|Home | About | Journals | Submit | Contact Us | Français|
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
HLA/MHC class II molecules show high degree of polymorphism in the human population. The individual polymorphic motifs have been suggested to be propagated and mixed by transfer of genetic material (recombination, gene conversion) between alleles, but no clear molecular basis for this has been identified as yet. A large number of MHC class II allele sequences is publicly available and could be used to analyze the sequence features behind the recombination, revealing possible basis for such recombination processes both in HLA class II genes and other genes, which recombination acts upon.
In this study we analyzed the vast dataset of human allelic variants (49 full coding sequences, 374 full exon 2 sequences) of the most polymorphic MHC class II locus, HLA-DRB1, and identified many previously unknown sequence features possibly contributing to the recombination. The CpG-dinucleotide content of exon 2 (containing the antigen-binding sites and subsequently a high degree of polymorphism) was much elevated as compared to the other exons despite similar overall G+C content. Furthermore, the CpG pattern was highly conserved. We also identified more complex, highly conserved sequence motifs in exon 2. Some of these can be identified as putative recombination motifs previously found in other genes, but most are previously unidentified.
The identified sequence features could putatively act in recombination allowing either less (CpG dinucleotides) or more specific DNA cleavage (complex sequences) or homologous recombination (complex sequences).
Over the last few years our knowledge of the mechanism of recombination has increased substantially. Still, the knowledge is to a large extent based on simple organisms such as E. coli and yeasts, as the vertebrate genome is not equally readily or rapidly monitored or manipulated. It is well known that homologous pairing and strand exchange involved in recombination in the eukaryotic cell is promoted by specific recombination proteins , and that recombination is tightly linked to DNA replication and repair. For example, double strand breaks are repaired by recombination using information from homologous DNA molecules. Moreover, stalled replication can be re-started by forming a recombination intermediate with assistance from recombination proteins at the replication fork . Recombination also generates diversity essential for, e.g., the vertebrate adaptive immune system (immunoglobulins and T-cell receptor genes) and long-term genome evolution. The term illegitimate recombination was coined to describe one type of "novel" recombination, which, in contrast to the classical (homologous) recombination, requires no or only short stretches of sequence homology [reviewed in [3-5]]. Despite recent advances in the investigation of eukaryotic recombination, little is known about the mechanisms of illegitimate recombination, except for some specific cases like the immunoglobulin gene rearrangements.
The major histocompatibility complex (MHC) class II loci encode heterodimeric cell surface receptors that present peptide antigens to helper T-cells so that an appropriate immune response can be induced. In man, the by-far most polymorphic MHC class II locus is HLA-DRB1; as of march 2008 the HLA-DRB1 locus had over 540 alleles [6,7] and is thus one of the most polymorphic loci in the human genome. A large number of low-frequency alleles is apparently maintained in the human population by balancing selection. The peptide fragments are bound by interactions with the peptide backbone and amino acid side chains in the second exon-coded part of HLA-DRB1 (DRB1-e2), termed antigen recognition sites (ARS). Each individual carries a maximum number of two different inherited alleles per locus (assuming heterozygocity), while the greater allelic diversity is present in the population, putatively allowing population adaptation to pathogens.
ARS polymorphisms are thought to be created by point mutations, which are propagated by some recombination events, e.g. gene conversion. This view is based on the observed patchwork pattern of apparently exchanged motifs and the fact that synonymous substitutions are also much elevated in the DRB1-e2 (hitch-hiking with the non-synonymous substitutions) [8-11]. However, there is little direct evidence for any recombination in MHC class II ARS, and no clear recombinogenic motifs or mechanisms have as yet been identified. Since the multiple ARS of DRB1-e2 are spread over a small region of 200 bp only, exchange of very small blocks of DNA is needed to create the pattern of polymorphism seen. This, again, is in sharp contrast to the classical (homologous) recombination, which requires significant stretches of sequence homology and exchanges relatively large blocks of generic material. Therefore, due to the apparent high activity of illegitimate recombination in DRB1-e2 and the large number of allelic sequences known, DRB1-e2 seems to be a uniquely suitable target for investigations of mechanisms behind illegitimate recombination. As it is known that specific DNA sequences can enhance or mediate recombination, we have in this study targeted the vast database of known human HLA-DRB1 alleles in the quest for possible sequence motifs that would enable recombination. The analyses identify strongly conserved sequence features as well as recombinogenic motifs previously recognized in other genes, which may thus lie at the basis of recombination events creating new alleles.
