BRPI0815543B1 - IN VITRO PROCESS FOR PREDICTING THE RESPONSE OF A PATIENT WITH NON-SMALL CELL LUNG CANCER TO TREATMENT WITH ERLOTINIB, AND IN VITRO USES OF A PTPRF GENE AND ERLOTINIB - Google Patents

Abstract

PREDICTIVE MARKER FOR EGFR INHIBITOR TREATMENT. The present invention provides a biomarker that is capable of predicting the clinical benefit of EGFR inhibitor treatment in cancer patients.

Description

The present invention relates to a biomarker that is capable of predicting the clinical benefit of treatment with an EGFR inhibitor in cancer patients.

Several human malignancies are associated with aberrant expression or overexpression of the epidermal growth factor receptor (EGFR). EGF, transforming growth factor-? (TGF-?) and several other ligands bind to EGFR, stimulating autophosphorylation of the receptor's intracellular tyrosine kinase domain. A variety of intracellular pathways are subsequently activated and these downstream events result in tumor cell proliferation in vitro. It has been postulated that simulation of tumor cells through EGFR may be important for both tumor growth and tumor survival in vivo.

Initial clinical data with Tarceva®, an EGFR tyrosine kinase inhibitor, indicate that the compound is safe and generally well tolerated at doses that provide the targeted effective concentration (which is determined by preclinical data). Phase I and II clinical trials in patients with advanced disease have demonstrated that Tarceva® has promising clinical activity in a range of epithelial tumours. In fact, Tarceva® has been shown to be able to induce durable partial remissions in patients previously treated with head and neck cancer and NSCLC (Non-Small Cell Lung Cancer) of a similar order to established second-line chemotherapy, but with the added benefit of a better safety profile than chemotherapy and greater convenience (pill rather than intravenous [i.v.] administration). A recently completed randomized double-blind placebo-controlled trial (BR.21) has shown that the agent Tarceva® alone significantly prolongs and increases survival in patients with NSCLC for whom standard therapy for advanced disease has failed.

Tarceva® (erlotinib) is a small chemical molecule; is an orally active potent selective EGFR tyrosine kinase inhibitor (EGFR-TKI).

Lung cancer is the leading cause of cancer-related death in North America and Europe. In the United States, the number of deaths secondary to lung cancer exceeds the combined total deaths from the second (colon), third (breast), and fourth (prostate) leading causes of cancer deaths combined. Approximately 75% to 80% of all lung cancers are NSCLC, with approximately 40% of patients having advanced disease locally and/or that cannot be surgically removed. This group typically includes those with bulky disease in stages IIIA and IIIB, excluding malignant pleural effusions.

The crude incidence of lung cancer in the European Union is 52.5, the mortality rate 48.7 cases/100,000/year. Among men the rates are 79.3 and 78.3, among women 21.6 and 20.5, respectively. NSCLC accounts for 80% of all lung cancer cases. Approximately 90% of mortality caused by lung cancer among men and 80% among women is attributed to smoking.

In the US, according to the American Cancer Society, during 2004, there were approximately 173,800 new cases of lung cancer (93,100 in men and 80,700 in women) and accounted for approximately 13% of all new cancers. Most patients die as a result of their illness within two years of diagnosis. For many NSCLC patients, successful treatment remains elusive. Advanced tumors are often not amenable to surgery and may also be resistant to tolerable doses of radiotherapy and chemotherapy. In randomized trials, the most active combination chemotherapies currently have achieved response rates of approximately 30% to 40% and a 1-year survival rate of between 35% and 40%. This is really an improvement over the 10% 1-year survival rate seen with supportive care alone.

Until recently, therapeutic options for relapsing patients after relapse were limited to better supportive or palliative care. A recent trial comparing docetaxel (Taxotere) with better supportive care showed that patients with NSCLC could benefit from second-line chemotherapy after first-line cisplatin-based regimens have failed. Patients of all ages and with an ECOG performance condition of 0, 1 or 2 demonstrated greater survival with docetaxel than those who had been refractory to prior platinum-based treatment. Patients who did not benefit from therapy included those with 10% weight loss, high lactate dehydrogenase levels, multi-organ involvement, or liver involvement. Additionally, the benefit of docetaxel monotherapy did not extend beyond the second-line adjustment. Patients who received docetaxel as a third-line treatment or beyond did not exhibit prolonged survival. Docetaxel as a single agent has become a standardized second-line therapy for NSCLC. Recently, another randomized phase III trial in second-line NSCLC therapy compared pemetrexed (Alimta®) with docetaxel. Treatment with pemetrexed resulted in clinically equivalent efficacy, but with significantly fewer side effects compared to docetaxel.

