So in this last session,

I'll talk about some recent results,

where we've extended the model to simulate human sleep.

Now, one thing about sleep on

the human time scale is that it happens over a 24-hour cycle.

So we need to include the circadian rhythm and, in particular,

the inputs from the suprachiasmatic nucleus to the sleep wake network.

Now, some details are known about how the SCN projects

to the sleep wake centers

that we've been talking about and that are included in the model.

So the SCN has direct synaptic projections to the SPZ,

the subparaventricular zone, and the DMH, the dorsomedial hypothalamous.

And then from the SPZ and the DMH,

the synapses diverge onto the sleep wake centers that we've been talking about,

the wake-promoting populations, the velpeau and the REM-promoting populations.

However, many of the details about those synaptic projections aren't worked out.

But, generally, it's been determined that the net effect of

a SCN activity seems to be promoting wake and inhibiting sleep in humans.

So, working on that,

how we included the SCN in our sleep wake network model was to just model,

say, the overall net effects of

the projections from the SCN to the different wake-promoting,

non-REM sleep promoting and REM-sleep promoting population.

So this schematic shows

a simplified version of our sleep wake network model on the bottom,

and how we extended the model was to include a SCN neuronal population.

And the synaptic projections we have,

we included were to have that the SCN activity had

an overall net effect of exciting

the wake-promoting population in exciting the REM-promoting population.

But it had a net effect of inhibiting

the Velpeau or the non-REM sleep promoting population.

What's also been determined is it seems is that the sleep wake

states and behavior feeds back onto the SCN.

So here are some experimental results showing where they recorded

neuronal activity in the SCN over an extended period of time to show that during the day,

SCN activity is high,

during the night SCN activity is lower.

So that's generally known,

but there's a lot of variability in the SCN activity during these states.

And when they looked closer,

what they found is that in these panels over here on the far right,

the bottom shows a hypnogram of, in this case,

it was a rat behavior transitioning between wake,

non-REM sleep and REM sleep.

And what they saw is that during wake and REM sleep,

that SCN activity increased during those states.

And so what that suggests is that there must be feedback synaptic projections

from the wake-promoting populations and the REM sleep promoting

populations that promote higher SCN activity during these states.

And so we included that in our model by including

these feedback projections from

the REM-promoting population and from

the wake-promoting population back onto the SCN, which we're excitatory.

So that's a schematic of our expanded model,

and we were able to,

by adjusting the time scale of

the homeostat and adjusting the strength of interactions between the populations,

tune this model, which has

a very similar structure to our model that can generate rodent sleep.

We could tune this mode to generate human sleep.

And so here are some numerical simulations of our model.

The top panel shows a hypnogram

representing what the predicted behavioral state of the model is,

and so we could tune it so that we have a 24-hour sleep

wake cycle with about a 16-hour wake period and eight-hour sleep period.

And with REM cycling during the night,

with about a 90-minute cycle of REM sleep with four REM bouts,

which is sort of typical of human sleep.

And underneath that top hypnogram trace,

what I'm showing is the firing rate in the SCN,

so would be our measure of the circadian rhythm,

where the SCN activity is high during wake and low during sleep.

And then in the last panel is our homeostatic sleep drive H,

which increases during waking and then decreases during sleep.

So we could tune the model to replicate normal human sleep.

But one thing that basically came out of the model,

just from the structure of

the reciprocal interaction mechanism for REM sleep generation and the effects of

the SCN activity is that the model

automatically generated the change in REM bout duration across the night.

So in these panels,

the panel on the left shows a close up of the model activity during the night,

where the different colored lines indicate the firing rates in the different population.

So green is the firing rate in the locus coeruleus, the wake-promoting population.

Red is the firing rate in the Velpeau,

the non-REM sleep promoting population,

and dark blue is the firing rate in the REM-promoting population.

And this is just showing that waking,

transitions to sleep when the locus coeruleus firing rate

drops down in the Velpeau firing rate increases.

And then during the night,

during the sleep episode,

we've got four REM bouts occurring during the night.

And what our model could

automatically generate is the change in the REM bout duration across the night.

So the panel on the right is showing the REM bout durations that are model generated,

and it reflects what's been experimentally observed that

the first REM episode is shortest,

and REM bout durations increase over the night but with

the last REM episode being a little bit shorter.

So that was great that our model was able

to basically verify these features of REM sleep,

but we wanted to basically push the model to see

what other kind of experimental results the model could account for.

And the one experimental protocol we decided to

investigate is that of spontaneous internal desynchrony.

So spontaneous internal desynchrony refers to

the desynchronization of the circadian rhythm and sleep wake behavior,

and this desynchrony has been invoked in these early experiments,

where they kept people under temporal isolation,

so complete isolation from all environmental time cues.

