Thank you so much to the organizers for inviting me to speak here, and thank you to all of you for being here. It's really a privilege to be able to speak to you all today. I'm going to be talking about scalable sleep bio-markers in neurodevelopmental disorders.
This is very much work in progress. This is also the first time I'm talking about these data in public, so it's very hot off the press and really represents a new direction for our lab, so very much looking forward to your questions and feedback. I'm going to put my thank you slide first.
I just need to spend a moment on a shout-out to these really incredible phenomenal people, the graduate students, postdocs, and the research assistants in the lab that have actually done all the really hard work that I just get to come here and talk about it. Also, to all of our collaborators around the world. Sleep difficulties are very common in autistic individuals across the lifespan with about up to 80 percent of autistic individuals reporting sleep problems.
These sleep disturbances are really wide-ranging, include disrupted sleep continuity, increased sleep latency, so increased time to fall asleep, sleep disruptions and fragmentation, increased time spent, a wake after sleep onset. The reports of both insomnia and circadian rhythm disruptions, so issues with both the circadian system and homeostatic sleep mechanisms. From polysomnography, there are reports of abnormal sleep physiology and differences in sleep staging or the architecture in autistic people.
Again, there appear to be all sorts of different sleep problems going on, and there's not a really precise understanding of how or why sleep is so disrupted, but importantly, these sleep disturbances have been associated with negative daytime outcomes, including worse behavioral problems and poorer quality of life, as well as increased anxiety. I think sleep is really important, given that it's a potentially modifiable intervention target, and there are a handful of studies, providing some preliminary evidence that their behavioral sleep interventions are maybe effective, not only in improving sleep in autistic people, but may also improve behavioral problems like hyperactivity, externalizing behaviors, emotional problems, and may also be able to improve quality of life in both the autistic person themselves as well as their caregivers. This is really promising, but the literature today is pretty small, and there are mixed results with some of these studies reporting really no effect of sleep therapy.
As I think probably most of you in this room are aware autism is incredibly heterogeneous, and so more research is really needed to understand the precise nature and biological underpinnings of sleep disturbances in autism. What do we know about the brain circuitry underlying sleep? Well, the thalamus, in particular, that little orange structure there in the middle of the brain.
The thalamus is a brain structure that's of a lot of interest in autism as it really reflects. It's a central communication hub in the brain that's responsible for relaying information from sensory receptors to proper brain areas in the cortex, the top part of the brain. These thalamocortical circuits, I'm going to be talking about a lot, these really represent a fundamental aspect of large scale brain organization across mammalian species, and they may be, we think, particularly sensitive to network level disturbances in the brain.
These circuits can be investigated in living human subjects, using a methodology called resting state functional MRI, and this reflects an index of the brain's intrinsic functional architecture, really reflecting these spontaneous fluctuations in neural activity that occur when one is not doing a particular cognitive task. The thalamus is also particularly relevant for understanding sleep because sleep spindles arise from thalamo-cortical inhibition that's mediated by, I think you can see my cursor, this little structure here called the thalamic reticular nucleus. The thalamus is a sub-cortical brain structure that has widespread connections to the cortex.
There is a converging evidence from both animal and human studies showing that there's evidence for thalamocortical dis-connectivity in people with autism, so it's different from typically developing folks. In fact, the overall finding that's been pretty consistent is there's actually increased functional connectivity or hyper connectivity in autistic people across the lifespan. You can see that here these are the autistic group shows an increase particularly in connectivity between the thalamus and sensory-motor cortex, both in toddlers and preschoolers, and that's also been found in children and adolescents.
In prior studies, this pattern of hyperconnectivity is also associated with worse sleep problems, so it's possible that thalamocortical disruption could be related to sleep disturbances and other autistic behaviors like sensory processing problems. But again, this literature on the neurobiological correlates of sleep disturbance and autism is really limited thus far. We're really interested in whether this thalamocortical disconnectivity between these brain regions may be related to sleep problems as well as other behaviors like sensory disturbances.