It has long been recognized that there is a need to develop individualized cancer treatment methods. With the development of targeted cancer treatments, there is particular interest in methodologies that could provide a molecular profile of the target tumor, (ie, one that can predict clinical benefit). A proof-of-principle has already been established for profiling gene expression in cancer with molecular classification of tumor types that are not evident based on current morphologic and immunohistochemical testing. Two separate disease entities have been identified with different prognoses from the single current classification of diffuse B-cell lymphoma using gene expression profiling.

Therefore, it is an objective of the present invention to provide expression biomarkers that can predict the clinical benefit of EGFR inhibitor treatment in cancer patients.

In a first objective, the present invention provides an in vitro method of predicting the clinical benefit of a patient with cancer in response to treatment with an EGFR inhibitor comprising the steps: of determining an expression level of a PTPRF gene in a tumor sample from a patient and comparing the PTPRF gene expression level with a representative value of a PTPRF gene expression level in tumors from a population of patients who do not derive clinical benefit from treatment, where an expression level higher concentration of the PTPRF gene in the patient's tumor sample is indicative of a patient who will derive clinical benefit from the treatment.

The abbreviation PTPRF stands for receptor-type protein tyrosine phosphatase, F. Seq. ID No. 1 shows the nucleotide sequence of human PTPRF, transcription variation 1 and Seq. ID No. 2 shows the nucleotide sequence of the human PTPRF transcript variant 2.

The term "a value representative of a PTPRF expression level in tumors from a population of patients who do not derive clinical benefit from treatment" refers to an estimate of the mean level of expression of the PTPRF gene in a population of patients who do not derive a clinical benefit from the treatment. Clinical benefit was defined as having an objective response or disease stabilization for >12 weeks.

In a further preferred embodiment, the PTPRF gene exhibits an expression level between 1.1 and 1.8, preferably 1.1 and 1.6 or more times greater in the patient's tumor sample compared to a representative patient population value. who do not derive clinical benefit from treatment.

In a further preferred embodiment, the PTPRF gene exhibits an expression level between 1.2 and 1.8 or more times greater in the patient's tumor sample compared to a representative population of patients who do not derive clinical benefit from the treatment.

In a preferred embodiment, the expression level of the marker gene is determined through microarray technology or other technologies that assess RNA expression levels such as quantitative RT-PCR or through any method that considers the expression level of the respective protein. , for example, immunohistochemistry (IHC). The construction and use of gene chips is well known in the art, see U.S. Pat. N—5,202,231; 5,445,934; 5,525,464; 5,695,940; 5,744,305; 5,795,716 and 15,800,992. See also, Johnston, M. Curr. Biol. 8:R171-174 (1998); Iyer VR et al., Science 283:83-87 (1999). Of course, the level of gene expression can be determined by other methods which are known to a person skilled in the art such as, for example, northern blots, RT-PCR, real-time quantitative PCR, primer extension, RNase protection, obtaining of RNA expression profiling.

The marker gene of the present invention can be combined with other biomarkers in biomarker sets. Biomarker sets can be constructed from any combination of predictive biomarkers to make predictions regarding the effect of EGFR inhibitor treatment in cancer patients. The biomarkers and biomarker sets described here can be used, for example, to predict how cancer patients will respond to therapeutic intervention with an EGFR inhibitor.

The term "gene", as used herein, comprises variations of the gene. The term "variation" refers to nucleic acid sequences that are substantially similar to the nucleic acid sequences provided by the GenBank accession number. The term "substantially similar" is well understood by one skilled in the art. In particular, a gene variation can be an allele that exhibits nucleotide changes when compared to the nucleic acid sequence of the most prevalent allele in the human population. Preferably, such substantially similar nucleic acid sequence has a sequence similarity to the most prevalent allele of at least 80%, preferably at least 85%, more preferably at least 90%, most preferably at least 95%. The term "variants" is also understood to refer to splice variants.

The EGFR inhibitor may be selected from the group consisting of gefitinib, erlotinib, PKI-166, EKB-569, GW2016, CI-1033 and an anti-erbB antibody such as trastuzumab and cetuximab.

In another embodiment, the EGFR inhibitor is erlotinib.

In yet another embodiment, the cancer is NSCLC.

