In this paper we investigate the degrees of freedom of layered two unicast Gaussian networks with destination-to-source feedback. The feedback model allows the destinations to feedback their received signals to their respective sources perfectly. Based upon the non-feedback degrees-of-freedom result by Shomorony and Avestimehr [1], we show that in general, feedback enlarges the degrees of freedom region for two unicast layered Gaussian networks. This is the first case in literature where feedback can provide a degrees-of-freedom gain in multi-user interference networks.