V_H replacement in rearranged immunoglobulin genes

Abstract

Examples suggesting that all or part of the V_H segment of a rearranged V_HDJ_H may be replaced by all or part of another V_H have been appearing since the 1980s. Evidence has been presented of two rather different types of replacement. One of these has gained acceptance and has now been clearly demonstrated to occur. The other, proposed more recently, has not yet gained general acceptance because the same effect can be produced by polymerase chain reaction artefact. We review both types of replacement including a critical examination of evidence for the latter. The first type involves RAG proteins and recombination signal sequences (RSS) and occurs in immature B cells. The second was also thought to be brought about by RAG proteins and RSS. However, it has been reported in hypermutating cells which are not thought to express RAG proteins but in which activation-induced cytidine deaminase (AID) has recently been shown to initiate homologous recombination. Re-examination of the published sequences reveals AID target sites in V_H–V_H junction regions and examples that resemble gene conversion.

Introduction

Faced with selective pressure to change the specificity of a rearranged immunoglobulin gene (or rescue itself from two non-functional rearrangements) a B cell may make a secondary rearrangement on the same chromosome.1–4 Some authors divide this into ‘receptor editing’ when it occurs centrally (in the bone marrow), and ‘receptor revision’ when it occurs in the periphery (in germinal centres) the former thought to be largely tolerance-driven and the latter mainly diversity-driven. Most rearranged light-chain genes will have both unrearranged V_L gene segments upstream and unrearranged J_L segments downstream, so it is possible for an upstream V_L segment to recombine with a downstream J_L segment to form a new V_LJ_L exon with deletion of the original rearrangement. At the heavy-chain locus the situation is different because the rearranged gene contains a D segment and in humans all the unused D segments are deleted during the D–J_H and V_H–DJ_H rearrangements. Several possible varieties of secondary rearrangement at the heavy chain locus have been suggested, but the ones for which there are the most evidence are replacement of all or part of the primarily rearranged V_H segment by all or part of an upstream V_H gene segment, and these comprise the subject of this article.

Two quite different mechanisms of V_H replacement have been reported. We refer to these as Types 1 and 2. We review the evidence for both types of replacement, describe examples of confusion between the two, and consider what is the most likely explanation for the Type 2 sequences in the light of other recent findings and our own re-examination of the sequences. We begin with a summary of current understanding of primary V(D)J recombination.

V(D)J rearrangement and non-homologous end joining

V(D)J recombination takes place in two stages. The first, recognition and cleavage of specific sequences within the immunoglobulin loci, is unique to V(D)J recombination. The second, the processing and joining of the cleaved ends, involves components of the ubiquitous nonhomologous end-joining (NHEJ) system. Understanding of the details of primary V(D)J rearrangement is developing rapidly.5–9 The functional V, D and J gene segments have recombination signal sequences (RSS), each consisting of a specific heptamer and nonamer separated by a spacer of 12 bp (roughly one turn of the DNA helix) or 23 bp (roughly two DNA turns). These distances are important for protein binding, and most efficient cleavage and rejoining occurs when one of each type of RSS are brought together in a protein complex. This ensures recombination of the correct coding sequences to make a functional product since, at the heavy-chain locus, the V and J segments all have RSS with a 23-bp spacer and the D segments have 12 bp spacers on both ends. The heptamer sequences are immediately adjacent to the ends of the coding sequences that are to be joined and have the same consensus sequence but are in opposite orientations. The sequence of the heptamer is more critical than that of the nonamer and the most important sequence is the three bases next to the cleavage point, CAC on the sense strand in RSS at the 3′ ends of coding segments, GTG in those at the 5′ ends.¹⁰ A complex formed by the products of the ‘recombination activating genes’, RAG-1 and RAG-2, and the high mobility group 1 (HMG1) protein binds the RSS and nicks the DNA precisely at the interfaces between the coding sequences and the RSS. The free –OH group on the 3′ end of the coding sequence at each nick then attacks the opposite DNA strand and the result is a closed hairpin loop on each coding sequence end and a blunt double-strand break at the end of each RSS.

The ends are then joined by NHEJ, present in all cells for repair of double-stranded DNA breaks. It frequently uses very short sequence-matches of typically 1–4 bp near the ends to be joined and frequently involves removal of nucleotides from the ends. In V(D)J recombination this imprecision is a benefit, conferring extra diversity to antibodies.

