This study explored the use of an innovative instructional approach called Productive Failure (PF) to design an educational game and its support. The study then examined the effects of two different types of instruction--PF vs. Direct Instruction (DI)--on learning genetics and relevant mathematical knowledge in a Game-Based Learning (GBL) environment. One hundred fifty-seven Year 10 students from two high schools participated in two quasi-experimental studies. The participants were divided into two treatment groups: one group learned targeted concepts using PF with GBL (PF-GBL), while the other group learned the same concepts using DI with GBL (DI-GBL). The results of the first study indicated that the PF-GBL group showed significantly higher learning gains than the DI-GBL group on explanatory genetics knowledge. In the second study, no group difference was found between the PF-GBL group and the DI-GBL group on learning genetics and relevant mathematical knowledge. Implications of findings, limitations, and future research are discussed.