Purification and Ligand Binding of a Soluble Class I Major Histocompatibility Complex Molecule Consisting of the First Three Domains of H-2Kd Fused to ,&-Microglobulin Expressed in the Baculovirus-Insect Cell System*

A recombinant baculovirus encoding a single-chain murine major histocompatibility complex class I molecule in which the first three domains of H-2Kd are fused to &-microglobulin (&m) via a 15-amino acid linker has been isolated and used to infect lepidopteran cells. A soluble, 391-amino acid single-chain H-2Kd (SC-Kd) molecule of 48 kDa was synthesized and gly- cosylated in insect cells and could be purified in the absence of detergents by affinity chromatography using the anti-H-2Kd monoclonal antibody SF1.1.1.1. We tested the ability of SC-Kd to bind antigenic peptides using a direct binding assay based on photo- affinity labeling. The photoreactive derivative was prepared from the H-2Kd-restricted Plasmodium berghei circumsporozoite protein (P.b. CS) peptide 253-260 (YIPSAEKI), a probe that we had previously shown to the first two domains of H-2K' and the third domain and the 3' sequences of the Q,,," gene and its introduction into mouse L cells has been described (Luescher et a/., 1992). Culture medium conditioned by clone NR37.15.2. was used to purify the K'Q,,, protein by immunoaffinity on an SF 1.1.1.1 column as described above.

nously derived antigenic peptides to CD8-positive cytotoxic T lymphocytes (Townsend and Bodmer, 1989;Rothbard and Gefter, 1991). The general features of class I molecule biosynthesis have been known for some time both in vitro (Dobberstein et al., 1979) and in vivo (Krangel et al., 1979). N-Linked core glycosylation occurs during biosynthesis of MHC class I heavy chains in the endoplasmic reticulum (see Kornfeld and Kornfeld (1985) for review), and after &-m association (Sege et al., 1981), the heterodimer is transported along the secretory pathway to the cell surface (Owen et al., 1980). The mechanism by which endogenous antigenic peptides associate with class I molecules is still incompletely understood, but current evidence suggests that it takes place in the endoplasmic reticulum (Nutchern et al., 1989;Yewdell and Bennink, 1989;Kvist and Hammann, 1990).
The three-dimensional structure of HLA-A2 has revealed that the a1 and a2 domains of the heavy chain fold in a manner constituting an extended, groove-like peptide binding site (Bjorkman et al., 1987a(Bjorkman et al., , 1987b. Class I molecules on cell surfaces contain short peptides (8-10 amino acids) with distinct allele-specific binding motifs (Falk et al., 1991;Jardetzky et al., 1991).The binding of antigenic peptides to MHC class I molecules is closely interrelated to the binding of Pz-m to the heavy chain. This is explained, on the one hand, by the intimate interactions of P2-m not only with the a3 domain but also with the a1 and a2 domains, which is likely to be important for the conformation of the peptide binding site (Saper et al., 1991). On the other hand, stable assembly of the class I heterodimer depends on occupancy of the heavy chain peptide binding site . Indeed, this hypothesis is supported by recent work with class I assemblydefective mutant cell lines   Townsend et al., 1990) and with purified human class I molecules (Silver et al., 1991), which has shown that suitable antigenic peptides promote class I molecule assembly.
Furthermore, there is evidence that peptide epitopes induce a conformational change of the heavy chain, resulting in increased affinity for Pz-m (Elliot et al., 1991), and that free heavy chains present on the cell surface are functionally inactive (Rock et al., 1991a(Rock et al., , 1991b.
Clearly, further understanding of these interactions will depend on the availability of reconstituted systems using purified molecules. In this respect, the insect-baculovirus expression system has proved particularly useful in recent years for producing, in relatively large quantities, recombinant proteins that are authentically folded, proteolytically processed, post-translationally modified, and biologically active.

