I got several e-mails yesterday about a new study about the molecular mechanism underlying circadian rhythms in mammals (“You gotta blog about this!”), so, thanks to Abel, I got the paper (PDF), printed it out, and, after coming back from the pool, sat down on the porch to read it.
After reading the press releases, I was in a mind-frame of a movie reviewer, looking for holes and weaknesses so I could pounce on it and write a highly critical post, but, even after a whole hour of careful reading of seven pages, I did not find anything deeply disturbing about the paper. Actually, more I read it more I liked it, my mood mellowed, and I am now ready for a long rambling post about it – I have no idea how is it going to end, but let’s go on a journey together….and let me start with a little background – the Big-Picture-kind of background – before I focus on the paper itself.
Period is Important
We live in a 24-hour world. No matter what the endogonous period of our clocks may be in the artificial environment of the laboratory, all of our biochemistry, physiology and behavior is forced into a 24-hour cycle. So, at first thought, the actual value of the freerunning period may not sound that important. But is that correct?
Circadian rhythm is defined as being “around 24 hours”. But, how much is “about”? The circadian range is usually taken to be between 16 and 32 hours, but extremes are rarely found in nature. Most organisms do exhibit cycles close to 24 hours. But, there are some fine distinctions. It would take a long, tedious and systematic review of the literature (and perhaps actual lab work!) to test this notion, but I have a hunch that ecology of the organism is correlated with how close the period is to 24 hours. So, this is my speculation (about animals – it may not work for plant, fungi or protists) and you do not have to believe me at all, but let me try to explain my reasoning.
For an animal that is nocturnal, non-migratory and spends most of the time underground in its burrow, it is advantageous to have a clock that is as close to 24 hours as possible. Thus, if it is forced to stay underground for several consecutive days and nights, it will still remain in sync with the planet. The clock of such an animal needs to be somewhat inert, i.e., very difficult to phase-shift by light. If every time it comes out and sees the moon its clock shifts by several hours, the poor animal would be constantly jet-lagged!
According to the math behind the Phase-Response Curves, a clock with a period very close to 24 hours will be harder to phase-shift. And, lo and behold, nocturnal burrowing non-migratory animals, e.g., mice, rats and hamsters have periods very close to 24 hours (i.e., 24 + 15 minutes or so) and their Phase-Response Curves (PRCs) are of the Type I, exhibiting a very low amplitude of the curve. The greatest phase-shifts induced by light pulses are about 2 hours. Also, a phase-shift of the light cycle does not immediately shift the clock – it takes several days of “transients” until the clock “catches up” with the shifted environmental cycle and locks onto the new phase.
On the other hand, diurnal, surface-dwelling, migratory animals need to have very flexible clocks. Especially during migration they need to be able to always be on the “local time” as they use it for Sun-Compass orientation. Thus, they need a much lower stimulus (i.e., the duration and/or intensity of the light pulse) to phase-shift and the shifts can be quite large – over 12 hours in some cases. Also, there are no transients – the complete shift is completed over a period of a few hours. Having such a (Type 0) PRC is easier if the period is farther away from 24 hours, e.g., 20 hours or so. The converse is also true – having a Type O PRC allows one to have a greater deviation from 24 hours, as a daily shift of 4 hours is easy to accomplish. Unfortunately, not too many PRCs were done on such animals, but Asian Migratory Quail, for instance, has a freerunning period of about 20-22 hours and exhibits a Type O Phase-Response Curve.
So, if I am correct, the actual value of the period matters. Although the organism is exposed to a 24-hour cycle every day and is thus entrained to this cycle (i.e., it is naturally cycling at 24 hours), the endogenous period is important – it determines the phase and determines the flexibility of the clock, i.e., its ability to resist phase-shifting effects of light. If I am correct in my speculations that the period correlates with ecology, this would suggest that the period is an evolved adaptation – a notion difficult to test in any trait, but not such an outlandish proposition after all.
Determination of Period
Period of a circadian rhythm is quite strongly determined by genetics. Most of the clock mutants are period mutants. In other words, naturally occuring mutations of clock genes usually result in changes of period. But, what are clock genes?
In vertebrates, there are four “core clock genes” coding for “core clock proteins”: PERIOD (PER1 and PER2), CRYPTOCHROME (CRY1 AND CRY2), CLOCK (CLK) and BMAL (sometimes also called MOP).
In fruitflies, BMAL is called CYCLE (CYC) and, instead of CRY, the core clock protein is TIMELESS (TIM). TIM has no known circadan function in vertebrates. In fruitflies, cryptochrome is part of the photoreceptive pathway into the central brain clock and may or may not be a part of the core clock machinery in peripheral oscillators. You can watch and compare cool animations of the Drosophila and mammalian clock mechanisms if you click here.
