On So's conjecture for integral circulant graphs