24223
In this system, the foreign gene is cloned downstream of a strong very late promoter and introduced via homologous recombination back into the genome of Autographa californica nuclear polyhedrosis virus (AcNPV) (Miller, 1988;Luckow and Summers, 1989;Maeda, 1989). We have previously used this approach to produce biologically active mouse &ma (Dargemont et al., 1989;Godeau et al., 1991) and a H-2Kd heavy chain which is able to associate with &m in an antigenic peptide-dependent manner and form a functional heterodimer able to bind peptides (Godeau et al., 1992). However, in doubly infected cells, the extent of association between the heavy and light chain is low and hence precludes isolation of larger amounts of assembled molecules. To circumvent this difficulty, we chose to express a fusion protein comprising the first three domains of H-2Kd heavy chain fused to mouse p2microglobulin. We describe here some characteristics of this purified soluble protein and demonstrate that its interaction with antigenic peptides is identical to that observed with natural H-2Kd molecules.

MATERIALS AND METHODS
Cell Culture and Transfection-Suspension or monolayer cultures of the Sf9 cell line from Spodoptera frugiperda were maintained a t 27 "C in TC 100 medium supplemented with 0.33% Yeastolate (Difco) and containing 4% heat-inactivated fetal bovine serum, 50 pg/ml penicillin, and 50 pg/ml streptomycin. Cell densities were routinely kept between 2.5 X lo5 and 1.5 X 106/ml by passaging the cells in fresh medium. DNAs from recombinant plasmids (20 pg) and viral DNA (10 pg) were co-transfected by electroporation using a standard cuvette with a 0.4-cm gap containing 1 ml of Sf9 cell suspension (0.3-

X lo7 cells/ml) in BNP (25 mM 2-[bis(2-hydroxyethyl)amino]
ethanesulfonic acid, pH 6.95, 140 mM NaC1, 1.5 mM NaH2P04, and 1 mM glucose) in the presence of 500 pg/ml herring sperm DNA and exposed to the exponential discharge of a 1030 pF capacitance charged under 220 V (Godeau, 1990;Godeau et al., 1992). For large scale recombinant protein preparations, viral stock was prepared and used to infect a t a multiplicity of infection >lo. Sf9 cells were grown in aerated spinner flasks and fed, after infection, with serum-free TC 100 medium. Seventy-two hours postinfection, the suspension was collected and infected cells were removed by centrifugation.
Isolation of AcNPV SC-H-2Kd Recombinant Baculouirus-All plasmids were constructed and purified according to standard recombinant DNA techniques (Sambrook et al., 1989) except that transformation of Escherichia coli hosts was performed by electroporation (Dover et al., 1988). Conditions for restriction endonuclease digestions were those suggested by the manufacturer. DNA fragments were isolated from agarose gels by electroelution. The transfer vector used in this study was the nonfusion pAcYMl (Matsuura et al., 1987), kindly provided by Dr. D. Bishop, which was cleaved by digestion with BamHI restriction endonuclease and filled in with E. coli DNA polymerase large fragment. The structure of SC-Kd chimeric cDNA has already been described (Mottez et al., 1991). Briefly, the entire coding sequence comprises the four H-2Kd exons encoding H-2Kd signal sequence, al, a2, and a3 domains fused to the mouse Ps-rn' with the sequence: 5'GGG GGG ATC GGA TCC GGA GGC GGT coding sequence lacking its signal sequence via a 45-base pair linker GGA TCC GGT GGC GGC GGT TCG3'. This DNA sequence, which encodes a 15-amino acid protein sequence predominantly composed of glycine and serine residues, was chosen for its expected flexibility and is represented in Fig. 1. To generate a recombinant baculovirus, SC-Kd chimeric cDNA was excised from pSC-Kd-15 (Mottez et al., 1991) by digestion with Hind111 restriction endonuclease, filled in with E. coli DNA polymerase large fragment, and subsequently bluntend ligated into the filled in BamHI site of pAcYMl transfer vector. Transformants were selected for proper orientation with respect to the polyhedrin gene transcriptional start site. Plasmid DNA was isolated and transfected together with viral DNA by electroporation, as already described (Chu et al., 1987;Godeau, 1990). Five to seven days after transfection, the viral progeny containing a mixture of wild type and recombinant viruses was submitted to a purification procedure based on limiting dilution in 96-well microtiter plates followed by hybridization to a H-2Kd probe. Five to seven days after infection, the virus-containing, adhering cell monolayer was directly dissolved in 200 p1 of 0.4 M NaOH and transferred to a nylon filter (Gene-Screen Plus, Du Pont) via a dot-blot manifold (Reed and Mann, 1985). The filter was hybridized (Church and Gilbert, 1984) to a 32Plabeled (Feinberg and Vogelstein, 1983) H-2Kd probe, and the recombinant virus-containing medium from positive wells was further purified by a subsequent repeat of this dilution-hybridization procedure leading to the isolation of AcNPV-SC-Kd. The viruses encoding murine P2-mn, AcNPV-&-m (Godeau et al., 1991), and H-2Kd heavy chain, AcNPV-Kd (Godeau et al., 1992), have already been described.
