Sunday, October 26, 2008

Broken Bottle Fight

Ya gotta love a broken bottle fight.

OK, that’s pure hype: I’ve heard no rumors of olfaction researchers actually resorting to smashing Pellegrino bottles and offering to rearrange each other’s faces at any recent symposia. But since Buck and Axel issued the first major installment in their Nobel-winning research on how the nose detects smells in 1991, two schools of thought in the science of smell have shared a major bone of contention.

Namely, is bigger necessarily better?

Here’s the rub: In 1991 Buck & Axel discovered a huge family of genes that encode the olfactory receptor proteins: A family of sensor molecules that sit on the outside of olfactory neurons in the main olfactory epithelium, deep in the nose, and detect the small molecules that constitute smells.

The OR family was one of those rare discoveries that suggest compelling answers to important questions. The sheer quantity of genes for the olfactory receptors, and the subsequently discovered fact that each olfactory neuron appears to express only one type of receptor, suggested how we detect so many complex smells: Olfactory neurons specialize in the specific odor-carrying molecules they respond to by virtue of carrying specialized detectors. Each receptor type matches a specific odorant — or, more likely, a small family of similar odorants [1]. The odorant fits the receptor much as a key fits into a car’s ignition; the receptor fires up its nerve cell, which in turn reports that odorant’s presence via an electrical jolt to the brain’s olfactory bulb.

The idea had legs for two reasons: First, in this model the receptors would act a lot like other receptor proteins and enzymes previously described in detail, and that’s always a reassuring sign you’re not going off the deep end. Second, it offered one explanation for why we use dogs to search for lost people: They can smell more things than we can.

Humans have about 900 olfactory receptor genes; dogs, around 1,200; and mice, 1,500. Now, by biological standards those numbers aren’t terribly different; it really gets interesting when we account for pseudogenes — genes that have a mistake in them that makes them non-functional. In humans, at least a whopping 63 percent of the receptor genes are such molecular dead weight — and so the actual numbers of working receptor types for each species are 350 for humans and roughly 1,000 apiece for dogs and mice.

Aha, a lot of us thought: That’s why they smell better than we do (after a fashion). They’ve got more receptors, ergo they’re detecting smells we can’t even imagine.

That’s when people started busting fizzy-water bottles against lecterns. For scientists love nothing more than to shoot down a simple idea for being too simple: People make careers from demolishing comfortable assumptions.

And some data seemed to do exactly that. Humans had far fewer total types of receptor; but if you looked at a more detailed structural level, much of the mouse "superiority" began to seem redundant. The mouse genes comprise 228 families of similar receptors, and it turned out that humans have a comparable number of gene families — we just have many more "only children" within each family. What this might mean is that mice carry around a bunch of receptors that detect pretty much the same stuff, and therefore, humans’ smaller repertoire of receptors can cover the same ground.

Enter Anna Lesniak and colleagues at the Polish Academy of Sciences: In the Sept./Oct. 2008 issue of the Journal of Heredity, they present a study of a group of 35 detector dogs from different scent specialties (human identification, explosives detection, drug detection, and tumor detection), that may rescue the more-is-better hypothesis after all.

The group tested these dogs for polymorphism — genetic diversity — in five randomly selected olfactory receptor genes. They asked: If a dog inherits two different versions of a receptor gene from its parents, does it carry out a scent-based task better than a dog who inherits the same version from each parent?

The situation turned out a bit differently for each of the five genes: For three of them, the version didn’t matter to the dog’s ability to carry out the scent task. But for the other two, an interesting, if not yet conclusive, pattern emerged [2].
* Dogs with two versions of one of these genes tended to be among the best performers for a scent task based on their specialty.
* The gene seemed to have a "better" version and a "worse" version: With two copies of the former, the dog tended to rank on average among the best performers; two copies of the latter put him among the slackers.
* It’s not clear whether having two versions is actually better than one: It may be simply that the existence of two versions gives you a better chance of having the good one.
* With the other gene, a similar pattern emerged, but only in the explosives dogs — lineup-identification dogs, tumor detectors, and drug dogs showed no difference.
* Taking the above together with the three genes that showed no difference, in no case were the dogs with two versions doing worse: They were either better off or indistinguishable.

So here’s the simple, if tentative, explanation: More receptors — even more versions of roughly the same receptor — give you a better chance of having a more-sensitive version that puts you ahead of the game in scenting ability. So the "redundant" receptors in the mouse and dog genomes may not be so redundant after all. (And yet another nail goes into the coffin of the quaint 19th-century notion that inbreeding produces superior working dogs, but that’s another issue.)

There are always provisos:
* The researchers aren’t completely satisfied with their scenting-ability tests, and they’ve got a point. The specialty based performance test hinges on the dog/handler relationship, and so doesn’t completely isolate pure scenting ability — though creating a more abstract, if lab-friendly, test might miss some important nuances in how the scent task unfolds.
* Worse, a second test, based on the dog’s ability to find treats, showed no effect for any of the genes, and its results didn’t correlate all that well with the specialty-specific tests. I think they might have gotten themselves into trouble here by using a poorly thought-out test, but they did use it, and its lack of a result is awkward.
* All the negative results could devolve from the simple expedient that, in the case of the negatives, those particular tasks simply don’t require those receptors. But we can’t rule out that the few positives were a fluke.
* For our larger, "is more better" question, we have to keep in mind there’s an apples-to-oranges going on here: We were talking about having more receptor types within each family of genes, while this experiment actually suggests what happens when you have more diversity within one type. My bet is that won’t make a difference to the validity of the argument, but it’s a possibility.

