This Article Abstract Full Text (PDF) Supplemental Research Data Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Takada, T. Articles by Lindahl, K. F. Search for Related Content PubMed PubMed Citation Articles by Takada, T. Articles by Lindahl, K. F. Social Bookmarking                   What's this?

LETTER

Species-Specific Class I Gene Expansions Formed the Telomeric 1 Mb of the Mouse Major Histocompatibility Complex

¹Howard Hughes Medical Institute, ²Center for Immunology, and ³McDermott Center for Human Growth and Development, University of Texas Southwestern Medical Center, Dallas, Texas 75390, USA

	ABSTRACT

Top ABSTRACT RESULTS DISCUSSION METHODS WEB SITE REFERENCES NOTE ADDED IN PROOF REFERENCES

 
We have determined the complete sequence of 951,695 bp from the class I region of H2, the mouse major histocompatibility complex (Mhc) from strain 129/Sv (haplotype bc). The sequence contains 26 genes. The sequence spans from the last 50 kb of the H2-T region, including 2 class I genes and 3 class I pesudogenes, and includes the H2-M region up to Gabbr1. A 500-kb stretch of the H2-M region contains 9 class I genes and 4 pseudogenes, which fall into two subfamilies, M1 and M10, distinct from other mouse class I genes. This M1/M10 class I gene-cluster is separated from the centromeric H2-T and the telomeric H2-M4, -5 and -6 class I genes by "nonclass I genes". Comparison with the corresponding 853-kb region of the human Mhc, which includes the HLA-A region, shows a mosaic of conserved regions of orthologous nonclass I genes separated by regions of species-specific expansion of paralogous Mhc class I genes. The analysis of this mosaic structure illuminates the dynamic evolution of the Mhc class I region among mammals and provides evidence for the framework hypothesis.

[Supplemental material is available online at www.genome.org. The sequence data from this study have been submitted to GenBank under accession nos. AC005413, AC005665, AF532111–AF532117. A preliminary draft sequence was earlier submitted as AC002615 and replaced this year by NT002615.]

The prototypical mammalian Mhc encodes highly polymorphic transmembrane proteins, the class I and class II MHC molecules, that bind peptide fragments derived from intra- and extracellular proteins, respectively, and present them on the cell surface for surveillance by T lymphocytes (Yewdell and Bennink 2001; Lennon-Duménil et al. 2002). Several other proteins involved in antigen processing and peptide loading and transport are also encoded in the Mhc, as are components of the complement system and several lymphokines. However, about half the genes in this 3.6-Mb (megabase) complex have no obvious immune function (Beck and Trowsdale, 2000).

Comparative physical mapping revealed extensive conservation of synteny within the Mhc, but in a patchwork pattern: so-called framework regions of near complete conservation of content and orientation of genes, mainly of nonimmune function, alternating with regions of species-specific expansion of rapidly evolving MHC class I and, to a lesser extent, class II genes (Amadou 1999; Amadou et al. 1999). All mammals have one or more polymorphic MHC class Ia genes that function in classical antigen presentation. The investigated species have another half dozen or so nonclassical, or class Ib, genes; rats and mice, however, have on the order of 30 or more intact class Ib genes (Amadou et al. 1999; Günther and Walter, 2001). The class Ib genes are generally distinguished by their low polymorphism and often have a limited tissue expression; the proteins they encode usually have shorter cytoplasmic tails, and some lack consensus residues associated with peptide binding (Stroynowski and Fischer Lindahl 1994). While several class Ib molecules are now known to bind peptides, their function is fully understood in only a few cases. HLA-E and its mouse homolog, Qa1 (encoded by H2-T23), display leader peptides from class Ia molecules and binds to inhibitory natural killer cell receptors (Yeager et al. 1997; Braud et al. 1998). The human MICA and MICB are induced by stress and are ligands for the activating immune receptor NKG2D (Spies 2002). Rodents have no MIC genes in the Mhc (Bahram et al. 1994), but they have other class Ib molecules not present in human, such as M3, which presents N-formylated peptides to cytotoxic T lymphocytes. M3 provides early protection against bacterial infection, and it can also serve as a histocompatibility antigen by presenting N-formylated polymorphic peptides from mitochondria (Fischer Lindahl et al. 1997).

The classical part of the mouse Mhc, including the H2-K and H2-D class I regions, were cloned and sequenced from strain 129 (haplotype bc; Kumánovics et al. 2002). The class I region of H2 contains more than 60 Mhc class I genes and pseudogenes, in about 2 Mb. It is traditionally divided into the H2-Q, H2-T, and H2-M regions, which were originally defined by recombination. The centromeric 1 Mb of the H2-M region is rich in class I genes, whereas the telomeric 1 Mb is rich in olfactory receptor genes and contains just two functional class I genes, H2-M3 and H2-M2 (Amadou et al. 1999). Here we report the finished sequence of the class I-rich part of the H2-M region.

