Cloning of HLA constructs

All genes were generated synthetically by consecutive overlap extension PCR reactions and QuikChange mutations, or purchased from GenScript, cloned into the pET28a+ (kanamycin resistance, IPTG induction) vector and transformed into DH5α E. coli cells using standard molecular biology techniques. The intended DNA sequences were verified by DNA sequencing (ABI3100, Perkin Elmer). Plasmids were purified and transformed into BL21(DE3) E. Coli cells for protein production. To allow for in vivo biotinylation, BL21(DE3) were co-transformed with a pASYC (chloramphenicol resistance, IPTG inducible) vector with the gene encoding the BirA biotinylation holoenzyme. Clones, which produced the fusion product upon induction with IPTG, were identified and used for large-scale expression. The full length sequences of produced protein constructs can be seen in additional file 1, Figure 1

Expression of HLA class II α and β chain proteins in E coli inclusion bodies

All α chains were produced in shaker bottles using an auto-induction method described by Studier [1].

All β chains were expressed in a fermentor using IPTG induction as previously described [2] (except for DPB1*0401, which was produced by auto-induction). Briefly, cells were expanded overnight and used to seed a 2.5 L Labfors^® fermentor. Cells were grown at 37°C to an OD of 25. The temperature was then raised to 42°C and IPTG added to a concentration of 1 mM. For in vivo biotinylation 0.5 mM d-Biotin was added at the time of induction. After 3 hours, cells were processed at 2.3 kBar in a cell disrupter (basic Z, Constant Systems Ltd Daventry, UK). Using centrifugation (Sorvall RC6, 17,000 g, 30 min, 4°C), the inclusion body pellet was washed twice in 0.5% NP40, 0.1% DOC in PBS.

Purification of denatured HLA α and β chain proteins

The washed pellet was dissolved overnight in 200 ml 8 M Urea, 25 mM Tris, pH 8, and any remaining DNA was precipitated with streptomycin sulphate (10 g/L). After centrifugation (Sorvall RC6, 17,000 g, 30 min, 4°C), the denatured protein solution was applied to an 800 ml Q Sepharose Fast Flow column. The column was washed with 8 M Urea, 25 mM Tris, pH 8 (Buffer A), and eluted with a two-step gradient (0–30% in 3 column volumes (CV), 30–100% in 1 CV of buffer B (A + 1 M NaCl). Fractions containing proteins of interest, as determined by SDS-PAGE, were pooled and concentrated to 100 ml using 10 kDa cut-off tangential ultrafiltration (Vivaflow 200, Vivascience AG, Göttingen, Germany). The concentrate was applied to a 3.5 L Superdex 200PG gel filtration column (GE Healthcare) and eluted with 8 M Urea, 25 mM Tris, 150 mM NaCl, pH 8. Fractions containing denatured monomer were pooled, frozen at -20°C, and used for further experiments.

Radiolabelling peptide with ¹²⁵I

For radio-iodination purposes the influenza hemagglutinin peptide, HA_306–318 (PKYVKQNTLKLAT), was extended N-terminally with a tyrosine generating (Y)HA_306–318. The peptide was ¹²⁵I iodinated using a chloramine method [3] to a specific activity of approximately 75 Ci/mmol corresponding to 500 cpm/μl at a final working concentration of 3 nM. We noted that only 30–40% of this particular reference peptide could bind regardless of how much MHC was added. We speculated that iodination of the tyrosine in the first anchor position of the HA_306–318 peptide might influence binding. Indeed, changing this tyrosine to phenylalanine, another accepted first anchor residue, increased the relative amount of radioactivity that could be incorporated into MHC complexes (data not shown).