PER and CRY (or PER and TIM in fruitflies) interact with each other. CLK and BMAL interact with each other. Let me ignore all the fine details and put it simply: the PER-TIM dimer regulates expression of the CLK-BMAL dimer and the CLK-BMAL dimer regulates expression of the PER-TIM dimer. Thus, two interlocking feedback loops are the underlying mechanims of the circadian clock.
But that is not all. There are a number of other proteins involved, in supporting roles. They regulate the dimerization between pairs of clock proteins, regulate the re-entry of clock proteins back into the nucleus, or regulate their degradation. Each one of these roles is important in determination of period as the rates of dimerization, nuclear entry and degradation are what really determines the rate at which the clock runs – its period.
One of the important proteins in a supporting role is Caseine Kinase 1epsilon (CKI). In fruitflies it is called DOUBLETIME (DBT). It occurs as a natural mutation in fruitflies, hamsters and humans. It was first described in 1988 by Ralph and Menaker in hamsters and they dubbed it TAU (Greek letter Tau is used as a symbol for freerunning period in mathematical models).
The period of a wild-type hamster is almost exactly 24 hours. In Tau mutants, the period is about 20 hours. Heterozygotes are in-between at 22 hours. The Tau-mutant hamster immediately became one of the most important laboratory models for studies of mammalian circadian rhythms (and photoperiodism) and has yielded much important information even during the decade or so it took to finally characterize the mutation at the molecular level.
In humans, the Tau mutation is the underlying cause of the Familial Advanced Sleep Phase Syndrome – whole families naturally wake up around 4am and fall asleep around 7pm.
So, the clock in Tau-mutants runs faster. It is to be expected that some part of the process (transcription, translation, post-translational modification, dimerization, nuclear entry or degradation) is speeded up. Yet, the earliest experiments on CKI suggested that degradation of Period slows down. It was shown that CKI phosphorilates PER protein, which allows it to re-enter the nucleus and get degraded. It was assumed that a mutation in CKI would result in a less phosophorilated PER, thus slower re-entry or degradation. Thus, for instance, clogging the nucleus with undegraded PER protein was supposed to phase-advance the clock.
This paper turns everything upside down
Yesterday’s paper (“An opposite role for tau in circadian rhythms revealed by mathematical modeling” by Monica Gallego, Erik J. Eide, Margaret F. Woolf, David M. Virshup and Daniel B. Forger) in PNAS challenges this understanding of the role of CKI.
What is really interesting about the paper is that the work was initiated by Daniel Forger who is a mathematician who models circadian clocks. His models predicted that Tau mutation has an opposite effect – it over-phosphorilates PER which then re-enters the nucleus and degrades faster. He happened to talk about it with the Utah group who had unpublished data from their cancer research that also indicated that Tau mutation over-energizes PER. So, together they set out to test this idea both mathematically and in the laboratory.
What they found is exactly what they expected – in cultured rat cells, mutated CKI added more phosphate groups to PER than the wild-type CKI, and as a result, PER degraded more rapidly.
So, how to reconcile these data with older data suggesting that CKI under-phosphorilated PER? Well, the old experiments were conducted in vitro with a neutral substrate – caseine. So, in this new paper, they repeated this experiment in their lab, but this time in cells, and, sure enough, CKI mutation reduced the phosphorilation of caseine. It also reduced phosphorilation of BMAL, another substrate. What is the difference then? CKI mutation has different effects on different substrates: PER on one hand and BMAL and caseine on the other.
There is another substrate for CKI – the Wnt pathway. Here, they found no effect of Tau mutation at all, which is a good thing as the Wnt pathway is a developmental pathway and the Tau mutation, if effective here, would result in serious developmental problems:
The fact that the tau mutation selectively increases phosphorylation of PER proteins may be under strong genetic selection. A broad gain-of-function mutation in CKI is unlikely to be tolerated because CKI regulates the activity of multiple pathways. If CKItau was a gain of function in the Wnt pathway, significant developmental abnormalities in the tau hamster would have been expected.
So, the Tau mutation is a loss-of-function mutation for caseine and BMAL, null-mutation for Wnt pathway, and gain-of-function mutation for PER. In the artifical world of the test-tube, CKI adds phosphates to PER in places in which it does not do so inside the cell. Thus, CKI is a site-specific phosphate donor – when particular sites on PER receive a phosphate group, the protein degrades more quickly, thus speeding up the clock:
If the CKItau mutation hinders phosphorylation of these nonphysiologic sites, that could explain why in vitro it has decreased overall activity. In vivo, the phosphorylation of irrelevant sites may be suppressed by (i) modification of CKI activity due to changes in its phosphorylation state and (ii) the fact that PER2 is in a multiprotein complex interacting with other circadian regulators, including cellular phosphatases. Our experimental data demonstrate that the tau mutation increases the rate of degradation of PER proteins by increasing site-specific phosphorylation, and the modeling studies indicate that this increase can be the cause of a faster circadian period.