	RESULTS

Top ABSTRACT RESULTS DISCUSSION METHODS WEB SITE REFERENCES NOTE ADDED IN PROOF REFERENCES

 
Genomic Sequencing of the Centromeric H2-M Region
We previously constructed a physical map of the H2-M region on Chr 17, based on a contig of bacterial artificial chromosome (BAC) clones from strain 129/SvJ (haplotype bc; Jones et al. 1999). Eight clones that should span the centromeric H2-M region were selected from this BAC contig. In the course of sequencing, it became clear that clones 585c7 and 10i1 did not overlap. The bridging 26,962 bp XhoI fragment from clone 255d16 (Jones et al. 1999) was sequenced to fill that gap.

The final 951,695 bp sequence assembly was verified in several ways, and no inconsistency was found. First, a sequence-based, virtual restriction enzyme digest was checked against restriction digests of the individual clones (data not shown); second, the position and sequence of previously mapped genes, pseudogenes, and STS markers were checked (Fig. 1). These markers included eight H2-M class I genes (M1 and M4-M10) and six nonclass I genes (Zfp173 or Trim26, Trim10 or Herf1, Mog, and Tctex4, -5, and -6), two microsatellite markers (D17Mit47 and D17Mit125), and all the suitable, nonrepetitive, BAC end sequences determined in the previous mapping study (Jones et al. 1999).

View larger version (25K): [in this window] [in a new window]   Figure 1. Gene, isochore, and repeat content in the centromeric H2-M region. Genes and markers are indicated on top by name, with already known genes in blue, newly identified class I genes in black, and microsatellites in green. The bar marks framework (FW) regions and class I (CI) expansion zones in light and dark blue. The GC-content in percentage was calculated using an 8-kb window/1-kb shift (grey line) and a 40-kb window/10-kb shift (red line) sliding window analysis. The class I gene-rich region from H2-M10.1 to H2-M10.8 forms an L2 isochore with an average of 42% GC, and it is surrounded by H1 isochores with an average of 45% GC. The distribution of repetitive elements is indicated in the bottom part of the figure. Red and blue boxes mark regions particularly rich in LINEs and SINEs, respectively. The class I gene orientation is indicated to help to understand the duplications and inversions that may have given rise to the present day H2-M1 (light and dark blue) and H2-M10 (light and dark orange) families.

The distribution of LINE (long interspersed nuclear element), SINE (short interspersed nuclear element), and LTR (long terminal repeat) genome-wide repeat-elements, as defined by the RepeatMasker program, is shown in the bottom part of Figure 1. The sequence contains 18.9% LINE repeats (18.68% LINE1, 0.17% LINE2, and 0.01% L3/CR1), 8.1% SINE repeats (3.59% B1, 4.18% B2-B4, 0.15% ID, and 0.14% MIR), 9.2% LTR retroposons (3.53% MaLR, 0.14% ERVL, 0.71% ERV class I, and 2.31% ERV class II), and 0.5% DNA transposons. Supplemental Research Data shows the repeat distribution in greater detail, including the position of simple sequence repeats.

Two distinct regions, boxed in blue on Figure 1, have a high density of SINEs and low density of LINEs. The 50 kb in the centromeric H1 isochore contains 17.2% SINEs and 1.8% LINEs; the 250-kb telomeric H1 isochore contains 16.3% SINEs and 3.4% LINEs. The central L2 isochore of 510 kb, boxed in red, has only 4.0% SINEs but 29.7% LINEs. The LTR retroposons and DNA transposons are uniformly distributed.

Gene Organization
Table 1 summarizes the gene and pseudogene content and the Supplemental Research Data shows the features of the sequenced region in detail. Exons of known genes were identified from mRNA data, whereas exons of predicted genes were identified by the prediction programs and from EST data, if available. We defined the subregions based on the gene content: The first 52 kb represents the telomeric part of the H2-T class I region. The transition from the H2-T to the H2-M region spans from 53 kb to 163 kb of the entire sequence (Supplemental Research Data). It contains several repeated fragments and presumed pseudogenes with similarity to genes encoding various nuclear proteins: Arl1, Doc-1, LSm5, Histone H3, and Hmg14. An Hmg14 (high-mobility-group chromatin protein) fragment was previously mapped to this region by hybridization (Johnson et al. 1992). We found a homolog of the human RPP21 subunit of the RNase P holoenzyme in this region, next to Trim39 (or Rnf23; Jarrous et al. 2001).

