Efficient generation of major histocompatibility complex class I-peptide complexes using synthetic peptide libraries.

The use of synthetic random peptide libraries is a powerful technology for the study of many aspects of antigen presentation and peptide selection by major histocompatibility complex (MHC) molecules. Here we have used them in conjunction with a recombinant system to determine the peptide binding motifs of three classical class I MHC molecules of the laboratory rat: RT1-Aa, RT1-Au, and RT1-A1c. Described is a method for producing large amounts of soluble class I heavy and light chains in bacteria. Refolding RT1-Aa heavy chain (HC) with rat beta2-microglobulin (beta2m) in the presence of a specific peptide and the subsequent purification of the complex yielded conformationally correct material. This was assessed by gel chromatography, SDS-polyacrylamide gel electrophoresis, isoelectric focussing gel electrophoresis, enzyme-linked immunosorbent assay, and fluorescence-activated cell sorter analysis employing a previously unreported method utilizing a His-Tag affinity silica. By refolding RT1-Aa HC and rat beta2m around a random nonapeptide library and subjecting the resulting complex to acid elution of the bound peptides and pool sequencing, the peptide binding motif for this MHC class I molecule was determined. Results corresponded well with those previously determined from naturally bound peptides and in addition gave a clear and unambiguous signal for the C-terminal anchor residue. This method was then applied to determine the previously undescribed binding motifs for RT1-Au and RT1-A1c. For both molecules, the whole motif was confirmed from naturally bound peptides. We propose this method as an alternative way to obtain the whole class I MHC peptide motif, particularly when a specific antibody is unavailable and/or natural expression of the class I molecule of interest is low.

Antigenic peptides generated principally by the proteasome are transported into the endoplasmic reticulum (ER) 1 via the heterodimeric transporter associated with antigen presentation (TAP) (1,2) where they associate with newly synthesized major histocompatibility complex (MHC) class I molecules. In the rat, the TAP2 subunit of the transporter exists as two main allotypes (TAP2-A and TAP2-B) (3,4), and it appears that class I MHC alleles, encoded by the classical class Ia locus RT1-A, are genetically associated with only one form of the transporter (4). In recent years, it has been shown that the two forms of the transporter display different peptide selectivities, such that TAP-A binds and transports peptides with many different amino acids present at the C terminus while TAP-B generally allows only hydrophobic and aromatic residues in that position (5,6). In addition, the TAP-B form of the transporter has been reported to be able to accept longer peptides than the usual 8 -11 amino acid lengths (7) although this is disputed (8). Despite extensive studies on TAP peptide selectivities, few peptide binding motifs for rat class Ia molecules (RT1-A) have been reported. Thus far, the motifs of only two TAP-A associated molecules, RT1-A l (9, 10) and RT1-A a (11), have been determined.
The main method used to determine a peptide binding motif for an MHC class I molecule, including those rat alleles already reported, is the pool sequencing method of Falk et al. (12), that consists of acid eluting peptides from a natural source of class I, separating the heavy and light chains from the peptides by reverse phase HPLC, pooling the predetermined peptide containing fractions, and subjecting the pool to automated sequence analysis. This method has proven extremely useful for obtaining peptide binding data for class I from different species. However, for its success, the method relies on a specific antibody for the purification of the class I under study, but this is not always available. Antibodies can often have cross-reactivities with other known or unknown class Ia or class Ib molecules (13). If expression of the molecule under study is low, then cloning and expression in a suitable cell line is needed. In addition, the C-terminal anchor residue is not always clearly defined by this sequencing method since class I alleles can bind peptides of varying lengths, thus spreading the C-terminal signal over several sequencing cycles (14 -16).
