Let’s start from first principles: most drugs mess something up. More elegantly, most drugs inhibit some enzyme’s activity or block some receptor’s binding site. Proteins are generally pretty well optimized at what they do, so it’s a lot easier to block their activities than it is to speed them up. (There are rare exceptions).
And if you’re going to target an enzyme with a small molecule inhibitor, you’ll do just that – find a small molecule that fits into the active site of the enzyme and gums up the works. In a few cases, we know of drugs that bind to other sites on the protein and mess up the active site indirectly, by altering the whole conformation of the protein, but most inhibitors are in or near the site where the natural substrates bind.
This background is what makes a paper in the latest Nature so odd. A large multicenter academic team has been studying inhibition of beta-amyloid formation by some known anti-inflammatory drugs. Beta-amyloid is cleaved out of a larger protein called APP, and the proteases that do the chopping have long been drug discovery targets. (Mind you, when I was working on Alzheimer’s disease in the early 1990s, we still didn’t know which enzymes those were, which made things rather difficult).
The key enzymes in that process are known as beta-secretase (or BACE) and gamma-secretase. The effect of the various known drugs has seemed to be more tied to the latter, although no one’s been sure just what the mechanism is, since none of them seem to be actual gamma-secretase inhibitors when you study them in isolated systems. The current work has turned some of these drugs into photoaffinity probes to try to find out what they’re really targeting.
(For those outside the field, photoaffinity probes are derivatives of some compound of interest, where some special UV-light-absorbing group has been attached off the back end. These photoaffinity groups are innocuous under normal conditions, but they turn into crazily reactive intermediates when they’re irradiated, and will then form a bond with the first thing they see. The idea is that you let your photoaffinity-modified compound find its usual protein targets, then you turn on the ultraviolet lamp. The reactive group does its werewolf thing and forms a permanent bond to the protein its next to. You can then search for the strangely labeled proteins, and you’ve found what the drug of interest was binding to. When it works, it works, although it’s a lot harder than I’ve made it sound).
When they labeled various gamma-secretase systems, all the way up to whole cell extracts, they found that the anti-inflammatories did not actually seem to bind to gamma-secretase at all: it wasn’t labeled. Based on earlier enzyme studies, that’s probably what they expected. But what was labeled was a real surprise: the APP protein, the substrate of the enzyme. Looking more closely, it appears that the compounds bind right to the part of APP that gets cleaved into beta-amyloid, and inhibit the enzyme’s action that way.
That, as far as I know, is pretty much a first. Update: the closest thing might be the mechanism of the antibiotic vancomycin, which binds to the weird D-Ala-D-Ala section of two of the components of the gram-positive bacterial cell wall and prevents them from being used.). This isn’t something that most drug discovery programs would try a priori, that’s for sure. For one thing, we have a hard time getting small molecule to bind to protein surfaces. Active sites inside proteins are our usual speed, because those are more defined cavities which are optimized to hold reasonably small substrates. But sticking to some outer part of a protein, while it does happen, is very hard to do in a targeted fashion. (We’d love to learn the trick, if there’s a trick to be learned – inhibiting protein-protein interactions with small molecules would open up a whole new world of drug targets).
Another reason that no one targets substrates instead of enzymes is that there’s generally a whole lot more substrate floating around than there is enzyme. Imagine someone throwing a hungry piranha into a pond full of goldfish. Which is the more efficient way to defuse the situation - armoring each goldfish, or disabling the piranha? That metaphor just occurred to me, and while a bit weird, it’s actually reasonably close to the situation you have with a protease enzyme and its substrates - if you want to get fancy, you can imagine that the piranha only likes certain types of goldfish, and only bites them in select spots.
But on the other side, there's also a reason why protecting the substrate might actually help out in some situations. Proteases tend to have multiple targets, so inhibiting them can also disrupt pathways that you didn't want to touch. Binding to the one substrate you care about might give you a much cleaner profile, compared to shutting down everything.
So you have to wonder what this result means. Have we been missing a whole range of potential enzyme inhibitors by ignoring things that bind to the substrates? I'm not convinced of that yet, but I am interested. I still have a hard time believing that we can do a good job targeting particular protein surfaces, at least at present, and I can't help wondering if there's something odd about that beta-amyloid sequence that makes it more likely to pick up small molecule interactions. (It certainly excels at picking up interactions with itself if it gets a chance, which is the whole problem). It's still going to be a lot easier to inhibit enzymes directly rather than bind to their targets, but it's worth exploring. We need all the ideas we can get.
College chemistry, 1983
The 2002 Model
After 10 years of blogging. . .
1. SP on June 17, 2008 9:55 AM writes...
Vancomycin blocks the incorporation of a cell wall substrate by binding to that substrate.
Permalink to Comment2. Mark M on June 17, 2008 9:56 AM writes...
I am reminded of Dervan's polyamides which are designed to bind to sequence specific DNA and modify the action of proteins (transcription factors) whose "substrates" are the DNA.
I may be wrong but, to my knowledge, agents targeting the DNA have made it farther in the clinic than those targeting the transcriptions factors (ie, with DNA mimetics). It seems you typically need to cover a lot of surface area though for specificity to accomplish this and I'm not sure if you are in the *small molecule* arena at that point (this is a subject term, I realize).
We are used to thinking of proteins as the binding target of drugs; but it is not the case with many bioactive natural products (eg, calicheamicin)
Permalink to Comment3. Chrispy on June 17, 2008 1:33 PM writes...
