This story originally ran on Nov. 18.
By Tony Fong
Name: Jamie Boehmer
Position: Research biologist, 2008 to date, FDA Center for Veterinary Medicine, 2008 to present
Background: Biologist, FDA Center for Veterinary Medicine, 2004 to 2008; faculty research assistant, University of Maryland Biotechnology Institute, Center for Biosystems Research, 2001 to 2004; research assistant, Institute for Molecular Medicine and Genetics, Medical College of Georgia, 1999 to 2001
Jamie Boehmer directs the proteomics biomarker research program at the Center for Veterinary Medicine at the US Food and Drug Administration, where she and her staff of two are trying to find and identify biomarkers for inflammatory diseases in cows as well as proteins that may serve as alternatives to antibiotics for use in food animals.
The goal of her work is to develop tools that will allow the FDA to evaluate and regulate new drugs for diseases for which no treatments exist, or better treatments than currently exist.
"The first and foremost [goal] is that those practitioners that are coming out to the farm … are going to have something, hopefully, that they are going to use to treat those diseases," she said.
While many of the issues and obstacles that she experiences are identical to those that continue to befuddle researchers in human-based proteomics research, the nature of her work also has its own set of unique challenges. The veterinary research community hasn't exactly wrapped its arms around proteomics, and so there is a poverty of proteomics-directed veterinary research from which she and her staff can draw, at times leaving them to figure out solutions to problems on their own.
ProteoMonitor spoke with Boehmer this week about her work and the challenges she has faced. Below is an edited version of the interview.
Give me an overview of the proteomics work that's being done at the Center for Veterinary Medicine at FDA.
Basically, for our laundry list, we have started with diseases of economic importance, and diseases for which we have fewer treatment options available. The big one that I have focused on is cow mastitis — that is infection of the mammary gland by a gram-negative [bacterial] species.
Why we focus on that in particular is that antibiotics are not effective in treating that disease. The problem with it is that the lipopolysaccharide … [is released] from their cell wall and not from the bacteria itself, so killing the bacteria does not alleviate the secondary effects that are the real problem.
So we've had some requests for relabels of some non-steroidal anti-inflammatory drugs, or NSAIDS, for use in food animals, but we don't have very good biomarkers to evaluate their efficacy in treating inflammation. And so that's been really a primary focus of my research group — looking for these biomarkers that we can use to evaluate treating inflammation in food animals specifically.
There's a two-fold problem with cow mastitis — that you get systemic inflammation as well as local inflammation in the mammary gland. But just thinking in terms of method development and getting at some of the real primary problems being right there local in the gland, we decided to focus on local before systemic. So we've been dealing with looking for inflammatory biomarkers in milk, and milk is a very complex sample.
If I had to sum it up to the two main things that we look at in my lab, one is dealing with very complex biological samples and being able to characterize individual proteins in those samples. And second, we've looked at actually quantifying or looking at the changes, temporal expression patterns, of those individual proteins.
That's something that makes what we do here at CVM very unique in the whole veterinary biomarker scheme. To begin with, there's not a lot of it that's going on. Most of what's going on is antibody-based — the evaluations, ELISAs, cytokine assays, looking at classic and alternative inflammatory pathways. Most of that is being done at the academic level.
We're kind of unique in that aspect in that we're really looking at the big picture, looking at everything. There's been very, very little work done in milk itself, in actually characterizing the bovine milk proteome. We’ve made a lot of progress in that alone.
[ pagebreak ]
Other diseases that we're looking at are pneumonia in beef animals, and we actually have a system here that we can [use to] access bronchial fluids directly, so we're not looking at bronchial alveolar lavage fluid. We are looking directly at bronchial fluid and we are characterizing specifically antimicrobial proteins, and trying to see if there are any therapeutic applications for those antimicrobial proteins.
Existing technology is not effective in diagnosing that?
Actually, we have more drugs to treat pneumonia than we do cow mastitis, because antimicrobials are actually efficacious to treat pneumonia. But …one of the problems with proteomics in general and getting a biomarker validated is being able to find something that's very specific to a disease and being able to handle these very complex samples — fractionating them, reducing sample complexity, depletion of high abundance proteins — all of those issues with human biomarker research carry over to veterinary [medicine].
The main difference is that we have fewer diseases to focus on. Cancer, for example, is a biggie in human research. If a food animal has cancer, it's going to be culled. We're not going to treat that.
What we're looking for [are] diseases more along the production line. We do have that serious advantage — that we have fewer [diseases] to focus on, which makes the aspect of specificity of that biomarker … a lot less complex than human medicine.
Is the work at CVM specifically focused on biomarkers?
Yes, to establish regulatory criteria.
It also sounds like the work is focused only on agricultural livestock.
Yes, food animals entirely.
So you're not doing any research on dogs or cats or pet animals?
