Proc glm, glimmix, genmod and many other regression procs will generate dummy variables for all variables described as class variables. You should add interaction terms to your model if you have the least doubt that there might be some interaction (i.e. that the effect of a variable might not be the same in every category of another variable).