A Novel Instance of Class I Modification (cim) Affecting Two of Three Rat Class I RT1-A Molecules Within One MHC Haplotype¹ Free

The number of genes for MHC class I molecules varies greatly between species: from ∼11 in pigs (1) to several thousand in the African pigmy mouse (2). Among the complement of class I genes in a species, some are designated classical class I, or MHC class Ia, genes. The protein products of the class Ia genes share certain characteristics. These include high cell surface expression levels relative to the levels of MHC class Ib genes, a major role in repertoire formation and Ag presentation for substantial numbers of α/β CD8⁺ T lymphocytes, the ability to provoke strong cellular CD8⁺ T cell responses in the context of allotransplantation, and a tendency toward a high degree of genetic polymorphism. MHC class I genes are found at several different positions within each MHC complex. A unifying description of mammalian MHC class I gene organization has been provided by Amadou (3), who pointed out that clusters of class I genes, varying in size from one species to another, are located at certain positions within a linear framework of conventional genes that maintains an evolutionarily conserved order.

The most detailed investigations of MHC class I genetics have involved studies of this system in humans and laboratory mice. Comparative studies in other species, however, have served to stress the remarkable plasticity of this system. The description of the MHC of the laboratory rat (Rattus norvegicus) is now well advanced (4). Close scrutiny of MHC class Ia gene expression in the rat has uncovered surprising differences from its fellow murine, the mouse. 1) Rat MHC class Ia genes are all located to one side (centromeric) of the MHC class II region, i.e., there is no evidence for an H2-D or -L equivalent in the rat (the rat class Ia region is called RT1-A). 2) The number of MHC class Ia genes per MHC haplotype is a variable (5, 6). Similar observations have been reported for cattle (7). 3) The TAP peptide transporter, which supplies peptides for MHC class I assembly, is functionally dimorphic in the rat (8, 9, 10). The rat MHC class Ia system has evolved alleles that apparently take advantage of the broader peptide specificity of the rat TAP-A transporter allele, while the mouse class Ia alleles reflect a more restrictive TAP, which has properties similar to those of rat TAP-B (6, 11).

The polymorphism of rat TAP described above was discovered on account of a phenomenon termed class I modification (cim).⁴ A rat MHC class Ia allele named RT1-A^a showed striking differences in expression and antigenicity for T lymphocytes depending on the genetic origin of the MHC class II region with which it was expressed (8, 12, 13). It emerged that allelism of the MHC class II-linked TAP2 gene was responsible for the observations. TAP heterodimers containing the TAP2A allele could readily supply the type of peptides required for RT1-A^a to assemble, but those containing TAP2B were deficient in this, such that RT1-A^a assembled slowly and with a different spectrum of peptides (9, 14). The relevant feature of the peptides required for RT1-A^a assembly was shown to be a carboxyl-terminal arginine residue (15, 16).

An RT-PCR-based method has proved extremely effective for the expression cloning of MHC class I molecules from the laboratory rat, and we have undertaken experiments to clone MHC class Ia cDNAs from all standard RT1-A regions (5, 6, 17). Here we describe expression cloning from three related rat MHC haplotypes, RT1^o, RT1^d, and RT1^m. We find that these haplotypes express three class Ia species and report that the MHC class Ia (RT1-A) molecules of the RT1^m haplotype, currently categorized as unique (18, 19), are, in fact, the same as those in RT1^o and RT1^d. This unexpected finding is explained as a previously undetected, and severe, instance of the cim phenomenon, affecting two of three class Ia molecules in an RT1-A haplotype.

Rats were bred and maintained in specific pathogen-free conditions at either the Babraham Institute or the central animal facilities of the Hannover Medical School.

The RT1^o fibroblast cell line was derived from cells that remained attached to a tissue culture flask after preparation of Con A-stimulated spleen lymphoblasts (CABs) from a 4.5-mo-old RT1^o stock female rat. The previously unpublished BBH2 fibroblast line (RT1^u haplotype) was a gift from Dr. B. Michelsen (Hagedorn Research Institute/Novo Nordisk, Copenhagen, Denmark). This was derived from cultures of fetal homogenates from diabetes-prone BB/H rats. After ∼75 passages cells ceased to divide: some surviving cells eventually started to divide and formed the basis for the immortalized fibroblast cultures. Both the RT1^o and the BBH2 fibroblast cell lines were grown in DMEM/10% FCS and passaged twice a week by trypsinization. Mouse L(tk⁻) and simian COS-7 cells were also maintained in DMEM/10% FCS. The IFN-γ used was of mouse origin and was in the form of supernatant from the X63-γ cell line (20).

Samples (0.5 μg of DNA and 4 pmol of appropriate primers) for DNA sequencing were submitted to the sequencing service of the Babraham Institute Microchemical Facility, where they were run on an ABI PRISM 373A automated sequencer (PE Applied Biosystems, Warrington, U.K.) after reactions employing AmpliTaq DNA polymerase and dsDNA.