MHC class II binding of radiolabelled peptide measured by gel filtration

Equimolar concentrations of denatured DRA*0101_1–191 and DRB1*0101_1–198 chains (4 μM each in 8 M urea) were diluted 50 times to a final concentration of 80 nM MHC. The refolding buffer consisted of 25% v/v glycerol, 50 mM Tris, pH8, PMSF (11.4 μg/ml), pepstatin (3.3 μg/ml), TLCK (3.3 μg/ml) and TPCK (3.3 μg/ml) – and in some cases (see below) further additives. Just before use, the refolding buffer was supplemented with 3 nM ¹²⁵I labeled (Y)HA_306–318 peptide. The reaction mixture was incubated for 24 h at 18°C. After incubation, free and MHC-bound ¹²⁵I labeled peptide was separated by gradient centrifugation spun column chromatography as previously described [4], and counted by gamma spectrometry (Packard Cobra 5010). The fraction of bound peptide was calculated as (peptide bound/(peptide bound + peptide free)).

Optimizing refolding and peptide binding conditions

Various buffer compositions were tested in an attempt to optimize refolding and peptide binding: Glycerol (range 0 to 25% v/v), Arginine (0 to 500 mM), Urea (0 to 1000 mM), Sucrose (0 to 500 mM), Sodium Chloride (0 to 500 mM), N-Laurylsarcosin (0 to 1,5% w/w), Dextran (0 to 0.5% w/w), Pluronic acid F68 (0 to 0.5% w/w), Deoxycholate (DOC) (0 to 0.5% w/w), NP40 (0 to 0.5% w/w), PEG6000 (0 to 0.5% w/w), PEG20000 (0 to 0.5% w/w), beta octylglycoside (0 to 0.5% w/w), and Tween20 (0 to 0.5% w/w). Similarly, various pH conditions were tested using a refolding buffer composed of protease inhibitors (as above), 25% v/v Glycerol, and 50 mM Tris/Morpholine Ethane Sulphonic acid (MES) adjusted to pH values between 5 and 10. After 24 h incubation at 18°C, the samples were analyzed by spun column chromatography (as above).

Time and temperature conditions were tested using a refolding buffer (as above) adjusted to pH 8. Aliquots were distributed into PCR tubes and incubated at 10, 20 or 30°C with temperatures controlled by a PCR thermocycler. At different time points, tubes were removed from the thermocycler and stored at -20°C until analyzed by spun column chromatography (the high glycerol content of the buffer allowed this kind of cryopreservation).

Unless otherwise stated, the following optimal MHC class II refolding and peptide binding conditions were subsequently chosen: a refolding buffer composed of protease inhibitors (as above), 0.01% Pluriol F68, 25% v/v Glycerol, and 50 mM Tris/Citrate pH 6–7.5, and reaction mixtures were incubated for 24 h at 18°C.

Refolding and purification of MHC class II standards

2 ml of 10 μM denatured MHC II α and β chain (DR1: DRA*0101_1–191/DRB1*0101_1–198(C30S); DR2a: DRA*0101_1–181/DRB5*0101_1–190; DR4: DRA*0101_1–181/DRB1*0401_1–190) was diluted drop wise into 60 ml refolding buffer pH 7.5 containing 2 μM HA_306-318H6 (PKYVKQNTLKLATHHHHHH). After 48 h incubation at 18°C, the refolding mixture was loaded onto a 6 ml Ni²⁺ charged IDA column. The column was washed with PBS (Buffer A) until the UV280 signal reached baseline followed by a two segment gradient (0–20% in 5 CV, and 20–100% B in 2 CV) with buffer B (PBS supplemented with 250 mM Imidazole). Fractions were collected and analyzed by reducing SDS-PAGE.

MHC class II binding measured by ELISA

Monoclonal murine antibodies LB3.1[5, 6] (mouse anti-DR α chain), D1.12 [7] (anti-DR), L243 [8] (anti-DR α chain) G8 [9] (anti-DR), 9.3F10 [6, 10] (anti-DR, DQ), 2.06 [11] (anti-DR), B7/21 [6] (pan specific-DP), and 14.4.4S [12] (anti I-E^d) were purified from culture supernatants (L243 and 14.4.4S), or ascites, by anion exchange or protein A chromatography.

