The Forgotten: Identification and Functional Characterization of MHC Class II Molecules H2-Eb2 and RT1-Db2

The mammalian MHC contains genes with pivotal roles in the immune system. Depending on the species, the MHC has different names: HLA in humans, H2 in mice, and RT1 in rats. The MHC is divided into three regions: class I, class II, and class III. Class I and II regions contain classical MHC molecules, which present Ag peptides to T cells. Classical MHC class I molecules present peptides to CD8 T cells, and classical class II proteins present peptides to CD4 T cells. In addition to these molecules, these regions encode many genes involved in the development of immune responses, in Ag processing and presentation, and in the presentation of Ags other than peptides. Among the products of these genes there are some that have a similar structure to the classical molecules, yet play different roles; these are known as nonclassical MHC molecules (1).

Classical MHC class II molecules are cell surface glycoproteins formed by an α- and a β-chain expressed on APCs, such as dendritic cells, macrophages, and B cells. Each chain contains two extracellular domains, a transmembrane, and a very short cytoplasmic region. Their Ag-binding groove is formed by the N-terminal domains of both chains (the α1 domain and the β1 domain). The high polymorphism, especially in their Ag-binding domains, allows them to present many peptide Ags. Binding to the N-terminal domain was also found in a number of cocrystals of superantigens (staphylococcal and streptococcal superantigens and of the Mycoplasma arthritidis superantigen bound to MHC class II molecules; however, their binding is clearly distinct from peptide-binding, because superantigens contact the outer surface of MHC class II molecules and not the Ag-binding groove. Moreover, the manner by which superantigens bind MHC class II molecules varies strongly among different superantigens (2). The two membrane-proximal extracellular domains (α2 and β2 domains) are Ig-like domains and are important for establishing contacts with CD4 (3).

The α- and β-chains of classical MHC class II molecules are transcribed and translocated into the endoplasmic reticulum (ER). There, they associate with the invariant chain (Ii; CD74), which acts as a chaperone assisting in the correct folding of the α− and β-chains and prevents the binding of peptides contained in the ER by accommodating an Ii–derived fragment, class II–associated Ii peptide (CLIP), in the binding groove. In addition, the Ii drives the newly synthesized class II molecules into the endosomal system, where proteases rapidly degrade the Ii and leave CLIP in the Ag-binding groove (4). In the endosomal compartments, a nonclassical MHC class II molecule, DM in humans and M in mice and rats, mediates the release of CLIP and stabilizes the open empty Ag-binding groove, allowing the accommodation of peptides derived most commonly from endocytosed proteins. Finally, class II molecules containing peptides from endosomal compartments are transported to the cell surface where they present these peptides to CD4 T cells. The conserved function of nonclassical MHC molecules is reflected by their lack of polymorphism.

Human MHC class II proteins are encoded by the DR, DQ, and DP loci. Individual gene expansions lead to different DRB region configurations (5), and peptide presentation was described for DRB1-, DRB3-, DRB4-, and DRB5-encoded β-chains. Rats and mice have two classical class II MHC molecules: RT1-B and RT1-D in rats and H2-A and H2-E in mice. These are the orthologs of the human HLA-DQ (RT1-B and H2-A) and HLA-DR (RT1-D and H2-E). RT1-D is encoded by the classical MHC class II genes RT1-Da (which encodes the α-chain) and RT1-Db1 (which encodes the β-chain) in rats and is localized on chromosome 20 (6). In mice, H2-E is encoded by the classical MHC class II genes H2-Ea and H2-Eb1 on chromosome 17. In addition to the genes that encode the classical β1 chain, rats and mice have a nonclassical β gene: RT1-Db2 in rats and H2-Eb2 in mice, which, in both cases, map between the classical α and β genes (7–9). Early reports that identified this nonclassical β-chain in mice described it as a gene with similar genomic organization to that of the classical H2-Eb1, but with limited polymorphism and an unusual pattern of transcription (9, 10). However, the complete coding sequences (CDSs) of H2-Eb2 and RT1-Db2 genes were not identified in these reports, and no evidence was found indicating that these genes were expressed at the protein level. Thus, many questions remained unanswered: Are these nonclassical β2 molecules expressed at the protein level? Do they pair with the classical RT1-Da and H2-Ea chains? Can they be expressed at the cell surface? Are these molecules able to present Ags?

The complete sequence of the rat MHC (6) and the availability of the mouse and rat genomes allowed us to identify the complete CDS of these nonclassical β2 genes and to study them by transducing them into cell lines. We found that the nonclassical β2 molecules share some features with the classical β1 chains, because they pair with the classical α-chains. The resulting RT1-D2 and H2-E2 molecules in transduced cell lines are expressed at the cell surface and present superantigens. Nonetheless, the nonclassical β2 molecules also show interesting particularities compared with their classical counterparts, such as very limited polymorphism in their β1 domain, lack of N-glycosylation and formation of SDS-resistant heterodimers, lower RNA abundance, and an extraordinary capacity to present a superantigen derived from Yersinia pseudotuberculosis.

C57BL/6J and BALB/c mice, as well as LEW/Crl and F344/DuCrl mice, were bred and kept in the animal facilities of the Institute for Virology and Immunobiology of the University of Würzburg. The procedures for performing animal experiments, as well as animal care, were in accordance with the principles of the German law. Permission to keep and breed the animals was given by the city of Würzburg, Germany (OA/he-wa07.12.1987). BN/SsNOlaHsd and PVG/OlaHsd mice were purchased from Harlan Laboratories, and DA/HanRj mice were from Janvier Laboratories. Single-cell suspensions from spleens were prepared as previously described (11). L929 (12), M12.4.1C3 (13), 53/4 (14), and P3/2 mouse fibroblasts expressing human HLA-DR1 (15) were cultured in RPMI 1640 supplemented with 10% FCS, 1 mM sodium pyruvate, 2.05 mM glutamine, 0.1 mM nonessential amino acids, 5 mM 2-ME, penicillin (100 U/ml), and streptomycin (100 μg/ml) at 37°C with 5% CO₂ and an H₂O-saturated atmosphere.