One challenge in research in the autism field is heterogeneity, and so both genetic and phenotypic or behavioral heterogeneity. In a typical research design, we're going to ascertain a sample of autistic and non autistic individuals based on certain behavioral criteria. But within this behaviorally defined sample, there's probably many different biological mechanisms that may be converging on the same behavioral criteria that makes research into the biology really difficult.
In our lab, we're taking instead, what we call a genetics first approach in which we take individuals with a shared common genetic etiology, which confers particularly large increases on one's risk or likelihood of developing autism, and so this is really an approach that we hope can decrease the heterogeneity and increase the signal. Taking this approach, we're also able to identify individuals very early in life based on this genetic diagnosis. I'll tell you a little bit about, and I think you heard from a really phenomenal previous speaker about a little bit about the autism genetics.
But we know that several genetic factors have been associated as being associated with autism, including rare single gene de novo variants, as well as copy number variants. Copy number variants are basically deletions or duplications of several genes together, as well as common genetic variation. What you're looking at up here is just, on this axis here, is how common a particular genetic variant is in the population, and on the y axis is the likelihood of having autism.
These common variants down here, you can see each of them confers a very tiny increase in risk, but maybe many of them acting together may increase risk for autism. Whereas with these what we call rare variants, each of them may confer a very high likelihood of developing autism. That's really where we're operating here.
Our lab has been particularly focusing on deletions and duplications at a particular chromosomal locus, the 22q11. 2 locus. Today, I'm going to be focusing on our work on the 22q11.
2 deletion. Let me tell you a little bit more about that. This is a genetic syndrome that's also known as DiGeorge or velocardiofacial syndrome.
It results from essentially a chunk of DNA that's missing on one copy of chromosome 22. It's actually a pretty large deletion, 1. 5-2.
6 megabase deletion. It occurs in one in approximately 4, 000 live births, so it is a rare disorder, but it's not as rare as you might think. It is, in fact, the most common contiguous gene deletion syndrome.
It has a greater than eight fold increase in likelihood of autism, and there are some associated medical conditions, including congenital heart defects, immunodeficiency, and craniofacial anomalies. These are just some pictures of some of the patients that have participated who gave their consent for their pictures to be shared. In addition to autism, there's also an increased likelihood of intellectual disability, ADHD, attention deficit hyperactivity disorder, anxiety disorder, as well as psychotic spectrum disorders.
While not all 22q deletion carriers meet the criteria for autism, this population is enriched for dimensionally measured phenotypes related to autism. For example, and this is data from our cohort, there is high endorsement of sensory processing problems, including sensory avoiding, increased sensory sensitivity, as well as low sensory registration. So on a group level, the 22q deletion carriers are scoring significantly higher on all of these measures, as you can see, but there's also really a lot of individual variability.
Now, like autistic individuals in the general population and I'm going to be using the term idiopathic to refer to behaviorally defined autism, but 22 deletion carriers also report severe aspects of sleep disturbance. I should mention this is work led by a really phenomenal graduate student in the lab, Kathleen O'Hora, who came into lab with a really wonderful background on sleep research and has really been the one to convince me that this is a really important thing to study. Essentially, if we divide up our cohort, so first of all the 22 deletion carriers have significantly greater sleep disturbance.
This is a pretty crude measure of sleep disturbance, just a clinician rated measure, but those who were within the 22q cohort, those who were poor sleepers, had significantly elevated scores on the social responsiveness scale, so a measure of reciprocal social behavior. Again, this was a very broad classification of sleep so it limits our understanding of mechanism, but suggests this was something important to follow up on. At the brain level, this is work led by Charlie Schleifer, who's also really incredible MD-PhD student in the lab.
