A team of Yale University researchers has used a protein microarray method to identify candidate protein biomarkers for ovarian cancer in a study they say supports the use of such a platform for discovery research.
When it comes to discovering and identifying candidate protein biomarkers, mass spectrometry is the dominant platform in proteomics, overshadowing the use of protein microarrays. But in a study published in the Oct. 30 edition of the Proceedings of the National Academies of Sciences, the authors say that “protein microarrays provide a powerful approach to identify proteins aberrantly expressed in disease states.”
While the identification of the proteins as candidate biomarkers may eventually prove clinically significant, the study’s lead author Michael Snyder, a professor of molecular, cellular, and developmental biology at Yale, told ProteoMonitor this week that, more importantly, “we can now use this type of approach, a microarray-plus-autoantibody platform, as a way to try to find things of interest.”
Currently, CA-125 is the best known test for detecting ovarian cancer, but it has not proven effective for early stages of the disease, necessitating the need for other tests, Snyder and his colleagues say in the study.
Several other strategies have already been used in recent years to identify markers for early detection, including mass spectrometry and gene-expression profiling.
“However, proteins identified by mass spectrometry have low reproducibility, and it remains unclear whether preferential expression of genes is reflected at the protein level,” they say.
Using serum autoantibodies is another method for protein discovery, but the identification of reactive proteins has proven difficult, they say. Similarly, the serological analysis of proteins expressed in E. coli from cDNA libararies has resulted in the discovery of several hundred autoantigens, including tumor-associated autoantigens found in ovarian cancer. The method, however is “limited to the interrogation of linear epitopes and the expression of some epitopes in bacteria is difficult,” Snyder and his colleagues say.
Their work is still in its early stages and further work is being done. But Synder said the study suggests that protein microarrays could be a robust platform for not only ovarian cancer markers, but cancer markers in general.
Digging for a Cancer Fingerprint
For their study, Snyder and his colleagues used a protein microarray with 5,005 proteins to incubate sera from 30 patients with ovarian cancer and 30 age-matched controls. A fluorescently labeled secondary antibody was used to detect autoantibodies that bound to each protein spot, and a control microarray was probed with only the fluorescently labeled anti-human IgG antibody.
Of the 5,005 original proteins, 1,845 were bound by autoantibodies in patients with cancer, while 1,441 were bound by autoantibodies in controls. Of that total of 3,286 binding proteins, 730 reacted only with sera from the cancer patients, and 326 reacted exclusively with the sera from controls.
Because the study did not at first reveal any antigens that “either solely or in combination were recognized by antibodies from either all diseases or all healthy patients,” the researchers used three statistical methods — pairwise t testing, ReliefF, and the Proteome Prospector informatics program — to identify those antigens that showed a “greater intensity and/or a greater frequency of positive signals with the sera of either diseased or healthy individuals.”
In total, 90 protein antigens were identified that were targeted by tumor-associated autoantibodies. About 10 proteins showed the strongest reactivity in cancer patients. From that, the researchers chose four antigens that were associated with cancer through the three statistical methods: lamin A/C; structure-specific recognition protein 1; Ral binding protein 1; and ZNF265.
Using immunoblot analyses as validation, the researchers determined that antibodies to lamin A/C and SSRP1 showed a stronger signal in the samples from cancer patients than those of healthy individuals. RALBP1 and ZNF265 were not evident in the dot blot assays, suggesting they are either low-abundance proteins or do not react well with their target proteins in the assays.
Snyder and his team then examined the expression of protein antigens in healthy tissues as well as ovarian cancer tissues by using tissue microarrays. Antibodies for each of the four proteins were used to stain 60 samples, evenly split between healthy patients and ovarian cancer patients. Enhanced staining was observed in most cancer patients for lamin A/C, SSRP1, and RALBP1, but not for ZNF265.
Quantification of the tissue staining results was done through a manual scoring system, and the different antigens were analyzed in detail in epithelial and stromal cells.
The researchers also examined lamin A/C, SSRP1, and RALBP1 in other types of cancer and healthy cells, including kidney, liver, breast, esophagus, and uterus cancer and found that the antigens were prevalent in the tissues of many cancer types and a subset of healthy tissues.
They conclude that the three antigens “produced a robust signature of cancer in tissue sections [and that] … these markers should be useful for tissue diagnostics and further characterization of the disease state. They may also be useful targets for therapeutic intervention.”
“We can now use this type of approach, a microarray plus autoantibody platform, as a way to try to find things of interest.”
Snyder said he did not want to oversell the potential of the candidate markers identified in the study, saying the field is littered with candidate markers that have never panned out clinically. At the same time, however, he said that the candidate markers have “some value for tissue staining in general … [and] will be of high interest to people because they may be informative about the disease state as well.”
In particular, he said, the study found proteins that are not specific to cancer but show differential staining between cancer cells and normal, healthy ones, suggesting that a shift may be necessary for future cancer research.
“I predict that that will be reflective of the kinds of things that people in the field will want to look for in general,” he said, “that we can probably find useful antigens just based on the fact that they’re normal but showing differential expression, whereas there has been a bit of an emphasis on trying to zoom in on cancer specific things — and I don’t think that’s going to be the most productive approach.”
More importantly, however, Snyder said, the study is intended as a proof-of-concept for the utility of protein microarrays. While the technology is not new, use of the platform for proteomics research pales in comparison to the use of mass spectrometry.
According to Snyder, that is more a result of “the herding effect, the fact that most people in the proteomics field like to do the same things.”
There has also been some skepticism around the reliability of the data generated by protein microarrays. Snyder said that it’s important to differentiate the two types of protein microarrays — antibody microarrays, and what he calls functional protein microarrays.
In the case of the latter, the field is much more developed, he said. His lab, for example, has made an array with 80 percent of the whole yeast proteome expressed. And a human microarray has as many as 8,000 proteins on it.
“So those microarrays in that field are much more mature,” Snyder said.
In the field of antibody arrays, however, there are issues, including the low number of antibodies that have been generated against proteins. Antibody arrays on the market contain fewer than 600 antibodies, and “if you think about it, that’s hardly anything compared to the size of the human proteome,” he said.
Another problem is that the antibodies are poorly characterized.
Overall, however, protein microarrays are a much more sensitive discovery and identification platform, particularly in comparison to mass spectrometers, which are notoriously weak at picking out low-abundance proteins.
“That’s one of the reasons we like this approach,” said Snyder. “This is an exquisitely sensitive system. One thing that may not be evident in the paper is that what we were initially looking for was an autoantibody signature that would be reflective of the cancer state, relative to healthy. And that actually didn’t pop out.
“In the end, it turned out to be better as a discovery tool for finding proteins of interest, rather than the antibody reactivity pattern.”