Analysis of MHC class II molecules mRNA expression by RT-PCR (shown in Fig. 2A) was performed using 1 μl cDNA with the following primer pairs (all primers are listed in Supplemental Table I): RT1-Db1, RT1DBexpfwd and RT1DBexprev; RT1-Db2, RT1DB2expfwd and RT1DB2exprev; H2-Eb1, Eb1EcoRIfo and Eb1BamHIre; and H2-Eb2, Eb2Bluntfo and Eb2bluntre. cDNA was produced using 500 ng RNA with the First-Strand cDNA Synthesis Kit (Fermentas). PCRs were run for 40 cycles at an annealing temperature of 60°C.

RT-PCR and direct sequencing were carried out to address nucleotide sequences of exons 2 and 3 of RT1-Db2 molecules. cDNAs were derived from splenocytes of several inbred rat strains. The primers used for the RT-PCR and direct sequencing were RT1DB2expfwd and RT1DB2exprev. Primer sequences are given in Supplemental Table I.

cDNA was synthesized, as described above, and purified using the DNA Clean and Concentrator-5 Kit (Zymo Research). Real-time PCR was conducted in a thermocycler iCycler (Bio-Rad) using Absolute QPCR SYBR Green Fluorescein Mix (Thermo Scientific). Relative expression of b1 and b2 genes was estimated by the ΔΔCt method normalizing the expression to HPRT genes. The following primer pairs were used: RN/MMHPRT1fwd and RNHPRT1rev to amplify rat HPRT, RN/MMHPRT1fwd and MMHPRT1rev to detect mouse HPRT, RT1Db1fwd and RT1Db1rev to amplify RT1-Db1, RT1Db2fwd and RT1Db2rev to detect RT1-Db2, H2Eb1fwd and H2Eb1rev to detect H2-Eb1, and H2Eb2fwd and H2Eb2rev to amplify H2-Eb2. Triplicates were run for each primer pair, in each of them, 1 μl the cDNA preparation (previously diluted 1/2) was used as template. Primer sequences are given in Supplemental Table I.

RNA sequencing (RNAseq) and ribosome profiling libraries (GSE62148) were reported by Diaz-Muñoz et al. (16). Libraries trimmed with Trim Galore! (http://www.bioinformatics.babraham.ac.uk/projects/trim_galore/) were mapped to the mouse mm10 genome using TopHat 2 (17) (v. 2.0.7 for RNAseq and v. 2.0.9 for ribosome profiling) and the annotation Mus_musculus.GRCm38.70.gtf as reference. Reads/kb of mapped million reads from RNAseq libraries were calculated using Cufflinks (18). Ribosome profiling reads were counted with HTSeq (19) and the Mus_musculus.GRCm38.73.gtf annotation. Counted reads were normalized using DESeq2 (20).

MHC class II molecules were cloned in diverse retroviral expression vectors (Supplemental Table II). For clarification purposes, we assigned abbreviated names for the different vectors generated (Supplemental Table II). Unless otherwise specified, in all cases the inserts were cloned into the EcoRI and BamHI restriction sites of the respective vectors. All accession numbers for the MHC class II molecules described in this section can be found at http://www.ncbi.nlm.nih.gov/nuccore/.

Da-H was constructed by cloning RT1-Da^l CDS amplified from Da-S (14) into pIH. The insert was prepared by PCR amplification with RT1Da_sense_k and RT1Da_antis_k primers. Db1-Z was cloned by inserting RT1-Db1^l CDS into pIZ. RT1-Db1^l insert was amplified by PCR with RT1Db_k_sense and RT1Db_antis_stop primers. Db1-S (14) was used as template. In Db1-EGFP-Z, RT1-Db1^l CDS was cut out with EcoRI, and BamHI from Db1-Z was inserted into the pEGZ vector. Before cloning RT1-Db2ⁿ CDS (accession no. KP012532) into retroviral vectors, it was cloned into the pDrive PCR cloning vector. This sequence differs in two synonymous nucleotide positions compared with the rat genome found at National Center for Biotechnology Information GeneBank database (accession no. NC_005119, annotation version Rnor6.0 constructed with pooled female animals of strain BN/SsNHsdMCW plus one male of strain SHR [also known as SHR-Akr]). Subsequently, the cloned RT1-Db2ⁿ CDS contained in pDrive was amplified by PCR with Sense_d-beta_2 and RT1Db2_antisense primers and, after digestion, was inserted into pIZ (abbreviated as Db2-Z). Db2-EGFP-Z was generated by cutting out RT1-Db2ⁿ cDNA from Db2-Z and inserting it into pEGZ.

YFP was fused to the C terminus of RT1-Db1 (cytoplasmic region) by inserting RT1-Db1 cDNA into pIZ-YFP. The insert was prepared by amplifying RT1-Db1 cDNA from the aforementioned vector Db1-S by PCR using the primers RT1-Db1k forwardYFP and RT1-Db1reverseYFP. Likewise, YFP was fused to the C terminus of Db2 after insertion of RT1-Db2ⁿ cDNA into pIZ-YFP. In order to fuse Db2 to YFP, RT1-Db2ⁿ cDNA was amplified by PCR with RT1-Db2k forwardYFP and RT1-Db2reverseYFP primers, using as template the Db2 cDNA contained in the pDrive vector, as described above.

The N-glycosylation site of RT1-Db1 was disrupted by site-directed mutagenesis involving nucleotide sequences encoding amino acids N19 and T21 to encode K19 and R21, respectively. The resulting amino acid sequence is the same as in Db2. Mutations were introduced in Db1-Z using the QuikChange II Site-Directed Mutagenesis Kit (Stratagene) and the Db1_NO-Ngly_fo and Db1_NO-Ngly_re primers. Then, the mutated Db1 was cut out and inserted into pEGZ (abbreviated as Db1-N EGFP-Z). Similarly, the N-glycosylation site of Db1 was transplanted to RT1-Db2 by mutating the nucleotides encoding K19 and R21 to those encoding N19 and T21 using the Db2_Ngly_fo and Db2_Ngly_re primers. The mutated Db2 also was cut out and inserted into pEGZ (Db2+N EGFP-Z).

