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 [1], 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 [2]. 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.

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) [31]. 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 [32]. CpG nucleotide could be preserved if the repair system had a bias towards G:C pairs instead of A:T pairs [33] as suggested for regions with high recombination activity [34]. 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 [35]. Conservation of CpG dinucleotides may therefore be a result of either low germ-line methylation or a specific selection against the loss of CpGs [36]. 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 [37], Endo.SceI [38] and homing endonucleases [reviewed in [39]], 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. 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.

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.