No, we're not, not here in the Office of Research. We are a small subdivision of the Center for Veterinary Medicine. All of the work that we do here is on food animals for the benefit of food animals. We don't do any companion animal work.
Can you describe the technology that you're using?
Pretty much what's happened classically in research on these diseases has been antibody-based.
I do a bottoms-up proteomics approach, mass spec-based, LC-MS/MS. We use 2D LC, do some gel-based assays, LC-MALDI, 2D gel to MALDI. We're entirely based on chromatography, protein separation, and identification using mass spectrometry.
You said that most of the problems and bottlenecks that researchers in human biomarker work see are seen also in your work. Are there specific bottlenecks in your research that aren't seen in human research?
I'm not too sure. I think the main problem is really …separating out quantification of these proteins versus statistical analysis of these proteins. …Probably the main problem with the food animals in terms of the proteins that we're analyzing [is] they're very highly conserved.
So if you have a group of proteins, let's say ion-binding proteins, which are very common in some of these bacterial diseases … when you're trying to quantify one versus the other, conserved peptides are a real problem.
I can't really think of anything that's unique, other than we could have more of an issue in terms of when you're looking into different species, those conserved peptides and highly conserved proteins are more of a problem, whereas with humans there's probably a handful of proteins that are conserved across different species, but really, you're dealing with the human population of proteins or base of proteins, whereas in animals, it's often very hard to differentiate some of these proteins, because they're so highly conserved, not only within the types of proteins but across different food animal species.
So it can make database searching a little bit tricky, especially because some of these genomes are not sequenced entirely, whereas … I think it's very easy to say that in terms of human medicine, the databases that are available are much more accurate and complete than what we have.
[ pagebreak ]
Do you end up having to do a lot of de novo sequencing?
I have, yes, in certain cases where things are novel, especially for publication purposes, when you really want to prove that you've got this protein. A lot of times we get an identification based on similarity. You've really got to dig into some of these databases and try a lot of different search engines. So then you've got to be really careful about what you claim in terms of your false discovery rates and decoy database searching, and that can be a bit challenging.
Is there any way to describe how much in funding your office devotes to proteomics research?
In terms of an actual dollar amount, I can't really give you that, primarily because I couldn't really tell you. But in terms of what we do here in our overall operating budget, it's pretty evenly divided among our three divisions here in the Office of Research. And I'm specifically in the office of animal research, and I can say that proteomics is top priority.
And we also take genomic approaches to this, as well, and I work very closely with the genomic scientists. That's all part of the ultimate validation process. It's really tying together the genomic and proteomic data.
It's very high priority, biomarkers in general, even for not necessarily disease-related [biomarkers]. It's one of our primary research focuses.
The proteomics itself, I've had very good support here in terms of being able to get instrumentation and having collaborations outside in the academic environment. We have good funding.
How many people do you have working specifically in this field?
I have two people right now who are working with me in my lab. This is a relatively new field here at the Office of Research, specifically for food animals. We've really just gotten underway in the past two years.
Is your office involved in the regulation of any tests for these animals?
No tests, but we are the office that establishes regulatory criteria. So if we were to be able to identify and validate biomarkers that could be used to evaluate the efficacy of non-steroidal anti-inflammatories, let's say, that would go into the development of guidance for industry — if a pharmaceutical applied for a relabel for an already approved NSAID or a newly developed drug.
Have you actually discovered any biomarkers yet?
We have some very good candidates, I will tell you that, specifically for mastitis, and all I can say is you can look for a publication, hopefully within the next couple of months, in [a prominent proteomics journal].
We're strictly in the discovery stage right now, so where we're moving toward is validation with some candidates.
In human research, there are still a lot of issues around the technology, its reproducibility, robustness, et cetera. Same issues for veterinary medicine?
Absolutely. I have tried a lot of different methodologies and different instrumentation systems in analyzing milk specifically, and I have narrowed it down — we may do a little bit more tweaking — but the specific instrumentation that I use is nanoflow liquid chromatography, high pressure.
Right now, I'm doing a 1D separation. We are working on 2D, but we are always using an ion trap mass spectrometer.
Because of the dynamic range of the proteins present in milk. It really helps us with the low abundance proteins at faster scanning speeds and that ability to accumulate some of those ions.
[ pagebreak ]
Of course, it's low resolution, so there are mass accuracy issues, but for just tracking changes, we've been using spectral counting … because we know, kind of, the time frame that we're looking for to target some biomarkers for inflammation. So spectral counting has been a very effective way for us to evaluate what's biologically relevant to this disease, what's changing at a time when we know that's what we want: to target the inflammation, just based on what we know about cytokines.
We've actually, for some proteins, compared spectral counting with ELISAs and we've gotten very good match-up of temporal expression. Spectral counting has been a really great thing for us. It works fine on low-resolution data.