I am not going to bore you with the details of a series of additional experiments they performed which strengthen one or another aspect of their hypothesis, but let me highlight a couple of their statements from the Discussion:
This finding supports the conclusion that nonintuitive predictions provided by quantitative analysis can provide important insights into basic biological mechanisms. It also provides a cautionary tale regarding extrapolation of in vitro phosphorylation sites to in vivo activities.
What is in vivo is pretty subjective, I guess. From my perspective, anything that is not done in whole, live animals is automatically called in vitro, thus test-tube, cell culture, tissue culture and organ culture experiments are all called in vitro in my world.
Modeling can lead to unexpected and testable predictions about mechanism. Here, modeling leads to the unexpected finding that, although the tau mutation creates a partial loss of function on standard substrates, it is a gain of function in the circadian clock that speeds up the clock by increasing the phosphorylationdependent degradation of PER proteins. Despite the rich history of experimental results in circadian rhythms that were inspired by, interpreted through, and incorporated into mathematical models, model-based experimental approaches are almost never used in studies of the genetic and molecular biology of circadian rhythms. The current study makes clear that incorporating modeling as part of a systematic approach to circadian research can strengthen both experimental design and conceptual analysis.
This is an interesting statement. On one hand, so much of the work in chronobiology is testing, or at least utilizing the mathematical models underlying Phase-Response Curves (and their various derivatives). Colin Pittendrigh’s modeling of PRCs, models based on limit cycles, and Arthur Winfree’s topology-based models, are the bread and butter of chronobiology. So, what is this about lack of use of mathematical models in current research? One possible answer is that times have changed and that today’s scientists pay less attention to math then yesterdays’s scientists did. Second possibility is that molecular geneticists are much less likely to look at math than the organismal and systems biologists.
But there may be another explanation and it depends on what counts as a mathematical model. The models I listed above were developed not by mathematicians, but by mathematically inclined biologists. They did the math, that’s for sure, but they then translated the math into plain English, drew simple graphic representations of the modesl and wrote simple instructions for “do-it-yourself” tests of the models which mathematically-challenged biologists could use in their daily work. Apart from being mathematical models those were also strong conceptual models. Finally, they were not typical mathematical models – no multiple lines of calculus formulae. They were general models that are useful for studies of all clocks in all organisms no matter how the molecular details may differ between them.
On the other hand, many calculus-based math models I have seen in the past 10 years or so are reactive, not predictive. When geneticists discover something, mathematicians insert new information into the model by tweaking the parameters until the model runs the way it should. Those models are also too focused on details of particular molecules in particular organisms – they are not depicting universal laws of nature, they are just describing the current knowledge about a particular system – there is really no need for geneticists to look at those as they are the leaders and mathematicians are followers.
So, it is exciting that a math model discovered an error in molecular work, but perhaps it has something to do with the way mathematicians approach modelling that makes them useless for the day-to-day work of molecular biologists.
One goal of the study of human circadian rhythms is the development of tools to manipulate the clock to relieve human suffering, in the form of depression, insomnia, and chronic fatigue. Inhibitors of CKI activity have been proposed as interventions in these disorders, but predictions of drug effects have been based on the models incorporating the hypothesis that kinase loss-of-function mutations speed up the clock. Our data suggest that CKI inhibitors may have effects on human circadian rhythms opposite from those predicted by previous models.
It is interesting that they do not mention Familial Advanced Sleep Phase Syndrome anywhere in the paper, although it is the expression of Tau mutation in humans. I also like the way they phrased this passage. There is no indication they are thinking about gene therapy. They are talking about drugs. Nice and sophisticated for a change! Just because they have discovered something about a gene does not mean that intervention has to be at the level of the gene. The knowledge about the molecules should have a ripple effect up the levels – informing the way we think about the effects of the mutation at the level of cells, tissues, organs and whole organisms interacting with the environment.
The same way, it can inform the way we try to design treatments. While gene therapy is potentially one such avenue, it does not appear to be in the cards in the nearest future. Development of drugs is much more feasible.
Higher up, better knowledge of the underlying mechanism may refine our treatments at the behavioral level – just what kinds of exposure to light, exercise, scheduled meals and/or pharmaceuticals (melatonin?) could help a person with Familial Advanced Sleep Phase Syndrome get in sync with the rest of the society.
Or, we should step back and think – is this really a disorder, or should it be treated as such in the modern world? Perhaps we can, as a society, accept that different people function on different schedules and learn some basic manners on how to deal with this diversity – not to expose your brother to noise or light while he is asleep, not to insist that everyone should be in the office from 9 to 5. Perhaps allowing people to set their own schedules can actually raise productivity as everyone is fully rested, lucid and happy.