An alternative method called positional scanning (17) has been shown to be extremely useful within the field of immunology (18 -20). Synthetic random peptide libraries coupled to single defined amino acids at set positions within the peptide have been used successfully to probe the effect of the * This work was supported by a Biotechnology and Biological Sciences Research Council Project grant (to E. J.), Biotechnology and Biological Sciences Research Council Competitive Strategic Grant funding (to G. W. B.), and funding by the Deutsche Forschungsgemeinschaft (SFB 323) (to K. H. W.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Here, we have combined the above methods to determine peptide binding motifs for class I MHC molecules. We report the use of a bacterial system to produce recombinant protein of three truncated rat class Ia MHC molecules and rat ␤ 2 -microglobulin (␤ 2 m). We have successfully refolded these rat class I heavy chains in the presence of rat ␤ 2 m and a random nonapeptide library and have successfully determined the binding motifs for all three molecules under study.
Bacteria were grown in LB containing 100 g/ml ampicillin. Protein expression was induced at mid-log phase for 3 h with 1 mM isopropyl-␤-D-thiogalactopyranoside. Heavy chain and ␤ 2 m proteins were found to be overexpressed as inclusion bodies. Bacterial pellets were resuspended in 10 mM Tris-HCl, pH 7.5, 1 mM EDTA, 100 g/ml PMSF, 0.1% (v/v) Triton X-100, disrupted by sonication, and centrifuged (25,000 ϫ g). The cell pellets were washed twice with resuspension buffer and then solubilized by resuspending and mixing for 1 h at room temperature in 8 M urea, 0.1 M NaH 2 PO 4 , 0.01 M Tris-HCl (buffer A), pH 8.0. The mixtures were then centrifuged (25,000 ϫ g) to remove any insoluble material. Refolding of urea-solubilized ␤ 2 m was achieved by extensive dialysis against 10 mM Tris-HCl, pH 7.5, at room temperature. The dialyzed mixture was centrifuged and any precipitate formed was resolubilized and redialyzed. Solutions were pooled and applied to a Q-Sepharose column (Pharmacia Biotech Inc.) equilibrated with 10 mM Tris-HCl, pH 7.5. Pure ␤ 2 m (Ͼ95% as determined by SDS-PAGE; results not shown) eluted from this column with 10 mM Tris-HCl, 50 mM NaCl, pH 7.5. Urea-solubilized heavy chain was purified in a denatured state using an Ni-NTA agarose column (Qiagen Inc., CA) according to the manufacturer instructions. Briefly, the urea-solubilized protein was mixed with the nickel-charged matrix for 1 h. The matrix was washed with buffer A at pH 8.0, followed by buffer A at pH 6.3. Pure protein (Ͼ95% as determined by SDS-PAGE; see Fig. 4) was eluted from the column by washing with buffer A at pH 5.9 and buffer A at pH 4.5. Protein concentration was determined by the BCA assay (Pierce).
DNA/Protein Sequencing-Samples for DNA sequencing were submitted to the sequencing service at the Microchemical Facility, The Babraham Institute. Purified peptides were applied to polybrene-coated glass fiber filters and sequenced by Edman degradation using an ABI model 492 sequencer with 610A data analysis software.
Peptide Synthesis-The specific 13-mer peptide, ILFPSSERLISNR (27), was synthesized by Alta Biosciences, The University of Birmingham, UK. The synthetic nonapeptide library was prepared by fully automated solid phase peptide synthesis using Fmoc/tert-butyl chemistry. For coupling of amino acids in randomized sequence positions (X), double couplings were performed with diisopropylcarbodiimide/butanol and an equimolar mixture of Fmoc-L-amino acids and Fmoc-Gly (cys-teine was excluded), used in an equimolar ratio with respect to the coupling sites of the resins. The peptide mixture was cleaved from the resins and side chain deprotected with trifluoroacetic acid:phenol:EDT: thioanisole (96:2:1:2, v/w/v/v) and filtered from the resin. The products were precipitated at Ϫ20°C by the addition of cold n-heptane:diethylether (1:1, v/v), washed twice by sonification, and lyophilized from acetic acid:water:t-butyl alcohol (1:10:50, v/v/v), yielding the completely randomized nonapeptide library. The amino acid composition of the peptide library was determined by pool sequencing (28), electrospray mass spectrometry (29), and amino acid analysis. Deviations from equimolar representation of the amino acids were found to be within the error limits of the analytical methods (30).
