Introduction
 Materials and Methods
 Results
 Discussion
 Disclosure of Potential...
 References
 
Resistance to targeted cancer therapies such as trastuzumab is a frequent clinical problem not solely because of insufficient expression of HER2 receptor but also because of the overriding activation states of cell signaling pathways. Systems biology approaches lend themselves to rapid in silico testing of factors, which may confer resistance to targeted therapies. Inthis study, we aimed to develop a new kinetic model that could be interrogated to predict resistance to receptor tyrosine kinase (RTK) inhibitor therapies and directly test predictions in vitro and in clinical samples. The new mathematical model included RTK inhibitor antibody binding, HER2/HER3 dimerization and inhibition, AKT/mitogen-activated protein kinase cross-talk, and the regulatory properties of PTEN. The model was parameterized using quantitative phosphoprotein expression data from cancer cell lines using reverse-phase protein microarrays. Quantitative PTEN protein expression was found to be the key determinant of resistance to anti-HER2 therapy in silico, which was predictive of unseen experiments in vitro using the PTEN inhibitor bp(V). When measured in cancer cell lines, PTEN expression predicts sensitivity to anti-HER2 therapy; furthermore, this quantitative measurement is more predictive of response (relative risk, 3.0; 95% confidence interval, 1.6–5.5; P < 0.0001) than other pathway components taken in isolation and when tested by multivariate analysis in a cohort of 122 breast cancers treated with trastuzumab. For the first time, a systems biology approach has successfully been used to stratify patients for personalized therapy in cancer and is further compelling evidence that PTEN, appropriately measured in the clinical setting, refines clinical decision making in patients treated with anti-HER2 therapies. [Cancer Res 2009;69(16):6713–20]


   Introduction
Top
 Abstract
 Introduction
Materials and Methods
 Results
 Discussion
 Disclosure of Potential...
 References
 
HER2-targeting receptor tyrosine kinase (RTK) inhibitors, such as trastuzumab and pertuzumab, show clinical efficacy in breast and ovarian cancer but measurement of HER2 protein expression or gene amplification status is a poor predictor of response with a very low positive predictive value (1, 2). The documented actual benefit of adjuvant trastuzumab combined with chemotherapy versus chemotherapy alone in terms of overall survival is only modest [96% versus 95%, respectively, at 1 year (1) and 91% versus 87%, respectively, at 4 years (2)]. A large proportion of patients therefore unnecessarily receive ineffective and expensive treatments with toxic side effects, and there is a need to be able to identify markers that can be used to predict therapeutic response and readily applied to the clinic. Activating mutations in the oncogene PIK3CA and inactivation of the tumor suppressor gene PTEN are known to regulate phosphatidylinositol 3-kinase (PI3K) signaling and cell proliferation and survival pathways, and empirically, these have been shown to play a role in resistance to anti-HER2 therapies in breast cancer (3, 4). However, the evidence supporting the role of PI3K pathway activation in RTK inhibitor resistance has only been tested in very small cohorts of patients retrospectively (n = 55 and n = 47, respectively; refs. 3, 4). Further evidence is required to support the use of PI3K pathway biomarkers in the clinic supported by robust preclinical and clinical evidence.

Kinetic (or dynamic) computational models offer the opportunity to cheaply and efficiently test the efficacy of targeted therapies in silico (i.e., computationally) as part of the preclinical testing process (5). However, this methodology is still perceived as esoteric and not clinically applicable because there are very few successful examples of when such modeling has changed practice (6). One notable exception is modeling of ion channels in the heart, which has helped explain the previously ill-defined action of the drug ranolazine (7, 8). These modeling data were subsequently used within a Food and Drug Administration submission as supportive evidence for mode of action (9).4 In oncology, systems biology has not yet made the same effect, but the use of therapies targeted against the products of cellular oncogenes in signaling pathways lends itself to this approach because these pathways can readily be modeled using ordinary differential equations (5). To date, canonical pathways such as the epidermal growth factor receptor pathway have only been modeled to explain and predict physiologic phenomena (10–16). However, such models have not been so helpful for understanding therapeutic interventions because they frequently fail to include important oncogene and tumor suppressor nodes, which are fundamental to carcinogenesis and proven resistance proteins [such as HER2, PTEN, and SRC in PI3K and mitogen-activated protein kinase (MAPK) signaling models].

The aim of this study was to address these deficiencies in current models by (a) developing a new kinetic model that could be interrogated to predict resistance to RTK inhibitor therapies and (b) directly test predictions in vitro and in clinical samples.


   Materials and Methods
Top
 Abstract
 Introduction
 Materials and Methods
Results
 Discussion
 Disclosure of Potential...
 References
 
Computational modeling. A kinetic model of heregulin-induced HER2/3 signaling through MAPK and PI3K pathways was developed based on the model proposed by Kholodenko and colleagues (13). The whole kinetic model contains 56 ordinary differentials equations (ODE; S1.1-S1.56) describing the change of the concentrations of 56 biological entities (proteins, receptors, and lipid second messengers) involved in MAPK and PI3K signaling (see Supplementary Table S1). A detailed modeling methodology and description of the model design is given in S1 and S2 (Supplementary Data—modeling methods and description of mathematical model). The abbreviations used in the text and ODEs are given in Supplementary Table S1. All network schemes were designed using the Edinburgh Pathway Editor (17). The computer model construction, fitting, and analysis were implemented using the DBsolve package for kinetic modeling (18) and SimBiology modeling software (MATLAB, The MathWorks, Inc.). The SBML file of the model is given in "Supplementary Data—SBML model.xml." The model parameterization procedure permitted us to describe satisfactorily experiments of activation in the different branches of the extracellular signal-regulated kinase (ERK)/AKT signaling network (see Supplementary Fig. S1) and determine a set of kinetic parameters of the model (see Supplementary Tables S2 and S3).

