[P004] Poster

1:00 PM - 1:50 PM JST(4:00 AM - 4:50 AM UTC)

[P02-216] Protein Language Model-Guided Discovery and Secretory Production of Novel GH28 Polygalacturonases with Superior Catalytic Performance

○Min Yoo¹, Seong Jin Jeong¹, Hyo Deok Seo¹, Young Kyoung Rhee¹, Jae Woong Choi¹ (1. Korea Food Research Institute (Korea))

Keywords:

polygalacturonase,protein language model,GH28,enzyme discovery,Bacillus subtilis,secretory expression,catalytic efficiency

Polygalacturonases (PGases) of glycoside hydrolase family 28 (GH28) are essential in food processing for juice clarification and viscosity reduction. However, conventional screening is labor-intensive. Protein language models (PLMs) offer a transformative alternative by predicting functional properties from sequence space without structural data. In this study, we constructed a GH28 database of 8,833 unique sequences from NCBI and UniProt. Using sequence similarity network analysis (90% identity) and Markov Cluster Algorithm (MCL), we identified 1,200 representative sequences. We employed CatPred, an ESM-based tool, for in silico screening of catalytic efficiency (k_cat/K_m) and solubility. Sixteen candidates were expressed in E. coli BL21(DE3) and purified via Ni-NTA chromatography. Screening identified FLCH_PG (Flavobacterium cheongpyeongense) and PAMA_PG (Paenibacillus massiliensis) as the most active, showing 5-6-fold higher activity than controls (pga1, pgaB, pgaII). Four enzymes (FLCH_PG, PAMA_PG, FLAG_PG, FLHY_PG) were selected for detailed characterization. Optimal temperatures ranged from 60°C to 70°C, and pH from 5.0 to 8.0. Notably, FLCH_PG showed remarkable pH tolerance (pH 5.0-9.0), while PAMA_PG demonstrated exceptional thermostability, retaining ~100% activity after 2 h at 50-70°C. Kinetic analysis showed FLCH_PG had the highest efficiency (k_cat/K_m = 7.73 s^-1・(mg/mL)^-1; V_max = 35.19 U/mg), while PAMA_PG had the highest substrate affinity (K_m= 1.41 mg/mL). Metal profiling suggested no metal cofactor requirement, as K⁺ and EDTA enhanced activity while Cu²⁺ and Ca²⁺inhibited it. Structural analysis via BioEmu revealed more rigid active-site architectures in these novel enzymes, correlating with enhanced performance. To establish a food-grade platform, enzymes were expressed in Bacillus subtilis using SecretoGen-optimized signal peptides. FLAG_PG with SP3 achieved the highest extracellular activity (0.040 U/mL), α1.5-fold increase over cytoplasmic control. This study demonstrates that integrating PLM-guided screening with GRAS-host secretory expression provides an efficient pipeline for accelerating enzyme development in the food industry.