Techniques for detecting and quantifying gene expression of the genes described by this invention include, but are not limited to, northern blots, RT-PCR, real-time quantitative PCR, primer extension, RNase protection, gene profiling, RNA expression and related techniques. These techniques are well known to those skilled in the art see, for example, Sambrook J et al., Molecular Cloning: A Laboratory Manual, Third Edition (Cold Spring Harbor Press, Cold Spring Harbor, 2000).

Techniques for detecting protein expression of the respective genes described by this invention include, but are not limited to, immunohistochemistry (IHC).

According to the invention, cells from a patient tissue sample, for example a tumor or cancer biopsy, can be analyzed to determine the expression pattern of one or more biomarkers. The success or failure of a cancer treatment can be determined on the basis of the biomarker expression pattern of cells derived from the test tissue (test cells), e.g. tumor or cancer biopsy, as being relatively similar or different from the expression pattern of a control set of one or more biomarkers. In the context of this invention, it was observed that the gene in table 3 is up-regulated, i.e., exhibits a higher expression level, in tumors from patients who obtained clinical benefit from treatment with an EGFR inhibitor compared to tumors from patients who did not. obtained clinical benefit from treatment with an EGFR inhibitor. Thus, if the test cells exhibit a biomarker expression profile that matches that of a patient who has responded to cancer treatment, it is highly likely or predicted that the individual's cancer or tumor will respond favorably to treatment with the EGFR inhibitor. . In contrast, if the test cells exhibit a biomarker expression pattern that matches that of a patient who has not responded to cancer treatment, it is highly likely or predicted that the individual's cancer or tumor will not respond to treatment with the EGFR.

The biomarker of the present invention, i.e. the gene listed in Table 3, is a first step towards an individualized therapy for cancer patients, in particular, patients with refractory NSCLC. This individualized therapy will allow treating physicians to select the most appropriate agent from existing drugs for cancer therapy, in particular, NSCLC. The benefits of individualized therapy for each future patient are: response rates/number of patients benefiting will increase and the risk of adverse side effects caused by ineffective treatment will be reduced.

In a further object, the present invention provides a therapeutic method of treating a cancer patient identified by the in vitro method of the present invention. Said therapeutic method comprises administering an EGFR inhibitor to the patient who has been selected for treatment based on the expression pattern of the predictive gene in Table 3. A preferred EGFR inhibitor is erlotinib and a preferred cancer to be treated is NSCLC.

Brief Description of Figures

Figure 1 shows the study design;

Figure 2 shows the sample processing scheme;

Figure 3a shows PTPRF expression levels versus clinical outcome for Genechip® profiling;

Figure 3b shows PTPRF expression levels versus clinical outcome for qRT-PCR and

Figure 3c shows the correlation between Genechip® and qRT-PCR measurements for PTPRF.

Experimental part Rationale for Study and Study Planning

Recently, mutations within the EGFR gene in the tumor tissue of a subset of NSCLC patients and the association of these mutations with erlotinib and gefitinib sensitivity have been described (Pao W, et al. 2004; Lynch et al. 2004; Paez et al. 2004). For the combined patients from the two studies, mutated EGFR was seen in 13 of 14 patients who responded to gefitinib and in none of the 11 gefitinib-treated patients who did not respond. The reported prevalence of these mutations was 8% (2 of 25) in unselected NSCLC patients. These mutations were seen most frequently in adenocarcinomas (21%), in tumors from women (20%) and in tumors from Japanese patients (26%). These mutations result in increased in vitro EGFR activity and increased sensitivity to gefitinib. The relationship of mutations to prolonged stable disease or duration of survival has not been prospectively evaluated.

Based on exploratory analyzes from the BR.21 study, it seemed unlikely that the observed survival advantage was due to EGFR mutations alone, as a significant survival advantage is maintained even when patients with an objective response are excluded from the analyzes (data in the file). Other molecular mechanisms must also contribute to the effect.

Based on the assumption that there are changes in gene expression levels that can predict a response/benefit to Tarceva® treatment, microarray analysis was used to detect these changes.

This required a clearly defined study population treated with Tarceva® monotherapy after failure of 1st line therapy. Based on the experience of the BR.21 study, the beneficiary population was defined as having objective response or disease stabilization over 12 weeks. Clinical and microarray datasets were analyzed according to a predefined statistical plan.

The application of this technique requires fresh frozen tissue (FFT). Therefore a mandatory biopsy had to be performed before starting treatment. The collected material was frozen in liquid nitrogen (N2).

