AN INFORMATIVE DIAGNOSIS OF PROSTATE AND OVARIAN CANCERS THROUGH SERUM ANALYSIS
The systems approach to biology and medicine promises to transform the practice of oncology over the next decade by moving it from a reactive discipline (responding after the patient is sick with advanced cancer) to a predictive, preventive and personalize modes. This will be, in part, achieved by using the blood as a window into health and disease. The realization of this goal constitutes the heart of this project. The blood bathes all of the major organs, and so proteins that are secreted by those organs are entrained in the blood. However, exactly because the blood contains so many proteins, cells, etc., it is also a very noisy environment. Thus, a first task of this project is to utilize systems biology approaches to identify an validate those proteins that contain the most information – i.e. those proteins that can help answer the following questions:
- Do I have cancer?
- If so, what organ is it in?
- Is it early stage or late stage?
- What is the appropriate therapy or combination of therapies?
- If I am on therapy, then what are my positive and adverse responses?
The need to answer many questions implies the need to do many measurements, and so it shouldn’t be surprising that it may take between 10 and 30 measurements to answer all of these questions for one particular cancer. To understand how we approach the problem of determining what to measure, it is important to first understand how we are using systems biology techniques to develop a picture of a given cancer.
The central idea is that biology is mediated by networks of proteins and other molecules that operate to execute the normal functions of life. Consider the figure below. This image shows a network of 17 proteins – a full network would include many thousand proteins, but this smaller snapshot captures the essence of what we are doing. Although this is a fictional network (used here for educational purposes), it bears many key similarities to the real networks that we have begun to elucidate for disease such as prion disease, prostate cancer, and ovarian cancer. Using systems biology measurement and computational techniques, such networks are constructed so proteins that work together to regulate, for example, some physiological process such as respiration, are grouped together. Two such ‘protein modules’ are encircled in the left-most image.
For cancer (and other disease), one or more of these networks becomes perturbed through either genetics (i.e. mother-to-daughter transfer of genetic information) or through environment (i.e. smoking, dietary habits, etc.) It is these altered networks that mediate the disease and as the disease progresses, those networks become increasingly perturbed. Some changes are detectable long before clinical symptoms appear, while others only appear during the late stages of disease.
It turns out that, at a given stage of disease progression, approximately 3000 of the potentially 30,000 expressed genes are up or down-regulated! Thus, cancer (or any other disease, for that matter) induces a very significant physiological response. Sorting through that response to identify the most informative diagnostic markers can be very challenging! The unique approach we have taken involves quantitatively measuring every single gene that is expressed, organ by organ, for healthy tissues. This involves on the order of 100 million measurements! Once we have all those measurements in hand, we can, through computational methods, compare the gene expression profiles for one organ against another, and we find that a certain number of genes are only expressed in particular organs, such as the prostate or the lungs or the ovaries. This reduces our number of 3000 differentially expressed genes to a much smaller number – about 200-300 differentially expressed, organ-specific genes. The graph below shows the measured expression level of one particular gene, organ by organ. As is clear from the graph, this gene is expressed only in the prostate.
We then further analyze, again using computational methods, those few hundred organ-specific genes to identify which ones can possibly encode for secreted proteins. That again reduces the numbers of differentially expressed genes to about 50 per organ. At this point, we must prepare the actual protein, prepare a capture agent against that protein, and then validate whether that protein can serve as a good blood-based indicator of disease. We do this initially by measuring the levels of the particular protein in a series of serum samples that are stored in National Cancer Institute sponsored serum and tissue banks.
Once we have complete network models constructed for various cancers, we will also be able to identify the specific physiological or metabolic processes that are associated with each of the particular proteins we are measuring. By measuring sufficient numbers of proteins (probably 10-30 per cancer) we hope to be able to eventually piece together a fairly detailed picture of the health state of the organ from which the proteins are secreted –all through blood analysis!
For both prostate and ovarian cancers, we have identified many organ-specific, secreted transcripts, and we have begun to validate the use of the secreted proteins for blood based diagnostics. The preliminary results so far are very promising, and indicate that we will be able to answer at least some, if not all, of the diagnostic questions posed above for specific cancers. One of the NSBCC corporate partners, Homestead Clinical Corporation (a Seattle company founded by Hood and Heath and others) is currently validating organ-specific, secreted proteins using the above described methods. As those proteins are validated, that knowledge will be shared with the NSBCC scientists and clinicians, and we will begin including those proteins within our diagnostics platforms.
