Circadian Metabolomics: Sleep, Food Timing & Human Clocks

[00:00:00]So what I'd like to do today is is focus more on uh the recent research we've been doing in my laboratory. Uh and that's really involved uh metabolic profiling or metabolomics and I would like to focus on human clocks and and sleep and food timing in human participants. Uh but indeed um we have also uh with collaborators looked at metabolomics in various animal models uh from the VO to the mouse the sheep and have done similar types of experiments sleep deprivation circadian rhythms and shifting the feeding fasting cycle. So if you're interested in those uh the references are are down below. Um right now I don't have to tell this audience that we have a master clock in the hypothalamus the super kismatic nuclei that is um capable of independent oscillation when you put it in a dish and of course we now know a lot more about um peripheral clocks and how they work as well. And uh just to see it as a whole concept here, of course we need uh synchrony uh between these internal uh peripheral oscillators and the central clock. Um and we also need synchrony uh between this this circadian timing system and the external environment. And the whole idea behind it of course is so that we can anticipate events uh but also optimize uh our function depending on the time of day.

[00:01:47]So that for example for humans uh we need to sleep at night and metabolize uh food during the day. Um now many uh reviews and much work has been done in the last um 15 years or so linking circadian clocks and metabolism.

[00:02:09]Uh and uh of course when we look at the peripheral clocks um in uh a bo in the body a lot of them are dealing with metabolism.

[00:02:22]uh and so uh they uh are set to uh coordinate and respond to the fact that humans uh we eat during the day. And if we think of um the circadian timing system, it's obviously involved in energy utilization and storage across the 24hour uh clock. And so it's important that we have these timed events during the wake time and our feeding time and then different uh metabolic pathways being activated here when we're sleeping and fasting. So it's it's quite um uh it's an orchestrated system. Uh and and because of this when we start disrupting this orchestrated system either um by disrupting circadian timing or um shortening our sleep and having sleep deprivation, these events have been shown in many studies now to be um associated with metabolic disturbances and metabolic disorders like obesity and type 2 diabetes. IBES and none so important as in the um work and the epidemiology studies about shift workers. Uh shift work of course involves sleep deprivation and circadian desynchrony and uh we see the increased risk of many metabolic disorders um when this occurs. So that is all fairly well known. I mean what isn't known though is the intricate mechanisms that go on behind uh these events. So what is linking metabolic disease, circadian misalignment and sleep uh shortening sleep deprivation and uh many uh people have started addressing this problem uh with OMIX technology that might give more of a um detail about the underlying mechanisms. And we opted about 10 years ago now to study uh metabolomics to see if this could elucidate these mechanisms. U now why metabolomics? Uh well uh when you see it goes from of course genomics, transcroics, proteomics, metabolomics. Metabolomics is a better representation of the functional phenotype uh than DNA, RNA and proteins and it uh reflects um homeostatic regulation and if the system is then disturbed. Now, if you've heard of metabolomics, it's usually in uh the context of looking for biomarkers, metabolic biomarkers to see if you could diagnose the disease, whether you can track therapeutics and it's often used in this um concept of personalized medicine. uh and you know there's a lot of papers about this but as chronobiologists you'll recognize that we needed to do some very basic experiments in the early days just to see whether um this u these profiles change across time of day. Uh then what is the effect of um sleep, wake, feeding, fasting cycles and meals etc on these profiles and also what is the endogenous circadian variation? Are there circadian rhythms in uh these metabolic processes in humans? Um and so a very simple first experiment was where we had an entrained protocol in the lab.

[00:06:32]So a bit of real life in the lab to assess time of day variation uh and also the effect of sleep and sleep deprivation.

[00:06:43]Now uh you can do unargeted or targeted metabolomics using liquid chromatography mass spectrometry and we opted to do both in the early uh experiment in this early experiment. Now we're fortunate enough at Surrey to have uh individual sleep rooms that are temperature, light and sound controlled. The other uh advantage here is that we have um if we don't want to disturb sleep uh we can uh uh pull blood through the wall from the volunteer without having to go in and disturb um the participant.

[00:07:23]And this was uh applied here. So here is an entrained um protocol. Uh I mean entrained because we have a light uh we have light during the day, we have dark at night, we have meals um and uh we have um the ability here to move around.

[00:07:44]So it's like what you would do in your own homes but in a very controlled lab condition. And following a full 24 hours or more of this day night cycle, the same people um were then submitted to sleep deprivation here. So we can directly compare um u sampling here during the sleep deprivation night compared to the sleep deprivation and they were wired up for polynography in this protocol.