Refolding Experiments-Refolding was performed by the dilution method (31,32). Briefly, 360 g of ␤ 2 m and 150 g of peptide were added to 15 ml of 50 mM Tris-HCl, pH 8.0, 400 mM arginine, 0.1 mM EDTA, 0.1 mM PMSF. Denatured heavy chain (465 g) was added to give a final molar ratio of 1:2:10 (heavy chain:␤ 2 m:peptide). For RT1-A u and RT1-A1 c , the peptide concentration was increased to give a final ratio of 1:2:100. (For RT1-A1 c , the results presented here were from an experiment using three times the amounts shown above for all three components). After 24 -48 h at 20°C, the refolding mixture was concentrated down to 100 l using a Centriprep-10 and Centricon-10 (Amicon). Refolded class I complex was purified by gel filtration on an FPLC Superdex 75 column (Pharmacia) equilibrated in 20 mM Tris-HCl, pH 8.0, 100 mM NaCl.
Peptides were eluted by first concentrating the purified complex to 50 l and then incubating the complex with 2 ml of 10% acetic acid for 5 min. This was then passed through a Centricon-3 (Amicon), and peptides were separated by reverse phase chromatography using an Applied Biosystems Aquapore (250 ϫ 1 mm) Brownlee C 18 column with an acetonitrile gradient (0 -90%) in 0.025% trifluoroacetic acid. Peptides eluted between 6.75% and 40.5% acetonitrile were pooled, concentrated, and submitted for sequencing. As described (11,12,33), a pmol yield Ͼ150% of the previous cycle was considered significant and is presented as boxed values in the results (Figs. 1-3). For cycles 7-10, this cut-off value was reduced to 120% to allow for cycle-to-cycle sample loss. For RT1-A1 c , because of the increased scale and amino acid yield, the cut-off values used for the random peptide refolding were increased to 300% for cycles 1-6 and 150% for cycles 7-10. Residues at cycle 1 were difficult to assign as anchors, mainly due to having no previous cycle with which to compare and that the first cycle is often prone to amino acid contamination. Amino acids presented at cycle 1 are the 3-4 amino acids with the greatest yield (glycine was ignored due to contamination).
Conformational Antibody Experiments-Successful refolding was determined by ELISA using a conformational RT1-A a specific antibody (MAC 30, Ref. 34), a general class I reactive antibody (MRC-OX18, Ref. 25), and a ␤ 2 m specific antibody (C9, Ref. 26). The complex to be tested was coated onto plates overnight at 4°C before addition of antibodies. A horseradish peroxidase-secondary antibody was used, and the plates were developed with TMB and read at 450 nm.
In addition to ELISA, FACS analysis of Ni-NTA silica beads (Qiagen Inc., CA) coated with refolded complexes was used. For coating, 40 g of the refolded complex was incubated with 50 g of silica beads preequilibrated in PBS, and the mixture was incubated at room temperature for 2 h with gentle mixing to allow the complex to associate to the matrix via the ␣-chains C-terminal His-Tag. After washing any excess protein away with PBS, 0.1% (v/v) Tween 20, the matrix was incubated with MAC 30, followed by a fluorescein isothiocyanate-coupled anti-rat-IgG antibody (DAKO A/S, Denmark) at a 1/100 dilution in PBS. After extensive washing with PBS, the matrix was subjected to FACS analysis using a FACSCalibur TM (Becton Dickinson).
