A Pacemaker is a Network

This is going to be a challenging post to write for several reasons. How do I explain that a paper that does not show too much new stuff is actually a seminal paper? How do I condense a 12-page Cell paper describing a gazillion experiments without spending too much time on details of each experiment (as much as I'd love to do exactly that)? How do I review it calmly and critically without gushing all over it and waxing poetically about its authors? How do I put it in proper theoretical and historical perspective without unnecessarily insulting someone? I'll give it a try and we'll see how it turns out (if you follow me under the fold).

Clock Genes - a brief history of discovery

Late 1990s were a period of amazing activity and rate of discovery in chronobiology, specifically in molecular basis of circadian rhythms. Sure, a few mutations resulting in period changes or arrhythmicity were known before, notably period in fruitflies, frequency in the fungus Neurospora crassa, the tau mutation in hamsters and some unidentified mutations in a couple of Protista.

But in 1995, as the molecular techniques came of age, flood-gates opened and new clock genes were discovered almost every week (or so it appeared).

First, Amita Seghal discovered timeless in fruitflies, Joe Takahashi discovered clock in the mouse, white collar was discovered in Neurospora, the kaiABC cluster in cyanobacteria, toc (originally named rigui by the Japanese team that first found it) in plants...and that was just the beginning.

Earlier attempts to find period in mammals by using primers against the fruitfly variant of the gene were unsuccessful in several labs, but by changing the approach and using clock as the entry into the system, it was discovered that mammals, too, have period and not just one but three copies of it. And clock was found in Drosophila. Both fruitflies and mammals were found to have another clock gene called cycle (Bmal in mammals) and another one (not a core clock gene but an important modifier) called doubletime in fruitflies which turned out to be our old friend the hamster's tau, actually known from before as kasein kinase A-epsilon. An early find in the Xenopus frog, nocturnin, turned out not to be as important as initially thought. It was found that although timeless is present in mammals, it has nothing to do with clocks - its role is taken by cryptochrome. This discovery prompted drosophilists to take a look and, lo and behold, fruitflies have cryptochrome as well, but, except perhaps in peripheral clocks, it does not work as a clock gene, but as a photoreceptive molecule that provides light information to the clock. This prompted mammalian researchers to check if cryptochrome also acts as a photopigment in mammals. This prompted others to look at other potential photopigments and they discovered melanopsin. Over a few years, there was a bitter and frantic research arms-race between the "melanopsin mafia" and the "cryptochrome mafia" which the former won and the latter gracefully conceded. All the genes found in mammals were also confirmed to be working in all the other vertebrate classes. The discovery of E-boxes paved the way for the study of outputs, i.e. how the clock genes time the expression of other genes in the cell. The fruitfly folks discovered dozens of secondary genes (some with nifty names like elvis and disco). It was a very exciting time and everyone was ebulient. And then it all stopped as suddenly as it all started. No more clock genes! It was over - we have apparently found them all. Now what?

The two ways of thinking about the Clock

By the 1990s, after decades of exciting and prolific chronobiological research, people were starting to bump into walls. As early as 1960 it was lamented that we did not know much about the inner workings of the clock inside the cells. In order to make much progress, we needed to get the handle on the genes. Without genes, some problems were becoming so untractable (e.g, tidal, lunar and circannual rhythms and photoperiodism) that by the early 1990s researchers were leaving them in droves. Instead, they learned the new molecular techniques as they became available and started discovering clock-related genes.

Genes are sexy - they generate buzz in the media. Clock-gene discoveries were touted as runners-up two years in a row in Science Magazine's lists of discoveries of the year. Buzz brings in money - new NSF Centers for Biological Timing were founded at several US universities and it was pretty easy to get NIH grants to study clock genes. Money brings in people...many new inductees into the field. Where did those new people come from? From genetics and molecular biology backgrounds. And with their backgrounds, they also brought in their own assumptions. There were thus two groups of people working on clock genetics: chronobiologists learning new techniques and molecular biologists learning new concepts. And those two groups had some very different ideas about the way clock works. I am not talking about the problem of the clock metaphor itself, so let me introduce a new metaphor to try to explain, however crudely, the two extremes (acknowledging that rare were the individuals that adhered to such extreme versions) representing the two attitudes of the two types of people.