View this table: [in this window] [in a new window]   Table 1. Genes and Pseudogenes in the Mouse Telomeric Class I Region

RING Finger and Other Nonclass I Genes
RING finger domains are defined by the consensus sequence CX2CX(9-39)CX(1-3)HX(2-3)C/HX2CX(4-48)CX2C, where the Cys and His residues bind zinc (Saurin et al. 1996). The RING finger proteins interact with other proteins in a range of biological processes, including development, oncogenesis, apoptosis, and viral replication. Five previously identified RING finger genes (Trim39, Trim26, Trim15, Trim10, and Trim31; Orimo et al. 2000; Reymond et al. 2001) were localized to the H2-M region, and two new RING finger genes (ZfpU and ZfpW) were predicted by GenScan and EST database searches (Table 1). Six of these form a cluster (Trim26, Trim15, Trim10, ZfpU, Trim31, and ZfpW) in the 120-kb region between H2-M10.8 and Ppp1r11 (or Tctex5; Supplemental Research Data). These genes are under further investigation (C. Amadou, T. Takada, A. Kumánovics, and K. Fischer Lindahl, in prep.). The corresponding 307-kb region around Trim26 (or Zfp173) from pig has been sequenced, and four RING finger genes were detected. The pig has an extra RING finger gene centromeric to Trim26 (or Zfp173) compared with mouse and man (Renard et al. 2001).

Trim39, which encodes a zinc-binding protein abundant in testis (Orimo et al. 2000), and Rpp21, which encodes a subunit of the ribonuclease P holoenzyme, span the 18 kb in the boundary between the H2-T and H2-M class I genes (Table 1).

The 20 kb between ZfpW and H2-M6 contains three "T complex testis-expressed" (Tctex) genes, Tctex5 (or Ppp1r11), -6, and -4 (Ha et al. 1991; Table 1). The 80-kb telomeric to H2-M5 contains a Krüppel-like zinc finger gene, Zfp57, and two previously mapped and analyzed genes, encoding the myelin-oligodendrocyte glycoprotein gene (Mog; Gardinier et al. 1992; Pham-Dinh et al. 1995), and a receptor for -amino-butyric acid (Gabbr1; Grifa et al. 1998).

Class I Genes
The sequence presented here contains three clusters of Mhc class I genes. In the first 52 kb of the centromeric H1 isochore (Fig. 1), which belongs to the H2-T region, there are two genes, three pseudogenes and some fragments. Because allelic relationships are difficult to establish for the H2-T class I genes, which differ in number between haplotypes and have been incompletely characterized, we have tentatively named these genes H2-Tw with a number counting from the telomeric end (w for work in progress). Tw5^bc is most similar to the H2-Bl (for blastocyst) gene previously sequenced from a strain 129/SvJ cosmid (Sipes et al. 1996) and mapped to the center of the H2-T region (Amadou et al. 1999). H2-Bl is expressed in mouse preimplantation blastocysts and in the placenta. Tw3^bc is very similar to the H2-T3 gene already sequenced from the b, d, and k haplotypes (Obata et al. 1985; Pontarotti et al. 1986; Mashimo et al. 1992), where it encodes the TL antigen expressed in the intestinal epithelium (Wu et al. 1991). Both Tw5 and Tw3 encode the consensus residues characteristic of peptide-binding class I molecules (Fig. 2), and they are more closely related to classical class Ia genes than to any of the H2-M class I genes (Fig. 3).

View larger version (47K): [in this window] [in a new window]   Figure 2. Alignment of the deduced amino acid sequences of two H2-T proteins and the H2-M1 family with their closest relatives. Tw5 is compared with H2-Blbc of strain 129 (accession no. NM_008199) and Tw3 with T3k of strain C3H (accession no. NM_008208). M1, M9, and M7.2 of haplotype bc are compared with M1 of the b haplotype (accession no. M20985); the purple box highlights the only allelic variation. M9 of the d haplotype (accession no. NM_008205) is identical to M9bc and not shown. The position of the highly conserved cysteine residues are marked in yellow; putative sites for N-linked glycosylation are marked in light blue; residues normally found at the ends of the peptide-binding site in Mhc class Ia molecules are listed below the sequences, with green highlights where conserved and pink where not.

View larger version (16K): [in this window] [in a new window]   Figure 3. Neighbor-joining tree, obtained by using MEGA2 (Kumar et al. 2001), is based on the amino acid sequence of the 3 domain of all H2-M class Ib molecules and Tw3 and Tw5. The classical class Ia proteins, H2-Kbc (accession no. AF100956) and H2-Db (accession no. AAA39601) were included for comparison. The sequence of H2-M3 is from the b haplotype (accession no. NM_013819) and that of H2-M2 is from the d haplotype (accession no. M26156); both genes are telomeric to Gabbr1 (Amadou et al. 1999).

The central L2 isochore (Fig. 1) contains nine H2-M class I genes, four pseudogenes and several fragments in 520 kb. These class I genes can be divided into two families, named after previously sequenced members: The M1 family of three genes, two pseudogenes and two fragments, flanked by five members of the M10 family on the centromeric side and three on the telomeric side (Fig. 1 and Supplemental Research Data). To clarify the relationships among the ten intact H2-M genes described here plus H2-M3 and H2-M2, which are located telomeric to the region discussed here, we have constructed a neighbor-joining tree based on the 3 domains encoded by exon 4 (Fig. 3). The genes from the L2 isochore formed two distinct families, a tight cluster of the M10 family and a more open cluster of the M1 family. M2, M3, and M5 were closer to the class Ia genes than the M1 and M10 families and did not form a tight family.