The proteins of the NHEJ system involved in V(D)J recombination include the Ku70-Ku86 heterodimer, Artemis, DNA-dependent protein kinase catalytic subunit, XRCC4 (X-ray cross-complementation group 4), DNA ligase IV, and others yet to be identified.^5–7,11 The cleaved ends bearing RSS are usually ligated together (signal-to-signal joint) by perfect blunt-ended ligation without loss or addition of bases. At the coding ends, the hairpins are opened by nicking and, if not central, this results in addition of palindromic (P) nucleotides. However, there is often excision of several nucleotides. Before joining there is usually non-templated (N) addition of nucleotides to the 3′ ends by terminal deoxynucleotidyl transferase (TdT). Complementary matches of one or a few bases at or near the ends of these extensions may then be used to pair the ends, non-matching bases are removed, gaps are filled in and nicks ligated. Where the coding segments joined are in the same orientation, as in the unrearranged human heavy-chain locus, ligation of the RSS causes the exclusion of the intervening DNA from the chromosome as a closed loop. Presentation of a functional B-cell receptor on the cell surface leads to down-regulation of the RAG genes, preventing further V(D)J rearrangement. It has yet to be established for certain whether RAG genes can be reactivated in mature cells in vivo or whether RAG-mediated secondary rearrangement only occurs in immature cells.¹² However, it has been shown that RAG expression and receptor revision can be induced in mature B cells in vitro.¹³

Type 1 V_H replacement

Type 1 V_H replacement is essentially the same mechanism as primary V(D)J rearrangement, using legitimate or ‘cryptic’ RSS in opposite orientations, usually involving deletion from the cleaved coding ends, addition of P and N nucleotides at the new junction, and sometimes a substantial change in length. It also frequently causes a change in frame, making a non-functional product from a functional one, or vice versa.

Because the D segment of the rearranged V_H–D–J_H exon no longer has its own 5′ RSS, cryptic RSS within the V_H are always involved in Type 1 V_H replacement. There are variations between RSS of different gene segments in the heptamer, spacer, and nonamer sequences^10,14 but the consensus sequence of the heptamer on the 5′ ends of the D segments is CACTGTG. Close to the 3′ end of nearly all of the functional human and mouse V_H segments is the sequence TACTGTG, only one base different from the consensus heptamer, and this can act as a surrogate for the original RSS on the 5′ end of the D segment.¹⁵ Within most V_H segments there are also more proximal cryptic RSS in both orientations, not as close to the RSS consensus as the 3′ cryptic RSS. These appear to be used very rarely in Type 1 replacement. Separated by 12 bases upstream from the 3′ cryptic heptamer (also known as ‘the internal heptamer’) is a nonamer sequence that varies a little between V_H genes. It does not resemble the consensus nonamer of the real RSS but, because of its conservation, Chen et al.¹⁶ considered that it might act as a signal sequence. Nadel et al.¹⁷ did not detect recombination in pre-B cells transfected with plasmids containing the complete nonamer–spacer–heptamer ‘embedded RSS’ present in the 3′ end of several of the mouse V_H genes, and concluded that V_H replacement must be an infrequent event, unlikely to contribute significantly to receptor editing. However, numerous examples of V_H replacement continued to be observed in pre-B-cell lines (endogenous genes and introduced substrates), lymphomas and leukaemias^15,18–29 and in transgenic mice.^16,30,31

The reported Type 1 V_H replacements fall into three subtypes. (i) As in the references above, a complete upstream V_H segment may be recombined into a rearranged V_H–D–J_H through the agency of its own RSS and the 3′ cryptic RSS of the rearranged V_H(Fig. 1a). This effectively increases the length of the V_H–D junction and has a (theoretically 1/3) chance of producing a functional heavy chain (depending on the reading frame). Most examples of Type 1 V_H replacement recorded so far are of this subtype. (ii) A complete V_H gene segment may be recombined into the V_H–D–J_H with its own RSS and a more 5′ cryptic RSS in the already-combined V_H (Fig. 1b). In the one example to date³² the incoming V_H was in a different frame from the original and this resulted in a stop codon being read in the original V_H. The nearest possible cryptic RSS 5′ of the cleavage point in the original V_H (now known as V_H3–49) is TATGGTG in codons 53–55. Only the four bases underlined are shared in common with the consensus RSS, which may explain why this is a rare event. (iii) Cryptic RSS in opposite orientations and different positions within both V_H might be involved (Fig. 1c). However, we have now shown (unpublished observation) that the three published examples³³ result from a duplication within one V_H.

Figure 1.

