Coomassie Blue forms strong but non-covalent complexes with proteins, most probably based on a combination of van der Waals forces and electrostatic interactions. Formation of the protein/dye complex stabilises the negatively charged anionic form of the dye producing the blue colour which may then be seen on the membrane or in the gel. The bound number of dye molecules is approx. proportional to the amount of protein present per band. However, binding of the Coomassie blue basic amino acids is much more efficient then to acidic amino acids. Glycosylation usually provides negative charge for protein, that is why highly glycosylated protein is difficult to be stained by Coomassie Blue.
I've not studied this in any great detail but based on my own observations highly glycosylated proteins such as carcinoembryonic antigen stain much more weakly than non-glycosylated proteins. See also Osset, M et al, Elecrophoresis,(1989) 10: 271-273.