Pharmacogenomics in Lung Cancer: Predictive Biomarkers for Chemotherapy

Measurement of Molecular Biomarkers

Another crucial consideration pertinent to the use of biomarkers, whether prognostic, predictive, or panoramic, is how they are measured and what is defined as positive or negative. For some biomarkers, it can be clearly discerned whether the marker is present or absent in the tumor; classic examples of this type of biomarker are mutations of the EGFR gene or translocations of the echinoderm microtubule-associated protein-like 4 ( EML4 ) and ALK genes. Because the molecular aberrations can be clearly and unambiguously measured, the effect of these mutations on patient outcomes is clear.

However, most biomarkers are present on a continuum in almost all tumors, and, as such, variations in measurement techniques and interpretation of values are more likely. Typically, a lower expression of these biomarkers is considered negative and a higher expression is considered positive. The challenge is that the level of expression corresponding to positive or negative is often arbitrary and, for ease of interpretation, the cutoff point is often the statistical median. This approach artificially renders a continuous variable into a discrete one, which potentially confounds the strength of the association being measured. The strength of the association may be particularly vulnerable around the median. Investigators have attempted to partially offset this problem by dichotomizing the results in a particular cohort into quartiles and examining the association between a marker and an outcome by comparing the highest quartile with the lowest quartile.

The measurement technique is crucial to successful incorporation of biomarkers into clinical decision-making. Discrete biomarkers (i.e., those that are either present or absent) are best measured at the DNA level. Mutations, translocations, and copy number gains fall into this category. Mutations and translocations are best measured by sequencing the gene of interest, preferably in its entirety, and this technique will identify common and rare mutations, as well as mutations that are as yet unidentified. Multiplexed polymerase chain reaction (PCR)-based techniques will identify common mutations but will only detect mutations for which the primers are included in the multiplex panel.

Nondiscrete biomarkers are best measured at the RNA or protein level. As an example, most epithelial tissue expresses the EGFR protein, with the expression higher in some tissues than in others. Increased EGFR expression is not a consequence of an abnormality of the EGFR gene at the DNA level but is most likely a consequence of increased transcription of the normal EGFR gene to RNA and then eventual translation of the RNA to protein. Thus, the increased expression of a particular gene may be measured at the RNA or the protein level.

Measurement at the RNA or protein level each is associated with advantages and disadvantages. Measurement at the RNA level is more technically complex and thus could be more challenging to accomplish in the routine clinical setting. The expression of a gene is measured relative to the expression of a housekeeping gene and expressed as a unitless ratio. The values derived also depend on the use of specific standardization techniques and procedures, which can vary from laboratory to laboratory. Because of this potential variation, numerically similar values may not be congruent across different laboratories and platforms. Additionally, the cutoff values for high versus low must be individually established for each laboratory and validated by clinical data. Nevertheless, measuring RNA through quantitative PCR, if done with proper controls and standardization procedures, is precise, reproducible, and quantifiable. Hence, despite the technical difficulties, quantitative PCR has been the favored approach by several investigators. Controversy also exists as to the optimal sample for quantitative PCR. Most investigators consider a fresh frozen sample to be ideal, but it is not practical to obtain fresh frozen biopsy specimens in the clinical setting, especially from patients with advanced NSCLC. However, most investigators now believe that good-quality mRNA can be extracted from formalin-fixed paraffin-embedded (FFPE) tissue, and thus, such samples can be used for quantitative PCR. However, the process used to make the FFPE samples may potentially alter the message, which raises questions about the relevance of quantitative PCR measurements in FFPE samples.

Immunohistochemistry (IHC) is most commonly used to measure protein levels in clinical samples. IHC has several advantages, including relative ease of use, widespread availability in most clinical pathology laboratories, and the capability of evaluating FFPE samples. The performance characteristics of IHC, however, are critically dependent on having a good antibody that effectively binds only to the antigen of interest. Additionally, the intensity of the staining is arbitrarily graded as 0 through 3 (0 = no staining, 1= weak staining, 2 = moderate staining, and 3 = strong staining) or by the H score (the H score is a product of staining intensity and the percent of cells stained; for example, if 50% of the slides show an intensity of more than 3, 20% show an intensity of more than 2, and 30% are negative, then the H score would be 150 + 40 = 190). Despite these scoring methods and the use of rigorous (positive and negative) controls, these techniques can still lead to variation in interpretation. The definition of positive is also arbitrary and if, for example, 2+ or higher is considered positive, the arbitrariness between a score of 1 or 2 thus jeopardizes the very definition of positive versus negative.

