Elman neural network training algorithm

If I use the train function to train an Elman neural network with the trainFcn parameter equal to "traingd", what kind of training algorithm will be performed: the static backpropagation (same used in multilayer perceptron) or a backpropagation through time?