Ea^d-H was cloned by amplifying H2-Ea^d CDS by PCR using cDNA derived from BALB/c splenocytes as template and the EaEcoRIfo and EaBamHIre primers. The H2-Ea^d insert was cloned into pIH. The sequence of the H2-Ea^d cDNA can be found under the accession no. KP012537. H2-Eb1^d CDS amplified from BALB/c splenocyte-derived cDNA with the Eb1EcoRIfo and Eb1BamHIre primers was first cloned into pIZ. Subsequently, it was amplified again with the same primers and, after digestion, was inserted into pEGZ (abbreviated as Eb1^d-EGFP-Z). H2-Eb1^d CDS sequence (accession no. KP012536) encodes the same amino acid sequence as Eb1 cloned from a BALB/c B cell lymphoma (21) (accession no. P01915.1). H2-Eb1^b CDS was cloned into pEGZ (abbreviated as Eb1^b-EGFP-Z) as H2-Eb1^d but using cDNA derived from C57BL/6 splenocytes. The obtained sequence is the same as the sequence published under accession no. NM_010382.2.

H2-Eb2^d CDS was amplified from BALB/c RNA derived from lymph nodes by RT-PCR using the Eb2Bluntfo and Eb2bluntre primers. H2-Eb2^d CDS insert was prepared by amplification with a second PCR using the Eb2EcoRIfo and Eb2BamHIre primers and subsequent digestion. H2-Eb2^d cDNA was inserted into pEGZ (the construct was abbreviated as Eb2^d-EGFP-Z). H2-Eb2^d cDNA sequence can be found under accession no. KP027332. H2-Eb2^b cDNA was amplified by RT-PCR using as template C57BL/6-derived RNA from lymph nodes and the Eb2Bluntfo and Eb2bluntre primers. H2-Eb2^b cDNA insert was also prepared with a second PCR using Eb2EcoRIfo and Eb2BamHIre primers and subsequent digestion. H2-Eb2^b cDNA insert was first cloned into the pIZ vector and, after excision, was subcloned into the pIEGZ retroviral expression vector. The H2-Eb2^b cDNA sequence obtained was the same as the sequence published under accession no. NM_001033978.

R150G mutation into H2-Eb2^b CDS was introduced using primers containing the desired mutation (H2Eb2Bl6fwd, H2Eb2Bl6R150Gfwd, H2Eb2Bl6R150Grev, and H2Eb2Bl6rev) and combining PCR products, as described elsewhere (11). Cloning of Ii was described previously (14).

Several cell lines (detailed in Supplemental Table III) were generated by transducing different combinations of MHC class II α- and β-chains into L929 (12) and M12.4.1C3 (13) cell lines, as previously described using retrovirus (11). To achieve expression of the desired molecules in all cells, cells were cultured in medium containing selecting antibiotics (hygromycin and phleomycin D) or sorted by FACS after transduction. To assure that differences in MHC class II molecule expression at the cell surface were not due to different transduction efficiencies, cell lines expressing either rat RT1-Da or mouse H2-Ea were first generated and then transduced with the different β-chains.

Flow cytometry analyses of cell lines were conducted by staining 200,000 cells in 100 μl FACS buffer. The following mAbs were used to analyze MHC class II cell surface expression: PE-conjugated 14.4.4S (BD Biosciences) (22); unconjugated OX-17 (contained in supernatant from OX-17 hybridoma culture) (23) detected with a polyclonal secondary donkey F(ab′)₂ fragment anti-mouse IgG (H+L) with minimal cross-reactivity to rat and other species serum proteins labeled with PE (Dianova) and unconjugated 3G9AD (contained in 3G9AD hybridoma culture supernatant) visualized with a polyclonal secondary PE-labeled donkey F(ab′)₂ fragment anti-rat IgG (H+L) with minimal cross-reactivity to other species serum proteins (Dianova). The mAb 3G9AD was a kind gift of Prof. Kurt Wonigeit (Medical School Hannover, Hannover, Germany) who developed the Ab with his team.

To address MHC class II molecule mRNA expression in different cell populations from the spleen, cells were sorted using a FACSAria III (BD Biosciences). Rat splenocytes were stained with anti-CD45RA (OX-33 labeled with FITC) to identify B cells and with anti-CD11b/c (OX-42-PE; both from BD Biosciences). Mouse B cells were stained with anti-CD19 (1D3 conjugated with allophycocyanin; BD Biosciences).

Biotinylation of cell surface proteins and immunoprecipitation were conducted as previously reported (11). Precipitates were resuspended in a 2% SDS-containing loading buffer (11) and divided into two aliquots. One aliquot was incubated for 5 min at 100°C (boiled), and the other one was incubated at room temperature for 1 h (not boiled). Precipitated proteins were separated on 12% polyacrylamide gels by SDS-PAGE. After blotting onto a Roti–polyvinylidene difluoride membrane (Roth), the membranes were incubated with streptavidin-HRP (BD Biosciences), and biotinylated proteins were visualized by ECL.

L929 cell lines expressing YFP fusion proteins were cultured on coverslips overnight at 37°C. M12.4.1C3 cell lines expressing YFP fusion proteins were added to poly-l-lysine–coated coverslips and incubated for 30 min at 37°C in media. After attachment to the coverslips, cells were fixed with 4% paraformaldehyde in PBS and permeabilized or not (for cell surface staining) with 0.3% Triton-X 100 in PBS. Ab staining was performed with unconjugated OX17 or R73 as isotype matched control contained in hybridoma culture supernatant and revealed with an Alexa Fluor 594–conjugated donkey anti-mouse IgG secondary Ab (Dianova). Before analysis, samples were embedded overnight at 4°C with Fluoromount G mounting medium (Southern Biotech).

Photographs were taken on a confocal laser-scanning microscope (LSM 780; Zeiss) and processed with ImageJ software.

YFP protein expression by Western blot was analyzed by running total cellular extract via SDS-PAGE under reducing conditions. After transfer to a Roti–polyvinylidene difluoride membrane, YFP protein was detected with an anti-GFP rabbit polyclonal Ab (A.v. Peptide Ab; Clontech), which recognizes YFP. Detection of the rabbit primary Ab was carried out with anti rabbit goat IgG conjugated to HRP and visualized by ECL.

Mycoplasma arthritidis superantigen was kindly provided by Prof. Lothar Rink (Rheinisch-Westfälische Technische Hochschule, Aachen University, Aachen, Germany). Mycoplasma arthritidis superantigen and Y. pseudotuberculosis mitogen (YPM) were purified as previously described (24–26).