What Charlie did was to look at resting stake functional connectivity in the brain in carriers of this 22q deletion compared to typically developing youth of the same age. We know that 22q deletion carriers have differences in thalamic structure, connectivity and development. Essentially what Charlie did was to take a seed, the little blue area in the thalamus and look at connectivity of that thalamic seed to the rest of the brain.
What you can see here is that the thalamus is over connected in the 22q deletion carriers to particularly to somatosensory cortical brain regions. I think I mentioned, there are several studies documenting similar patterns in idiopathic autism. Over here on the right, we're just seeing what that looks like quantified, so essentially, you can see this.
Shift of the over connectivity of the thalamus to somatosensory regions in the 22q deletion carriers compared to typically developing youth. In studies of idiopathic autism, there's evidence that this thalamocortical disruption is related to sleep disturbances and so here we took this same measure. First, we're just looking at the relationship between time it takes to fall asleep, and that was associated in idiopathic autism with overconnectivity of the thalamus to auditory cortex.
We took our same measure of thalamic to sensory motor connectivity and compared it to our measure of sleep disturbances that we see and we found that this pattern of thalamocortical hyperconnectivity was also associated with sleep disturbances in these kids with 22q deletion, even again, using this pretty coarse measure of sleep disturbance. Given all this put together, we thought it was really important to further investigate sleep neurophysiology in 22q deletion carriers. We launched this study pretty recently to better characterize sleep disruptions associated with the 22q deletion and its relationship to cognition and autism related behaviors using a multimodal approach.
We used subjective reports, both from parent and child questionnaire measures, as well as sleep diaries and then we did actigraphy. This is essentially a wristwatch that is an estimate of sleep and wake through movement, so it's measured by this wrist accelerometer. Actigraphy provides pretty good estimates of sleep behavior, but it's not a direct measure of sleep process, so it doesn't provide information on sleep depth or sleep staging.
For this, we used a pretty new wearable EEG device that people can wear at home. It's a pretty exciting development. Then because sleep is also important for memory consolidation, we looked at a sleep dependent memory task, this motor sequence task that I'll tell you a little bit more about.
Traditionally, sleep research in the lab relies on polysomnography. This is a kid undergoing polysomnography. You can see that it is incredibly burdensome.
Lots of sensors you have to be hooked up to and this is really challenging for children with neurodevelopmental disorders and sensory sensitivities. You can imagine you're really not going to get a good night's sleep like this. This is a pretty exciting recent development that these wearable devices that can allow in-home assessment of things like sleep and activity.
We're using this EEG headband. It's called the Dram 3 to remotely measure sleep architecture and sleep neurophysiology across several nights. Essentially, it includes frontal and occipital, dry EEG sensors for folks that are interested in the technology.
It gives you automated sleep staging. On the participant side, it's really simple to use. They basically just have to press the button to turn it on when they go to sleep.
I'll tell you a little bit about our actigraphy findings. This is just over there on the left is just to show you what a single 24 hour actigraph looks like. We compared six different phenotypes between the 22q deletion carriers and controls.
Again, this is work in progress but so far, we have 29 in our 22q deletion group, 30 age and sex match controls with good usable data. You can see it's a really wide age range. We control for sex, age as well as weekend status in the analyses.
Our protocol was to get seven days and nights of data. We excluded anyone that we didn't get four days and nights of usable data. What we're finding, actually, is interesting.
A little bit unexpected. We actually found that the 22q deletion carriers had significantly increased sleep duration. They were actually sleeping longer.
Now, we don't know. No. Maybe that sleep isn't very efficient, but they were spending longer time in bed.
Now, they were not different in terms of the amount of week time after sleep onset. The nap duration, the number of naps, the number of awakenings or the sleep regularity index, they did not differ from controls. That's a measure of how consistent their sleep is across nights.
That's what we're seeing so far with actigraphy. But to actually study the brain during sleep, we're using and this is, again, from typically studied with polysomnography. On the EEG, there's really two important measures of sleep neurophysiology, the slow waves shown on the left and the sleep spindles shown on the right.
