Name: Denis Hochstrasser
Position: Chairman of the Genetic & Laboratory Medicine Department of Geneva University Hospital
Background: MD, University of Geneva; Founder, Swiss Institute of Bioinformatics, GeneBio, and Geneva Proteomics.
Denis Hochstrasser is the chairman of Geneva University Hospital's Genetic & Laboratory Medicine Department and vice dean of the university's faculty of medicine. He is also a member of the executive committee of the Human Proteome Organization, founder of the Swiss Institute for Bioinformatics and GeneBio, and a scientific founder of Geneva Proteomics.
At the Human Proteome Organization's 10th annual meeting in September, Hochstrasser spoke about "missed opportunities" in clinical proteomics, raising the question of whether, in its emphasis on identifying blood-based protein biomarkers for diseases like cancer, the field might have passed over lower-hanging clinical fruit.
This week, ProteoMonitor spoke with Hochstrasser about some of the questions he raised in his presentation and discussed where he felt proteomics would ultimately prove most useful clinically.
Below is an edited version of the interview.
During your talk at HUPO you noted the challenges of blood-based protein biomarker work. Do you think right now the technology simply isn't there to move such assays into clinical practice?
I believe that, currently, among scientists who do multiple-reaction monitoring [mass spec], they do not see so well the major difficulties of entering the clinical world. When you digest the protein it is very tricky to measure it with high accuracy and sensitivity, but, more importantly, reproducibly. To measure the presence of the peptide in a blood specimen is a very tricky issue.
One of the reasons for the difficulty is the amount of anti-trypsin enzymes that we have in the blood. Depending on the disease of the patient, those proteins that block the effect of trypsin can be dramatically different in concentration. Some of them are acute-phase reactants, and when you have a fever, when you have inflammation, you have a lot of those anti-trypsin agents, and so I think that the reproducibility of doing an MRM assay in the blood of sick patients in the hospital will not be easy.
Another argument is the fact [that] if you draw the blood and cool [it] immediately you modify the blood. Platelets are built by nature to release into the blood all their contents when the temperature of the blood drops. When you cut yourself, the blood gets to the surface and the surface is cooler, and so the platelets realize that something is wrong and that you need to stop bleeding, and so the platelets release their contents, [which triggers coagulation].
On the other hand, if you do not cool the blood, then what's happening is that outside the body the cells continue to live and so they modify their environments: pH drops and lactate builds up, and you have many changes. Doing an MRM study on control or well prepared serum or plasma samples is far from the clinical reality where the nurse draws the blood, the blood sits on the bench for a few hours, the tube eventually gets to the lab, and people start to do the work. I think there's a long way to having a practice for using it in the clinic.
Wouldn't many of these problems apply to blood-based assays in general? Not just assays using mass spec?
Yes and no. There are several publications saying that ultimately MRM-MS studies are not that different from immunoassay studies done in the past. That is partially true. But when we have an immunoassay that recognizes an epitope that is not modified by whatever happens in the blood sample, I think the test is robust.
Andy Hoofnagle at the University of Washington's Department of Laboratory Medicine, for instance, is using SISCAPA-MRM assays to measure thyroglobulin as a backup in cases where results from ELISAs don't line up with patients' clinical situations.
Yes, I discussed that test with Broad Institute researcher Steve Carr. He just sent it to me after my presentation at HUPO. The immunoassay that we have for thyroglobulin has a lot of difficulties, and so we might get around those using MRM-MS. So, maybe; I need to check that. I didn't say that it won't happen. I said that it's quite difficult and that it won't be that easy to get it into clinical practice for every protein.
So, mass spec could potentially offer an option for proteins where a good immunoassay doesn't exist?
That's exactly what I was thinking about in terms of missed opportunities. Missed opportunities to apply mass spec-based proteomics where there is a need for a good test. Not only in the cancer area, although I realize that thyroglobulin is a marker of thyroid cancer. But I think when I was talking about missed opportunities, I felt we should apply proteomics to areas where the problem is not that tricky, as opposed to cancer markers that are difficult to identify and follow.
What is an example of a clinical problem you feel has been overlooked?
Typically the example I give is the analysis of disorders of hemoglobin, doing it by mass spec using electron-transfer dissociation. The first reason for this is that hemoglobin is in such a large concentration that you don't have to worry about any sensitivity issues. You can inject the red blood cells directly into the mass spec – you don't even need chromatography up front.
The second reason is that today when you study hemoglobin disorders, you have two types of disorders. You have quantitative disorders, and you have qualitative disorders. Quantitative disorders are like thalassemias in which too few globin chain types are synthesized. Qualitative disorders are diseases like sickle cell anemia caused by mutations and improperly functioning globins.
