We are researching and examining genes associated with ALS (Amyotrophic Lateral Sclerosis) to study how motor neurons and behavior are affected in fruit flies (Drosophila melanogaster). ALS is a progressive neurodegenerative disease that breaks down nerve cells leading to the loss of motor function, yet the specific genetic and cellular mechanisms behind it are still not fully understood. By using Drosophila as a model organism, we can express ALS-related genes, such as GGGGCC repeat expansions and GR.PO proteins in motor neurons and observe their effects on behavior. This helps us explore the relationship between gene expression and motor function, potentially identifying early behavioral changes that reflect neurodegeneration. Our research aims to fill the gap in understanding how these genes impact movement and could contribute to disease progression.
The illustration depicts the predicted genotypic outcomes of our genetic crosses and highlights the selection process required to isolate offspring carrying the desired transgenes. For the UAS line GR.PO (UAS #1), 100% of the progeny are expected to carry the UAS transgene, allowing us to retain all offspring for further analysis. In contrast, for UAS line GC49x (UAS #2), only 50% of the progeny are expected to carry the desired UAS gene, while the remaining 50% carry the TM6SB balancer chromosome, identifiable by the stubble phenotype. Similarly, in the control cross using line GC8x, only 50% of the offspring are suitable for retention, as the remaining individuals also carry the TM6SB balancer. Flies expressing the TM6SB balancer are excluded from analysis, as their presence indicates the absence of the desired transgene, which is confirmed by our Punnett square predictions.
Set up parent crosses: We collected virgin females or males from D42-gal4 and collected the opposite sex (virgin females or males) from one of our UAS genes (typically 2-3 of each per cross) and placed them in a fresh vial. Each vial is labeled with the genotypes used, the number of flies of each sex, and the date of setup
Flip out the parent flies: After 8 days, we flip the parent flies into a new vial. This prevents mixing between the parent generation (F0) and the offspring from Flip 1 (F1) and ensures that the progeny can be tracked by generation.
Collect F1 offspring: Once the F1 flies have eclosed, we anesthetize them with CO₂, check if they are stubble for our UAS genes GC repeat, and sort them by sex under a microscope. Males and females are separated into labeled vials for further analysis and to prevent unintended breeding.
Check for stubble in the offspring (shorter, bristly dorsal hairs on the back characterize Stubble). Only flies with the TM6, Sb balancer should show the stubble phenotype. Both UAS #2: GC49x(Toxic) x D42-gal4, Control: GC8x x D42-gal4 offspring have stubble we need to check. We want the non-stubble offspring to record behaviors.
After the offspring are collected, the behavior is recorded every 7 days from the collection. To track the behavior, the flies are flipped into empty vials and placed on ice to sedate them. They are then placed onto a 3D-printed “dome,” and a petri dish is set on top to prevent the flies from flying away and from moving onto the lid.
The flies are left on the 3D-printed “dome” for 2 minutes to regain consciousness. The fly's behavior is then recorded for around 4 minutes. Once the recordings are finished, they are anesthetized with CO2 and placed back into a vial. This process is then repeated for all of the genes available and for each milestone needed to collect, i.e., Day 1, 7, 14, 28, 35, etc. (Flies are collected every 7 days for as long as their lifespan.)
Import: Load the ~4 minute video to HPC using the terminal app on your laptop
Launch SLEAP Interface: Open SLEAP on our HPC or local machine. Then, open the mate terminal and type in the necessary commands to launch the AI interface
Open Project: Click on File at the top Right Hand corner, click on Open New Project, and upload a copy of Dr. B's prediction files into the project. Then, proceed to upload your videos to the same project.
Label Flies: Generate 20 random frames using labeling suggestions. Manually label each fly in a video. For every frame we label, we place 11 key points per fly (head, thorax, abdomen, forelegs, etc.). If a specific body part is not recognizable, we gray out the respective label to avoid errors in our training model.
Run Training: Click predict and start training on 20 random frames in current video. Stop early if necessary when training the centroid instance. Check to see if your model has precision score of 0.95 or above. If not relabel, and run training again until you reach the 0.95 threshold.
Run Inference: Use the trained model to predict and label your videos in your project
Export: After running inference for a video, export the HDF5 file
Repeat: Add videos to the original project you created, run inference, and export the HDF5 file
Make a copy of the template code used to analyze thorax velocities from the Spring Outline Google Document
Run the first line of code and connect to your Google Drive
Create a folder in your Google Drive to store your groups' HDF5 files
In Google CoLab, following the instructions in the copied template code, copy your.h5 file's path from your drive and input it into the template
Click on Runtime, and select Run all.
Modify and/or delete any code you won't need for data analysis
The illustrated tables summarize data from our genetic crosses. We analyzed three primary genotype lines: UAS-GR.PO × D42-Gal4, D42-Gal4 × UAS-GC49x, and UAS-GC8x × D42-Gal4. Each table includes the genotype labeling, the date each cross was established, the number of progeny collected, and the recorded death dates of individual offspring.
During the collection process, we observed a notable result in the D42-Gal4 × UAS-GC49x cross. All resulting progeny exhibited the stubble phenotype, indicating that the GC49x transgene was not being expressed. Due to this lack of expression, we were unable to collect viable offspring from this cross for further analysis and deemed that the genotype is toxic to express. We also noticed on a weekly basis while recording behaviors for flies expressing UAS-GR.PO, that their wings became more wrinkled and shriveled up as they aged.
Based on our limited data on the thorax tracks of GR.PO flies and GC8x files, we can logically infer that the affected flies ability to move across space diminishes at a faster rate on weekly intervals compared to the control flies. However, more data on thorax tracks and thorax tracks by magnitude of fly speed should be collected to establishing causation between the expression of GR.PO and fly locomotion.
Our data on the relationship between the mean thorax velocity and age of flies is unpredictable and variable for flies expressing the ALS-related gene and control gene. The variability in our data may indicate the presence of confounding variables and lack of consistency in the way our group executed procedures for collecting fly behavior and locomotion data.
It is well-established that ALS affects men more than women, men experiencing a shorter lifespan after diagnosis. In the ALS test flies, sex differences in lifespan are also observed showing male flies have a shorter survival rate.
Moving forward, our primary objective is to strengthen the validity of our conclusions through continued analysis of the collected data. We aim to refine our datasets by utilizing SLEAP and Google Colab to track and analyze behavioral patterns associated with each genotype. This will enable us to make more precise comparisons of genotype-specific behaviors and gene expression profiles.
Limitations encountered in this study include:
Tracking accuracy with SLEAP: Challenges with pose estimation and model training affected consistency in behavioral tracking.
Technical issues with SLEAP software: Problems related to user permissions, model training, and system compatibility hindered progress.
Time constraints: Limited time for data collection, processing, and analysis impacted the scope of the study. Forced prioritization of certain tasks led to data being under-explored.
Unverified gene expression in progeny: Some offspring did not exhibit expected phenotypes, likely due to incorrect stock usage or misidentification of parent-offspring relationships.