Type 1 V_H replacement. (a) Usual subtype: recombination occurs after cleavage at the conventional RSS of a new (upstream) V_H and at the 3′ cryptic RSS in the V_H portion of the recombined V_H–D–J_H. (b)Cleavage occurs at the conventional RSS of the incoming V_H but at an upstream cryptic RSS of the V_H of the VDJ. (c) Cleavage occurs at cryptic RSS in opposite directions and different positions in the two V_H genes. In this and all other figures, the conventional RSS are indicated by filled triangles and cryptic RSS by open triangles. Small arrows indicate cleavage positions and deletion products are omitted. ‘∼’ indicates the typical random small deletions from either or both coding ends and added P and N bases.

Other types of RAG-mediated rearrangement, D→V_HDJ_H and V→D→V_HDJ_H may occur in artificial systems.^16,34 Taki et al.³⁴ found that of 53 RSS-like motifs present in the monoclonal mouse system that they were using, only two were used, suggesting that other sequences are involved in the targeting of RAG-mediated recombination in addition to the RSS.

The sequence between the 3′ cryptic RSS and the 3′ end of the V_H is 5–9 (usually 7) bases long in human V_H genes. This may be truncated from its 3′ end in the original rearrangement and from its 5′ end after the 3′ cryptic RSS is cleaved off. Therefore the V_H-D junction, after Type 1(i) V_H replacement, will usually be longer than the original V_H-D junction with the same sequence towards its 3′ end, new sequence at its 5′ end, and possibly a remnant of the tail of the original V_H in the middle, as shown in Fig. 2(a). This is easy to identify by comparing sequences from members of the same B-cell clone but almost impossible to identify in isolation.

Figure 2.

Comparison of Types 1 and 2 V_H replacement. (a) Detail of the usual form of Type 1 involving the conventional RSS of the incoming V_H and the 3′ cryptic RSS in the V_H portion of the pre-existing V_H–D–J_H. Diagram conventions are as in Fig. 1. Just as in normal V(D)J recombination, the new junction in the VDJ is a coding-to-coding junction and the deletion product has a signal-to-signal junction. (b) A Type 2 V_H replacement scheme to explain the published hybrid sequences by the investigators' hypothesis of RAG-mediated double-strand cleavage at cryptic RSS and rejoining without end modification. Reaction products shown are a V_HDJ_H with the hybrid V_H of the kind published and the deduced circular deletion product, both with cryptic RSS signal-to-coding junctions. Alternative mechanisms are homologous recombination following RAG-mediated single-strand nicking at a cryptic RSS and our preferred hypothesis of AID-initiated homologous recombination unrelated to RAG or cryptic RSS. Homologous repair in these cases need not involve any deletion.

Recently, Zhang et al.³⁵ demonstrated Type I V_H replacement conclusively in an acute lymphocytic leukaemia cell line that undergoes continuous pro-B- to pre-B- to B-cell differentiation and serial V_H replacement. Different members of the cell line contained rearranged genes with the same D–J_H junction but different V_H segments, and remnants of previous V_H segments in some of the V_H–D junctions. The investigators demonstrated double-strand DNA breaks at 3′ cryptic RSS, V_H-replacement excision circles, binding of copurified recombinant RAG1/RAG2 to cryptic RSS, nicking at the heptamer ends, and coding-end hairpin formation. Furthermore, they found that the 3′ cryptic heptamer enhanced hairpin formation with true RSS with both 12 bp and 23 bp spacers. They also found evidence for Type 1 V_H replacement in healthy human neonates and adults. A search of the V–D junctions of 343 functional immunoglobulin H (IgH) sequences from 18 subjects identified 16 (∼5%) with 5 consecutive matches to a 3′ V_H sequence. When they lowered the stringency to six bases with one mismatch, this increased to ∼12%, and the true frequency of Type I replacement is likely to be much higher, because, as we mentioned earlier, there will frequently be little or no trace of the previous V_H segment.

Type 2 V_H replacement

Type 2 V_H replacement has been described much more recently than Type 1.^36–38 The result gives the appearance of homologous recombination between two V_H gene segments (Fig. 2b). Because the exchange occurs in a stretch in which both segments have the same sequence, the exact point of exchange cannot be determined for certain; it is only delimited by the closest positions of difference between the two sequences in either direction. It has been proposed to occur at cryptic RSS but the recombination process is quite different from normal V(D)J rearrangement because the supposed cryptic RSS are in the same orientation, one of them remains in the recombined product, there is no addition of nucleotides at the point of exchange, and (with a single exception³⁶) no nucleotides are deleted from the gene sequence. The specificity of the antibody is changed, the reading frame of the gene is unchanged and the V_H–D junction is always unchanged. If this type of V_H replacement is guided by cryptic RSS, the joints formed must be signal-to-coding joints, known as hybrid joints. These have been shown to occur but, when they do, there is usually deletion of nucleotides from the coding end and addition of new nucleotides.³⁹

