See also this thread: Getting Coding Strand (Cds) Using Uniprot Ids
And in particular my comment (from the UniProt helpdesk):
"I would like to point out that in the vast majority of cases, there is no single nucleic acid reference sequence for a given UniProtKB/Swiss-Prot protein sequence.
The canonical protein sequence is the outcome of thorough curation work, which often involves the merge of various sequences encoded by the same gene (in one species). In the annotation process, the most correct amino acid sequences are chosen and discrepancies are analyzed and documented.
cf this FAQ for more details: http://www.uniprot.org/faq/35 "