A Theory of Sleep

yan king yin (no spam please) y.k.y at lycos.com
Mon Aug 13 03:54:53 EST 2001

Hi All,

This is the first draft of my paper.
Critiques or comments welcome.

The HTML version with a few figures is at:

A Theory of Sleep

The Generative theory of sleep is proposed here. It is based on synaptic
selectionism, summarized below. The new theory construes REM sleep as the
generator of "random" synaptic connections; and slow-wave sleep as the stage
where inadaptive synapses are eliminated. Molecular mechanisms are
speculated upon.

1. Synaptic Selectionism

Synaptic selectionism, advocated by Edelman and Changeaux [2001], is based
on an analogy between the brain and the immune system, particularly on the
mechanism whereby antibody diversity is generated through genetic
recombination. Following the main tenets of this theory, the following
assumptions are made:

    1) During the development of an individual, extensive variability of
neural circuit is generated by "random" creation of synaptic connections;
    2) Subsequently, these synapses are subjected to a selectional process,
the constrains of which are provided by general neurotrophic factors
(including depolarisation), which are the physical correlates of drives and
    3) Memory is a by-product of this selectional process where
early-established synapses are spared from elimination.

This is schematically summarized in figure 1:

2. Mature Neurons Undergo Morphological Changes

According to this view, life-experience is consolidated into memory through
remodelling of synaptic connections, which includes changes in both axonal
and dentritic morphology. Remodelling of neurite structure of the developing
Xenopus tadpole neuron has been recorded in vivo with time-lapse imaging
[Cline 1999]. Rapid dendritic rearrangements can occur over 2-h and even
10-min intervals (figure 2). The rearrangements include creation, extension,
and (complete or incomplete) retraction of branches. As the neuron matures,
it has slower growth rates and fewer dendritic branch rearrangements.

Similar changes were also observed in mature mouse neurons in vivo, by
Purves et al [1985], over a period of one month (figure 3). Remodelling of
mature neurons might be the cellular basis for cortical reorganization (eg
when part of a blind person's visual cortex is re-adapted for hearing).

3. Neuronal Remodelling Might Be The Function of Sleep

The observation of dendritic branch retraction suggests that the process of
neurite outgrowth is random. During early development, an excessive neuronal
network is generated and then refined through programmed cell death and
neurite elimination. The same mechanism might also operate through the adult
stage. Such a scenario, of neurons constantly changing morphology, might not
be totally compatible with wakeful activities, which are usually precise and
which rely on electrical signalling among populations of neurons. Although
it is possible that mental processes are averaged neural activities in space
and time and that small fluctuations do not matter, the massive generation
of randomness might significantly affect normal function. Therefore, it is
plausible that a synchronized cycle exists, which consists of:

    1) elimination of previous synapses during sleep;
    2) generation of new synapses during sleep; and
    3) activity and evaluation of synapses while awake.

Some more evidence support that neuronal growth does not occur while awake.
Firstly, growth hormone secretion occurs in episodic bursts and manifests a
diurnal rhythm. During most of the day, there is little or no growth hormone
secreted, while during sleep, there are pulsatile releases of growth hormone
from the anterior pituitary. This release is controlled by a negative
feedback loop involving the hypothalamus and anterior pituitary. Neuronal
growth might also follow this diurnal pattern. For instance, in the
hippocampus and cerebellum, the tissue concentration of BDNF (brain-derived
neurotrophic factor) protein is higher during the dark phase [Pollock et al

Secondly, during the early stages of development, longer periods of sleep
also correspond to the more rapid neuronal growth during development [Jouvet

4. Serotonin and REM Sleep

J Allan Hobson discovered that the brainstem locus ceruleus (noradrenergic)
and raphe nuclei (serotonergic) are active during wake, but almost inactive
during REM sleep, and intermediate during slow-wave sleep. Both nuclei have
extensive innervations to virtually the entire CNS. Virtually every cell in
the brain is in close proximity to a serotonergic fiber and is capable of
responding to serotonin by the process of volume diffusion [Azmitia et al].