A second tumor sample was collected at the same time and stored in paraffin (embedded in formalin-fixed paraffin, FF-PE). This sample was analyzed for changes in the EGFR signaling pathway.

The ability to perform tumor biopsies via bronchoscopy was a prerequisite for this study. Bronchoscopy is a standardized procedure to confirm the diagnosis of lung cancer. While generally safe, there is a remaining risk of complications, for example bleeding.

This study was a first step towards individualized therapy for patients with refractory NSCLC. This individualized therapy will allow treating physicians to select the most appropriate agent from existing drugs for this indication.

Once individualized therapy is available, the benefit for each future patient will outweigh the risk patients have to take in the present study: • response rates/number of patients benefiting will increase, • the risk of Adverse side effects due to ineffective treatment will be reduced.

Rationale for Dosage Selection

Tarceva® was given orally once daily at a dose of 150 mg until disease progression, intolerable toxicities, or death. Selection of this dose was based on pharmacokinetic parameters, as well as the safety profile and tolerability of this dose observed in Phase I, II, and III trials in heavily pretreated patients with advanced cancer. The observed drug levels in the plasma of cancer patients receiving the 150 mg/day dose were consistently above the mean plasma concentration of 500 ng/mL targeted for clinical efficacy. BR.21 exhibited a survival advantage at this dose.

Study Objectives

The primary objective was the identification of differentially expressed genes that may predict the benefit (CR, PR or SD ? 12 weeks) of treatment with Tarceva®. Identification of differentially expressed genes that can predict "response" (CR, PR) to Tarceva® treatment was an important additional objective.

Secondary objectives were to assess changes in EGFR signaling pathways in relation to benefit from treatment.

Study Planning

Overview of Study Design and Dosing Regimen

This was an open-label Phase II study of identifying markers that can make a prediction. The study was conducted at approximately 26 sites in approximately 12 countries. 264 patients with advanced NSCLC after failure of at least one previous chemotherapy regimen were enrolled over a 12-month period. Continuous oral Tarceva® was provided at a dose of 150 mg/day. Dose reductions were allowed based on ability to tolerate drug therapy. Clinical and laboratory parameters were analyzed to assess disease control and toxicity. Treatment continued until disease progression, unacceptable toxicity, or death. The study planning is represented in figure 1.

Tumor tissue and blood samples were obtained for molecular analysis to assess the effects of Tarceva® and to identify subgroups of patients who benefited from the therapy;

Forecast Marker Evaluations

Tumor biopsies were performed within a 2-week period prior to initiation of treatment. Two different samples were collected:

The first sample was always immediately frozen in liquid N2.

The second sample was fixed in formalin and embedded in paraffin.

Flash-frozen tissue had the highest priority in this study.

Figure 2 shows a schematic of sample processing.

Microarray Analysis

The snap-frozen samples were used for laser capture microdissection (LCM) of tumor cells to extract tumor RNA and RNA from tissue surrounding the tumor. RNA was analyzed on Affymetrix microarray chips (HG-U133A) to profile the patients' tumor gene expression. Quality Control of the Affymetrix chips was used to select samples of adequate quality for statistical comparison.

Analyzes of Isolated Biomarkers in Formalin-Fixed Paraffin Embedded Tissue

The second tumor biopsy, the FFPE sample, was used to perform DNA mutation, IHC and ISH analysis as described below. Similar analyzes were performed on tissues collected at the initial diagnosis.

The DNA mutation status of genes encoding EGFR and other molecules involved in the EGFR signaling pathway was analyzed by DNA sequencing. Gene amplification of EGFR and related genes was studied by FISH.

Protein expression analyzes included immunohistochemical [IHC] analyzes of EGFR and other proteins within the EGFR signaling pathway.

Response Ratings

RECIST (Unidimensional Tumor Measurement) criteria were used to assess response. These criteria can be found at the following link: http://www.eortc.be/recist/

Note that:

To determine a CR or PR status, changes in tumor measurements must be confirmed by repeat assessments at least 4 weeks apart at any time during the treatment period.

In the case of DS, the follow-up measures have to have satisfied the criteria for DS at least once after entry into the study at a minimum interval of 6 weeks.

In the case of sustained SD, follow-up measures have to have met the criteria for SD at least once after entry into the study with a maintenance duration of at least 12 weeks.

Survival Assessment

A regular status check every 3 months was performed either through the patient's visit to the clinic or by telephone. All deaths were registered. At the end of the study, a definitive confirmation of survival was required for each patient.