Appropriate MHC class II standards were serially diluted in PBS and added to a Streptavidin plate (NUNC, Denmark), which had been blocked in 5% skim milk powder (SMP). After 1 h incubation at RT, the plate was washed 3 times in PBS with 0,01% Tween-20, and monoclonal anti-MHC class II antibodies were added (50 μl/well of a 10 μg/ml dilution in PBS with 2% SMP). After 1 h incubation at RT, the plate was washed 3 times in PBS with 0.01% Tween-20, and HRP conjugated goat anti-mouse antisera (A9917, Sigma Aldrich) was added (50 μl/well of a 1/12,000 dilution in PBS with 2% SMP). After 1 h incubation at RT, the plate was washed 3 times in PBS with 0.01% Tween-20, and 3,3',5,5'-tetrametylbenzidin (TMB) "One-Step Substrate" (DAKO) was added (50 μl/well). After 30 min incubation at RT in the dark, sulfuric acid (50 μl/well of a 0.3 M solution) was added to stop the color reaction, and the plates were read at 450 nm in a Powerwave microplate spectrometer (Bio-Tek instruments).

To identify optimal concentrations of MHC class II α and β chains (DR1: DRA*0101_1–191/DRB1*0101_1–198(C30S); DR2a: DRA*0101_1–181/DRB5*0101_1–190; DR4: DRA*0101_1–181/DRB1*0401_1–190), six titrations of α chains, and eight of β chains, were diluted in 8 M Urea, 25 mM Tris, pH 8; and the two titration series were combined to create a checkerboard of 48 different combinations of α/β concentrations. These were further diluted 33 times into a refolding buffer pH 6–8 with, or without, 2 μM HA_306–318. After 24 h incubation at 18°C, the plates were developed in an ELISA (as described above) using appropriate anti-MHC class II monoclonal antibodies as detection antibodies. Signals and signal/noise ratios were calculated for each combination of α/β concentrations.

Using optimal α/β concentrations, denatured MHC class II α and β chains were diluted into refolding buffer with a titration of test peptide, and incubated for 48 h at 18°C. The ELISA was used to measure the resulting peptide-MHC class II complexes. The absorbance values (450 nm) were graphed vs. the concentrations of peptide offered, and the data analyzed by GraphPad Prism as described below.

MHC class II binding measured by Luminescent Oxygen Channelling Immunoassay (LOCI)

LOCI (marketed by Perkin Elmer as AlphaScreen) is a two-bead assay system: donor beads contain a photosensitizer compound, which upon illumination with laser light at a wavelength of 680 nm converts ambient oxygen to energy-rich, short-lived singlet oxygen; and acceptor beads, which can respond to singlet oxygen with a luminescence/fluorescence cascade leading to an amplified signal in the 520–620 nm range. The oxygen released by a donor bead will only excite acceptor beads within a distance of ~200 nm, and this allows a proximity-based homogenous assay quantifying interacting biomolecules. We have recently described a LOCI-based HTS assay for measuring peptide-MHC class I interaction [13].

Here, donor beads coated with streptavidin, and acceptor beads for custom protein coupling were both purchased from PerkinElmer. Specific anti-MHC class II monoclonal antibodies were coupled to acceptors beads following the manufacturers recommendation (L243 for HLA-DR; B7/21 for HLA-DP; 9,3F10 for HLA-DQ; and 14-4-4 for I-E^d). Peptide-MHC class II reaction mixtures were generated and incubated 48 h at 18°C as described above for the ELISA. The peptide-MHC class II reaction mixtures were then mixed with equal volumes (12–25 μl) of a solution containing streptavidin donor beads and anti-MHC-II monoclonal antibody conjugated acceptor beads (both at 10 μg/ml in PBS containing 0,1% Pluronic F68). The plates were incubated for 18 h at 18°C and then read in an ENVISION reader (Perkin Elmer). As for the ELISA, the optimal combination of α and β chain concentrations was identified in pilot experiments. Then, peptide titrations and the resulting peptide-MHC class II formations were determined in a LOCI assay calibrated with a known MHC class II standard.

