Targets of somatic hypermutation within immunoglobulin light chain genes in zebrafish

Abstract

In mammals, somatic hypermutation (SHM) of immunoglobulin (Ig) genes is critical for the generation of high-affinity antibodies and effective immune responses. Knowledge of sequence-specific biases in the targeting of somatic mutations can be useful for studies aimed at understanding antibody repertoires produced in response to infections, B-cell neoplasms, or autoimmune disease. To evaluate potential nucleotide targets of somatic mutation in zebrafish (Danio rerio), an enriched IgL cDNA library was constructed and > 250 randomly selected clones were sequenced and analysed. In total, 55 unique VJ-C sequences were identified encoding a total of 125 mutations. Mutations were most prevalent in V_L with a bias towards single base transitions and increased mutation in the complementarity-determining regions (CDRs). Overall, mutations were overrepresented at WRCH/DGYW motifs suggestive of activation-induced cytidine deaminase (AID) targeting which is common in mice and humans. In contrast to mammalian models, N and P addition was not observed and mutations at AID hotspots were largely restricted to palindromic WRCH/DGYW motifs. Mutability indexes for di- and trinucleotide combinations confirmed C/G targets within WRCH/DGYW motifs to be statistically significant mutational hotspots and showed trinucleotides ATC and ATG to be mutation coldspots. Additive mutations in VJ-C sequences revealed patterns of clonal expansion consistent with affinity maturation responses seen in higher vertebrates. Taken together, the data reveal specific nucleotide targets of SHM in zebrafish and suggest that AID and affinity maturation contribute to antibody diversification in this emerging immunological model.

Introduction

A hallmark of the adaptive immune system is the capacity to mount a heightened memory response to pathogens encountered upon sustained or recurrent infection. Within mammalian models, an integral part of this protection stems from the ability of the immune system to fine-tune its diverse repertoire of antigen receptors over time through mutation and selection.¹ In the case of B cells and antibody responses, initial receptor diversity is created by V(D)J rearrangement of immunoglobulin gene segments to form functional immunoglobulin genes. The modular nature of the immunoglobulin segments, imprecise joining, and addition of nucleotides at V(D)J junctions generate an initial repertoire of naïve B cells with membrane-bound antigen receptors (B-cell receptors).

When a pathogen triggers an immune response, B cells with specificity to antigen rapidly relocate to the T/B-cell interface of lymphoid tissues where they are stimulated to proliferate through interactions with T helper (Th) cells and appropriate cytokines. The cellular outcome of the resultant daughter cells is thought to bifurcate down one of two pathways. Cells either terminally differentiate into post-mitotic short-lived antibody-producing plasma cells or infiltrate germinal centres of lymphoid follicles or other comparable structures and undergo massive and rapid clonal expansion. During this expansion, B cells can be subject to somatic hypermutation (SHM) in which point mutations are introduced into variable (V) regions of rearranged immunoglobulin DNA. This targeted mutation is thought to be largely responsible for generating subsets of B cells with slightly different affinities to the antigen.

In mice and humans, the mutation rate in V regions is estimated to be quite high, at 10⁻³ per base per generation, that is, 10⁶-fold higher than the rate of background mutations in DNA.^2,3 In mammals, mutations in rearranged immunoglobulin gene segments have been found to cluster in CDRs of V_H and V_L, which together encode the regions of the antibody most likely to influence affinity for antigen. Following hypermutation, B cells encoding receptors with greater specificity for antigen can undergo positive selection⁴ through a process referred to as ‘affinity maturation’. Selected B cells may further differentiate into either long-lived plasma cells or memory B cells which contribute to enhanced antibody responses upon reinfection by pathogen.

At present, bony fishes represent one of the most ancestral groups of animals known to produce antibodies for which high-quality genome sequences have become available. The availability of genomic sequences has recently facilitated genome-wide annotations of the germline segments available for V(D)J rearrangements. Prior to such annotations, the repertoire of immunoglobulin segments available for rearrangements was often unknown, making alignment of expressed transcripts with concordant genomic segments difficult to ascertain. Recently, using a full annotation of the zebrafish (Danio rerio) IgH locus,⁵ Weinstein et al.⁶ analysed an expressed repertoire of immunoglobulin H (IgH) gene segments in zebrafish to determine V_H family usage. Previous work in our laboratory⁷ coupled genomic annotation with expression data to reveal VJ-C expression of IgL loci from five different chromosomes in zebrafish. In the present study, we constructed an enriched cDNA library of zebrafish IgL in order to generate a sufficient quantity of VJ-C transcripts with which to ascertain potential nucleotide targets of somatic hypermutation. Collectively, the data presented herein reveal several patterns of mutation in the IgL of zebrafish and extend understanding of the processes underlying the generation of antibody diversity in this emerging immunological model.

