A new analysis approach of epidermal growth factor receptor pathway activation patterns provides insights into cetuximab resistance mechanisms in head and neck cancer
© von der Heyde and Beissbarth; licensee BioMed Central Ltd. 2012
Received: 11 April 2012
Accepted: 1 May 2012
Published: 1 May 2012
The pathways downstream of the epidermal growth factor receptor (EGFR) have often been implicated to play crucial roles in the development and progression of various cancer types. Different authors have proposed models in cell lines in which they study the modes of pathway activities after perturbation experiments. It is prudent to believe that a better understanding of these pathway activation patterns might lead to novel treatment concepts for cancer patients or at least allow a better stratification of patient collectives into different risk groups or into groups that might respond to different treatments. Traditionally, such analyses focused on the individual players of the pathways. More recently in the field of systems biology, a plethora of approaches that take a more holistic view on the signaling pathways and their downstream transcriptional targets has been developed. Fertig et al. have recently developed a new method to identify patterns and biological process activity from transcriptomics data, and they demonstrate the utility of this methodology to analyze gene expression activity downstream of the EGFR in head and neck squamous cell carcinoma to study cetuximab resistance. Please see related article: http://www.biomedcentral.com/1471-2164/13/160
KeywordsHNSCC EGFR cetuximab drug resistance matrix factorization GSEA pathway signature
Gene expression microarrays are a widely used tool to measure genomewide transcription within cell lines or tissues under varying conditions. Usually, gene-wise statistical tests, for example employing linear models, are then performed to determine differentially expressed genes . Methods to find overrepresentation of functional gene sets or pathway genes, so called gene set enrichment analysis (GSEA), are employed in order to interpret the resulting long lists of differential genes [10–12]. To monitor the activity of certain pathway parts or transcription factors (TFs), gene sets of TF target genes, as they can be retrieved from databases like TRANSFAC, are of special interest . Another aspect of data analysis is revealing gene expression patterns of patient or gene groups by clustering or dimension reduction techniques . A number of specialized methods have been proposed previously, for example, clustering genes and patients simultaneously into biclusters , applying predefined gene signatures in guided clustering approaches  or signal flow reconstruction in pathways from downstream effects of perturbation experiments .
Fertig et al. have proposed the new method Coordinated Gene Activity in Pattern Sets (CoGAPS)  and made it available as add-on for the popular free statistical computing software R . It combines a matrix factorization technique with GSEA of downstream transcriptional targets to determine patterns of pathway activity. They now demonstrate its utility to study cetuximab resistance in HNSCC by analyzing gene expression patterns downstream of EGFR .
Fertig et al. present a modeling approach of cetuximab resistance mechanisms applying the CoGAPS algorithm to infer gene expression signatures, distinguishing five variants of HaCaT cell lines under different media conditions concerning serum starvation and addition of EGF or TNF-α. These immortalized keratinocytes are chosen as model systems as they are well characterized and their genetic aberrations reflect early oncogenic events in HNSCC. The detected pathway signatures are then used to compare two isogenic HNSCC cell lines, that is, UMSCC1 and 1CC8, of which the latter is known to be cetuximab resistant in contrast to the sensitive UMSCC1 cell line.
The CoGAPS method
Analysis of EGFR downstream activation patterns on HNSCC data
The HaCaT variants include transfected cell types overexpressing EGFR, NF-kappa-B p65 subunit or mutant HRAS. Transcriptional targets of sub-pathways under investigation belong to STAT, AKT, RAS, Notch and TGF-β due to their implication in HNSCC. Applying CoGAPS to the HaCaT gene expression data reveals six patterns, which separate the samples well according to their experimental conditions. Thus, the patterns are attributed to baseline HaCaT activity, HaCaT-HRASVal12, HaCaT-vector control, HaCaT-EGFRWT, serum and HaCaT-p65WT. Afterwards, the activities of downstream transcriptional targets are calculated based on the Z-scores. This confirms upregulation of expected pathways but also indicates potential cross-talk mechanisms. The method is compared to a standard linear model approach with outcomes less consistent with prior knowledge. For example, CoGAPS reveals RAS and STAT overrepresentation for forced HRAS and EGFR expression in HaCaT cells and assigns Notch activity to the baseline pattern. Finally, the CoGAPS patterns are projected to the gene expression data of UMSCC1 and 1CC8 with and without cetuximab treatment. The most interesting finding here is that the pathway signature associated with HaCaT-HRASVal12 could predict the cetuximab treatment response, that is, treatment reduces the signature amplitude in sensitive UMSCC1, but not in resistant 1CC8. This is interpretable in such a way that cetuximab fails to repress the hyperactive RAS pathway in resistant HNSCC cell lines. A possible extension of this for the future would be to apply the learned signature to patient data and test whether it is likewise able to predict clinical parameters such as treatment response.
