Expression of recombinant HLA-DR2 molecules. Replacement of the hydrophobic transmembrane region by a leucine zipper dimerization motif allows the assembly and secretion of soluble DR alpha beta heterodimers.

Major histocompatibility complex (MHC) class II molecules are membrane-anchored heterodimers that present peptides on the surface of antigen presenting cells to T cells. Soluble HLA-DR2 molecules were expressed for structural and functional characterization of the MHC/peptide/T cell receptor recognition unit. The α and β; chains of DR2 (encoded by the DRA, DRB1*1501 genes) did not assemble in mammalian or insect cell lines when the transmembrane regions of one or both chains were truncated. The hydrophobic transmembrane regions of DRα and DRβ; facilitate assembly of the heterodimer and were therefore replaced by the leucine zipper dimerization motifs from the transcription factors Fos and Jun, which assemble as a soluble, tightly packed coiled coil structure. The DRα-Fos and DRβ;-Jun constructs were expressed in a methyltrophic yeast, Pichia pastoris, using the α-mating factor secretion signal to direct expression to the secretory pathway. DR αβ; heterodimers were purified from supernatants using an antibody specific for the DR αβ; heterodimer. Kinetic and quantitative peptide binding experiments demonstrated that recombinant DR2 molecules were efficiently loaded with an antigenic peptide. Soluble DR2 molecules can be used to define structural aspects of the MHC/peptide/T cell receptor interaction and to study the signals induced by T cell receptor recognition of soluble DR2·peptide complexes.

MHC 1 molecules determine the specificity of T cell-mediated immune responses by binding peptides from foreign antigens in an intracellular processing compartment and by presenting these peptides on the surface of antigen presenting cells to T cells (for review, see Strominger and Wiley (1995)). MHC genes are highly polymorphic; with 137 known alleles, the DR␤ chain gene (DRB1) is the most polymorphic human gene that has been identified (Marsh and Bodmer, 1995). The polymorphic residues are clustered in the peptide binding site and thereby define the repertoire of peptides that are presented to T cells (Bjorkman et al., 1987;Stern et al., 1994). Some alleles of MHC class II genes confer susceptibility to autoimmune diseases, probably through the presentation of pathogenic self-peptides. For example, HLA-DR2 confers an increased risk for multiple sclerosis, while subtypes of HLA-DR4 confer susceptibility to rheumatoid arthritis (for reviews, see Todd et al. (1988) and Wucherpfennig and Strominger (1995b)).
Soluble, empty MHC class II molecules are required for crystallographic studies of single MHC⅐peptide complexes and for studying the biochemical interaction of MHC⅐peptide complexes with the T cell receptor. Structural characterization of the MHC⅐peptide/T cell receptor recognition unit will provide important insights into the mechanisms by which MHC molecules confer susceptibility to autoimmunity. Soluble MHC⅐peptide complexes may also be useful for the treatment of autoimmune diseases. Studies in the experimental autoimmune encephalomyelitis model have demonstrated that an autoimmune disease can be treated by the administration of soluble MHC⅐peptide complexes (Sharma et al., 1991).
MHC class II molecules can be purified by affinity chromatography following detergent solubilization of membranes (Gorga et al., 1987); however, MHC molecules purified from B cell lines have passed through the MHC class II peptide loading compartment and are therefore already loaded with a diverse set of peptides (Chicz et al., 1992). Soluble HLA-DR1 and HLA-DR4 molecules were expressed in insect cells using cDNA constructs of the DR␣ and DR␤ extracellular domains (Stern and Wiley, 1992). Expression of other MHC class II molecules has been difficult due to a failure of MHC class II ␣ and ␤ chains to assemble and/or due to a strong tendency for molecules to aggregate, even in the presence of a peptide ligand. These results may be explained by the observation that the transmembrane regions of the MHC class II ␣ and ␤ chains facilitate the proper assembly of the ␣␤ heterodimer, presumably through the interaction of the two ␣-helical transmembrane segments (Cosson and Bonifacino, 1992). Transmembrane interactions are also important for the assembly of other multiprotein complexes, such as the T cell receptor complex (Manolios et al., 1990).