Materials and methods

Targeted IgL loci

Rearrangements from concordant immunoglobulin segments from BACzK158E13 depicted in Fig. 1 were targeted for this study. Our decision to target exons that could be anchored to this BAC was based on the lack of gaps and the fact that these IgL loci were isolated in a single region on chromosome 19 far from other potential IgLs. Our previous per cent matrix and tree analyses⁷ showed that zebrafish IgLs were confined to a single chromosomal region group together, indicating that loci of a single chromosomal region are more likely to be similar to one another than those found in other parts of the genome. By targeting a single V_L, it was predicted that sufficient numbers of unique transcripts could be obtained to facilitate both identification of potential nucleotides targeted for mutation and groups of nucleotides comprising mutational hotspot motifs. In addition, by targeting a single V_L, each VJ-C transcript could be aligned to both targeted and neighbouring V_L to ensure that concordant assignments were made and that deviations ascribed to mutations were not alterations of closely related gene segments.

Figure 1.

Targeted zebrafish immunoglobulin L (IgL) germline reference sequence. The existing genomic annotation of zebrafish IgL was extended by assembling overlapping BAC inserts. BACs with overlapping end reads were prioritized for sequencing at the Sanger Center (http://www.sanger.ac.uk/Projects/D_rerio). The 923-kb contiguous chromosomal tiling path was manually assembled using the Artemis Annotation Software package. The IgL gene segments targeted are clustered in a single region on zebrafish chromosome 19. Immunoglobulin segments at this locus are divergent, enabling alignment of cDNA sequences with germline sequences to determine patterns of somatic hypermutation (SHM). Each BAC is designated by its corresponding NCBI accession number and drawn approximately to scale, while the IgL locus is expanded with exon sizes exaggerated.

The zebrafish V7 locus

Initial polymerase chain reaction (PCR) studies (data not shown) revealed that primers in V1 and V7 leader sequences in combination with primers to C1, C2, C3, C4 or C5 C regions more readily amplified VJ-C sequences from a panel of cDNA derived from zebrafish of various ages. Because the V1 heptamer (CACAGTA) deviates at the seventh base while the V7 heptamer and nonomer are canonical, we selected V7 as the target locus in which to analyse potential mutational events in adult zebrafish. Our decision to target an IgL locus with a canonical recombination signal sequences (RSS) was made on the assumption that such a locus would be likely to yield the highest number of transcripts. This assumption is supported by previous work by Ramsden and Wu,⁸ who found that synthetic substrates of mouse origin showed favoured expression of V segments flanked by RSS that more closely resemble canonical heptamer (CACAGTG) and nonamer (ACAAACC) sequences. By focusing on a single rearranged IgL gene (V7) and constructing an enriched cDNA library from a single individual, direct comparison of sequences could be utilized to discern mutational hotspots and potential lineages of clonal expansions could be examined.

Animals and RNA extraction

Zebrafish embryos (Tübingen line) were obtained from the Zebrafish International Resource Center (Eugene, OR) and raised under standard conditions⁹ to establish laboratory equilibrium of normal immune function. The Tübingen line of fish was chosen as reference genomic segments from the BAC zK158E13 clone were derived from fish of the Tübingen strain. A single adult zebrafish (2 years of age) was anaesthetized with MS222 (Sigma Chemicals, ST. Louis, MO) from which organs were harvested. Upon removal, the haematopoietic tissues (pronephros, mesonephros and spleen) were pooled, snap-frozen in liquid N₂, and held at −80° prior to RNA extraction.