The main drawback of established techniques to infer activity of gene sets, clustering for example, is that they are neglecting multiple regulation of genes, that is, gene re-usage and co-regulation by diverse pathways and TFs as well as coordinated activity of gene sets, for example, pathway cross-talk, which actually constitutes a specific phenotype. To overcome this disadvantage, the CoGAPS algorithm focuses on gene sets instead of isolated genes for inferring biological processes based on transcriptional data. The multitude of computational methods and tools analyzing activity patterns of (interacting) pathways should be further developed and compared to each other in the future. The presented results indicate the potential of the CoGAPS algorithm to detect transcriptional signatures as biomarkers for individual drug sensitivity or resistance, respectively. These signatures will have to be tested and prove their value in clinical practice in the future.
SH is a research scientist focusing on network reconstruction from proteomics data and systems biology of the EGFR pathway in breast cancer. TB is an associate professor for statistical bioinformatics in the Department of Medical Statistics at the University Medical Center Göttingen. His main research focus is on the development of methods for the analysis and interpretation of high-throughput genomics data and on network reconstruction algorithms. He leads the multidisciplinary consortium BreastSys with systems biological analysis of the EGFR pathway as key aspect.
List of abbreviations
epidermal growth factor receptor
Coordinated Gene Activity in Pattern Sets
head and neck squamous cell carcinoma
gene set enrichment analysis
mitogen-activated protein kinase
We thank Dr. Stefan Wiemann, the handling deputy section editor for BMC Genomics, for pointing out this manuscript to us. We thank Frauke Henjes for inspiring the layout draft of Figure 1.
- Hynes NE, Lane HA: ERBB receptors and cancer: the complexity of targeted inhibitors. Nat Rev Cancer. 2005, 5 (5): 341-354.View ArticlePubMedGoogle Scholar
- Sahin O, Fröhlich H, Löbke C, Korf U, Burmester S, Majety M, Mattern J, Schupp I, Chaouiya C, Thieffry D, Poustka A, Wiemann S, Beissbarth T, Arlt D: Modeling ERBB receptor-regulated G1/S transition to find novel targets for de novo trastuzumab resistance. BMC Syst Biol. 2009, 3: 1.PubMed CentralView ArticlePubMedGoogle Scholar
- Lai SY, Johnson FM: Defining the role of the JAK-STAT pathway in head and neck and thoracic malignancies: implications for future therapeutic approaches. Drug Resist Updat. 2010, 13 (3): 67-78.View ArticlePubMedGoogle Scholar
- Jones RB, Gordus A, Krall JA, MacBeath G: A quantitative protein interaction network for the ErbB receptors using protein microarrays. Nature. 2006, 439 (7073): 168-174.View ArticlePubMedGoogle Scholar
- Vermorken JB, Mesia R, Rivera F, Remenar E, Kawecki A, Rottey S, Erfan J, Zabolotnyy D, Kienzer HR, Cupissol D, Peyrade F, Benasso M, Vynnychenko I, Raucourt DD, Bokemeyer C, Schueler A, Amellal N, Hitt R: Platinum-based chemotherapy plus cetuximab in head and neck cancer. N Engl J Med. 2008, 359 (11): 1116-1127.View ArticlePubMedGoogle Scholar
