A recombinant approach to generate MHC class II reagents would offer many advantages in the analysis of the highly polymorphic MHC class II system. These include ease of production, manipulation, purification, and a high yield at a modest cost. Since the first report of recombinant MHC class II expression was published in 1992, many different approaches to recombinant MHC class II production have been suggested and all of them appear to be in current use i.e. there is no consensus on how to generate recombinant MHC class II molecules. Many variations have been tried. E. coli [8, 20, 21, 22, 35, 36, 37, 38, 39, 40], insect cells [36, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62], yeast cells [63, 64, 65], and eukaryotic cells [66, 67] have all been used as production cells. Natural inter-chain interactions – sometimes including the transmembrane domains [51, 52, 65, 66, 67] sometimes excluding the transmembrane domains (i.e. truncating the chains after the α2 and β2 domains) [20, 36, 39, 40, 43, 68, 69] – have been tried. Alternatively, assembly of MHC α and β chains have been facilitated by phosphatidyl inositol membrane anchoring , by fusion to leucine zippers [46, 49, 50, 53, 63, 70, 71, 72, 73, 74], by fusion to IgG chains [42, 45, 58, 75, 76], or enforced through the generation of single chain fusion constructs oriented either as α1 β1- or β1α1 [37, 38, 65]. Molecules have been produced in cells with an intact antigen processing machinery leading to molecules pre-loaded with a collection of naturally offered high-affinity binding peptides , or in cells deficient in antigen processing leading to putatively empty molecules. Alternatively, molecules have been loaded with a single predefined high-affinity binding peptide fused to the β chain [43, 45, 46, 47, 49, 56, 57, 58, 64, 76, 78]. These strategies have aimed at improving one or more aspects of MHC class II production; however, several have involved principles that significantly limit versatility. Systems that require co-expression of α/β heterodimers limit utility since cells expressing each α/β combination of interest will have to be generated one by one. Systems that require linking of a given peptide to the β chain limit utility since the stable binding of the linked peptide compromises subsequent binding experiments and/or replacement with any other peptide. Some reported class II expression systems have had low refolding yields [21, 22], some associated binding assays have depended upon a low pH incubation to release endogenous peptides [6, 79, 80, 81, 82, 83], and others have had to use high concentrations of reporter peptide to detect interaction thus precluding detection of high-affinity interactions [66, 73, 80, 81, 82, 84, 85, 86, 87].
Protein expression systems based upon E. coli expression are potentially fast, versatile and high-yield. Unfortunately, it would appear that many attempts to express class II in E. coli have failed . Major drawbacks of E. coli expression include lack of proper folding, disulfide bond formation, and glycosylation leading to aggregate deposition of these non-functional proteins in inclusion bodies. However, a few class II molecules, capable of binding any appropriate peptide offered, have already been successfully produced as isolated subunits in E. coli (DR1 [20, 22], DR2a  DR4  and I-Ek ). This demonstrates that it might be possible to express the two chains as isolated subunits and recombine them to generate any desired heterodimer capable of binding any appropriate peptide. This should lead to considerable savings, in particular for DP and DQ molecules, where a limited number of α and β chains can be combined to generate thousands of different receptors. Here, we illustrate this latter point by making HLA-DP and DQ molecules composed of polymorphic α chains (DP1A*0301 and DQA1*0501) paired with the polymorphic β chains (DPB1*0401 and DQB1*0201) respectively.
Here, we have generated an efficient E. coli -based expression system for MHC class II molecules. Our approach to E. coli production of MHC class II molecules differs in several respects from those described in the literature. We have used dimerizing modules to facilitate class II α/β pairing and refolding. To the best of our knowledge this has never before been attempted for class II molecules produced in E. coli . We have also used a pre-oxidized refolding principle. To our knowledge, all past attempts at producing class II in E. coli have involved extraction of class II proteins from inclusion bodies using denaturant solutions containing a reducing agent followed by refolding by dilution into a buffer containing a suitable redox pair to facilitate disulphide bond formation. Such refolding approaches are frequently plagued by low yields. We have successfully produced functional class I molecules in high yield from E. coli exploiting the fact that correctly pre-oxidized class I heavy chain molecules can be extracted from inclusion bodies in the absence of reducing agents (note, we don't know whether these disulfide bonds are generated in vivo in the bacteria, or in vitro during the protein extraction process). Upon dilution into refolding buffer such pre-oxidized class I molecules refolds rapidly and completely. Indeed, our preparations of MHC class II α and β chain proteins contain pre-oxidized species, and they appear essential for the efficiency of the refolding process. We speculate that this is the main reason why all nine MHC class II molecules, that we have produced, have been successfully refolded and been useful in studies involving binding of soluble peptide.
One of our primary motivations to generate recombinant MHC class II molecules is to examine their peptide binding specificity and eventually generate accurate predictors of this event. To this end, biochemical assays should be able to provide large amounts of detailed quantitative binding information. Initially, we developed a standard assay detecting binding of radio-labeled peptide by gel filtration, and used this to systematically vary a number of parameters such as refolding additives, pH, temperature and time etc. A few parameters seemed universally beneficial; most pronounced was the addition of glycerol to the refolding buffer. Stern et al have also noticed this . Other parameters such as length (i.e. where to truncate the α and β chains, additional file 1, Figure 4) and pH were more variable and had to be optimized for each MHC class II heterodimer in question.
Ideally, high-throughput binding assays should be developed to deal with the large number of potential peptide-MHC class II combinations of interest. In the additional information, we demonstrate that a homogenous Scintillation Proximity Assay (SPA) is an attractive high-throughput method for those who consider assays based upon radioactivity as an option. We have also developed assays that do not depend upon radioactivity. One is based on the interaction of the peptide-MHC combination in question followed by a standardized sandwich ELISA to measure the concentration of bound peptide. This can be implemented in most laboratories and yields robust results capable of measuring binding affinities in the low nM range. We have recently developed a non-radioactive, high-throughput homogenous assay for peptide-MHC class I interaction. This assay is based upon Luminescent Oxygen Channeling Immunoassay (LOCI). Here, we demonstrate that this detection mode also works for peptide-MHC class II. In fact, the signal to noise ratio is even better for the LOCI-based assay than it is for the ELISA based assay. The LOCI-based assay has been automated for large-scale screening for MHC class II restricted epitopes (Table 1).