These slow waves are high amplitude, low frequency oscillations like this that you can see them what they look like on the EEG, that occur during non-rapid eye movement sleep. These are occurring in this Delta band frequency, so 1-4 hertz and they are dependent on sleep history. In other words, they reflect homeostatic sleep regulation, meaning that after you're sleep deprived, then there's increased slow waves in the beginning of the night and then that dissipates throughout the night as basically that homeostatic drive decreases.
These reflect corticothalamic interactions. Now, on the other hand, these sleep spindles, which are the little purple squiggles over there, these are high frequency bursts and these reflect oscillations in the Sigma band or 11-16 hertz. Sleep spindles tend to be robust to sleep history, so they tend to be pretty stable within an individual.
They're also highly heritable and these reflect thalamocortical interactions. Just to go into this in a little bit more detail, I mentioned these slow waves are the low frequency isolations that are generated by corticothalamic interactions and they are propagated throughout the cortex of the brain. They have this upstate and a downstate that arise from hyperpolarization and depolarization of these populations of what are called cortical pyramidal neurons.
You can see, again, the thalamocortical neurons there and the relationship to the cortex. Now, the sleep spindles are a signal of thalamocortical inhibition that's mediated by that little red structure there, that's called the thalamic reticular nucleus. The sleep spindles are present during Stage 2 and 3, non rapid eye movement sleep and they are actually arising from inhibition of those thalamocortical projections by voltage gated calcium channels within that structure, the thalamic reticular nucleus.
This inhibition is leading to burst firing. It's going to show up on the EEG, like that little purple squiggle as a sleep spindle. This nucleus is able to generate sleep spindles in isolation, but their reflection.
What we're looking at on the EEG is really the reflection across the cortex of the brain. It's really reflecting these reciprocal interactions in those thalamocortical feedback loops. Now, one critical function of sleep and sleep spindles in particular is memory consolidation.
During non-rapid eye movement sleep, the sleep spindles are temporally coordinated with cortical slow waves and these hippocampal sharp wave ripples to transfer memories from you have this temporary indexing in the hippocampus of the brain and then they get transferred to more permanent distributed representation in the cortex. Now, sleep spindles, interestingly are also responsible for sensory inhibition during sleep. You're not moving around constantly when you're sleeping during non-rapid eye movement sleep.
They preferentially occur during slow wave, these upstates. Now the hippocampal ripples are shown here. They're occurring in the troughs of these spindles.
But hippocampal ripples require invasive recording, so we're not actually able to measure them. We're really focusing here on the slow waves and the sleep spindles, as what we can measure with sleepy EEG. One method of testing sleep dependent memory consolidation is this motor sequencing task.
This is a task where participants complete a training phase and a test phase that's 24 hours apart. In the training phase, it's a fun task actually. In the training phase, you see this string of five digits and you're instructed, you just have to type it over and over again as quickly and accurately as possible for 30 seconds.
You get a rest period, 30 seconds and then there's 12 trials. Then 24 hours later, you get a night of sleep, 24 hours later, you do it again, the same sequence as quickly and accurately as possible. What's really interesting is that in typically developing controls, healthy individuals, participants get significantly better at this task the next morning after sleeping because that motor memory of the sequence gets consolidated during this non-rapid eye movement sleep.
This is from a study of pretty large study of college undergraduates, really showing. That's the black bars are Day 1 and then the striped one is Day 2 after a night of sleep. You can see that people get quite a bit better at this task after the night of sleep.
Importantly, the degree of overnight improvement, how much better you get on the task is directly related to the percentage of Stage 2 sleep, which is where the sleep spindles are most active. Really suggesting the sleep spindles are particularly critical for this. Don't worry about all this methods details, if anyone's interested, this is a lot about the way we did this analysis.
This is just our preliminary analysis of non RM sleep physiology. Here we had 2022 deletion patients, 18 controls so far again, this study is still ongoing. In our 20 deletion group, just to show you clinically about the sample, half of them have an autism diagnosis.