Sequences were edited using GCG software (University of Wisconsin Genetics Computer Group, Madison, WI); FASTA and BLAST were used for comparison with sequences in the EMBL database. The novel or published sequences are available from the EMBL nucleotide database with the following accession numbers: RT1-A^a (M31018), clone 3.6 (M31038), clone cc23 (AJ005025), RT1-A^u (X82106), RT1-A1^c (X90370), RT1-A2^c (X90371), RT1-A1° (X90373), RT1-A2° (X90372), RT1-A3° (X90374), RT1-A^f (Y14014), RT1-A2^f (Y13579), and RT1-A3ⁿ (AJ277139).

mAbs that have not been described previously were generated through the following immunizations, using the Y3-Ag1.2.3 cell line (21) as a fusion partner: GY38/7, LOU/ola anti-RT1^o stock; and GY40/87 and GY40/267, SHR anti-RT1^o stock. A number of the mAbs used have been described previously: JY/, R2/, and R3/series, AO anti-DA (22, 23); YR2/series, LOU/ola anti-RT1^o stock; YR5/10, AO anti-PVG; MAC 30, PVG-RT1^u anti-PVG.R8 (all in Ref.24); GN3/2 (MAC161) and GN4/4 (MAC162), PVG-RT1ⁿ(BN) anti-PVG-RT1^u; and GN6/9 (MAC166) and GN7/5 (MAC169), PVG-RT1^l(LEW) anti-PVG-RT1^u (5, 24). The origins of the three mouse anti-rat MHC class I mAbs, F16.4.4, MRC-OX18, and HAM-2, were described previously (5). The 4C9 mouse hybridoma against rat β₂-microglobulin (25) was supplied by Dr. P. Bjorkman (Caltech, Pasadena, CA).

This was performed as previously described (5, 17). Briefly, mRNA was purified from unstimulated splenocytes from a 5-mo-old RT1^o stock female rat or a 10-mo-old male BDIX rat or from CABs generated from DA.1M spleen and lymph node cells. After oligo(dT)-primed RT, PCR was conducted between oligonucleotides 5′-GCTCTAGAGTCCAGGCAGCTGTCTTCA-3′ and 5′-TGCTGCTGGCGGCCGCCCTGG-3′ before cloning into pCMU-D^b via NotI and XbaI sites. Plasmid DNAs from transformant colonies were subsequently analyzed by restriction digests and flow cytometric analysis of transiently transfected COS-7 cells (26). Those plasmids expressing MHC class I molecules as indicated by staining with an initial panel of mAbs were then selected for more detailed serological characterization on transiently transfected COS-7 cells and were subsequently DNA sequenced. On account of the high frequency of PCR-generated mutations (5, 17, 27), it is essential when using such an approach to sequence multiple, independently generated, clones to obtain the native sequence. If the sequences differ, then sequencing of at least three clones is necessary to determine a consensus sequence.

BBH2 cells were transiently transfected using FuGene 6 (Roche, Meylan, France) following the manufacturer’s instructions, using 2 × 10⁵ cells, 5 μl of FuGene, and 1 μg of each expression plasmid. TAP2A and -B were in pHβApr-1-neo (9), and RT1-A cDNAs were in pCMU (17). FuGene/DNA mixes were added to the cells, which were returned to 37°C for 48 h before harvesting by trypsinization and staining for FACS analysis.

Stable L(tk⁻) cell transfectants were generated by calcium phosphate transfection using pSV2neo as a cotransfection marker. Transfectants were selected using 0.5 mg/ml G418 (Life Technologies, Cergy Pontoise, France), and cells expressing the MHC molecule of interest were selected by flow cytometry. Stable BBH2 transfectants originated from the transient transfections described above. Surplus cells from one experiment were kept in their tissue culture flasks and returned to 37°C in culture medium. The following day (72 h after the transfection), 0.7 mg/ml G418 was added. Ten to 14 days later, multiple clones of G418-resistant cells were seen in all four transfection flasks (A1° and TAP2A, A2° and TAP2A, A1° and TAP2B, A2° and TAP2B), but not in the negative control flask. Cells were trypsinized and further amplified as oligoclonal populations for another 10 days before immunoselection of the cells expressing the transfected RT1-A molecules using JY1/132 mAb. The cells were first enriched via magnetic selection using the autoMACS technology (Miltenyi Biotec, Paris, France), and positively selected cells were then amplified further in culture before performing another round of selection by cell sorting using a Coulter HSS ALTRA flow cytometer (Beckman Coulter, Paris, France).

Cells were trypsinized and washed once in culture medium before incubation with primary Ab on ice for 30 min. They were then washed three times by centrifugation using cold PBS containing 2% (v/v) FCS and 0.1% (w/v) sodium azide (PFN) before incubation with secondary Ab (FITC-conjugated rabbit-anti-rat IgG diluted 1/100 in PFN) for 30 min on ice. After two washes in PFN, cells were resuspended in PFN containing 2 μg/ml propidium iodide and analyzed on a FACScan (BD Biosciences, San Jose, CA) within 1 h of staining. For cell sorting, PFN was replaced by sterile tissue culture medium.

PVG-RT1^l anti-RT1^o effectors were generated by primary in vitro immunization. For this, lymph node cells (∼1–1.5 × 10⁶/ml) from PVG-RT1^l(LEW) rats were incubated in round-bottom microtiter trays with equal numbers of irradiated RT1^o stock lymph node cells in RPMI 1640 medium containing 5% FCS, 2.5 μM 2-ME, and 10% supernatant from Con A-activated rat lymphocytes (8). After 5 days at 37°C in a humidified and CO₂-controlled atmosphere, these cells were washed and then used in a standard 5-h ⁵¹Cr release assay, using 10⁴ target cells/well. Tests were performed in triplicate, with 3-fold serial dilutions starting at an E:T cell ratio of 100:1. Spontaneous release of ⁵¹Cr was determined by incubating targets with medium only. Specific lysis (percentage) was calculated by the formula: [(experimental counts − SR)/(total input counts − SR)] × 100.

