I think you could  pellet the 800 kDa protein at 100,000 x g in a short time, leaving the 20 kDa protein in the supernatant.

You could use a sucrose gradient. The 800 kDa protein would travel down the gradient, and the 20 kDa protein would stay at the top.

Gel filtration chromatography could also be used. The size differential is so big that the two proteins could easily be separated.

Hi Adam,

Thank you for your reply. Do you think without sucrose or glycerol gradient, 100,000 g ultracentrifugation would sufficiently separate 800 and 20 kDa proteins?

SEC/GPC is a good option. unfortunately, in my specific case, i found the two molecules somehow interact each other because the I can detect the presence of small protein (20kDa) from the early peak (small retention time). 

Hi Andrei, I tried ultrafiltration, but the separation is not good. I tried many columns from different venders. Ammonium sulfate precipitation sounds good to me. thank you.

If the proteins interact with each other tightly enough to come through the gel filtration column together, separating them will be difficult without using chaotropic conditions such as high salt or low pH. If that works, then any of the above-mentioned techniques may separate them.

Of course, if the two proteins are tightly bound together, it may be because they function as a complex and should stay together.

Another possibility is that the 20 kDa protein is not bound to the 800 kDa protein but is just aggregated into a large particle that elutes at the same place as the large protein.

Hi Adam,

thank you for your suggestion. it is helpful.

I thought about playing about pH and salt concentration a while ago, but didn't have a clue how to optimize/test two variables simultaneously. any tips? 

the aggregation is a good assumption. I will test it out.

BTW, I am doing sucrose density gradient ultra centrifugation now. do you think the high osmosis would somehow attenuate the interaction?  

To keep things simple, I would optimize each variable independently. Sucrose has high osmolarity but is non-ionic. I don't think it will disrupt the binding interaction, unless that interaction is based on recognition of carbohydrate. Make sure you use the highest purity of sucrose, since the concentration is high enough that minor contaminants could cause problems.

I have the reverse question: what is the reference for the 20 kDa protein staying in the supernatant after 2h at 100,000 RCF?

That depends on how far it has to travel, its shape, and the viscosity of the medium.