mediation does not require a specific direction. If you have not yet collected the data, rather invest some effort to strengthen the support for what mediation really aims to express (=a causal chain).
Hence, understand the assumptions inherent in that chain:
a) thinking about potential confounders (i.e., common causes) of the X-Y, M-Y, and X-M relationships (sounds complicated but will often involve the same candidates)
b) thinking about potential instruments for X and M