[00:08:18]Right? So this is the unargeted uh uh profile. Uh we isolated 367 features. So this is a profile of all of these features and how they vary across time of day. So I think you would agree with me that you can see a clear time of day variation there. And then when um uh the uh participants were sleepdeprived here, we got this reduction in amplitude uh during this night of sleep deprivation. And this was uh 12 subjects here, 12 participants, all healthy young men.

[00:09:01]Now the bottleneck in unargeted metabolomics is that once you you identify a feature but you don't really know what it is until you do a lot more work. So you have to get the accurate mass of this feature. You can then search databases and when you think you know what it is, you have to if they're available, buy a standard, put it on to your instrument and be absolutely sure that it is uh the feature is confirmed as the standard. And this can take an awful long time, a lot of time. And it requires um you know quite a lot of expertise. And we were able uh with these features to identify about 40 known metabolites out of that 367 and these were mainly amino acids, acetyl carnitines, phospholipids, billy rubin and cortisol and cortisone.

[00:10:05]Um now because it is reverse phase chromatography, we weren't able to pick up these polar metabolites. So it is a restricted view of the metabolites that exist.

[00:10:20]Um and because the metabolites that we identified were so close um to this targeted uh metabolomics kit that you could purchase. Um we opted to move over to what we call targeted metabolomics.

[00:10:38]So in this case um you buy the kit and it is set up to measure about 180 metabolites.

[00:10:47]Um the good news is you don't need much plasma. So we can do this in mice and rats and small rodents as well as humans. But the other uh really nice point about targeted analysis is you have standards and you can set up standard curves uh and you can semi-quantify or quantify the metabolites and you can have quality controls. So it improves the reproducibility and reliability of your technique. So here we have the same um sample set different alleyquats of the samples in a different mass spec but you see a very similar time of day variation and again a reduction in amplitude during the sleep deprivation and when we compared the targeted and the unargeted metabolomics together. Um so these as I say different uh sample alquat different massspec instruments but all tracking these amino acids on top and some carnitines here you see it beautifully uh went together. So we were pleased uh w with that analysis.

[00:12:06]So just to show you some of these uh what I'd call dal or daily rhythms in metabolites. Um you can see glutamate uh you can see SDMA and uh you of course you have these two peaking at different times of the day and not much change in amplitude here um during the sleep deprivation shown here in the yellow.

[00:12:35]Here's some more just to get give you a sense of of these rhythms. an an acetylc carnitine, a lyso phospholipid, one of these glycerero phospholippids here and this is a sphingo lipid. So the we get this clear daily rhythm in metabolites.

[00:12:59]The other thing we can do is we can directly look at the metabolic profile here uh during sleep and compare this with during sleep deprivation.

[00:13:11]And this was in our males. And uh when you do supervised um uh uh opalsda so discriminate analysis where you can group uh your data into either uh sleep or sleep deprivation you can see a clear separation between uh these two groups.

[00:13:36]uh that's a validated model and you can then also do parametric um or uh what I'd call univariat analysis here two-way annover um and you can see that we had some metabolites that were significantly affected in these two conditions and all of the metabolites only 27 but all of them were increased during sleep deprivation and most of them were lipids. And this agrees with uh animal work that then came after us uh after the study also showing um lipid and fatty acids um uh changing during sleep deprivation. Now that was 24 and the other three metabolites caught our interest. There was one amino acid here tryptophan and you can see that is increased compared to during sleep deprivation as was serotonin and torine and uh what I didn't say is that we also measured melatonin classically using our radio assay and this also increased during sleep deprivation compared to during sleep. And for those of you who know um the uh pathway of um synthesis of melatonin in the pineal gland uh elptophan, serotonin and melatonin here all increasing some sort of activation of that pathway during sleep deprivation.

[00:15:19]uh which of course needs uh confirming and uh now that we know this we could directly more uh do studies to to to investigate this in more detail.

[00:15:34]Right now you're probably saying yes and we should have said and we did say what about females and also another um uh addition to the protocol was what about h what happens during recovery sleep. So we did both of those uh in our next series of experiments and this is also published. So the details of all of this the intricate details please um move um see see that. So here we have um the normal 24-hour light dark cycle then sleep deprivation here followed by a recovery sleep. So we had um one to three days of blood sampling here across the time and as I said this has uh been published. Um we only use the targeted metabolomics approach in this study and uh just to deal with some of the similarities uh we found clear daily rhythms in metabolites like we had seen in the men and most of these rhythms continue to be rhythmic during sleep deprivation and uh we then also So redid the male analysis in a similar way and here we see again uh that the males have these clear rhythms and most of these are maintained during sleep deprivation. So it seems that the rhythms aren't greatly affected by total sleep deprivation or sleep recovery.