Different L929-transduced cell lines generated in this study, L929 transduced with RT1-B^l (14), or P3/2 mouse fibroblasts expressing human HLA-DR1 (15) were seeded at a density of 10⁴ cells/well in a flat-bottom 96-well culture plate. One day later, 5 × 10⁴ 53/4 T cell hybridoma cells (14) were added to the cultures, together with superantigens at the indicated concentrations. After 22 h, mouse IL-2 contained in the culture supernatants was determined using a commercial mouse IL-2 ELISA Kit (BD Biosciences). Superantigen-mediated stimulation by primary splenocytes was carried out similarly, but using 4 × 10⁵ splenocytes and culturing them with the 53/4 T cell hybridoma and the superantigens on the same day that they were prepared.

In MHC class II–inhibition experiments, the following mAbs were added to the cultures: anti–HLA-DR (HB55) (27) (1.7 μg/ml), anti–RT1-B (OX-6) (28) from ascites fluid diluted 1:500, anti–RT1-Da and H2-Ea (14.4.4S) from ascites fluid diluted 1:500, and anti–RT1-DaDb1 (3G9AD) diluted 1:5.

The entire sequence of the rat MHC (6) allowed us to identify the complete CDS of RT1-Db2, which maps between the RT1-Db1 and RT1-Da genes and encompasses ∼20 kb (Fig. 1A). Exons 1–4 of RT1-Db2 and of the classical RT1-Db1 gene are homologous. They encode signal peptide, extracellular domains, and the transmembrane region, respectively. However, RT1-Db2 lacks homologs of exons 5 and 6 of RT1-Db1 as the result of a deletion of the region encoding exon 5 and inactivation of exon 6 (Fig. 1A). Nevertheless, we observed an insertion of a region containing a processed pseudogene of Arbb2, as well as a novel exon with a 3′ untranslated region (UTR) and a poly(A) signal. This novel exon does not show homology to other exons in MHC molecules. Together with the 3′UTR, it is indeed processed into mRNA and encodes the complete cytoplasmic region. RT1-Db2 mRNA is properly processed at a novel poly(A) signal site included in the novel 3′UTR (data not shown).

Analysis of mouse H2-Eb2 in the mouse genome (genome version GRCm38 based on the strain C57BL/6J available in Ensembl) shows the same genomic organization as RT1-Db2, mapping between H2-Eb1 and H2-Ea and also containing the novel exon encoding the complete cytoplasmic region and the 3′UTR. The promoter regions are highly conserved between RT1-Db2 and H2-Eb2 and show high similarity to the corresponding promoters of RT1-Db1 and H2-Eb1 (Fig. 1B): all contain Y, X, and S boxes, and the distance between the S and X boxes has been maintained.

We analyzed these β2 molecules in different haplotypes (Fig. 1C). Rat and mouse β2 predicted that amino acid sequences are highly conserved compared with classical β1 molecules, yet they also differ in important aspects. β2 molecules contain all domains of MHC class II β molecules: a signal peptide, β1and β2 domains, and transmembrane hydrophobic and cytoplasmic regions. The four cysteines responsible for the disulfide bridges in the Ag-binding domain (β1 domain) and in the Ig-like domain (β2 domain) of classical β molecules are conserved in mouse and rat β2 molecules. The degree of conservation of the β1 and β2 domains in rat RT1-Db2 and mouse H2-Eb2 is such that when the predicted amino acid sequences were interrogated for the presence of known domains with CD-search (29), the results were Ag-binding domain for the β1 domain and Ig-like domain for the β2 domain. Nevertheless, mouse and rat β2 molecules show interesting differences compared with the classical ones. First, the cytoplasmic region of β2 molecules is completely different from that of β1 molecules, which is to be expected because they are encoded by nonhomologous exons. Blast searches in the Mouse Genomic Plus Transcript, Reference RNA Sequenced, and Reference Proteins databases with the amino acid and nucleotide sequences of these novel exons did not find other homologous exons in any other gene, suggesting that the exon encoding the cytoplasmic region is only present in β2 molecules. Second, the β1 domain of β2 molecules is extremely conserved, whereas it is the most polymorphic domain in classical β1 molecules (Fig. 1C). We detected only a single amino acid substitution among the studied mouse haplotypes (b, d, a/k, and g7) and none among the rat haplotypes (n, l, lv1, and c). Furthermore, the interspecies degree of conservation between the β1 domain of RT1-Db2 and H2-Eb2 is much greater than between the β1 domain of classical RT1-Db1 and H2-Eb1. In contrast to the β1 domains, the β2 domains of rat RT1-Db2 and mouse H2-Eb2 contain polymorphic residues, and the degree of variability is comparable to that of the classical β1 domains. The amino acids of the classical β2 domain described to make direct contacts with the α2 domain of the α-chain to form MHC heterodimers that are annotated in the Conserved Domain Database (29) are conserved in β2 molecules, with the exception of position 153 where β2 molecules have a tyrosine and β1 molecules have a phenylalanine (Fig. 1C, bold type). A tyrosine at this position is also found in DMb molecules of rat, mouse, and human. Moreover, the mouse H2-Eb2 molecule of the b haplotype contains a glycine-to-arginine substitution at position 150. Last, both RT1-Db2 and H2-Eb2 lack the N-glycosylation site (NXS/T), which is present in all classical MHC β-chains (Fig. 1C, box).

RT-PCR analysis showed that RT1-Db2 and H2-Eb2 mRNAs are present in rat and mouse splenocytes, although the levels detected appeared lower compared with those of RT1-Db1 and H2-Eb1 (Fig. 2A). To obtain a more quantitative estimation of the relative expression levels of these molecules among different cells of the immune system, semiquantitative real-time PCR was carried out with sorted cells from the spleen (Table I). The abundance of RT1-Db2 transcripts shows a similar pattern as RT1-Db1, with higher levels in B cells than in total splenocytes and CD11b/c⁺ cells (containing monocytes, granulocytes, macrophages, and dendritic cells). RT1-Db2 mRNA was also detected in T cells. However, its levels were much lower compared with total splenocytes. Similarly, the mRNA levels of H2-Eb1 and H2-Eb2 were also higher in B cells than in total splenocytes.

