From the code:

# This parser will read Direct Xrefs from a simple tab-delimited file.
# The columns of the file should be the following:
#
# 1)    Accession ID
# 2)    Ensembl ID
# 3)    Type (one of 'transcript', 'gene', or 'translation',
#       will default to 'gene')
# 4)    Display label (will default to the Accession ID in column 1)
# 5)    Description (will default to the empty string)
# 6)    Version (will default to '1')
#
# Columns 1 and 2 are obligatory.