Hint: See if you can compute the covariance of X1 and X2.
I think you are supposed to assume that Z1 and Z2 have a joint bivariate normal distribution. Under that assumption, 2Z1-Z2 and Z1+Z2 have a joint bivariate normal distribution. I think you can compute the means, standard deviations, and correlation of 2Z1-Z2 and Z1+Z2 from the information given. Then you should be able to find the MGF from a known formula for the MGF of a bivariate normal.