To partially counteract the arbitrary nature of IHC grading methods, the automated quantitative analysis of in situ protein expression (AQUA) method was developed. The AQUA method involves the use of fluorescent microscopic technology that measures the expression of proteins of interest by quantifying the intensity of antibody-conjugated fluorophores within a specific cellular compartment (such as the nucleus or cytoplasm) in a tumor. A quantitative score is thus generated based on the intensity of immunofluorescence. The AQUA method thus provides a more continuous scoring of protein expression in tissue samples. Even though the AQUA method eliminates some of the subjectivity in the interpretation of IHC, the method is still associated with some of the same challenges as IHC. For example, AQUA also depends on an antibody that binds only to the protein of interest and the cutoff point to define positive and negative is arbitrary.

Driver Mutations in Lung Adenocarcinomas

The recognition of EGFR and ALK oncogenes as predictive biomarkers in lung cancer has led an ongoing investigation to identify additional oncogenic drivers with predictive and prognostic importance. Certain subsets of NSCLC can now be further defined at a molecular level by driver mutations in multiple oncogenes that lead to constitutive activation of mutant signaling proteins, causing induction and sustaining tumorigenesis. Mutations can be detected in all NSCLC histologies, including adenocarcinoma, squamous cell carcinoma, and large cell carcinoma, and in current-, former-, and never-smokers (defined as individuals who smoked fewer than 100 cigarettes in a lifetime).

It is estimated that at least 50% of patients with lung cancer have actionable driver mutations. Actionable driver mutations are defined as molecular abnormalities with downstream effects that initiate or maintain the neoplastic process, which can be negated by agents directed against each genomic alteration. Some of the evidence of the driver mutations of significance in lung cancer comes from research conducted by the 14-member Lung Cancer Mutation Consortium (LCMC), which has investigated metastatic lung adenocarcinomas since 2009 to identify and study driver genomic alterations. Between 2009 and 2012 more than 1000 patients underwent genotyping to determine the frequency of oncogenic drivers in lung cancer and demonstrate the practicality of using routine genetic analyses to inform treatment with targeted therapies.

In the LCMC patient cohort actionable driver mutations were found in 64% of tumors from patients with lung adenocarcinomas. Table 47.1 lists the driver mutations the LCMC investigators identified. Most of these driver mutations were found in a small percentage of patients. The most common driver mutations detected were in the EGFR, KRAS, and ALK genes.

The ALK fusion oncogene and sensitizing EGFR mutations have become accepted predictive biomarkers in lung cancer. The National Comprehensive Cancer Network (NCCN) recommends genotyping for EGFR mutations and ALK rearrangements in its algorithm for patients with metastatic disease.

Epidermal Growth Factor Receptor

EGFR (also known as HER1) is a transmembrane receptor for the epidermal growth factor with intrinsic tyrosine kinase activity. It is encoded by a gene located on chromosome 7. EGFR belongs to a family of receptor tyrosine kinases, which upon activation result in stimulation of multiple downstream pathways within the cell, including those involved in cell survival, proliferation, and resistance to apoptosis.

In normal cells the tyrosine kinase activity of the EGFR is strictly regulated, and therefore cell growth is controlled. Although the EGFR signaling cascades are complex, tyrosine kinase in the EGFR intracellular domain is the key factor that triggers signaling. If tyrosine kinase activity is blocked (i.e., via a molecular targeted agent), EGFR is unable to transduce signals to the cell nucleus. In cancer cells, various mechanisms of EGFR activation have been identified, including receptor overexpression, ligand overexpression, and EGFR gene amplification.