Of the intact members of the M1 family, H2-M1 (Singer et al. 1988) and H2-M9 (Arepalli et al. 1998) were described previously, as was the pseudogene H2-M7 and the fragment H2-M8 (Wang and Fischer Lindahl 1993). The newly identified open reading frame, H2-M7.2, is most closely related to the pseudogene, H2-M7; hence the name. Figure 2 aligns the deduced amino acid sequence of the bc allelic forms of these three molecules with M1^b from C57BL/6 and M9^d from BALB/c. M9 is identical between the bc and d haplotypes, and M1 differs between b and bc by only one amino acid (Gly135 to Ala), predicted to be located in the loop before the 2 helix. We chose to name the members of the M10 family by their linear order along the chromosome, just like genes of the H2-Q and H2-T regions were previously named (Klein et al. 1990). The first M10 gene to be cloned came from strain BALB/c (Arepalli et al. 1998). It is now clear that this gene represents the d allele of M10.2; the alleles differ by only three amino acids.

Class Ia molecules share canonical amino acids that interact with the amino- and carboxy-termini of bound peptides and are located at the ends of the peptide binding groove. Of these, only Trp147 is conserved in the M1 family (Fig. 2). Substitution of Trp167 is seen in class I-like molecules, such as M3 and FcRn, in which the A pocket is closed (Burmeister et al. 1994; Wang et al. 1995a). Therefore, if the molecules of the M1 family bind peptides at all, it is not in the classical manner. Like the M1 family of molecules, the M10 proteins have a substitution at position 167 (Trp to Arg), and they lack most consensus residues involved in standard peptide binding. All four cysteine residues that form the essential disulfide bridges in the 2 and 3 domains of class I molecules are conserved in the M1/M10 families.

Expansion of Class I Genes
To understand the organization and evolution of the H2-M region, we compared the 951,695 bp sequence (after masking of genome-wide repeats) to itself as a dot matrix (Fig. 4A). Many regions of similarity were found, represented by dots and lines parallel (direct repeats) or perpendicular (inverted repeats) to the diagonal. Because the Mhc class I genes are similar to each other, they show up as short diagonal lines. For example, in Figure 4A, the similarity between the H2-T and H2-M class I sequences is visible; the H2-T and H2-M class I genes are separated by a 90-kb region of extensive duplication of "chromatin-related" pseudogenes (Figs. 4A,B).

View larger version (40K): [in this window] [in a new window]   Figure 4. Dot matrix analysis of the 951,695 bp sequence of the H2-M region against itself. (A) Segmental duplication and inversion of the class I genes is evident in the central box. This plot was generated with PipMaker, set to compare both strands and show all matches. (B) Self-comparison of the 70-kb region rich in pseudogenes for nuclear proteins. (C) Forward and (D) reverse self-comparison of the 520-kb M1-M10 region. The genome-wide repeats are masked in plots (C) and (D), whereas plots (A) and (B) are not masked.

Comparison of the Telomeric Mhc Class I Region of Mouse and Man
We identified an 853-kb segment from the sequenced human Mhc (accession no. NT_001520; MHC Sequencing Consortium 1999), which represents the region of conserved synteny to our mouse M-region sequence. The dot matrix comparison of these human and mouse sequences (Fig. 5) recognized three well conserved regions: 18 kb including Rpp21 and Trim39 (or Rnf23), 130 kb including Trim26 (or Zfp173) to Tctex4, and 80 kb including Zfp57 to Gabbr1 genes. These regions appear as long diagonal lines, boxed in red, green and blue in Figure 5.

View larger version (26K): [in this window] [in a new window]   Figure 5. Dot matrix analysis of the telomeric class I region of mouse and man. The 951,695-bp sequence of the centromeric H2-M region is compared with 853,185 bp of the HLA region (accession no. NF_001520). Class I genes are red, class I pseudogenes are blue; other genes and pseudogenes are black. The red, green, and blue boxes mark the three framework regions that are conserved syntenic regions of orthologous nonclass I genes, which yield clear diagonal lines on the dot matrix. The class I genes give only short lines of similarity between paralogous genes.

	DISCUSSION

Top ABSTRACT RESULTS DISCUSSION METHODS WEB SITE REFERENCES NOTE ADDED IN PROOF REFERENCES

 
The entire mouse Mhc (H2) has been cloned from strain 129/Sv (Amadou et al. 1999), the strain of choice for gene targeting and deletion. Almost two megabases of sequence, from the H2-K region through the class II/III regions to the H2-D/Q region is already determined (Kumánovics et al. 2003). Here we provide the complete genomic sequence of 951 kb of the telomeric part of the mouse Mhc, from H2-Tw5 through Gabbr1. Another 879 kb of sequence from Gabbr1 through the olfactory receptor region is being analyzed (C. Amadou, R.M. Younger, S. Sims, L.H. Matthews, J. Rogers, A. Ziegler, S. Beck, and K. Fischer Lindahl, in prep.; Younger et al. 2000). The sequencing of the remaining 800 kb, including the H2-T region, is hindered by the highly repetitive nature of the region, but it should, nonetheless, soon be completed (A. Kumánovics, E. P. Jones and K. Fischer Lindahl, unpubl.).