DRB1-e2 displays much higher degree of sequence diversity than the other exons in DRB1 (Fig. (Fig.1A),1A), independently of whether gross-diversity or non-synonymous (aminoacid-changing) diversity is analyzed . As seen in earlier studies [see e.g. ], the synonymous diversity is also elevated in exon 2 (Fig. (Fig.1B),1B), supporting the view that recombination, e.g. gene conversion is involved in creating polymorphism in this exon [13,14,10]. Consequently, synonymous substitutions would "hitch-hike" within the same exchanged DNA blocks as the non-synonymous substitutions and remain conserved due to selection forces acting on the non-synonymous substitutions. Indeed, synonymous substitutions were mainly found either in the same codons as the non-synonymous ones (Fig. (Fig.1C,1C, the "complex" trace in its entity and the overlap of the non-synonymous and synonymous traces) or in their close vicinity (Fig. (Fig.1C1C).
The higher diversity of DRB1-e2 is also reflected by a higher frequency of both transitions (T↔C, A↔G) and transversions (T↔A, T↔G, C↔A, C↔G), as compared to the rest of the DRB1 exons (Fig. (Fig.2).2). In general, transitions occur at higher frequencies than transversions in our genome . However, while the transition/transversion-ratio was about 2 in the fused other exons (between 2 and 3 in the separate exons 1, 3 and 4, which have a length comparable to exon 2), it was 0.8 in exon 2. The results using the full dataset of 374 complete DRB1-e2 (transitions = 3.4 ± 0.0%, transversions = 4.2 ± 0.1%, ratio = 0.8) were similar to the 49 complete coding sequences (Fig. (Fig.2).2). The transition/transversion ratio near unity in DRB1-e2 is logical in the light of previous studies, which show that when sequences diverge and mutations accumulate, the transition/transversion-ratio decreases finally approaching 1 due to transition saturation [16,17].
The G+C level was similar across all DRB1 exons (Fig. (Fig.3).3). At the determined G+C content of the whole of DRB1 of 60%, the theoretical level of CpG should under neutral conditions be 9% (see below). When present in the CpG-dinucleotide, cytosine is often methylated. Methyl-cytosine can then deaminate to uracil, and thus lead to the transition C→T or, if occurring in the complementary strand, a G→A transition. Thus, CG may mutate to TG or CA, and genes regularly have a lower than mathematically expected level of CpG dinucleotides. The high propensity of CpG to mutate may also effectively engage DNA repair. DNA repair induces double strand breaks and may support recombination events, which may explain why CpG-rich sequences have been identified to display high recombination activity.
Theoretically, CpG content = ((G+C content/100%)/2)2 * 100% (e.g. for G+C content of 50%, CpG content = 6.25%). The actual CpG content depends on the age of the gene, but in average the content would be below one third of the theoretical (e.g. for G+C content of 50%, CpG content < 2%) . As expected, the determined CpG content of DRB1-e1, and -3-6 was well below the theoretical level of 9% (Fig. (Fig.4A;4A; in average ~1/6 of the theoretical level [dotted line in Fig. Fig.4B]).4B]). In contrast, CpG content of DRB1-e2 was surprisingly high, about 8% (Fig. (Fig.4A),4A), which suggests that the CpG level is almost fully preserved in DRB1-e2 (see also Fig. Fig.4B).4B). In addition, the distribution of CpG dinucleotides in DRB1-e2 was to a very high extent conserved in all alleles (Fig. 5AB).
There is a number of sequence features or motifs proposed to be recognized by specific nuclease complexes, resulting in double strand breaks and increased recombination rate [19,20]. We further analyzed the sequence data sets to explore the possibility that specific recombination motifs are involved in creating polymorphism in DRB1-e2 (Table (Table11).
Recombination signal sequences (RSSs) are involved in the diversification of antibody genes, initiated by DNA double-strand breaks introduced in the vicinity of RSSs. The RSSs are composed of conserved heptamer and nonamer motifs separated by a spacer of 12 or 23 bp . The heptamer motifs, especially the first three bases (CAC), are the most influential on recombination efficiency and are usually the most conserved . We found a heptamer-like motif (5'-CACGGTG-3', the bold letter is replaced in 7% of the alleles) at position 254–260 in exon 2.