The transmembrane regions were important for the assembly of DR2 molecules since the ␣ and ␤ chains did not assemble in mammalian or insect cells when the transmembrane regions were truncated. The leucine zipper dimerization motifs from the transcription factors Fos and Jun were therefore used to replace the hydrophobic transmembrane regions. Synthetic peptides of the Fos and Jun leucine zipper dimerization motif are known to assemble as stable, soluble heterodimers (O'Shea et al., 1989). The leucine zippers are characterized by five leucines that are spaced periodically at every seventh residue (heptad repeat); each heptad repeat contributes two turns of the ␣-helix (3.5 residues/turn). The leucine residues have a special function in leucine zipper dimerization and form the interface between the two ␣-helices in the coiled coil. The Fos/Jun heterodimer is soluble due to charged residues on the outer surface of the coiled coil (O'Shea et al., 1989(O'Shea et al., , 1991Hu et al., 1990). Use of the Fos and Jun leucine zipper dimerization motifs as a replacement of the hydrophobic transmembrane regions allowed the assembly and secretion of soluble HLA-DR2 in a yeast expression system, Pichia pastoris.

MATERIALS AND METHODS
DNA Constructs-The extracellular domains of DR␣ and DR␤ were expressed as fusions with the leucine zipper dimerization motifs of Fos or Jun, respectively. The extracellular domains of DR␣ and DR␤ as well as the Fos and Jun dimerization motifs were generated by polymerase chain reaction (PCR); primers were designed such that a 7-amino acid linker (VDGGGGG) with a SalI restriction site was included between the DR and the Fos/Jun segments. The DR segments were joined with the Fos or Jun segments through the SalI restriction site. These constructs were reamplified by PCR to permit cloning into the XhoI-EcoRI sites of pPIC9 as in-frame fusions with the ␣-mating factor secretion signal; the in-frame cloning preserved the Lys-Arg2Glu recognition sequence (cleavage C-terminal to Arg) required for cleavage of the ␣-mating secretion signal by the KEX2 gene product (Brake, 1990). The following oligonucleotides were used for the construction: DR␣ forward primer 5Ј GTA TCT CTC GAG AAA AGA GAG ATC AAA GAA GAA CAT GTG ATC 3Ј, XhoI site underlined; DR␣ reverse primer 5Ј GTC ATA GAA TTC TCA ATG GGC GGC CAG GAT GAA CTC CAG 3Ј, EcoRI site underlined (encodes 3Ј-end of Fos segment, stop codon and EcoRI restriction site); DR␤ forward primer 5Ј GTA TCT CTC GAG AAA AGA GAG GGG GAC ACC CGA CCA CGT TTC 3Ј (XhoI site underlined); DR␤ reverse primer 5Ј GTC ATA GAA TTC TCA ATG GTT CAT GAC TTT CTG TTT AAG 3Ј EcoRI site underlined (encodes 3Ј-end of Jun segment, stop codon and EcoRI restriction site). PCR products were cloned into the XhoI-EcoRI sites of pPIC9 and were verified by restriction mapping and dideoxy sequencing.
Transformation of P. pastoris with DR␣ and ␤ Chain Constructs-pPIC9 plasmid DNA was purified on CsCl gradients and digested with BglII to release the expression cassette (5Ј-end of AOX1 gene-DR␣ or DR␤ chain construct-polyadenylation signal-HIS4 gene-3Ј-end of the AOX1 gene). Transformations were done by spheroplasting of the GS115 strain (following the procedure provided by Invitrogen). Briefly, GS115 cells were grown to mid-log phase in YPD media (1% yeast extract (w/v), 2% peptone (w/v), 2% dextrose (w/v)), and spheroplasts were prepared by limited digestion of the yeast cell wall with zymolase (approximately 70% of spheroplasting) (Cregg et al., 1987). Cells were transfected with 5 g of DR␣ and DR␤ plasmid DNA, and transfectants that expressed the HIS4 gene (present in the pPIC9 expression cassette) were selected on HIS Ϫ plates. Integration of plasmids into the AOX1 locus was confirmed by replica plating of colonies on minimal media plates with methanol or dextrose as the sole carbon source. Transformants that had integrated the plasmid DNA into the AOX1 locus showed little or no growth on methanol plates due to disruption of the alcohol oxidase gene.
Identification of Recombinant Colonies-Integration of DR␣ and DR␤ chain constructs was examined by PCR analysis of genomic DNA isolated from individual Mut s colonies. Replica colonies were transferred into 200 l of lysis buffer (2.5 M LiCl, 50 mM Tris, pH 8.0, 4% Triton X-100, 62 mM EDTA) using a sterile toothpick. Acid-washed glass beads and an equal volume of phenol/chloroform (1:1) were added, and samples were vigorously vortexed. Following centrifugation, the upper phase was transferred to a clean tube, and genomic DNA was precipitated by addition of 2.5 volumes of cold EtOH. Following incubation at Ϫ20°C for 20 min, the pellet was collected by centrifugation, washed with cold 70% EtOH, and air-dried. DNA was resuspended in 40 l of sterile water and denatured at 94°C for 10 min; 10 l of DNA was used for each PCR reaction. DR␣ and DR␤ chains were amplified by PCR for 35 cycles (94°C 1 min, 55°C 2 min, 72°C 2 min) using the oligonucleotides that had been used to generate the DNA constructs; PCR products were resolved on 1% agarose gels stained with ethidium bromide.
