Why is the presence of autoantibodies against GAD associated with a relatively slow progression to clinical diabetes?