The trouble you will have with removing the non-CSS parts from that string is that it's very hard to determine which parts are CSS and which aren't.
You say you want to be left with aaaa bbbbb cccccc ddddd eeeeee, but in fact from your original string, aaaa bbbbb cccccc would be valid parts of the CSS selector. They would select elements named <aaaa> or <bbbbb> or <cccccc>.
Granted, these are not valid HTML elements, but CSS can be used to apply styles to arbitrary XML as well as HTML, so they could very easily be valid elements. If you're using xhtml, they could appear in your page quite legitimately under a custom name-space.
But it gets worse. I assume that the text wouldn't actually be aaaa bbbbb cccccc, but would be an arbitrary string of words. In that case, consider that it may be a sentence like 'I am strong'. In this case, strong would be part of the string you want to remove, but <strong> is also a valid HTML element (as is <I> for that matter), so even if you are just sticking with it would be impossible to tell in the above string whether it was intended to be a CSS selector or part of the text string to keep. You simply couldn't write a regex that would be completely reliable in all cases.
As you say, you can remove everything inside the {} braces fairly easily, but the selectors outside the braces would be very hard to separate reliably from surrounding text.