A LOCI-based competitive assay was also developed. In this assay, a binding reaction was set up between a trace concentration of a biotin labeled agonist peptide (biotin attached via a PEG linker to HA_306–318) and non-biotinylated MHC-II molecules. For many HLA-DR molecules one could use low nanomolar (in casu 4 nM) concentrations of agonist peptide. The resulting agonist-MHC-II interactions were developed using a LOCI assay as described above. Once an agonist-MHC-II interaction assay had been established, competition assays using titrations of any test peptide of interest could be conducted.

Peptide affinity calculations

The formation of peptide-MHC complexes was calculated from LOCI generated data, which had been calibrated using standard curves obtained with purified peptide-MHC complexes of known concentrations.

For direct binding experiments, the concentrations of peptide-MHC class II complexes formed were graphed versus the concentrations of peptide offered, and analyzed by non-linear regression (GraphPad Prism). The peptide concentration resulting in half saturation, the half maximal effective concentration (EC₅₀), was estimated by fitting the experimental data to the equation Y = Bmax*X/(K_D+X), where Y is the concentration of peptide-MHC-II complexes formed and X is the concentration of ligand (peptide) offered. The EC₅₀ approximates the K_D as long as the receptor concentration used is less than the K_D thus avoiding ligand depletion.

For the competitive binding assay, the concentrations of peptide-MHC class II complexes formed were graphed versus the logarithm of the concentrations of inhibitory peptide used, and analyzed by non-linear regression (GraphPad Prism). The peptide concentration resulting in half inhibition, the half maximal inhibitory concentration (IC₅₀), was estimated by fitting the semi-log transformed experimental data to the equation Y = Bottom + (Top-Bottom)/(1+10^(X-Log(IC₅₀))), where Y is the concentration of agonist-MHC-II complexes formed, X is the logarithm of the concentration of inhibitor offered, Top is the upper plateau of the curve (i.e. without inhibitory peptide) and Bottom is the lower plateau of the curve (i.e. with saturating concentrations of the inhibitory peptide). The IC₅₀ approximates the K_D of the inhibitor peptide as long as the receptor concentration used is less than the K_D's of both the agonist and inhibitor peptides [14].

Design of MHC class II α and β chains

HLA class II α and β chain polypeptide sequences were obtained from the IMGT/HLA Sequence Databasehttp://www.anthonynolan.org.uk/HIG/. The I-E^d α and β chains were obtained from the Swiss-Prot database http://beta.uniprot.org/; accessions P01904 and P01915, respectively. Gene sequences were codon-optimized for E coli expression and truncated to remove the transmembrane regions. Most chains were produced in two different truncations; a short and a long version of the α chain comprising amino acids 1–181 and 1–191, respectively; and a short and a long version of the β chains comprising amino acids 1–190 and 1–198, respectively (see additional file 1, Figure 4).

We have previously demonstrated how unpaired cysteines can seriously reduce functional yield [15, 16]. Unpaired cysteine residues in positions 30, 47 and 123 of DRB1*0101, DQA1*0501 and DPA1*0103, respectively, were therefore mutated to serine. Position 30 of DRB1*0101 and 47 of DQA1*0501 is located in the binding groove and might affect the peptide binding specificity, albeit a cysteine to serine might be considered a conservative substitution in an MHC context. Position 123 of DPA1*0103 is located in the second domain and would not be expected to influence the specificity of peptide binding.

Initially DQA1*0501 expressed poorly in bacteria, however N-terminal fusion of the natural histidine affinity tag (HAT) increased expression.