During polymerase chain reaction (PCR) amplification of DNA prepared from many lymphocytes, it is possible for the reaction to produce a rearranged V_H–D–J_H sequence with a hybrid V_H just like the postulated Type 2 replacement (Fig. 3). Double strand breaks (DSBs), targeted to RGYW·WRCY motifs have been found to occur in unrearranged V gene segments and in rearranged genes in the absence of activation-induced cytidine deaminase (AID), a protein essential for hypermutation.^40–42 On AID expression, additional staggered breaks occur at the same target.⁴³ All these breaks will result in incomplete single-strand PCR products, ripe for hybrid artefact formation, whose likelihood may thus be increased in mutating cells. It is therefore necessary to establish that Type 2 replacements occur in vivo.

Figure 3.

Construction of a sequence resembling Type 2 V_H replacement by PCR artefact. The N and P bases between V_H and D and between D and J_H are omitted for simplicity. (a) Copies of cDNA of two rearranged immunoglobulin heavy-chain genes occurring at an early stage in the same reaction mixture. ‘p’ is a PCR product made with a specific family V_H leader primer and a specific constant-region primer. ‘q’ is another rearrangement with a different D and J_H. The V_H differs from that of p at various points indicated by the pink spots in p and blue spots in q. q may be another PCR product with the same leader sequence, or it may be a cDNA template with a different leader that would not have been amplified with the primers in use. (b) During one of the PCR cycles, ‘n’, a copy of p, is incomplete when melting occurs. (c) In the next cycle n anneals with q at a point at which q has the same sequence as p. A polymerase molecule attaches and copying continues, giving a product (the extended ‘n’) that has the leader and 5′ end of the V_H the same as in p but the 3′ end of the V_H and the VDJ junction the same as q. This new molecule will now be amplified in all following cycles.

Wilson et al.³⁶ examined sequences from purified IgD⁺ IgM⁻ CD38⁺ tonsil B cells. In one group of 65 clonally related sequences (with the same CDR3) there were 7 that all had the same partial V_H replacement but differed in their mutations. In a phylogeny of the complete sequences these hybrids would be bound to come together on the tree, but these sequences all came on the same branch of a phylogenetic tree based only on the unreplaced segment, strongly suggesting that a single replacement event occurred, followed by further somatic mutation. There are at least three ways that this appearance might be produced by PCR artefact. Firstly, a single PCR hybridization could have occurred, followed by polymerase errors in some of the copies. These sequences would come together on the tree because they only differed by a single nucleotide. Experience in our laboratory has shown that even with a high-fidelity polymerase system this can happen when many copies of the same template are cloned and sequenced (error frequency ∼1/1200 bp in nested PCR or ∼1·1 × 10⁻⁵/changes/bp/cycle⁴⁴). Discounting this, one might postulate that seven separate PCR hybridizations had occurred between the same regions of copies of the same two rearranged genes but these would have to have mutation patterns so similar that the products came together on the tree despite the presence of many differently mutated copies of these genes (and many other rearranged templates) in the reaction. The chance of this seems negligible. In another group of 77 clonally related sequences there were two with the same Type 2 V_H replacement and they also came side-by-side on their phylogenetic tree, apparently compounding the likelihood that this was in vivo recombination.

There is, however, another possibility. If, after the first PCR hybridization, further hybridization occurred not to native templates but to the hybrid template, these secondary hybridizations would be much more likely to occur with sequences having closely related patterns of mutation. On both trees there were non-hybrid sequences beyond hybrid ones on the same branch. This is clearly impossible. The trees were produced by parsimony, and the findings might be a demonstration of the inadequacy of parsimony alone in constructing phylogenetic trees, or due to use of incomplete sequences, but it could also be an indication that the results were artefactual. The investigators³⁶ took the presence of somatic mutations before as well as after the replacements to indicate that replacement must have taken place in the periphery, i.e. that they were examples of receptor revision, and this must surely be the case if their explanation is correct.

Four more hybrid V_H genes were found in the two clones already mentioned, and a further 10 amongst >700 sequences from different B-cell subsets from tonsils of four other individuals, but they were all single examples of partial V_H replacements and so could not be distinguished from PCR artefacts. However, cryptic RSS were found in all nine different cross-over regions in the 15 different examples of V_H exchange. As some of these regions were quite long (up to at least 43 bp) and some of the identified cryptic RSS were only CAC or GTG, there was a high probability of some of these being present by chance, but the presence of an RSS in all nine regions cannot be ignored.

