Saturday, July 23, 2011

To sleep, perchance to renormalize synapses

Flies sleep too, and give us novel access to the nature and function of sleep.

Why do we sleep? Why do even whales and dolphins sleep, alternating hemispheres so that they can stay at the surface, breathing reliably? Why does sleep deprivation eventually lead to insanity and death? Incidentally, isn't sleep deprivation as commonly used by our military and other security organizations equivalent to torture?

A recent paper supports one theory (originally developed by the same authors, Tononi and Cirelli, in 2003) of why we sleep, which is that our brain's neurons build up connections through the day, which the special characteristics of deep, slow wave sleep prune back to a manageable condition. Apparently fruit flies also sleep, allowing the enormous tool chest of experimental fly biology to be deployed on this fascinating question. That flies sleep at all is quite remarkable, since a fly's life is extremely short- a perilous few weeks at most. Yet they snooze away a third of it, admittedly at night when they might not want to be active anyhow.

The idea behind this theory is that learning happens by the classic Hebbian theory- cells that fire together wire together. That is to say, their synaptic connections increase in strength and number. Nerve cell processes (axons and dendrites) grow and make new contacts, and in a few cases, nerve cells may even divide and multiply. So during a day of intense activity and learning, we (or a fly) are continuously building up synapses, and perhaps not breaking them down ... the balance continues to rise and rise, until eventually the brain just seizes up and doesn't work anymore.
"The hypothesis predicts that the more one learns and adapts, (i.e. the more intense is the wake experience), the more one needs to sleep."
Sleep -specifically, the deepest slow wave forms that happen relatively early each night- is then theorized to put this process into reverse, erasing the weakest synapses and thus "renormalizing" the system back to some "normal" level that exchanges some old or unimportant crud for newly learned connections. Thus you might wake up and suddenly, the piano piece you were working on with diminishing returns the night before now seems a great deal easier, while yesterday's lunch is wiped clean away. Incidentally, later stages of sleep that involve dreaming may selectively reinforce older memories, instinctive imperatives, and stray connections in a completely different tuning process.

The original paper outlining this theory is thankfully available, and here are a few choice quotes from it:
"The slow oscillation occurs at a frequency that is ideally suited to induce depotentiation/depression in stimulation paradigms, namely <1 Hz [31]. Thus, from a frequency perspective alone, slow-wave sleep would be a good candidate for promoting depotentiation/depression.
The close temporal pairing between generalized spiking at the end of the up-phase and generalized hyperpolarization at the beginning of the down-phase may indicate to synapses that presynaptic input was not effective in driving postsynaptic activity, a key requirement for depression." 
(Here, depression means the opposite of classic Hebbian learning. Now cells that fire together under the slow wave paradigm unwire from each other. "Downscaling" is another term for synaptic renormalization.) 
"Thus, at least at the molecular level, sleep may not just be unfavorable to synaptic potentiation, but specifically conducive to generalized synaptic depotentiation/depression. More direct tests of this prediction can be envisaged. It is already known that sleep altogether favors dephosphorylation in the brain. One could further measure phosphorylation levels in sleep and wakefulness of residues of the AMPA channel [sensitive to the neurotransmitter glutamate and widely present in the brain] involved in potentiation/depotentiation and depression/dedepression, as well as indices of AMPA receptor internalization."
"Finally, the reduced activity of the noradrenergic system during sleep would ensure that only downscaling occurs, and not potentiation."
"According to the hypothesis, during sleep the strength of each synapse would decrease by a proportional amount, until the total amount of synaptic weight impinging on each neuron returns to a baseline level. Provided there is a threshold below which synapses become ineffective or silent, synapses contributing to the noise, being on average weaker than those contributing to the signal, would cease to interfere in the execution, and the SNR [signal to noise ratio] would increase. "

