January 1, 2011 (Vol. 31, No. 1)
Developing Effective Targeted Therapies Requires Elucidation of Multigene Signatures
It is no longer believed that dysregulation of limited amounts of oncogenes or tumor-suppressor genes causes cancer. Cancer cells show aberrant regulation of many other genes directly or indirectly involved in carcinogenesis and tumor progression. Therefore, observation of a single protein or gene is no longer meaningful by itself: it must be observed in the context of a molecular signature with multiple variants.
Only at the confluence of various molecular signatures can we begin to comprehend where the root cause of cancers may lie. Then we can start to design tailored drug therapies using active doses specifically directed at tumors with a defined molecular profile.
Currently, too many tumors are diagnosed at an advanced stage, when available cancer therapies are effective in only a minority of patients. An effective package of tumor markers would facilitate individualized treatment whether through early detection, better determination of risk or prognosis, or more rational selection of treatment.
The goal of CHI’s upcoming “Cancer Molecular Markers” conference is to showcase novel approaches to cancer diagnostics based on molecular signatures. Topics will include miRNA signatures, circulating tumor cells, cancer stem cells, and personalized cancer diagnostics, and the spectrum of research presented will run the gamut of genetic, epigenetic, proteomic, glycomic, and imaging biomarkers that can be used for cancer diagnosis, prognosis, detection, and prediction.
“The field of cancer biomarkers has particular challenges,” says Leif Ellisen, M.D., Ph.D., co-executive director, Massachusetts General Hospital (MGH) Cancer Center Translational Research Laboratory. “The field is still struggling with determining which analytes would provide the most straightforward information. And in some cases it may take years before we understand exactly how a new test will fit into medical practice.”
Mass General is taking the first steps toward providing its clinical practice with additional information that may influence the treatment decisions. “It is now widely accepted that tumor genetic testing is an important component of choosing an appropriate disease-management plan,” continues Dr. Ellisen.
“We provide our clinicians with molecular analysis of tumors for 130 mutations in 18 oncogenes. Physicians order the tests if they feel that the derived information can be potentially actionable.”
The Translational Research Laboratory at MGH utilizes an adaptation of Applied Biosystems’ SNaPshot® pyrosequencing kit to determine genetic alterations in regions that are known to cause abnormal activation. The genetic material is extracted from standard paraffin blocks using a completely automated robotic platform. Then the regions in question are amplified and analyzed in a multiplex format.
“We do not know all the oncogenes that exist in human cancer, but as cancer gene discovery efforts continue, we will continue incorporating relevant mutations into our panel,” says Dr. Ellisen. The MGH Cancer Center is part of a national collaborative effort called the Lung Cancer Mutation Consortium funded by the American Recovery and Reinvestment Act of 2009. Together with 12 other cancer centers across the country, Mass General collects linked clinical, pathologic, and genomic data on over 1,000 patients.
The MGH Cancer Center is also involved in several clinical trials with investigational drugs directed against abnormally activated proteins or pathways. Currently, only a few FDA-approved cancer therapies are based on a genetic profile. Nevertheless, information that Dr. Ellisen has collected has already teased out a few unexpectedly activated pathways, especially in tumors that became resistant to the first line of therapy. This information may serve as a basis for development of new genetically targeted drugs, and additional studies may identify molecular predictors of response to therapy.
Patient outcome could be improved by using targeted therapies based upon the molecular signature of the original tumor. However, establishing the tissue of origin is quite difficult for many metastatic malignancies. Even after extensive immunohistochemistry profiling, the primary site remains unidentified for 30–60% of cases.
“Metastatic undifferentiated tumors retain at least some of the molecular signatures of the parent tissues, but the search for such signatures using mRNA chips was not initially very successful,” comments Federico A. Monzon, M.D., medical director of molecular diagnostics at The Methodist Hospital in Houston.
“Using high-density microarray platforms for clinical diagnostics proved to be challenging due to high interlaboratory preanalytical variability and computational complexity arising from interpreting thousands of data points,” he says. “The Pathwork TOO test developed by Pathwork Diagnostics seems to address both issues. In our studies it demonstrated the ability to issue unambiguous calls for 15 tissue types. It also showed good reproducibility in blinded studies conducted by four independent labs.”
The test is based on interpretation of the expression patterns of 1,550 genes in comparison with the patterns established for selected tumor types such as bladder, breast, and colorectal. In each signature, a single gene is only predictive in combination with other genes, requiring pairwise comparisons between 1,550 genes on the chip.
For each specimen, the test reduces this highly complex expression data into 15 numerical similarity scores (0 to 100). In order to develop the scoring system based on this multiplex pattern, the whole chip was trained on a large set of 2,039 specimens. Dr. Monzon and collaborators further validated the test in a blinded multicenter study with 547 metastatic or poorly differentiated primary tumor specimens.
The validation study resulted in 87.8% accuracy of determination, with only 5% of specimens with indeterminate results. The Pathwork test has since been validated on paraffin-embedded tissues and approved by the FDA.
“The paradox of validation studies for molecular profiling of unknown primary tumors,” continues Dr. Monzon, “lies in the fact that we cannot use tumors of unknown origin. By definition, these tissues lack the clear origin that could be defined by other means and thus cannot be used as a standard to evaluate performance. Future studies, thus, are focusing on the demonstration that clinical management based on expression profiling can improve treatment outcomes for patients with cancer of unknown primary.”
Protein Biomarker Identification
Current diagnostic methods for lung cancer include CT or PET followed by biopsy and surgery. However, sporadically discovered small nodules in lungs are typically inflammatory in character. Therefore, the standard of care often involves simply waiting to see if they grow in size.
