You should normalize to the non phosphorylated protein. The best option should be to strip the membrane and use the same membrane for the phosphorylated and the non phosphorylated protein.

I prefer not to strip

Alternative 1: via the IR-linked secondary antibodies you can work with two colors and visualize the phospho-porteins X and all protein X at the same time

Altenative 2 : load two seperate gels. 1 for the phospho-protein X + a standard loading control and a second gel for all protien X + a standard loading control

I would also suggest to work with one blot- regardless whether you strip or work with two colours (in case you have the required equipment). If you use two separate blots there is always the possibility of technical variations between the two, which could generate missleading data.

Thank you for our answers.

We have small human samples that's why I could load only one or two gel but we a some protein of interest a the same molecular weight.

We will use IR antibodies.

You can normalise to total protein per lane by coomassie staining your blots after youve probed with antibodies, or if you have reliable Ponceau S staining after transfer( this works better with nitrocellulose and semi dry transfer, for PVDF/wet transfer try coomassie after all antibody detection) . This means you can control for uneven transfer and we have found it more reliable than tubulin/actin etc, especially in muscle  where these can vary wildly between samples. The coomassie stain protocol we use is from this paper. http://pubs.acs.org/doi/abs/10.1021/pr1011476.

Good luck!

Dear Kate,

Staining protocols for normalization are indeed fine for normal proteins.

However, for a post-translational modification of a particular protein, you should preferably also take that particular protein in the equation.