For both cell types, the cell suspension was centrifuged and the pellets were washed with buffer A (10 mM Tris, pH 7.7, 150 mM NaCl). The cell pellets were resuspended in 100 ml of buffer A before addition of 100 ml of buffer B (10 mM Tris-HCl, pH 7.7, 150 mM NaCl, 2% Nonidet P-40, 1 mM PMSF). The suspension was mixed at room temperature for 30 min and then spun at 20,000 ϫ g for 30 min at 4°C. The supernatants were decanted into fresh bottles, and 3 ml of non-immune rat serum-Sepharose was added. This was mixed for 1 h at 4°C before the matrices were removed by centrifugation/filtration. U9F4-Sepharose and YR5/310-Sepharose (1.5 ml for both) were then added to their respective lysates and mixed at 4°C for a further 90 min. The Sepharose was collected and washed thoroughly with buffer B followed by buffer A. Peptides were eluted by incubating the antibody-coupled Sepharose with 15 ml of 10% acetic acid for 5 min at room temperature. After filtration, the solutions were then passed through Centriprep-3 units and concentrated, and the peptides were separated by reverse phase chromatography using the same method described above for the peptides released from the refolded complex. The cut-off values used to determine binding motifs were a 300% increase over the previous cycle for cycles 1-6 and 150% for cycles 7-10.

RESULTS
Expression and Purification-Using PCR, we have engineered expression plasmids to produce three different soluble rat class I RT1-A molecules in E. coli, RT1-A a , RT1-A u , and RT1-A1 c . Under isopropyl-␤-D-thiogalactopyranoside induction, both ␤ 2 m and the heavy chains were expressed as inclusion bodies and therefore required urea solubilization before further purification could be performed. Induction of the heavy  Fig. 1. B, RT1.A u complexes were isolated by immunoprecipitation using a U9F4-Sepharose matrix, and their bound peptides were purified by acid elution and reverse phase chromatography. Pooled peptides were sequenced by Edman degradation. Anchor residues were determined as described under "Experimental Procedures" and are boxed while other increases are underlined.
chain constructs produced material consistent in size, by SDS-PAGE, with that expected for the truncated, 1-276-amino acid heavy chains coupled at their C-termini to an 8-amino acid stretch containing a His-Tag motif (LEHHHHHH) coded by the pET-22b(ϩ) plasmid. The resulting class I heavy chains were further purified on an Ni-NTA-agarose column to Ͼ95% purity  Fig. 1. B, RT1.A1 c complexes were isolated by immunoprecipitation using a YR5/310-Sepharose matrix, and their bound peptides were purified by acid elution and reverse phase chromatography. Pooled peptides were sequenced by Edman degradation. Anchor residues were determined as described under "Experimental Procedures" and are boxed while other increases are underlined. (Fig. 4). The yield of heavy chain was 10 -20 mg/liter of bacterial culture. This eluted off the Ni-NTA column as two species, one in pH 5.9 and the other in pH 4.5 buffer ( Fig. 4; last two lanes). SDS-PAGE in the presence/absence of reducing agent confirmed the mixture eluting in the pH 5.9 buffer to be mainly composed of monomers, while protein eluting in the pH 4.5 buffer comprised large molecular weight multimers (results not shown). Only the monomer fraction was used for refolding work (yield 5-10 mg/liter). Protein sequencing of the monomer fraction revealed that the bacteria had successfully cleaved the initiation methionine to give the expected N-termini (results not shown).
The ␤ 2 m construct was cloned into the plasmid with its own stop codon, thereby producing only the native protein (99 amino acids in length). Successful purification was achieved first by refolding the urea-solubilized material by dialysis against 10 mM Tris-HCl, pH 7.5, and subsequently by ion exchange chromatography of the soluble fraction. Yields could be increased by subjecting the insoluble fraction, produced during dialysis, to a further round of solubilization/dialysis. Ion exchange chromatography of ␤ 2 m was performed on a Q-Sepharose (HiLoad 16/10) column. Protein (Ͼ80% purity) composed of ␤ 2 m aggregates eluted in 10 mM Tris-HCl, pH 7.5, while the monomer fraction (Ͼ95% purity) appeared in the 50 mM NaCl wash (SDS-PAGE results not shown). The yield of ␤ 2 m was 40 -50 mg/liter of bacterial culture of which 20 -25 mg/liter were monomers. Only the monomer fraction was used for refolding work. Protein sequencing revealed the expected N terminus (MIQKT-) but the bacteria had only deformylated and not cleaved the initiation formyl-methionine (results not shown). This is probably due to inhibition of bacterial methionine aminopeptidase by the adjacent isoleucine (first residue of ␤ 2 m), as discussed by Hirel et al. (36). The methionine, however, should not interfere with subsequent refolding experiments since, in all the reported structures of class I molecules, position 1 is away from the binding groove and the ␣-chain/␤ 2 m interface (37).
