This is a nasty problem. There is no nice answer, i.e., nice form.
IF BOTH X and Y are normal or BOTH are exponential then you can use the MGF. BUT the MGF here won't be recognizable otherwise and that only works for the sum and difference too. So that's worthless. I would just do a 2-2 tranform and integrate out the dummy variable. BUT I wonder if the instructor wanted both of the same form.