As stated above, it will likely require 10-30 measurements to capture one cancer through serum analysis, and it may require on the order of 1000 measurements to capture all cancers. There currently 2 protein measurement methods – antibodies and proteomics – and neither one will scale to the level that is required here. Thus, we are developing a series of nanotechnologies, microfluidics technologies, and chemical technologies that will allow us to scale to such large numbers of measurements, to do such measurements routinely and inexpensively, and to do them all out of a fingerprick of blood.
TECHNOLOGIES FOR REALIZING AN INFORMATIVE DIAGNOSIS OF CANCER THROUGH MULTIPARAMETER SERUM ANALYSIS
To meet the requirements that emerge in the drive towards systems medicine, we are working toward developing chip platforms that include on the order of 103 or more nanoelectronic sensors, each encoded for the highly sensitive and specific detection of genes (mRNAs) or proteins of interest across a broad dynamic concentration range (106). Microfluidic platforms that can integrate these nanotechnologies together with tissue and blood handling capabilities are also being developed. There are a number of non-trivial and fundamentally important scientific and engineering bottlenecks that must be solved for the full development of these platforms, and they include:
1) Fabricate 1000’s of highly reproducible protein and gene sensors on a chip in a scaleable and cost-effective manner. Challenges associated with the preparation of high-affinity antibodies and other protein-capture agents require these sensors should be label-free (i.e. one probe molecule for one target molecule) and quantitative, and this necessarily implies that the readout should be electronic. Furthermore, many biologically important molecules are present in low abundance. For example, 68% of the 18,000 different types of mRNA transcripts in a prostate cancer cell line are present at 10 copies or less per cell—and many of these transcripts encode important molecules in signal transduction pathways and gene regulatory networks. Thus, being able to detect mRNAs and proteins at low-concentrations (and over broad dynamic ranges) is key for systems approaches to biology and medicine. While there are multiple biosensor technologies, only semiconductor nanowire (NW) sensors thus far meet the requirements for sensitive detection, dynamic range, and scaling. However, NW sensors must have diameters within the range of 10-20 nanometers for optimized sensitivity, those diameters must be controlled to within 1-2 nanometers for reproducibility, and they must have excellent and controllable conductivity properties. Finally, they must operate in biologically relevant (e.g. serum) environments. The required NW dimensions imply that alternatives to traditional semiconductor patterning methods are required that can produce and reproduce nearly atomically perfect and similar semiconductor NWs time and time again. Finally, most examples of nanoelectronic technologies have consisted of single or few-device demonstrations. Manufacturing approaches for producing nanoelectronic technologies with near atomic perfection, at molecular-type dimensions, and doing so reliably and inexpensively, are only beginning to emerge.
A related set of problems are the chemical issues surrounding the nanosensor/biology interface. In order to do many measurements from small amounts of blood or tissues, the nanosensor circuits must be patterned at a high density. This means that biofunctionalization methods such as spotting or photolithography, because they can not achieve required densities of information spacing, will probably not be useful for encoding the individual nanosensor elements with the appropriate molecular probes.
2) Integrate the nanochip with microfluidics sample delivery system (that can ultimately supply multiple gene and protein detector arrays simultaneously). Multiple types of microfluidics systems have been developed for at least small number gene and protein assays, but the requirements that those microfluidics systems be integrated with the nanoelectronic devices, the biological systems of interest, and the chemistry that serves as the interface between the nanosensors and the biology, presents fundamentally new materials and surface chemistry challenges. In general, integration of new technologies with each other is often more difficult than the development of the individual technologies themselves. NSBCC partners, Materia Inc. and Siemens Biomarker Solutions, are working with Caltech scientists to develop new materials that can solve many of these problems.
3) Develop new approaches for rapidly and inexpensively preparing high affinity protein-capture agents. High affinity antibodies against protein biomarkers are very expensive and time-consuming to prepare, often costing a more than $100,000 and taking several years. This implies that antibodies cannot serve as the long-term solution for 1000s of protein-capture agents, and that alternatives must be developed. As described in Project 3, NSBCC researchers are utilizing chemical concepts known as template-driven synthesis approaches for the preparation of high-affinity protein capture agents. For these approaches, the protein itself actually catalyzes the formation of the high affinity capture agent.
Significant progress in all of these areas has been made, although most of those results are either in press or undergoing scientific review prior to publication. As that work becomes publicly available, we will update this site with those results.