Methods Preparation of RNA Samples and Quality Control of RNA Samples

All processing of biopsy specimens was performed by a pathology reference laboratory; Fresh frozen tissue samples were shipped from the research sites to the Clinical Sample Operations facility at Roche Basel and from there to the pathology laboratory for further processing. Laser capture microdissection was used to select tumor cells from surrounding tissue. After LCM, RNA was purified from enriched tumor material. The pathology laboratory then performed a number of steps to obtain an estimate of RNA concentration and quality.

RNases are RNA-degrading enzymes and are found everywhere and thus all procedures in which RNA will be used have to be strictly controlled to minimize RNA degradation. Most mRNA species by themselves have particularly short half-lives and are thus considered quite unstable. It is therefore important to perform RNA integrity checks and quantification prior to any assay.

The RNA concentration and quality profile can be evaluated using an instrument from Agilent (Agilent Technologies, Inc., Palo Alto, CA) called a 2100 Bioanalyzer®. The instrument's software generates an RNA integrity number (RIN), an estimate of the quantification (Schroeder, A. et al., The RIN: an RNA integrity number for assigning integrity values to RNA measurements. BMC Mol Biol, 2006. 7: p. 3) and calculates ribosomal proportions of the total RNA sample. The RIN is determined from the entire electrophoretic trace of the RNA sample and thus includes the presence and absence of degradation products.

RNA quality was analyzed by a 2100 Bioanalyzer®. Only samples with at least one rRNA peak above the added poly-l noise and with sufficient RNA were selected for further analysis on the Affymetrix platform. Purified RNA was sent to the Roche Center for Medical Genomics (RCMG; Basel, Switzerland) for microarray analysis. 122 RNA samples were received from the pathology laboratory for further processing.

Targeted Labeling of tissue RNA samples

Targeted labeling was performed according to the Affymetrix Two-Cycle Target Labeling Amplification Protocol (Affymetrix, Santa Clara, California), following the manufacturer's instructions.

The method is based on the Eberwine standard linear amplification procedure, but uses two cycles of this procedure to generate sufficient cRNA for hybridization, but uses two cycles of this procedure to generate sufficient labeled cRNA for hybridization to a microarray.

The total RNA input used in the labeling reaction was 10 ng for those samples where more than 10 ng of RNA was available; if less than this amount was available or if there was no amount data available (due to very low RNA concentration), half of the total sample was used in the reaction. Yields from labeling reactions ranged from 20-180 µg of cRNA. A normalization step was introduced at the hybridization level where 15 µg of cRNA was used for each sample.

Human Reference RNA (Stratagene, Carlsbad, CA, USA) was used as a control sample in the workflow with each batch of samples. 10 ng of this RNA was used as an input along with the test samples to verify that the labeling and hybridization reagents were working as expected.

Hybridizations in Microarrays

Affymetrix HG-U133A microarrays contain over 22,000 sets of probes that target approximately 18,400 transcription products and variations representing approximately 14,500 well-characterized genes.

Hybridization for all samples was performed according to the Affymetrix instructions (Affymetrix Inc., Expression Analysis Technical Manual, 2004). Briefly, for each sample, 15 μg of biotin-labelled cRNA was fragmented in the presence of divalent cations and heat and hybridized overnight to the Affymetrix HG-U133A genome-wide oligonucleotide arrays. The next day, the arrays were stained with streptavidin-phycoerythrin (Molecular Probes; Eugene, OR) according to the manufacturer's instructions. The arrays were then scanned using a GeneChip Scanner 3000 (Affymetrix) and the signal intensities were calculated automatically by GeneChip Operating Software (GCOS) Version 1.4 (Affymetrix).

Statistical analysis

Affymetrix® data analysis consisted of five main steps.

Step 1 consisted of quality control. The goal was to identify and exclude data with a substandard quality profile from this analysis arrangement.

Step 2 consisted of pre-processing and normalization. The goal was to create a normalized and graded "analysis dataset" sensitive to interchip comparison. This included estimating and subtracting background noise, compacting and classifying the probes.

Step 3 consisted of exploration and description. The goal was to identify the potential trend and sources of variability. It consisted of applying multivariate and univariate descriptive analysis techniques to identify covariates that exerted influence.

Step 4 consisted of modeling and testing. The goal was to identify a list of candidate markers based on statistical evaluation of the difference in mean expression level between patients with "clinical benefit" and "no clinical benefit". It consisted of fitting a suitable statistical model for each set of probes and deriving a measure of statistical significance.