The V_H in a rearranged V_HDJ_H can only be replaced by a V_H upstream on the same chromosome, because the downstream V_H segments have all been deleted in the original rearrangement, but in 8/15 of the hybrids the incoming V_H was downstream in the existing V_H locus maps, which again could indicate PCR artefact. The authors considered the most likely possibility to be that mapping of the V_H region of more individuals will reveal more differences in order since the only two complete maps differ considerably. Less likely, they considered, was the possibility that the incoming V_H sequence had come from the other chromosome. Zhang et al.³⁵ noted that 4 of the 16 possible V_H replacement products, that they found by searching human V–D junctions for matches to 3′ ends of V gene segments, disagree with the gene order known so far, but these were not proven V_H replacements, and the authors did not find violation of the order in their serially V_H-replacing cell-line.

If the hybrids were formed by RAG-mediated recombination at cryptic RSS, one would expect recombinations between V_H segments from different families, just as with Type 1 V_H replacements. The PCRs were carried out with V_H Family 4 leader and C_µ, C_δ and C_γ primers and both parts of each hybrid were derived from V_H4 segments, as would be expected in PCR artefact. However, if the sequences do represent in vivo recombined genes, the finding strongly favours homologous recombination since this is much more likely to occur between members of the same family. Another argument against RAG-mediated recombination is that it should have produced some hybrids shorter and others longer than normal V_H segments by recombining at RSS at different positions in the two V_H segments. No such products were found, though this could have been because they were non-functional and the cells bearing them were thus not amongst the selected IgD⁺ IgM⁻ cells.

If the hybrids of Wilson et al.³⁶ were formed in vivo by RAG mediated recombination, then DNA double-strand breaks should be found at the proposed recombination junctions (from cleavage at cryptic RSS) as well as circular recombination products but all attempts to detect them failed or were inconclusive. The authors suggested that the Type 2 V_H replacements they found might be the tip of the iceberg of all receptor revision because complete (i.e. Type 1) functional replacements of V_H or J_H would be invisible and non-functional rearrangements would be inactivated, allowing rearrangement on the other chromosome, or eliminated.

Itoh et al.³⁷ amplified IgM, IgG, and IgA heavy-chain sequences with Family 1 V_H segments from synovial tissue samples from patients with rheumatoid arthritis. They too found hybrid sequences in some clones but no examples of the same hybrid with different mutations. Again, cryptic RSS were present in all the cross-over regions and again were always at the same position in the two recombining V_H segments. As with the other investigators' results, in half of their hybrids the 5′ V_H was from downstream of the original V_H on existing genomic maps. The authors gave the following reasons why they thought that their results were genuine: (i) hybrid rearrangments were found in different tissues from the same patient and in different patients; (ii) they were found in PCRs from both cDNA and genomic DNA (using a consensus J_H primer); (iii) RSS-like sequences were always found in the cross-over regions; (iv) rearrangements were always in-frame and coded for translatable proteins; (v) two different sets of related sequences showed members with different partial V_H replacements to the same original VDJ either in the same heptamer-containing region or in different ones. Reasons (i) (ii) and (v) could all be accommodated by PCR artefact. Reason (iii) we have already discussed and will return to later. Reason (iv) could be explained by PCR hybridization or homologous recombination, but would be less expected in a RAG-mediated process in which one might expect some diversification of the junctions. However, there is a better reason for believing that their results represented in vivo recombination. Though most of the receiving segments were of the same family as the incoming segments, there were two different hybrids in which a Family 1 segment was hybridized to the same original rearrangement with a Family 3 V_H segment. Their V_H primer was specific for Family 1 and they did not amplify any rearranged genes that started with a Family 3 V_H segment, making the chances of generating artefactual hybrids with such a gene very small. Further support seemed to come from the following additional investigations.

First, they looked for double strand breaks occurring at the cryptic RSS by blunt-end ligation followed by PCR, using human bone marrow cells as positive controls and fibroblasts as negative controls. This showed clearly that double strand breaks (DSBs) occurred in the synovial tissue samples and the positive control but not in the fibroblasts and there were two distinct bands corresponding to cleavage at two V_H heptamers, one of them the 3′ cryptic RSS seen in most V_H segments. Sequences of the PCR products were said to show that the breaks were at or within heptamers. RAG-mediated cleavage should be precisely at the edge of a heptamer. A break within one might suggest that the DNA had been chewed back but none of the examples of recombination in this paper showed any deletion. However, DSBs have since been found to occur even in unrearranged V gene segments^40–42 which somewhat diminishes the significance of the finding, though the breaks tend to occur at RGYW·WRCY motifs. Second, Itoh et al. found evidence of RAG activity in several of their synovial tissue samples and the positive but not the fibroblast control. Taken together, these findings suggested that RAG-mediated recombination was occurring in the tissue but not whether it was Type 1 or Type 2. Type 1 V_H replacement with the 3′ cryptic RSS would sometimes code for a functional product and the sequence would be identified as having secondary rearrangement only if a sequence with the original rearrangement were also cloned. Type 1 recombination with the more 5′-cryptic RSS at which they saw cleavage would not be expected to have produced a functional antibody (see Fig. 1b. The product in this particular case would have been about 70 bp longer than the original). The authors considered the possibility that their hybrids could have been formed by RAG-mediated cleavage at cryptic RSS followed by homologous recombination.