Expression and Purification of HLA-DR2 Heterodimers-Recombinant molecules were purified from the supernatant of a recombinant strain that was grown to high density in a fermentor. Induction of high density cultures was carried out using a Inceltech LH series fermenter equipped with monitors and controls for pH, dissolved O 2 , agitation, temperature, and air flow. A 100-ml YNB-glycerol overnight culture was used to inoculate the fermentor, which contained 10 liters of fermentation basal salts medium (0.93 g/liter calcium sulfate 2 H 2 0, 18.2 g/liter potassium sulfate, 14.9 g/liter magnesium sulfate 7 H 2 O, and 6.5 g/liter potassium hydroxide) containing 4% glycerol (w/v) plus 43.5 ml of PTM 1 trace salts (24 mM CuSO 4 , 0.53 mM NaI, 19.87 mM MnSO 4 , 0.83 mM Na 2 MoO 4 , 0.32 mM boric acid, 2.1 mM CoCl 2 , 0.15 mM ZnCl 2 , 0.23 mM FeSO 4 , and 0.82 mM biotin) at 30°C. Dissolved O 2 was maintained above 20% by adjusting aeration and agitation, and pH was maintained at 6.0 by the addition of 28% (v/v) ammonium hydroxide. Growth was continued until the glycerol was exhausted (20 h). A glycerol-fed batch phase was initiated by the limited addition of 50% (w/v) glycerol and 12 ml of PTM 1 salts/liter of glycerol at 18.15 ml/h/liter initial fermentation volume until the culture reached a wet cell weight of 200 g/liter (22 h). After the glycerol-fed batch phase, the culture was induced by replacing the glycerol feed with a methanol batch feed (100% methanol containing 12 ml of PTM 1 trace salts/liter of methanol) at 1 ml/h/liter. The methanol feed was gradually increased in 10% increments every 30 min to a rate of 3 ml/h/liter, and the fermentation continued for a duration of 96 h.
Supernatants were concentrated by ultrafiltration on a YM30 membrane (Amicon) and passed over an anti-DR (mAb L243) affinity column at a flow rate of approximately 10 ml/h. Following extensive washing with PBS, heterodimers were eluted with 50 mM glycine, pH 11.5. Eluates were immediately neutralized by the addition of 2 M Tris, pH 8.0, dialyzed against PBS and concentrated by ultrafiltration. Protein concentrations were determined by Coomassie plus protein assay (Pierce) using bovine serum albumin as a standard.
Peptide Binding Experiments-A biotinylated myelin basic protein peptide (biotin-SGSGENPVVHFFKNIVTPR) that binds with high affinity to DR2 (Wucherpfennig et al., 1994) was used to examine peptide binding to recombinant DR2; the nonbiotinylated MBP peptide (ENPV-VHFFKNIVTPR) was used as a competitor in these binding assays.
For immunoprecipitation experiments, DR2 (400 nM) was incubated with biotinylated peptide (2 M) in a 50-l volume in PBS, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, pH 7.2, for 24 h at 37°C. DR2⅐peptide complexes were precipitated with streptavidin-agarose beads. Beads were first blocked with 3% bovine serum albumin in PBS, 0.1% Nonidet P-40 for 1 h at 4°C; beads were then pelleted, and the DR2⅐peptide samples were added. Following a 1-h incubation, beads were washed 3 times with blocking buffer. DR2⅐peptide complexes were eluted from streptavidin beads by heating in 1 ϫ SDS-PAGE buffer at 94°C for 3 min. Samples were resolved on a 12.5% SDS-PAGE and transferred to Immobilon membrane (Millipore). Blots were blocked overnight with 5% nonfat dry milk in 50 mM Tris, pH 8.0, 150 mM NaCl, 0.2% Tween 20 (TBST buffer). Precipitated DR␣ and ␤ chains were detected with a polyclonal DR antiserum (CHAMP, 1:50,000 in blocking buffer for 90 min). Blots were washed in TBST buffer and incubated for 30 min with a peroxidase-conjugated anti-rabbit IgG antibody (1:10,000 in blocking buffer). Following extensive washing in TBST, bands were detected by enhanced chemiluminescence (Amersham Corp.).