Specific Peptide Refolding-Refolding tests were initially performed using the RT1-A a heavy chain because, thus far, this is the only rat class Ia molecule for which a specific peptide has been identified and reported to bind (27). The peptide used for refolding was ILFPSSERLISNR, a 13-mer peptide from the rat mitochondrial A chain of ATP synthase (SwissProt accession number P05504). Results from a typical refolding experiment, using a 10-fold excess of peptide over the heavy chain, are shown in Fig. 5A. Gel filtration of the refolding mixture through a Superdex 75 column gave three peaks eluting at Ͼ100 kDa (Peak A), ϳ40 kDa (Peak B), and 12 kDa (Peak C). These corresponded to aggregates, refolded complex, and ␤ 2 m, respectively. Peak B increased in the presence of peptide, and SDS-PAGE analysis showed that there were both heavy chain and ␤ 2 m bands present (Fig. 5B). When refolding is performed without peptide, the observed peak at ϳ40 kDa could be an empty heavy chain/␤ 2 m complex, but this has yet to be characterized. Isoelectric focussing gel electrophoresis and a subsequent Western blot of the peptide/MHC complex revealed bands with an approximate pI of 6.7 for free ␤ 2 m (C9 positive) and with pI 5.5 for the heavy chain/␤ 2 m complex (MRC-OX18 and C9 positive), and at pI 5.1, two almost indistinguishable bands (MRC-OX18 positive) were revealed that could be heavy chain with/without peptide (results not shown).
To assess whether the purified complex was truly conformational, we performed ELISA assays and FACS analysis. For the ELISA, three antibodies were tested: MRC-OX18 (general antirat MHC class I), MAC 30 (anti-RT1-A a ), and C9 (anti-rat ␤ 2 m). Fig. 6A shows the results of the ELISA. MRC-OX18 was positive for all forms where the heavy chain was present, consistent with its broad specificity. The reactivity of MAC 30 increased with the addition of peptide and was positive in the purified complex, which was also positive for C9 (anti-␤ 2 m).
FACS analysis with MAC 30 on complex bound to Ni-NTA silica recorded an increase in the mean fluorescence (Fig. 6B), further suggesting the purified complex was in its correct conformation. Finally, acid elution of the peptide bound in the complex and sequencing of the eluted material, Ͻ3 kDa, gave the expected sequence of the 13-mer peptide.
Random Peptide Refolding Experiments-Direct pool sequencing of the random nonapeptide library showed no significant enrichments for any amino acids at any position (results not shown). Therefore, any enrichment seen in the material eluted from the refolded RT1-A molecules was considered specific.
For RT1-A a , a 10-fold molar excess of random peptide:heavy chain was sufficient to give equivalent successful refolding compared with the specific 13-mer peptide (results not shown). Acid elution and subsequent pool sequencing of the released peptides yielded the results and motif presented in Fig. 1A. Dominant residues (boxed values) as determined previously (12,33) were seen at positions 2, 3, 4, 8, and 9. These results are comparable with the RT1-A a motif results already published by Powis et al. (11) and reproduced in Fig. 1B. It can be noted that, in their results obtained from natural isolates, the residues which increased at positions 3 and 4 (Phe and Pro, respectively) were not seen as large enough increases to be considered anchor residues although both are present in the natural mitochondrial peptide ligand.

