Good question, the key distinction is deciding what you want to see. Denaturing a protein does two things: it (usually irreversibly) destroys the activity of a protein, and it turns the protein from a structured mess of squiggles into an unstructured mess of squiggles:

But fortunately, those squiggles are so small, that you can't distinguish the two by eye anyways. A normal brain slice looks something like this:

Those dark areas? They're dark because they are denser in fat. Zoom in as much as you want, and you're still mainly seeing the fat scattering light. This method (well, the one its derived from)? It clears that fat, leaving you with something much more transparent:

Want to see neurons? You'll have to selectively brighten them somehow. Here's a slice with neurons expressing a fluorescent protein. Put them under the right laser light, and they'll show up brighter than the rest of the tissue:

Brainbow? Just a combination of added fluorescent proteins:

In both of those two examples, denaturing the proteins will destroy their fluorescence and make it again impossible to see anything.

The workaround is to add fluorescence by another method: tag a dye to an antibody that binds a small section of your (denatured, separated) protein. This is pretty common as it's a million times easier to find a good antibody as it is to genetically modify an organism. Inject a rabbit / mouse / rat / goat with a section of your protein, extract some of its blood, and filter that for antibodies and you'll usually have something workable. Plus there's already large catalogs of these antibodies already for the major proteins of interest (neuron, synapse, axon markers, etc) and 80% of them still worked after applying this method to their samples.

So now you have a 4-5-fold increase in resolution when you go to image the samples, and there's no pesky physical limit to using some other gel that expands to even larger after adding water.