Because these methods do not allow a direct comparison of the mRNA levels of different genes, we analyzed classical and nonclassical MHC class II RNA abundance using RNAseq libraries of B cells from C57BL/6 mice, which were prepared directly after B cell isolation (ex vivo) or after activation with LPS for 48 h. These libraries were carried out and published by Diaz-Muñoz et al. (16). Importantly, the RNA levels of H2-Eb2 are 31.8-fold (± 1.22 SD) and 46.01-fold (± 2.21 SD) lower than H2-Eb1 in ex vivo–isolated B cells and LPS-activated B cells, respectively (Fig. 2B). Although C57BL/6 mice (b-haplotype) do not express H2-Ea as a result of a 627-bp deletion encompassing its promoter and first exon (30), the RNA levels of H2-Eb1 are very similar to those of H2-Ab1. Therefore, we expect that the expression of H2-Eb2 transcripts will not be affected by the lack of a functional H2-Ea gene in C57BL/6 mice. All nonclassical MHC class II molecules have lower RNA levels compared with the classical ones, with H2-Eb2 being the least abundant. Moreover, as with all of the other MHC molecules, the RNA levels of H2-Eb2 are reduced in B cells after LPS activation. Ribosome profiling libraries, also published by Diaz-Muñoz et al. (16), showed that H2-Eb2 transcripts are actively translated in mouse primary B cells (Fig. 2C).

To assess functional features of RT1-Db2 and H2-Eb2, we generated various cell lines by transducing rat MHC class II chains into L929 and M12.4.1C3 cells (Fig. 3). L929 are mouse fibroblasts and do not express MHC class II molecules (12). In contrast, M12.4.1C3 cells are derived from a BALB/c B cell lymphoma, which was selected for the absence of H2-A and H2-E on the surface (13). Their H2-E molecules are not expressed at the cell surface as a result of mutations in the H2-Eb1 chain; however, H2-Ea^d and the molecules involved in the MHC class II Ag processing and presentation pathway, such as Ii, are expressed. We generated cell lines stably expressing RT1-Dl or RT1-D2 by first transducing RT1-Da^l and then either RT1-Db1^l or RT1-Db2ⁿ (which has the same amino acid sequence as RT1-Db2^l) chain genes. H2-E^d or H2-E2^d was expressed by transducing H2-Ea in L929 cells and then either H2-Eb1 or H2-Eb2 chains from BALB/c mice or only the β-chains in M12.4.1C3 cells. We also generated an H2-E2 of mixed haplotypes by transducing the mouse β2 chain of the b-haplotype (C57BL/6). Once the cell lines were generated, expression at the cell surface of the interrogated molecules was assessed by flow cytometry (Fig. 3).

RT1-Db2 molecules are expressed at the cell surface when cotransduced with RT1-Da (Fig. 3A), because the mAbs OX-17 (specific for RT1-Da) and 14.4.4S (which binds to rat RT1-Da and mouse H2-Ea) detected Da at the cell surface after transduction of RT1-Db2. As expected, RT1-Da was also detected at the cell surface when the classical RT1-Db1 chain CDS was cotransduced, but not when no β-chain CDS was transduced. In addition to OX-17 and 14.4.4S, we stained these cell lines with 3G9A mAb, which detects heterodimers formed by the classical α- and β-chains (RT1-Dl) but not those formed by Da and Db2 (RT1-D2) (Fig. 3A, 3B). We consistently observed that RT1-D2 was detected at the cell surface at slightly lower levels compared with RT1-Dl in L929 cells, whereas the differences were more pronounced in M12.4.1C3 cells (Fig. 3B). To ensure that the differences in cell surface expression were not due to different transduction efficiencies, total cellular expression of the β-chains was monitored with an EGFP reporter gene.

Mouse H2-E2 heterodimers formed by H2-Ea and H2-Eb2 of the d haplotype were detected as well at the cell surface in both cell lines with the 14.4.4S mAb. In L929 mouse fibroblasts, Ea^dEb2^d expression was higher compared with Ea^dEb1^d, whereas expression of Ea^dEb1^d was >2-fold higher than Ea^dEb2^d in M12.4.1C3 cells, which is reminiscent of our findings with rat molecules. In sharp contrast to this, the expression of Ea^dEb2^b heterodimers of mixed haplotypes was barely detectable, if at all (Fig. 3C).

Because H2-Eb2^d has a glycine-to-arginine substitution at position 150 within the motif described to make contact with the α-chain (3), we hypothesized that this residue could be responsible for its extremely low cell surface expression. Indeed, we observed a cell surface expression that was similar to that of H2-Eb2^b when we mutated the arginine to glycine in this position in H2-Eb2^d (H2-Eb2^{d R150G}) and expressed it with H2-Ea in L929 cells (Fig. 3D).

To address the influence of Ii on the cell surface expression of the investigated MHC class II molecules, L929 cells expressing DaDb1 or DaDb2 were subsequently transduced with the Ii CDS. The vector used for transduction contains an EGFP reporter 3′ to an internal ribosomal entry site, which enabled the comparison of MHC class II cell surface expression in cells having (EGFP⁺) or lacking (EGFP⁻) Ii (Fig. 3E). In this case, the vectors encoding the β-chains used to generate the MHC class II–expressing L929 cells did not contain the EGFP reporter. Detection of both heterodimers at the cell surface was increased after expression of Ii. However, the increase observed in DaDb1 was greater than that of DaDb2 (average fold increase ± [SD] in cell surface expression upon Ii expression of three independent experiments was 1.45 ± 0.01 and 1.21 ± 0.08, respectively, when stained with OX-17 (p = 0.0062, t test) and 1.46 ± 0.03 and 1.26 ± 0.07, respectively, when stained with 14.4.4S (p = 0.0089, t test). In this context, it is important to point out that changes in cell surface MHC class II expression after transduction of Ii in L929 cells are clearly weaker for RT1-D^l (DaDb1) than for the other MHC isotype of the same haplotype, RT1-B^l (BaBb) (14).