In a separate set of experiments, peptide binding to recombinant DR2 was quantitated by capturing DR2 molecules to ELISA plates with an immobilized DR antibody. Standard binding conditions were as follows: 37°C for 24 h in PBS, pH 7.2, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride. Following peptide binding, bound peptide was quantitated by ELISA. Plates were coated with 200 ng/well of purified L234 in 0.1 M bicarbonate, pH 9.6, overnight at 4°C. Nonspecific binding sites were blocked with 3% bovine serum albumin in PBS, 0.05% Tween 20 for 2 h. Samples were diluted in blocking buffer and added to the wells (1 h). HLA-DR2 bound biotinylated peptide was quantitated with streptavidin-peroxidase using ABTS as a peroxidase substrate; absorbance was read at 405 nm.

Replacement of the Transmembrane Region with the Leucine Zipper Dimerization Motif Allows Assembly and Secretion of
HLA-DR2 Heterodimers-Soluble HLA-DR1 and HLA-DR4 have been expressed in insect cells using cDNA constructs for the extracellular domains of DR␣ and DR␤ (Stern and Wiley, 1992). When this strategy was attempted for the expression of HLA-DR2 (using the DRA, DRB1‫1051ء‬ genes), DR ␣␤ heterodimers could not be detected in cell lysates or supernatants of infected Sf9 cells, even though separate DR␣ and ␤ chains could be readily detected in cell lysates by Western blot analysis. The requirements for the assembly of DR2 heterodimers were therefore examined by transfecting mammalian cells (Chinese hamster ovary cells) with different cDNA constructs. The DR␣ and DR␤ transmembrane regions could be replaced with a glycan-phosphatidyl inositol anchor from human placental alkaline phosphatase (Wettstein et al., 1991); however, DR2 molecules were not assembled when the transmembrane region of one chain was truncated (data not shown). These results strongly suggested that interaction of the transmembrane regions of DR␣ and DR␤ was important for the assembly of DR2.
The transmembrane regions of DR␣ and DR␤ were replaced with the leucine zipper dimerization motifs from the transcription factors Fos and Jun, which form stable heterodimers in solution ( Figs. 1 and 2). The extracellular domains of DR␣ (residues 1-191) and DR␤ (residues 1-198) were fused in frame with a 7-amino acid linker (VDGGGGG that contained a SalI restriction site) and the 40-amino acid leucine zipper domains of Fos (DR␣-Fos) or Jun (DR␤ -Jun) (van Straaten et al., 1983;Angel et al., 1988). The extracellular domains ended with DR␣ 191 (E) and DR␤ 198 (K) because charge-charge interactions between these two residues are thought to facilitate assembly (Cosson and Bonifacino, 1992). Since the leucine zipper has a tightly packed coiled coil structure, a (Gly) 5 spacer was included between the DR and leucine zipper segments to allow for rotational freedom of the chains (Fig. 2). These constructs were first tested in Chinese hamster ovary cells (using the native DR␣ and ␤ chain signal peptides). Chinese hamster ovary transfectants were found to assemble and secrete DR ␣␤ heterodimers, indicating that the Fos/Jun leucine zipper promoted the proper assembly of DR2 molecules (data not shown).
Expression of Soluble HLA-DR2 in P. pastoris, a Methyltrophic Yeast-For protein production, the DR␣-Fos and DR␤-Jun constructs were expressed in P. pastoris under the control of the alcohol oxidase (AOX1) promoter. P. pastoris was chosen because stable transformants can be rapidly generated and screened; in addition, several secreted proteins have been produced at very high levels in this system (Cregg et al., 1987). To direct expression to the secretory pathway, DR␣ and ␤ chains were cloned into P. pastoris expression vector pPIC9 as inframe fusions with the ␣-mating factor secretion signal (Fig. 3) (Brake, 1990). The ␣-mating factor secretion signal is cleaved by the KEX2 gene product (Leu-Glu-Lys-Arg2Glu; cleavage C-terminal to Arg). Although this design results in the addition of a glutamic acid residue to the N terminus of the mature DR␣ and DR␤ chains (see "Materials and Methods"), the N termini of these chains are located in a manner that this additional residue should not affect the assembly of the heterodimer. Molecules expressed as fusions with the ␣-mating factor secretion signal were efficiently secreted, while usage of the PHO1 secretion signal (vector pHIL-S1, Invitrogen) resulted in little or no secretion.
