Using FlyVISTA to Study Sleep in Fruit Flies

The brain has the important job of taking in information from both the body and the outside world to help us act appropriately. Researchers study how animals behave to find out more about this process. New technology, especially in computers and machine learning, has created a new field called computational ethology. This field focuses on measuring different aspects of animal behavior automatically. Most of these studies have looked at larger movements over short periods of time.

For over 150 years, scientists like Darwin and Tinbergen have pointed out that even small movements of body parts or changes in posture can signal important feelings such as fear or anxiety, especially in animals like cats and gulls. These small movements, known as “microbehaviors,” are very important but difficult to study. The challenge comes from accurately detecting these tiny movements and analyzing high-quality images over long periods.

To tackle this issue, we created FlyVISTA, a system using machine learning to describe and measure these microbehaviors in fruit flies, specifically Drosophila. This system uses high-definition video to track the flies while also allowing researchers to introduce changes in their environment using infrared lasers. FlyVISTA employs a machine learning model to label 35 different body parts of the flies. After labeling, it extracts meaningful behaviors from these labels over time.

We used FlyVISTA to study Sleep in fruit flies. Sleep is a basic behavior found across many species, but its purpose is still not fully known. Most research on sleep has focused on mammals, but studying sleep in simpler organisms like fruit flies can provide valuable insights. Analyzing the sleep of fruit flies may help us understand more about the essential functions of sleep and how it has evolved.

Traditionally, researchers have looked at sleep in fruit flies by checking for periods of inactivity lasting more than five minutes in a small tube. However, this method has its flaws. It might miss short movements or small behaviors that are also part of sleep. Additionally, keeping flies in narrow tubes can cause stress and affect their natural behavior. Therefore, we aimed to develop a system that provides a more accurate view of sleep behavior in freely moving fruit flies.

Sleep is generally considered to be a quiet state. Different microbehaviors can be observed during sleep. For example, in mammals, the lack of movement or certain quick eye movements can indicate deeper sleep or a specific sleep phase. Previous studies on bees and cockroaches have shown similar small movements connected to sleep.

With FlyVISTA, we found a variety of microbehaviors during sleep in fruit flies, including relaxation of the body posture, drooping antennae, and rhythmic movements of the proboscis. We also tested how activating two specific neural circuits involved in promoting sleep affects microbehaviors. One neural circuit, called dFB neurons, was linked to increased small movements rather than sleep, while another, called R5 neurons, promoted proboscis extensions followed by sustained sleep.

Interestingly, we identified a new microbehavior named “haltere switch” which refers to a specific movement of halteres. Halteres are small organs in flies that function similarly to the vestibular system, which helps in balance in mammals. This switch movement was only seen during quiet times and seemed to mark a deeper level of sleep.

#High-Resolution Video Imaging

To study these behaviors better, we designed a setup that allows flies to move freely and be recorded in high detail. In our chamber, we used a side view to capture more behavior, as research on other insects showed that important changes, like lowered antennae, are better viewed from the side. This setup allowed us to see the fly in much more detail compared to other methods.

To track the fly's movements, we used DeepLabCut, a software tool that helps label and analyze body parts by using a trained model. With this tool, we tracked 35 points on the fly’s body. This information helps us understand how different body parts move in relation to one another over time.

We recorded individual flies for long periods and examined their activity. As expected, we noticed peaks in movement during certain times of day. At night, flies showed long bouts of inactivity, which we linked to sleep.

During these quiet periods, we identified various microbehaviors. After becoming still, a fly’s body posture would relax under gravity. The antennae would droop, and rhythmic movements of the proboscis were also observed. These movements were not exclusive to sleep, as they would also occur when the flies groomed or fed.

Additionally, the haltere switch behavior, which involves a downward movement of the halteres, was consistently observed during quiet times. This particular behavior has not been noted in previous studies and illustrates the variety of microbehaviors present during sleep.

#Assessing Arousal Thresholds

A critical aspect of defining sleep is the idea of an increased arousal threshold. To test this, we used a special infrared laser that flies cannot see. By gradually heating the flies after they had been inactive for a while, we measured how much energy was needed to get them to move again. Each fly's response was unique, so we standardized the measurements for comparison.

As expected, we found that the amount of energy required to disturb the flies varied throughout the night. The threshold was low at certain times and much higher during other periods, suggesting dynamic changes in sleep depth across the night.

Most studies have defined sleep as periods of inactivity lasting over five minutes. However, our findings suggest that shorter periods of inactivity of three minutes are more meaningful and should be considered as indicative of sleep. Flies were seen to stay close to their food during these quiet moments, indicating that they might prepare for sleep nearby.

Given that our system is precise and does not involve moving the flies, we could observe how their microbehaviors changed as they woke up. We noted that certain movements of the halteres and antennae often occurred just before the flies became active again, suggesting a connection between these small movements and the transition from sleep to wakefulness.