Class switch recombination (CSR) refers to the event when a lymphocyte changes the type of immunoglobulin it produces . Also CSR involves recombination via DNA double strand breaks at switch regions containing repetitive elements (predominantly, 5'-GAGCT-3' and 5'-GGGGT-3'). The immunoglobulin heavy chain class switch repeat GAGCT was present at position 145–149 and 248–252 in all but one (DRB1*0423) of 374 DRB1 alleles (Table (Table2).2). Another immunoglobulin heavy chain class switch repeat (5'-TGGGG-3') was present in all alleles except for alleles in the lineage DRB1*07 (Table (Table2).2). This immunoglobulin heavy chain class switch repeats was also present in the exon 2 sequences excluded from the analysis due to missing bases in either the 3' or 5' end (see Additional file 1).
Chi (crossover hotspot instigator, χ) is an octamer recombination hotspot (5'-GCTGGTGG-3') of the major recombination pathway in E. coli [reviewed in ]. Recombination by this pathway is initiated by double-strand breaks occurring at chi sequences. Variants of this motif are suggested to have partial recombinogenic activity, and chi-like sequences have been speculated to be involved in both deletion and translocation events in man [25-27]. A chi-like sequences at nucleotide position 143–149 in exon 2 was found in all alleles except for the DRB1*07 allelic lineage (Table (Table2).2). The chi-like sequence in DRB1-e2 was overlapping with a motif reported to be a deletion hotspot consensus sequence (5'-TGRRKM-3'), suggested to be involved in illegitimate recombination [28,29]. We located this hotspot sequence at positions 145–150 in all alleles except for the allelic lineage DRB1*07 (Table (Table2).2). Moreover, this sequence was present in the non-coding strand of all alleles at coding strand position 37–42 (Table (Table22).
Several types of the recombination motifs screened for were also found in the other exons of DRB1 (the dataset of 49 complete coding sequences) (not shown).
Despite the high degree of variability in DRB1-e2 we could, to our surprise, find 19 stretches of a length 3–13 bp that have no variation at all among different DRB1-e2 (Table (Table33 and Fig. Fig.6).6). Some of these fully conserved bases corresponded to the known motifs as identified above (Table (Table33).
In this study we identify several distinct features of exon 2 of DRB1. One of these is its high CpG content, possibly leading to a high degree of a) point mutations and b) DNA repair. However, not only is the CpG level high in DRB1-e2, but also the CpG pattern is highly conserved in DRB-e2. It therefore appears unlikely that CpG-dinucleotides would support ARS polymorphism by point mutations. More likely is that the conserved CpG pattern is explained by frequent DNA repair, which, by introducing double-strand DNA cleavage followed by non-homologous end-joining, is one of the suggested mechanisms of gene conversion [reviewed in [3,5,30]]. Earlier studies of MHC class I nucleotide sequences in mice have proposed that regions with high levels of CpG dinucleotides are involved in non-reciprocal recombination (gene conversion) . Analyzes of human MHC class I and II sequences also have reported increased CpG dinucleotide levels in regions suggested to be involved in gene conversion . CpG nucleotide could be preserved if the repair system had a bias towards G:C pairs instead of A:T pairs  as suggested for regions with high recombination activity . However, it should be born in mind that unmethylated CpG dinucleotides, in contrast to the cytosine-methylated CpG:s, mutate at normal rates and regions with high CpG contents may have low levels of methylation . Conservation of CpG dinucleotides may therefore be a result of either low germ-line methylation or a specific selection against the loss of CpGs . However, the highly significant pattern of conserved CpG in HLA-DRB1-e2 can be considered unlikely even if the CpG dinucleotides were unmethylated and mutated at the rate of other bases.
A few eukaryotic endonucleases with specific DNA recognition sequences involved in DNA recombination, such as topoisomerase I , Endo.SceI  and homing endonucleases [reviewed in ], have been identified. The enzymes have in common that they recognize a more or less strictly defined DNA sequence and cleave at it or some distance from it. In addition, a number of other conserved sequence motifs associated with high recombination activity (such as the chi-like sequences) but without a pinpointed endonuclease/recombinase have been recognized [40,19,41]. In this study, we screened DRB1-e2 for "known" recombination, translocation and deletion motifs. A heptamer-like motif was found in all investigated DRB1 alleles and an immunoglobulin heavy chain class switch repeat was found in all but one of the DRB1 alleles. Moreover, a chi-like sequence and a deletion hotspot consensus sequence were present in all alleles except for the *07 lineage. It is thinkable that the DRB1*07 allelic lineage, which contains least alleles of all the lineages, may have lost one of these motifs and therefore gained a limited ability to recombine. This may also be true for other, even less frequent motifs, also including conserved CpGs. Currently, not enough is known about the function of the specific motifs found in order to speculate further on their function. Interspersed among the highly polymorphic areas of DRB1-e2, we found multiple short stretches of bases that have no variation at all between the DRB1 alleles in the dataset. Conserved amino acid motifs can be important for the maintenance of the overall structure of the antigen-binding groove, but as these stretches also lack synonymous substitutions they may have a function in allowing recombination between alleles via illegitimate recombination. This could occur either by offering homology for recombination, by allowing cleavage by some specific enzymes or by stabilization of DNA's secondary structure. Comparison of these sequences to known sequence motifs associated with recombination (see above) produced no hits, which is by no means surprising as indeed only few motifs are known and even fewer verified.
