University of Pittsburgh
Who: Thomas Conrads
Position: Associate professor, University of Pittsburgh, department of pharmacology, 2006 to present; co-director Clinical Proteomics Facility, University of Pittsburgh Cancer Institute, 2006 to present
Background: Deputy director, Biomedical Proteomics Program, National Cancer Institute, 2001 to 2006; director, Mass Spectrometry Lab, SAIC-Frederick-NCI, NCI-Frederick, 2003 to 2006; PhD, Ohio State University, biochemistry, 1999; postdoc with Richard Smith, Pacific Northwest National Laboratory, 1999 to 2001
Formalin-fixed, paraffin-embedded tissue samples are widely used for tissue processing prior to light microscopic evaluation. However, the technique has not been used for mass spectrometry-based proteomics research, mainly because the formaldehyde used in preparing the samples creates crosslinks that render proteins insoluble, and, therefore, unsuited for routine biochemical analysis.
Thomas Conrads has been conducting research to develop technologies and methodologies to change this. Below is an edited version of a conversation ProteoMonitor had with Conrads this week about his work.
Why is it so difficult to use mass spectrometry on FFPE tissues?
The typical route in proteomics for conducting biomarker investigations has been to take the tissues or samples, in general, and digest them first into peptides. The reason is that peptides behave better in the separations, and they’re readily identifiable using the types of mass spectrometry technologies that are currently employed where we do tandem MS and we get partial sequence information [and] match that back to proteins.
Just conceptually, when one thinks about what a formalin-fixed, paraffin-embedded tissue is, the first thing that comes to mind is that it’s likely that the sample is completely crosslinked and, therefore, intractable to the typical digestions that we employ. The concept is you have this web, mesh, crosslinked stuff — proteins, DNA, all the biomolecules in the cell are basically thought to be crosslinked and intractable.
What we’ve been doing over the last three, four, five years is trying to develop sample-preparation protocols in collaboration with Expression Pathology to make essentially a kit that you can buy off the shelf and then render these samples amenable to the typical workflows that have been so widely used in proteomics for biomarker discovery, which is again predicated on this [model of] digestion-first, and then analysis of peptides.
It’s really the development of this type of technology that now is opening up the doors of these pathology labs that have scores of these slides basically just sitting very stable at room temperature.
Why would the use of FFPE tissues be advantageous for proteomics research?
The primary reason is that when we talk about doing biomarker discovery experiments, we generally talk about doing relatively minor populations, but nonetheless populations of patients who have either cancer or not, or different stages of cancer. And trying to access or design new prospective clinical trials to set up gathering these samples is extremely expensive. It’s very unlikely that you’ll get such a biomarker discovery, prospective gathering funded by the NIH because it is so expensive [and] the outcome isn’t known. We’re not talking about hypothesis research where we can guarantee a biomarker at the end of the day. To set up a clinical trial to gather prospectively is almost out of the cards.
I think the value is really in the ability to go back [and] retrospectively access what are typically unused blocks of resected tumor, and then design retrospectively the biomarker investigation using these [tissues] that most people have forgotten about.
So I think the power is in the ability to design retrospective proteomic biomarker investigations.
Have people actually tried using them for proteomic research, trying to find biomarkers?
It’s a very limited cadre of folks. We’ve been working in this field for quite some time now, maybe three, four, five years. Really, when you look around the field, there’s maybe another two or three groups around the world who’ve tried it even.
Can you talk about the research you’re doing now?
We’ve got a couple of programs where we use and are designing these types of retrospective investigations. One of the ones we’re most excited about is where we’re looking to quantitate HER2 expression.
HER2 is a protein whose overexpression is related to very invasive breast cancer. Of the patients who are diagnosed with breast cancer, about 20 to 25 percent of them have this protein overexpressed, and it’s related to a very poor prognostic outcome.
We’ve got some therapeutic intervention with this molecule called Herceptin. Herceptin is an antibody that can be given to the patients. It’s a molecularly targeted therapeutic, so it targets HER2 proteins, and the antibody prevents this protein whose overexpression in over 20 percent of patients is responsible for driving the very aggressive breast cancer phenotype.
The issue, though, is detecting and quantitating how much HER2 there is in the breast cancer. We’re developing technologies using very targeted mass spectrometry analysis to quantitate HER2 expression in breast cancer samples, both from formalin-fixed and from fresh frozen, the idea being we want to understand better essentially the clinical outcome and how it relates to HER2 expression — because the real molecular correlate hasn’t been demonstrated. All the studies to date have been done with antibodies, which don’t give you very good quantitation. Certainly they don’t directly detect the molecule.