For two-dimensional gel electrophoretic analysis, immunoprecipitation of MHC class I molecules was performed as previously described (28). Briefly, 1 × 10⁷ RT1^o CABs and 5 × 10⁶ L cell transfectants were metabolically radiolabeled with 3.7 MBq of ³⁵S-Translabel (ICN Pharmaceuticals, Basingstoke, U.K.) for 30 min. Cells were then lysed in detergent buffer (1% Nonidet P-40, 150 mM NaCl, 10 mM Tris (pH 7.5), and 1 mM PMSF) for 20 min, and insoluble material was removed by centrifugation at 20,000 × g for 10 min. Samples were precleared for 1 h with a mixture of proteins A and G/Sepharose (Sigma-Aldrich, Dorset, U.K.), followed by immunoprecipitation for 1 h with 2 μg of purified Ab or 100 μl of Ab-containing supernatant as required, with a mixture of proteins A and G/Sepharose. Samples were washed three times in lysis buffer. Two-dimensional electrophoresis was performed on Bio-Rad Mini Protean equipment following the manufacturers’ instructions using a mixture of 2.5% ampholytes with pI range 3–10 and a 2.5% pI range 5–7 (Sigma-Aldrich). Second-dimension SDS-PAGE was performed with 10% Prosieve (Flowgen, Ashby de la Zouch, Leics, U.K.) gels using Tris-tricine buffers according to the manufacturers’ instructions. Dried gels were exposed to Kodak Biomax MS1 film with Kodak Transcreen LE intensifying screens (Eastman Kodak, Rochester, NY) at −80°C.

For pulse-chase analysis of RT1-(A1° and A2°), CABs were suspended at ∼3 × 10⁷/ml in methionine-free DMEM (without serum but supplemented with 2 mM l-glutamine) and incubated at 37°C for 30 min. They were then pulsed for 10 min with 3.7 MBq ³⁵S-Translabel. To terminate the incorporation, cells were centrifuged and then resuspended in DMEM containing 10% FCS and 2 mM unlabeled methionine. Aliquots of cells were removed at 0, 30, and 90 min. Cells were lysed in detergent buffer, and insoluble material was removed by centrifugation. Samples were then precleared for 1 h at 4°C by rotating with beads of Sepharose 4B (cyanogen bromide-activated) coupled to rat serum. Specific immunoprecipitation was performed by addition of an aliquot of Sepharose 4B beads coupled with rat alloantibody MAC30, followed by rotation for 1 h at 4°C. After four washes with detergent buffer and one with 150 mM NaCl and 10 mM Tris (pH 7.4), the beads were resuspended in loading buffer and heated at 85°C for 3 min, and samples were electrophoresed on a 7–16% gradient SDS-polyacrylamide gel.

To characterize MHC class I molecules encoded by the RT1^o haplotype, we employed methods developed and described previously (see Materials and Methods). RT-PCR products from spleen mRNA were cloned into pCMU-D^b, and bacteria were transformed. Plasmid DNA from 100 colonies was analyzed, firstly by restriction digest and secondly by transient transfection into COS-7 cells in combination with flow cytometric analysis using several xeno- and allospecific mAbs recognizing MHC class I determinants. Of these 100 plasmids, 60 determined the expression on COS-7 of a class I molecule recognized by the majority of RT1-A°-reactive mAbs available to us, which we called A1°. We obtained another species, albeit in only three clones, that was recognized by a nonidentical set of the RT1-A°-reactive mAbs in our panel; this we named A2°. A third species of cloned cDNAs was identified in 14 of 100 plasmids, characterized in restriction digest by a distinctive internal PvuII site. Transfection of these plasmids led to the efficient expression of an MHC molecule that we call RT1-A3°. This molecule was recognized by several allospecific mAbs, notably by cross-reactions of some RT1-A^a-specific mAbs that did not recognize A1° or A2°. mAbs that reacted selectively with each of the three A° molecules, viz., YR2/51 for A1°, GN7/5 for A2°, and R3/13 for A3°, were identified (see Table II).

Serological analysis of COS-7 cells transiently transfected with the remaining 23 plasmids resulted in either no staining or weak staining with only the monomorphic, xenospecific anti-rat MHC class I Ab MRC-OX18. From previous experience we believe that these clones encoded either 1) MHC class Ib heavy chains that could not be expressed efficiently in COS-7 cells or were not recognized by the panel of Abs used in our screen, or 2) versions of the first three species described above carrying PCR-generated mutations preventing their expression.

Independently derived plasmids were chosen for sequencing (seven for A1°, three for A2°, and four for A3°). Plasmids carrying consensus sequences were obtained for all three species. These sequences have been deposited in the EMBL database under accession numbers X90373 (RT1-A1°), X90372 (RT1-A2°), and X90374 (RT1-A3° = V°), and Fig. 1 shows the translated polypeptide sequences aligned with those of other selected rat MHC class I heavy chains. Analysis of the nucleotide sequences obtained for A1° and A2° showed that they were most closely related to other rat RT1-A sequences. We reported earlier that when the regions comprising exons 4–8 of RT1-A heavy chain sequences are compared, they fall into two clusters. We proposed that these clusters correspond to two RT1-A loci, A1 and A2 (6). Thus, A1° falls into the A1 cluster, and A2° into the A2 cluster.

