By Malorye Allison

Breast cancer is considered one of biomarker research’s biggest successes, largely because of Roche/Genentech/Chugai’s trastuzumab (Herceptin)—a breakthrough therapy that has helped at least a subset of these patients to live much longer. But that specific example also highlights how complex the challenge is. Last week at Cambridge Healthtech Institute’s Beyond Genome meeting, the search for biomarkers to guide breast cancer diagnosis and treatment was one of the major topics.

Beyond HER2Neu

Despite the widespread availability of trastuzumab and the HER2Neu test needed to prescribe it, only about 10% of all breast cancer patients actually benefit from this drug. That's because only about a quarter of women with this disease have elevated levels of HER2Neu. Of those who do test positive, only about 40% actually respond to the treatment, according to speaker Michael Liebman, executive director of the Windber Research Institute in Windber, Pennsylvania.

Liebman and others worry that many women who could benefit are just not getting the drug. Women must have a positive HER2Neu test result to receive trastuzumab, and that test has some serious limitations: Data from groups such as Windber suggests that the two most common tests used to determine HER2Neu status—one using FISH, the other IHC—often give conflicting results. This led Liebman and colleagues to wonder, “Are we measuring the right functional form of HER2Neu?”

The group decided to start by studying a set of samples that came from patients with conflicting results from FISH and IHC HER2Neu testing. Working with the ExPlain software tool from BIOBASE, the scientists looked for additional factors, beyond those test results, that might help them make better decisions about who should receive the drug. The software let them look at the interacting proteins in the entire signaling network, to find those proteins and genes that represent key elements in the pathways.

“BIOBASE’s ExPlain helped us to look upstream,” Liebman said. “We already found something that we would not have found with genomics and proteomics alone.” The group has filed a patent application on at least one marker, which they hope can be used as an adjunct to HER2Neu testing for selecting patients for trastuzumab therapy. They are now also “Starting to couple the ability to do this type of analysis with modeling of pathways.” In collaboration with BIOBASE, the Windber scientists are building a model of how steroid synthesis and metabolism occurs across a woman’s lifetime. This model will help them mine these interactions even more effectively.

Clinical Data

In addition, the Windber group now has enough samples from patients who have responded well to trastuzumab to evaluate those samples for biomarkers too. “We’re looking for causal biomarkers—markers that are actually linked to the mechanism of disease,” Liebman said. Ideally, the markers will also be available from serum.

Windber is an ideal place for such studies, since the Institute already has more than 25,000 samples from breast tumors alone, which have been carefully characterized. The Institute also boasts an unusual wealth of clinical data, since many of the patients are from military families and all now have electronic health records. “We have an enormous amount of patient data collected at the personalized health level,” Liebman explained.

To do such biomarker mining successfully, Liebman advised researchers to look at any patient’s condition as “a series of temporal elements.” Diseases like breast cancer usually occur in the context of many risks, and the patient is likely to have other conditions as well. Windber researchers use software from InforSense to select patient cohorts using temporal characteristics as well as the usual features. They can also model this data to explore possible interactions with other disorders.

“This is particularly important when looking at the interrelationships of diseases,” Liebman said. Breast cancer, for example, may be preceded by inflammatory breast disease. Understanding how these conditions relate biologically will help scientists find more relevant, and more powerful, biomarkers for this disease.

Leads from Functional Proteomics

Proteomics is also beginning to deliver some intriguing breast cancer marker leads. Harvard Institute of Proteomics, for example, is working on high-throughput functional proteomics: Rather than identifying proteomic signatures that are associated with disease, these scientists are looking for high-throughput ways to uncover how proteins act in key types of cells, such as the breast epithelium.

To do that, they have been building a unique set of tools, and the Institute’s “dream is now more of a reality,” said Joshua LaBaer, director, Institute of Proteomics, Harvard Medical School. The group has built an automated storage system to store clones, each with a 2D bar code, and all uniquely associated with a specific gene in their Oracle database. The researchers can go to a computer, dial up the gene of interest, and then those clones will be automatically retrieved from storage. The genes are all stored in inert vector systems from Invitrogen (Gateway) and Clontech (Creator).

Harvard Institute of Proteomics researchers have also developed a nucleic acid programmable protein array (NAPA). Instead of attaching proteins to a slide, the researchers developed a system using DNA mini preps through which the proteins are expressed in situ when needed. “The minute they are expressed, the proteins are captured by an epitope tag on the surface of the array,” LaBaer said. “Most have a GST moiety—so if you see the protein it is full length.”

According to LaBaer, this system is both easier and more effective than earlier types of protein arrays, where proteins are expressed and then attached to the slides and stored for later use. “One thing we worried about with protein arrays is what the shelf life would be,” he said. “With these, we don’t have to worry about how long they have been sitting around.” In a validation of the system, the researchers created an array which contained proteins from all the genes known to be involved in human DNA replication. “Of the known interactions in this system, we detected 85%,” La Baer said. “But we also found 65 novel ones.”

The group is now expanding the use of these arrays into biomarker studies. “We are looking for proteins that will indicate presence or absence of disease,” LaBaer said. One study is looking at autoantibodies in the serum of breast cancer patients. While many patients will have autoantibodies to key antigens such as P53, others will not. Using such an array, scientists might be able to find more diagnostic autoantibodies. “The thinking in proteomics now is that we should be testing panels of proteins as markers,” LaBaer said.

An initial study shows the system is specific. Antibodies to common proteins, such as Epstein Barr Virus, are used as controls. Possible new markers are already being discovered from serum samples using this unique, high-throughput proteomics platform. “This NAPA micro Elisa approach works the same as a standard ELISA,” LaBaer said.