The body-clock is a cluster of neurons connected to the eyes that dictates the rhythm of most bodily functions.
For example, the body-clock, technically known as the Suprachiasmatic Nucleus (SCN), regulates the release of a hormone that determines when we wake up and when we feel sleepy. Research into the function of the body-clock is important: the 2017 Nobel prize for physiology or medicine was awarded to three scientists (Jeffrey Hall, Michael Rosbash and Michael Young) for their significant contributions to the understanding of the inner workings of the body-clock.
The body-clock, which is a cluster of neurons in the brain, has the same structure in all mammals, which is remarkable. It consists of two groups: two-communities of neurons that are strongly linked within each community and less strongly linked between the communities.
The neurons "fire" in a periodic way. This means that they pass through a variety of states, culminating in the release of electricity before the process starts again. In order for the body-clock to function optimally, the firing of the neurons needs to be synchronized to some degree. This is achieved because each neuron reacts to the neurons firing around it, by adjusting the moment at which it fires itself. Many experiments have been done to understand the structure and functioning of the SCN better. These experiments must also be supplemented with mathematical models, which can guide the experiments. One of the mathematical models used to describe synchronization was proposed by Kuramoto in 1975.
Many experiments have been conducted to understand the influence of the light-dark cycle of the sun on the body-clock. These typically involve rats or mice that are exposed to a regular light-dark cycle, before being put in a changing environment. The goal is to see how the behavior of the animals changes as a result of this change. The activity of the animals is used as a measure of the synchronization of the body-clock. How does the experiment work? The day is divided into time slots and the animal is provided with a running wheel, or something of that type, which measures when the animal is active (running in its wheel). Each time slot in which the animal is active is marked appropriately. When the light-dark cycle is regular, the animal will display a single period of activity, represented by a number of consecutively marked active time slots, followed by a period of inactivity or sleep. Since the body-clock regulates when the animal is active, a regular pattern is a sign of a synchronized body-clock.
After the animal displays a regular pattern, one can change its environment, for example, by exposing it to constant light. This leads to an interesting phenomenon. In some cases the animal will start being active in two periods during the day. Recently, some authors have argued that this is due to the neurons in each community of the body-clock being synchronized within the community but anti-synchronized with the neurons in the other community. This state of the body-clock is called the phase-split state, and the activity pattern observed during the transition to this state can be either smooth or chaotic.
We can represent the interactions between the neurons in the body-clock as a network. In our case the network has two communities, where each community has many edges between the nodes, while there are only few edges between the nodes in different communities. The network describes which neurons interact with each other, but it does not yet tell us how they respond to one another. This is where the Kuramoto model comes in. The state of each neuron at each point in time can be thought of as the position of a dot on a circle, or equivalently as an angle between and . Here the subscript tells us which community the neuron is in (so for community or for community ), and the subscript tells us which neuron in that community it is, where we assume that there are neurons in each community, (so ). For the interaction we want that a neuron adjusts the time at which it fires as a result of some other neuron firing before it, either by firing a little earlier, or by firing a little later. The easiest way to do this is to adjust the neuron represented by angle as a response to the neuron represented by the angle via the sine of the difference between the angles, i.e., . Since a neuron is responding to all the neurons it is connected to simultaneously, it will adjust its angle by an amount that equals the sum of all the adjustments it would make for each neuron separately. In order to model the fact that the neurons are positioned in a fluid, we add a small random term to the total adjustment.
The Kuramoto model has been widely studied since it was conceived in 1975. Most studies have focused on the case where all angles interact with all other angles (i.e., the network is on the complete graph). Since this is clearly not reflecting the structure of the body-clock, we analyze what happens if we consider the Kuramoto model on a two-community network. For this purpose, we take the neurons in the same community to interact with a strength given by a parameter and the neurons in different communities to interact with a strength given by parameter , which can be positive, zero or negative. When is positive the two communities try to align with each other, when is zero there is no interaction and when is negative the two communities repel each other. We are interested in studying which states are possible for the system as a whole when we fix and and consider the number of neurons in each community to be very large.