To assist in α/β chain assembly, a pentaglycine linker followed by the fos and jun leucine zipper was added to the C-terminal of the α and β chain, respectively. To allow for specific enzymatic biotinylation, the jun leucine zipper of the β chain was further extended by a triglycine linker and a biotinylation substrate peptide [17]. Using the E. coli GrpE chaperone to direct dimerization, an alternative α/β chain assembly system was successfully developed (see additional file 1).

In all cases, the β chains were produced with a C-terminally added biotinylation substrate peptide (BSP) allowing in vivo biotinylation using co-induced BirA holoenzyme to couple d-biotin specifically to the BSP sequence [18].

The resulting polypeptide sequences are shown in additional file 1, Figure 1

Expression and Purification of MHC class II α and β chains

After induction, the recombinant proteins were contained in bacterial inclusion bodies. Inclusion bodies were processed by high-pressure cell disruption, harvested and washed by centrifugation. The recombinant proteins were extracted from the inclusion bodies into 8 M Urea and purified by Q Sepharose anion exchange chromatograhy followed by Superdex 200PG gel filtration chromatography. Pure preparations were obtained as demonstrated by non-reducing SDS-PAGE (exemplified for DRB5*0101, Figure 1A). The yields were typically 50–70 g/L for autoinduction and 80–120 g/L for IPTG induction of pure α and β chain, respectively.

Denaturing, but non-reducing, conditions were maintained throughout the extraction, purification and subsequent storage procedures. We have previously described that a large fraction of MHC class I molecules produced in this manner is correctly oxidized and refolds very efficiently [2, 16, 18]. Being structurally related to MHC class I, we reasoned that MHC class II molecules might also be produced in a highly active, pre-oxidized form. Indeed, non-reducing SDS-PAGE showed slightly smeared MHC bands (Figure 1A), or in some cases even distinct and closely appositional bands, which disappeared upon reduction and generated a single band (Figure 1C). Overloading the SDS-PAGE only visualized few contaminants; mostly high-molecular aggregates of MHC since they disappeared upon reduction while the staining of the MHC band increased (Figure 1C). Storage at -20°C for a year gave rise to modest formation of high-molecular weight aggregates and complexes (Figure 1A), which disappeared upon reduction. Together these data suggest that our preparations predominantly contained MHC proteins, and as expected consisted of different di-sulfide isomers.

We have previously reported that in vivo enzymatic biotinylation can be very efficient. For a large cohort of MHC class I molecules we routinely obtained biotinylation efficiencies in excess of 90–95%. Similar high biotinylation efficiencies were obtained for MHC class II as illustrated by a SDS-PAGE gel-shift assay where an excess streptavidin led to a mobility shift of almost the entire population of monomers (Figure 1B).

Characterization of peptide-MHC class II complexes

A high-affinity binding peptide with a hexa-histidine affinity tag was used to purify functional DRA*0101/DRB5*0101 (DR2a) heterodimers. Purified denatured DRA*0101_1–181 and DRB5*0101_1–190 were diluted drop-wise into a refolding buffer containing a histidine-tagged influenza peptide, HA_306-318H6, incubated for 48 h at 18°C, and then purified by immobilized Ni²⁺ affinity chromatography (Figure 1D). Samples were taken from each fraction, reduced and boiled, and analyzed by SDS-PAGE (Figure 1D insert). Since empty α and β chains bound with low affinity to the Ni²⁺ column, a two-segment gradient was used to separate empty chains from complexes. The chromatogram therefore showed two elution peaks, which both contained MHC molecules (labeled II and III in Figure 1D and insert).