Purification of SC-Kd-Immunoaffinity matrices were prepared by using anti-H-2Kd monoclonal antibodies SF1.l.l.l from ATCC and 34-1-2 (Ozato et al., 1980). Ten mg of pure IgG2. mAb were covalently coupled to protein A-Sepharose (Pharmacia) via dimethylpimelimidate following the procedure of Schneider et al. (1982). After saturating unreacted amino groups with ethanolamine and washing the column extensivsly, 1 liter of medium conditioned by AcNPV-SC-Kd-infected Sf9 cells was passed over the column during a 6-h period, after which the column was washed with 20 column volumes of phosphate-buffered saline. Bound material was eluted with 50 mM diethylamine, pH 11.5, and the material was neutralized with 200 mM triethanolamine/HCl, pH 7.5, and concentrated. For affinity chromatography with concanavalin A, the cell-free conditioned medium was first ultracentrifuged (100,000 X g for 1 h), concentrated to onetenth of the original volume, and dialyzed against 10 mM Tris, pH 7.5, 100 mM NaCl. The dialysate was passed through an immobilized concanavalin A column (IBF, France), eluted with a-methylmannoside, and analyzed by immunoblot using an anti-bz-m antiserum.
Biochemical Characterization of Purified SC-Kd-N-terminal sequence analysis of purified soluble SC-Kd from insect cells was performed by automated Edman degradation after electrophoresis on SDS-PAGE and transfer to PVDF, as described (Matsudaira 1987), on an Applied Biosystems 470 gas-phase peptide sequanator. Phenylthiohydantoin amino acids were detected with an on-line Applied Biosystems 120 A analyzer. For internal peptide sequence determination, the purified protein was reduced and alkylated, transfered on PVDF membranes as described above and digested in 100 mM Tris-HCl, pH 8.5, with 1 pg of porcine trypsin (Sigma) in a 1:50 enzyme to substrate ratio for 4 h a t 37 "C. The digest was separated by HPLC on a C18 Vydac column (250 X 2.1 mm) using a 0-60% acetonitrile gradient in 0.1% trifluoroacetic acid. A well resolved peak was chosen for sequence determination.
Antibodies and Immunoblots-The class I H-2d-specific monoclonal antibodies used were 34-1-2 (IgG 2a) (anti H-2Kd and H-2Dd) (Ozato et al., 1980) and SF1.l.l.l (IgG 2a) from ATCC. All antibodies were purified from ascites fluid according to established procedures (Ey et al., 1978;Harlow and Lane, 1988). The rabbit serum raised against a fusion protein comprising the second half of the a2 and all of the a3 domain of H-2Kd will be described,* and the rabbit antiserum antimouse &m was from Dr. N. Tanigaki (Natori et al., 1976).
Peptide Synthesis-Peptides were synthesized using an Applied Biosystems model 430A peptide synthesizer and the standard tbutoxycarbonyl strategy (Merrifield, 1986). The deprotected peptides were purified on a C-18 reverse phase HPLC column, dissolved in phosphate-buffered saline, and stored frozen.

Photoaffinity Labeling with the ["'I]IASA-YIPSEAK(Biotin)l
Peptide Probe-The photoprobe was prepared by reacting freshly iodinated [1251]IASA-ONSu with the orthogonal biotinylated octapeptide from mouse malaria Plasmodium berghei circumsporozoite protein YIPSAEKI(biotin)I and purified by C-18 reversed phase HPLC as previously described  (1 X lo7 cpm) for 6 h at 27 "C, and UV-irradiated at 4 "C for 10 min (15 watt lamp with an emission maximum a t 312 nm). After labeling, F. Godeau and H. Ploegh, manuscript in preparation. the cells were washed twice in cold medium and once in phosphatebuffered saline, detergent solubilized in 0.7% Nonidet P-40, and immunoprecipitated using SF1.l.l.l. mAb. After 1 h the lysate was centrifuged (15,000 X g for 3 min), and the supernatant was filtered through disposahle membrane filters (Millipore Nihon Kogyo, Yonezawa, Japan). Immunoprecipitation was performed on nitrocellulose membrane filters in a dot-hlot manifold as described (Luescher, 1987). For each immunoprecipitation, 30 pg of pure SFl.I.l.l mAh was used. The immunoprecipitates were subjected to SDS-PAGE (10% acrylamide) under reducing conditions and the dried gels were exposed to XAR 5 (Kodak) films.
