NEW YORK (GenomeWeb) – A team led by scientists at the University of Pennsylvania has identified a pair of protein markers that could prove useful for diagnosing pancreatic cancer.
In a study published today in Science Translational Medicine, the researchers found that the proteins, thrombospondin-2 and CA19-9, could identify patients with pancreatic cancer with a specificity of 98 percent and sensitivity of 87 percent, suggesting that they could aid early detection of the disease, particularly in high-risk populations, said Kenneth Zaret, professor of genetics at the University of Pennsylvania and senior author on the paper.
He added that he and his colleagues are now continuing validation of the markers using prospectively collected blood samples provided by the National Cancer Institute and collaborating with clinicians at the University of Pennsylvania School of Medicine and the Mayo Clinic to use the test in high-risk patients.
They are also exploring additional candidate markers identified but not pursued in the STM study, Zaret said.
Pancreatic cancer is notoriously difficult to detect early with most cases diagnosed at advanced stages, at which time removing the cancer via surgery is no longer possible. This contributes significantly to the generally poor prognosis of pancreatic cancer patients, whose overall five-year survival rate is around 7 percent.
The typically late-stage diagnosis of pancreatic cancer has also presented a challenge to researchers looking for markers of early-stage disease, Zaret said. "If you looked at how people have developed or tried to develop biomarkers, they used what they could get, which is usually late-stage pancreatic cancer tissue handed to them by the surgeon after a resection. And party's over at that point."
To identify potential early-stage markers, Zaret and his colleagues used a stem cell-based model developed at Penn's Institute for Regenerative Medicine, where he is director. The researchers took pancreatic cancer cells and reprogrammed them using the commonly used approach for generating induced pluripotent stem cells from adult cells. When grown in mice, this reprogrammed cell line produces stage II and III pancreatic intraepithelial neoplasia that progress to invasive pancreatic ductal adenocarcinoma (PDAC).
"I thought if we could take a late-stage cancer cell and make it into an early stem cell, maybe it would redevelop the cancer," Zaret said. "In other words, it would undergo the progression and not just go to a late-stage right away. … Using that model we were able to develop early-stage pancreatic cancer tissue and have live cells that were actively secreting and releasing proteins from this human cell model," he said.
This stem cell reprogramming work, initially published in 2013, was part of larger effort by Zaret and his colleagues to develop stem cell-based models of early-stage disease for a variety of cancers. However, he said, of the more than 20 different tumor types the researchers have tried the technique with, the PDAC line is their only success thus far.
"I had hoped that this would be a generally applicable mechanism, but there's something about late-stage tumor cells that usually resist the kind of reprogramming that we do," he said. "So, we basically got lucky once. And truth be told, that's one of the criticisms of our [2013] study, that we only developed one line that was useful. I understand that. But, nonetheless, it was useful, and there are many examples of cases of individual cell lines that have been useful for study."
Using their PDAC model, the researchers identified 107 proteins secreted by the stage II and III tumors. They then winnowed this set down to a group of 53 proteins with low abundance in the plasma of healthy subjects, ultimately focusing on three proteins, matrix metalloproteinase 2 (MMP2), MMP10, and THBS2, that are part of cellular networks linked to PDAC progression and for which good ELISAs existed.
Testing these markers in a 20-sample discovery phase looking at PDAC cases and controls, they found that THBS2 could identify patients across all stages of pancreatic cancer with an area under the receiver operating curve of .76.
They then moved to looking at the performance of THBS2 in combination with the protein marker CA19-9, which is used to track progression in patients with PDAC. In a validation study looking at 189 patients (58 with stage I or II PDAC, 23 with stage III or IV, 28 pancreatitis patients, and 80 healthy controls), they found that THBS2 and CA19-9 could distinguish between cases and controls with a UC of .96. A second validation study looking at 537 patients (88 with stage I or II PDAC, 109 with stage III or IV, 115 with an intraductal papillary mucinous neoplasm, 30 with a pancreatic neuroendocrine tumor, 55 with pancreatitis, and 140 healthy controls) found that the two markers could distinguish cases from controls with specificity of 98 percent and sensitivity of 87 percent.
Zaret noted that, due to the low prevalence of pancreatic cancer, this level of performance is not sufficient for a general screening test, but it could be a useful rule-out test in high-risk populations — for instance, in patients with newly diagnosed diabetes resulting from a pancreatic injury or carriers of certain relevant germline mutations.
"If you have a positive result with our test, it would be very warranted to get further analysis [such as imaging] to determine if you are harboring a tumor," he said. "And if you are negative with our test and are in the high-risk population, you probably are fine. The statistics would say that."
"Of course, we have to do further tests," he added. "We've done over 700 samples, and now we have to do 7,000, right? As more and more contexts are studied and this marker is looked at in more and more situations, maybe it will find that it also reports out other kinds of pathologies that don't necessarily relate to cancer, and that will reduce the confidence in the model. But, for the time being, it seems useful for the high-risk population."
Zaret said that he and his colleagues are also exploring proteins from the original set of 107 that they did not describe in the STM paper. Mass spec-based analyses suggest that several from this set could be "as good or better candidates," but, he said, more extensive follow-up would have to wait until the researchers had funds to develop high-quality ELISAs to these targets.