Recently Ercolani and McLaughlin proved that the zeros of the biorthogonal polynomials with the weight function w(x, y) = exp(−V(x) − W(y) − 2τxy) are all real and distinct, and Mehta has extended their argument to the weight function w(x, y) = e−x−y/(x + y) and to the more general case of the convolution (w1 * w2 * ... * wm)(x, y), where wi are functions of the same form as above. Using the concept of total positive and sign-regular functions, we further extend the argument to a large class of weight functions. Many examples are presented, including several whose pair of biorthogonal polynomials turn out to come from different families of classical orthogonal polynomials.