Did you dialyze the protein after the Ni column to reduce the salt concentration? the ionic strength of 500 mM NaCl and 200 mM imidazole would prevent almost any protein from binding to an ion exchange column.

I would I try SEC in high salt buffer to see if you can get rid of the DNA.

Hi Zhenning, in addition to Adam's suggestion of dyalisis, did you run fractions on agarose gels to see if there was DNA present in your protein? 

I have done quite a bit optimization with salt condition in loading. My protein elutes from HiTrap SP column at 700 mM NaCl. That's why I tried to increased salt up to 500 mM NaCl to disrupt the interaction between DNA and protein. I also tried lower salt, 150, 200, 250 mM NaCl. They don't have any effect on improving binding. 

What I don't understand is that 1) at such high salt concentration, there are still high A260/A280 ratio in my eluted protein fraction, 2) Most of my protein is in flow through.

There seems to be a tradeoff for the salt concentration. My goal of this IEX is to remove nucleic acid. 

This is a nucleic acid binding protein and we expect this protein to bind DNA in expression and that's demonstrated by the high A260/A280 ratio of fraction from His tag column. 

I load my sample at 1M NaCl, 20 mM Sodium Phosphate, 60 mM Imidazole for Ni column. I don't understand why the protein fractions have high A260/A280 ratio.

What is the salt concentration of the sample loaded onto the SP column?

Like Adam said, high salt inhibits protein binding to both cation and ion exchange resins.

Protein loading buffer has 500 mM NaCl (250, 150 mM in other trials). SP buffer A has 250 mM NaCl, buffer B has 1.5 M NaCl. The small portion of bound protein elutes at 700 mM NaCl

If I understand correctly, the evidence you have that the protein has DNA bound is the high A260/A280. This could also be caused by the protein being aggregated, which causes turbidity, which is greater at shorter wavelengths. As Steingrimur suggested, try running an agarose gel to check for DNA directly. Also try doing dynamic light scattering to check for aggregation.

You could also try treating your protein with DNase to digest the DNA, then put the protein over a desalting column to remove the resulting nucleotides. Of course, this adds a contaminating DNase, which is going to be a problem later, but it can either be removed by chromatography or inhibited by EDTA.

The presence of DNA bound to your protein might change the charge of this protein complex. That might explain why only small fraction of the protein binds to the resin. I agree with Adam that the best way is to use DNase to hydrolyze the DNA bound to the protein to get your protein back to its natural isoelectric point, then remove the nucleotides and the salts by desalting (gel-filtration or dialysis using any buffer with 50-100 mM NaCl and pH 6.0 or 7.0), and then load your protein to the column equilibrated in the same buffer. The buffer of your column is equilibrated with should be the same as the buffer of your sample loaded (both salt concentration and pH). If you use the pH ~ 6.0 as the equilibration buffer and your sample buffer (and your protein Pi ~ 10.0) you will get better binding of your protein (at the low salt concentrations like 50 mM - 100 mM NaCl).