cDNA synthesis, library construction and cloning of VJ-C rearrangements

Total RNA was extracted with Trizol (Life Technologies, Carlsbad, CA) and poly(A)-enriched mRNA was isolated from total RNA using the Micro-Fast Track mRNA Isolation Kit (Invitrogen, Carlsbad, CA). Poly(A) mRNA was reverse-transcribed into cDNA according to the manufacturer’s protocol using an RLM-RACE kit (Ambion, Carlsbad, CA). Briefly, first-strand cDNA synthesis was initiated with an oligo-dT adapter primer. cDNA was then subjected to PCR using forward primers that span the V7 leader/exon junction of BAC ZK158E13 and reverse primers anchored to the adapter sequence added using the RACE kit. One-sided nested PCR was then carried out using V7 primers and reverse primers designed to amplify cDNA corresponding to conserved regions of C1, C2, C3, C4 or C5 identified on BAC clone ZK158E13 (primers listed in Table 1). Products were run on agarose gels, bands of appropriate sizes were excised using a QIAquick Gel Purification kit (Qiagen, Valencia, CA), and fragments were cloned into TOPO T/A pCR2·1 cloning vectors (Invitrogen) and transformed into TOP10 cells (Invitrogen). Colonies were picked by blue/white screening, expanded, and maintained in agar stabs. In total, plasmids from > 250 colonies were purified (Qiagen miniprep) and EcoRI (New England Biolabs, Beverly, MA) restriction digests were performed to identify clones with inserts.

Table 1.

Primers and polymerase chain reaction (PCR) or reverse transcription conditions

Targeted transcript	Primer	Conditions
V7	5′-TGACTGTAGTGACTCAGAGTCC-3′	5 min at 94°; 30 cycles (30 seconds at 95°, 30 seconds at 50–55° and 60 seconds at 72°); 10 min at 72°
C1/C2/C3/C4/C5	5′-GCTCAGGCTGCTGCTCCAGC-3′
C5	5′-TGTACAGTCCATCCTC-3′
EF1α	FWD: 5′-CCTGGTGACAACGTTGGCTT-3′RVS: 5′-GAACGGTGTGATTGAGGGAA-3′	4 min at 94°; 30 cycles (30 seconds at 95°, 30 seconds at 56°, 60 seconds at 72° and 10 min at 72°); 10 min at 72°
Random hexamers	Invitrogen (product no. 48190-011)	10 min at 25° and 30 min at 42°
Oligo-dT(20-VN)1	Invitrogen (product no. 12577-011)	30 min at 42°
Oligo-dT adapter	5′-GCGAGCACAGAATTAATACGACT-3′	1 hr at 42°
3′ RACE adapter	5′-GCGAGCACAGAATAATACGACTCACTATAGG(dT)-3′	3 min at 94°; 35 cycles (30 seconds at 94°, 30 seconds at 60° and 30 seconds at 72°); 7 min at 72°

¹Oligo-dT_(20-VN) is mixture of 12 primers, each a string of 20-dT residues followed by two additional variable nucleotides (VNs). The VN ‘anchor’ targets primer annealing at the 5′ end of the poly(A) tail.

EF1α, elongation factor 1α; RACE, rapid amplification cDNA ends.

DNA sequencing

Clones with inserts were sequenced bi-directionally using universal M13 forward and M13 reverse primers at the Clemson University Genomics Institute (Clemson, SC). Plasmid vectors and PCR primers were trimmed from sequences, and overall sequence quality and automated sequencing calls were verified by inspection of each sequence chromatogram. In total, 220 VJ-C clones were identified to have high-quality forward and reverse complement sequencing reads.

Alignment of expressed VJ-C sequences to genomic regions

The 220 VJ-C sequences were compared against the non-redundant nucleotide database at national center biological information (NCBI) using the megaBLAST algorithm. Sequences were subsequently compared against V_L, J_L, and C_L identified⁷ in zebrafish using the Matrix Global Alignment Tool.¹⁰ All of the 220 VJ-C clones had highest identity to the targeted V7 gene segment with at least 95% identity. The stringent 95% requirement of IMGT/V-QUEST¹¹ was employed, as the existence of additional V_L cannot be ruled out from the genome. Given that the per cent variability in nucleotide sequences of identified zebrafish V_L ranges from 43 to 93% overall, with V_L on Chr 19 (V1–V8) ranging from 49·5 to 83·6%, a 95% criterion is suitably rigorous. The resultant 220 VJ-C sequences were aligned to germline IgL segments using ClustalW,¹² and CDRs and frameworks (FRs) were defined using the rules of Kabat.¹³ In total, from the 220 VJ-C clones, 55 unique sequences containing a total of 125 mutations from concordant germline immunoglobulin segments were identified. Because any somatic mutation could in theory be carried during the clonal expansion of a single B cell or be amplification of the same transcript by reverse transcriptase (RT)-PCR, identical VJ-C sequences were deemed to represent a single B-cell population and therefore counted only once in the mutational analyses. In addition, as mutations can be additive if resultant B cells are derived from a common founder, the 55 unique VJ-C sequences were scored for mutations exclusive for that clone and with all mutations included. Unique VJ-C sequences were submitted to NCBI (accession numbers in Table 2).