MHC class II molecules isolated from cell extracts vary in their capacity to enter a conformation that resists dissociation in 2% SDS–containing loading buffer when incubated at room temperature (31, 32). To assess whether heterodimers formed by β2 molecules can enter such an SDS-resistant state, as well as to interrogate cell surface expression of the investigated molecules, we immunoprecipitated cell surface–biotinylated MHC class II molecules of M12.3.1C4-transduced cell lines and divided the precipitates into two parts: one part was incubated at 100°C for 5 min (B), and the other part was kept for 1 h at room temperature (NB) (Fig. 4). The immunoprecipitates of the classical heterodimers (rat DaDb1 and mouse EaEb1) that had been incubated at 100°C (B) show two bands corresponding to the α-chain (higher band) and the β1-chain (lower band). The B lanes of cells expressing DaDb2 or EaEb2 heterodimers show one band corresponding to the α-chain and a band that runs faster than that of the classical β1-chains. The predicted amino acid sequences of β2 molecules are longer than those of classical β1 chains; however, they lack N-glycosylation sites and, therefore, they can be predicted to run faster in SDS-PAGE than β1 chains. These results confirm that rat RT1-Db2 and mouse H2-Eb2^d molecules pair at the cell surface with RT1-Da and H2-Ea^d, respectively, because the proteins visualized are derived from the cell surface, and all of the immunoprecipitations were carried out with Abs that bind to the α-chains.

Furthermore, SDS-resistant dimers (∼55 kDa, in NB lanes) were readily detected in M12.3.1C4 cells transduced with the classical rat RT1-DaDb1 and mouse H2-Eb1^d molecules, but not with β2 molecules.

To study Da intracellular localization when paired with Db1 or Db2 molecules, we stained transduced L929 and M12.4.1C3 cells on the cell surface or intracellularly with the mAb OX-17 and analyzed them by confocal microscopy. OX-17 only recognizes Da when paired to β-chains (Fig. 5A). The localization of Da appears very similar, regardless of the β-chain that was cotransduced on the cell surface and intracellularly. Intracellular staining of L929 cells, which have a much larger cytoplasm than M12.4.1C3 cells, revealed that Da localizes in an intracellular compartment whose appearance is that of the endoplasmic reticulum.

In addition to these experiments and because of the lack of a specific Ab that detects Db2, but not Db1, molecules, we fused YFP to the cytoplasmic region of both β-chains. Using this method, we observed that the cellular localization of the Db1 and Db2 chains differs (Fig. 5B). Db2-YFP fusion proteins show a more condensed localization in particular intracellular compartments compared with Db1-YFP in both cell lines. The identity of this intracellular compartment remains to be elucidated.

Intracellular staining revealed that Db1-YFP colocalizes with Da in both cell lines. The colocalization of Db2-YFP with Da is also apparent via intracellular staining of M12.4.1C3 cells, whereas it is not obvious in L929 cells.

Cell surface flow cytometry analyses of the cells transduced with the YFP fusion proteins showed that both β-chains are also expressed at the cell surface, together with Da, in both cell lines (data not shown). In addition, we ruled out a misinterpretation of the YFP signal observed by confocal microscopy due to proteolytic cleavage of the YFP molecules from the β-chains, because YFP protein detection by Western blot showed bands with molecular weights that correspond to the complete fusion proteins (Fig. 5C, data shown for L929). However, because of the lack of a reagent that directly detects β2, but not β1, molecules in primary cells, we cannot be certain that the intracellular localization of β2 molecules revealed by the YFP fusion protein is the true localization of endogenous β2 molecules or whether it is a consequence of expressing YFP.

To test the capacity of the heterodimers composed of rat and mouse β2 molecules to present superantigens and to compare it with that of the classical class II molecules, we cultured L929 transduced cells with the Vβ8.2⁺ T cell hybridoma 53/4 (14) and different concentrations of two superantigens; 24 h later, we measured IL-2 secreted into the culture media by ELISA (Fig. 6). We used L929 cells transduced only with α-chains as controls. The superantigens added to the cultures were Mycoplasma arthritidis superantigen, a strain originally isolated from an arthritic rat, which causes arthritis in this species (33, 34), and YPM, a virulence factor produced by the pathogenic Y. pseudotuberculosis (35).

Fig. 6A shows the response to Mycoplasma arthritidis superantigen or YPM presented by DaDb1 or DaDb2 heterodimers. Their response to Mycoplasma arthritidis superantigen was very similar. In contrast, the response to YPM varied strongly between the presenting heterodimers. EC₅₀, as well as the differential threshold of the response, revealed a 10,000-fold higher sensitivity of the T cell hybridoma to YPM presented by DaDb2 compared with the classical DaDb1 MHC molecule.

A very similar picture can be drawn for the Mycoplasma arthritidis superantigen and YPM response to mouse heterodimers. The response to Mycoplasma arthritidis superantigen presented by EaEb1 and EaEb2 from the d haplotype is very similar. In the case of heterodimers of mixed haplotypes (d and b), the poor response to Mycoplasma arthritidis superantigen presented by Ea^dEb2^b is most likely due to its very low cell surface expression (Fig. 3C), because when cells transduced with the Ea^dEb2^{b R150} mutant were used as APCs, the response to Mycoplasma arthritidis superantigen was similar to that of EaEb2 from the d haplotype (Fig. 6C). As for rat molecules, the response to YPM was much superior when the superantigen was presented by EaEb2 heterodimers compared with EaEb1, which failed to present YPM at the highest concentrations in our assays (regardless of the haplotypes). Remarkably, even the YPM response to heterodimers formed by Ea^dEb2^b of mixed haplotypes was much better than that of the classical heterodimers, despite its extremely low cell surface expression (Fig. 3C).

The predicted amino acid sequences of rat RT1-Db2 and mouse H2-Eb2 revealed an intriguing feature: the lack of N-glycosylation. To address whether this is the reason why cell surface expression of heterodimers with β2 molecules was generally lower compared with heterodimers with the classical β1 chains in the studied transduced cell lines, we transplanted the N-glycosylation site from Db1 into Db2 (Db2 +N) and the absence of N-glycosylation of Db2 into Db1 (Db1 −N) by site-directed mutagenesis. These mutant β-chains were transduced into L929 cells, and cell surface expression was analyzed by flow cytometry (Fig. 7A). The loss of N-glycosylation in Db1 results in ∼2-fold reduced cell surface expression compared with intact Db1, whereas, for Db2, introduction of N-glycosylation resulted in enhanced cell surface expression compared with the wild-type. The total protein expression of all β-chains was monitored with the expression of EGFP as described above, ensuring that differences observed in cell surface expression did not reflect variations in total protein expression of the investigated molecules.

