So, the antibodies recognize folded proteins in an ELISA, but don't work on Westerns? On a blot, the protein is typically in a denatured and unfolded state. If your antibody recognizes an epitope created by protein folding (rather than consecutive amino acids in the protein sequence), the antibody may not be able to bind the protein on a blot. The context is just too different. The same thing can happen with immunofluorescent staining and Western blots, or any other technique that changes the epitope context. This problem should be less severe with a polyclonal, but can still occur.

This could arise from a problem with the blocking buffer. Blocker choice can have a profound effect on antibody binding. Have you tried doing the blots with the same blocker you use for ELISA? That might help. Or you can test a variety of different blockers to see which ones are most appropriate for your antibodies. This is easiest to do if you use a mutiplex manifold on the blot, or run a prep gel (one big well) and then cut the blot into strips and use a different blocker for each strip. I can send you a few applications notes if you are interested.

The interesting fact is that for some antigens our recombinant poly-clonal antibodies work in very different assay, with folded and unfolded proteins. Some others just work with folded targets, and we know that there are different Abs in the population thanks to sequencing analysis. I would like to try your approach to address the "blocking buffer" issue, so if you can send me your notes I will appreciate. Thanks.

Our group published a study in 2008 in Proteomics that showed how blocker choice affects antibody specificity (http://www.ncbi.nlm.nih.gov/pubmed/18563731). I've also attached a couple of application notes that discuss blocking buffer optimization. Hope these are helpful!

http://www.ncbi.nlm.nih.gov/pubmed/18563731

Here are the application notes.