Scientists at the University of Edinburgh have discovered a protein signature in urine that was shown in early-stage studies to identify the presence of carcinomas of the upper gastrointestinal tract with sensitivity and specificity over 90 percent.
Using SELDI-TOF mass spectrometry, the researchers, led by Holger Husi of the University's School of Clinical Sciences and Community Health, identified a pattern of proteins that they believe has the potential to eventually serve as a diagnostic for upper GI cancers.
The group confirmed two of the proteins, S100-A6 and S100-A9, by Western blotting, and identified several others as likely candidates by their molecular mass. They published the study in the June issue of Proteomics Clinical Applications.
The study examined a total of 120 urine specimens — 60 from patients with upper GI cancer and 60 from healthy controls — to establish the biomarker pattern, and then another 59 samples from 33 control and 26 uGI patients for a blinded validation of the signature.
"It was almost too nice to believe," Husi told ProteoMonitor this week. "There was nothing clear cut when we began, so starting the project it could have all fallen to bits," but the markers performed "exactly as predicted" in the validation set, he said.
According to Husi, though the research is in a very early stage, the group's end goal is to create a diagnostic test to catch upper GI gastric cancers in their early stages, allowing the possibility of more successful and less invasive treatment.
For the discovery stage, the group tested a number of SELDI chip types, choosing eventually to use CM10 chips to analyze the entire cohort.
"What I was given by the clinicians was sort of the raw samples … then I took the whole thing and … basically threw it at the mass spectrometer, and tried to find the particular pattern," Husi said.
Mass spec of the biomarker discovery specimens revealed several peaks, which the researchers were able to translate into a predictive model using Biomarker Pattern software. In the learning sample set, the model performed with 98 percent sensitivity and 95 percent specificity. In the validation set it reached 96 percent sensitivity, and 72 percent specificity.
Husi said the researchers then matched the measured peaks to four likely proteins, validating two – S100-A6 and S100-A9 – with Western blotting.
While much research has been devoted to finding protein markers in blood and serum, urine has several advantages, the study authors wrote, including its relative stability in terms of protein composition and fragmentation state, normally unrestricted obtainable quantities, and non-invasive, easy sampling.
Urine is also a more simple fluid to analyze — especially for mass spectrometry, which can have difficulty detecting low-abundance markers among the background noise of much higher-abundance proteins in serum or blood. Other researchers have recently examined urine for gastric cancer markers, but struggles with mass spec led them to use an array-based discovery method (PM 5/06/2011).
Interestingly, Husi said, the proteins identified in his study are most likely "global markers" of cancer, not just of uGI carcinomas.
"These S100 molecules, not much is actually known about those to be honest," he said. "All that is really known is that they are supposed to bind calcium. Where they are involved and what they are doing, it goes all over the place."
"They are certainly associated [with cancer] based on previous work by other people which looked at different cancer types, though, again, not in urine," he said.
There is also a chance, Husi said, that the S100 proteins are associated not with cancer necessarily, but more generally with inflammation. However, he explained, his continuing research over the last few months has shown that the protein signature, at least, does not correlate with inflammation.
"That has been something I have been addressing over the last several months — trying to figure out whether there is anything that might go hand-in-hand with systemic inflammation. None of the people in my cohort actually had just systemic inflammation without cancer, so I [couldn't] answer that particular question [at the time]. However, it was raised by other people as well that it might be to do with inflammation," he said.
Husi said he used C-reactive protein, which is associated with inflammation and cross-mapped that with his markers. "I could find no correlation whatsoever," he said. "So at this stage I can be quite confident that none of the markers have anything to do with inflammation per se."
The researchers also built a peak model that included only S100-A6, finding that this single-protein predictor also had high sensitivity and specificity (86 percent and 80 percent respectively), demonstrating the importance of this single molecule.
However, Husi said, having at least two markers combined into a signature only increased the predictive accuracy, and is clearly preferable.
According to the study, the group also generated a "full SELDI-TOF-MS data set" using another chip type, the IMAC30. The authors wrote that their initial analysis using the same study approach has shown a comparable number of candidate markers. However, "the molecular identity of those markers, solely based on m/z distribution, is potentially different from the ones identified here." This suggests that there could be additional biomarkers beyond those identified in the group's study.
Moving forward, Husi said that he plans to try to narrow his search to proteins that could serve as markers for specific cancer types.
"I've been looking at … whether I can sub-define, or look at more specific markers which are unique to the different cancer types included in my study — let's say gastric cancer or pancreatic cancer. I have some lead candidates already identified for that," he said.
In addition, more validation work is required on the signature reported in the study.
The next stage would be to open up the cohort to a large random sampling, Husi said, regardless of whether the group examines one cancer type or a global marker, "and then, of course, to see if I can find in confirmed cancer cases whether they are still following the paradigm, or whether it falls apart," he said.
"And at the same time then we [also need to] look at longitudinal studies… to see for high-risk people, for example, for a period over five years, who is going to develop disease and when any markers start cropping up … And then leading on from that, to use this information to actually establish a bona fide clinical test, that would be the ultimate aim," Husi said, though he did not provide a timeline for reaching that goal.
The researchers propose in their paper to establish a database of urinary SELDI-TOF-MS results, called UPdb, to "allow the scientific community to assess published findings as well as incorporate existing data into new studies. All spectra used in this study will be deposited in this database," they wrote.
Have topics you'd like to see covered in ProteoMonitor? Contact the editor at abonislawski [at] genomeweb [.] com.