Gene Content: The Class I Genes
Strain 129, haplotype bc, is similar in the centromeric 1.5-Mb, classical part of H2 from H2-K through H2-D to haplotype b (e.g., C57BL/6). For example, the H2-K and H2-D proteins are identical between haplotypes b and bc (Kumánovics et al. 2002). But it was also long known that haplotype bc differs in the telomeric H2, for example, the H2-Q, -T and -M regions, from haplotype b, hence the designation bc (Snell et al. 1971). The number of class I genes in the Q and the T regions differs between these two haplotypes (Fischer Lindahl 1997; Kumánovics et al. 2002). The H2-M region, on the other hand, appears to be stable among the investigated strains (Wang and Fischer Lindahl 1993; Jones et al. 1995; Jones et al. 1999). Moreover, the class I genes are orthologous between the rat and mouse M regions (Wang et al. 1995b; Günther and Walter, 2001).

By sequencing, we identified 12 class I genes from H2-Tw5 to Gabbr1 that have complete open reading frames and are therefore predicted to be functional. In the H2-T region segment, we found two genes and three pseudogenes. H2-Tw3 appears to be an allele of H2-T3, which encodes the serologically defined TL (for thymus-leukemia) antigen (Old et al. 1963; Wu et al. 1991). TL has been studied quite extensively, yet the function remains elusive. The strong interaction between the TL molecule expressed by intestinal epithelial cells and the CD8 homodimer expressed by intraepithelial lymphocytes (IEL) could regulate the behavior of IELs (Leishman et al. 2001).

H2-Tw5 is most similar to H2-Bl (Fig. 2). H2-Bl is expressed in preimplantation embryos and the placenta (Sipes et al. 1996). H2-Bl maps to the center of the H2-T region in haplotype bc (Amadou et al. 1999). Tw5 and Bl are presumably related to each other by one of the strain-specific large-scale duplications that are characteristic of the H2-T region (Teitell et al. 1994). We have to wait for the complete sequence of the H2-T region to define this duplication in strain 129.

The class I genes of the H2-M region fall into three groups, based on the comparison of their 3 domains (Fig. 3). The M1 family has three potentially functional members (M1, M7.2 and M9), and the M10 family has six (M10.1, M10.2, M10.3, M10.5, M10.7 and M10.8; Figs. 2, 3). The rest of the M region class I genes (M2, M3, and M5) do not form a close family (Fig. 3). In the M6–M4–M5 class I cluster, only M5 has the potential to encode a class I molecule; the others have frame-shifts and in-frame stop codons rendering them pseudogenes (Wang and Fischer Lindahl 1993). Unlike in mouse, the rat M4, M5, and M6 all have open reading frames, and transcripts can be detected by RT-PCR (D. Lambracht-Washington and K. Fischer Lindahl, in prep.). In mouse, M5 transcripts have been detected in the thymus by the same method.

The M1 and M10 families encode novel and divergent Mhc class I proteins. Restricted expression pattern and low polymorphism define the nonclassical or class Ib molecules (Stroynowski and Fischer Lindahl 1994). The M1/M10 families conform to both. Their expression appears to be strictly regulated. M1 expression appears to be actively repressed in adult tissues (Howcroft et al. 1996). We did not find convincing M10 expression in mouse embryos or major adult mouse organs (E.P. Jones, unpubl; Arepalli et al. 1998). There is only limited information on the M1/M10 polymorphism (Singer et al. 1988; Arepalli et al. 1998), but based on the three sequences from haplotype d, the M1/M10 genes appear to be oligo- or monomorphic (Fig. 2). Most of the residues forming the M1/M10 antigen binding groove do not follow the class I consensus (Fig. 2), suggesting a function different from the classical antigen presentation. M1/M10 have no equivalent in human or any other species, so far, except in rat.

Gene Content: The Nonclass I Genes
One of the main interests in the Mhc is its association to hundreds of diseases. Sequencing and detailed mapping of the Mhc revealed that the class I regions from human, mouse, rat, and pig encode a large number of nonclass I genes (this study; Amadou et al. 1999; MHC Sequencing Consortium 1999; Chardon et al. 2001; Ioannidu et al. 2001; Renard et al. 2001). Mapping of corneodesmosin (CDSN or S gene), a nonclass I gene expressed in skin, to the class I region provided the first hint that not all class I-linked diseases are necessarily mediated by the class I genes themselves (Zhou and Chaplin 1993). Psoriasis vulgaris, a skin disorder, is associated with alleles of HLA-C, but the gene (or genes) responsible for the disease most likely maps telomeric to HLA-C, where, among others, corneodesmosin maps (Oka et al. 1999; Nair et al. 2000). The presence of nonclass I genes in the class I region not only helps to explain the evolution of the Mhc, but provides new clues to the molecular bases of a large number of diseases.

