Skip to main content
Premium Trial:

Request an Annual Quote

Robert Bowser on Starting a Company Based Upon SELDI-Based Biomarkers for Lou Gehrig s Disease

Premium
Robert Bowser
Visiting assistant professor of pathology
University of Pittsburgh

At A Glance

Name: Robert Bowser

Position: Visiting assistant professor of pathology, University of Pittsburgh, since 1994. Co-founder and consultant for Knopp NeuroSciences, a company that investigates ALS, since 1994. Director of University of Pittsburgh's Amyotrophic Lateral Sclerosis Tissue Bank, since 1997.

Background: Postdoc, Peter Davies' laboratory, Albert Einstein College of Medicine department of pathology, 1992-1994.

Assistant professor of pathology, University of Pittsburgh, 1995-2001.

PhD, Peter Novick's laboratory, Yale University department of cell biology, 1991.


Last year, Robert Bowser co-founded Knopp NeuroSciences based upon a panel of biomarkers for Amyotrophic Lateral Sclerosis, or Lou Gehrig's disease, that he discovered using Ciphergen's SELDI technology. ProteoMonitor caught up with Bowser to find out more about his background in SELDI research, his new company, and the ALS test in development.

How did you get into using SELDI technology?

I'm interested in mechanisms of neurological diseases and doing predominantly biochemical methods and approaches to uncovering proteins involved in mechanisms of neurodegeneration. I heard a talk on SELDI in November 2002 for cancer based biomarkers and thought it could be a technology that could be applied to neurologic disorders. And I thought that ALS, or Lou Gehrig's disease, would be a good disease to try with this technology for a few reasons.

One, there really aren't any biomarkers for the disease, especially not a panel that's a real good diagnostic tool for the disease. Another reason I thought this would be a good neurological disease to try is that unfortunately for patients with this disease, the disease rapidly progresses. It's not like some of the other neurological diseases like Alzheimer's and Parkinson's where patients can live for a decade or more with the disease and it slowly progresses. This is one that rapidly progresses. By applying this type of technology to this type of disease both at early stages, when you're trying to find putative diagnostic markers, and also following throughout the course of the disease, which unfortunately is only a few years, one might be able to identify protein biomarkers.

The ALS Association has supported and funded much of my research on ALS.

Why did you think it would be better to study a rapidly progressing disease?

Because with the slowly progressing diseases, it's hard to truly identify progression within the patient population when you have a mixed bag, so to speak. So you have patients that are slowly progressing, and the rate at which each individual patient progresses is very random. For ALS, the time period is much contracted. So there's still variability in individuality of the disease, but the time period is much contracted, so it's easier to track in a way.

Before you got into SELDI, what kinds of technologies were you using to study ALS?

Really protein biochemistries, column fractionations, purifications, immunohistochemistry. I run a tissue banking program for ALS; I harvest brains and spinal cords from patients after time of death so we can do a diagnosis and confirm the clinical diagnosis and also have the material for research purposes.

How did you learn how to do research using SELDI?

Some of it was with the assistance of Ciphergen, and much of it was pretty much on our own. At the time, we had access to a SELDI mass spec through our proteomics core facility at the University of Pittsburgh, and we pretty much just played with the machine, so to speak, to learn how to best utilize the samples that we were using for the study.

The other advantage I saw in our study was we also, instead of using blood, serum or plasma, we were using cerebral spinal fluid. The complexity of the study material is about 1000-fold less than serum or plasma. We certainly were able to obtain reproducible data using the platform.

Once you got into SELDI, how long did it take you to start finding something interesting?

Probably about six to eight months. It was good in that it took very small quantities of material, and it was fairly rapid. Obviously, your samples are critical for any type of biomarker study. We actually are now finishing up a paper describing how we used SELDI instruments to do an in depth study of the stability of the proteome in cerebral spinal fluid.