However, the late peak was only seen when the HA_306-318H6 peptide was present during refolding, whereas the early peak was generated even in the absence of the peptide (data not shown). To further support that the late peak contained relevant peptide-MHC class II complexes, a non-reducing SDS-PAGE gel was performed with and without boiling the samples. It has previously been shown that many peptide-MHC class II complexes are resistant to SDS-PAGE as long as they are not boiled [19]. The hexa-histidine tagged HA_306-318H6 peptide was used to generate and purify three different DR molecules, DR1, DR2a and DR4, as described above, and samples from the late peak subjected to boiling and SDS-PAGE. For all three DR molecules, the late peaks contained SDS-resistant, but heat sensitive, complexes (Figure 1E), whereas the early peaks did not contain SDS-resistant complexes (data not shown). We conclude that our recombinant MHC-II molecules exhibit the known peptide-dependent SDS-sensitivity of native MHC-II molecules.

We also noted that the folding efficiencies of our recombinant DR1, DR2a and DR4 molecules interacting with the Ha_306-318H6 peptide, under the conditions of these initial experiments, were quite high: 12, 23, and 14%, respectively.

Development of a peptide-MHC class II binding assay

To optimize the refolding conditions of MHC-II molecules, we initially employed a robust and easily interpretable assay involving binding of radioactively labeled peptide to MHC class II molecules. Briefly, a dose-titration of equimolar denatured DR1 α and β chains were diluted into refolding buffer containing 3 nM of a ¹²⁵I labeled reference peptide, (Y)HA_306–318, and incubated (24 h, 18°C). Binding was determined by gradient-centrifugation spun column gel-filtration separating bound vs. free peptide [4]. An MHC class II dose-dependent binding was observed with a half maximal binding occurring at approximately 20 nM (Figure 2A). In an attempt to optimize the binding reaction, we tested the effects of various commonly used refolding additives such as arginine, glycerol, sucrose, detergents, alcohols etc. Glycerol was found to have the a pronounced positive impact upon refolding yields (Figure 2B); no other investigated refolding additive had any significant positive effect upon refolding and many had a negative effect (additional file 1, Table 1). The positive effect of glycerol upon MHC class II folding has been noticed by others [8, 20, 21, 22]. Refolding of proteins can be very pH dependent and the effect of pH was therefore also examined. The optimal pH for refolding and incorporation of radio labeled (Y)HA_306–318 was found to be around pH 7–8 (Figure 2C). Then, the kinetics of peptide binding to de novo diluted recombinant MHC class II α and β chains at different temperatures were investigated. The rate of peptide complex formation was found to be highly temperature dependent (Figure 2D). Steady state bindings were reached after 8, 11 and 48 h incubation at 30, 20, and 10°C, respectively. For all temperatures, a decline in bound peptide was observed for incubation periods longer than 20–30 h. This is probably due to the stability of one or more of the assay components being compromised during long-term incubation.

Competition experiments were performed to ascertain that the observed peptide-MHC class II bindings were saturable and specific. Increasing concentrations of competing peptide were added to a reaction involving binding of ¹²⁵I labeled (Y)HA_306–318 to DR1. The resulting binding was measured in the spun column assay and depicted as an inhibition curve (Figure 3). Using GraphPad Prism, non-linear fittings of the data were performed in all cases obtaining regression coefficients better than 0.99. IC₅₀ values could be calculated. The HA_306–318, the Invariant chain (positions 97–120, the CLIP peptide), and the Invariant chain (positions 73–208), were all good binders to DR1, whereas a C-terminal Invariant chain fragment devoid of CLIP (pos 118–208) was a poor binder. The fragment was previously shown to interact with empty HLA-DR1 molecules, augmenting rather than blocking binding of peptide [23].

Thus, using isolated recombinant MHC class II α and β chains it is possible to obtain specific peptide-MHC class II interaction, and measure the affinity of interaction.

Subsequent binding experiments were conducted with pre-oxidized MHC-II chains diluted into a refolding buffer containing peptide and 25% glycerol, 50 mM Tris/Citrate (or MES), a protease inhibitor cocktail, 0.01% pluronic acid F68, pH 8. This reaction mixture was incubated for 24 h at 18°C, and then analyzed for complex formation.