"Finally, the hypothesis triggers some further questions. For example, can anesthetic agents also produce synaptic downscaling, to the extent that they promote slow-wave activity comparable to that of NREM [non-rapid eye movement] sleep? ... How does the hypothesis apply to other brain structures where sleep rhythms are different, such as the hippocampus? Or to other species, such as the fruit fly? And finally, what about REM sleep? Could it be, for example, that with its steady depolarization and high spontaneous activity, REM sleep might pro- mote the insertion of AMPA receptors in the synaptic sites that are still effective after the downscaling of NREM sleep, and thereby favor their consolidation? Such “polishing” of synapses after the “cleansing” action of NREM sleep would agree with the regular alternation between NREM and REM sleep and the reported cooperativity between the two stages of sleep in certain procedural tasks"
Summary of model, from a later Tononi and Cirelli paper.
"Evidence for a relationship between synaptic strength or density and SWA [slow wave activity during sleep] also comes from developmental studies. SWA changes during the lifespan in a way that seems to follow cortical synaptic density, as indicated directly by electron microscopy on post-mortem tissue and by MRI estimates of the amount of gray matter. Thus, both synaptic density and SWA reach a peak in adolescence, after which they decline rapidly, and continue a slower decline into old age. Pathological decreases in synaptic density, as observed in neurodegenerative disorders and schizophrenia, are also associated with reductions in SWA. Moreover, after visual deprivation during the critical period—a procedure associated with synaptic depression, slow waves are reduced by 40% in the absence of changes in sleep architecture."

The synapse is a complicated place, with special proteins expressed on both sides, (i.e. presynaptic and postsynaptic), including storage systems for neurotransmitters and neuropeptides, secretion apparatus, receptors to detect them, as well as the usual ion channels that run electrical conduction along membranes all over the nerve cell.

The researchers use flies altered to express neuron- and synapse-specific genes fused to a gene fragment encoding the fluorescent protein GFP, lighting up the resulting neurons. The researchers also use straightforward sleep/wake control over their flies, keeping them in individual glass tubes (+ food) mounted in automated activity monitors, and shaken when needed to keep the flies sleep-deprived.

Flies don't make beds, put on pyjamas, or even so much as curl up to sleep, so researchers have to define it in a slightly more indirect way. Immobility for over five minutes is called sleep, while activity within a one-minute interval is defined as wakefullness (called "wake", for convenience). Nevertheless, they tend to sleep in a familiar pattern, for an eight-hour night, of which the first 2.5 hours are the deepest and most immobile. Flies are kept up by caffeine, noisy neighbors, etc., and catch up on missed Z's as soon as possible, but only for a fraction of the missed time. All of this is very much like our own sleep.

Below is the observed sleep pattern for typical female flies, where W marks the wake period (lights on), S marks the sleep period in the dark, and SD marks sleep deprivation, which appears to be very effective.

The weak part of the paper is that the researchers get their data by microscopically visualizing the synapses and neurons in selected areas of the fly brains and counting them or measuring their overall brightness by eye, which, even if done by neutral ("blind"!) observers, as they claim, is inherently noisy and subjective. But the point was to judge the effect of sleep on synapse structure and proliferation as directly as possible, (rather than, say, measuring the level of synaptonemal proteins over the entire brain in a gross way, which has been done repeatedly and agrees with the hypothesis on that level), so for the moment, this might be the best method available.

Below is shown a typical set of data, where they measure the volume of presynaptic structures (the half of the synapse that comes from the upstream axon) with two different marked proteins (syt-e/synaptogamin, and a neuropeptide called PDF) in a type of neuron involved in circadian rhythms. There are unequivocal differences when the flies were sampled after the sleep (S) condition or after the other two conditions- sleep deprived (SD) and wake (W).

Tha shake things up a little, they use mutant flies with no circadian rhythm- the Per (Period) gene is knocked out, so the flies have random bouts of sleep, (shown below), at least until they were all sleep deprived (SD7, seven hours), followed by either more sleep deprivation (SD12) or sleep (+S5). The data shows the despite that lack of endogenous rhythm, flies still need to catch up on sleep after being deprived.

The corresponding graphs of visualized synaptonemal proteins is below, showing, as expected, that sleep lowers these proteins significantly, while continued sleep deprivation raises them. These flies are cranky!

In the next set, the experimenters test another type of neuron, and look at postsynaptic dendrite elaboration in flies expressing fluorescent actin, which should show up pretty much everywhere. In this case, the dentrites grow "spines" as they contact upstream neurons and make synapses, so the spines are counted, shown in the image below as the very small balls. They also measured dendrite branch lengths. After showing that sleep deprivation leads to slightly elevated numbers of spines, they ask a new question- whether environmental enrichment during the day correlates with number of dendritic connections. While the control flies are left in their boring one-fly-per-tube hotels (W), the experimental flies are unleashed for twelve hours into a fly "mall" with a hundred other flies (Wm). Whether terror or happiness ensues, the researchers probably can't tell!