By contrast, nucleotide sequence analysis of RT1-A3° placed it outside the RT1-A1 or -A2 groups. When exons 4–8 were first compared among a broad collection of rat class I cDNA sequences (29), the nearest relative for A3° (termed RT1-V° in Ref.29) was identified as clone 3.6 (30), with which it had 92% homology. RT1-A1 and -A2 sequences are only 81–82% homologous to A3° and 3.6 over the same region. While the latter two sequences have clearly diverged by ∼10% at the nucleotide level throughout the coding sequence (overall, 92%; α1, 93%; α2, 87%; α3, 95%; TM/IC, 92%), many unique sequence features common to them can be identified all along the open reading frame. The cDNA 3.6, which was isolated first from a library derived from a DA rat (RT1^av1), is expected to produce a soluble class I molecule, possibly that described by Spencer and Fabre (31), on account of a 13-nt deletion near the beginning of exon 5 (30). RT1-A3° does not carry such a deletion and is therefore predicted to express transmembrane and intracytoplasmic domains similar to those of other class I molecules, except that it lacks the final 16 aa encoded by exon 7. In this region the DNA sequences of A3° and 3.6 are, however, still closely related, as can be seen from the line labeled 3.6* in Fig. 1, which shows the translation predicted for clone 3.6 if the frameshift caused by the 13-nt deletion was corrected. More recently, other nucleotide sequences related to A3° and 3.6 have been identified. These include the genomic A3ⁿ sequence (discussed below) and the cc23 cDNA (32), another putative soluble form. Searching the nucleotide database for related sequences in other species did not reveal matches with high homology scores.

An unusual feature of the RT1-A3° molecule lies in the transmembrane and intracytoplasmic domains. The predicted amino acid sequence for this region of the molecule is divergent from any other MHC molecule reported to date (with the exception of the presumptive allele RT1-A3ⁿ), with a particularly striking stretch of eight successive arginine residues located just after the transmembrane domain (Fig. 1). A similar stretch of seven arginines is predicted for A3ⁿ.

The high levels of homology of A1° and A2° with other A1 and A2 sequences provided a very strong argument for their mapping to the RT1-A region, although no intra-MHC laboratory recombinants involving this haplotype were available to address this issue using formal genetics. The case for RT1-A3^o, however, was not as clear, so further haplotypes were investigated. Classical immunogenetic typing has suggested that rats of the RT1^d and RT1^o haplotypes share the same classical MHC class I Ags (class Ia) but differ with respect to MHC class II Ags (RT1-B/D) and, probably, MHC class Ib Ags mapping at the telomeric end of the MHC (RT1-C/E/M) (33). We next undertook a careful serological and expression cloning analysis of the BDIX rat strain (RT1^dv1 = RT1-A^d/oB^dD^dC^l?) in comparison with the RT1^o stock (RT1-A^d/oB^aD^aC^o).

We used the three mAbs selectively reactive with A1° (YR2/51), A2° (GN7/5), and A3° (R3/13) to investigate the expression of these molecules on spleen lymphocytes and RBC. We found similar levels of expression of the three types of molecules on RT1^o stock lymphocytes. On RBC, however, the expression of A2° and A3° was low compared with that of A1° (Table I); the low, but unambiguous, binding of R3/13 to RT1^o RBC was reported previously (22). Table I also shows results obtained with cells from BDIX rats. The close similarity of the results obtained with the RT1^o stock and BDIX rats lent credence to the idea that these strains carry the same RT1-A regions. At the same time there is evidence that the BDIX strain possesses an RT1-C/E/M region that is possibly identical with that of the LEW strain (RT1^l). Since the R3/13 mAb does not react with LEW cells, it seems probable that the A3° gene does not map in the RT1-C/E/M class I region, but, rather, in the centromeric class I region of the rat, i.e., the RT1-A region.

The hypothesis that RT1-A° and RT1-A^d class I Ags are identical was tested directly by conducting expression cloning from BDIX spleen cells. Among 50 clones screened, we isolated plasmids expressing molecules serologically identical with A1°, A2° and A3°, which were in relative proportions comparable to those obtained from RT1^o stock (12 A1^d, one A2^d, and two A3^d). Sequencing confirmed that the cDNAs for these molecules were indistinguishable from those obtained from the RT1^o haplotype. The finding that the RT1-A3 molecules from these two rat strains have identical sequences again supports the suggestion that this gene maps to the RT1-A region. This hypothesis has been further strengthened by recent examination of genomic PAC clones for the RT1-A region of the BN rat (RT1ⁿ). In addition to A1ⁿ and A2ⁿ, this work identified a gene, the exons of which display a high degree of similarity to A3° and which has been named A3ⁿ (4).

