In addition to the excellent answer above, it is definitely worth trying to secrete the protein, particularly if the protein is normally secreted (i.e. produced by endoplasmic reticulum in eukaryotic cell), because often those proteins have disulfide bridges which stabilise the protein in a soluble conformation, but disulfide bridges can't be formed in the reducing environment of the E.coli cytosol. However, adding a bacterial signal peptide will certainly not increase expression levels, as there are limited translocation pores in the E.coli inner membrane. The amounts you will obtain from the periplasm are going to be smaller, but it could indeed lead to a soluble protein. However, there are specific protocols for the selective extraction of periplasm, which can be considered a purification step on its own, and together with using an affinity tag on the protein, you can certainly compensate poor expression with large culture volumes. But every protein is different, and you can only try.

Alternatively, if you just want to generate an antibody and you don't need biological activity, you can use the formation of inclusion bodies as a purification step. First extract soluble proteins and remove them from the pellet, then boil the pellet and run on preparative SDS PAGE and stain with coomassie blue. You will have lost a very large proportion of contaminants, and you will have only membrane proteins and your inclusion body. We have successfully cut out sharp coomassie blue stained bands (avoid overloading of the gel), ground them up with liquid nitrogen to obtain a fine powder, and used this for immunisations to get great antibodies. It just depends on what you want.

Best wishes, Jurgen 

Thanks  for Joke's and Jurgen's advices,

I have tried the E.coli strains BL21(DE3)  and induced at 37℃ or at 25℃,but the protein expressed as inclusion bodies. Then I used the ArcticExpress(DE3) and  induced at 12℃, unfortunately, the protein expressed as inclusion bodies also. 

I need the protein have biological activity,so solubility expression is important. currently, I am to reconstruct the expression vector and to fused the signal peptide with the protein, in orde to secrete expression. As Jurgen's saying, the poor expression can be  compensate with large culture volumes, but I am not sure this way can be work or not.

If you need biological activity, then you have to keep trying different expression systems and conditions. Also with a signal peptide, you should explore different temperatures, as well as minimal versus rich growth medium. It is trial and error. Start with the signal peptide and ext5raction from the periplasm, and if that fails, you can try insect cells perhaps. If all fails, you have to do it the hard way and extract the protein from its native context.