RT1-A u Peptide Motif from Recombinant and Natural
Sources-Having obtained successful results for RT1-A a , this method was used to refold around the TAP2-B-associated class I molecules, RT1-A u and RT1-A1 c (see below). Initial experiments with a 10-fold excess of random peptide:heavy chain failed to yield sufficient refolded complex. Only when the peptide ratio was increased to 100-fold did the complex refold sufficiently (see "Discussion"), with the peak area being 41% of the total area of the aggregate and complex (Fig. 7). With the higher ratio of peptide to heavy chain, the peptide motif for RT1-A u was determined (Fig. 2A). RT1-A u revealed a motif with strong preferences for hydrophobic and uncharged residues. Dominant residues appeared at positions 2, 3, 7, and 8 and, in particular, showed a C-terminal preference for Tyr at position 9.
In addition, this RT1-A u motif was confirmed by performing an immunoprecipitation experiment from the C58 cell line using the RT1-A u specific monoclonal antibody, U9F4 (Fig. 2B). Results revealed dominant residues at positions 2, 3, 7, and 9, which were in agreement with the results from the in vitro refolding. Other positions in the motif showed some strong increases that were not present with the recombinant system. At position 4, there were strong increases for Asp and Arg, whereas an Ala was dominant at position 6.
RT1-A1 c Peptide Motif from Recombinant and Natural Sources-As for RT1-A u , successful refolding with recombinant RT1-A1 c was achieved only with a 100-fold excess of peptide over the heavy chain. The motif obtained from pool sequencing for recombinant RT1-A1 c is shown in Fig. 3A. Dominant residues appeared at positions 2, 3, and 8 and with an almost singular preference for proline at position 2. Interestingly, there was an observed increase in tyrosine at position 8 while no obvious C-terminal anchor residue was observed at position 9, unlike for the two other class I heavy chains that we have used in this study. The motif derived from the rat splenocytes ( Fig. 3B), revealed similar anchor residues to the recombinant system. The strong proline at position 2 is confirmed as is the preference for positive residues, particularly His and Arg at position 3. An increase for tyrosine at position 8 was observed, and in addition, a C-terminal (P9) anchor, leucine, was indicated.

DISCUSSION
In this study we have successfully used a random nonamer peptide library to determine the peptide binding motifs for RT1-A a , RT1-A u , and RT1-A1 c using a recombinant system. We describe a method to produce large amounts of rat MHC heavy and light chains. We have cloned, expressed, and purified truncated versions of three rat RT1-A class I heavy chains containing only the ␣1, ␣2, and ␣3 domains. Engineering a His-Tag at the C-terminal end of the heavy chains enabled a quick and simple purification strategy while leaving the protein denatured in 8 M urea and ready for refolding. Purification in the absence of reducing agents yielded both monomers and aggregates with interdisulfide bonds. The use of the Qiagen Ni-NTA matrix allowed for the enrichment of monomers from aggregates. Attempts to denature aggregates in the presence of reducing agents and to refold them subsequently were unsuccessful (results not shown). The ␤ 2 m was purified by urea solubilization, followed by renaturation by dialysis. Ion exchange chromatography with a Q-Sepharose column successfully purified monomers from aggregates.