For transformation, the expression cassette of pPIC9 (Fig. 3) can be excised as a BglII fragment; the cassette carries 5Ј-and 3Ј-sequences of the AOX1 gene to allow for integration into the cDNA constructs were generated by PCR in which the extracellular domains of DR␣ (residues 1-191) and DR␤ (residues 1-198) were fused in frame with the Fos or Jun dimerization motifs (40 amino acids). A flexible 7-amino acid linker with a SalI restriction site and (Gly) 5 was inserted between the extracellular domains of DR␣ or ␤, and the Fos or Jun leucine zipper dimerization domains to allow for rotational freedom between the tightly packed Fos/Jun coiled coil and the DR␣ and ␤ chains. The leucine zipper dimerization motifs of Fos and Jun are both characterized by 5 leucine residues at every seventh position (heptad repeat, two turns of a helix with 3.5 residues/turn). The leucine residues have a special function in the formation of the leucine zipper and are located at the hydrophobic interface of the coiled coil. The Fos/Jun heterodimer is soluble due to charged residues on the outer surface of the coiled coil.

FIG. 3. Constructs for the expression of DR␣ and DR␤ in P.
pastoris, a methyltrophic yeast. DR␣-Fos and DR␤-Jun constructs were cloned into the P. pastoris expression vector pPIC9, which carries the ␣-mating factor secretion signal. cDNA constructs were under the control of the alcohol oxidase promoter (AOX1), which can be strongly induced by addition of methanol to the growth medium. For transfection of yeast cells (strain GS115, HIS Ϫ ), the expression cassettes (AOX1 promoter-cDNA construct-polyadenylation signal-HIS4 gene-3Ј-end of AOX1 gene) were excised with BglII. With this approach, genes integrate into the AOX1 locus by homologous recombination (5Ј-end and 3Ј-end of AOX1 gene); disruption of the AOX1 gene by integration of the expression cassette leads to a methanol utilization deficient (Mut s ) phenotype.
AOX1 locus as well as the HIS4 gene that allows for selection of transformants in histidine deficient media. Genes integrate into the AOX1 locus by homologous recombination; integration into the AOX1 gene disrupts the gene and leads to slow growth if methanol is the only carbon source (methanol utilization deficient phenotype, Mut s ) (Cregg et al., 1987).
A major advantage of the P. pastoris system is that transformants can be readily identified. Integration into the AOX1 locus confers a methanol utilization deficient (Mut s ) phenotype that can be determined by comparing the growth of duplicate colonies on plates with methanol or dextrose as the sole carbon source. Mut s colonies obtained after cotransformation of plasmids carrying the DR␣ and DR␤ chain constructs (see "Materials and Methods") were tested by PCR analysis of genomic DNA for the integration of DR␣ and ␤ chain genes. 27 of 28 colonies with a Mut s phenotype carried DR␣ and/or DR␤ chain genes; four of these colonies (14.2%) had integrated both genes (Fig. 4).
Assembly and Secretion of Soluble HLA-DR2 Molecules-The four transformants that carried both DR␣ and ␤ chain genes were examined for the expression of DR2 heterodimers. Cells were grown for 2 days in media containing glycerol as the sole carbon source and were then switched to media containing 0.5% methanol. Supernatants and cell lysates were examined by sandwich ELISA using a mAb specific for the DR ␣␤ heterodimer (mAb L243) for capture and a polyclonal DR antiserum (CHAMP) for detection. DR ␣␤ heterodimer was detected in the cell lysates and supernatants of DR ␣␤ transfectants. Transformants that carried only DR␣ or DR␤ chain genes were used as controls; cell lysates and supernatants from these cells were negative in the assay (Fig. 5). These experiments demonstrated that the DR ␣␤ heterodimer was assembled and efficiently secreted. The four Pichia clones showed similar expression levels; this is not surprising since all four transformants had integrated the genes into the AOX1 locus.
For large scale expression, cells were grown in a high density fermentor, and DR2 molecules were purified from concentrated supernatants by affinity chromatography. The mAb used for purification (L243) binds to the DR␣ chain but only when properly assembled with the DR␤ chain. Affinity purification yielded approximately 300 -400 g of HLA-DR2/liter of culture. SDS-PAGE revealed two bands (Fig. 6A); the identity of these bands (upper band DR␣, lower band DR␤) and appropriate cleavage of the ␣-mating factor signal peptide were confirmed by N-terminal sequence analysis following separation of DR␣ and ␤ chains by SDS-PAGE and transfer to a polyvinylidene difluoride membrane.
