Principle (classical liberalism) – no one should work on command. Who is “one” – humans capable of mutual recognition, moral conflict, ethical life.
Animals are more free (with respect to each other) than a human incapable of mutual recognition. Freedom is not ability to carry out my contingent desires (Mill), but rather ability to recognize another’s ability to carry out his or her contingent desires – and modify one’s desires or actions according to the needs of living together.
It might be morally acceptable to enslave a human which has inner life but no social recognition because they would be incapable of differentiating between working out of their own will and working on command – because they cannot recognize an external stimulus as a command (as the will of another).
Whether or not an animal can be morally enslaved hinges on this point – can they recognize command as command?
Regardless of the technical aspect of enslaving, in all cases we have duty to treat sentient beings’ sentience as morally relevant. However, this does not mean we cannot act in such a way as any sentient being might suffer. Rather, we need to consider such suffering on a par with human suffering.
Even if a being has inner life and freewill, there is no duty to treat beings also as ends in themselves if they cannot recognize another as having free will. Treating another as an end in itself means traeting them as the kind of being that can have themselves (as a moral being) as the end of their actions – but a being incapable of mutual recongition is incapable of moral action, and thus can not have themselves as their own end.