[00:17:20]And at the same time the increase in uh melatonin that we saw in the men, we also saw here in the women and this then uh went back down to baseline levels during the recovery recovery night. So a confirmation of the data we saw in the men and we also saw an increase of torine and tryptophan in these women uh but was more accentuated in in the males.

[00:17:57]But when we came to looking at the metabolites that were discriminatory between the sleep and the sleep deprivation, these were the metabolites in the men that were different and these were the metabolites in the females. And this was a striking difference. First of all, because all the metabolites in the men increased during sleep deprivation, all positive here and here all but one of the metabolites decreased during sleep dep uh sleep deprivation. So remarkably different, not a single common metabolite change and theonin was the only one that was increased significantly. Um but again no overlap with the males.

[00:18:59]The sex differences uh we really need to study this now in more detail. Um it has been suggested in animal studies that the uh they respond differently to sleep deprivation and a recent study a human study by in Frank Shar's lab had shown sex differences in energy regulation. Um but we need to really look in more detail what the underlying basis of these differences might be. Of course, it might reflect the sex differences in metabolic rate that men and women have and how they utilize energy.

[00:19:39]Right? So, at this point, um we've got daily rhythms um in the metabolites and uh we've looked at time of day, the sleepwake site, the light dark cycle.

[00:19:52]The next phase was to look at what is the contribution of the indogenous circadian timing system. And to do that we have to employ the well-known uh gold standard protocol the constant routine protocol where we remove or minimize the effect of all of the external um uh things that could affect a rhythm. So the confounders so to speak. So no knowledge of clock time, constant dim light, less than five um lux at the in the direction of gaze, minimal social interaction, no big meals, only these very small isoc caloric snacks.

[00:20:38]And people uh previously had looked at this. So we weren't the first to look at this. Uh the first original study was uh by Robert Dolman and Steve Brown's group. uh but they they had pulled their samples uh across subjects and across time points. Um and so and uh again Japanese group had studied six people um but but no women. So we wanted to u continue this work. We went for very high resolution sampling hourly and two-hourly. We wanted to assess whether there were any sex differences and we also looked at urine which I won't talk about today. Right? So here's your constant routine protocol. So following an adaptation of just them getting used to the lab a meal and sleeping over uh they then start the constant routine. So it's a 40hour constant routine classic.

[00:21:42]Um and from about uh 1,500 hours we started hourly blood sampling in both men and women all young and not medicated no extreme chronotypes uh but uh at least uh we did both sexes and again we recording uh EEG polyomnob free throughout to check that they're not sleeping because this is a sleep deprivation constant sleep deprivation here. Okay.

[00:22:17]So, since then we've been employing a lot of metabolomics platforms. Um, our reverse phase is what we normally do at Surrey. Uh, we did targeted two-hourly.

[00:22:29]We also did unargeted hourly. Um and then in collaboration um with uh our colleagues in Birmingham, they've looked at lipidomics unargeted and also normal phase. So this is what picks up polar metabolites. And so we're trying to get the whole extension of the metabolom coverage here. And I have to thank uh for this work and this is uh mainly unpublished work. Um this is my PhD student Namata Chowry from Surrey and my collaborative PhD student Thomas Hanox who is in Birmingham.

[00:23:10]Right. So here we have clear sex differences uh even at the PCA well not quite at the PCA level uh males and females but when you supervise this analysis a clear sex difference in the unargeted uh analysis and similarly um in the targeted metabolomics analysis clear sex differences and you can even see that in um the PC1 versus PC2 component here.

[00:23:49]Right now, when we're looking at what is causing the sex difference, you can look at that by looking at a loading plot.

[00:23:56]And you can see that the amino acids shown here in blue are higher in males, higher biogenic amines, those are the green and acetylc carnitines in green uh and the lyso pcs. Whereas in females mainly phospholipids and sphingo lipids and this has been shown before. So this isn't novel. It's just what's determining the difference between the sexes to remind you that you have to control for sex when you're doing metabolomics. Uh and the other thing that uh we are now seeing and this is largely uh Tom Hancock's work. He's looking at the rhythms um in these metabolites and where their peak times are. and he's also getting sex differences in um his lipidomics and helic metabolomics profiles between males and females and you see that here.