The question is, if you take CSF or even blood in a clinical study, how long does it take to get from the patient into your freezer? Different clinics operate in different ways. Given whatever particular day it is, it could be random. It could be an hour, or it could be many hours. And so we did a study looking at the how stable the proteome is in CSF at two different temperatures. That was sort of a first study for that type of an analysis using a SELDI mass spec. There's one or two reports fairly recently using proteomic stability in CSF using either 1D- or 2D-gel approaches. But then again, you're looking for proteins of fairly high abundance. But we couldn't apply that to mass spec, where you're looking at proteins of low abundance in an approach that's more sensitive than a gel-based approach.

What did you find with the two different temperatures?

We found that at room temperatures, you've got two to four hours from harvesting the patient to when it's got to be in the freezer. So we've been telling this to our collaborators, saying, 'Look, if you send us samples even six to eight hours out, and we include those in our study, we're going to create artifacts in our mass spec.'

What happens after two hours?

It starts degrading in some regions of the spectrum. Some peaks appear quite stable, and others start to fall apart quite rapidly.

How many candidate ALS biomarkers did you find using SELDI?

We have a nice panel of about 19. That's using specific algorithms to mine the data. There's one that we have used fairly extensively that was created by some of our collaborators for data mining spectral data sets. The algorithm is called Rule Learner. It's a machine learning algorithm that seems to work quite well for this type of data set.

Is that a Predictive Diagnostics software?

It is. So it learns rules from a training set, and then can apply it to test groups. It's not a commercially available algorithm. It's not BAMF, which is commercially available.

I don't think there's going to be one algorithm that is 'the best.' I think that in this type of work, using multiple algorithms becomes valuable because you start to see peaks of interest from multiple algorithms. And if you start to get that over the common group of protein peaks, I think that's really trying to tell you something — that those actually are quite informative and valuable.

Of the 19 biomarkers, did you identify some of them, or try to narrow down the group?

Yes. We narrowed down to a group of 12 of them, and we've been trying to sequence some of them. That's the hard part of the technology — if the peaks are small, and you've got to purify and enrich and determine the amino acid identity, that can take considerable time, depending on the peak. So out of the panel of 12, we know three of them. Right now we're putting the finishing touches on revisions to the paper about this panel of biomarkers. It's been submitted to the Journal of Neurochemistry.

The ALS test involves all 12 of the biomarkers. Not everyone has all 12. As you might imagine, some people have certain markers, and others have other markers. Hence, you need a panel. For these types of diseases, I don't think you're going to find one biomarker that works for everyone.

How well has the ALS test done in terms of sensitivity and specificity?

In the training set, it was fantastic, and in the coded sample set, it worked great, but it was a small sample set. The blinded samples that we had at the time were only 12. It was 96 percent accurate, but there were only 12 samples. So we've just finished an analysis of another almost 200 samples. We're working through that data right now, and we're hoping, obviously, to confirm and validate the initial study.

In terms of sensitivity and specificity for the small set of blinded samples — specificity was 100 percent, and sensitivity was 92 percent.

We went back to the three protein biomarkers that we've identified, and validated the mass spec results using other technologies. So we can do western blots on our cerebral spinal fluid samples. I went back and did immunohistochemistry on lumbar spinal cord samples from our tissue bank. So we could validate the mass spec results using other technologies.

So, especially for the neurologic disorders, this is a study that not only used a SELDI/mass spec approach to identify biomarkers, but also confirmed them using two different approaches — using both samples used in the study, and other coded blinded subjects in a different patient cohort.

What would be the next step after you identified somebody as having ALS using this panel of biomarkers?

Well, for ALS, there's only one FDA approved drug. So they would start that drug fairly quickly. That drug, unfortunately, only improves the life span by about two to three months. If you give it sooner, it'll probably work better but now no one can really identify the patients quickly. Diagnosis for this disease is typically almost a year.