Serotonin (5-HT) is also a potent inhibitor of neurite outgrowth. Single
isolated filopodia is observed to contract when exposed to serotonin.
Stimulation of the 5-HT1A receptors of rat neurons in vitro specifically
decreases the branching of neurites by 70% and reduces total neuritic length
by more than half [Sikich et al 1990]. Neurite outgrowth may actually be
inhibited by serotonin in a paracrine, non-localized manner in a
surprisingly large percentage of snail neurons [Goldberg 1998]. In the
prefrontal cortex, extracellular levels of serotonin gradually increase up
to 450% during wakefulness (as compared to slow-wave sleep), and decreases
toward stable levels during the next slow-wave sleep episode; During REM
sleep serotonin levels are about -60% with respect to slow-wave sleep [de
saint Hilaire et al 2000].  These findings suggest that REM sleep might be
the stage for synaptic generation where serotonin inhibition is low or

The effect of serotonin on neurite retraction might be due to Ca2+ influx.
The effects of Ca2+ influx on neurites will be discussed in the next
section. 5-HT1A receptors are present (postsynaptically) in high density in
the hippocampus, septum, amygdala, hypothalamus, and neocortex.

Some evidence is also suggested from antidepressants that increase the
availability of serotonin in the brain. MAOIs (monoamine oxidase inhibitor)
almost completely eliminate REM sleep. SSRIs (selective serotonin reuptake
inhibitors) and TCAs (tricyclic antidepressants) also reduce REM sleep by
30~85% [Vertes and Eastman 2000]. From my theory, it would seem that some
patients might be unable to learn new tasks that require extensive brain
reorganization (eg playing a musical instrument or learning a foreign
language). One such case was reported by Vertes and Eastman [2000]. However,
this must be taken in view of various confounding factors, including the
auto-regulation of serotonin release. Serotonin is also reported to induce
growth of new cells in the CNS.

The random generation in REM sleep might be the reason why dreams are
bizarre. Also, immediately after waking, the brain needs a period of
adaptation before being fully awake. This indicates that there might be
reorganization during sleep.

The noradrenergic locus ceruleus also innervates the cortex extensively,
particularly densely at the somatosensory cortex. Its inactivation may
account for the diffused state of consciousness during sleep, because of
withdrawal of its normal contributions to action potentials (Think of
"integrate and fire"). Thus, sensory stimulation has weaker effects in

Phylogenetically, neurotransmitters such as serotonin, actylcholine, and
catecholamines are not found in coelenterates such as hydra [Shaw 1996,
Grimmelikhuijzen et al 1996]. In the lowest animals such as cnidarians (eg
hydra, sea anemones, corals, jellyfishes) the nervous system is strongly
peptidergic. This suggests that serotonin is only involved in higher
cognitive functions and complex adaptative behavior. Sleep-like states have
been observed in mollusca and the seahare aplysia, but evidence seems to
support that there is no REM sleep in invertebrates [Frank 1999].

5. Synapse Elimination May Occur During Slow-Wave Sleep

If REM sleep generates connections in a random manner, then there must be a
mechanism to eliminate inadaptive synapses. This process should take place
after daytime activity and before the next phase of generation. Slow-wave
sleep precedes REM sleep and is the first stage of sleep. It represents
approximately 75% of total sleep time.

The calcium "set-point" hypothesis prosits that there is an optimal [Ca2+]i
(intracellular calcium concentration) level for neurite outgrowth. Segal et
al [2000] suggested that dentritic spine remodelling is regulated such that
a moderate rise in [Ca2+]i causes elongation of dentritic spines, while a
very large increase causes fast shrinkage and eventual collapse of spines.
Studies in LTP and LTD show that small increases in [Ca2+]i cause activation
of phosphatases, while a larger ones lead to activation of kinases [Bear
1995]. It has also been suggested that the high-voltage delta-wave (~1 Hz)
EEG pattern (figure 4) in slow-wave sleep (SWS) is likely to alter [Ca2+]i
with respect to REM sleep and wake. The extracellular levels of Ca2+ in the
cortex has been observed to oscillate at the same frequency as in slow-wave
sleep, with a baseline level at 1.1 mM and amplitude of about 0.2 mM
[Massimini and Amzica 2001]. This oscillation might activate particular
proteins, kinases, or phosphotases, leading to neurite elimination. For
example, the microfilament-severing protein gelsolin is activated by Ca2+.

Neurites that make the correct connections are prevented from elimination
possibly through redistribution of calcium channels such that [Ca2+]i is
reduced [Neely and Nicholls 1995].

One evidence for synapse elimination in slow-wave sleep is that when people
experience great expectations followed by great disappointment, they often
fall sleepy. From the selectionist perspective, this is when inadaptive
synapses must be eliminated. It could be that these neurites fail to receive
some form of target-derived factors.

(13 August 2001) contact author: y.k.y at lycos.com

More information about the Neur-sci mailing list