Problem with endogenously biotinylated proteins

[ranen aviner]

I am using a biotinylated probe for in-vivo biotinylation of target proteins
in HeLa cells. My initial experiments consist of pulldown of biotinylated
targets using Pierce's streptavidin agarose resin followed by boiling in SB
and SDS WB using Vector's ABC kit (HRP-based streptavidin detection). The
ultimate goal is to identify the proteins in the pulldown using MS.

The problem is this: when I load 5% of my lysate (treated vs. untreated HeLa
cells lysed in 0.5% NP-40), I see a definite enhancement of biotinylated
proteins in the treated sample (over a faint background of endogenously
biotinylated proteins). However, the pulldown results of both samples seem
identical; there is an extreme enrichment of biotinylated proteins of all
MWs and I do not detect the same pattern of proteins I see in the 5% total
treated lane.

So, my questions are as follows:
- Why might I get such different results between the total and pulldown?
- Does anybody know of a way to reduce this background signal? Perhaps a
cell line with less endogenously biotinylated proteins?