Expression plasmids encoding A1°, A2°, or A3° heavy chains were transfected into mouse L cells. As with the previous transient transfection into simian COS-7 cells, expression was observed with all three plasmids, as detected by immunofluorescent staining and flow cytometry. Stable expressors were then selected, and these were used to document the specificity of a panel of RT1-A°-reactive mAbs (Table II). These results showed that many of the mAbs cross-reacted on two or more of the A° molecules, although mAbs selective for each of the three A° molecules were present. It was also noticeable that all the mAbs raised by direct immunization against the RT1^o haplotype reacted with A1°. A fibroblast cell line derived from an RT1^o stock rat was also analyzed. Interestingly, the levels of surface expression of A2° and A3° in these RT1^o fibroblasts (see mAbs GN7/5, JY3/50, R3/13, and JY2/73), as in RBC, were lower than that of A1°. Treatment with γ-IFN resulted in a 10- to 20-fold induction of cell surface expression of all three molecules. This can be seen most clearly with the mAbs selective for each of the three A° molecules, viz., YR2/51 for A1°, GN7/5 for A2°, and R3/13 for A3°.

We also examined the expression of RT1^o MHC class I molecules using two-dimensional gel analysis (SDS-PAGE/isoelectric focusing) of immunoprecipitates obtained from metabolically labeled cells (Fig. 2). The Abs we used were 4C9, specific for rat β₂-microglobulin; MRC-OX18, reactive with a broad range of rat MHC class I heavy chains; R3/13, reactive with the A3° molecule; MAC30, reactive with both A1° and A2°; and GN7/5, reactive with A2°. Attempts to use the A1°-specific mAb YR2/51 for immunoprecipitation were unsuccessful. First (Fig. 2,A), we immunoprecipitated from L cell transfectant lines expressing each of the A° molecules to establish the locations of the three class I heavy chain species in our gel system. The relative positions of the heavy chain spots were as expected from the predicted pI values of these polypeptides (calculated using the program ProtParam tool). These experiments established that the A3° heavy chain spot resolved well away from A1° and A2°, but that the latter two spots were in overlapping positions. 4C9 and MRC-OX18 immunoprecipitates from RT1^o lymphoblasts (Fig. 2,B) displayed predominant species at two positions corresponding to A3° (left side) and to A1° and A2° (right side), but no significant sign of other heavy chain species. The identities of these major bands were confirmed by immunoprecipitation with the more selective Abs R3/13, MAC30, and GN7/5. Furthermore, immunoprecipitates obtained with an anti-rat TAP2 antiserum also revealed heavy chain spots at positions corresponding to A3° and to A1° and A2°, implying that molecules of these types interact successfully with the peptide-loading machinery in the endoplasmic reticulum (see Fig. 3 c in Ref.28). Together, A1°, A2°, and A3° account for the majority of the MHC class I heavy chains detectable in RT1^o lymphoblasts under the conditions of labeling and immunoprecipitation employed.

The identification in the rat RT1^o and RT1^d haplotypes of three well-expressed class I molecules, probably all of which are encoded in the class Ia (RT1-A) region of the MHC, stands in contrast to studies of other rat haplotypes in which MHC class Ia function appears to be mediated by one (30) or two (5) RT1-A region species. We were interested to know whether these three different A° molecules had similar capacities to act as target structures for T cell recognition. We generated alloreactive CTLs against RT1° Ags by primary in vitro immunizations of lymphocytes from both PVG-RT1^l(LEW) and PVG-RT1^lv1 (F344) rats (data not shown for the latter) and tested them against L cell transfectants expressing A1°, A2°, or A3° (Fig. 3). Although in both cases cross-reactivity on the control L cells was evident, it was nevertheless clear that all three A° species increased the level of CTL killing. In both experiments the killing of the A1°- and A2°-expressing targets was greater than that of the A3°-expressing target.

Messenger RNA was prepared from Con A-stimulated lymphoblasts from the DA.1M (RT1^m) rat strain, and MHC class I expression cloning by RT-PCR was conducted as described in Materials and Methods. The class Ia Ags of the RT1^m haplotype have previously been categorized as unique. We were therefore surprised to find that the principal MHC class I species that we cloned from DA.1M cells were the same three species identified in the RT1^o and RT1^d haplotypes, namely, A1°, A2°, and A3°. These were present as, respectively, five, five, and four clones of 20 analyzed. The different relative proportions compared with the cloning from d and o haplotype cells may be related to the use, in this case, of mitogen blasts, rather than spleen cells, as the source of mRNA.

Previous immunogenetic studies had demonstrated that the RT1^m haplotype was distinct from the RT1^o and RT1^d haplotypes with respect to the MHC class II region, including the alleles expressed at the TAP2 locus (10, 34, 35). We now hypothesized that the previous classification of RT1^m class Ia Ags as separate and distinct from those of the RT1^o and RT1^d haplotypes (36) was a consequence of this TAP polymorphism, and represented a new example of class I modification (cim) (see introduction). One of the defining features of the cim phenomenon was the systematic variation in the level of surface expression of the RT1-A^a allotype on primary cells in the presence of the TAP2A or the TAP2B allele (37). We therefore assessed MHC class I expression on RT1°- and RT1^m-bearing cells by immunofluorescent staining and flow cytometry using the RT1-A1/A2-reactive mouse mAb F16.4.4. It can be seen in Fig. 4 that the level of staining obtained with cells of the RT1^o haplotype (DA.1O) was far greater than that with congenic cells of the RT1^m haplotype (DA.1M). The differences observed were ∼16-fold in the case of RBCs (Fig. 4,A and Table III) and 1.7-fold in the case of unstimulated lymphocytes (Fig. 4,B). Similar results were obtained with the A1^o,m/A2^o,m-reactive mAb MAC30 (Table III). Using this mAb for immunoprecipitation from lymphoblasts pulse-labeled with methionine/cysteine, we were also able to demonstrate a profound difference in the rate of intracellular assembly and transport of these class I molecules in the two haplotypes, with little, if any, maturation being detectable in the RT1^m cells (Fig. 5).