Photoaffinity labeling of purified SC-K' protein was performed using a similar procedure. One pg of purified protein was incubated with ["'IIIASA-YIPSEAK(biotin)I (1 X 10' cpm) in 10 pl of phosphate-huffered saline. Aliquots were incubated in polypropylene tubes (500 ml Eppendorf, Hamburg, Germany) a t 24 "C for 3 h. The tubes were UV-irradiated in the reflector of the described UV lamp for 3 min. The samples were diluted with 450 pl of phosphate-buffered saline containing 0.5% Nonidet P-40 and the labeled protein was immunoprecipitated as described above.
For competition experiments, 1 pg of purified protein was incubated with the photoprobe in the absence or presence of graded amounts of competitor peptides in :<-fold dilution steps up to 1000-fold molar excess with respect to the photoprobe. The competitor was added 15 min before addition of the photoprobe.
The K'Q,,, plasmid construct leading to the synthesis of a soluble chimeric heavy chain between the first two domains of H-2K' and the third domain and the 3' sequences of the Q,,," gene and its introduction into mouse L cells has been described . Culture medium conditioned by clone NR37.15.2. was used to purify the K'Q,,, protein by immunoaffinity on an SF 1.1.1.1 column as described above. fusion protein after transfection in COS-1 cells (Mottez et al., 1991). To express this molecule in the higher amounts required for biochemical studies, we chose the recombinant baculovirus system. Thus, the use of the selection procedure described under "Materials and Methods" led to the isolation of a recombinant baculovirus containing the SC-K' coding sequence under the transcriptional control of the polyhedrin promoter (AcNPV-SC-K"). Preliminary experiments showed that a 48-kDa polypeptide (the expected apparent molecular mass for SC-K' if one assumes that the two N-glycosylation sites are used) was present in the supernatant of metabolically labeled insect cells infected with this virus. This band could be immunoprecipitated by an anti-mouse P2-m antiserum as well as by the 34-1-2 mAb recognizing the native conformation of the H-2K' heavy chain (Godeau et al., 1992) (results not shown). Taken together, these results indicate that the SC-K' polypeptide is exported as a soluble protein in insect cells. In this fusion protein, the first three domains of the H-2K' heavy chain associate with mouse &-m in a manner that mimics the native conformation of the heterodimer.

Biochemical
To purify this soluble class I fusion protein, immunoaffinity chromatography with an immobilized SF1.l.l.l. mAb column was carried out using crude cell-free conditioned medium from infected cells as starting material. Whereas no prominent Coomassie Blue stainable band could be detected in an SDS-PAGE analysis of the crude supernatant, the eluate from the affinity column showed a single major band of approximately 48 kDa (Fig. 2). This procedure routinely yielded approximately 500 pg of SC-Kd per liter of conditioned medium. Immunoblot analysis was carried out to further characterize the purified protein. In addition to reacting with the H-2K'specific rabbit antiserum raised against a fusion protein containing half of the a2 and the entire a3 domain of H-2K' (Fig.  3, panel A, lane a), this polypeptide also reacted with the antimouse P2-m antibody (Fig. 3, panel A, lane b ) , indicating that the 48-kDa polypeptide contained both MHC Class I heavy and light chain sequences. We then analyzed the glycosylation of this molecule. Conditioned medium from serum-free cultures was thus passed over a concanavalin A column, and the a-methylmannoside eluate was analyzed by immunoblot using Sfl.l.1. mAb immunoaffinity column, eluted at pH 11.5. and freeze-dried. Five pg of protein were loaded on an SDS-PAGE which was stained with Coomassie Blue. Lane n, eluate lrom the immunoaffinity column; /rrnP b, molecular weight standards. the anti-&m antiserum.