You can see that there's also pretty high rates of ADHD as well as anxiety disorder. Here this is just so you can see what this spectrogram looks like. This is just an example from and this is from a 32-year-old male with 22 deletion.
Alpha power, if you're looking at the beginning of the night. Alpha power really reflects the wakefulness where you're basically, you see those red in the beginning of the night where this person is still awake. Now, Delta power is a biomarker of restorative sleep and is associated with sleep duration and intensity.
It's highest after you've been awake and then it decreases as sleep progresses. Now, Theta waves are this EEG power between 5 and 8 hertz. Theta is associated with deep relaxation and sleep and actually predicts this subsequent homeostatic increase of slow wave activity.
Now, Sigma power is a measure of sleep spindle activity. Again, this hallmark of non-rapid eye movement sleep. Preliminary analysis, we just looked at group differences in each of these EEG power spectrum.
For sleep spindle detection, this is still work in progress, again, work being led by Kathleen O'Hora in the lab. But to assess the ability to detect sleep spindles in this data, again, this really hasn't been done before with these wearable devices. We applied an algorithm that's been validated in polyspography to these electro channels and essentially after removing artifacts, and then these beige boxes are showing you where we were able to identify the sleep spindles.
We were pretty excited that we were actually able to identify these sleep spindles using these devices. Just to show you the night to night stability of sleep spindles, as I mentioned, they've been shown to be a stable measure with very little variability from night to night. We wanted to see if this was true in our data.
We calculated intraclass correlation coefficients just to see how consistent it was over time. We did find pretty good reliability within subjects, 0. 77 across all nights and when we cut out that first night, we got excellent reliability, 0.
91. That first night really indicating we should treat that as a night of acclimation because they're getting used to wearing this device. What are we seeing?
We did see, interestingly, a trend towards increased absolute power specifically within that Theta band in the 22q deletion. Because we saw that increased absolute power for the rest of the analyses, we're looking at relative power, which is basically we're taking the absolute power divided by the total EEG power across all the frequency spectrum. Here, again, with this relative Theta power, and the Theta power, again, is this power between 4-8 hertz.
We are seeing a really substantially increased Theta power. Difference in this particular frequency band in the 20 deletion carriers. Then overall, so just to point out, these are just showing the differences overall between the 20 deletion carriers in blue and the typically developing controls in yellow.
We do see 22 deletion carriers, again, this pretty large effect on increased Theta power, but also interestingly decreased power in the very low frequencies, what we call slow power, less than one hertz. Interesting. We're still trying to figure out what to make of it, we also saw this increase in the relative Alpha power, which is that early stage where they're still awake, but no differences in Sigma and Delta thus far.
What does this all mean? Well, we're still trying to figure that out ourselves. But we think this is pretty interesting because differences in sleep neurophysiology may reflect disrupted neurobiological mechanisms during non-RM sleep in 22 deletion carriers.
Now, we know that non-REM Theta typically declines from early adolescence to young adulthood reflecting the synaptic pruning process. This is really interesting because in a 22 deletion, we see that this is from our longitudinal structural MRI study where the X axis is age and the Y axis is cortical thickness. The purple line is those 22q deletion carrier, the patients.
We're seeing is that the cortex is thicker in 22q deletion patients through adolescence. Then that actually switches after adolescence so their cortex starts more rapidly thinning later in life but essentially, that typical process of cortical thinning is delayed. Essentially, there's a delayed cortical maturation in 22q deletion carriers.
This suggests that aberrant synaptic pruning mechanisms may be at play here and may be related to this abnormal theta that we're seeing. Now, of course, this is pretty speculative at this point, so we really need more research to understand what we're seeing, but it's interesting so far. The last bit of data, I'm going to tell you about is just with this motor sequence task analysis.
