On insulin action in vivo: the single gateway hypothesis.