As shown in Fig. 3, panel R, a prominent 48-kDa and &m immunoreactive band could be visualized in medium conditioned by AcNPV-SC-Kd-infected cells which was absent from that of the wild type control, demonstrating that SC-Kd contains mannose-rich carbohydrates.
To further confirm the identity of the purified protein, the 48-kDa polypeptide was sequenced following transfer onto PVDF membrane (thus further purifying the input protein). As shown in Table I, N-terminal sequencing yielded the sequence Gly-Pro-His, the expected N-terminal sequence of the mature H-2K' protein. This proves that the signal sequence of this polypeptide was cleaved in this expression system, as expected for a soluble protein recovered from a cell-free supernatant. Internal sequences were also obtained after tryptic digestion and separation of the digest by HPLC. One well separated peak was sequenced and shown to contain two distinct sequences in molar ratios differing by no more than a factor of 2. The first sequence, Tyr-Tyr-(Asp)-Gln, belongs to the H-2K' heavy chain and encompasses an Nglycosylation site. As expected for a glycosylated protein, the Asn residue is lacking in this sequence and replaced by a minute quantity of Asp. The second sequence, Ile-Pro-Lys, belongs to the mouse &m coding sequence, further indicating that the P2-m polypeptide was present in the SC-K" 48-kDa polypeptide. In view of the purity of the 48-kDa protein, no other peak was sequenced. Thus, among the three sequences obtained, all of them were derived from the expected sequences. Taken together, these results demonstrate that a soluble, 48-kDa, glycosylated SC-Kd fusion protein with a correct N terminus is expressed efficiently in the baculovirusinsect cell system and can be isolated by a simple one-step procedure. In this polypeptide, the first three domains of H- 2K' are physically linked to mouse Pr-m and both sequences seem to fold to achieve a native conformation.
The Binding of IASA-YIPSAEKfiiotin)I to SC-Kd Is Not Affected by Soluble p2-m-We then determined whether the single-chain class I molecule retained the ability of the parental molecule to bind antigenic peptide ligands. We thus used the recently described H-2Kd-specific photoprobe derived from the octapeptide 253-260 of the mouse malaria P. berghei circumsporozoite protein IASA-YIPSAEK(biotin)I. In as much as photo-cross-linking is several orders of magnitude faster than the interaction of peptides with class I molecules (De Graff et al., 1974;Luescher et al., 1992), it directly reflects the binding of the peptide to class I molecules or any other cell component at the time of UV irradiation. We had previously shown that the binding of this photoprobe to the H-2Kd produced in the baculovirus-insect cell system is dependent on the presence of &-m (Godeau et dl., 1992). We first examined the ability of the probe to label living insect cells infected with AcNPV-SC-Kd. Thus, insect cells infected with various recombinant viruses were incubated with a 3 nM concentration of [""IIIASA-YIPSAEK(biotin)I and UV-irradiated, and following detergent solubilization, the various Kd constructs were immunoprecipitated and analyzed by SDS-PAGE and autoradiography. As shown in Fig. 4 ( (Godeau et al., 1992).
We then wished to examine the binding properties of the SC-Kd protein by incubation of the purified protein with the photoprobe. As a control, we used the soluble class I molecule H-2K'Qlt,, purified in the same manner, in which the heavy and light chains are produced as individual polypeptides and are therefore noncovalently associated. As shown in Fig. 4, panel B , this molecule was able to bind the photoprobe (lune c), but maximal labeling intensity required the addition of an excess of exogenous human Pn-m (lune d). In contrast, purified SC-Kd bound the photoprobe equally well in the absence or presence of an excess of human &m. Taken together, these results indicate that the SC-Kd fusion protein can bind peptides in a manner similar to that of the parental class I heterodimer.