Imagine a small town. Imagine in that small town a clockshop. Inside the clockshop there are thousands of clocks of different kinds. Every morning the owner of the shop comes in to unlock. Being an idiot savant, or having a great computer program, he takes a look at all of the thousands of clocks in his shop and immediatelly calculates their average time. About half are a little slow, the other half a little fast, some more, some less, but their average is likely to be the exactly right time of day. The shopkeeper then resets all the clocks to that mean time. Finally, he also resets a digital clock in the shop-window to that same mean time.

The inhabitants of that little town pass by the clockshop every morning and reset their own wrist-watches and pocket-watches to match the shop-window digital clock. That way, everyone's clocks are synchronized with each other every day and everyone goes to work and leaves work at the same time, has meals at the same time, goes to bed at the same time, etc.

Now, you, the chronobiologist, are trying to figure out how a clock works! All you can see is the behavior of the inhabitants of the town. You cannot see the digital clock in the window, you are not allowed inside the shop and you are certainly not allowed to open up any individual clock in the store. All you can do is send some vast calamity to the store. You can smite it with a lightning, or flood it with water, or cover it with thick black paint, or some such thing. Then you sit back and watch the behavior of the citizens and from that you try to infer how the clocks work.

So, as you do those nasty things to the clockshop, sometimes you observe an interesting effect - the citizens drift out of phase with each other over a few days. Everyone gets up at different time, goes to work and eats meals at a different time...no mater what kinds of statistics you use you cannot detect any rhytmicity in the activities of the town as a whole. What can you conclude? There are three possibilities:

a) whatever you did to the shop resulted in all the clocks in it breaking and stopping so the shopkeeper, honest as he is, has to switch off the digital clock in the window to show no time at all.

b) whatever you did to the shop resulted in shopkeeper dying, or shopkeeper loosing his ability to calculate the mean, or shopkeeper loosing his key so being unable to open up the shop, or the digital clock breaking down (or having no power to run on). All the individual clocks are working just fine, but are rapidly getting out of phase with each other.

c) whatever you did to the shop resulted in citizens not being able to see the digital clock in the window. Perhaps its display is broken, or the glass of the shop window became opaque. All the clocks are working, the digital clock is working, yet the citizens have no access to the information about the correct time of day and start drifting out of phase with each other.

Now, all of us trained in classical chronobiology, in labs of people like Pittendrigh, Aschoff, Daan, Menaker or their students, think in these terms. We are aware that we are observing and monitoring the final output of the clock, not a direct machinery of the clock. Many things can go wrong in between the clock and the observable behavior.

If you take the SCN out of a mammal or a pineal out of a sparrow or eyes out of quail kept in constant conditions in the lab, the rhythms of locomotor activity, body temperature, blood pressure, heart rate, plasma melatonin etc. become arrhythmic.

In the mammals, removal of the SCN makes the animal immediatelly arrhytmic and this cannot be reinstated in any way. This indicates that the pacemaker itself has been removed. But people working in non-mammalian vertebrates cut their teeth on complexity. Removal of the pineal in sparrows or eyes in quail results in arrhythmicity only after a couple of more weeks! This means that some additional pacemaker is still inside the animal. If there is another pacemaker there, why is animal becoming arrhytmic? See the a, b and c above - we were trained to think in those three terms: either every cellular clock has stopped, or cell-clocks are doing fine but lost synchronization with each other within the tissue, or the pacemaking tissue is doing fine but has lost the ability to instruct the target organ, e.g., a brain center regulating body temperature or heart rate. Perhaps a two-way feedback loop between the pacemaker and the target is broken. If you put such a bird back in light-dark cycles it shows rhythms again. If you then again put it in constant darkness it will again take a few days before it becomes arrhythmic. What did the light cycle do? Restart the individual cellular clocks, synchronize the cells within the pacemaker, or couple the clock to the effector that drives the observable behavior?