Staining of RBCs with mAbs selectively reactive with each of the three A^o,m molecules revealed that the quantitative MHC class I expression differences between the RT1^o and the RT1^m cells applied to both the A1^o,m and the A2^o,m molecules, but not to A3^o,m (Table III). The staining of RT1^m RBC by mAbs YR2/51 (anti-A1^o,m) and GN7/5 (anti-A2^o,m) was almost indistinguishable from that in the negative control samples. By contrast, the staining with R3/13 (anti-A3^o,m), although at a relatively low modal value, was as good on RT1^m as on RT1^o cells.

Finally, we compared the staining of RT1^m and RT1^o cells by rat mAb JY3/84. This mAb was found previously to recognize the RT1-A^a molecule in a cim-dependent manner, displaying much greater reactivity in the context of the TAP2A than the TAP2B allele (8, 12). Since JY3/84 also reacts with A^d/o molecules (12) Table II), it was interesting to compare its staining of RT1^m and RT1^o cells. The result was clear-cut: JY3/84 stained RT1^o, but not RT1^m RBCs (Table III).

We next sought direct evidence that the rat TAP-A and TAP-B alleles could differentially influence the expression of RT1-A^o,m molecules. In most aspects of the cim phenomenon as it affects the expression of the RT1-A^a molecule (9), TAP-A is dominant and TAP-B is recessive. We therefore designed a transient cotransfection experiment in which cDNAs for the three RT1-A^o,m molecules and the two rat TAP2 alleles were introduced in various combinations into the TAP-B (recessive) fibroblast cell line BBH2 (RT1^u). Transfectant cells were then examined 48 h later for RT1-A^o,m expression using flow cytometry. The rat mAb JY1/132 was selected because it reacts on all three RT1-A^o,m species (Table II) and could therefore be used to assess both transfection efficiency and the levels of expression attained. mAb JY3/84 was used because it detects a highly cim-dependent MHC class I epitope (see above) and could be used to assess whether the cim-like effects observed between the RT1^o,d and RT1^m haplotypes were mediated by TAP. The results are shown in Fig. 6,A. It can be seen that transfection of either A1° or A2° in combination with TAP2A led to substantially higher JY1/132 staining than in combination with TAP2B. It is unlikely that this difference resulted from inadequate expression from the TAP2B plasmid, since a different pattern was evident in the A3° transfectants, where JY1/132 staining was higher in cells transfected with TAP2B than TAP2A. The staining of the transfectants with the cim-dependent mAb JY3/84 confirmed the impact of TAP2A. JY3/84 staining of both A1° and A2° transfectants was much higher in the presence of TAP2A than TAP2B; indeed, the JY3/84 epitope on A2° appeared completely TAP2A dependent in this experiment (N.B. JY3/84 is not reactive with A3° molecules; Table II).

Next, G418-resistant cells were selected from the A1° and A2° transfections, and expressing cells were selected by one round each of immunomagnetic selection and FACS sorting using JY1/132, briefly cultured and then stained with mAbs JY1/132 and JY3/84. The results with these stable A1° and A2° transfectants (Fig. 6 B) were in accord with the transient expression experiments, namely, quantitative dependence of JY1/132 staining and nearly complete dependence of JY3/84 staining upon the cotransfected TAP allele. Overall, these data demonstrate a direct cim-like effect of TAP2 acting on the RT1-A1^o,m and RT1-A2^o,m class I molecules.

The MHC is a genetic region that evolves rapidly under the constant selective pressure of infectious microorganisms. The number of MHC class I genes expressed by an individual is an important parameter of this process; the expression of more molecules will allow presentation of a broader range of Ags, but may also result in a wider deletion of functional T cells through negative selection to achieve self tolerance (38). Mathematical modeling suggests that the optimal number of T cell-restricting MHC class I molecules is between two and three per haploid genome, which are indeed the numbers found in mouse and human (39). Our previous immunogenetic analysis of rat MHC class Ia expression has indicated that RT1 haplotypes express in some cases one and in other cases two molecular species of this type (5, 6) mapping to the RT1-A region, named RT1-A1 and -A2. The findings presented here add to the picture of genetic complexity and plasticity of RT1 evolution in that we found the expression of an RT1-A3 molecule as yet another polymorphic component of the RT1-A region in some haplotypes (4).