There are 14 nonclass I genes in our sequence (Table 1). Three of them have known function: Rpp21 is a subunit of the nuclear ribonuclease P holoenzyme, the tRNA processing enzyme (Jarrous et al. 2001). Mog encodes the myelin-oligodendrocyte glycoprotein. Mog protein is specific to the central nervous system and it is the major autoantigen in experimental autoimmune encephalomyelitis (EAE), the widely used animal model for autoimmune demyelinating diseases such as multiple sclerosis (Gardinier et al. 1992; Pham-Dinh et al. 1995). Gabbr1 is a receptor for the neurotransmitter -amino-butyric acid (GABA). Intriguingly, susceptibility loci for multiple sclerosis, epilepsy, and schizophrenia have been suggested to map in the region of GABBR1 (Grifa et al. 1998). Herf1 (hematopoietic RING finger 1) is required for terminal differentiation of erythroid cells (Harada et al. 1999). ZfpRingR or TRIM31 might participate in the retinoid-induced growth arrest of MCF-7 breast carcinoma cells (Dokmanovic et al. 2002). Tctex5 is a regulatory (inhibitor) subunit of protein phosphatase 1 (Ppp1r11; LocusID: 76497).

The other nonclass I genes in our region have no identified function (Table 1). But the Trim26 (or Zfp173) toTctex4 cluster (Table 1, Fig. 5) is an interesting one even without known functions: (1) Overlapping and oppositely oriented genes generate possible sense–antisense gene pairs, such as ZfpW (or HZFw) and Ppp1r11 (Tctex5 or HCGV) at their 3' end UTR (untranslated region), and Tctex6 (or HTEX6) and Tctex4 (or HTEX4) at their 5' end (Coriton et al. 2000). (2) The region has a high gene density generated by the overlapping genes; and, finally, (3) alternative splicing has been detected for most of them (Lepourcelet et al. 1996; Lepourcelet et al. 1998; Coriton et al. 2000). Sense–antisense transcript pairs can form double-stranded RNA duplexes, which are increasingly thought to be an important part of the gene regulatory process. Natural antisense transcripts play a role in RNA interference, genomic imprinting, translational regulation, alternative splicing, X chromosome inactivation, and RNA editing, and finally, sense–antisense gene pairs can exhibit reciprocal expression patterns (Lehner et al. 2002; Shendure and Church 2002). Natural antisense transcripts have been predicted in the Mhc, in the class III region, which is also conserved between human and mouse and has a high gene density (Beck and Trowsdale 2000; Lehner et al. 2002).

Evolution of the Mhc
The Mhc class I and II proteins present antigens. An important question about the Mhc is how the Mhc can maintain the ability to present antigens from a large variety of infectious agents in diverse environments. Analysis of the class I genes showed that the basic structure and function of the Mhc and the class I and class II proteins are conserved throughout evolution (Trowsdale 1995; Ohta et al. 2000). Despite this fundamental conservation, the class I and class II gene sets are replaced with every major radiation as a result of a dynamic birth-and-death process (Klein et al. 1992; Nei et al. 1997; Klein et al. 1998). Orthologous genes are separated by speciation, as opposed to paralogs that are separated by gene duplication. Among mammals, orthologous class I genes are only found within the same order, such as in primates or in rodents (Hughes and Nei 1989; Hughes 1991). The human and mouse class I genes group separately in a phylogenetic tree, in a species-specific manner (Hughes and Nei 1989; Hughes 1991; Yeager et al. 1997; Kumánovics et al. 2003).

The mapping of nonclass I genes in the class I region complements the previous studies on class I evolution by providing evidence that the class I regions are orthologous, even if the class I genes are not (Amadou 1999; Amadou et al. 1999). We compared about half of the Mhc class I region of mouse (951 kb) and human (853 kb), and showed that, unlike the class I genes, all the nonclass I genes are conserved between the two species (Fig. 5). Without the nonclass I genes, the two class I regions cannot be aligned. The use of these nonclass I genes helps to understand the genomic organization and evolution of the Mhc class I region. The species-specific expansions of class I genes occurred in the same framework of nonclass I genes in both man and mouse (Amadou 1999; Amadou et al. 1999), and, most likely, in all other mammals too (Chardon et al. 2001; Ioannidu et al. 2001; Renard et al. 2001; Di Palma et al. 2002). In other words, the class I genes expanded in the same position in the genomes, but from different ancestors. For example, there are class I genes between Trim39 (or Rnf23) and Trim26 (or Zfp173) in both man and mouse, but the expanding gene in mouse was the ancestor of the M1/M10 families occupying 520 kb, whereas in human the sole HLA-92 pseudogene in 120 kb is unrelated to the mouse M1/M10 genes. The 390-kb long HLA-A region expansion occurred between HTEX4 and ZFP57. In mouse, the 37-kb Tctex4–Zfp57 region contains the H2-M4, 5, 6 genes, which are unrelated to HLA-A, G, F and to the pseudogenes from the corresponding human region.