We also analyzed Mycoplasma arthritidis superantigen and YPM superantigen presentation capacity of these N-glycosylation mutants (Fig. 7B). Importantly, the extraordinary response to YPM presented by Db2 molecules was not hindered by the addition of N-glycosylation, because the differences between the mutants and the original forms are only minimal. The slightly better presentation by the N-glycosylated forms compared with their nonglycosylated counterparts could well reflect their higher cell surface expression (Fig. 7A).

The nonclassical RT1-Db2 and H2-Eb2 MHC class II genes were identified in early studies of molecular genetics of the MHC class II region of mice and rats (7–9). Nonetheless, the lack of homology between the exons that encode the cytoplasmic region of the nonclassical β2 genes and those of the classical MHC class II β1 chains precluded the identification of the complete CDS of these nonclassical molecules at that time. The availability of the complete rat MHC (6) and the rat and mouse genomes, as well as the expressed sequence tag database, allowed us to identify the complete CDSs of these nonclassical β2 chains.

Nonclassical RT1-Db2 and H2-Eb2 molecules share important features with their classical counterparts; however, they also show significant differences, which might indicate a function distinct from that of a classical Ag-presenting molecule. In this study, we transduced these molecules in cell lines to study them. We show that RT1-Db2 and H2-Eb2 chains pair with their respective α-chains and are expressed at the cell surface of the transduced cell lines where they can present superantigens to T cells. In a recent publication, Tuncel et al. (36) carried out immunoprecipitation with the OX-17 mAb using whole-cell extracts from lymph nodes from DA rats, and the congenic DA.H1R61 strain as well, and analyzed the precipitated molecules by mass spectrometry. OX-17 mAb precipitates Da molecules bound to β-chains (23). Importantly, these experiments identified peptides of the classical Db1 chain, as well as Db2-specific peptides. Thereby, they demonstrated the expression of RT1-Db2 protein and the formation of DaDb2 heterodimer (RT1-D2) molecules in rat primary cells, although it remains to be demonstrated that RT1-D2 (or mouse H2-E2) molecules are expressed at the cell surface of primary cells. Mouse (with the exception of mouse H2-Eb2 of the b haplotype) and rat β2 molecules have the residues described for classical β1 molecules to make contacts with the α-chain and with CD4 (3). Of these, the only nonconserved residue in β2 molecules is a tyrosine at position 154, where classical β1 molecules have a phenylalanine. This substitution results in the presence of a hydrophilic amino acid (tyrosine) in β2 molecules where the β1 chains have a hydrophobic residue (phenylalanine) and could affect the interaction with the α-chain. This might be a reason why we observed a reduced cell surface expression of heterodimers formed by β2 molecules compared with the classical ones in the transduced cell lines. We studied β2 molecules of two mouse haplotypes: d (BALB/c strain) and b (C57BL/6 strain). Although H2-Eb2^d was expressed with H2-Ea^d to a similar extent as the rat heterodimer formed by RT1-Da and RT1-Db2, H2-Eb2^b expression at the cell surface with H2-Ea^d was very low. This reduced cell surface expression could be rescued by mutating the arginine in position 150 to a glycine (the residue present in all of the other studied mouse and rat β2 haplotypes, as well as classical β1 molecules).

We chose the M12.4.1C3 cell line (derived from BALB/c mice, d haplotype) to study the transduced class II molecules in a cell that contains other molecules important for Ag processing and presentation. These cells were generated by negative immunoselection using Abs against H2-A and H2-E (34-5-3, 14-4-4, and 34-1-4) after inducing mutations by irradiating the cells (13). These cells do not express H2-E molecules at the cell surface as a result of a mutation in their H2-Eb1 chain, although their H2-Ea chain does not contain any mutations that precludes them from being expressed at the cell surface, but in the presence of the mutated H2-Eb1 chain remain intracellular. We analyzed the CDS and the sequences adjacent to the start and stop codons of H2-Eb2 in M12.4.1C3 by RT-PCR and direct sequencing (data not shown), which confirmed that H2-Eb2 is expressed at the RNA level in these cells and that its predicted amino acid sequence is the same as the H2-Eb2^d haplotype. Based on our data using the transductant cell lines, we expected that if H2-Eb2^d is not mutated in M12.4.1C3 cells, it would have the potential to be expressed at the cell surface with the endogenous H2-Ea chain. However, this was not the case, as shown by flow cytometry analysis (Fig. 3A) and by YPM superantigen presentation assays in which we used untransduced M12.4.1C3 cells as APCs and observed a minute or no response by the T cell hybridoma (data not shown). A possible reason why M12.4.1C3 cells do not express their endogenous H2-Eb2 molecules at the cell surface is that, although H2-Eb2 polypeptides could be expressed, their levels would be much lower compared with the classical mutated Eb1 (as expected from the data on RNA abundance in primary B cells), and these low levels would not be enough to outcompete the mutated Eb1, which still binds to H2-Ea but cannot be translocated to the cell membrane (37). The system that we use to transduce class II molecules is driven by the retroviral promoter and contains only the CDS of the transduced molecules; therefore, any gene that we transduce is overexpressed and, thus, in our cell lines, Eb2 molecules could outcompete the endogenous mutated H2-Eb1 molecules for H2-Ea binding.

Despite their similarities to their classical counterparts, nonclassical β2 molecules differ in very important aspects. First, the RNA levels of β2 molecules in primary cells are much lower compared with the classical β1 molecules. Braunstein and Germain (9) also showed that expression of the mouse β2 molecule was not induced after stimulation with IFN-γ. The promoter regions of RT1-Db2 and H2-Eb2 are highly conserved between them and show only few differences compared with the classical β1 promoters. These few differences could be responsible for the lower RNA levels of β2 genes compared with β1 genes. Nonetheless, the β2 genes differ completely in their 3′UTR compared with the classical β1 genes, because the cytoplasmic regions and 3′UTRs are encoded by nonhomologous exons. Therefore, the fact that these genes have completely different 3′UTRs could also account for their different RNA abundances.

