How can I embed a Dense layer inside of an RNN cell in tensorflow/keras?

Question

Inside of a custom RNN cell, is it possible to add a Dense layer so that each LSTM cell at each timestep looks like something like this?

(input, prev_state) -> L