Purification of caveolae in gradients prepared from OptiPrep™
Early methods for the purification of lipid-rich plasma membrane domains largely relied on their insolubility in Triton X-110 (or some other non-ionic detergent) relative to that of the bulk plasma membrane or that of all the other subcellular membranes.
Sometimes detergent was added to the whole homogenate or more frequently a partially-purified plasma membrane fraction was first isolated before treating with detergent. Smart et al  however pointed out that, while use of a non-ionic detergent did permit the isolation of a lipid-rich membrane domain, that some characteristic caveolar proteins can be lost in the procedure. These workers therefore developed a method that avoids the use of Triton X-100.
After isolation of a plasma membrane fraction from either human skin fibroblasts or MA104 cells, the caveolae are released by sonication in a standard cell homogenization medium. The first part of the isolation procedure is a flotation through a continuous iodixanol gradient (0-20%); this gradient is essentially a resolving gradient in which the caveolinrich vesicles are concentrated in the top third of the gradient, while the predominantly caveolin-poor vesicles band in denser regions.
A second discontinuous gradient is essentially a concentration gradient to band the caveolin-rich vesicles sharply at an interface. Smart et al  used a Percoll™-based method for the initial purification of the plasma membrane, but there is no obvious requirement that such a method must be used. Kumanogoh et al  for example used a sucrose gradient to purify a synaptic plasma membrane before using the method devised by Smart et al .
There are many examples in the literature of iodixanol gradients being used to purify plasma membrane from a homogenate. The McDonald and Pike  method for the isolation of lipid rafts incorporates elements of both the detergent and the sonicated plasma membrane approaches. It involves performing two rounds of homogenization of CHO cells using multiple passages through a syringe needle.
A post-nuclear supernatant is then adjusted
to 25% (w/v) iodixanol and loaded under a 0-20% iodixanol gradient for floating the lipid-rich plasma
membrane fragments. In effect it resembles a Smart et al  method without a plasma membrane purification
step. It also omits the final iodixanol caveolae concentration gradient. Occasionally the Percoll™ gradient is
omitted, for example ref 4.