Single-cell PCR avoids the problem of hybrid product formation that can occur when DNA or RNA are extracted from many cells together. In order to determine whether V_H replacement accompanies somatic hypermutation in the germinal centre Goossens et al.⁴⁵ looked for hybrid V_H segments in single-cell PCR products from 124 IgD-only and 49 IgD⁻ IgM⁻ peripheral blood B cells and found no examples of Type 2 or Type 1. However, recently Lenze et al.³⁸ have discovered very strong evidence in another way. They looked for Type 2 V_H replacement in two human mucosa-associated lymphoid tissue lymphomas. Using DNA from many cells and nested PCR of rearranged heavy-chain sequences, they found a major clone of related VDJs (with identical CDR3 and J_H) and a polyclonal background of many unrelated sequences in each tumour. In each case, the related set contained members with hybrid V segments. However, 23/26 related sequences from three independent sets of PCRs from one tumour were V_H3-23(5′)/V_H3-07(3′) hybrids. Only a single related sequence had the complete parent V_H3–07, but this contained several mutations that were not present in the hybrids but were in the region shared in common with them. This showed that it could not have been a PCR template for a hybrid artefact. The remaining two members of the clone had V_H3–30 hybridized to the same original V_H3–07 rearrangement, and shared several mutations with the other hybrids that were not present in the sequence with the complete V_H3–07. These findings alone are very good evidence that this type of V_H replacement is a real in vivo event. Furthermore all the hybrids could be amplified using primers for the common CDR3, and the different CDR2s. In the other case, 49/53 related sequences had a complete V_H4–59. The other four sequences were hybrids with three different V_H segments. All the mutations in the complete V_H were also present in the hybrids so PCR artefact cannot be excluded on these grounds. However, both alleles of V_H4–61 were present in hybrids and therefore, if this occurred in vivo, there must have been exchange between chromosomes. The other hybrid in this case was with V_H4–04, which is downstream of V_H4–59 and therefore should have been deleted in the initial VDJ recombination. The authors suggest that hybridization with V_H4–04 was the initial trans-chromosomal event and that this was subsequently replaced by the V_H4–61 allele upstream on the same chromosome. This is borne out by the mutation pattern.

In both lymphomata, all the hybrids contained donor and acceptor V_H segments from the same family. This agrees with our conclusion from the earlier papers that homology is involved in Type 2 recombination. In the first case, the likely point of exchange does not contain a cryptic RSS. In the second case, there are several potential cryptic RSS, one of which is shared by all hybridizing genomic segments. Thus this paper provides strong evidence that Type 2 V_H replacement is a real in vivo phenomenon, that it depends on sequence homology, and that, though cryptic RSS are frequently present in the cross-over region, this is not always the case.

Cleavage at RSS by RAG proteins is affected by the DNA sequence in the coding flank^46,47 and non-favourable sequences tend to receive a single-strand nick but not double-strand cleavage. Poor cryptic RSS away from the standard 3′ position may therefore be more likely to be nicked. Though the RAG proteins direct DSBs to repair by NHEJ, these nicks are liable to engender repair by homologous recombination⁴⁸ so, if there were any RAG activity, cryptic RSS could well be sites of origin of Type 2 replacement by homologous recombination. However, most of the examples of Type 2 replacement come from mutating cells in which most investigators believe RAG genes are not expressed.

One of the hybrids of Itoh et al.³⁷ shows no mutation before or after recombination. Recently further examples of V_H–V_H hybrids have been published⁴⁹ in human immature B cells that have developed in non-obese diabetic/severe combined immune deficiency mice and show no somatic hypermutation. The five hybrid sequences in this paper are all between the same two family 4 V_H segments, with one exchange region in two and another in the other three, and again cryptic RSS have been suggested in each region. One of these has a heptamer GACTGTT which seems very unlikely to act as an RSS because it contains neither CAC nor GTG. Though the authors experimented with producing PCR hybrids and concluded that the published sequences were probably not artefacts, there is nothing in the sequences themselves that can verify this.