This strategy seems to run contrary to the fact that survival rates are considerably better when lung cancer is diagnosed at early stages. “A blood biomarker that is able to detect lung cancer at early stages would guide the treatment decision at the time when the patient’s chances for survival are high,” comments Laszlo Takacs, M.D., Ph.D., CSO of Biosystems International (BSI).
While many approaches include finding a blood biomarker by mass spectrometry based on differences between affected and healthy individuals, Biosystems International decided on a different strategy. Its approach is based on generating antibodies against the whole plasma proteome of lung cancer patients.
BSI prepares an immunogenic fraction of whole blood plasma by removing the most abundant proteins and then normalizing concentrations of the remaining proteins. The fraction is then injected into mice to generate nascent hybridomas.
After multiple rounds of screening, 3,000 hybridomas were narrowed down to 13 clones with good discriminatory power between patient and normal samples. “The preliminary data shows that our hybridomas recognize five proteins specific for lung cancer,” says Dr. Takacs.
“Moreover, some of these are cancer-specific isomers, also detected by immunohistology. We see significant potential for our diagnostic assay in guiding treatment decisions at the initial encounter of the stage I cancer patient with a primary-care physician or pulmonologist. Definitive management of the disease could be introduced earlier and in more specific ways.” The company is preparing for the first clinical trials in symptomatic patients with an ELISA-based diagnostic kit.
Single-Patient Comparisons
Blood-based protein biomarkers would make a significant difference in clinical practice. Even though proteomics is expected to play a leading role in clinical biomarker discovery, translation of MS-based assays into clinical diagnostic tests has been lagging.
The human proteome content varies significantly even within a single individual based on food intake, rest, and other physiological factors. Finding differences between blood proteomes of healthy individuals and cancer patients that could indicate the presence of cancer rather than merely a physiological bias has been problematic.
The strategy proposed by Josip Blonder, M.D., head of clinical proteomics at the National Cancer Institute, utilizes combined plasma/tissue proteome analysis from a single patient. This strategy relies on the assumption that salient proteins present in the tumor are detectable in blood. A three-way comparison of plasma, tumor, and normal tissue would help to narrow down the selection of relevant biomarker candidates.
Comparison of samples from the same patient minimizes the effect of normal biological variances. First, Dr. Blonder’s team identified tumor-specific proteins by discounting common proteins found in renal cell carcinoma (RCC) and adjacent normal tissue. Next, the tumor-specific proteins were compared with the plasma, looking for overlapping sets.
“This strategy is not devoid of typical proteomics deficiencies, such as the mismatch between the dynamic ranges of MS instrumentation and the human proteome. This issue is significantly alleviated, however, by the use of shotgun proteomics coupled with high-resolution and high-accuracy mass measurements,” says Dr. Blonder.
“Analysis resulted in the identification of 202 proteins that belong only to the tumor. This set was compared with 179 proteins from plasma, revealing 8 overlapping proteins, whose spectral count was higher in the tumor than in the plasma.” All eight proteins were specific to the tumor tissue, and four of these proteins were cross-validated in the blood of this patient and in the blood of four additional patients diagnosed with RCC.
“This pilot study was a proof of concept demonstrating that the comparative analysis of blood and tissue may result in the identification of tumor proteins in blood and may yield a realistic biomarker panel,” continues Dr. Blonder. “We hope this schema will be adopted and popularized by other laboratories.
“I do believe that current state-of-the-art clinical proteomics has matured enough for comprehensive and coordinated effort focused on creation of the cancer proteome atlas by relying on shotgun proteomics to sequence human cancers individually. The resulting atlas will be used as a complement to the cancer genome atlas and accelerate our understanding of the molecular basis of cancer.”
miRNA Signatures
miRNAs hold much promise as cancer biomarkers. These are small, noncoding RNAs that affect gene expression on the post-transcriptional level. Over 1,200 miRNAs are identified to date, and the expression profiles for many of these have been established.
“miRNA profiles, like protein or mRNA profiles, reflect the functional, physiological state of a cell, tissue, or tumor, says Bernard Andruss, Ph.D., director of collaborations and business development at Asuragen (www.asuragen.com). “A particularly valuable application for miRNAs will be to augment the information obtained from other classes of biomarkers. The combination of miRNA profiles with other marker classes has the potential to drive higher overall biomarker performance.”
miRNAs are known to regulate oncogenes and tumor suppressor genes. What makes them promising as biomarkers is that they are stable in a variety of biofluids (serum, saliva, cerebrospinal fluid) as well as in paraffin-fixed specimens, and their expression profile does not seem to be affected by normal physiological variations to the same extent as proteins.
“I envision that miRNAs will become valuable biomarkers for cancer diagnostics, prognostic for survival, or predictive of disease recurrence or response to therapy,” says Glen Weiss, M.D., associate investigator, Translational Genomics Research Institute and director, thoracic oncology, at Scottsdale Healthcare’s Virginia G. Piper Cancer Center.
“My research is focusing on using miRNAs as prognostic predictors of brain metastasis from lung cancer. By combining miRNA expression values with brain imaging and other validation techniques, our initial results show that we may be able to identify miRNAs that indicate increased probability of the future onset of brain metastasis,” says Dr. Weiss.
In addition to other statistical measurements, the researchers utilized an algorithm called in silico conditioning. The algorithm identifies a set of miRNAs with a consistent expression pattern when brain metastasis are present. “Some of the tumors were collected months before the metastasis developed in the brain,” adds Dr. Weiss. “Validating the miRNA biomarkers predictive of the metastasis could have high impact on treatment decisions.”