Fast 3D Confocal 2-Channel Fluorescent Microscope
From sensing to behavior, biological information processing in C. elegans
Core Francisco Park
Vladislav Susoy, Aravinthan D.T. Samuel
Abstract
Biological organisms have evolved for millions of years to adapt, survive and reproduce in the wild. In order to forage resources, maintain homeostasis and avoid predation, these beings must efficiently and accurately process sensory information and produce a suitable behavior. In this work, we dive into the thermotaxis behavior of Caenorhabditis elegans (C. elegans).
When cultivated on food at a certain temperature, Tc, C. elegans remembers that temperature and navigates back to this temperature when placed on a thermal gradient. The neuron AFD acts as the main “thermometer” of this behavior, sensing thermal signals as low as a change of 0.01 °C.
Previous works have demonstrated that AFD stimulates its main postsynaptic partner AIY in order to generate thermotactic behavior. Recent works has revealed the role of the worm’s intestine in controlling the thermotaxis behavior. However, most of these works rely on immobilized animals with a temporal thermal stimulus which are informative but are somewhat not similar to the stimuli the animal will experience in a natural environment.
In this work, we discuss the experimental setup to enable the measurement of Ca2+ signals of many neurons from a freely behaving worm performing a thermotaxis behavior. Our goal is to reveal how a thermal stimulus, an external variable, is processed in a biological network to produce the desired behavior. We discuss the transgenic worm required for the experiment, the spinning disk confocal microscope to capture volumetric images in real time and the convolutional neural network based algorithm to track individual neurons.
Acknowledgements: CFP thanks VS and ADTS for academic and research support. CFP thanks Helena Casademunt, David Zimmerman, Stanislav Lazopulo, Jade Ho for useful discussions and help in daily lab work. CFP acknowledges Vivek Venkatachalam for the original microscope setup. CFP thanks Sahand J. Rahi, Mahsa Barzegarkesheli, Ariane Delrocq for co-developing the GUI. CFP thanks Hyerin Cho for discussion and support.
Tracking & Analysis
Results
Conclusion
Organisms have evolved for millions of years in nature to efficiently process sensory information by their neural networks to produce the adequate behavior. The nematode C. elegans, which has its whole brain connectome mapped out, is a perfect organism to understand the very basics of how these information processing functions.
In this work, we discussed the experimental setup and data analysis pipeline together with the CNN based tracking algorithm which, altogether, enables in vivo whole brain recording from a freely behaving C. elegans under thermal stimuli. Until very recently, neuroscience in C. elegans was focused on many single neuron recordings. Having simultaneous recordings of multiple neurons will help reveal the causal relation of neuron activity and take a step closer to understanding the neural basis of behavior.
C. elegans thermotactic behavior
Figure 3. Design of the fast 3D 2-Channel Fluorescent Microscope�(Original Design: Venkatachalam et al. 2016)
`
x40
Piezomotor
Spinning Disk 5000 RPM
Trigger
Ethylene Glycol Cooler
Motorized Stage
Cooled CMOS Camera 2
Cooled CMOS Camera 1
PID controller
LASER
488 nm
561 nm
Peltier
Agarose
Immersion Oil
DAC
Stage Controller
Workstation
GPU
USB
PCIe
PCIe
Real-time image processing, motor control, laser control, temperature control software
Figure 4. Data Processing pipeline for Freely Moving C. elegans whole brain Ca2+ imaging.
Signal Analysis
Difference of Gaussian Filtering
Multi-Channel Least Square Alignment
Raw Data
XYZT Annotation GUI
Gaussian Fitted Signal Extraction
Neuron ID determination
Posture Tracking CNN
Neuronal Segmentation CNN
Ratiometric Ca2+ dynamics measurement
Figure 7. Figure from Park et al. 2022: Posture space of the worm’s behavior required for targeted augmentation of neural annotations (Park et al. 2022)
Figure 5. Figure from Park et al. 2022: Architecture of the neuron tracking Convolutional Neural Network. (Park et al. 2022) Beige cubes are preceded by 3D convolutions while blue cubes are preceded by Atrous Spatial Pooling Pyramid layers [3]
Figure 2. Figure from Kimata et al 2012
Figure 6. Annotation and neural network assisted tracking software. (Park et al. 2022)
Figure 8. Correlation between the brain wide neural activity between all time slices. The transition of behavioral state is visible and displays complex pattern.
Figure 7. Neural activity of freely behaving C. elegans under a oscillating temperature signal. The calcium activity of 24 neurons in the C. elegans brain are shown. Neurons 6 shows clear temperature rise dependent increase of intracellular calcium, and is thus determined to be AFD. Neuron 7 is shown to correlate with backward motion (data not shown).
Figure 1. The transgenic C. elegans strain ZM9624. This animal expresses GCaMP6s and mNeptune in all its neurons under the panneuronal rgef-1 promoter. (Susoy et al. 2021)
Image Credit: Vladislav Susoy
Temperature homeostasis is one of the most important mechanisms required for a survival of an organism. Small animals like nematodes cannot afford to be endothermic, simply due to their surface to volume ratio. They must thus navigate to their preferred temperature.
While many animals have a set prefered temperature, the nematode C. elegans has an interesting behavior where they navigate towards the temperature they have been in the last 3~4 hours [1], as seen in Figure 2. A.
Neuron ablation studies, calcium recordings of neurons and intuitions from the connectome has established a model of flow of information from sensory stimuli to behavior[2], as seen in Figure 2. B.
However, the exact mechanism of this circuit in animals performing thermotaxis remains to be determined.
The microscope used to record Ca2+ activity of neurons in vivo. The microscope simultaneously acquires from two channels. The 488 nm laser activated the calcium dependent fluorophore GCaMP6s while the 561 nm laser activates mNeptune used to reject motion artifacts. The trigger signal from the camera is fed into a piezomotor controlling the focal plane of the microscope objective.
The PID controlled peltier device attached to a ethylene glycol cooler enables control of the temperature experienced by the worm. The motorized stage enables tracking of the freely behaving worm.
All devices are connected to a workstation with a multiprocessed software providing simultaneous control and recording.
The data from the microscope is processed by the pipeline shown in Figure 4. Among these steps, tracking individual neurons in 3D recordings of freely moving worms is the biggest challenge. The animal moves, rotates, deforms, shrinks while motion blur and signal noise affects image quality.
We have developed a custom software enabling fast annotation and tracking of individual neurons with minimal error. Our algorithm consists of a specially designed convolutional neural network (CNN) (Figure 5.) and a trick we call targeted augmentation. We embed each frame in a posture space (Figure 7.) and deform each ground truth annotation into a new unexplored posture. The new postures are chosen by selecting the furthest posture from the existing set of postures.
With the setup described in this work, we have successfully obtained 10+ annotated recordings with multiple neurons. As we see in Figure 7, whole brain activity is complex and spans different time scales. The sensory stimulus, temperature, here is only well correlated with the AFD neuron. As we can see in Figure 8, the high dimensional neural activity can be correlated between different times to represent the neural state similarity between behavioral states. How the temperature signal is processed in interneurons to orchestrate thermotactic behavior remains to be elucidated with this data.
[1] Russell & Hedgecock 1975
[2] Ryu&Samuel 2002,Clark 2006, Luo 2014
[3] Chen 2012