Cell culture and collection of lysates. PE04 and BT474 cells were grown as monolayer cultures in DMEM supplemented with 10% heat-inactivated FCS and penicillin/streptomycin (100 IU/mL) in a humidified atmosphere of 5% CO2 at 37°C. Time course experiments were set up by plating cells into 10-cm ø Petri dishes and leaving for 48 h. Cells were then briefly washed in PBS before transferring to phenol red–free DMEM containing 5% double charcoal-stripped serum supplemented with penicillin/streptomycin (100 IU/mL) and glutamine (0.3 mg/mL) for a further 48 h before treatment. Paired lysates were prepared by first treating relevant dishes with pertuzumab (100 nmol/L) ± bpV(pic) (50 nmol) immediately followed by the addition of heregulin-β (1 nmol/L). Samples were collected at time points of 1, 2, 5, 10, 30, 45, or 60 min; washed in PBS; and immediately lysed in ice-cold isotonic lysis buffer [50 mmol/L Tris-HCl (pH 7.5), 5 mmol/L EGTA (pH 8.5), 150 mmol/L NaCl, 1% Triton X-100] supplemented with aprotinin (10 µg/mL) and a protease inhibitor cocktail (Roche). Lysates were centrifuged for 6 min at 13,000 x g and protein concentrations of supernatants were subsequently determined using the bicinchoninic acid assay (Sigma). All cell culture and time course experiments were performed at least three times, and representative reverse-phase protein array (RPPA) and Western blot curves were used for the model fitting procedure described above and to validate model predictions.

Western blotting. Protein lysates were electrophoretically resolved on either 10% or 12% SDS-PAGE and transferred overnight onto Immobilon-P membranes (Millipore). After transfer, membranes were blocked with 1% blocking agent (Roche) in TBS before probing overnight at 4°C with the appropriate primary antibody made up in 0.5% blocking agent. Primary antibodies used for Western blotting were as follows: anti–phospho-AKT (pAKT; Ser473; Cell Signaling Technology) at 1:1,000, anti-AKT (Cell Signaling Technology) at 1:1,000, anti–phospho-p44/42 MAPK (Thr202/Tyr204; Cell Signaling Technology) at 1:1,000, anti-p44/42 MAPK (Thr202/Tyr204; Cell Signaling Technology) at 1:1,000, anti-HER1 (Cell Signaling Technology) at 1:1,000, anti-HER2 (Cell Signaling Technology) at 1:1,000, anti–phospho-HER2 (Tyr877; Cell Signaling Technology) at 1:1,000, anti-HER3 (Cell Signaling Technology) at 1:1,000, anti–phospho-HER3 (Tyr1289; Cell Signaling Technology) at 1:1,000, anti-HER4 (Cell Signaling Technology) at 1:1,000, anti-PI3K (Cell Signaling Technology) at 1:1,000, anti-PTEN (Cell Signaling Technology) at 1:1,000, and anti–phospho-PTEN (Ser380/Thr382/383; Cell Signaling Technology) at 1:1,000. Immunoreactive bands were detected using enhanced chemiluminescent (ECL) reagents (Roche) and Hyperfilm ECL film (GE Healthcare). Bands were scanned using an Epson Perfection 4990 scanner and Integrated Optical Density absorbance values were obtained by densitometric analysis using Labworks gel analysis software (UVP Life Sciences).

Reverse-phase protein arrays. Denatured and reduced protein lysates were spotted onto nitrocellulose-coated glass slides (Eurogentec), as previously described (19). Three replicates were spotted per sample in eight 2-fold dilutions. Slides were hydrated in Whatman wash buffer for 5min and Li-Cor blocking buffer for 1 h (LI-COR Biosciences) and then incubated with primary antibodies overnight at 4°C. The following day, slides were washed in PBS/T at room temperature for 5 min (thrice) before incubating with far-red fluorescently labeled secondary antibodies diluted in Li-Cor Odyssey Blocking Buffer (1 µL/2 mL) at room temperature for 45min with gentle shaking. Slides were then washed in excess PBS/T (thrice)/PBS (thrice) and allowed to air dry before reading on a Li-Cor Odyssey scanner at 680 and 780 nm, and images were exported as TIFF files for further analysis. Slides were stained using the above primary antibodies (matched total and phosphoproteins duplexed on each slide).

RPPA analysis was performed using MicroVigene RPPA analysis module (VigeneTech). Spots were quantified by accurate single segmentation, with actual spot signal boundaries determined by the image analysis algorithm. Each spot intensity was quantified by measuring the total pixel intensity of the area of each spot (volume of spot signal pixels), with background subtraction of 2 pixels around each individual spot. The mean of the replicates was used for normalization and curve fitting. Curve fitting was performed using five-parameter logistical nonlinear regression using a joint estimation approach ("supercurve method"). The quantification y0 (intensity of curve) or rsu (relative concentration value) of sample dilution curves was normalized by corresponding total protein.

