Taxol is a "blockbuster" antitumor drug produced by Taxus species with extremely low amount, while its analogue 7-β-xylosyl-10-deacetyltaxol is generally much higher in the plants. Both the fungal enzymes LXYL-P1-1 and LXYL-P1-2 can convert 7-β-xylosyl-10-deacetyltaxol into 10-deacetyltaxol for Taxol semi-synthesis. Of them, LXYL-P1-2 is twice more active than LXYL-P1-1, but there are only 11 significantly different amino acids in terms of the polarity and acidic-basic properties between them. In this study, single and multiple site-directed mutations at the 11 sites from LXYL-P1-1 to LXYL-P1-2 were performed to define the amino acids with upward bias in activities and to acquire variants with improved catalytic properties. Among all the 17 mutants, E12 (A72T/V91S) was the most active and even displayed 2.8- and 3-fold higher than LXYL-P1-2 on β-xylosidase and β-glucosidase activities. The possible mechanism for such improvement was proposed by homology modeling and molecular docking between E12 and 7-β-xylosyl-10-deacetyltaxol. The recombinant yeast GS115-P1E12-7 was constructed by introducing variant E12, the molecular chaperone gene pdi and the bacterial hemoglobin gene vhb. This engineered yeast rendered 4 times higher biomass enzyme activity than GS115-3.5K-P1-2 that had been used for demo-scale fermentation. Thus, GS115-P1E12-7 becomes a promising candidate to replace GS115-3.5K-P1-2 for industrial purpose.