Efficient refolding of pre-oxidized, denatured MHC-II proteins

The above peptide-binding assay was used to investigate the basic premise of our approach: that pre-oxidized MHC class II molecules refold more efficiently than fully reduced MHC class II molecules. SDS-PAGE analysis of reduced vs. non-reduced α and β chain proteins clearly demonstrated that pre-oxidized species are present in the non-reduced protein preparations (Figure 4A). Preparations of denatured β chain proteins were reduced with graded concentrations of DTT, and then diluted into an excess refolding buffer containing various redox pairs as well as non-reduced denatured α chain and radiolabeled (Y)HA_306–318 peptide, and incubated. The resulting complex formation was analyzed by spun column gel filtration as described above. Whereas the pre-oxidized species were highly active with respect to peptide binding, exposing the denatured β chain proteins to reduction with as little as 0.6 mM DTT lead to a significant loss of peptide-binding capacity, and none of the tested concentrations of redox pairs could regain full peptide-binding capacity (Figure 4B)

High-throughput screening assays of peptide-MHC class II binding (ELISA)

We have recently generated a quantitative ELISA assay for measuring peptide-MHC class I interaction [24]. In this assay, titrations of peptide and a fixed low concentration of refolding MHC molecules are co-incubated. The concentrations of the resulting complexes are determined, and it is possible to calculate the K_D values directly from the saturation curves. An important advantage of this approach is that there is no need to identify a reference peptide for each MHC to be addressed. However, this type of assay relies on the identification of a conformation-dependent MHC class II-specific antibody capable of discriminating between empty and peptide-loaded complexes.

To construct a quantitative sandwich ELISA for HLA-DR molecules, we decided to take advantage of our MHC class II molecules being biotinylated. Streptavidin coated plates were used to capture biotinylated HLA-DR molecules. A panel of anti-MHC class II antibodies were tested as detection antibodies using a refolded and purified biotinylated peptide-DR1 complex as target (additional file 1, Figure 3) The monoclonal anti-DR antibodies LB3.1, D1.12, L243 and G8 all yielded reasonable signals and were subsequently tested for their suitability as a detecting antibody.

Since the final assay would be based upon the correct pairing and refolding of separate MHC class II α and β chains, each potentially refolding with their own efficiency, it was important to determine the optimal concentration of each chain needed to generate peptide receptive complexes. Two-dimensional titrations were performed using pre-oxidized denatured α and β chains diluted to each their specified concentration and incubated in the presence, or absence, of an excess of binding peptide. After 48 h incubation at 18°C, the samples were analyzed in an ELISA using streptavidin as capture reagent, and L243, D1.12, LB3.1, or G8 as detection antibodies. An example of DR4 complex formation analyzed using L243 as the detection antibody is shown in Figure 5. For each of the 48 combinations of MHC class II α and β chains concentrations a signal (+ peptide, Figure 5A) and the corresponding noise (-peptide, Figure 5B) was determined, from these values a signal to noise ratios could be calculated (Figure 5C). In this case, final concentrations of 31 nM α and 1 nM β chains gave the highest signal to noise ratio. In a similar way, optimal concentrations were determined for other DR molecules (data not shown). The four selected antibodies were compared using these optimal MHC class II α and β chain concentrations to generate the different DR complexes in the presence, or absence, of binding peptide. Overall, L243 gave the best signal to noise ratios for the DR proteins tested; D1.12 and LB3.1 gave less discriminatory power; and G8 gave no or very little discriminatory power (the latter is perhaps not surprising as G8 is specific for the superantigen binding site [9]). L243 recognizes an epitope on HLA DR α chain [25] and has previously been used in DR peptide binding assays [8]. In subsequent DR experiments, L243 was used as a detection antibody. In the same manner, suitable antibodies for peptide-dependent DQA1*0501/DQB1*0201 (in casu 9,3F10) DP401 (in casu B7/21) or murine I-E^d (in casu 14.4.4s) complex formation were identified.