[00:24:59]So when we drill it back down to actually looking at the rhythms and this is the targeted analysis I wanted to show you some um metabolites that are rhythmic in constant routine. So if something is rhythmic in constant routine, we can say it has a circadian rhythm because we've removed all the external um factors that may cause that rhythm. Um and here you see in males and females also glutamate and proline lysopcs are rhythmic as is tryptophan only um significant in the females. So around 20% of the metabolites that we can measure um were rhythmic in constant routine.

[00:25:52]Now because we've done these different protocols where we've done this dial in train protocol that I've already shown you as well as the constant routine and um we can now put these profiles together and this allows us to assess the effect of the food and the sleep and the fasting on the metabolite profiles and this is what my student Namata has been doing. So if you just look at melatonin for a moment and cortisol, these are the SCN driven rhythms, you can see no changes in cortisol and melatonin um uh when uh you are in a constant routine or an entrained protocol.

[00:26:44]However, let's look at the amino acid alanine. Now if you look at the red profile here, this is what we saw um in the entrain protocol. So you can see a clear effect of the meal because these uh dotted lines show the meal. So you can see it goes up after each meal, breakfast, lunch and dinner. And then you see a clear effect of the fasting.

[00:27:14]And of course we can't say if it's effect only of the fasting or the fact that they're sleeping. Um but when you look at the blue profile it's a completely different rhythm and this is the rhythm of alanine the circadian rhythm of alanine that has a different peak time to um the daily profile.

[00:27:41]Similarly here, just to show you a few more, it's mainly the amino acids that are showing an acute response to meals.

[00:27:50]So if we look at the red lines again, clear effect of meals in this female protocol. Um, and the same with phenile alanine, an effect of meals um on the profile. And when you don't have these big meals rather these sort of hourly sandwiches then you see a very different profile in blue.

[00:28:19]Um just to show you some of the acetyl carnitines and phospholipids. Um the beautiful daily rhythm shown in red of one of these phospholipids and this lingo lipid. beautiful daily profile where you get this reduction during sleep and fasting and then this is completely uh abolished here. No rhythm at all when we in constant routine. So no circadian rhythm in these compounds um when you remove the external um factors.

[00:28:56]Right? So in summary of this, we get sex differences. We have clear daily rhythms in plasma metabolites. But these seem to be predominantly driven by our environment, our feeding, fasting, our sleepwake, our light, dark, and our rest activity. And only about 20% of the metabolites are exhibiting circadian rhythmicity, endogenous circadian rhythmicity.

[00:29:26]So now that we've done this uh these experiments, the whole point really was to have some knowledge and databases to then be able to apply this technology to studies in shift workers. And so I was helped here by Hans Vanongan's team at Washington State uh University uh because they were doing a simulated shift work study in the laboratory and they had collected samples and we did our targeted um metabolomics approach on their samples.

[00:30:04]So let me show you the protocol. Um and here you see a day night uh a day shift condition uh sleeping at night and eating during the day and then the night shift condition where they have three nights here of night shift. Now important part about this protocol is that it was only following the shift work conditions that we sampled um the participants both in a constant routine. So when the samples were taken, they were taken in the same environment, but what they had had previous to that environment were was different. And out of um the metabolites that displayed rhythmicity, um we had about 65 of them that showed rhythmicity. And we basically got three principal clusters here. Um we had some the biggest group was 27 metabolites that were rhythmic in both the day and the night and I'll come back to that.

[00:31:15]Some were only rhythmic in the day shift. That's shown here in orange 19.

[00:31:20]And then 19 were only rhythmic after the night shift.

[00:31:26]Now coming back to uh this um biggest cluster and and this you know kind of blew our minds because here we have a metabolite that's rhythm is completely reversed after three nights of night shift. um the night shift profile here in purple.

[00:31:49]Um and the group this happened to quite a lot of 24 um of these metabolites were mainly glycerero phospholippids, fingolipids and amino acids. I'll show you a bit more. Now, why we were so shocked is because we and others have shown for many years that if you have three nights of night shift, your melatonin and your cortisol hardly move. SCN driven rhythms in humans are sluggish and do not shift uh easily when you go into night shift.

