I may be way off base here, but it seems to me that there is no way to stop the columns from seeing a cyclic load, and therefore no way to stop them from fatigue. With no preload, the load goes from 100,000 N to -100, 000 N. Suppose you preload them with a 100,000 N load. The load will go 200, 000 N to 0. You have shifted the mean stress, but the alternating stress hasn't changed. I believe you would get the best fatigue life with no preload, since the stress can go as high as the fatigue strength of the material.