In the rat, unlike mouse and human, all the class Ia activity is concentrated to one side of the MHC, closely linked to the TAP genes, which encode the transporter associated with Ag processing. We have suggested that rat TAP and RT1-A alleles have coevolved in functionally optimal cis pairings (6, 40). Discordant TAP/RT1-A pairings have been observed in the laboratory, notably the association of rat TAP-B with the allotype RT1-A^a in the intra-MHC recombinant strains PVG.R1 and PVG.R8. In these strains the A^a molecule, which prefers to assemble with peptides bearing arginine at the C terminus, is supplied only poorly with such peptides on account of the restrictive specificity of the TAP-B transporter allele (11, 41). As a consequence, A^a is assembled less efficiently and expressed less well in TAP-B than in TAP-A cells and may also assemble with distinct peptides, leading to differences in antigenicity detectable by T cells. The structural basis of the preference of A^a for C-terminal arginine is a triad of negatively charged residues in the F pocket of the peptide binding groove, namely, D77, E97, and D116 (6, 15, 42). Both RT1-A1° and -A2° also possess this triad; indeed, they share with RT1-A^a (and A^f) the entire polypeptide sequence from positions 74–160, i.e., the complete right end of the α₁α₂ codomain (Fig. 1). Furthermore, experiments exploiting anhydrotrypsin to analyze peptides with basic C-terminal residues revealed a high proportion of such peptides bound to the MHC class I molecules of cells of the RT1^o haplotype (6). This is consistent with their genetic association with the TAP-A allele in the RT1^o and RT1^d haplotypes, since this transporter is able to ferry arginine-ended peptides into the endoplasmic reticulum, while the TAP-B allele is very deficient in this function. (The shared sequence (aa 74–160) is strongly suggestive of recent intergenic recombination events both between and within RT1 haplotypes and may also underlie the serological cross-reactivity seen between A1° and A2° in Table II.)

RT1-A/TAP discordance is therefore a likely explanation for the poor expression of the RT1-A1° and -A2° allotypes in rats of the RT1^m haplotype, which carries the TAP-B allele. This explanation was directly supported by the results of the RT1-A/TAP cotransfection experiments shown in Fig. 6. These experiments also demonstrated the TAP dependence of the expression of the JY3/84 epitope on A1^o,m and A2^o,m molecules. From its first detection by Kren (43), the RT1^m haplotype presented difficulties for major histocompatibility Ag typing. Opinions have differed over whether its MHC class Ia and class II regions/Ags are unique or correspond to those in standard haplotypes (34, 35, 36, 44). The most thorough serological analysis was performed by Stark and colleagues (36), who concluded that the RT1^m class I Ags are serologically closely related to, but distinct from, those of the a, d, o, and f haplotypes. Of particular significance to the present report, the latter authors were able to generate an RT1-A-specific alloantiserum between rats of the m and o haplotypes (MNR anti-MR) (36), and we speculate that this antiserum was of a specificity similar to that of the JY3/84 mAb.

It is interesting to note that the impact of the RT1-A1^o,m/TAP and RT1-A2^o,m/TAP discordances appears even more profound than in the original cim phenomenon involving RT1-A^a. Firstly, the effect on cell surface expression of A1°/A2° is greater. Fig. 4,A, and Table III show an ∼8- to 16-fold difference in expression on RBCs between o and m haplotype cells (mAbs MAC 30 and F16.4.4), while the corresponding difference for RT1-A^a is ∼3-fold (Table III, PVG.R19 vs PVG.R1). Secondly, while, as mentioned above, RT1-A-specific Abs could be generated between rats of the m and o haplotypes (MNR anti-MR) (36), equivalent antisera could not be raised in studies of the cim effect on RT1-A^a (G. W. Butcher, unpublished observations). It seems, therefore, that RT1-A1° and -A2° are even less compatible with TAP-B-transported peptides than is A^a. It is possible that features of their peptide anchor motifs other than the C-terminal arginine requirement are responsible for this, since it has been shown that internal residues in peptides can influence the efficiency of peptide transport by TAP (45). The deficient assembly of RT1-A1° and -A2° in RT1^m cells raises the question of whether this haplotype would be sufficiently well adapted to be maintained in wild populations of rats. It has to be considered that it may have arisen as a result of an intra-MHC recombination event in laboratory rat stocks between an RT1^d or RT1^o haplotype (both TAP-A) and another haplotype bearing an RT1^c-like class II region (including TAP-B). We should note here that unpublished work from one of our laboratories (D. Wedekind and H. J. Hedrich) suggests that the class II region of RT1^m is not identical with that of RT1^c, since some differences are detectable in the sequences of the RT1-Bβ molecules. Finally, the poor surface expression of RT1-A1° and -A2° on mouse L cells (Table II) is probably also a consequence of class I/TAP discordance, since mouse TAP has been shown to behave similarly to rat TAP-B and unlike rat TAP-A (11, 41).

The level of expression of the RT1-A3° molecule, as detected by the R3/13 mAb, was similar in cells of the RT1^o and RT1^m haplotypes. Furthermore, structural predictions of the RT1-A3° molecule do not suggest a preference for arginine-ended peptides. A3° lacks the acidic triad discussed above and instead bears 77S, 97L, and 116S, a combination more suited to interaction with uncharged side chains. At present, therefore, we have no reason to propose adaptive pairings with TAP for alleles at the A3 locus. Since an A3 gene has now been detected in the RT1ⁿ haplotype that carries the TAP2B allele, it is likely that this locus predates the divergence of the two rat TAP alleles. Expression of A3° was found to be intermediate on a fibroblast cell line (Table II) and very substantial in lymphoblasts (Fig. 3), indicating a considerable investment of protein synthesis in this class I isoform. This is suggestive of a functional role for this molecule in the organism, rather than the status of an evolutionary relic. An unusual feature of the predicted polypeptide sequence of A3° is a stretch of eight successive arginine residues located intracytoplasmically, just after the transmembrane domain. The function of this is unknown. We eliminated the possibility that this molecule acquires a glycosylphosphatidylinositol anchor post-translationally, since the expression of A3° by RT1^o fibroblasts was unaffected by digestion with the enzyme phospholipase C (data not shown).

