Lactose is produced in large amounts as a by-product from the dairy industry. This inexpensive disaccharide can be converted to more useful value-added products such as galacto-oligosaccharides (GOSs) by transgalactosylation reactions with retaining beta-galactosidases (BGALs) being normally used for this purpose. Hydrolysis is always competing with the transglycosylation reaction, and hence, the yields of GOSs can be too low for industrial use. We have reported that a beta-glucosidase from Halothermothrix orenii (HoBGLA) shows promising characteristics for lactose conversion and GOS synthesis. Here, we engineered HoBGLA to investigate the possibility to further improve lactose conversion and GOS production. Five variants that targeted the glycone (-1) and aglycone (+1) subsites (N222F, N294T, F417S, F417Y, and Y296F) were designed and expressed. All variants show significantly impaired catalytic activity with cellobiose and lactose as substrates. Particularly, F417S is hydrolytically crippled with cellobiose as substrate with a 1000-fold decrease in apparent k(cat), but to a lesser extent affected when catalyzing hydrolysis of lactose (47-fold lower k(cat)). This large selective effect on cellobiose hydrolysis is manifested as a change in substrate selectivity from cellobiose to lactose. The least affected variant is F417Y, which retains the capacity to hydrolyze both cellobiose and lactose with the same relative substrate selectivity as the wild type, but with similar to 10-fold lower turnover numbers. Thin-layer chromatography results show that this effect is accompanied by synthesis of a particular GOS product in higher yields by Y296F and F417S compared with the other variants, whereas the variant F417Y produces a higher yield of total GOSs.