The process of generating a soluble nanoscale membrane assembly involves a self-assembly of the constituent parts: phospholipids, the protein of interest and the membrane scaffold protein, from a detergent micellar state. Hence one needs to have enough of the correct detergent to generate this micellar starting state. A description, together with a partial phase-diagram is contained in our earliest publications (e.g. NanoLetters 2, 853-856.)
As discussed in individual publications and the review articles cited earlier, one of the most critical factors to success is having the correct stoichiometry of phospholipids (PL), scaffold protein (MSP) and target protein. For the assembly of bare discs, we know the precise number of phospholipids necessary to assemble homogenous, monodisperse discs with constant size. A detailed summary of assembly protocols is available through the following link:
If one is slightly off ideal stoichiometry of lipid:MSP:target then one finds various other aggregates in the void volume of a size exclusion chromatographic separation. If further away from ideal stochiometry, various other sized entities (some of which are discoidal, others of which are various other structures) form. If very far off ideal, there are too many other aggregates to separate or characterize and one cannot isolate the desired nanolipoprotein particles. The problems many seem to be having in generating homogeneous nanodisc structures is most probably due to incorrect lipid:MSP ratios as well as using other amphipathic helical MSP sequences which have not been optimized to produce homogeneous and monodisperse entites (for instance see J. American Chemical Society 126, 3477-3487).
When one is including a detergent solubilized target protein in the self-assembly mixture, the general problem is that you do not know a priori how many PL the target protein will displace in the assembly – or in other words – what is the net number of PL in the resultant target protein incorporated into Nanodiscs. While we have pretty good idea of how many lipids are displaced from the bare discs for 7-TM proteins, anchored P450s and some other systems (see publications and review articles), the best approach is to try 3-5 different rations of PL:MSP:target.
Another comment relates to the protocols for incorporating more than one protein into a Nanodisc. We know a fair bit about this for the assembly of bR (Protein Science 12, 2476-2481; Archives Biochemistry and Biophysics 450, 215-222; J. Biol. Chem, 282, 14875-14881; J. Biological Chemistry 282, 7066-7076) for some PLs, but in general to get monomeric proteins into Nanodiscs we often self-assemble with an excess of PL and MSP over the target to generate bare discs and those with the target which can then be separated (most ideally with a tag on the target).
We are often asked if the Nanodisc assembly works for all proteins. Although one cannot guarantee anything, we have been able to assemble an amazing variety of membrane proteins and membrane protein assemblies into Nanodiscs. Where it will not work is if the target protein is already in an aggregated soluble state. Here the target is already happy in aqueous solution, although probably inactive. Interestingly, it is clear that you can sometimes get direct insertion into pre-formed Nanodiscs (e.g. simple anchored proteins like the P450 reductase and cytochrome b5) but in general, integral membrane proteins must be assembled from the detergent solubilized state.