Again, we did this 24 hours apart with the task, so far, we have 20 deletion carriers and 19 controls in this analysis. Their age and sex match, and we controlled for age and sex in our analysis, because as you can see, there's a pretty wide age range here. What we're looking at, those first lines are their performance on day one.
With the motor sequence task, the deletion carrier is shown in blue. What you can see is that as you would expect, they're doing more poorly on this motor sequencing task performance. With the controls shown in yellow, they get substantially better just as we would have expected on this task after a night of sleep.
For the 22q deletion carriers, they do get a little bit better, but not nearly as much as the controls. There's a significant interaction basically showing that there was a lower magnitude of overnight improvement in the 22q deletion cases relative to these typically developing controls. In other words, they didn't see as great of a sleep-related memory benefit as the controls did.
We're just looking at the same data here, just so you can see the individual data points and how different they are in terms of their overnight change in performance. Putting it all together, we see that people with the 22q deletion syndrome have increased sleep duration, as well as increased theta power from our sleepy EG. But also decrease slow oscillatory power and decrease sleep-dependent memory consolidation.
Basically, they are getting reduced sleep-related cognitive benefits, which may be due to these differences in sleep neurophysiology. As I mentioned, this is still work in progress and particularly the measurement of the amplitude and density of sleep spindles and slow waves to determine whether something about diminished thalamocortical sleep spindles, maybe mediating these memory consolidation differences. As I mentioned, the sleep spindles really had this interesting sensor gaiting function during sleep that we think maybe related to sensory processing difficulties associated with autism.
As well as perhaps the memory consolidation and downstream psychiatric symptoms. We are in the process of investigating thalamocortical connectivity using resting-state functional MRI in the same patients that are getting these sleepy EGs to see how connected these things are and how both of these measures are related to sensory processing difficulties and autism diagnosis. Lastly, in terms of future directions, a lot of this has collaborations with UC San Diego I'm really excited about.
We got a grant with UCSD with Jonathan Sabat and others from the Simons Foundation to expand our sleep biomarker data collection to other autism link genetic variants. A 22q duplication, as well as deletions and duplications at 16P11. 2, we'll be doing all of these same measures across these different autism-linked copy number variants.
Then, lastly, we recently got a grant from the California Institute for Genitive Medicine, essentially focused on generating stem cells from patients with these disorders who will also have the sleepy EG resting-state functional MRI. It's going to be really exciting to be able to put all of these pieces together and just a little tidbit. This was from a very recent study from our collaborators at UCSF using stem cells, organoid models, little mini brains derived from stem cells from our patients.
Essentially what they're finding is that the 22q deletion appears to promote thalamocortical axon overgrowth in both organoids, as well as a mouse model. They were actually able to track it to a particular gene, the FOXP2, which seems to be basically over-expressed in these thalamic organoids. These organoid models, I think, are really, really incredible, and I'm really excited to start working with folks at UCSD on the SRM grant.
I think this is a really interesting finding that we're going to be following up on FOXP2 is a high-confidence autism risk gene, and just to see how all these things are potentially related to each other. With that, I just want to profoundly thank all of our funding sources and particularly all the patients and their families that participated, and, of course, what this work is really all about. I will stop there for questions.
Thank you. [APPLAUSE]. I was just wondering if the participants had, levels of autism.
Yeah, the question was about whether participants at all levels, whether there was a variability in the severity of autism, I guess. Yes, absolutely. These are participants that were able to wear this headband at night, although some of them did have intellectual disability.
We're finding that people find it pretty comfortable and are able to do it, even if they're not, particularly what we call high functioning cognitively. There was a range, and some of them had a lot of sensory sensitivity but were able to wear the headband. I'm curious about your diet and your control.
Like, their diet, their exercise between the people that are neurodiverse. This other group. Those are factors that are, as you probably know, very difficult to control.
We get sleep diaries. We get information on medication, on daily exercise, and, not for the whole lifespan, but, we do try to collect as much information as possible on what people are doing when they're participating in the study. Thank you.