HPLC gel filtration analysis demonstrated that HLA-DR2 eluted as a single symmetric peak, demonstrating that the recombinant protein was not aggregated (Fig. 6B). HLA-DR1 expressed in the Baculovirus system was found to aggregate unless these molecules were loaded with a high affinity peptide (Stern and Wiley, 1992). These data demonstrated that the DR ␣␤ heterodimer was assembled and secreted in the P. pastoris expression system. Importantly, the purified molecules did not aggregate even though they had not been loaded with a high affinity peptide.
Loading of Soluble HLA-DR2 Molecules with a High Affinity Peptide-A human myelin basic protein (residues 85-99) that is recognized by DR2-restricted T cell clones from multiple sclerosis patients was previously shown to bind with high affinity (IC 50 of 4.2 nM) to detergent soluble DR2 purified from L cell transfectants (Wucherpfennig et al., 1994(Wucherpfennig et al., , 1995a. A biotinylated peptide (biotin-SGSG-ENPVVHFFKNIVTPR with SGSG as a spacer between the biotin moiety and the MBP sequence) was used to examine the specificity of peptide binding to recombinant DR2. Peptide binding was assessed by in- FIG. 4. Identification of P. pastoris transformants that carry DR␣ and DR␤ chain genes. Integration of DR␣ and DR␤ chain genes into Mut s transformants was examined by PCR amplification of genomic DNA using the oligonucleotides that had been used to generate the constructs. PCR products were resolved on a 1% agarose gel and stained with ethidium bromide. Four colonies that carried both DR␣ and DR␤ chain constructs were identified. MW, molecular weight marker (100-base pair ladder); C, control reaction (no genomic DNA).
FIG. 5. Assembly and secretion of DR2 by P. pastoris. Expression of DR2 was examined by sandwich ELISA using a mAb specific for the DR␣␤ heterodimer (L243) for capture and a polyclonal DR antiserum for detection; binding of the secondary antibody was quantitated with a peroxidase conjugated anti-rabbit IgG antiserum and with ABTS as a peroxidase substrate. Pichia colonies were grown for 48 h in media containing glycerol as a carbon source and were then switched to growth medium containing 0.5% methanol. Colonies carrying only DR␣ or DR␤ chain genes were used as controls. Ⅺ, DR␣; ࡗ, DR␤; E, DR␣ and -␤. cubating DR2 molecules with the biotinylated peptide for different periods of time; nonbiotinylated peptide was used as a competitor to demonstrate the specificity of binding (Fig. 7). DR2⅐peptide complexes were then captured on an ELISA plate using the L243 mAb, and the amount of bound biotinylated peptide was quantitated using peroxidase-labeled streptavidin.
Peptide binding to DR2 was strongly dependent on the pH, with a maximum observed at pH 7-8; relatively little binding was observed at pH 5. A similar pH optimum had previously been observed for binding of the MBP peptide to detergent soluble DR2 (Wucherpfennig et al., 1994). Binding of peptide was dependent on the relative molar ratio of DR versus peptide, with a maximum of binding at a 10-fold molar excess of peptide over DR2 (Fig. 7A). Binding was specific because it could be blocked by an excess of nonbiotinylated MBP(85-99) peptide but not by an analog peptide in which the P1 anchor residue of MBP(85-99) (valine 89) had been substituted by aspartic acid (Fig. 7B).
To determine what fraction of recombinant molecules could be loaded with a single peptide, complexes of DR2 and the biotinylated MBP peptide were precipitated with streptavidin beads (Fig. 8). Following precipitation, DR␣ and ␤ chains were resolved by SDS-PAGE and detected by Western blotting using a polyclonal DR antiserum. Approximately 50% of the molecules were precipitated with streptavidin beads (Fig. 8, lane 3), FIG. 6. Affinity purification of DR2 from supernatant of a P. pastoris transformant. A P. pastoris clone that carried both DR␣ and DR␤ genes was grown to a high cell density in a fermentor. Supernatant was concentrated by ultrafiltration, and DR2 was purified by affinity chromatography using a mAb specific for the DR ␣␤ heterodimer (L243). DR2 was eluted with 50 mM glycine, pH 11.5, dialyzed and concentrated by ultrafiltration. A, SDS-PAGE of purified DR2. The identity of DR␣ (upper band) and DR␤ chains (lower band) was confirmed by N-terminal sequence analysis following transfer of proteins from the SDS gel to a polyvinylidene difluoride membrane. B, high performance liquid chromatography gel filtration analysis of recombinant DR2. 10 g of DR2 were run on a BIO-SIL SEC-250 gel filtration column (300 ϫ 7.8 mm, Bio-Rad) in PBS, pH 6.8, at a flow rate of 1 ml/min. The injection time is marked with an arrow, the positions of molecular mass standards are marked with bars (molecular mass of 670, 158, 44, 17, and 1.3 kDa).