If you put an animal in bright constant light, it also become arrhythmic (no need for surgery). Are cellular clocks stopped, or are they out of phase with each other, or is teh output inhibited? When one places a quail in constant light of just over 5 lux, it becomes arrhythmic. But constant light is also the longest possible photoperiod which will stumulate the development of the gonads. In female quail, after about two weeks, the ovaries will be big enough and the bird will start ovulating and laying eggs every day. On the very first day of ovulation all the circadian rhythms come back - activity, temperature, feeding, etc. Does ovulation restart the clocks in cells, or resynchronize the cellular clocks, or remove inhibition on the overt rhythms?

A quail exposed to some very strange light-dark cycles (e.g., LD6:30) shows nice rhythms of body temprature but no rhythms of locomotor activity. Obviously the c) is the solution - decoupling of the clock from the effector.

So, when a classically trained chronobiologist dicovers a genetic mutation that eliminates overt rhythms, what does s/he think? Either a, b or c, of course are possibilities.

When someone with a genetics background discovers such a mutation, only interpretation a) comes to his/her mind. While it took them a while to learn (and they did eventually), initially the newcomers to the field produced hundreds of papers with great data and shoddy (to be polite) interpretations. Their conception of the clock was like a small town with a single clock on a tower. Thus, if the population is going about its business in an arrhythmic fashion, it must mean that the spire clock is broken. In other words, they ignored the complexity of the organism and the intervening levels of organization and assumed a one-to-one mapping of genotype to phenotype. They made an incredible assumption that the period and phase of the rhythms will be exactly the same in each cell no matter what perturbations or noise may intrude into the system. The faith in strong genetic determinism was unwavering - perhaps they were attracted to the clock research in the first place because it appeared in the beginning that this trait is strongly determined by genes. Nothing wishy-washy - an apparently great system to study without too much worries about the effects of the environment.

Swimming back up for air

As early as 1999 at the conference in Washington DC, people like Erik Herzog - with a fresh PhD yet already a star (come on, you knew I was going to mention him somewhere in this post, didn't you?), Low-Zeddies from Joe Takahashi's lab and a few others showed data suggesting that the picture is not that simplistic, i.e., that SCN as a tissue may behave differently than just a simple assembly of independent clock cells. They presented their data in a session dedicated to (and if I remember correctly, titled appropriately) the idea that higher levels of organization play a role in rhythm generation, not just genes in a cell.

Discovery of peripheral clocks provided the avenue for targeting this goal. It appeared they had all of the circadian molecular machinery acting exactly the same as in the central pacemakers, yet their rhythmic properties were different. What was different then?

In my 'basics' post defining the biological clock I described the way the understanding of the distinction between pacemakers and peripheral clocks is now in flux, and I was very careful how to define the differences between pacemakers and "just clocks". I simply stated that the former get input from the outside environment (e.g., light) and have a means (i.e., a signalling molecule) to send out temporal information into the body to entrain the other clocks, while those other clocks were dependent on that signal for their own timing. But there was something else I wanted to put there but was too unsure. There is another difference between pacemakers and 'mere' clocks that I wanted to point out but not all the evidence was available yet. There were many suggestive studies (some of which I reviewed here, here, here, here, here and here), but no final nail in the coffin that would let me state that peripheral clocks are bunches of individual cells while pacemakers are networks of intercommunicating neurons.