The prevalence of the A3 locus in different haplotypes has yet to be fairly assessed. It is a recent discovery and may not be serologically detectable with currently available reagents in several haplotypes. A feature of the available data on A3 is the remarkable divergence of the alleles known to date. Thus, the predicted polypeptide sequences of A3° and A3ⁿ differ at 46 positions, 43 of which are located in the α₁α₂ codomain. In addition, it is most intriguing that the closest published relative of A3° and A3ⁿ is the 3.6 putative soluble class I molecule of the RT1^av1 haplotype, which is probably the soluble molecule purified by Spencer and Fabre (31). More recently, cDNAs encoding further putative soluble homologues of clone 3.6 have been derived from LEW.1F rats (f haplotype; clones 9.5 and 9.6) and from PVG.R19 rats (av1-av1-c; clones cc22 and cc23) (32). Furthermore, the evidence in the av1 haplotype suggests duplication of the A3-like locus, with both genes, namely 3.6 and cc22/cc23, encoding soluble products (32).

We have shown here that the RT1^o haplotype encodes three expressed RT1-A cell surface class I molecules and that the same three molecules are expressed by cells of the RT1^d and RT1^m haplotypes. To what extent do molecules of the A1, A2, or A3 series fulfil the role of fully fledged classical MHC class Ia molecules? Among RT1 haplotypes that use just a single MHC class Ia allele, one finds representatives of both the A1 series (RT1-A^l and RT1-A^a) and the A2 series (RT1-A^u) displaying full class Ia function. When considering RT1 haplotypes with two or three expressible RT1-A species, the issue arises of whether these are functioning equivalently or whether there is functional dominance of one over the others. When considering the human MHC system, HLA, it is possible to discern a hierarchy among the class Ia species with respect to their involvement in CD8⁺ T cell restriction in which HLA-B may be considered dominant over, in order, HLA-A and HLA-C (38).

In the present study the A1° molecules assume fully their roles as class Ia molecule and Ag-presenting molecules by three criteria mentioned previously: 1) high degree of sequence similarity to other known rat class Ia sequences; 2) recognition by alloreactive RT1-A specific mAbs, and 3) recognition by primary in vitro CTLs. By serological analysis, A1° expressed singly in mouse L cells (L-A1°) displayed wider and stronger reactivities with the panel of mAbs tested compared with L-A2° and L-A3° (Table II). Furthermore, it is noticeable that all the mAbs raised by immunization against RT1° cells react with A1°, while some, but not all, reacted on A2° or A3°. Functional studies in the form of CTL assays revealed a better efficiency in killing A1° than A2° or A3° targets (Fig. 3). These differences in antigenicity and immunogenicity suggest a hierarchy in which A1° performs classical MHC class I function more prominently than A2° or A3°. Indeed, it is conceivable that A1° is the only effective stimulator of alloreactive Abs or CTLs in our studies and that the reactions observed against A2° and A3° are due to cross-reactions of A1°-specific Abs/receptors.

The apparent dominance of A1° over the others may be related to their levels of expression. Firstly, among 100 RT1^o expression clones derived from spleen mRNA, the proportion of A1° clones was ∼60%, compared with 3% for A2° and 14% for A3°. A similar distribution of clones was obtained from a smaller screen of the BDIX strain (RT1^dv1) that shares the same RT1-A region as the RT1^o stock. Secondly, the levels of cell surface expression of A1° appear to be higher than those of A2° or A3°, most noticeably on RBC (Table I); the expression of all three molecules was increased substantially after IFN-γ treatment of RT1° fibroblasts, a feature common to conventional MHC class I molecules. Thirdly, in the detailed immunoprecipitation studies of RT1^o lymphoblasts, all detectable MHC class I heavy chains associated with β₂-microglobulin (precipitated by mAb 4C9) could be accounted for by the species A1°, A2° (weak), and A3°. Collectively, these data suggest that A1° is the principal, functional, MHC class Ia species in these related RT1 haplotypes (d,o,m) and that A2° and A3° play subsidiary, but significant, roles.

It may be a general rule that if an RT1 haplotype expresses any allele at the RT1-A1 locus, then this allele will function as a classical class Ia molecule. RT1-A2 alleles, by contrast, either may exhibit full MHC class Ia function, as do A^u in the RT1^u haplotype and A2^c in the c haplotype (5), or may not, as in the cases of A2° (described here) and A2ⁿ (A. González-Muñoz and G. W. Butcher, unpublished observations). It will be important to determine whether these differences are dependent upon issues of expression level, which, in turn, may depend either on rates of transcription or, alternatively, on post-translational control of assembly/expression by factors such as peptide stringency, peptide availability, and interactions with chaperones.

We are grateful to Eberhard Günther and Lutz Walter for discussions, and to Birgitte Michelsen for providing the BBH2 cell line. We thank Nigel Miller for help with flow cytometry, and Sieglinde Eghtessadi for technical assistance.