You could use a negative lookbehind assertion for the beginning of line anchor.
That matches any MU that is not preceded by the beginning of a line. That essentially does the same thing as racklet's !/^MU/ but lets you put other things into the regex if you need to (it negates only the position, not the entire regex). If you don't need anything else in the regex, I'd suggest his method.
The [\s|\w]+MU (the + or any quantifier is really redundant, since all you're looking for is a single character before it to ensure it's not at the beginning) wouldn't catch things like "$MU" because $ isn't whitespace or a word character. More so, if you're using multi-line input, where there could be a newline before MU, the \s contains a newline (hey, it's whitespace) and would match when you don't want it to. Using the . is better (it, by default, doesn't match newlines, just everything else), but I think the cleanest way is to use a short, simple regex and negate it.
Also, you don't need the OR operator | inside a [character class]. In fact, it actually inserts it as a character, adding it to the defined class of words and spaces, and now a pipe, which is probably not what was intended.
Ugh, it's 2am. I think I get cranky in the wee hours.
----------------------------------------------------------------------------------
...but I'm just a C man trying to see the light