[00:32:26]And if you look here on average it was less than 30 minutes a day that melatonin was moving. Uh likewise here maybe 40 minutes a day that the cortisol was moving. And also uh per three expression in lucasytes the white blood cells again um here probably about 50 minutes per day shifting. So this was very different to um the metabolites and you can see it better here because we've got the 27 metabolites the peak time in the day shift and the peak time in the night shift. Now anything along this line is hardly moving. And here you see dilmo melatonin and here is cortisol and three other metabolites torine, serotonin and sarazine that did not shift. So they likely are more aligned and perhaps more driven by um melatonin, cortisol or the SCEN. But these other metabolites are really much more um further away in the different conditions. And so what um what we interpreted from this is first of all the rhythms in the metabolites mostly dissociate from the SCN pacemaker rhythm that is uh exemplified by the melatonin the cortisol and we are thinking that by tracking uh plasma metabolites we've got more of a window in how the um uh the shift work moved uh in time and so we are um excited by the fact that we can maybe by um measuring metabolites in plasma that we're tracking possibly the peripheral clocks uh and of course this will give us a lot more insight later on so I've added there are a lot of advantages of um uh metabolomics And these are your classic ones. If you read any papers about metabolomics, it's highly conserved.

[00:34:50]Most of the metabolites are known. Uh but we've added here useful may be useful for tracking peripheral clocks.

[00:34:58]So watch that space. Uh but of course you could say well when you do shift work you shift the feeding fasting, you shift the sleepwake, you shift the rest activity, and you shift the light dark cycle. And so the critical question is you know what is the relative contribution of these um changes you know which one is shifting that metabolite rhythm so much uh and why do we want to know that well if because this is a circadian misalignment of course because you've got your SC CN clock melatonin in one place and you've got your metabolite rhythms completely uh peaking at another time. So you've got this internal desynchrony and if we know what is causing that then we can try and minimize that by designing uh better shift schedules.

[00:35:57]We don't know yet and I'd love to do it and we're trying to get money to determine the relative contribution because you can design experiments where you just change one of these and not all all of it at once. Uh and the closest we've got to that at the moment is an experiment that was run uh by Jonathan Johnston in our lab. Here's the team and we were the first in this study to show that if you shift the meals only, you can have an effect. So this this experiment was a fivehour delay in meal times. Um, and you can see here's breakfast, lunch, and dinner. And here's a constant routine. And then we shifted the whole breakfast, lunch, and dinner by 5 hours. And then a constant routine.

[00:36:49]So then we could compare the constant routine one and the constant routine two. And hypothesis uh was proven that we showed no effect on the SCEN driven rhythms in this way.

[00:37:04]Look at them. They're completely superimposable.

[00:37:07]Uh melatonin and cortisol perhaps a small something there but not significantly uh different. So the SCN was not shifted by changing food and that agrees with the animal work of course as most of you would know. But when we came to looking at the glucose in the constant routine, we saw a 5hour delay in the glucose rhythm. So that is changing following the the 5hour delay in meal patterns.

[00:37:42]The triglycerides didn't change and nor did insulin. So of course we wanted to see what happened about the metabolomics.

[00:37:53]Um, oh, hang on. Before I go on to that, I should say that we also looked at the same time at um per two and per three rhythms in um adipost tissue. Um and these were from adipost tissue biopsies in our participants. And here we had um clock gene expression in the lucasytes.

[00:38:21]And here no change in the blood lucasite um uh pattern but about a 1hour delay here in the per two and per three in adipost tissue. So clearly and this has also been shown in animal experiments there's a differential response to um time feeding.

[00:38:47]Sorry about the uh the background noise.

[00:38:50]I live in London so it's constant. Um and so what we've done in uh recent experiments is to look at the plasma meta metabolite profile here. And what you can see similar type of acryphase in constant routine one versus constant routine two. And most of the metabolites that were rhythmic in both conditions we had 29. And most of them are shifted by at least 2 hours. That's 72%.

[00:39:24]And half of them are shifted by four. So yes, um the um this protocol is shifting the rhythm in the metabolites the same way it is shifting the rhythm in plasma glucose. But this is unpublished, right?

[00:39:41]And so this is why we're so excited uh that we can uh provide a baseline for future studies. I think I'm going to go straight on to um uh I just want to go on to my final slides here um to say that um in conclusion then um I think that metabolomics has a role to play in trying to find mechanisms involved in in these processes. We've shown that tryptophan serotonin metabolism might be important in sleepwake regulation as well as the carnitine system and this has been supported now by other studies.

[00:40:26]Um I think it's a powerful tool to look at mechanisms uh promise being able to track peripheral clock function uh when we are in simulated shift work. Um uh of course the holy grail is to find biomarkers uh for this uh disruption monitor uh recovery from circadian misalignment or sleep deprivation and uh this is the part I didn't have uh time to talk to you about but uh the potential here to have ambulatory sampling uh in real life conditions because of course all these experiments were done in the highly controlled lab conditions. And of course, we're all moving out now to try and do this type of monitoring in the field. Uh, and this has has good potential. So, I will end there. I would like to thank all my collaborators that have contributed to the data that I've shown you today and thank you for your attention.