PIK3CA mutation analysis and copy number quantification. PIK3CA mutation analysis was performed using the ARMS/Scorpions multiplexed PCR assay as previously described (20) to detect the four most common mutations in PIK3CA [H1047L, H1047R (exon 20), E545K, and E542K (exon 9)]. PIK3CA copy number quantification was performed with primers for PIK3CA and the glucokinase (GCK) gene as the reference gene as previously described (21). Quantitative PCR was performed on the Rotor-Gene 6000 real-time detection system (Corbett Life Sciences) in 15 µL reaction volumes containing 2x Power SYBR Green PCR Master Mix (Applied Biosystems) and 1 µmol/L forward and reverse primers. PCR conditions were 10 min at 95°C followed by 40 cycles consisting of 10 s at 95°C, 15 s at 60°C, and 20 s at 72°C. {Delta}Ct was calculated as GCK Ct – PIK3CA Ct and {Delta}Ct > 3 was defined as amplified.

Samples and tissue microarray construction. The population characteristics of the trastuzumab-treated cohort are summarized in Table 1 . HER2 gene amplification status was confirmed by fluorescence in situ hybridization (DAKO HER2 FISH PharmDx, Ely). The study was approved by the Lothian Research Ethics Committee (08/S1101/41). Following H&E sectioning of representative tumor blocks, tumor areas were marked for tissue microarray (TMA) construction and 0.6 mm2 cores were placed into three separate TMA replicates for each sample, as previously described (22).


View this table:
[in this window]
[in a new window]

  		Table 1. Clinicopathologic characteristics of patients treated with trastuzumab and association with survival (log-rank test)

 
Immunofluorescence. Immunofluorescence for PTEN was performed using methods previously described (23). Briefly, 3-µm TMA slides were deparaffinized and antigen retrieved by pressure cooking in 0.15 mmol/L sodium citrate buffer at pH 6. Endogenous peroxidases were blocked with 2.5% hydrogen peroxide for 15 min and nonspecific binding was blocked with serum-free protein block for 15 min. Slides were then incubated with primary antibodies diluted in 0.025% PBST for 1 h at room temperature [AE1/AE3 mouse monoclonal cytokeratin antibody, rabbit polyclonal to PTEN (Cell Signaling) diluted 1:100 and 1:25, respectively]. After washing in 0.025% PBST, sections were incubated for 1 h at room temperature with secondary antibodies, which included an Alexa Fluor 488–conjugated goat anti-mouse antibody diluted 1:100 in 0.1 mol/L TBS and prediluted goat anti-rabbit antibody conjugated to a horseradish peroxidase–decorated dextran polymer backbone (EnVision, Dako). Slides were then incubated for 10 min with Cy5-tyramide, which is activated by horseradish peroxidase, to visualize HER2 expression. 4',6-Diamidino-2-phenylindole (DAPI; Molecular Probes) was used to stain the nuclear compartment.

AQUA automated image analysis. A detailed description of the AQUA methodology is available elsewhere (23). Pan-cytokeratin antibody was used to identify infiltrating tumor cells and normal epithelial cells, DAPI counterstain to identify nuclei, and Cy5-tyramide detection for target (PTEN) for compartmentalized (tissue and subcellular) analysis of tissue sections. Monochromatic images of each TMA core were captured at 20x objective using an Olympus AX-51 epifluorescence microscope, and high-resolution digital images were analyzed by the AQUAnalysis software. Briefly, a binary epithelial mask was created from the cytokeratin image of each TMA core. If the epithelium comprised <5% of total core area, the core was excluded from analysis. Similar binary masks were created for cytoplasmic and nuclear compartments based on DAPI staining of nuclei. PTEN expression was quantified by calculating the Cy5 fluorescent signal intensity on a scale of 0 to 255 within each image pixel, and the AQUA score was generated by dividing the sum of Cy5 signal within the epithelial mask by the area of the cytoplasmic compartment.

Statistical analysis methods. AQUA scores were averaged from replicate cores. To reduce type I error that can result from using the minimum P value method for determining the cutpoint value for PTEN expression in Kaplan-Meier analysis (24), we used X-Tile, which allows determination of an optimal cutpoint while correcting for the use of minimum P statistics (25). Two methods of statistical correction for the use of minimal P approach were used: the first calculation of a Monte-Carlo P value and, for the second, the Miller-Siegmund minimal P correction (24). Overall survival was subsequently assessed by Kaplan-Meier analysis with log rank for determining statistical significance. Relative risk was assessed by the univariate and multivariate Cox proportional hazards model. Comparison of differences in means in vitro was performed using the Student's t test. All calculations and analyses were two tailed where appropriate and done with SPSS 14.0 for Windows (SPSS, Inc.).


   Results
Top
 Abstract
 Introduction
 Materials and Methods
 Results
Discussion
 Disclosure of Potential...
 References
 