In recent years, large scale protein expression of MHC molecules in bacteria has mainly been directed toward crystal production for structure determination, when a binding peptide is already known (31, 32, 38 -42). Here we present results showing that the same method can be adapted successfully to elucidate the binding motifs of rat RT1-A MHC molecules by using a random peptide library. Refolding with a fixed length random peptide library has a clear advantage over immunoprecipitation of MHC class I molecules expressed in mammalian cells. Unlike class II MHC molecules, peptides are usually locked into the groove of class I molecules at their N and C termini. In a natural system, peptides of different lengths can still bind to class I molecules (7, 14 -16), and when eluted and subjected to pool sequencing, the C-terminal residues are often "diluted" away because of the different lengths of the popula-tion. This should not often occur with a fixed length random peptide library although it is possible for a nonamer library to bind with an end extending out of the groove. In addition, immunoprecipitation requires a rich source of class I and a specific antibody to extract a homogeneous population of the class I under study. This is often unavailable, especially with the poorly expressed non-classical (class Ib) molecules. The cloning scheme described here, coupled to refolding around random peptides, could also produce conformationally correct material for raising anti-class I antibodies. Raising antibodies against a complex with a specific peptide bound within the groove of a specific class I is also possible with this system. Injection of a single peptide⅐MHC complex, rather than a cell presenting the MHC with the peptide as was done by Porgador et al. (43), should produce more specific antibodies.
Only the monomer fractions for both subunits were used during the refolding experiments presented here. This enrichment produced reproducible refolding of between 35 and 45% of the starting material, when using both the specific 13-mer peptide and the random peptide library. This value is higher Conformational experiments for RT1-A a refolding with a specific 13-mer peptide. A, an ELISA assay was performed using the antibodies described. There is an increase in MAC 30 reactivity (A a -specific) when peptide is present. The purified complex is both MAC 30 and C9 (anti-␤ 2 m) positive. B, the complex was further analyzed by FACS analysis by coupling the complex to Ni-NTA silica via its His-Tag and probing the matrix with MAC 30 and a fluorescein isothiocyanate-coupled rabbit anti-rat-IgG secondary antibody. than any previously reported for refolding with bacterial protein; 20% was the maximum yield reported by Reid et al. (32), who "pulsed" the refolding mixture with aliquots of heavy chain. We have yet to determine whether such pulsing could improve yields in our system. The purified complex was shown to contain both heavy and light chains, by SDS-PAGE and isoelectric focussing gel electrophoresis, and to have conformationally correct material, by ELISA and FACS. In this report, we have used Ni-NTA silica (Qiagen Inc.) for FACS analysis. The product, sold as a solid support for FPLC, HPLC, and microspin miniprep applications, has a bead size of 16 -24 m, which was small enough to be unhindered in the tubing of the FACSCalibur™ (Becton Dickinson). Binding via the His-Tag allows complete access of the antibodies to the native complex and produced an increase in fluorescence with the RT1-A a specific antibody, MAC 30. Such material could provide a replacement for TAP-deficient cell lines such as RMA-S and T2, which are routinely used for providing "empty" class I molecules.
Pool sequencing results for acid-eluted random nonamer peptides from RT1-A a , when compared with the motif reported by Powis et al. (11), show a comparable motif with identical amino acid preferences at positions 2, 8, and 9. In addition to these positions, our results suggest a strong preference for serine at position 1 and for phenylalanine and proline at positions 3 and 4, respectively. The latter two were both observed in the previous report but not as anchor residues. This discrepancy may be explained by reduced availability of such peptides in a natural system compared with the random library. Similarly, results for the TAP2-B-associated RT1-A u from both random peptide refolding and immunoprecipitation were comparable and revealed a general preference for aliphatic and aromatic residues at positions 2, 3, 6, and 7 and a strong preference for tyrosine at position 9 (even to the exclusion of Phe). As for RT1-A a , there was also a preference for serine at position 1 in addition to threonine, alanine, and glutamic acid. The major differences between the two systems occur at positions 3 and 4, which once again may be due to a more stringent peptide availability in the natural system. Fig. 8 shows a comparison of the hydrophobic residues in the theoretical structures of RT1-A a , RT1-A u , and RT1-A1 c . The circled region clearly shows an increase in hydrophobic residues for RT1-A u in the ␤-sheets at the bottom of the groove. This provides a likely explanation for the hydrophobic residues observed at positions 6 -8 of the immunoprecipitation motif (Fig. 2), which would fit into this area if placed in the groove.