We’re using mass spectrometry and very targeted analyses of HER2 peptides from these tissues to, again, almost clinically stage patients — whether they have HER2 or not — and then make decisions about treatment and what best route to take. Obviously, if they don’t have high HER2 levels, you’re less inclined to treat it with Herceptin because not only is it a cardiotoxic molecule, so if you get Herceptin and you don’t need it, there’s cardiotoxicity related to it, plus, it’s very expensive treatment.
What are you finding? Are you having the problems with protein solubility that have been described as one of the issues of using FFPE samples?
Actually no, not at all. The technology and the kit basically that Expression Pathology is now marketing is wonderfully suited to both soluble proteins and membrane proteins, which HER2 is. So we get very good recovery of all sorts of types of proteins across the spectrum of solubility.
What about standards. There are a lot of groups trying to develop standards for proteomics research, including tissue sample standards. How does that apply to FFPE?
One of the studies that we’ve been doing has been in prostate cancer, and one of the nice things that tissue affords you is essentially the ability to do laser-capture microdissection. Using this dissection technique, you can capture tumor cells and you can capture adjacent healthy cells. So essentially, your standard is built into the project because you’re comparing the same person’s tumor to the same person’s normal tissue and so you have essentially the ultimate control.
Has your research looked at how to overcome some of the hurdles of dealing with this type of tissue, or are you looking specifically at just identifying the proteins and biomarkers?
A lot of the initial development that we did … really we were strongly working to develop essentially a technology, that is a buffer system and a kit, to get very standardized recovery of proteins across all FFPE tissues. We think we certainly have demonstrated through a number of publications that we think we do get very standardized recovery.
Now, we’re at the point where we’re using the technology to do biomarker discoveries and to develop the ability to quantitate these molecules from tissues. We’re really at the stage where we’re applying what we developed over the last five years to clinical samples.
When you’re talking about technology, you’re talking about basically mass spec-based technology, right?
Primarily, but in the laboratory we certainly have our eyes and our intellects focused on the application of protein microarrays as well, recognizing that mass spectrometry is great for discovery, and it may have its use also in validation. But I think most of the field recognizes a strong presence and a strong utility of using certainly microarrays, specifically protein microarrays, whether they be forward arrays or reverse-phase arrays.
We do have our eyes set on the use of high-affinity reagents for certainly targeted analyses based on the discoveries that we make using mass spectrometry.
Are there certain types of mass specs that work better with this type of tissue than others?
Most of our data has been gathered on essentially a basic linear ion trap, so it’s completely [doable] by even your average core lab.
Are there certain methods of proteomics techniques that work better with FFPE tissues, high-throughput or top-down, for example?
What we’re analyzing using the product of this developed technology are peptides, so top-down may not be so useful. We use a typical bottom-up strategy where we’re analyzing peptides and searching for information regarding not only identification, modification, but also quantitation of these emergent peptides from the FFPE tissues, across, for example, case-control patient cohorts.
What hurdles still exist that your research has not addressed and how do you plan to address them, if you plan to address them?
I think the real hurdle when you talk about tissue, is in getting enough material to do a very good, in-depth analysis with replication and the proper statistics behind the analyses. When we talk about clinical tissues, we’re talking about very limited amounts, so I think the major hurdle that we all face when we’re dealing with clinical proteomics, as opposed to when we’re dealing with cell lines or what have you, is in getting enough to do a very good, in-depth analysis.
And that has to do both with our efficiencies of extraction at the sample preparation level, efficiencies of digestion, efficiencies of sample handling. It also strongly has to do with essentially resolution, dynamic range, sensitivity of the mass spectrometer.
The field right now, where we’re going is in trying to integrate and optimize every step of what is a relatively complex workflow, especially when we talk about using laser capture microdissected cells from tissue when maybe all we have are 15,000 to 20,000 cells, and what we want is a good, in-depth, broad coverage of the proteome.
The challenge we have in clinical proteomics with these types of tissues is not necessarily in identifying lots of proteins. We can do a pretty good job at that. What we don’t do very well right now, especially with those types of cells, is in enumerating the post-translational modifications, those things that decorate proteins that actually modulate and give them activities.
Certainly when we’re talking about cancer, we’re exquisitely focused on understanding not necessarily what the circuits are … we can identify pretty well what the circuits are. I think the challenge is in understanding how the circuits are misfiring or how they’re cross-talking in different phenotypes than we would see in normal cells.