How did the class I expansion happen? The self dot-plot comparison of the M region clearly shows that the M1/M10 segment is the result of a series of segmental duplications followed by a large-scale inversion (Fig. 1 and 4D). The region is rich (29.7%) in L1-like transposons, including not only the usual fragmented L1 sequences (Smit 1999), but also one full length L1. In human, HERV-16 retroviral sequences are thought to mediate duplications by recombination between homologous retroelements (Kulski et al. 1997; Dawkins et al. 1999; Shiina et al. 1999b). Similarly, in the M1/M10 segment the L1 elements are abundant and may have contributed to the expansion.

There is one major difference between man and mouse in this respect: In human, the same duplication unit can explain most of the class I region, whereas in mouse, the segments undergoing expansion are entirely locus specific. That is, we do not find common elements, outside of the class I genes, shared between the K/D/Q, T, M1/M10 and M4/5/6 expansion (this paper; Kumánovics et al. 2002), whereas in human, an ancient duplication unit for the entire class I region can be deduced as a segment containing a MIC gene, a HERV element, and the ancestor of the human class I genes (Dawkins et al. 1999).

The human class I duplication unit includes nonclass I genes, such as the HCG (hemochromatosis candidate gene) series (Pichon et al. 1996). In mouse, the H2-Q region duplication included archain pseudogenes (Kumánovics et al. 2002). The M1/M10 expansion contains no nonclass I genes or gene fragments. There is a 70-kb, highly repetitive segment (Fig. 4B) between H2-Tw1 and Rpp21 (Fig. 4A), which contains only gene fragments and pseudogenes derived from class I and nonclass I genes. Despite the plentiful L1s and duplicated segments, the M-region appears stable among the mouse strains and is largely shared with rats (Jones et al. 1995; Wang et al. 1995b; Jones et al. 1999; Günther and Walter, 2001).

	METHODS

Top ABSTRACT RESULTS DISCUSSION METHODS WEB SITE REFERENCES NOTE ADDED IN PROOF REFERENCES

 
Sequencing of BAC Clones
BAC clones were all from the CITB-CJ7-B library of strain 129/SvJ, distributed by Research Genetics. The minimal overlap fragment between clones 585c7 and 10i1 was isolated from clone 255d16 by treatment with XhoI and used to prepare sequencing templates.

Mechanically sheared BAC DNA fragments, 1–4-kb long, were ligated into M13 or pUC18 vectors (Bankier et al. 1987) and sequenced with M13 reverse primer until greater than 6x random shotgun coverage was achieved, based on the estimated BAC clone size. The DNA sequence was determined by the enzymatic dideoxy chain termination chemistry with automated ABI 377 or ABI 3700 sequencers (Applied Biosystems). Sequence base calling, contig assembly, quality clipping and screening for vectors were performed with PHRED/PHRAP and CONSED software (Ewing and Green, 1998; Ewing et al. 1998; Gordon et al. 1998).

We used two strategies in finishing to fill gaps and clarify ambiguities. Selected clones were sequenced with M13 forward or custom primers. We also used custom primers to generate PCR products, which were then sequenced directly.

To guard against sequence misassembly, we digested each BAC clone with restriction enzyme (BamHI, BglII, EcoRI, HindIII or XbaI), separated the fragments by pulsed-field gel electrophoresis (CHEF-MAPPER, BioRad, Hercules, CA), and compared the pattern to a virtual digest of the consensus sequence. In all cases, the predictions were consistent with the digest pattern (not shown).

The sequences reported here have been submitted to GenBank under the following clone name and accession numbers: Citb585c7–AF532116; Citb255d16–AF532113; Citb10i1–AF532111; Citb592j14–AC005413; Citb76k14–AC005665; Citb592j14–AF532114; Citb9k22–AF532117; 553n23–AF532115; 20h22–AF532112; and 544e14–AF532114. The final sequence assembles as follows: Citb585c7 is from 1 to 192,871, 255d16 fragment is from 188,743 to 215,704, 10i1 is from 212,656 to 390,909, 592j14 is from 390,371 to 513,741, 76k14 is from 402,819 to 527,004, 9k22 is from 499,424 to 672,711, 553n23 is from 630,719 to 792,447, 20h22 is from 703,411 to 875,609, and 544e14 is from 797,824 to 951,695.