FIG. 7. Specificity of peptide binding to recombinant DR2.
Peptide binding was examined using a biotinylated MBP peptide that was previously shown to bind with high affinity to detergent-soluble DR2. A, binding of a biotinylated peptide to rDR2. DR2 (50 -400 nM) was incubated for 24 h at 37°C with biotinylated peptide (2 M) at pH 7.2. DR2⅐peptide complexes were captured on ELISA plate using an DR-specific mAb; DR-bound biotinylated peptide was quantitated using peroxidase-labeled streptavidin. Maximum binding was seen at a 10fold molar excess of biotinylated peptide over DR2. B, specificity of binding. Binding of the biotinylated MBP(85-99) peptide was inhibited by an excess of the nonbiotinylated MBP(85-99) peptide, but not by an analog peptide (89 aspartic acid) in which the P1 anchor residue of MBP(85-99) (residue 89, valine) was substituted by aspartic acid. DR2 (200 nM) was incubated for 24 h at 37°C with biotinylated peptide (2 M) at pH 7.2 in the presence of competitor peptide (0 -100 M). DR2bound biotinylated peptide was quantitated as described above.

50% remained in the supernatant (lane 4).
Control experiments demonstrated that precipitation of the DR2⅐peptide complexes was specific as the molecules were not precipitated when control agarose beads (lane 5), an unlabeled MBP peptide (lane 7), or an excess of unlabeled peptide over biotinylated peptide were used (lane 8); rather the DR2 molecules remained in the unbound fraction (lanes 6, 8, and 10).
Kinetics of Peptide Binding to Soluble HLA-DR2 Molecules-The kinetics of peptide binding by detergent soluble DR2 purified from an Epstein-Barr virus-transformed B cell line and by recombinant DR2 were compared (Fig. 9). The kinetics of peptide binding were strikingly different. With recombinant molecules, the kinetics of binding were much faster, and a much larger fraction of the molecules were loaded (50% maximum binding after only 3 h). In contrast, the kinetics of peptide binding to DR2 from B cells were slow; the fraction of peptide loaded molecules slowly increased over a 48-h period without reaching a plateau (Fig. 9). These results are explained by the fact that the majority of DR molecules purified from B cells are already occupied with high affinity peptides, as demonstrated by peptide elution studies and crystallization of HLA-DR1 (Chicz et al., 1993;Brown et al., 1993). In contrast, the peptide binding site of a large fraction of the recombinant DR2 molecules is readily available for binding by a high affinity peptide. Recombinant DR2 molecules will therefore be useful for studying the peptide binding specificity of HLA-DR2 and for generating complexes with defined antigenic peptides.

DISCUSSION
Soluble HLA-DR1 has been expressed in the baculovirus system using cDNA constructs for the extracellular domains of DR␣ and DR␤. These molecules were assembled and secreted but had a tendency to aggregate unless they were loaded with a high affinity peptide (Stern and Wiley, 1992). Other MHC class II molecules (such as the product of the DRA, DRB5*0101 genes) showed a strong tendency to aggregate when this approach was attempted, even when high affinity peptides were added. 2 The expression of mouse I-A molecules (I-A u and I-A g7 , which confer susceptibility to experimental autoimmune encephalomyelitis and to diabetes, respectively) has also been difficult. The native transmembrane region of these molecules could be replaced with a glycan-phosphatidyl inositol anchor; however, following cleavage of these molecules from the surface of transfected cells by phospholipase C, irreversible aggregation occurred even if the cells had been incubated with I-A binding peptides prior to cleavage. 3 Soluble DR2 molecules could not be expressed in the Baculovirus system because the extracellular domains of DR␣ and DR␤ chains did not assemble. These results indicated that the transmembrane regions were important for the proper assembly of DR2. Replacement of the hydrophobic transmembrane regions by the Fos and Jun leucine zipper dimerization motifs resulted in the assembly and secretion of ␣␤ heterodimers, both in Chinese hamster ovary cells and in yeast. The recombinant DR2 molecules were purified from supernatants of stably transformed yeast cells (P. pastoris) and could be efficiently loaded with an antigenic peptide. In contrast to other recombinant MHC molecules that have been expressed, these molecules were stable and did not aggregate, even in the absence of a high affinity peptide. These results demonstrate that the Fos-Jun heterodimer facilitates assembly and that it stabilizes the DR2 heterodimer.