Now that final nail in the coffin has been published - last week in Cell. Here is the abstract of the paper itself: Intercellular Coupling Confers Robustness against Mutations in the SCN Circadian Clock Network:

Molecular mechanisms of the mammalian circadian clock have been studied primarily by genetic perturbation and behavioral analysis. Here, we used bioluminescence imaging to monitor Per2 gene expression in tissues and cells from clock mutant mice. We discovered that Per1 and Cry1 are required for sustained rhythms in peripheral tissues and cells, and in neurons dissociated from the suprachiasmatic nuclei (SCN). Per2 is also required for sustained rhythms, whereas Cry2 and Per3 deficiencies cause only period length defects. However, oscillator network interactions in the SCN can compensate for Per1 or Cry1 deficiency, preserving sustained rhythmicity in mutant SCN slices and behavior. Thus, behavior does not necessarily reflect cell-autonomous clock phenotypes. Our studies reveal previously unappreciated requirements for Per1, Per2, and Cry1 in sustaining cellular circadian rhythmicity and demonstrate that SCN intercellular coupling is essential not only to synchronize component cellular oscillators but also for robustness against genetic perturbations.

The press release (e.g., on EurekAlert and ScienceDaily) is actually very well done as well.

Cells vs. Networks

Let me now try really hard not to go into the details of every experiment as cool as they all are. If you are really interested, read the paper - it is very clearly written.

What did they do? They picked a couple of core clock genes to monitor and a few clock genes to knock out. When I say "picked", I really mean "picked"! They chose genes that are easy to monitor, only as markers of rhythmicity of the molecular clock inside cells. They chose genetic knock-outs in mice that are easy to obtain and have the required effects (i.e., complete abolishment of rhythms in cells). In the course of experiments (and the methods varied a little bit between experiments due to technical requirements), they confirmed some of the old observations and made some new ones and those are nice, but those are all of quite secondary importance.

What is really important is that this was the first systematic study that compared excatly the same genetic perturbations in:

- whole animals (locomotor wheel-running activity)
- intact SCN pacemakers (in brain slices)
- individual pacemaker neurons (dispersed SCN cells)
- intact peripheral clocks (slices of lung, liver and cornea)
- individual peripheral clock cells (dispersed fibroblasts)
- cyberclocks (a mathematical model)

And what they found, over and over, is that particular genetic knock-outs eliminate rhythms in individual cells (both SCN and peripheral) and in peripheral tissues, yet intact SCN tissue remains rhythmic and whole animal even more so. A peripheral clock is a collection of cells, a pacemaker is a network of cells.

Remember that mammals have three copies of the period (per) gene and two copies of the cryptochrome (cry) gene. When both cry genes are knocked out, everything is arrhytmic. Some other Per/Cry knock-out combinations also make everything arrhythmic. But deleting only one of them (e.g., only Per1 or only Cry1) leaves rhythms in whole animals and in whole SCNs while eliminating rhythms in whole peripheral clocks, in individual SCN cells and in individual peripheral clock cells.

Here, on the left, are rhythms of cells (each line is one cell, x-axis is time) without Per1 gene in SCN slices, and on the right are equivalent dispersed SCN cells:

This shows that assembly of SCN neurons into a tissue can buffer against such genetic disasters. Peripheral clocks cannot do that. And this is the most important difference between them.

But, you may ask, how do oscillations get started in an SCN in the first place, if one of the core genes is missing? After all, one can see how a network of interacting neurons can remain rhythmic after a perturbation, but how does one start from scratch? These are all classical knock-outs (not inducible), i.e., they were conceived, developed and grew up lacking the genes of interest.

Well, 10% of Per1- and 6% of Cry1- dispersed SCN cells actually still showed some oscillations of very low amplitudes. It is not known where they come from but it is possible that the machinery can utilize the other copies (Per2 or Cry2 respectively) to generate, if fortunate to once in the early life "hit" just the right phase, weak circadian rhythms. These weak oscillations then serve as seeds that gradually recruit more and more cells into the oscillating network. As the network grows, the cells feed back at each other and the amplitude of oscillations and the robustness of rhythms increase until the genetically defective SCN attains a completely normal function.

Why do peripheral clocks have no ability to communicate time with each other? The press release version is actually pretty good about this question:

The lack of networked interactions in peripheral tissues may actually be an adaptive feature in most circumstances. SCN cells in vivo must synchronize not only to light-dark cycles but also to one another to coordinate circadian behavior. Lack of coupling may allow peripheral oscillators to anticipate and respond rapidly not only to the synchronizing cues emanating from the SCN but also to physiological signals related to feeding and behavior.

