Discovering a Palette for Tracing the Brain

Q&A with Chester Mathis

September 27, 2010

Mathis thumbnailChester Mathis, a professor of radiology at the University of Pittsburgh, co-invented Pittsburgh Compound B (PIB), the first radioactively-labelled tracer compound that allows doctors to image amyloid plaques in the brains of living patients, using a positron-emission tomography (PET) scan. Mathis spoke to Jim Schnabel about the origin of PIB and the possibility of developing new PET tracers for Alzheimer’s and other neurodegenerative diseases.

JS: When did you start thinking about the development of PET tracers for Alzheimer’s?

CM: When I was a postdoc at the University of California at Berkeley in the early 1980s, I worked with a UC-Davis neurologist named Robert Friedland on the first FDG-PET study of how metabolism is altered in different brain regions in Alzheimer’s. Afterwards, Rob asked me to think about radio-labeling the dye Congo Red, because neuropathologists use Congo Red to identify amyloid plaques in brain tissue at autopsy. The idea was that we could use a radio-labelled Congo Red compound to image plaques in living people.

So I radio-labeled Congo Red and injected it into a rat—and it did not get into the rat’s brain. At that time, I was young and naïve and I didn’t really understand all the intricacies of blood-brain-barrier penetration.

But I learned about it. And in the 1980s and early ’90s I developed some PET tracers that bound to receptors in the brain for several major neurotransmitters, including serotonin and dopamine. And so in about 1994, I was at the University of Pittsburgh, and Bill Klunk, an assistant professor of psychiatry who was interested in doing some Alzheimer’s imaging work, introduced himself to me. He had some compounds that he thought could get into the brain and bind specifically to amyloid. It turned out we both liked fly-fishing, too. So we started discussing all this, and working on these agents, with very little funding and almost in our spare time. And we found that these compounds of his got into the brain much better than Congo Red, but not enough to make good imaging agents.

By 1999 we had tested hundreds of compounds, and had established detailed criteria for what makes a good imaging agent. But nothing we tested had worked very well. So we decided to look at thioflavin S and T, which like Congo Red are dyes used by pathologists to stain amyloid plaques in brain tissue at autopsy. Thioflavin T is structurally simpler than thioflavin S, and has only one charge [which impedes transport through the blood-brain barrier] so we started with that one, modified it to remove the charge, and tested it. And we soon found that it got into the brain five times more easily than anything we had previously tested. To our surprise, it also bound to amyloid with an affinity that was 100 times higher than ordinary thioflavin T.

So then with some NIH and other funding [including a grant from the Dana Foundation], we started to make derivatives that would meet all the criteria on our list. And out of that effort came Pittsburgh Compound B (PIB).

What’s the most important property that one of these imaging compounds should have?

It turns out that it’s not the efficiency with which it crosses the blood-brain barrier. It’s not its high affinity for amyloid. It’s not even a high specificity for amyloid versus other molecules. The most important property—and the hardest to find as we develop these things—is what we call the non-specific binding clearance, which is the speed with which it disappears from parts of the brain that don’t have amyloid in it.

Until that “background noise” goes down to an acceptable level, you can’t image amyloid effectively. You don’t have a high enough signal-to-noise ratio. And you can’t wait for more than 30 to 90 minutes after injection because the radioactive isotopes you’re using in these compounds have very short half-lives.

And even today, although we and others have developed amyloid-imaging compounds that are easier to work with clinically [because of their longer radioactive half-life], PIB remains the champion. It still has the best signal-to-noise characteristics following injection into humans.

PIB and the other amyloid-beta imaging agents now in clinical trials bind to the insoluble amyloid fibrils in plaques. But what about soluble and smaller clumps of A-beta, which are now thought to be more directly toxic to neurons?

We have thought about developing imaging agents for soluble A-beta, and other labs have too. But when GE Healthcare [the licensee for Mathis and Klunk’s amyloid imaging agents] asked my opinion of that, I told them to save their money and do something else. The amount of soluble A-beta in the brain of a cognitively normal person or even an Alzheimer’s patient is very low—only 1 or 2 percent of the amount in the insoluble fibril form. And most of the soluble A-beta is in monomer [single-copy] form, whereas many scientists claim that the true toxic species is the soluble dodecamer [12 stuck-together copies of A-beta]. The concentration of A-beta dodecamer in the brain is about 1/40,000th the concentration of insoluble aggregated A-beta in the plaques.

In principle you could develop an antibody that would be sufficiently selective for the free floating A-beta dodecamer. The problem is that antibodies don’t cross from the bloodstream to the brain quickly enough to use them for imaging with short-lived radionuclides. I’m not saying it’s impossible to do PET imaging of soluble amyloid species, but I would be astounded if anybody overcame this problem within the next few decades.

What about other proteins involved in Alzheimer’s, such as the phosphorus-modified tau protein that is more associated with the later stages of the disease?

We and many others are trying to develop selective imaging agents for tau. That’s the new frontier. But other than in the hippocampus and entorhinal cortex, tau fibrils are at a very much lower concentration than A-beta fibrils. And you can’t target them with something that binds generically to fibrillar proteins, because such an agent would bind also to the more numerous A-beta fibrils.

So far, in the literature anyway, no one has identified an agent that will bind selectively to tau, except for antibodies, which we can’t use. So we’re looking for chemical compounds that can get across the blood-brain barrier and show some selectivity for tau, and we and others are starting to develop some of these leads and radio-label them.

Fibril-forming proteins have been implicated in other neurodegenerative diseases, such as alpha-synuclein in Parkinson’s disease. Couldn’t you target them as well?

People do want to image alpha-synuclein in Parkinson’s and in other synuclein-associated diseases. But alpha-synuclein’s concentration in the brain is an order of magnitude less than tau’s—and tau is a tough enough challenge. We and others continue to work on this, but so far no alpha-synuclein-specific leads have been developed.

Can we look forward to amyloid- or tau- or synuclein-tracers for MRI, using an MRI-contrast agent such as gadolinium?

There’s been a big effort to do that over the past ten years. But the problem with gadolinium and other MRI contrast agents is that, to get them through the blood-brain barrier into the brain, you have to modify them in such a way that you tend to interfere with their ability to bind to the target of interest.

Gadolinium has seven different places that you have to bind to it and neutralize its charges. The end result is that it’s extremely difficult to make gadolinium-based small-molecule compounds that retain their ability to bind tightly and specifically to, say, amyloid plaques. Also if the gadolinium gets free of all these modifications it becomes toxic, particularly at the high concentrations you would need for amyloid imaging.

So gadolinium-based compounds are among what you might call the high hanging fruit—the hardest to reach. With radio-labeled PIB we picked the low hanging fruit. And we’re very happy we did.