The peptide binding site of MHC class II molecules is formed by the N-terminal domains of the MHC class II ␣ and ␤ chains; each chain contributes half of the floor and one of the two ␣-helices that flank the peptide binding site (Brown et al., 1993). The fact that the two chains have to be properly paired 2 K. Vranovsky and J. L. Strominger, unpublished results. 3 L. Fugger and H. McDevitt, personal communication.
FIG. 8. Precipitation of DR2 molecules loaded with a biotinylated peptide using streptavidin beads. DR2 (400 nM) was incubated with biotinylated peptide (2 M) for 24 h at 37°C. DR2-peptide complexes were precipitated with streptavidin-agarose beads, beads were washed, and DR2-peptide complexes were eluted by heating in 1 ϫ SDS-PAGE buffer. Samples were resolved on a 12.5% SDS-PAGE and transferred to a polyvinylidene difluoride membrane. Precipitated DR␣ and ␤ chains were detected with a polyclonal DR antiserum, followed by a peroxidase-conjugated anti-rabbit IgG antibody and ECL. Lanes 1 and 2, 200 and 500 ng of rDR2; lanes 3 and 4, DR2 incubated with biotinylated peptide, pellet (streptavidin bound, lane 3), and supernatant (not streptavidin bound, lane 4). Lanes 5-10 were controls: lanes 5 and 6, precipitation with agarose beads (no streptavidin); lanes 7 and 8, incubation of DR2 with non-biotinylated peptide; lanes 9 and 10, incubation of DR2 with biotinylated peptide and a 50-fold molar excess of nonbiotinylated peptide. In the controls, DR2 was not precipitated with streptavidin beads but stayed in the supernatant (lanes 6, 8, and 10).
FIG. 9. Kinetics of peptide binding to recombinant DR2. The kinetics of peptide binding were compared for recombinant DR2 and for detergent-soluble DR2 purified from an Epstein-Barr virus-transformed B cell line. DR2 (200 nM) was incubated with biotinylated MBP peptide (2 M) at 37°C for different periods of time; the amount of DR-bound peptide was examined by ELISA using a DR-specific antibody for capture and streptavidin peroxidase for quantification of bound peptide. Recombinant DR2 molecules were rapidly loaded (50% maximum signal after 3 h with a plateau after 18 h); in contrast, binding to detergent-soluble DR2 did not reach a plateau during the 48-h incubation period. Also, a much larger fraction of recombinant molecules could be loaded with the biotinylated peptide. These results demonstrate that recombinant molecules can be efficiently loaded with a single peptide, while the peptide binding site of DR molecules from B cells is already occupied with high affinity peptides.
for the peptide binding domain to be formed may explain some of the difficulties that have been encountered with the expression of recombinant MHC class II molecules. This report demonstrates that a MHC class II molecule can be appropriately assembled and secreted by P. pastoris. This eucaryotic expression system has the advantage that yeast cell transformants can be generated in a similar time frame as bacterial transformants; colonies appear 3-4 days following transformation and can be readily screened for integration of plasmid DNA by PCR. Positive colonies can be rapidly expanded in liquid media and tested for expression of the recombinant protein following induction of the AOX1 promoter. The AOX1 promoter has a low basal level of transcription; high level transcription can be rapidly induced by addition of methanol to the growth media (Cregg et al., 1987). Some secreted proteins have been expressed at very high yields (Ͼ1 g/liter) in this system; the expression level of HLA-DR2 was lower (300 -400 g/liter), possibly because heterodimer assembly was the rate-limiting step. Availability of (low affinity) peptides in the secretory pathway may be the limiting factor in the assembly of DR ␣␤ heterodimers; this may be overcome by coexpressing a DR-binding peptide in the secretory pathway (under the control of the AOX1 promoter as fusion with the ␣-mating factor secretion signal). The invariant chain peptide (CLIP) may be suitable for this purpose since it could be later exchanged with other DR2 binding peptide(s) (Avva and Cresswell, 1994). 4 Soluble DR2 molecules will allow a detailed characterization of the peptide binding specificity of HLA-DR2 in an effort to define the mechanisms by which this MHC molecule confers susceptibility to multiple sclerosis.