Often in the process of making a robust screen we eliminate parts which could actually be targets. So we use enzymes instead of cells and we use a peptide instead of the real substrate. This is classic HTS: narrow down the variables so you minimize hard-to-explain outliers.
One of the things we miss in doing this is the opportunity to get lucky in a missed-the-target-but-hit-a-different-one kind of way. We in drug discovery have become completely arrogant: "If we just had a small molecule which specifically hit X..." Small molecules really aren't specific, so we settle for compounds which hit the target enzyme very hard and don't hit problem targets like HERG channels too much. The vast majority of exosite-type inhibitors (and I'll bet the amyloid-binding compound, too) have real issues with specificity because they tend generally to bind protein. Active sites are nice because of their built-in selectivity for their substrates which can translate (in theory) into selectivity for small molecules.
In contrast, antibodies have really nice specificity. I wonder if the chemists are jealous of the biologists? Antibody folks don't even worry about off-target effects. Of course, they see enough which is unexpected from the on-target effects to keep them scratching their heads...
Permalink to Comment4. Retread on June 17, 2008 2:48 PM writes...
It's a very nice story, that aggregates of misfolded proteins cause neurologic disease -- the senile plaque (beta amyloid) and the neurofibrillary tangle (phosphorylated tau protein) cause neuronal dysfunction and death in Alzheimer's, the Lewy body (composed mostly of alpha-synuclein) in dopamine containing neurons causes Parkinsonism, superoxide dismutase aggregates cause motorneuron degneration in familial amyotrophic lateral sclerosis. E.g. the aggregates are the smoking gun causing disease.
This is extremely simplistic thinking (but has been characteristic of the field until recently). One can regard an abnormal structure seen on a microscope slide in at least two other ways -- (1) as a pile of spent bullets, used by the cell to defend itself (2) as a tombstone -- part of the dying process of the neuron, but unrelated to the cause of death.
The evidence for the smoking gun theory in all 3 diseases mentioned is at best controversial. Morever there is evidence for (1) -- see below. So efforts to find small molecules to break up the aggregates (which is ongoing) might be successful but ineffectual therapeutically or actually harmful.
See [ Proc. Natl. Acad. Sci. vol. 104 pp. 3591 - 3596 '07 ] for the protective effects of the neurofibrillary tangle in Alzheimer's disease.
See [ Proc. Natl. Acad. Sci. vol. 101 pp. 17510 - 17515 '04 ] for the protective effects of the Lewy body in (an animal model of) Parkinsonism.
See [ Cell vol. 104 p. 586 '01 ] for the protective effect of axonal spheroids in motor neuron disease.
See [ Proc. Natl. Acad. Sci. vol. 105 pp. 7206 - 7211 '08 ] for the protective effects of amyloid formation (but in the rather removed yeast prion system).
There's a lot more coming out but this should give you an idea.
None of this should be construed as denigrating the sort of work cited above -- but it is something to keep in mind. Alzheimer's is an awful disease and the drugs we had for it when I retired in '00 were lousy, despite a lot of drug company and academic hoopla to the contrary. Desperate families were led to expect a lot from the drugs (Cognex, etc. etc) but the effects (perhaps a slightly slower rate of decline) were clinically meaningless. The results were always the same -- nothing dramatic. It was very hard to watch and very cruel for the families. This was the experience of every clinical neurologist of my acquaintance.
So as Mao said "let a hundred flowers bloom". Even if only one proves useful, it will be worth it.
Permalink to Comment5. SP on June 17, 2008 3:22 PM writes...
We're more heavily weighted towards phenotypic screens these days- whole cell, tissue, or even organism assays. Chrispy sounds like another Chris (Lipinsky) who made the same points when I saw him a couple years ago.
Permalink to Comment6. Bill on June 18, 2008 8:27 AM writes...
Greatly appreciate your discussions. Speaking to some of the concerns in #4 comment, there is at least one study that demonstrates an appropriate activity (reduced oligomerization of soluble beta amyloid), does not affect APP levels, and, perhaps most importantly, demonstrates cognitive performance improvements. OK, it's just mice. I bring it to your attention on the off chance you're not familiar with the work, and in the hope you might venture an enlightening comment about it. See: Pasinetti, et al., "Valsartan lowers brain β-amyloid protein levels and improves spatial learning in a mouse model of Alzheimer disease" (The Journal of Clinical Investigation Volume 117 Number 11 November 2007; http://www.jci.org)
Permalink to Comment7. retread on June 30, 2008 8:13 AM writes...
12 June Nature finally arrived. It's an intriguing paper. The work should be repeated with completely different enzymes and their known inhibitors and substrates. If the substrates are also labeled by similar biotinylated photoactivatable enzyme inhibitor derivatives, the results might be an artifact of the technique. Looking at their figure #1, it's clear that the derivatized compounds (Fenofibrate, Tarenflurbil) have at least twice the mass and size of the parent.
Also gamma secretase is an enzyme activity found within the membrane, a class we've only recently been able to study (because we didn't realize they existed). Almost all enzyme work up to now has been done with soluble (e.g. non-membrane embedded) enzymes. A good test case for the idea in the previous paragraph would be another membrane localized enzyme such as the one which cleaves SREBP
Bill -- I've never been a fan of animal models of human learning and intelligence. There was a paper over 30 years ago (whose citation I've lost) showing that gerbils with their entire cerebral cortex removed can still build a nest. Also someone made the crack that Newton and Einstein probably wouldn't have been great at running mazes.
That's what makes animal research on neurologic diseases so difficult -- we're trying to study exactly what separates us from them. It's all we've got unfortunately, but I don't put much stock in the results.
Permalink to Comment