Development of a mathematical model to predict responses to RTK inhibitors. To systematically assess resistance factors to anti-RTK therapy in cancer biopsies, we developed a kinetic mathematical model of MAPK/PI3K signaling and used it to predict consequences of anti-HER2 monoclonal antibody therapeutic intervention (a full kinetic mathematical model description is given in Supplementary Data—modeling methods and mathematical model description). We developed it to describe HER2 inhibitor antibody/receptor binding, HER2/HER3 dimerization and inhibition, AKT/MAPK cross-talk, and the kinetic and regulatory properties of PTEN (Fig. 1 ; Supplementary Figs. S2–S4 and S10). The design of our model wasalso based on modeling studies of the HER signaling network (13, 26–28). The inclusion of the tumor suppressor protein PTEN was deemed particularly important because it is a key negative regulator of the PI3K signaling pathway (29), and it has been implicated as a resistance mechanism to trastuzumab in breast cancer (3, 4); however, the effect of quantitative decreases in protein expression, which is more commonly seen than mutation or epigenetic regulation, is unknown. The new model, including a PTEN subsystem, explicitly describes the competition between the lipid and protein phosphatase activities of PTEN, autodephosphorylation, and the balance and interchange between the active (PTEN) and inactive (pPTEN) forms of the protein using experimentally derived kinetic parameters (Fig. 1; Supplementary Fig. S10; ref. 30). The kinetic model (Fig. 1 and submodels shown in Supplementary Data) was then fit to high-resolution in vitro dynamic phosphoprotein expression generated from RPPAs and normalized to total protein expression (Fig. 2A ) as well as supportive Western blot data (Supplementary Figs. S1, S12, and S13). Time course data were generated from PE04 ovarian cancer cell lines stimulated with heregulin-β in the presence or absence of the HER2 dimerization inhibitor pertuzumab (Fig. 2A; Supplementary Fig. S12) to derive kinetic parameters, as described in Supplementary Data—modeling methods and description of mathematical model. Additional time course profiles of BT474 breast cancer cell lines were studied, but not used for parameterization of the mathematical model, to establish qualitative similarities between cell lines and the generality of the approach. Additional model validation was provided by heregulin dose-response experiments (Supplementary Fig. S13). The model satisfactorily reproduced the response kinetics of heterodimerization of HER2/HER3, phospho-ERK (pERK), and pAKT activation stimulated by heregulin-β and the inhibitory effect of pertuzumab (Supplementary Fig. S1). A dominant signaling effect through PI3K with sustained steady-state changes in PI3K signaling in response to heregulin-β after 1 hour in multiple cell lines was observed (Fig. 2A).


Figure 1
View larger version (24K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 1. Network schema of MAPK and PI3K signaling. Schematic representation of mathematical model used to interrogate sensitivity and resistance factors to RTK inhibitors. Abbreviations used within the scheme are shown in Supplementary Table S1.

 

Figure 2
View larger version (30K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 2. Computational model validation and predictions. A, time course plots of PE04 ovarian carcinoma cell line experimental data used to parameterize the computational model. Data were generated from RPPA (phosphoprotein normalized by total protein) and are representative of at least three replicate experiments. Blue, heregulin-β; red, heregulin + pertuzumab. Bars, 95% confidence intervals of technical replicates. B, the theoretical dependence of therapeutic resistance R80 on the ratio {gamma} of PTEN and PI3K* ({gamma} = PTEN/PI3K*; see Supplementary Data—modeling methods and description of mathematical model S3 for details). The index of therapeutic resistance (R) was defined as follows: R80 = pAKT80/pAKT0, where pAKT80 is the value of pAKT signal when 80% of receptors are inhibited by the drug and pAKT0 is the pAKT signal in the absence of RTK inhibitors. Thus, R reflects the nonresponsiveness (or resistance) of the pAKT signal to the inhibition of receptors by RTK inhibitors. When R80 = 1, the resistance is the highest, meaning that 80% inhibition of RTK causes no inhibition of pAKT signal. High resistance is observed on decreasing {gamma} three to five times in relation to {gamma}o set in the model. C, predicted effect of PTEN loss on pAKT kinetics (left) ± pertuzumab and resulting experimental data (right). D, comparison of computational predictions and unseen experiments of steady-state levels of pAKT at 30 min. Kinetic and steady-state profiles of pAKT by inhibition of PTEN function with bpV(pic) closely match computational simulations.

 
The balance of the enzyme activities in the PI3K/PTEN cycle dictates sensitivity and resistance to RTK inhibitor therapy. To predict which pathway components most influence therapeutic resistance to anti-HER2 therapy, we used the kinetic model to interrogate which parameters most influenced downstream pathway activation in response to endogenous ligand stimulation and inhibition with dimerization inhibitors. Computational analysis of the model showed that HER2 receptor levels, total PTEN levels, and activation of PI3K by mutation led to dimerization inhibitor resistance measured by sustained AKT phosphorylation after treatment with the dimerization inhibitor pertuzumab (Supplementary Fig. S15); however, because the activation of PI3K required to induce resistance was supraphysiologic (x10 required to induce resistance in simulations versus x2 activation by mutation as seen in vitro; ref. 31), and patients are always selected based on high HER2 expression for RTK inhibitor therapy, we focused on PTEN as potential resistance mechanisms to offer a quantitative explanation for therapeutic resistance observed in vivo (3). PTEN exists in a signaling cycle with PI3K such that the balance between the activities of PTEN and PI3K enzymes, PTEN:activated PI3K (integrated resistance factor {gamma}; Fig. 2B; Supplementary Data—modeling methods and description of model S3), dictates the sensitivity of AKT activation to ligand stimulation or inhibition. High {gamma} results in effective inhibition of AKT activation in response to pertuzumab, whereas low {gamma} predicts sustained AKT activation and insensitivity to pertuzumab (Fig. 2B). Thus, we found that the resistance to inhibition is strongly {gamma} dependent. At low {gamma}, the system amplifies small signals from membrane receptors and becomes insensitive to receptor inhibition. In Fig. 2B, this regimen corresponds to high values of the therapeutic resistance index R80({gamma}). In the range of {gamma}o (corresponding to the computational results in Fig. 2A), signal transduction through PI3K/PTEN cycle is sensitive to inhibition by pertuzumab (pAKT signal can be inhibited by 30–70%). Further increase in {gamma} (>10 times higher than {gamma}o) leads to strong inhibition of pAKT signals due to significant weakening of the signal transduction through the PTEN/PI3K cycle (see Supplementary Data—modeling methods and description S3). Our results are in a good agreement with ultrasensitive and signal-transducing regimens of signaling cascades described earlier for covalent modification (32, 33) and kinase-phosphatase cycles (34).