The quality score for each clone was calculated by CONSED before manual finishing: 585c7: 0.23 x 10^–4, 10i1: 0.06 x 10^–4, 20h22: 0.04 x 10^–4, 544e14: 0.12 x 10^–4, 255d16: 13.45 x 10^–4, 592j14: 2.6 x 10^–4, 76k14: 4.3 x 10^–4, 9k22: 9.17 x 10^–4 and 533n23: 3.7 x 10^–4. Each BAC was finished separately, but in overlap regions, use was made of the information from the clearest reads. Small discrepancies were found in the overlap between 20h22 and 544e14. Three bases differ between them: T at position 107,794 and AC at position 113,728–113,729 of 20h22 are missing in 544e14. All these differences are located within repeat sequences. We used 20h22 for analysis, because it had a better quality score than 544e14.

544e14 is the most telomeric BAC in our contig and it overlaps with 573k1 (Younger et al. 2000), which was sequenced and assembled independently in the Sanger Institute (accession no. AL078630). There are five discrepancies in the 35,882 bp overlap between 544e14 and 573k1 (listed as 544e14 to 573k1, and the positions are given in 544e14): A to C at position 119,514; G insertion at position 119,964; G to N at position 121,847; CTC to GCT at position 122,414–122,416; GA insertion at position 151,811–151,812.

Sequence Analysis Programs
Interspersed and simple repeat sequences were identified and masked by RepeatMasker (http://ftp.genome.washington.edu/cgi-bin/RepeatMasker). The masked sequence was compared against various databases using the BLAST programs (http://www.ncbi.nlm.nih.gov/BLAST; Altschul et al. 1997). GC-content was calculated with the ISOCHORE program (http://www.emboss.org). Coding regions and the gene structure were predicted by GENSCAN (http://genes.mit.edu/GENSCAN.html; Burge and Karlin 1997). EST alignments were also used to determine the exon–intron boundaries of predicted genes. CLUSTALW (Thompson et al. 1994) was used for multiple alignments, and the neighbor-joining trees were constructed with MEGA2 (Kumar et al. 2001). Dot matrix comparisons were carried out with PipMaker. To compare the human and mouse sequence we searched "both strands" and "showed all matches" in the output. To compare the sequence to itself we used "both strands" and "single strand" searches (http://bio.cse.psu.edu/pipmaker; Schwartz et al. 2000).

	WEB SITE REFERENCES

Top ABSTRACT RESULTS DISCUSSION METHODS WEB SITE REFERENCES NOTE ADDED IN PROOF REFERENCES

 
http://ftp.genome.washington.edu/cgi-bin/RepeatMasker; RepeatMasker home page.

http://www.ncbi.nlm.nih.gov/BLAST; NCBI BLAST home page.

http://www.emboss.org; EMBOSS (European Molecular Biology Software Suite) home page.

http://genes.mit.edu/GENSCAN.html; GENSCAN home page.

http://bio.cse.psu.edu/pipmaker; PIPMAKER home page.

	NOTE ADDED IN PROOF

Top ABSTRACT RESULTS DISCUSSION METHODS WEB SITE REFERENCES NOTE ADDED IN PROOF REFERENCES

 
The March 7, 2003, issue of Cell describes expression of M10 and M1 families in the vomeronasal organ (Loconto, J., Papes, F., Chang, E., Stowers, L., Jones, E.P., Takada, T., Kumánovics, A., Fischer Lindahl, K., and Dulac., C. Functional expression of murine V2R pheromone receptors involves selective association with the M10 and M1 families of class Ib molecules).

	Acknowledgements

 
This work was supported in part by NIH grants AI37818 and NHGRI HG00202. We thank Dr. Clive Slaughter and the HHMI Biopolymer Facility, in particular Teresa Bannister and Brad Carmical, as well as members of the Genome Science and Technology Center at University of Texas Southwestern Medical Center, in particular Pablo Aguayo, Cynthia English, Meiyu Loo and Jeff Schageman for their technical support.

The publication costs of this article were defrayed in part by payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 USC section 1734 solely to indicate this fact.

	Footnotes

 
⁴ Laboratory Animal Science, The Tokyo Metropolitan Institute of Medical Science, 3-18-22, Honkoagome, Bunkyo-ku, Tokyo 113-8613, Japan;

⁵ Genaissance Pharmaceuticals, Five Science Park, New Haven, Connecticut 06511, USA;

⁶ Egea Biosciences, Inc., 6759 Mesa Ridge Road, Suite 100, San Diego, California 92121, USA

Present addresses:

⁷ Corresponding author.

E-MAIL kfl{at}chop.swmed.edu; FAX (214) 648-5453.

Article and publication are at http://www.genome.org/cgi/doi/10.1101/gr.975303.

	REFERENCES

Top ABSTRACT RESULTS DISCUSSION METHODS WEB SITE REFERENCES NOTE ADDED IN PROOF REFERENCES

 

Grifa, A., Totaro, A., Rommens, J.M., Carella, M., Roetto, A., Borgato, L., Zelante, L., and Gasparini, P. 1998. GABA (-amino-butyric acid) neurotransmission: Identification and fine mapping of the human GABAB receptor gene. Biochem. Biophys. Res. Commun. 250: 240-245.[CrossRef][Medline]