It would be difficult to guess as to what a universally best method would look like - do you know the nature/properties of your target protein and what is its relative proportion in the the supernatant?

What can said is that you have to grow the bacteria in protein-free medium. Otherwise your life is going to be amazingly difficult, as your sample will be saturated in broth proteins.

what is required here is a method that concentrates the protein which usually is very dilute in the broth. a good approach is to directly bind the protein to an ion exchange material. all you need to know is the pI of your protein. use a buffer which is at least 1 pH unit above or below the proteins pI and bind it to the column material (for cation or anion exchange as appropriate) in batch fashion. shake or stir (larger volumes) for 1 h  and the transfer to a column. elute with NaCl gradient as usual. 

alternatively, precipitation (with ammonium sulphate or organic solvents) is also advisable. however, the method above actually constitutes the first purification step, precipitation is in most cases does not enrich the protein. 

If concentration of the protein in the medium is low, (NH4)2SO4 will not be a good idea.

This why cleavable tagging is commonly used. For example, one can direct the protein  to the periplasmic space. This will facilitate recovery in a small volume without much cytosolic contamination when bacteria are harvested and the cell wall is digested with N-acetyl muramidase. Alternatively, the secretory signal can also include features like His tags so the entire material can be passed through a Ni-resin column to retain the protein in one step.

you can do the following:

Salting out

Dialysis

Gel filteration

Thank you all for your advise. Unfortunately, I don't yet know the nature of the protein. I'll keep you up to date :)