Step 5 consisted of a strength analysis. The goal was to generate a qualified list of candidate markers that do not rely heavily on preprocessing methods and statistical assumptions. It consisted of repeating the analysis with different methodological approaches and crossing the list of candidates.

All analyzes were performed using the R program package.

Step 1: Quality Control

The assessment of data quality was based on checking several parameters. These included the Affymetrix GeneChip® quality parameters, in particular: Scale Factor, Percentage of Present Request and Average Background. This step also included visual inspection of virtual chip images to detect localized hybridization problems and comparison of each chip to an intermediate virtual chip to detect any deviation from the median behavior. Interchip correlation analysis was also performed to detect outliers. In addition, ancillary measures of RNA quality obtained from analyzing RNA samples with the Agilent Bioanalyzer® 2100 were taken into account.

Based on these parameters, data from 20 arrays were excluded from the analysis. Thus, data from a total of 102 arrays representing 102 patients were included in the analysis. The clinical description of this set of 102 samples is reported in Table 1.

Table 1: Description of clinical characteristics of patients included in the analysis

Step 2: Data pre-processing and normalization

The rma algorithm (Irizarry, R.A. et al., Summaries of Affymetrix GeneChip probe level data. Nucl. Acids Res., 2003. 31(4): p. e15) was used for preprocessing and normalization. The mas5 algorithm (AFFYMETRIX, GeneChip® Expression: Data Analysis Fundamentals. 2004, AFFYMETRIX) was used to detect requests against individual probe sets. Probe sets called "missing" or "marginal" in all samples were removed from further analysis; 5930 probe sets were removed from the analysis based on this criterion. The analysis dataset therefore consisted of a matrix with 16353 (out of 22283) probe sets measured in 102 patients.

Step 3: Data Description and Exploration

Descriptive exploratory analysis was performed to identify potential trends and major sources of variability. A set of covariates with a potential impact on gene expression profiles was verified. This comprised both technical and clinical variables. Technical covariates included: RNA processing date (later referred to as batch), RIN (as a measure of RNA quality/integrity), Operator, and Sample Collection Center. Clinical covariates included: Histology type, smoking status, tumor grade, performance score (Oken, M.M. et al., Toxicity and response criteria of the Eastern Cooperative Oncology Group. Am J Clin Oncol, 1982. 5(6): p. 649-55), demographic data, responsiveness condition, and clinical benefit condition.

Analysis tools included univariate ANOVA and principal component analysis. For each of these covariates, univariate ANOVA was applied independently to each set of probes.

A significant effect of the batch variable was identified. In practice, the batch variable captured differences between sample processing dates and the Affymetrix chip batch. After verifying that the batch variable was almost independent of the variables of interest, the batch effect was corrected using the method described in Johnson, W.E., C. Li and A. Rabinovic, Adjusting batch effects in microarray expression data using empirical Bayes methods. Biostat, 2007. 8(1): p. 118- 127.

The normalized dataset after correcting for batch effects served as the analysis dataset in subsequent analyses.

Histology and RIN were two additional important variables 5 highlighted by the descriptive analysis.

Step 4: Modeling and Testing the Data

A linear model was independently fitted to each set of probes. The variables included in the model are reported in Table 2. The model parameters were estimated using the maximum probability 10 technique. The parameter corresponding to the variable "Clinical Benefit" (X1) was used to assess the difference in expression level between the group of patients with clinical benefit and the group without clinical benefit.

Table 2: Description of variables included in the linear model.

For each set of probes i, the purpose of the statistical test was to reject the hypothesis that the mean expression levels in patients with clinical benefit and in patients without clinical benefit were equivalent, taking into account the other adjustment covariates listed in Table 2. Formally, the null hypothesis of equality was tested against a two-sided alternative. Under the null hypothesis, the t-statistic distribution for this test follows a Student's t-distribution with 92 degrees of freedom. Corresponding p values are reported in Table 3.

The choice of the linear model was motivated by two reasons. First, linear modeling is a versatile, well-characterized, and robust approach that allows adjustment of confounding variables when estimating the effect of the variable of interest. Second, given the sample size of 102 and normalization and scaling of the data set, the assumption of normal distribution was reasonable and justified.

For each set of probes, the hypothesis of homogeneity of variance was evaluated using Fligner-Killeen tests based on model residuals. The analysis consisted of 3 steps: 1. Testing each of the categorical variables for residual variance homogeneity 2. Observing the variable V with the minimum p-value 3. If the minimum p-value is less than 0.001, readjust the model that allows the different level of V variables to obtain a different variance.

