Integrated statistical and machine-learning optimization for enhanced heparosan production by Lactococcus lactis

Area of Science:

• Biotechnology and Synthetic Biology
• Microbial Engineering
• Bioprocess Optimization

Background:

• Heparosan is a valuable glycosaminoglycan with therapeutic potential.
• Current production methods may rely on animal-derived components or have low yields.
• Engineering Lactococcus lactis offers a promising platform for animal-product-free heparosan production.

Purpose of the Study:

• To engineer Lactococcus lactis SH6 for enhanced heterologous heparosan production.
• To optimize the growth medium for maximum heparosan yield using a combined experimental design and machine learning approach.
• To establish a scalable and efficient process for producing animal-product-free heparosan.

Main Methods:

• Initial medium screening using one-factor-at-a-time experiments.
• Statistical optimization using Plackett-Burman and Central Composite Design (CCD).
• Machine learning (Gaussian Process Regression) for fine-tuning medium composition.
• Bioreactor studies with optimized medium, early nisin induction, and fed-batch strategies.

Main Results:

• Initial optimization identified glucose and yeast extract as key substrates.
• CCD predicted and validated a significant increase in heparosan production.
• Machine learning further refined the medium, leading to 85.28 mg/L heparosan at flask scale.
• Bioreactor scale-up with optimized conditions (early induction, glucose feeding) achieved a final titer of 133 mg/L.

Conclusions:

• The integrated design-of-experiments and machine learning approach effectively optimized the Lactococcus lactis medium for heparosan production.
• Heparosan yield was nearly doubled compared to unoptimized conditions, setting a new benchmark for L. lactis.
• This study presents a viable strategy for high-yield, animal-product-free heparosan production using a Generally Regarded as Safe (GRAS) microbe.