#Activating Neural Circuits and Behavior Analysis

To further study sleep behaviors, we manipulated two neural circuits known to influence sleep in fruit flies. One known group of neurons has been linked to rapidly inducing sleep-like states, while another group is connected to overall sleep need.

When we activated the first group of neurons, we noticed a decrease in movement but not actual sleep. Instead, we saw that the flies exhibited more micro-movements and grooming behaviors, indicating that activating these neurons does not promote sleep.

However, activating the other group of neurons led to increased probing movements of the proboscis during stimulation, followed by lasting sleep. This response indicates that these neurons may play a crucial role in the need for sleep and the physiological processes connected to it.

Our study with FlyVISTA allowed us to classify different behaviors before, during, and after these neural activations. The results showed distinct behaviors linked to the activation of the different circuits. Flies whose dFB neurons were activated did not enter a sleep state despite their reduced movement, whereas activation of R5 neurons led to increased proboscis extensions and a following sleep state.

#Automated Behavior Recognition

One of the main goals in studying animal behavior is to measure it automatically. Although tools like DeepLabCut have made it easier to label movements, extracting meaningful behaviors from these labels remains a challenge, especially for subtle microbehaviors related to sleep.

To address this, we developed a system within FlyVISTA that classifies behaviors from unannotated video data based on a set of manually annotated examples. The system assigns labels to five behavioral categories based on the observed movements of the flies. However, due to their subtle and varied nature, classifying some behaviors proved more difficult than others.

Through our novel computational pipeline, we were able to distinguish flies’ behaviors with relative success. For instance, our system accurately identified proboscis extension and feeding behaviors more than others. However, we found that the behavior involving haltere movements was harder to track reliably due to their subtlety and the difficulty of observing them consistently.

#Measuring Sleep with FlyVISTA

Using FlyVISTA, we quantified sleep in fruit flies based on our criteria of observing periods of inactivity lasting over three minutes, excluding grooming and feeding. Our findings showed that sleep patterns in freely moving flies are more consolidated than previously reported in tethered flies.

We analyzed sleep amounts and durations during specific periods of the day. Female flies generally maintained higher sleep levels throughout the night, whereas male flies showed a significant decrease in sleep toward the later night hours.

To compare our results with traditional approaches, we used another method of measuring sleep in small tubes. We found that the amount of sleep detected using conventional methods was higher than what we observed with FlyVISTA. Using FlyVISTA led to a narrower range of sleep duration, indicating that flies likely have a standard length for sleep bouts.

After mechanically depriving the flies of sleep, we observed increased sleep duration and amounts in both male and female flies, confirming that our approach using FlyVISTA is effective in measuring sleep behavior.

#Proboscis Extension Behavior in Fruit Flies

We also investigated how the microbehaviors of fruit flies, specifically proboscis extensions, change over time and under different conditions. Under normal circumstances, the frequency of proboscis extensions remained steady throughout the night, without showing a clear decrease as had been previously reported.

Interestingly, these extensions mainly occurred during sleep, suggesting a strong connection between this behavior and sleep states. After sleep deprivation, there was a noticeable increase in proboscis extensions, indicating homeostatic control.

This relationship provides insights into how such behaviors might be regulated differently under various conditions. It highlights the importance of studying these microbehaviors to fully understand the sleep architecture in fruit flies.

#Haltere Switch Behavior: A New Insight

Finally, we focused on the haltere switch behavior which shows promise in defining deeper sleep states in flies. Evidence suggests that these haltere movements are closely linked to sleep, occurring almost entirely during periods of sleep.

We found that the first occurrence of haltere switches typically happens shortly after the onset of sleep, indicating that this behavior can signify a transition into deeper sleep quickly. Additionally, we discovered that when the halteres were in the “down” position, the flies had a higher arousal threshold, suggesting these switches indicate a deeper sleep state.

Furthermore, we noticed various types of haltere movements, including rhythmic oscillations that were often accompanied by other subtle movements in the fly's body. These rhythmic movements point to the idea that there could be different sleep phases that need further exploration.

#Conclusion

In summary, the use of FlyVISTA has opened new avenues for studying sleep and behavior in fruit flies. By combining high-resolution imaging and advanced algorithms, we can better understand the complex nature of sleep and identify subtle behaviors that were previously overlooked.

Our findings about sleep-related microbehaviors, the role of different neural circuits, and the dynamics of sleep using FlyVISTA provide meaningful insights into the nature of sleep in fruit flies. As we continue to refine our techniques and analysis methods, we will uncover even more about the intricate relationships between behavior, sleep, and the underlying neural mechanisms. This work not only enhances our understanding of fruit fly behavior but also contributes to broader discussions about sleep and behavior across species.