It should be born in mind that the specific sequence motifs screened for are for the most part very short and may thus appear in a random fashion in any sequence analyzed. Indeed, some motifs were found in the other exons of HLA-DRB1, not subject to recombination. However, the fact that exon 2 is subject to high rate of recombination – in contrast to the other exons which are highly conserved – makes random conservation of such stretches unlikely, especially in such a large pool of allelic sequences. Yet the most remarkable features of DRB1-e2 are, rather than the known recombinogenic motifs found, a) the fully conserved sequence stretches and b) high CpG content and the conserved CpG pattern.
We have identified in DRB1-e2 both some known recombination motifs and multiple putative motifs. The latter include both the conserved CpG pattern and other fully conserved sequence motifs. Although the role of these sequence features in the recombination processes in DRB1 is speculative, it is obvious that the known recombination motifs identified here cannot be enough to support the full spectrum of recombination. 22 variable and 15 conserved DRB1-e2 ARS-coding codons, spread over 245 bp (Fig. (Fig.6),6), are known, and each of the variable ARS codons should probably be able to recombine separately from the others, theoretically requiring 23 recombination breakpoints. Whether this indeed is the case, will be deduced from full mapping of the DRB1-e2 recombination profile, which is currently in progress. If the conserved sequence motifs identified here indeed are important in recombination, they would likely be present in other regions of the genome with high recombination activity. This will also be addressed in future studies.
For the analysis, sequences from the IMGT/HLA database [6,7] were used. The datasets analyzed were the 374 complete exon 2 sequences and 49 complete coding sequences (exons 1–6). Full descriptions of the datasets can be found in the Additional files 2 and 3.
The sequences were aligned using ClustalW . The mean synonymous and non-synonymous diversities (d) were estimated by pairwise comparison of the number of nucleotide substitutions using the Jukes-Cantor method  with the MEGA3.1 software . Sliding-window analyses of the nucleotide diversities were performed using DnaSP 4.10.9  Analyses of the transition/transversion-ratios and the G+C and CpG contents were done with SWAAP 1.0.2  and MEGA3.1. The sliding window analyses of the CpG content were performed using Microsoft Excel.
Recombination has been suggested to be promoted by common sequence features or motifs , known or postulated to be recognized by specific nuclease complexes, leading to double strand break and increased recombination rate. We screened DRB1-e2 (coding and non-coding strands) in MEGA3.1 for sequence motifs previously shown to be involved in recombination, to explore the possibility that specific motifs are involved in creating new polymorphisms. The motifs screened for are listed in Table Table11.
ARS: antigen-recognition site(s); bp: basepairs; Chi: crossover hotspot instigator; CpG: CG-dinucleotide (in DNA); CSR: class switch recombination; d: nucleotide diversity; G+C content: content of guanine and cytosine nucleotides (in DNA), DRB1-e2: exon 2 of the HLA-DRB1 gene; HLA: human leukocyte antigen; MHC: major histocompatibility complex; RSSs: recombination signal sequences.
JPK designed the study, JvS and JPK performed the analyses and wrote the manuscript. Both authors read and approved the final manuscript.
The 98 incomplete HLA-DRB1 exon 2 sequences excluded from the analyses.
The 374 complete HLA-DRB1 exon 2 sequences used in the analyses.
The 49 complete HLA-DRB1 coding sequences used in the analyses.
This study was supported by grants from the Novo Nordisk Foundation, the Sigrid Jusélius Foundation, the Magnus Ehrnrooth Foundation, the K. Albin Johansson Foundation, the Swedish Research Council, Uppsala University, Åbo Akademi University and the University of Helsinki Research Funds.