Step 5: Strength

The goal of power analysis was to reduce the risk that the results of the analysis could come from artifacts and a result of pre-processing steps or hypotheses underpinning the statistical analysis. The following three aspects were considered: a) inclusion or exclusion of some extra chips in the quality control step; b) pre-processing and normalization algorithm; c) statistical hypotheses and test approach.

The list of candidate markers was defined as the subset of genes consistently declared significant with different analysis settings. The different analysis options applied were as follows: a) An additional subset of 8 chips was identified based on stricter quality control criteria. A "reduced dataset" was defined by excluding these 8 chips. b) MAS5 was identified as an alternative to rma for pre-processing and normalization. MAS5 uses different methods for background estimation, probe summarization and normalization. c) Two additional statistical tests were employed. The. A Wilcoxon test for the difference between clinical benefit and no benefit and b. a likelihood ratio test (LRT) testing the logistic regression model in which clinical benefit was taken as the response variable and gene expression as the covariate. These two additional tests are based on a different set of supporting statistical hypotheses. For each set of probes, the LRT was performed after a Chi-square with 1 degree of freedom.

In summary, two sample sets (the "full" data set and the "reduced" data set) and 2 pre-processing algorithms (mas5 and rma) were considered; this resulted in four different analysis datasets. For each of these four data sets, three different statistical tests were applied. Therefore, for each set of probes, three p values were calculated. In each analysis dataset, a composite criterion was applied to identify the list of differentially regulated genes. This composite criterion was defined as: the maximum p-value is less than 0.05 and the minimum p-value is less than 0.001. Strength analysis using criterion 1 for identifying marker genes provided PTPRF as a predictor marker for EGFR inhibitor treatment.

Table 3: Gene Marker for Clinical Benefit Based on Strength Analysis after Application of Composite Criteria.

Column 1 is the Affymetrix identifier of the probeset. Column 2 is the GenBank accession number of the corresponding gene sequence. Column 3 is the corresponding official gene name. Column 4 is the corresponding adjusted multiplication mean change in expression level between patients with clinical benefit and without clinical benefit, which is estimated from the linear model. Column 5 is the p-value for the test for the difference in expression level between patients with clinical benefit and without clinical benefit that is derived from the linear model. Column 6 is the 95% confidence interval for the change in adjusted mean multiplications at the expression level.

Additional Statistical Analysis

For the selected candidate marker PTPRF, the following additional analyzes were performed in an environment validated by independent statisticians: • Univariate Cox Regression for PFS (Survival Without Progression) starting from the Affymetrix Primary Analysis, • Univariate Logistic Regression for Clinical Benefit starting from from the Primary Affymetrix Analysis and • Univariate Cox Regression for Survival from the Primary Affymetrix Analysis 15

The results of these analyzes are presented below. These are consistent with the results of the primary analysis and confirm the choice of the selected marker.

Results: Univariate Cox Regression for PFS (Progression Free Survival) from the Affymetrix Primary Analysis:

Results: Univariate Cox Regression for Clinical Benefit from the Primary Affymetrix Analysis:

Results: Univariate Cox Regression for Survival from the Primary Affymetrix Analysis:

qRT-PCR

The cDNA was synthesized using SuperScript® III First-strand Synthesis SuperMix for qRT-PCR (Invitrogen, CA, USA) according to the manufacturer's instructions, but without the inclusion of an RNase H digestion product.

Quantitative PCR was performed using TaqMan® Gene Expression Assays on an ABI PRISM® 7900HT Sequence Detection System according to the manufacturer's recommendations (Applied Biosystems, CA, USA). All assays were performed in triplicates.

The primers and probes used crossed exon boundaries or were within the sequence of interest of the Affymetrix Genechip® probe. Two house-keeping genes were included as endogenous controls: beta-2-microglobulin (B2M-, Assay Hs99999907_m1) and hypoxanthine phosphoribosyl transferase (HPRT; Assay Hs99999909_m1).

All runs included a calibration sample (MVP® total RNA from adult human lung; Stratagene, CA, USA) and a standard curve. Universal Human Reference total RNA (Stratagene, CA, USA) was used as a template for PTPRF standard curves. All samples were measured in triplicates.

Relative quantification was performed using the - 5 ΔCt method.

