Investigating Carboxysome Morphology Dynamics with a Rotationally Invariant Variational Autoencoder

Area of Science:

• Microbiology
• Biophysics
• Systems Biology

Background:

• Carboxysomes are essential bacterial microcompartments crucial for CO2 fixation in cyanobacteria.
• Understanding carboxysome assembly and dynamics is vital for photosynthetic metabolism.
• Current microscopy-based analysis of carboxysome morphology is subjective and low-throughput.

Purpose of the Study:

• To develop and apply deep learning techniques for quantitative analysis of carboxysome dynamics.
• To evaluate the impact of carboxysome shell remodelling on microcompartment morphology over time.
• To accelerate the study of carboxysome assembly and dynamics in response to perturbations.

Main Methods:

• Utilized a Rotationally Invariant Variational Autoencoder (rVAE) for image analysis.
• Employed a system to control endogenous protein levels in *Synechococcus elongatus* PCC 7942.
• Analyzed fluorescence microscopy images of cyanobacteria with a carboxysome reporter.

Main Results:

• rVAEs quantitatively assessed changes in carboxysome number, shape, and size.
• Demonstrated the ability to tune carboxysomal protein levels in real-time.
• Successfully correlated shell remodelling with morphological trends.

Conclusions:

• rVAEs offer a powerful, high-throughput tool for analyzing carboxysome dynamics.
• This approach can accelerate research into carboxysome assembly and response to stimuli.
• The methodology may be applicable to other bacterial microcompartments and regulatory processes.