We usally use 1*10^6 cells for WB. Depending on the cell line and how much you load on a gel it typically is enough for 3-4 Blots. However, when I had only a small amount of cells (e.g. after transfection) 200.000 cells were enough for one WB. In this case I resuspenden the cells directly in 25µl of loading buffer, boiled them and sonificated them for 1 min.
I have been using 20 000 cells (overexpressing the protein) for one lane, the cells were lysed (boiled) directly in SDS loading buffer and loaded on a 10% gel without sonication. I would recommend optimization because the signal depends on the expression level of the protein, on the affinity of the antibody, transfer efficiency and so on...
I have approximately 1-1.5 x10^6 cells. But in response to Maren and Elzbieta, lysing the cells in loading buffer is a great idea that I never came across, but I will most certainly try. Just wanted to know whether you get allot of background with this approach?
Your "background" will also be determined by specific conditions confined to your experiment/choice of reagents, i.e. choice of blocking buffer, methodology (IR vs. chemi), transfer efficency etc.
If you have a moderately expressed protein and the ideal conditions surrounding western blot are met (good protein isolation, transfer etc), then your background should be minimal because your signal is relatively higher.