We've been able to increase the number of low-abundant, diagnostically significant proteins that we've been able to identify using the ion traps.
Can you talk a little about using milk as a biofluid and how it compares to something like blood, which is probably the most difficult fluid in human biomarker research?
One of the things that happens during disease, and I'll talk about mastitis specifically, is that you get cytokine expression very early. The two main ones are tumor necrosis factor alpha and interleukin 1-beta. And why this becomes a problem is, they act on vasculature, so things that typically wind up in the blood wind up in the milk, as well.
For milk from a healthy cow, the main issue there is, there are basically four main casein proteins that comprise about 80 percent of the total protein concentration of milk. And there are two main whey proteins that comprise anywhere from 15 to 18 percent of the total protein concentration.
And then there's that little 2 to 3 percent that's left over that's comprised of, theoretically, hundreds of proteins, those low-abundance proteins.
So, you've got that going on when you're talking about a cow that's healthy. That doesn't change, but when an animal is sick, you then basically have everything that's in the blood in the milk, and the primary component is the big, nasty serum albumin.
Then you have the two-fold problem of your serum albumins in there, and then it's a transport protein, so it's binding up a whole bunch of those little diagnostically significant low-abundance proteins, so if you try to selectively deplete it, you lose a lot of information.
The only effective way to deplete the caseins is either by immunoaffinity or what a lot of people do is acid precipitation. But then you have to dialyse your sample, and so again, you're losing a lot of your small low-abundance proteins.
How I've handled it is just basically trying to fine-tune the inline separation, the chromatography aspect of it — I don't do any sample depletion, I don't do fractionation — and take the whole sample. And that's another reason the ion trap is so helpful at being able to accumulate some of those low-abundance proteins.
In terms of quantification, normalization becomes a big issue because you have such huge changes, not only in the number of proteins but the amount of them from a healthy to a diseased state. It's very complicated; it's almost like dealing with two complex samples in one.
Is this a strategy that you borrowed from human-based research or is this something that you arrived at on your own?
I've had to figure it out on my own. I think that there were maybe two publications prior to the work that I did that even looked at anything in milk at all. And they were definitely based on what's done in human medicine.
[ pagebreak ]
But if you look at some of the previous work, there's been an attempt to deplete high-abundance [proteins] and they're basically identifying three, four proteins.
Unfortunately, when you've got so few in terms of the number of proteins that are high abundance, what happens is, when you fractionate, you carry over. You just spread out your identification of the casein and the whey proteins.
There's a very elegant paper where they actually tried to fractionate normal milk from a healthy animal using 2D gel. They did an inline separation based on cation exchange and then they ran these fractions on 2D gels, and every single fraction was either characterized by a casein or a whey.
It was just a tremendous workload to basically identify maybe five to 10 low-abundance proteins.
Is your method one that can also be translated to human research?
I would hope so, with a little more fine-tuning. Like I said, one of the things that we're doing now is, we have the technology to do nano 2D LC, and that's one thing that we're working on, refining, really slowing down the flow rate, and doing some different online separations prior to direct injection onto the mass spectrometer.
We're still trying to tweak to get a little bit lower, but I've had very good success just taking the kit and caboodle, not trying to deplete anything out. In terms of discovery, I've had a lot of success there.
What's the advantage of using milk as your biofluid over something else, say blood?
A lot of the detrimental effects of the inflammation affect the mammary gland directly, so if we can target it before it hits the mammary gland, or be able to detect those initial signs that it's going to happen in the gland, that would be advantageous because we can alleviate a lot of the secondary effects that are what cause billions and billions of dollars lost in the dairy industry.
Cows end up having to be culled, and because of the issue of the breakdown of the blood-milk barrier, it is often through the milk, and that exchange between the blood and the milk, that the cow goes septic and dies.
As well, it's easy to obtain a lot of sample, and if in the long-term future, someone outside of FDA wanted to try to develop a diagnostic for a particular biomarker that a farmer could measure, it would be much easier for that farmer to measure that marker in milk.
What's the general state of proteomics as applied to animal science? You said before there isn't very much work being done, but is that changing? Or do you see that people don't really trust the technology and so are reluctant to adopt it as a research method?
If I had to point the finger at one thing, I would just say it's the state of funding in terms of academia. A lot of academics, the type of biomarker research that they're doing is, they're trying to get a lot of bang for their buck, so a lot of them have gone with [genomics-based research] because chips are a little less expensive now and that technology has been around for a little bit longer, and most academic environments have a core facility where they can get a genome sequenced.
Proteomics has been a little bit slower to catch on because I think mass spectrometry is very daunting … and a lot of people don't understand the advantages of what it can offer over antibody-based strategies and even over genomic strategies.
But a lot of people don't have that instrumentation available to them. In terms of getting funding to do it, the emphasis is definitely on the human side of things unless it's using an animal model for a human disease