What may we expect? In the Kuramoto model with the all-to-all interaction i.e., all neurons interact with all other neurons with strength , synchronization appears as a threshold phenomenon. This means that for values of below a critical value the neurons fire completely randomly, while for values of above the neurons start to fire in a synchronized way. When exceeds , synchronization is measured by what is called an order parameter. The order parameter for the synchronization is , which takes values between and , where represents the case where the neurons are completely unsynchronized and represents the case where the neurons are completely synchronized. We may therefore expect that synchronization in the two-community Kuramoto model will also appear as a threshold phenomenon.
If the environment in which the body-clock resides remains constant, then we may expect that the system will reach a stable and stationary synchronization level after some time. This means that the synchronization level will remain the same as long as the environment does not change.
Of course, in the two-community case we do not only want to know the level of synchronization in the body-clock, but also the synchronization in each community. In order to characterize all the possible stationary states, we also want to know the difference between the average angles in the two communities, which we call the phase difference and denote by . We found that the phase difference can only be or in the stationary state. In both of these cases, we can split the possible stationary solutions into three categories. In category (I) both communities are unsynchronized; in category (II) both communities are synchronized to the same degree (symmetrically synchronized); in category (III) both communities are synchronized, but one community is more synchronized than the other (nonsymmetrically synchronized). The two-community version of the model has two parameters, and , which means that we do not have a single critical value, like in the one-community version, but have a critical line. The critical line, splitting the unsynchronized state (category (I)) from the symmetrically synchronized state (category (II)), is . The nonsymmetrically synchronized state (category (III)) is only possible in a part of the symmetrically synchronized section of the plane and requires when and when . The critical line dividing the symmetrically synchronized state and the nonsymmetrically synchronized state is identified as the solution to an equation involving special functions, which means that we cannot write down the solution explicitly. In order to visualize the situation, we plot the critical lines in Figure 1 below and indicate in which regions which solutions are possible.
The previous section dealt with the possible stationary states. In the experiments with rats and mice one of the most interesting features that was observed was the possibility to see two different types of transitions to the phase-split state, which is certainly not something that can be understood by studying stationary states. In our model we did, however, see that stationary states exist in which the two communities have a phase difference of , which corresponds to the phase-split state observed in experiments. To study the transitions between these states (the one with and the one with ) is a mathematically intractable problem at the moment, but it is possible to see what sort of transitions are possible. The possible transitions that we observed are shown in Figures 2 and 3 below. Here, the average angle for community 1 and community 2 are given by and , and the synchronization levels are given by and .
The simulations also show two different transitions between the symmetrically synchronized state with and the symmetrically synchronized state with , and back. In the first type of transition, both communities move to a higher level of synchrony together while their average phases either move further apart or move closer together. In the second type of transition, one community is forced to a much lower level of synchrony before it moves to the higher level of synchrony, while the other community moves directly to the higher level of synchrony. The first situation could possibly correspond to the smooth transition observed in the experiments, while the second transition could correspond to the chaotic transition.
There are two factors that seem to play a role in determining which transition occurs. The first is the initialization of the system and the second is the existence of the nonsymmetrically stationary state. When we start the simulation, we have to chose an initial state for the angles of the neurons. In experiments this state is difficult to interpret, but we can think of it as the situation that prevails immediately after a change of the environment in which the animals live has occurred (e.g. switch from light-dark cycle to constant light). The second factor is more easy to interpret: the strength of the interactions in the network is given by and , and if these are such that the nonsymmetrically synchronized state is possible, then the system might be forced to the nonsymmetrically synchronized state before it can move to the phase-split state.
The body-clock is a complex network of neurons that is difficult to model mathematically, due to its many intricacies. Above we have seen that we can gain some understanding by using mathematics and simulations, but also that experiments uncovering interesting phenomenon can play an important role in guiding mathematicians in making choices about which questions are interesting and worthwhile to study.
Specifically, we have found a possible mechanism for the different types of transitions to the phase-split state observed in experiments. Which transition occurs will depend on the interactions strengths, which may be different in different individuals. Much remains to be done in studying the two-community Kuramoto model, and even more when it comes to understanding the body-clock.