Computational simulations are predictive of unseen experiments of therapeutic resistance caused by resistance factor {gamma}. We tested the effect of changing the PTEN:PI3K ratio by stimulating with heregulin-β after pretreating PE04 cell lines with bpV(pic), a potent and specific inhibitor of PTEN, which acts by binding to the active phosphatase CX5R motif (35, 36). The model was entirely predictive of these unseen in vitro experiments, with dynamics (compared by curve comparison and integration of kinetic profiles) and steady-state levels of pAKT activation matching computational simulations and abrogating inhibition of pAKT with pertuzumab (Fig. 2C). Likewise, increasing the ratio of PTEN:activated PI3K with the PI3K inhibitor LY294002 resulted in dose-dependent inhibition of pAKT, but not pERK, in PE04 cells (Supplementary Fig. S14). This is also in agreement with our theoretical results (see Supplementary Data—modeling methods and description S3). Therefore, the model suggests that resistance to RTK inhibitors is dependent on the quantitative balance of both the protein expression of PTEN (which is common in breast cancer; ref. 37) and PI3K activating mutations, which are commonly amplified and mutated in many breast and ovarian cancers (38, 39).

The resistance factor {gamma} dictates resistance to pertuzumab in cellular models in vitro. Further validation of the capacity of measurement of PTEN:activated PI3K ratio to predict therapeutic response was sought from cellular models of therapeutic sensitivity and resistance. We measured total PTEN, PI3K mutation status of the four most common mutations in the PIK3CA gene (H1047L and H1047R in exon 20 and E545K and E542K in exon 9, representing ~90% of PIK3CA mutations), and PI3K amplification status in a panel of 13 ovarian carcinoma cell lines to investigate the association of PTEN:activated PI3K with response to pertuzumab (Fig. 3A ). Total PTEN protein expression levels were strongly negatively correlated with both pAKT and percentage growth inhibition with pertuzumab (Pearson's correlation coefficient = 0.68 and 0.74, respectively; Table 2 ). Two of the cell lines, SKOV3 and OAW42, showed activating mutations in PI3K, but none of the cell lines harbored PIK3CA amplifications. Activation by mutation was assumed to increase the activity of PI3K 2-fold (31), and calculation of PTEN:activated PI3K resulted in small increases in correlation coefficients with pAKT and growth to 0.73 and 0.76, respectively, suggesting that PTEN expression is the dominant factor in sensitivity to HER2 dimerization inhibition, consistent with recent reports (40), and the relatively weak effect of x2 activity of PI3K on resistance seen in simulations. However, measurement of HER1, HER2, HER3, pAKT, or pERK was not highly correlated with sensitivity to pertuzumab (Table 2).


Figure 3
View larger version (28K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 3. Quantitative PTEN protein expression is associated with pertuzumab sensitivity in vitro and trastuzumab resistance in vivo. A, relative sensitivity of 13 ovarian carcinoma cell lines to the HER2 dimerization inhibitor pertuzumab was determined by growth assay [sulforhodamine B (SRB) assay], and expression of PTEN and other MAPK and PI3K pathway components was assessed by semiquantitative Western blotting. B, left, AQUA fluorescent analysis of PTEN expression in a TMA core, showing mainly cytoplasmic localization of PTEN (red) and masking of tumor areas for quantification by cytokeratin (green); middle, calculation of optimal cutpoint for PTEN expression. X-tile calculates all possible combinations of results for different cutpoints (top) and corresponding relative risks (bottom). Cutpoint is selected on the best-corrected Monte-Carlo P value. Right, Kaplan-Meier survival curves for patients treated with trastuzumab for low (blue) and high (red) protein expression of PTEN. Overall survival is calculated from time of initial diagnosis to date of death.

 

View this table:
[in this window]
[in a new window]

  		Table 2. Pearson's correlation coefficients of association of protein expression of pathway components and resistance factor {gamma} with pAKT and growth (SRB assay)

 
The quantitative expression of PTEN is associated to trastuzumab resistance in the clinic. Because both computational and cellular modeling strongly suggest that quantitative PTEN protein expression most influences ultimate cellular response to HER2 inhibition, we quantified expression of PTEN in 122 trastuzumab-treated primary breast tumors using the AQUA quantitative image analysis system (Fig. 3B; Table 1) on primary breast tumors arrayed on TMAs. Increasing PTEN expression was proportional to decreased relative risk of death when treated with trastuzumab (Fig. 3B, middle), mirroring the effect on the therapeutic resistance index R80({gamma}) by changes in {gamma}. A rigorously defined cutpoint for PTEN expression (see Materials and Methods) resulted in a mean reduction in overall survival of 21.6 months in PTENlow versus PTENhigh patients (uncorrected P = 0.0003, Miller-Siegmund P = 0.0099, Monte-Carlo P = 0.01; Fig. 3B), equivalent to a 3.0 times increase in relative risk of dying from breast cancer after treatment with trastuzumab in PTENlow cases (relative risk, 3.0; 95% confidence interval, 1.6-5.5; P < 0.0001). In univariate analysis, tumor size, ER status, chemotherapy regimen, and PTEN expression levels were all associated with significant survival differences (P < 0.05, log-rank test; Table 1), but PTEN remained the only significant predictor of survival in multivariate analysis (P = 0.01, Cox logistic regression) and corresponds well to the parameter {gamma} when PI3K expression level assumed to be constant. HER2 expression and PIK3CA mutation status [mutations found in 26 of 119 (21.8%) available samples] were not predictive of survival.