Early detection would help in getting them on the drug quick. Also, we know that there are some quality-of-life issues that would be greatly improved. We'd be able to slow down the time in which a patient would have to be in, say, a wheelchair or a ventilator. So that's increased standard of life during that time. So there are some good, valid reasons for identifying them quickly, even with only one not-so-wonderful drug out there.

Obviously, we're looking also to see if our biomarkers can help identify pathways for new therapeutics. That's part of where the company fits in.

When did you start your company, Knopp NeuroSciences?

In 2004, not very long ago. I co-founded the company along with Greg Hebrank, who is currently president of the company, and Thomas Petzinger, who is the current CEO of the company. I have an academic appointment, so I'm not allowed to hold a true title position. I'm just a consultant for the company.

Was it just based on the preliminary data from that small, blinded set?

Pretty much. And in the larger, 200-set, we can see these peaks again — they keep popping up.

What are the next steps to take for the company?

The company is interested in putting out a diagnostic test — something that I was not going to create in my academic lab, so that was one of the reasons I wanted to create something outside of the academic world. So they're talking with Ciphergen to be a potential partner for such a diagnostic test. It could be that they market the test, or it could be that they have a CLIA-certified lab in which to perform the test.

Initially, I think it could be a mass spec-based approach. There really aren't mass spec-based diagnostics out there, but the time is nearing, I think, where that will happen. Ultimately, you'd like to make it an ELISA-based approach, because that's what every lab and hospital around the country does. So that's still also a goal for the company — to create an ELISA-based test. But I think a mass spec-based approach could be initiated more rapidly.

To make an ELISA-based test, you'd probably have to identify more proteins, and it'd be a multiplexed ELISA.

So the diagnostic test is the main focus of the company. They also have an option on a drug that's in Phase I for ALS. So they're doing a bit of due diligence for that.

Is that drug based on the same set of proteins?

No, it is not. That's different. But even for the Phase I, we're going to monitor the biomarker panel in those patients. We're interested in seeing how this will work in testing drug effectiveness.

What other projects are you working on for the future?

With SELDI research, we're working on the animal model of the disease as well, and trying to look for common markers and pathways in the transgenic animal model and humans. Right now every drug that's been tried in the rodent and shown to be of some benefit, and then tried subsequently in humans, has failed. We don't know why. So the idea is to find common pathways between the two, then target those within the rodent. Then if it works in the rodent, we might at least have some idea that it'll work in humans.

So we're doing that. We're also doing plasma samples in humans. We have our common set of both CSF and blood from the same patients. We want to find out if the markers are the same in blood and CSF also.

Then finally, I have a cohort of subjects that I take CSF and blood from every four months. So we're monitoring clinically the progression of the disease, and looking to see how the biomarkers change as the disease progresses.

The SELDI platform has worked for us. We're using other mass spec instruments too, but the SELDI has performed well. Especially with CSF, it's performed reproducibly and given us some good data and some good panels and proteins to go after.

Have you started discussions with the US Food and Drug Administration about this ALS test?

Only cursory. The goal is the definitely go down there and set up a more formal meeting within the next few months.

I think they're still undecided about this type of technology. I think they have an open mind, but I think they're undecided about these types of technologies because there's been nothing approved using them. So what will be the sort-of threshold level of information and data you need in order for them to provide that approval is unclear at the moment.

File Attachments
The Scan

Gone, But Now Reconstructed SARS-CoV-2 Genomes

In a preprint, a researcher describes his recovery of viral sequences that had been removed from a common database.

Rare Heart Inflammation Warning

The Food and Drug Administration is adding a warning about links between a rare inflammatory heart condition and two SARS-CoV-2 vaccines, Reuters reports.

Sandwich Sampling

The New York Times sent tuna sandwiches for PCR analysis.

Nature Papers Describe Gut Viruses, New Format for Storing Quantitative Genomic Data, More

In Nature this week: catalog of DNA viruses of the human gut microbiome, new dense depth data dump format to store quantitative genomic data, and more.