In both the motifs for RT1-A1 c , we find a strong preference for proline at position 2. This anchor residue has already been reported in other class I molecules such as H-2L d (12), HLA-B7 (44), HLA-B*3501 (45), and HLA-B53 (45). It is of interest to note that the TAP/peptide translocation work of Neefjes et al. (46) found that a proline at position 2 or 3 of the peptide negatively influenced transport in both the rat TAP-A and TAP-B transporters (also observed with human TAP, Ref. 22). Our data do not necessarily contradict the TAP data since longer, Pro-containing peptides could be transported and amino-terminal trimming in the ER would make them suitable for binding to RT1-A1 c .
Other interesting observations from the recombinant RT1-A1 c are the significant increase in tyrosine at position 8 and surprisingly, no observed specific increase at the ninth position, whereas from the natural system (Fig. 3B), a leucine was observed at position 9. The C-terminal anchor residues for the other motifs reported, RT1-A a and RT1-A u , were determined FIG. 7. Peptide concentration dependence for successful RT1-A u refolding. RT1-A u was refolded around different concentrations of random peptide library. Gel filtration analysis revealed a different concentration dependence when compared with RT1-A a . A 100-fold excess of peptide is required for similar yields of refolding seen with only a 10-fold excess for RT1-A a .
accurately and unequivocally to be an arginine and tyrosine, respectively. Previous attempts to determine MHC peptide binding motifs using recombinant protein and a random peptide library, with mouse H-2K b from insect cells (47) and human HLA-A2, B8 and B53 from bacteria (48), were only partially successful as both were unable to report any of the C-terminal anchor residues. Interestingly, a C-terminal leucine appears in all of the motifs for the above examples (except HLA-B53 whose motif is yet to be completed). It may be possible that a C-terminal leucine is lost to a "washout" effect during the Edman degradation. Closer examination of Fig. 3A does show a hint of a leucine at cycle 9 since it shows only a small drop in signal compared with the other residues from the previous cycle. A larger scale prep may circumvent this problem.
RT1-A u and RT1-A1 c , therefore, display anchor motifs with strong preferences for aromatic and hydrophobic residues, respectively. The restrictive TAP-B transporter to which RT1-A u and RT1-A1 c are normally genetically linked is known to be able to transport these types of peptides (5, 6), so our results are consistent with the possibility that genetic associations of alleles at RT1-A and TAP2 may reflect the related specificities of transporter and presenting molecule (4). However, since the permissive TAP-A transporter can also transport such peptides, the allelic associations may be based on other unknown factors.
It is also interesting to note that, whereas only a 10-fold ratio of peptide:heavy chain was required for refolding with RT1-A a , a 100-fold ratio was required for RT1-A u and RT1-A1 c . Since pool sequencing of the untreated random peptide library showed no significant enrichments for any amino acids at any position, indicating a random population, there is an implication that both of these molecules have a more stringent requirement for particular peptides, possibly as the groove requires specific types of residues at more positions than does RT1-A a . Alternatively, the increased preference for hydrophobic and aromatic residues in RT1-A u raises the possibility of reduced solubility for the preferred peptides and therefore reduced availability for refolding. But RT1-A1 c also requires a 100-fold ratio for efficient refolding, and this has charged residue preferences at position 3. In vivo, before peptides can bind to class I molecules for presentation, at least two points have been identified where they are processed: the proteasome where the peptides are produced and the TAP transporter which transports peptides into the endoplasmic reticulum lumen. Therefore, peptide presentation in natural systems is subject not only to what peptides can bind to the class I but also to how proteins are processed and what peptides are able to be transported into the ER lumen. The method we have used here is free from the cellular processing and transport and as such, measures only the specificity of the groove of the class I molecule itself. Comparisons of the motifs from natural and recombinant sources for both RT1-A a and RT1-A u do indeed show some differences, suggesting that the proteolytic and TAP transporter machineries may be imposing significant restrictions on the availability of peptides including positions other than the C-terminal anchor. If this were the case, then the method we describe here could prove extremely valuable for identifying such restrictions.