DISCUSSION
In the insect cell-baculovirus system, recombinant proteins often retain' their biological activity (Miller, 1988;Luckow and Summers, 1989;Maeda 1989) and are produced in considerable amounts, thus permitting their purification and detailed analysis. Indeed, our previous work using this system has provided evidence for a significant production of biologically active mouse &ma (Godeau et al., 1991) and of a functional H-2Kd heavy chain which was able to associate with &ma in a peptide-dependent manner, so as to give rise t o a functional heterodimer (Godeau et al., 1992). Since our ultimate goal is to study a reconstituted system with purified molecules, the lack of complete heterodimerization observed in doubly infected cells prompted us to seek a more efficient system to produce a soluble molecule. The only studies reporting the successful production of soluble class I molecules have used a truncated heavy chain associated with the substitution of the transmembrane domain of the Blob molecule, leading to the production of a soluble class I molecule upon association with &-m (Margulies et al., 1986;Schneck et al., 1989). We have used another approach consisting of the creation of a fusion molecule of a truncated heavy chain linked t o the light chain via a flexible 15-amino acid linker (Mottez et al., 1991). Although the transiently transfected COS-1 cells do not constitute an expression system suitable for preparative purposes, we have previously shown that the single-chain H-2Kd molecule produced in COS-1 cells bears some resemblance to the natural class I molecule (Mottez et al., 1991), based on the presence of a conformational epitope recognized by the 34-1-2 mAb, which we have shown to be conformationspecific (Godeau et al., 1992). However, the peptide binding assay (Bouillot et al., 1989) used to initially assess the functionality of the fusion protein was subsequently reported to be of degenerate allele specificity (Chen et al., 1990;Frelinger et al., 1990).
We have expressed the truncated mouse class I major histocompatibility glycoprotein H-2Kd in baculovirus-infected insect cells as a soluble, secreted fusion protein with mouse &-ma. Here we provide direct evidence that SC-Kd, produced in the baculovirus expression system, is able to bind H-2Kdrestricted peptide epitopes, since photoaffinity labeling of the purified soluble SC-Kd by a H-2Kd-restricted peptide derivative was specifically inhibited in the presence of antigenic peptides known to be presented by the H-2Kd molecule. The competitor activity exhibited by a given peptide in the direct binding assay correlated with the activity determined in functional competition experiments  and with the relatedness of its respective sequence to the allelespecific consensus motifs described by Falk et al. (1991). Other peptides presented in the context of other class I molecules competed weakly if at all with the photoprobe. Thus, the binding properties of this fusion protein seem qualitatively indistinguishible from those of the cell-associated H-2Kd molecule. Because we have shown that &-m is absolutely required for the binding of this ligand to the H-2Kd heavy chain under the present experimental conditions (Godeau et al., 1992), we can infer that the P2-m linked covalently to the heavy chain plays a role identical to that of native Pz-m in stabilizing the native conformation of the heavy chain. Furthermore, in contrast to the soluble version of H-2Kd (KdQlo), in which p2m is not covalently linked to the heavy chain, maximal peptide binding to the single-chain molecule can be achieved without addition of &-m, suggesting that the covalently linked murine Pz-m is effective at providing the heavy chain conformation necessary for peptide binding (Elliot et al., 1991).
Other heterodimeric molecules of immunological interest have been produced as fusion proteins between the two chains of the original heterodimer and expressed in E. coli mostly as insoluble molecules. After denaturation and renaturation, however, it has been possible to recover a single-chain antibody Fv fragment retaining some of the affinity of the original immunoglobulin for its antigen (Huston et al., 1988). Similarly, a variable fragment of a T-cell receptor, rendered soluble by site-directed mutagenesis, showed some antigen-combining properties after renaturation (Novotny et al., 1991). Thus, this molecular design can lead to the synthesis of functional heterodimers and could possibly be used for other heterodimeric molecules.
As demonstrated in the present study, the baculovirusinsect cell system yields directly soluble molecules which are functional. Interestingly, the human class I1 MHC protein HLA-DR1 has recently been expressed in the baculovirusinsect cell system as a soluble heterodimer produced in an "empty" conformation able to bind peptide antigens (Stern and Wiley, 1992). It would be of interest to determine whether the soluble class I molecule herein described is also empty.
Peptide binding and kinetic studies are presently in progress to settle this issue.
In conclusion, expression of the mouse class I histocompatibility antigen H-2Kd as a fusion protein in insect cells is a convenient source of soluble mouse MHC class I heterodimer, which behaves as the native protein and should be particularly useful for peptide binding studies aimed a t a biochemical and biophysical characterization of peptide-class I molecule complex formation.