Range of valid character for a base 64 encoding

I am interested in the following:
Is there a list of characters that would never occur as part of a base 64 encoded string?
For example *. I am not sure if this would occur or not. If the original input actually had * as part of it would that be encoded differently?