12: Liquid-liquid Immiscibility in Lipid Membranes Containing Cholesterol
At Stanford in the early 80’s I encountered a new graduate student, Sriram Subramanian, who wanted to join my research group. I was giving a course in chemical physics at the time and somehow learned that Sriram knew about Wigner rotation matrices. I wanted him to give a lecture on the subject. He said no he wouldn’t. I said he should. After several back and forth yeses and noes, I asked Sriram if he wanted to join my research group. He said “yes” and gave a fine talk on Wigner rotation matrices.
For a number of years I was interested in the theoretical possibility of liquid-liquid immiscibility in lipid membranes, as I originally proposed in 1981 (252). Very close to the end of Sriram’s PhD work I suggested he do a simple experiment: put a phospholipid and cholesterol on the surface of water and look at the mixture with an epifluorescence microscope (all the equipment was readily available). He said NO. I said, “you want a PhD?” The experiment was done (a day’s work at most) and was a resounding success, leading to hundreds – if not more – papers in this field. See (325, 427). (Sriram Subramanian is now Chief of the Biophysics Section in the Laboratory of Cell Biology at NIH and maintains a visiting faculty appointment in the Johns Hopkins School of Medicine.)
Immiscibility in lipid membranes has turned out to be an extensively studied subject. The first hint that this immiscibility might be present in cell membranes containing cholesterol arose from an observation of immiscibility in monolayers of lipids extracted from red cells (431). This was followed by observations of immiscibility in bilayers by other laboratories in including that of Sarah Keller, a former postdoc in my lab. A former student of Sarah Keller, Sarah Veatch and collaborators then went on to discover this immiscibility in blebs from the membranes of living cells. Two outstanding questions remain. What is the function of this potential immiscibility in a cell, and what is the physical chemical origin of this immiscibility? In the first case I have shown that a lipid mixture near a critical point may be particularly effective at solubilizing membrane proteins in 2D (479). In the second case Arun Radhakrishnan and I have made extensive calculations showing how complex formation between cholesterol and saturated fatty acids together with a repulsive interaction between the complexes and unsaturated fatty acids can lead to immiscibility (471). These interactions clearly play a role in determining the chemical activity of cholesterol, which may play a direct role in regulating the rate of cholesterol biosynthesis in cells.
The idea of complexes between cholesterol and phospholipids dates back many years. We used this concept together with a general theoretical model of liquids by John Wheeler and collaborators to account for immiscibility in membranes containing cholesterol. In retrospect, our use of the term “complex” between cholesterol and saturated phospholipid may have been a poor choice of terminology. (The theory not only includes pairwise interactions between molecules, but also mean field effects. It might therefore have been better to have used the term “extended complexes”.) Chemists often think of complexes as entities with well-defined molecular structures that can sometimes even be isolated and purified. (However, molecular complexes in liquids have been described with relatively well defined structures and very short lifetimes. See Zheng et al. Our view of cholesterol-phospholipid complexes is very general: if in a mixture of molecules A, B, C, … the probability of A being close to B is higher than random, we refer to this preferred proximity as complex formation. The idea of a specific structure for phospholipid-cholesterol complexes is further made implausible by our observation that at least in monolayers a saturated chain phospholipid can be replaced by two saturated fatty acids, or one fatty acid and one saturated chain lysophospholipid, all resulting in similar phase diagrams (462). For lattice models of intermolecular interactions and phase separations (critical phenomena) see Wheeler. Decorated lattice models for ternary systems yielding liquid-liquid immiscibility and ternary critical points have been given by Clark and Neece and are applicable to the lipid system considered here. These models have the advantage that they can be mapped onto the Ising model which through the work of Sarah L. Keller and collaborators is known to be appropriate to lipid bilayers.
We have also been criticized for trying to extrapolate from phase diagrams of monolayers to phase diagrams of bilayers. This criticism is perfectly reasonable since the thermodynamic conditions for equilibrium are so different. For a free, planar bilayer the equilibrium condition and area per molecule is one in which the surface tension is essentially zero. For monolayers the surface tension and the area per molecule are defined in part by the geometry of the trough containing the monolayer. And the phase diagrams under discussion have rarely been compared at equal molecular areas. Our rebuttal to such criticism is qualitative: Cholesterol–dependent liquid-liquid immiscibility was discovered in 1987 in monolayers (325) and only much later in bilayers (See leading references to Veatch and Keller in (478)). Cholesterol-dependent liquid-liquid immiscibility in three component lipid mixtures was also first discovered in monolayers in 2000 (450) and only later in bilayers. (Again, see Veatch and Keller for leading references.) What is true is that we were mistaken in holding to the belief that liquid – liquid immiscibility in binary mixtures of cholesterol and DPPC held in both monolayers and bilayers, based in part on some thermodynamic bilayer data and the monolayer immiscibility (See (478) for references to this erroneous conclusion.) But finally, the idea of “complexes” between saturated fatty acid phospholipids and cholesterol is in our view the simplest way of accounting for immiscibility in both monolayers and bilayers. A relevant comment of Linus Pauling to me was “an approximate theory is better than none.”