Meaning is not something that is particularly an end goal. That is why there is specific differenciation between the syntax (structure) and its semantics. Once you isolate them, one could think of the structure itself and manipulate its related symbols, without worrying about their meanings. Sometimes surprisingly, some of these results map back to the meaning of the structure, but that is purely a byproduct.
The manipulation of symbols it itself meaningful, because the symbols are real things that exist in nature. You may be working at a different layer of indirection, but it makes no difference to the usefulness of your work, because nature itself does no indirection (as far as we can tell); your results are always applicable to something real. Even if many mathematicians are so poor at explaining it that they'll just p…
I also agree that the manipulation operates at an indirection but has meaning, I think the greatest example to that is problem reduction, that operates at a higher level of indirection but eventually maps back to a completely different meaning.
But it is often times that the manipulation of the symbols is simply inspired by meaning, in fact, you can define abitrary operations on any system (by doing S -> S in a relation) without any useful semantic relevence to anything. Maybe we are cherrypicking the meaningful ones? But maybe the fact that you can do so implies a different meaning. Well I guess we will never know. If we define symbols to be part of nature then surely "The opposite of nature is impossible."
Mathematics is effectively the study of things, therefore it has purpose, it can even study itself!