Exceptions obscuring the difference between types 1 and 2

Though there is usually deletion and addition of nucleotides in Type 1 V_H replacement and, in the common subtype, alteration of the V_H–D junction, there are exceptions. One such was seen in one of the first cases reported.⁵⁰ Two sequences from a mouse B-cell lymphoma line were found to have different V_H segments but to be identical from the beginnings of their V_H–D junctions onwards so that this might appear to be a case of Type 2 V_H replacement in which the cross-over was within the last 3 bp at the 3′ ends of the V_H segments. The authors, however, believed that the replacement had occurred by the mechanism that we now call Type 1. In the V_H gene segments concerned, the distance between the 3′ cryptic heptamer and the 3′ end of the segments is 5 bp and there is a difference in the second base to the 3′ of the heptamer. For the replacement to have occurred by the Type 1 mechanism described above, the new and old V_H would have had to have been modified and joined in such a way that the V_H was replaced seamlessly with neither frame-shift nor any change in the V_H-D junction. Later the same authors¹⁹ published the same replacement and a replacement of yet another complete V_H to the same original, again with no frame-shift nor change in the junction. The finding of two such replacements to the same substrate is very unlikely unless either they were Type 2 or some other mechanism was operating to ensure perfect match.

It is now known that there is another mechanism that can explain their observations. In mice, particularly fetal and neonatal ones^51,52 and in fetal and neonatal humans⁵³ pairs of coding ends to be joined in V(D)J rearrangement may be overlapped and spliced at regions of sequence identity as little as 1–5 bp in length. This is more frequent in the absence of TdT (references in Ref 52) or Ku86, which also seems to be involved in the addition of extra nucleotides.⁵⁴ In the light of knowledge of the place of non-homologous recombination in V(D)J rearrangement, the explanation is clear. After non-templated addition of nucleotides to the coding ends, the formation of the join by the matching of ‘microhomologies’ in these sequences is not apparent because these nucleotides are not present in the genomic sequences. However, when there is no addition of nucleotides, the matching of microhomologies is apparent because these sequences belong to the genomic segments. After cleavage at the internal heptamer in the V_H portion of a rearranged V_HDJ_H, the stump of the old V_H is effectively the new 5′ end of the D that is to be joined to the new V_H so the substrate is there for the type of recombination described above. This mechanism could not account for most of the reported sequences that appear to be Type 2 V_H replacement because cleavage of the two V_H gene segments at equivalent positions would leave no homologous ends to overlap. However, it has also been reported that, in transfected V(D)J recombination substrates in Ku86 deficient mouse cell lines, hybrid joints are abundant, and that the majority of them have no nucleotide excision from either of the ends and no nucleotide addition.⁵⁵

The timing of events

Most of the examples proposed as Type 2 V_H replacement have been in cells undergoing somatic hypermutation (SHM). Therefore the contention that cryptic RSS are involved requires that RAG proteins are re-expressed in mature, mutating cells (unless there is another, unknown, protein that recognises RSS). Though Itoh et al.³⁷ found RAG mRNA in several of their synovial tissue samples, they did not demonstrate that it was expressed as protein or that it came from mutating cells, and Zhang et al.³⁵ concluded that Type 1 V_H replacement does not occur later than the immature B-cell stage.

AID is essential not only for SHM but also for immunoglobulin gene conversion and class-switching, and the initiating event in each case is believed to be the deamination of deoxycytosine to uracil.⁵⁶ Gene conversion is a non-reciprocal form of homologous recombination in which a section of a gene is replaced by copying from another without alteration of the template gene. This is the primary method of producing antibody diversity in birds and several investigated mammals but not primates and rodents.⁵⁷ Although gene conversion has not been observed in rearranged human V_H segments, it has recently been shown that the expression of AID in human B cells is associated with the appearance of double-strand breaks with single-stranded protruding ends which are repaired by homologous recombination.⁴³ Therefore, if Type 2 V_H replacement occurs by homologous recombination it could well appear in mutating cells and the mechanism might be gene conversion rather than reciprocal exchange.