When you have a patient with anemia [that is not caused by] iron, vitamin B12, or folate deficiency, you have to look at all of these disorders. To do so right now you need to use several techniques. For the quantitative disorders you need to do electrophoresis or chromatography. If you're trying to diagnose a genetic disease of hemoglobin, then you need to do PCR, genetic testing, and sequence the gene.
With only one test — that is, mass spec analysis of hemoglobin using ETD — you are able to analyze the hemoglobin quantitative disorders, because you can very nicely see alpha chain, beta chain, and most likely all the other chains; you can directly sequence because with ETD you can almost completely sequence the hemoglobin chain and see any mutations; and, most likely – we are not there yet, but I am convinced we can do it – we can see modifications, like glycated hemoglobin. And so with one mass spec test we can do several tests faster and cheaper that we are currently doing in clinical practice.
I have a PhD student working on this test, and we have a collaboration with industry, and I hope to get it in routine clinical practice within six months.
So you think opportunities like this have been overlooked by the field?
Clearly in microbiology! For some reason people [for many years] have looked at the Holy Grail as being cancer tests. And I believe that there are tests in other fields in clinical practice where we would benefit from mass spec.
Any examples other than hemoglobin?
I think once you have started with hemoglobin one could look at other proteins that have a good diagnostic opportunity, but it's a bit too early to say. I want to finish with hemoglobin and see how it goes.
In the diagnostic area there is clearly a fine line, and you have to define: Do you want to do it on the gene side, on the transcript side, on the protein side, or on the metabolite side? For example, with diabetes, the best diagnosis is [testing an individual's] glucose [level]. Now, you could do a diagnosis or a predictive diagnosis of diabetes from the gene side, and you could do it from the protein side with glycated hemoglobin. I'm convinced that in several diseases you will have to decide where or what you should measure – the gene, the transcript, the protein, and/or the metabolite.
Typically, for toxic effects I would guess it would be more on the protein side, although it could affect DNA or its surroundings. When you talk about cancer, I think that it will be deep sequencing on the gene or transcript side, and I doubt that the cancer diagnosis will really come from the protein side. The treatment follow-up, on the contrary, could come from the protein side.
You think proteomics will prove more useful for patient monitoring than for making initial diagnoses?
So far from the protein side, protein markers have been useful in monitoring treatment. For example, if you have colon cancer, you measure carcinoembryonic antigen. If CEA is up, you treat the patient with surgery and you measure CEA after surgery. If CEA is zero after surgery, you feel good and you can then follow the patient. If after surgery CEA is still up, there is a metastasis somewhere and you have to find it. However, you cannot use CEA as a diagnostic marker to find colon cancer initially, because you can have colon cancer without CEA, and you can have CEA elevated from other cancers or practices such as smoking.
Do you think then it is misguided to search for protein biomarkers as diagnostic biomarkers for cancer?
To diagnose cancer early, the questions are, 'Is there any abnormal protein floating around telling you that the cancer is there? Are there any abnormal lipids or sugars or modifications of proteins that tell you that the cancer is there? Are there any abnormal cells that are already circulating?' Of course there are many, many types of cancer, and in some cancers, when there are a few hundred cells present, it's already too late because they have already disseminated and there is no way we can detect just a few hundred cells in the body. Then there are cancers where you have a tumor of a few centimeters in diameter and it's not too late. You can remove it because the behavior of the tumor is OK. The likelihood that you could find a protein biomarker [for the first kind of tumor] before it has already spread is relatively small. I don't know if it is possible.
I think in the blood there are very interesting biomarkers, but many of them are non-cancer biomarkers. Recently procalcitonin was introduced, and it is an excellent biomarker for sepsis. Another biomarker that came out relatively recently is amino-terminal pro-B-type natriuretic peptide, which is a very good biomarker with negative predictive value for heart failure. It is very good for clinical management of heart failure.
I will give you another example, and it is a very interesting example. I think the first success of the clinical application of proteomics is the Bruker MALDI Biotyper. It is spreading like crazy. There are more than 150 labs in Europe using it, I believe. You can identify a bacterial colony within seconds by shooting the protein with a laser and getting the proteomic profile of the bacteria. It's cheaper than genetic testing by PCR, and it's very fast. We save days, and we save a lot of costs as well. That is clearly a very successful clinical proteomic application. And in a sense it was a missed opportunity for 10 years because it was known 10 years ago that you could profile bacteria by proteomics, and it was not introduced in the clinic until almost a year ago.
Have topics you'd like to see covered in ProteoMonitor? Contact the editor at abonislawski [at] genomeweb [.] com.