   Discussion
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
Disclosure of Potential...
 References
 
Our data have shown the successful application of kinetic modeling of the PI3K and MAPK signaling pathways in the selection of resistance factors to RTK inhibitors. We have shown that PTEN is the dominant resistance factor to RTK inhibition. As far as we are aware, this is not only the largest series of breast cancers treated with trastuzumab to be analyzed for PTEN to date but also the first to show strong association of the level of PTEN expression with overall survival rather than simply time to progression (3). Our data strongly suggest that PI3K inhibition in PTEN-low tumors should be combined with RTK inhibitors, as recently shown for lapatinib invitro (41). Similar to quantitative estrogen receptor positivity in breast cancer (42), our data show that approximately a quarter of HER2+ patients may be spared ineffective, and potentially toxic, treatment if stratified by PTEN expression.

The PI3K pathway mediates key hallmarks of cancer, such as growth, proliferation, survival, motility, and angiogenesis (43). PI3K phosphorylates the lipid second messenger PI, which recruits AKT and PDK1 to the cell membrane. PTEN is a key negative regulator of the PI3K signaling pathway (29). PTEN mutations are responsible for the hereditary disease Cowden disease, partial loss-of-function mutations and loss of heterozygosity are common in the majority of cancers, and loss of expression by techniques such as immunohistochemistry or Western blotting is frequently seen. Although recent reports have suggested that PIK3CA mutations are important on a PTEN mutant background (44), and PTEN expression may cooperate with PIK3CA mutations in breast cancer and mediating trastuzumab resistance (3, 40), our data show that PTEN exerts dominant control in downstream pathway activation and resistance to RTK inhibitors. PIK3CA mutations exert a relatively small influence on PTEN:PI3K signaling cycle and resistance and would have to have a 5- to 10-fold increase in enzymatic activity to mediate resistance, which may explain the lack of effect of PIK3CA mutation on trastuzumab resistance seen herein. Previous reports implicating PIK3CA mutations in HER2 inhibitor resistance have only been in vitro or are very small cohorts of patients (3, 4). Other RTK inhibitors, such as pertuzumab, have also been shown to inhibit PI3K pathway activation (45, 46), suggesting that constitutive activation of this pathway through PTEN down-regulation may also contribute to pertuzumab resistance in vitro and in vivo. Given the quantitative nature of these phenomena, a mathematical description and systems analysis of the PI3K pathway has been useful for understanding therapeutic responses and allows better selection of patients expected to respond to RTK inhibitors.

In this study, we have successfully shown how a systems biology approach can generate hypotheses that can be tested experimentally in preclinical models and which can then be applied to clinical evaluation. Predictions from this model are consistent with known findings, have been extended within this study, and add weight to the use of PTEN as a biomarker for stratifying patients for HER2 inhibitor or combinatorial therapy, particularly a RTK inhibitor and PI3K inhibitor in cancers with low {gamma}. This model provides a foundation for further development and inclusion of more variables that may affect RTK inhibitor response. Network topology, kinetics, and quantities of molecules dictate cellular and ultimately clinical outcome. Systems biology approaches, particularly deterministic kinetic models based on experimental data, offer a new approach for integrating molecular pathology and computational modeling to more rationally interrogate cancer pathways and predict responses to therapy.


   Disclosure of Potential Conflicts of Interest
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
 Disclosure of Potential...
References
 
No potential conflicts of interest were disclosed.


   Acknowledgments
 
Grant support: Breakthrough Breast Cancer, Breast Cancer Campaign, Scottish Funding Council (Strategic Research Development Grant), and Biotechnology and Biological Sciences Research Council. Molecular pathology on tissue was supported by the Edinburgh CRUK Experimental Cancer Medicine Centre.

We thank A. Larionov for technical assistance, G. Kerr for help with statistics, and Minzi Ruan for help with MicroVigene RPPA analysis.


   Footnotes
 
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).

D. Faratian and A. Goltsov contributed equally to this work.

4 http://www.fda.gov/NewsEvents/Newsroom/PressAnnouncements/2006/ucm108587.htm Back

Received 2/28/09. Revised 5/11/09. Accepted 5/29/09.


