If you don't want to muck around directly with RColorBrewer
(a lovely package), you can reverse the levels of the factor in the original data.frame, and then plot it:
dsamp <- diamonds[sample(nrow(diamonds), 1000), ]
# Reverse the levels of the factor associated with color, here 'clarity'
# (Might be safer to assign this to a new column named, e.g., 'clarity2')
levels(dsamp$clarity) <- rev(levels(dsamp$clarity))
d <- qplot(carat, price, data = dsamp, colour = clarity)
d + scale_colour_brewer(breaks = levels(dsamp$clarity))
And if you want to print the key in the same order as before the reversal, just do this:
d + scale_colour_brewer(breaks = rev(levels(dsamp$clarity)))