So, in these experiments,

people were basically spending

several months inside of a kind of a bunker room that was completely

blacked out from the environmental day night cues

and any other kind of cues from the outside world,

and they were on their own kind of internal schedule.

And what was observed is that after a number

of weeks that their circadian rhythms in their sleep wake behavior desynchronized,

they were no longer synchronized.

And so what these panels show are some of the recordings of behavior,

where the dark lines indicate the time that these subjects were sleeping.

And what we can see is that at the start of the experiment,

people's sleep wake cycles were a little bit longer than 24 hours indicated by

the rightward trend of these dark lines indicating the sleep episodes.

And that after a couple of weeks,

there'd be some variation in the timing of when sleep started,

and then, eventually, after, say,

day 35 in the panel on the left and maybe day 50 on the panel on the right,

that sleep wake behavior was very desynchronized,

and basically, sleep was occurring at all different circadian phases.

So we've stimulated these experiments in our model and

the way we simulated the isolation from

temporal time cues is to assume that

the intrinsic period of the sleep wake cycle lengthened gradually over time.

And so here are some model results

of evolution to spontaneous internal desynchronization,

where the panel on the left shows a similar kind of behavioral actogram,

where the dark bars indicate the sleep episode,

and what we've also included in this panel are the purple circles,

which are the minimum of SCN firing rates or the minimum of the circadian rhythm.

And so what we see in our simulation is very similar

to what was shown in the experiments is that,

in the beginning over the first two weeks that

the sleep wake cycle period lengthens but the sleep cycle and

the circadian cycle are still synchronized with a minimum of

the SCN firing rate occurring during sleep as normally occurs.

Then after a little longer,

there's some variation in the timing of sleep,

which is this phase trapping kind of behavior,

which was seen experimentally.

And then, eventually, the two rhythms,

the circadian rhythm in

the SCN and the sleep wake cycle become completely desynchronized,

where we've got sleep happening independent of the circadian cycle.

And so the panels on the right show some more details of the changes

in the wake bout durations and the sleep bout durations across the simulation.

And the model was actually able to

replicate the experimentally observed circadian variation in REM sleep,

which is shown in the bottom panel on the right.

Where we're showing the percent of REM sleep as a function of total sleep time.

And we can see that that starts to vary in the cyclical pattern as

sleep is occurring at different circadian phases.

So, this was also a sort of validation of our model that our model

we could generate this spontaneous internal desynchrony,

and that it could also replicate the changes in

REM sleep behavior at different circadian cycles as has been observed experimentally.

But, the one thing with a model is that,

what else can it tell us?

Can it actually tell us something new?

Could the model tell us something that we didn't know

about spontaneous internal desynchrony,

or synchronization of the sleep-wake cycle and the circadian rhythm?

So we did a lot of investigating of

the dependence of this kind of behavior on different components of the model.

And one thing that we found,

which was a bit surprising.

And which I'll say is a prediction of our model is that,

the SCN or circadian input to the REM-promoting populations

in the LDT and the PPT actually had

a strong influence on the synchronization

of the sleep-wake cycle with the circadian rhythm.

The simulation shown here,

show the same kind of

behavioral actograms where the sleep-wakes in each of the two panels,

the sleep-wake cycle period is increased in the same manner.

So, this panel on the very far right shows

that what we did in these simulations was increase the sleep-wake cycle period linearly.

And we did it the same in both the left panel and the middle panel,

but what was different was the strength of

the excitation to the REM-promoting population,

the LDT and the PPT, from the SCN.

In the simulation on the left,

that excitation was weak,

and in the panel on the right,

that excitation was strong.

And so what we found is that with

the stronger SCN excitation to only the REM-promoting population,

that's the only thing that's different.

Is that, synchronization was maintained between

the sleep-wake cycle and the circadian rhythm for a much longer period of time,

than when that SCN excitation was weak.

And now this was a surprising prediction,

because the transitions into sleep and the transitions into

wake are dictated by the wake-promoting population and the sleep-promoting population,

not by the REM-sleep promoting population.

But what the model is showing here is that this circadian input just to

the REM-promoting population can have

a big influence on basically the sleep-wake flip-flop switch.

And so in our model,

we could analyze exactly what it was.

And just to give you a brief idea,

it was basically that the stronger SCN excitation to

the REM-promoting population had an effect on the REM sleep cycles,

and the changing dynamics and timing of the REM sleep cycles actually then

affected when the sleep-wake flip-flop switch could transition.

And so in a way, it's kind of a network-propagated effect of the circadian rhythm through

the REM-promoting mechanisms to

the sleep-wake transition mechanisms that are in play here.

And so we present this as a prediction of our model,

and that perhaps it might prompt experimental investigators to look at the effects of

REM sleep on the synchronization of the circadian rhythm and sleep-wake

03.09 - Diniz Benh & Booth Model - Part II

Sleep: Neurobiology, Medicine, and Society

Conheça os instrutores