Story ovarian cancer

We are beginning to see the first signs of fundamental change in protein-based cancer diagnostics. The field has long been dominated by the search for highly predictive, single biomarkers that could reveal the state of the patient's health, much as hCG can indicate pregnancy or the detection of viral DNA or RNA can reveal the presence of certain pathogens. The difficulty of the search is complicated by the breadth and subtlety of the clinical questions that arise in the management of cancer patients: Does this asymptomatic patient have ovarian cancer? How aggressive is it? Are there distant metastases? What is the preferred treatment? Has the cancer recurred? Not surprisingly, the search for individual proteins with sufficient information content to answer these types of questions has generally been frustrating. Although immunohistochemical analysis of tumor proteins is common, and serum biomarkers are used in a number of cancers (Table 1), (1) performance is not always adequate, and significant unmet needs r emain, particularly in the areas of screening and post-treatment surveillance.

There is general agreement that earlier detection is preferable, since it is associated with improved survival (Figure 1). (2) Unfortunately, early detection is the exception for most cancers. For example, patients diagnosed with localized nonsmall-cell lung cancer (NSCLC) are approximately 20 times more likely to survive five years than patients whose disease is diagnosed with distant metastases (50.5 percent survival vs. 2.6 percent). Fewer than 20 percent of NSCLC cases are diagnosed at the localized stage. Certainly, there is the potential for lead-time bias, and some cancers that are diagnosed at later stages are very aggressive -- earlier detection will not always improve survival. Nevertheless, the magnitude of the survival benefit associated with earlier-stage diagnosis is striking and presents an extraordinary opportunity. Recognition of the potential impact of early detection has led to the establishment of the NCI's Early Detection Research Network (EDRN) (3,4) and a proposal for the appropriate de velopment processes for early-detection biomarkers. (5) The current debate over screening for breast and prostate cancers is really about the quality of the available screening methodologies, rather than the merits of screening per se.

For the patient who has been treated for cancer, the need for improved surveillance tools is profound. While tracking PSA is highly informative for the patient who has undergone prostatectomy (prostate cells should not be present in a such a patient), the tools are quite limited for other cancers. For women who have been treated for breast cancer, recommended surveillance and follow-up procedures include periodic histories and physical exams; mammography; and pelvic exams (for women taking tamoxifen who have not undergone hysterectomy). Routine measurement of biomarkers is not recommended. (1) Not surprisingly, patients and physicians turn to tools without proven value in the desperate hope of detecting recurrences or metastases earlier. One study found that approximately 62 percent of surveillance costs for breast cancer survivors were attributable to "excess testing" that exceeded recommendations and was unlikely to be effective. (6) A separate study found that only 22.6 percent of recurrences were detected at scheduled follow-up surveillance, and that symptoms were the primary indicator of relapse for 57.6 percent of cases. (7)

The single-marker approach has neglected the fact that tumors, particularly early in their development, differ from related normal cells in many subtle ways rather than in a few large ways. Moreover, tumors induce a variety of changes in their environment, including inflammation and angiogenesis. The challenge for next-generation cancer diagnostics is to measure a broad range of potential biomarkers and combine their information content to reach clinically meaningful conclusions. Genomics, protein microarrays and bioinformatics are providing the foundation for such an approach, and these technologies are driving a new approach to diagnostics.

Driver No.1: Genomics

The release of the first-drafts of the human genome (8,9) has drawn attention to the fact that current therapeutics and diagnostics address only a tiny proportion of the range of available targets. The two genomes (one provided by the Human Genome Project, the other by Celera) are said to. describe about 35,000 genes, though the total number is likely to be greater. Given that many proteins exist as stable isoforms that result from molecular events, such as alternative transcriptional initiation and termination, alternative translational initiation and termination, alternative mRNA splicing and post-translational modifications (e.g., phosphorylation, glycosylation, sulfation), the number of distinct potential protein analytes almost certainly exceeds 100,000. Obviously, we are today drawing on only a minimal proportion of the information that the organism makes available to us. For example, the Abbott AxSYM and JMx systems (Abbott Laboratories, Abbott Park, IL) offer five and six cancer-specific assays, respe ctively. (10)

Genetic analyses are already assuming an important role in clinical practice. Examples include the assessment of the Bcr-Abl oncogene in leukemias, (11,12) Her2/neu gene amplification in breast tumors, (13) and the analysis of CFTR mutations in cystic fibrosis. (14) In addition to these diagnostic and prognostic applications, genetic analyses can provide an indication of disease risk and provide an important input to clinical decision making (e.g., BRCA-1 and BRCA-2 mutations in the case of breast and ovarian cancers). Beyond these and other emerging applications, genomics is creating the potential for new diagnostic approaches by providing the "parts list" of proteins that might be usefully measured, as well as an ever-increasing sense (through gene expression studies) of proteins that might (or might not) go up or down with disease'5'6 (see Table 1, p.14).

Despite the value of genetic analyses and gene expression studies, in many clinical settings they face substantial limitations that protein-based tools can address. Particularly for screening and surveillance, gene expression analyses present the challenge of sample acquisition, since mRNA must be obtained from the cells of interest. Routine ovarian biopsies certainly would not be acceptable and would be unlikely to yield material from a comparatively small tumor. The problem is the same for post-treatment surveillance for recurrence. Beyond a few exceptions such as colorectal cancer (17) and hematological malignancies, the challenge of sample acquisition can be a significant barrier to effective screening and surveillance. Moreover, the measurement of proteins should be preferred in many cases, since these molecules are generally the effectors of physiology and pathophysiology and since the correlation between mRNA levels and the concentrations of the corresponding proteins is far from perfect. (18,19) In ad dition, tumor-associated necrosis and possibly apoptosis can be detected through protein analyses, but not genetic analyses. This is the concept behind Matritech's (Newton, MA) use of nuclear matrix proteins as markers for the detection of cancer, including an FDA-approved test for bladder cancer. (20) Finally, to the extent that many genetic analyses reflect disease risk, they cannot detect the actual onset of disease.

The genomic "parts list" will provide the foundation for recombinant protein expression and purification, and the subsequent development of specific "capture agents" such as antibodies, antibody mimics, (21) affibodies (22) and aptamers. (23,24) These capture agents will allow highly sensitive and specific measurements that, when combined in multiplexed formats, will provide unprecedented information and serve as the basis for an entire generation of new tests.

Driver No. 2: Multiplexed protein measurements

Tumors are accompanied by myriad subtle changes at the molecular level. The clinician, patient and laboratorian need tools for examining a large number of potentially relevant proteins with sufficient resolution to detect even minor changes reproducibly. Today, tools such as two-dimensional polyacrylamide gel electrophoresis with mass spectroscopy analysis (2-D PAGE-MS) meet the throughput needs embodied in the former criterion, while quantitative tools in routine laboratory use, such as ELISAs, meet the latter. Multiplexed protein measurements, defined here as the simultaneous quantitative measurement of 10 or more analytes (ultimately, hundreds or thousands), will combine the best qualities of both of these technologies.