In other words, peripheral clocks need more flexibility - they sometimes need to phase-shift much faster (in response to a meal, for instance) than the SCN is capable of.

Finally, the rhythms in whole animals are even more resistant than the SCN slices. This is something that future research needs to address - a feedback loop between the pacemaker and the effector are most likely culprits in stabilizing rhythms.

What next?

It is not surprising that these people published this paper. Liu and Welsh have made exciting new discoveries on peripheral clocks before. Steve Kay got his start in plant phytochromes and is the Mister Big Science of chronobiology - always the first one to introduce new molecular techniques into the field (e.g., the first one to use DNA-chips some years ago). He is also one of the smoothest, most polished speakers I have ever seen. Sometimes it almost does not matter (oh, it always matters with him!) what he says - one can just sit back and enjoy the perfection of the delivery of a scientific talk.

Joe Takahashi is old school - he started with birds in Menaker's lab. And he took his classical education with him to the Big Lab and Big Money at Northwestern University (and is now moving to University of Virginia). He does not publish a quickie every week, but when he publishes something, people stop and read, as it is likely to be a big milestone in the field. He discovered the first mammalian clock gene clock and in the process showed that by using forward genetics one can find behavioral genes in vertebrates. He (with Keith Barrett) showed everyone how to use the Limit Cycle model to drive one's research. He (with Low-Zeddies) made mice chimeras with mixes of Clk- and wildtype cells in the SCN (a direct intellectual precursor to this study). And now this. Quiet and gentle, with a sense of humor and unassuming posture, he is like Yoda of chronobiology, there is a force field around him and one always knows when he is in the room. Many people deserve a share in it, but if anyone ever gets a Nobel for clock research, Joe should be one of the three recepients.

This paper will rewrite the textbooks, close the book on simplistic bean-bag genetics in the field, and usher in a new exciting era of research. Nobody will, from now on, be able to publish a paper or get a grant funded without explicitely taking care of the network-like properties of the clock (not just paying it lip service in the Discussion section).

After this paper, we can all go back to the exciting questions of chronobiology - lunar, tidal and circannual rhythms, photoperiodism, sun-compass orientation, or whatever other crazy ideas we may have, in whichever organism strikes our fancy. We'll now know how to go about it.

Addendum: As a blogger, I was free to make large sweeping statements. The press release is somewhat more constrained, but right on target.

The language of the paper itself is much more reserved - it only suggests that the work fills some holes due to methodological gaps in the past, namely the use of slices instead of cells, the use of whole-organism response to infer the role of single genes, and the tendency to monitor gene expression for too brief periods of time.

But those methodological gaps come from people's backgrounds as well. They choose methods they are used to, but they also choose methods they think will answer their questions, sometimes revealing blind spots in their assumptions.

Chronobiologists have been studying SCN slices for many years - but those were neuro types, sticking electrodes into slices of brains from wildtype rats and hamsters, not mouse mutants.

On the other hand, geneticists tend to use dispersed cell cultures which they can monitor for long periods, or they probe tissues taken from animals sacrificed every 3-4 hours over a single 24-hour period - which is not long enough, as that is not sufficient time for cells to drift out of phase with each other, nor does it provide sufficient resolution (to figure out if lack of rhythm is actually a lack of synchrony, not the stopping of clocks). They also tend to use overt animal rhythms as direct reporters of the cellular function.

This paper demonstrates where the holes are in this approach.

I hope that the people in the field will realize what this paper has really done and not just focus on the small molecular details that arose out of each little experiment in it. I also hope that more people will focus on studying the mechanisms of cell-cell coupling in the SCN (and in other pacemakers in other organisms) and factors that can influence the strength of such coupling (e.g., hormonal environment) and not leave it to just a couple of people like Herzog to work out on their own.

Is there a similar coupling in the pineal of a sparrow and the retina of a quail, but not in the SCN of those species? I'd like to know...