In an attempt to reduce the background further, different detergents, a chaotrope (urea) or an elevated temperature, were attempted included in the detection step. Both DR1 and DR4 were tested. In all cases, the inclusion of detergents, urea or an increased temperature, increased the signal to noise ratios; unfortunately at the cost of overall signal. The most pronounced signal to noise improvement (about three fold) was obtained with the mildly denaturing detergent DOC, however, at a concomitant loss of half the signal. None of these manipulations were subsequently used.

Peptide titrations were then carried out on the DR molecules, DR1, DR2a and DR4, using optimized α and β chain concentrations (Figure 6A) followed by the above described L243-driven ELISA analysis, and curve fitting the results. The affinities were approximated from the peptide concentrations leading to half-saturation of the MHC; the EC₅₀'s (Figure 6B). The ELISA-based assay appeared capable of measuring peptide-binding affinities for all three DR molecules, and affinities in the low nanomolar range could be determined. From the titration curves it is apparent that the HA_306–318 peptide gives a higher B_max values compared to the CLIP peptide 97–120. The reason for this is not known, but could be caused by different complex stabilities affecting the number of complexes surviving the ELISA processing and development procedures.

High-throughput screening assays of peptide-MHC class II binding (direct LOCI)

We have recently developed a HTS assay for peptide-MHC class I binding using a non-radioactive bead-based homogenous proximity assay: Luminescent Oxygen Channeling Immunoassay [13]. As one tag, we used the biotin group engineered unto the β chain. As the other tag, we used conformation-dependent MHC class II specific antibodies (L243 for DR; 9,3F10 for DQ; B7/21 for DP and 14.4.4s for I-E^d). As for the development of the ELISA, two-dimensional titrations of MHC class II α and β chain concentrations were performed and diluted into refolding buffer containing, or not containing, an excess concentration of appropriate binding peptide. After 24 h incubation at 18°C, a mixture of streptavidin-donor beads and antibody coupled acceptor beads were added to the reaction wells and analyzed (see Figure 7). It was evident that peptide-MHC class II complexes required both MHC class II α and β chains, and were generated in a strictly peptide-concentration dependent fashion. In this case, optimal α and β chain concentrations were found to be around 6 nM and 1 nM, respectively. Optimal conditions in terms of whether long or short versions of the different molecules should be used, at what concentrations the α and β chain proteins should be used, and the optimal pH were subsequently determined (additional file 1, Table 2 and Figure 4). Strikingly, signal to noise ratios observed with the LOCI technique were consistently better than those observed with the more traditional ELISA technique (e.g. compare Figure 5C with Figure 7C).

In some cases, poor signal to noise ratios caused by rather high background, compromised the direct binding assay described above. To solve this problem, a competitive assay was developed using a trace concentration of a biotin tagged agonist peptide (in casu HA_306–318 at 4 nm) and non-biotinylated β chains molecules. Complex formation was determined using a LOCI assay detecting the biotinylated peptide with streptavidin donor beads and the peptide-dependent MHC-II conformation with appropriately monoclonal antibody coated acceptor beads. Any test peptide could now be titrated into a binding reaction involving the biotinylated agonist peptide and refolding MHC-II proteins. The resulting inhibition curves could be fitted by non-linear regression allowing the half maximum inhibitory concentration, IC₅₀, to be determined (GrapPad prism, see Materials and Methods). Figure 8 shows affinity measurements of a promiscuous malaria-derived peptide binding to four different HLA molecules. To test the direct and competitive assay formats, a panel of peptides was titrated on different MHC II alleles (Table 1). The direct and competitive formats were compared for HLA-DR2a, and found to yield similar results. Note also that known peptide-MHC-II binding interactions reported in the literature could be reproduced [26, 27, 28, 29, 30, 31, 32, 33, 34].