Results

As previously reported, Affymetrix Genechip® gene expression profiles were determined for 102 patients enrolled in this study. Among these patients, qRT-PCR results were obtained for 75 (Table 4). The demographic and clinical characteristics of patients with qRT-PCR results were similar to those of the entire population (n=264) and of patients with available Genechip® gene expression profiles.

Table 4: Baseline characteristics: patients with qRT-15 PCR analysis (n=75)

Of the 75 patients with qRT-PCR results, 4 (5%) had a partial response (PR), 23 (31%) had DS, 39 (52%) had PD, and 9 (12%) could not be evaluated. These results were very similar to those observed in the entire study population (n=264).

Figure 3 shows relative mRNA levels for PTPRF in individual patients, which are assessed by profiling with Affymetrix Genechip® and qRT-PCR. Figure 3a shows expression levels versus clinical outcome for Genechip® profiling and Figure 3b shows expression levels for qRT-PCR.

There was a good relationship between the Genechip® and qRT-PCR measurements of the PTPRF transcript (Figure 3c; Pearson p=0.76, p<0.01). As seen with Genechip® profiling, PTPRF mRNA levels assessed using qRT-PCR appeared to correlate with response to erlotinib, with higher levels being seen in responders compared to responders.

Discussion

By analyzing tissue samples with high-density oligonucleotide microarray technology and applying statistical modeling to the data, it was possible to identify genes whose expression levels may be able to predict patients who derive a clinical benefit from treatment with erlotinib.

A composite criterion (defined above) was applied. This resulted in PTPRF as a predictive marker for EGFR inhibitor treatment.

The PTPRF gene, located on chromosome 1p34, encodes a protein member of the protein tyrosine phosphatase (PTP) family. This has an extracellular region, a single transmembrane region and two intracytoplasmic catalytic domains in tandem and thus represents a receptor-type PTP. The extracellular region contains three Ig-like domains and nine non-Ig-like domains, similar to those of neural cell adhesion molecules.

In this study, it was observed that PTPRF was relatively up-regulated in patients who derived a clinical benefit from erlotinib treatment. This finding can be interpreted in the context of published reports demonstrating the potential role of this gene in important mechanisms other than tumorigenesis.

First, it has been clearly established that EGFR is a substrate of PTPRF. In a detailed investigation, the interaction between EGFR and PTPRF was further characterized and shown to be complex and tightly controlled. These observations led the present inventors to postulate that PTPRF plays an important and direct role in controlling signaling downstream of the EGFR receptor. In another line of evidence, it was observed that PTPRF has tumor suppressor activity, acting through an inhibitory effect on cell migration and possibly inducing apoptosis. The mechanism through which PTPRF controls the cell migration process was additionally obtained. Two studies have shown that this protein works through a complex interaction with the E-cadherin complex, mediated by a direct regulation of beta-catenin activity.

Direct interaction with EGFR and well-characterized tumor suppressor activity are two prominent features that make PTPRF a particularly convincing marker of erlotinib response.

Claims (6)

1. In vitro process for predicting the response of a patient with non-small cell lung cancer (NSCLC) to treatment with erlotinib, characterized in that it comprises: determining an expression level of a PTPRF gene in a sample of patient's tumor and comparing the PTPRF gene expression level to a value representative of a PTPRF gene expression level in tumors from a population of patients who do not derive clinical benefit from treatment, where a higher level of PTPRF expression PTPRF gene in the patient's tumor sample is indicative of a patient who will derive clinical benefit from treatment, where clinical benefit was defined as having an objective response or disease stabilization for 12 weeks or longer. 2. Process according to claim 1, characterized by the fact that the expression level is determined through microarray technology. 3. Process according to claim 1 or 2, characterized in that the PTPRF gene exhibits an expression level between 1.1 and 1.8, preferably between 1.1 and 1.6 or more times greater in the sample of patient tumor compared to a representative value of a PTPRF gene expression level in tumors from a population of patients who do not derive clinical benefit from treatment. 4. Process according to any one of claims 1 to 3, characterized in that the PTPRF gene exhibits an expression level between 1.2 and 1.8 times or more in the patient's tumor sample compared to a representative value population of patients who do not derive clinical benefit from treatment. 5. In vitro use of a PTPRF gene, characterized in that it is to predict the response of a patient with NSCLC to treatment with erlotinib. 6. Use of erlotinib, characterized by the fact that it is in the preparation of a drug for the treatment of a patient with NSCLC identified through a process, as defined in any one of claims 1 to 4.