   References
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
 Disclosure of Potential...
 References
 

  1. Piccart-Gebhart MJ, Procter M, Leyland-Jones B, et al. Trastuzumab after adjuvant chemotherapy in HER2-positive breast cancer. N Engl J Med 2005;353:1659–72.[Abstract/Free Full Text]
  2. Romond EH, Perez EA, Bryant J, et al. Trastuzumab plus adjuvant chemotherapy for operable HER2-positive breast cancer. N Engl J Med 2005;353:1673–84.[Abstract/Free Full Text]
  3. Berns K, Horlings HM, Hennessy BT, et al. A functional genetic approach identifies the PI3K pathway as a major determinant of trastuzumab resistance in breast cancer. Cancer Cell 2007;12:395–402.[CrossRef][Medline]
  4. Nagata Y, Lan KH, Zhou X, et al. PTEN activation contributes to tumor inhibition by trastuzumab, and loss of PTEN predicts trastuzumab resistance in patients. Cancer Cell 2004;6:117–27.[CrossRef][Medline]
  5. Faratian D, Moodie SL, Harrison DJ, Goryanin I. Dynamic computational modeling in the search for better breast cancer drug therapy. Pharmacogenomics 2007;8:1757–61.[CrossRef][Medline]
  6. Henney A, Superti-Furga G. A network solution. Nature 2008;455:730–1.[CrossRef][Medline]
  7. Belardinelli L, Shryock JC, Fraser H. Inhibition of the late sodium current as a potential cardioprotective principle: effects of the late sodium current inhibitor ranolazine. Heart 2006;92 Suppl 4:iv6–14.
  8. Noble D, Noble PJ. Late sodium current in the pathophysiology of cardiovascular disease: consequences of sodium-calcium overload. Heart 2006;92 Suppl 4:iv1–5.
  9. Cho CR, Labow M, Reinhardt M, van Oostrum J, Peitsch MC. The application of systems biology to drug discovery. Curr Opin Chem Biol 2006;10:294–302.[CrossRef][Medline]
  10. Fuss H, Dubitzky W, Downes CS, Kurth MJ. Mathematical models of cell cycle regulation. Brief Bioinform 2005;6:163–77.[Abstract/Free Full Text]
  11. Hatakeyama M, Kimura S, Naka T, et al. A computational model on the modulation of mitogen-activated protein kinase (MAPK) and Akt pathways in heregulin-induced ErbB signalling. Biochem J 2003;373:451–63.[CrossRef][Medline]
  12. Hendriks BS, Cook J, Burke JM, et al. Computational modelling of ErbB family phosphorylation dynamics in response to transforming growth factor {alpha} and heregulin indicates spatial compartmentation of phosphatase activity. Syst Biol Stevenage 2006;153:22–33.[Medline]
  13. Kholodenko BN, Demin OV, Moehren G, Hoek JB. Quantification of short term signaling by the epidermal growth factor receptor. J Biol Chem 1999;274:30169–81.[Abstract/Free Full Text]
  14. Markevich NI, Hoek JB, Kholodenko BN. Signaling switches and bistability arising from multisite phosphorylation in protein kinase cascades. J Cell Biol 2004;164:353–9.[Abstract/Free Full Text]
  15. Shankaran H, Wiley HS, Resat H. Modeling the effects of HER/ErbB1-3 coexpression on receptor dimerization and biological response. Biophys J 2006;90:3993–4009.[CrossRef][Medline]
  16. Steuer R. Computational approaches to the topology, stability and dynamics of metabolic networks. Phytochemistry 2007;68:2139–51.[CrossRef][Medline]
  17. Sorokin A, Paliy K, Selkov A, et al. The Pathway Editor: a tool for managing complex biological networks. IBM J Res Dev Sys Biol 2006;50:561–75.
  18. Goryanin I, Hodgman TC, Selkov E. Mathematical simulation and analysis of cellular metabolism and regulation. Bioinformatics 1999;15:749–58.[Abstract/Free Full Text]
  19. Spurrier B, Ramalingam S, Nishizuka S. Reverse-phase protein lysate microarrays for cell signaling analysis. Nat Protoc 2008;3:1796–808.[CrossRef][Medline]
  20. Board RE, Thelwell NJ, Ravetto PF, et al. Multiplexed assays for detection of mutations in PIK3CA. Clin Chem 2008;54:757–60.[Abstract/Free Full Text]
  21. Miyake T, Yoshino K, Enomoto T, et al. PIK3CA gene mutations and amplifications in uterine cancers, identified by methods that avoid confounding by PIK3CA pseudogene sequences. Cancer Lett 2008;261:120–6.[CrossRef][Medline]
  22. Kononen J, Bubendorf L, Kallioniemi A, et al. Tissue microarrays for high-throughput molecular profiling of tumor specimens. Nat Med 1998;4:844–7.[CrossRef][Medline]
  23. Camp RL, Chung GG, Rimm DL. Automated subcellular localization and quantification of protein expression in tissue microarrays. Nat Med 2002;8:1323–7.[CrossRef][Medline]
  24. Altman DG, Lausen B, Sauerbrei W, Schumacher M. Dangers of using "optimal" cutpoints in the evaluation of prognostic factors. J Natl Cancer Inst 1994;86:829–35.[Free Full Text]
  25. Camp RL, Dolled-Filhart M, Rimm DL. X-tile: a new bio-informatics tool for biomarker assessment and outcome-based cut-point optimization. Clin Cancer Res 2004;10:7252–9.[Abstract/Free Full Text]
  26. Birtwistle MR, Hatakeyama M, Yumoto N, et al. Ligand-dependent responses of the ErbB signaling network: experimental and modeling analyses. Mol Syst Biol 2007;3:144.[Medline]