Second, the β1 domain of β2 molecules shows minimal polymorphism [this study; (9, 36)], whereas most of the polymorphic residues of the classical β1 molecules are concentrated in this domain. In contrast to the conservation of the β1 domain of β2 molecules, we observed a similar degree of polymorphism in the β2 domains of β2 molecules as in classical β1 molecules, at least in the analyzed haplotypes. Moreover, the β1 domain of β2 molecules is more conserved between rat and mice than is the β1 domain of β1 molecules (Fig. 1C) and, interestingly, the β1 domain of H2-Eb2 is not more similar to the β1 domain of the classical H2-Eb1 molecules than it is to other class II β1 chains (9). One striking feature of nonclassical β2 molecules is the loss of histidine 81, which was found to be crucial for high affinity binding of peptides (38), as well as for zinc-coordinate binding of the superantigen staphylococcal enterotoxin A (39). Furthermore, histidine 81 is conserved in classical MHC class II β-chains (14), as well as in the β-chains of HLA-DM and HLA-DO. Taken altogether, this high degree of conservation of the β1 domain of β2 molecules and the divergence from the β1 domain of their classical β1 could be the result of a selective pressure to ensure the presentation of a particular class of Ags by these nonclassical molecules or it could reflect a conserved role in a different function, such as Ag processing and presentation, similar to the role of the nonclassical nonpolymorphic DM molecules. If the lack of polymorphism of nonclassical β2 molecules is a reflection of the level of conservation needed to perform their biological function and the need to form a heterodimer with the classical α-chain, this would be consistent with the known lack of polymorphism of classical RT1-Da and H2-Ea chains and loss of the capacity to be expressed at high levels at the cell surface in the H2-Ea⁻ b-haplotype. It would be interesting to look for H2-Eb2 alleles in other H2-Ea⁻ strains and to determine whether similar correlations can be found, which then might be interpreted as a further indicator of a functional relationship between H2-Ea and H2-Eb2 molecules.

Third, these nonclassical class II molecules have a completely different cytoplasmic region compared with their classical counterparts, which is also longer than the classical cytoplasmic region. Importantly, Tuncel et al. (36) also detected peptides specific for the cytoplasmic region of RT1-Db2 molecules in their immunoprecipitation and mass spectrometry analysis of rat lymph node protein extracts. As mentioned in the 13Results, we have not found homologs for the cytoplasmic region in any other molecules, and its function remains to be elucidated. One possible function is the involvement of this region in signaling transduction because it was shown for classical MHC class II cytoplasmic regions (40, 41), although, as a result of the lack of homology with the classical cytoplasmic regions, a role in the activation of different factors could be expected. Another possible function of the cytoplasmic region is to dictate the different intracellular localization of the β2 molecules compared with β1 molecules, because it was reported that amino acid residues in the cytoplasmic region of Ii constitute the motif responsible for driving their localization in endocytic compartments (4).

Fourth, expression of Ii on L929 fibroblasts resulted in a greater cell surface expression of classical RT1-D1 compared with RT1-D2. Keeping in mind that we overexpress all of the molecules in our system, this differential increase at the cell surface might reflect the fact that β2 molecules do not function as classical Ag-presenting molecules, despite the fact that we observe them at the cell surface, and the coexpression of Ii in L929 enhances its translocation to the cell surface. Moreover, the increase in RT1-D2 at the cell surface when the Ii is present could be an indirect Ii-independent effect.

Fifth, RT1-Db2 and H2-Eb2 molecules lack the N-glycosylation site that is present in most molecules expressed at the cell surface, classical β1 molecules, and in the β-chains of nonclassical H2-Ob chains and Mb chains of human and mice (H2-Mb1). The rat RT1-Mb chain lacks N-glycosylation. The absence of N-glycosylation could be the reason why we observed generally lower cell surface expression levels of heterodimers composed of DaDb2 and EaEb2 chains compared with the classical heterodimers in the different transduced cell lines, because we observed a higher cell surface expression when Db2 mutant molecules containing an N-glycosylation site were expressed. Given the unknown biological function of Db2 and Eb2, it is also possible that the lack of N-glycosylation is important for executing such functions.

Last, surprisingly we found that heterodimers formed by the classical α-chains and the nonclassical β2 chains had an extraordinary capacity to present YPM superantigen to a T cell hybridoma compared with the classical RT1-D1 and H2-E1 molecules using various transduced cell lines. The functional relevance of this finding remains to be determined, but it suggests that, if RT1-D2 and H2-E2 are expressed at the cell surface under physiological conditions, they might be important players during infection with Y. pseudotuberculosis or present other unknown Ags or superantigens.

The molecular basis of the very low threshold of T cell activation upon YPM presentation by RT1-D2– and H2-E2–expressing cell lines remains unclear, given the lack of structural information about β2-containing class II molecules. Even speculation about a possible binding mode of YPM to MHC class II and its interaction with TCR is nearly impossible, given that results from YPM mutagenesis and experiments with YPM-derived peptides did not allow the identification of YPM surface regions especially critical for YPM-mediated T cell activation (31). Indeed, despite the fact that YPM has a slight structural similarity to the TNF family and some virus capsid proteins (42), it represents its own conserved domain superfamily, as indicated by a recent search of the National Center for Biotechnology Information’s Conserved Domain Database (29) (result from BLAST search September 2015).

Humans possess several DRb genes; however, compared with the nonclassical β2 chains examined in this study, they differ in terms of genetic structure (none possesses a paralogous exon that encodes the novel cytoplasmic region) and protein sequence (they contain N-glycosylation sites and are highly polymorphic in their β1 domain). Therefore, it is expected that their function differs considerably from Db2 and Eb2. Nevertheless, the identification of the latter two is of importance given that mice and rats are widely used as model organisms in biomedical research and to study the immune system. Thus, our findings are important to better understand these animal models, at least in studies involving Y. pseudotuberculosis. Further studies might identify other Ags that are presented by these newly characterized MHC class II molecules. Finally, it is also possible that these molecules might have taken over tasks beyond Ag presentation or even an immune function.

E.M.-C. and T.H. thank the reviewers for helpful comments.

The authors have no financial conflicts of interest.