By homologous recombination, one double- or single-stranded break in the rearranged V might be repaired by crossing over of strands, using as the template an unrearranged V segment on the same chromosome, the sister chromatid (in the G₂ phase of the cell cycle), or even on the homologous chromosome, in which case a downstream V segment might be used. Such a repair would not generate a deletion circle (as shown in Fig. 2b) and this could be the reason why Wilson et al. failed to detect circular recombination products, though it is not the only possibility.³⁶

There are however, two problems with this: (1) either the length of the repair must always be so long that only one cross-over point is within the V_H or there must be unrecognized cases, in which both ends of the piece of new V-segment are within the old V segment, as in other species; (2) how to explain the fact that cryptic RSS are found in the cross-over region in almost all cases if they are not involved. We have therefore re-examined the published examples of Type 2 V_H replacement for AID recognition sites in the V–V cross-over regions that could explain replacement by AID-initiated homologous recombination, and discovered answers to both problems.

It has recently been shown⁵⁸ that AID prefers dA or dG at the −1 position to the targeted cytosine, and dA or dT at the −2 position. This gives WRC – reverse complement GYW – which corresponds to part of the previously identified RGYW/RGY hypermutation hotspot motif (reverse complement WRCY/RCY). Of the different V_H–V_H junctional regions shown by Wilson et al.³⁶ 13/15 have the complete WRCY·RGYW motif in the hybrid, or in the region of the germline sequence where hybridization occurred, and the others contain WRC·GYW or RCY·RGY. Likewise, the sequences of Itoh et al.³⁷ and Lenze et al.³⁸ have AID target motifs in the exchange regions. Furthermore, those of Itoh et al.³⁷ include four examples suggestive of gene conversion, i.e. an internal section of the V_H resembles a different germline segment while both ends resemble the parent rearrangement. There is insufficient room for a detailed analysis here, and we shall discuss it more fully elsewhere, but give a brief description here of one of these examples. Two rearranged genes with the same HCDR3 were presented (their Fig. 5B), one (ST1R G1) resembling a V_H1–02/V_H3–09 hybrid and the other (ST1R G2) a V_H1–69/V_H3–09 hybrid, which we shall refer to as G1 and G2. On closer inspection, both sequences resemble V_H1–69 at their 5′ ends and the section of G1 that exclusively matches V_H1–02 is confined to nucleotides (nt) 148–222. G1 therefore seems to have resulted from a substitution of a section of V_H1–02 into the G2 rearrangement followed by further hypermutation. The matching parts of both sequences contain several mutations not present in the other, which seems to exclude the possibility of PCR artefact. This and the other three examples are therefore consistent with gene conversion.

RGYW motifs are common in V_H gene segments and the hybrid sequences of Kolar and Capra⁴⁹ have the motif in their exchange regions though the hybrids are from immature cells that would not be expressing AID. However, PCR artefact cannot be ruled out for these sequences, and could have occurred from template DNA with double strand breaks at RGYW motifs.

Concluding remarks

There are many examples of recombination of V_H gene segments into previously existing VDJ sequences by a process that we have named Type 1 V_H replacement. Before its recent decisive investigation³⁵ this was already clearly a RAG-mediated process involving the RSS of the incoming V_H and a cryptic RSS within the 3′ end of the originally recombined V_H gene segment. It occurs in immature B cells in which the RAG genes are normally active. Evidence has also been presented^36–38,49 for V_H replacement by another mechanism, that we have called Type 2, of which most examples are from actively mutating B cells. It resembles homologous recombination between two V_H segments, and does not bear the hallmarks of NHEJ. It has not gained acceptance so far because similar hybrid sequences can be produced as PCR artefacts and it was believed to be mediated by RSS and RAG proteins, though RAG genes usually are suppressed in hypermutating cells. Those who would dismiss the evidence on the grounds that a single specific search⁴⁵ has failed to find this type of V_H replacement should remember that Type 1 V_H replacement was also previously dismissed on the basis of failure to find it in a specific search.¹⁷ Detailed examination of the sequences shows that (a) some of them are extremely unlikely to have been produced by PCR artefact (b) AID target motifs can be found in the regions of exchange between the V_H segments (suggesting that the finding of cryptic RSS may have been a red herring), and (c) in some cases the region replaced does not reach either end of the original V_H. Therefore, though hybrids undoubtedly occur as an artefact during PCR, and may account for some of the published examples, we believe that Type 2 V_H replacement probably does occur in vivo and results from AID-mediated homologous recombination. Whether the exchange is reciprocal homologous recombination or gene conversion can only be determined experimentally but, in view of the results from other species, the most reasonable model is gene conversion with replacement often including the 5′ end of the V_H.

Abbreviations

AID

activation-induced cytidine deaminase

DSB

double-strand break

NHEJ

nonhomologous end-joining

nt

nucleotide

RSS

recombination signal sequence(s)

SHM

somatic hypermutation

TdT

terminal deoxynucleotidyl transferase