  27. Moehren G, Markevich N, Demin O, et al. Temperature dependence of the epidermal growth factor receptor signaling network can be accounted for by a kinetic model. Biochemistry 2002;41:306–20.[CrossRef][Medline]
  28. Schoeberl B, Eichler-Jonsson C, Gilles ED, Muller G. Computational modeling of the dynamics of the MAP kinase cascade activated by surface and internalized EGF receptors. Nat Biotechnol 2002;20:370–5.[CrossRef][Medline]
  29. Leslie NR, Downes CP. PTEN function: how normal cells control it and tumour cells lose it. Biochem J 2004;382:1–11.[CrossRef][Medline]
  30. McConnachie G, Pass I, Walker SM, Downes CP. Interfacial kinetic analysis of the tumour suppressor phosphatase, PTEN: evidence for activation by anionic phospholipids. Biochem J 2003;371:947–55.[CrossRef][Medline]
  31. Carson JD, Van Aller G, Lehr R, et al. Effects of oncogenic p110{alpha} subunit mutations on the lipid kinase activity of phosphoinositide 3-kinase. Biochem J 2008;409:519–24.[CrossRef][Medline]
  32. Goldbeter A, Koshland DE, Jr. An amplified sensitivity arising from covalent modification in biological systems. Proc Natl Acad Sci U S A 1981;78:6840–4.[Abstract/Free Full Text]
  33. Stadtman ER, Chock PB. Superiority of interconvertible enzyme cascades in metabolic regulation: analysis of monocyclic systems. Proc Natl Acad Sci U S A 1977;74:2761–5.[Abstract/Free Full Text]
  34. Gomez-Uribe C, Verghese GC, Mirny LA. Operating regimes of signaling cycles: statics, dynamics, and noise filtering. PLoS Comput Biol 2007;3:e246.[CrossRef][Medline]
  35. Schmid AC, Byrne RD, Vilar R, Woscholski R. Bisperoxovanadium compounds are potent PTEN inhibitors. FEBS Lett 2004;566:35–8.[CrossRef][Medline]
  36. Posner BI, Faure R, Burgess JW, et al. Peroxovanadium compounds. A new class of potent phosphotyrosine phosphatase inhibitors which are insulin mimetics. J Biol Chem 1994;269:4596–604.[Abstract/Free Full Text]
  37. Wood LD, Parsons DW, Jones S, et al. The genomic landscapes of human breast and colorectal cancers. Science 2007;318:1108–13.[Abstract/Free Full Text]
  38. Shayesteh L, Lu Y, Kuo WL, et al. PIK3CA is implicated as an oncogene in ovarian cancer. Nat Genet 1999;21:99–102.[CrossRef][Medline]
  39. Levine DA, Bogomolniy F, Yee CJ, et al. Frequent mutation of the PIK3CA gene in ovarian and breast cancers. Clin Cancer Res 2005;11:2875–8.[Abstract/Free Full Text]
  40. Stemke-Hale K, Gonzalez-Angulo AM, Lluch A, et al. An integrative genomic and proteomic analysis of PIK3CA, PTEN, and AKT mutations in breast cancer. Cancer Res 2008;68:6084–91.[Abstract/Free Full Text]
  41. Eichhorn PJ, Gili M, Scaltriti M, et al. Phosphatidylinositol 3-kinase hyperactivation results in lapatinib resistance that is reversed by the mTOR/phosphatidylinositol 3-kinase inhibitor NVP-BEZ235. Cancer Res 2008;68:9221–30.[Abstract/Free Full Text]
  42. Hartwell L, Mankoff D, Paulovich A, Ramsey S, Swisher E. Cancer biomarkers: a systems approach. Nat Biotechnol 2006;24:905–8.[CrossRef][Medline]
  43. Hennessy BT, Smith DL, Ram PT, Lu Y, Mills GB. Exploiting the PI3K/AKT pathway for cancer drug discovery. Nat Rev Drug Discov 2005;4:988–1004.[CrossRef][Medline]
  44. Jia S, Liu Z, Zhang S, et al. Essential roles of PI(3)K-p110β in cell growth, metabolism and tumorigenesis. Nature 2008;454:776–9.[Medline]
  45. Agus DB, Gordon MS, Taylor C, et al. Phase I clinical study of pertuzumab, a novel HER dimerization inhibitor, in patients with advanced cancer. J Clin Oncol 2005;23:2534–43.[Abstract/Free Full Text]
  46. Mullen P, Cameron DA, Hasmann M, Smyth JF, Langdon SP. Sensitivity to pertuzumab (2C4) in ovarian cancer models: cross-talk with estrogen receptor signaling. Mol Cancer Ther 2007;6:93–100.[Abstract/Free Full Text]



This article has been cited by other articles:


Home page
 Mol Cancer ResHome page
Y. Nagumo, D. Faratian, P. Mullen, D. J. Harrison, M. Hasmann, and S. P. Langdon
Modulation of HER3 Is a Marker of Dynamic Cell Signaling in Ovarian Cancer: Implications for Pertuzumab Sensitivity
Mol. Cancer Res., September 1, 2009; 7(9): 1563 - 1571.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Supplementary Data
Right arrow All Versions of this Article:
0008-5472.CAN-09-0777v1
69/16/6713    most recent
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Google Scholar
Right arrow Articles by Faratian, D.
Right arrow Articles by Harrison, D. J.
PubMed
Right arrow PubMed Citation
Right arrow Articles by Faratian, D.
Right arrow Articles by Harrison, D. J.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
Cancer Research Clinical Cancer Research
Cancer Epidemiology Biomarkers & Prevention Molecular Cancer Therapeutics
Molecular Cancer Research Cancer Prevention Research
Cancer Prevention Journals Portal Cancer Reviews Online
Annual Meeting Education Book Meeting Abstracts Online