The good news is that Lesniak and her crew are well aware that their result is preliminary, and are planning to sharpen up their methods in future reports. So stay tuned.

But I wouldn’t blame conference organizers if they went for the plastic Pellegrino bottles, at least for now ...

[1] The number that people have thrown around is about 10 odorants per receptor, for a total of about 10,000 odorants. The only problem is that it doesn’t seem to come from any published scientific findings (even though the Nobel press release cited it). But here’s a wild one that suggests the whole issue is moot (or at least a vast oversimplification): In a 1993 theoretical biophysics paper that I only just found out about, Doron Lancet and posse calculated — calculated, mind you — that the olfactory system needs 300 to 1,000 receptors with the average sensitivity of the olfactory receptors to recognize, well, everything.

[2] For the wonks, they carried out their analysis with an ANOVA and subsequent Tukey testing, so they’ve presumably ruled out false discovery.


Malin Sandström said...

"Namely, is bigger necessarily better?"
The obivous answer should be: it depends. It depends on the type of task, the difficulty of the task and on which odorants are there.

I see it as an issue of resolution. To be able distinguish between two very similar odorants, you must have at least one receptor that recognises the difference - that is, binds just to one of the odorant versions or perhaps binds sufficiently worse to one of them that upstream parts of the brain can tell the difference.

If you have more receptors, on average you will have a better chance at having at least one that can make the difference for any given task. BUT. The receptor repertoires are evolved differently for different animals, presumably to optimise detection and discrimination of behaviourally relevant odors.

To really be able to say that you're comparing across sizes (which would mean different sorts of animals), you'd have to find and use an odorant that was equally important, evolutionary speaking, for all of their species - in all concentrations. And you'd probably have to train them on it as well, since training improves performance for the individual and some of them may have encountered it more often for others.

The above is perhaps not impossible, in theory, but at least really hard... and I'm not aware of any studies/papers where this has been done. And even if it had been tested, it would not necessarily be true for any other odor that the tested one, no?

Coming back to mice and humans, I would guess that humans are better generalists, but mice are far better at predator detection and tricky discrimination tasks for their relevant odours - essentially, that they have better resolution within the odor space that they have use for. So those extra receptor siblings that the mice have are probably not useless at all, quite the opposite.

The polymorphism study seems very interesting, but I think they're mixing up their cards a bit since they have both "detection" dogs and "discrimination" dogs. Finding very small amounts of a drug or explosive against a background that contains no confusing similar substances just reqires you to be very sensitive (i.e., if you have two copies of the best receptor, ypu're likely better off). In finding one specific human in a lineup of others - that is discriminating between complex, similar odors with lots of odorants - one slightly better receptor variant might not help you very much, but better resoluting might. All depending on the receptors of course, since they picked just five out of a thousand I would not take their results as a proof in either direction. (Also, the latter task is way more cognitively demanding - you'd have to hold the different
person odors in workning memory to be able to compare them)

Well, that was long :-) (I'm a PhD student in computational neuroscience modelling olfaction, I've had some time and reason to think about this). Hope some of it made sense, it's quite early here yet and I'm drinking my morning coffee as I write...

Ken Chiacchia said...

Well, I think you're right about the complexities. But stay tuned -- that biophysics paper I cited in the footnote will be the subject of a near-future blog. The trick is not thinking about olfactory receptors as a static bunch of locks awaiting specific keys, but as a spectrum of locks awaiting unidentified odorants that will be important to that animal's life experience.

Thanks for reading,

Malin Sandström said...

Actually, I think the "lock and key" metaphor is a bad one and in fact can be rather harmful for understanding (I've seen quite a few examples of people getting their understanding of olfaction very wrong based on it). After all, keys and locks fit each other very closely and in most people's experience one key fits one lock only and vice versa. While the biological situation is more like several 'keys' fitting in several 'locks' and 'locks' being opened by several 'keys' (or perhaps opened by several partly overlapping parts of the same 'key'). The population code is the important part, not the fitting of a single odorant into a single receptor.

There are other interesting papers about the properties of theoretical olfactory receptor sets, besides the Doron Lancet one. Let me know if you want some pointers :-)

Anonymous said...

I cringe a bit thinking about it, but they may be able to nail down how the dogs are reacting better with a bunch of dogs straightjacketed and put in an MRI machine looking at their brains while a jet sprays a particular odor past their nose once every 30 seconds or so. Forgive the run-on sentence. There's a lot of this going on with humans recently; e.g. Kiehl's attempt to locate what makes sociopaths different by giving prison inmates MRIs. (I think the name is Kiehl. Problem is, there's also a climatologist named Kiehl, and I may be conflating two people here.)