- Uribe P, Gonzalez S: Epidermal growth factor receptor (EGFR) and squamous cell carcinoma of the skin: molecular bases for EGFR-targeted therapy. Pathol Res Pract. 2011, 207 (6): 337-342.View ArticlePubMedGoogle Scholar
- Hatakeyama H, Cheng H, Wirth P, Counsell A, Marcrom SR, Wood CB, Pohlmann PR, Gilbert J, Murphy B, Yarbrough WG, Wheeler DL, Harari PM, Guo Y, Shyr Y, Slebos RJ, Chung CH: Regulation of heparin-binding EGF-like growth factor by miR-212 and acquired cetuximab-resistance in head and neck squamous cell carcinoma. PLoS One. 2010, 5 (9): e12702.PubMed CentralView ArticlePubMedGoogle Scholar
- Freudlsperger C, Burnett JR, Friedman JA, Kannabiran VR, Chen Z, Waes CV: EGFR-PI3K-AKT-mTOR signaling in head and neck squamous cell carcinomas: attractive targets for molecular-oriented therapy. Expert Opin Ther Targets. 2011, 15: 63-74.PubMed CentralView ArticlePubMedGoogle Scholar
- Smyth GK: Linear models and empirical bayes methods for assessing differential expression in microarray experiments. Stat Appl Genet Mol Biol. 2004, 3: Article3.PubMedGoogle Scholar
- Beissbarth T: Interpreting experimental results using gene ontologies. Methods Enzymol. 2006, 411: 340-352.View ArticlePubMedGoogle Scholar
- Wu D, Lim E, Vaillant F, Asselin-Labat ML, Visvader JE, Smyth GK: ROAST: rotation gene set tests for complex microarray experiments. Bioinformatics. 2010, 26 (17): 2176-2182.PubMed CentralView ArticlePubMedGoogle Scholar
- Bauer S, Robinson PN, Gagneur J: Model-based gene set analysis for Bioconductor. Bioinformatics. 2011, 27 (13): 1882-1883.PubMed CentralView ArticlePubMedGoogle Scholar
- Wingender E: The TRANSFAC project as an example of framework technology that supports the analysis of genomic regulation. Brief Bioinform. 2008, 9 (4): 326-332.View ArticlePubMedGoogle Scholar
- Li L: Dimension reduction for high-dimensional data. Methods Mol Biol. 2010, 620: 417-434.View ArticlePubMedGoogle Scholar
- Sill M, Kaiser S, Benner A, Kopp-Schneider A: Robust biclustering by sparse singular value decomposition incorporating stability selection. Bioinformatics. 2011, 27 (15): 2089-2097.View ArticlePubMedGoogle Scholar
- Maneck M, Schrader A, Kube D, Spang R: Genomic data integration using guided clustering. Bioinformatics. 2011, 27 (16): 2231-2238.View ArticlePubMedGoogle Scholar
- Fröhlich H, Tresch A, Beissbarth T: Nested effects models for learning signaling networks from perturbation data. Biom J. 2009, 51 (2): 304-323.View ArticlePubMedGoogle Scholar
- Fertig EJ, Ding J, Favorov AV, Parmigiani G, Ochs MF: CoGAPS: an R/C++ package to identify patterns and biological process activity in transcriptomic data. Bioinformatics. 2010, 26 (21): 2792-2793.PubMed CentralView ArticlePubMedGoogle Scholar
- R Development Core Team: R: A Language and Environment for Statistical Computing. 2011, Vienna: R Foundation for Statistical ComputingGoogle Scholar
- Fertig EJ, Ren Q, Cheng H, Hatakeyama H, Dicker A, Rodeck U, Considine M, Ochs MF, Chung CH: Gene expression signatures modulated by epidermal growth factor 1 receptor activation and their relationship to cetuximab resistance in head and neck squamous cell carcinoma. BMC Genomics. 2012, 13: 160.PubMed CentralView ArticlePubMedGoogle Scholar
- The pre-publication history for this paper can be accessed here:http://www.biomedcentral.com/1741-7015/10/43/prepub