You can see that the briefly socialized flies have a lot more going on in their brains, growing longer dendrites and more dendritic synapses (spines).

All these values fell back to normal levels after sleep (below), showing a correlating cycle of more neuronal connections after waking activities and declines after sleep. Incidentally, the enriched flies also slept longer (I), as you might expect after a day of partying, including naps taken during daylight.

Lastly, the researchers use an interesting gene (FMR1), which, when defective in its homologous form in humans, causes mental retardation and other problems, called fragile X syndrome. Prior work indicated that flies lacking this gene product have over-elaborated neurons with lack of pruning, along with learning and other problems, and that overexpression of the gene can cause the reverse effect: well-pruned neurons even in the absence of sleep. These flies sleep 30% less than normal, and sure enough, the experimental protocols of sleep deprivation didn't significantly alter the dendrite counts and volume compared to wild-type flies.

Genetic variation in such a gene might be responsible for variable amounts of sleep need seen naturally, in flies as well as in humans. So it remains slightly puzzling why most animals need so much of it. Perhaps, given the astronomically-imposed day/night cycle, checking out for eight hours is not much worse than checking out for two. We may also have only scratched the surface of what sleep does for us, whether physiologically or psychologically. Certainly Jungians, among others, set great store by dreaming, which happens during a separate sleep phase. Do flies dream?

This hypothesis is well on its way to becoming a compelling theory of what is likely the principal reason why we need deep slow-wave sleep and run into serious problems if we don't get it. The underlying mechanisms remain under investigation, (especially the functions of the various brain waves, and the dynamics of synapse growth and regression), and the hunt goes on for the mechanisms and rationale of other significant processes that happen during sleep.

  • We have a gambling problem.
  • On the power of the placebo effect.
  • What happened to the Republicans?
  • How's that hearts-and-minds operation going in the Middle East?
  • A local peace dividend- open space on Mt Umunhum.
  • Krugman tracks employment.
  • Economics quote of the week, George Packer in the New Yorker, on Goldman Sachs board member Rajat Gupta's insider trading conspiracy with Galleon's Raj Rajaratnam.
"On the afternoon of September 23rd [2008], Rajat Gupta, former head of McKinsey, joined members of the Goldman Sachs board on a conference call. They discussed Warren Buffet's proposed investment of five billion dollars in the investment bank, which had been imperiled by the crash. The conference call ended at 3:54 P.M. Sixteen seconds later, Gupta called Rajaratnam's office. At 3:58, just two minutes before the markets closed, Rajaratnam gave an order to buy three hundred and fifty thousand shares of Goldman stock, worth forty-three million dollars. That night, the world learned of the Buffett investment. At the peak of the crisis, Gupta the Goldman board member's first thought was to make sure that his investment partner Raj Rajaratnam could exploit the deal. A month later, the drill was repeated: as Goldman prepared to announce an unexpected quarterly loss, Gupta called Rajaratnam, and Rajaratnam sold all his Goldman stock before the announcement."


  1. "The hypothesis predicts that the more one learns and adapts, (i.e. the more intense is the wake experience), the more one needs to sleep."

    No matter what, I love a hypothesis that gives me an excuse to sleep in! ;-)

    In all seriousness, I had no idea that flies slept. How utterly fascinating. I'm also interested in the idea touched upon in the original paper - how do these conclusions affect our use (or non-use) of anasthetics? If we can chemically induce someone into a sleep where the brain repairs and renormalizes connections, can we speed up learning? Can we brainwash? It's all very strange.

  2. Hi, Kelly-

    Thanks- I found it all pretty neat too. I think it would be hard to improve on the natural process, though there are many ways we can screw it up. Feeding people bad, emotionally charged data ("brainwashing") is definitely one of them. One could make the case that induced coma is one way of using these processes therapeutically for healing, though I am not sure what the brain wave situation is there- learning is probably not a big factor.

    The depressing part (for me, at least) is how learning and sleep decline in tandem into middle and old age. I would very much like some way to stem that ebb.