Table 2.

Distribution of immunoglobulin L (IgL) mutations in zebrafish VJ-C cDNA clones

Accession number	Region1	Mutation2,3	Type4	Codon5
EU795304	VL-CDR1 (50)	AGT→AGC	T	Silent (Ser)
	VL-CDR2 (118)	GGA→GAA	T	Gly→Glu
	VL-CDR2 (158)	AGT→AGG	V	Ser→Arg
	VL-CDR3 (259)	TGC→TAC	T	Cys→Tyr
EU795305	VL-CDR2 (137)	TCT→TCC	T	Silent (Ser)
	VL-FR3 (186)	CTG→TTG	T	Silent (Leu)
EU795306	VL-CDR3 (259)	TGC→TAC	T	Cys→Tyr
EU795308	VL-CDR3 (259)	TGC→TAC	T	Cys→Tyr
	CL (126)	GGC→AGC	T	Gly→Ser
EU795309	VL-CDR3 (259)	TGC→TAC	T	Cys→Tyr
	CL (112)	C, insertion	NA	Frameshift
	CL (154)	C, insertion	NA	Frameshift
EU795310	VL-CDR1 (65)	GAC→GAT	T	Silent (Asp)
	VL-CDR3 (259)	TGC→TAC	T	Cys→Tyr
EU795311	VL-FR3 (196)	CGT→ CAT	T	Arg→His
	VL-CDR3 (259)	TGC→TAC	T	Cys→Tyr
	CL (138)	C, insertion	NA	Frameshift
EU795313	VL-CDR3 (259)	TGC→TAC	T	Cys→Tyr
	CL (17)	GGC→GC	NA	Frameshift
EU795314	VL-FR2 (87,88)	CCT→CTG	T,V	Pro→Leu
	VL-FR2 (92)	GGA→GAA	T	Gly→Glu
	JL (18)	T, insertion	NA	Frameshift
	CL (23)	GCG→GCA	T	Silent (Ala)
	CL (34)	CCC→CTC	T	Pro→Leu
EU795315	VL-CDR3 (259)	TGC→TAC	T	Cys→Tyr
	CL (22)	C, insertion	NA	Frameshift
EU795316	VL-FR2 (80)	TTG→TTA	T	Silent (Leu)
	VL-CDR3 (259)	TGC→TAC	T	Cys→Tyr
EU795318	VL-CDR2 (166)	GGA→GAA	T	Gly→Glu
	VL-CDR3 (259)	TGC→TAC	T	Cys→Tyr
EU795319	VL-FR2 (82)	CAG→CGG	T	Gln→Arg
	VL-FR2 (93)	AAA→GAA	T	Lys→Glu
	VL-CDR3 (259)	TGC→TAC	T	Cys→Tyr
	CL (54)	AAG→GAG	T	Asp→Gly
EU795320	VL-CDR3 (259)	TGC→TAC	T	Cys→Tyr
	CL (17)	GGC→GC	NA	ORF intact
	CL (22)	C, insertion	NA	‘ ’
EU795321	JL (14)	GGC→TGC	V	Gly→Cys
EU795322	VL-CDR3 (259)	TGC→TAC	T	Cys→Tyr
	JL (24)	C, insertion	NA	Frameshift
EU795323	VL-FR1 (14)	GCA→GCC	V*	Silent (Ala)
	VL-FR1 (15)	GGG→AGG	T	Gly→Arg
	VL-FR1 (19)	GAT→GTT	V	Asp→Val
	VL-FR1 (21)	TCT→CCT	T	Ser→Pro
	VL-FR1 (27)	TCT→CCT	T	Ser→Pro
	VL-FR1 (30,31)	ATC→TCC	V,T	Ile→Ser
	VL-FR1 (33)	AGC→GGC	T	Ser→Gly
	VL-CDR3 (259)	TGC→TAC	T	Cys→Tyr
EU795324	VL-CDR3 (259)	TGC→TAC	T	Cys→Tyr
	CL (154)	C, insertion	NA	Frameshift
EU795326	CL (63)	GCT→ACT	T	Ala→Thr
EU795328	VL-CDR3 (259)	TGC→TAC	T	Cys→Tyr
	CL (95)	TTT→TAC	T	Silent (Phe)
EU795329	VL-FR3 (217)	GCA→GAA	V	Ala→Glu
EU795330	VL-FR3 (181)	TTT→TAT	V	Phe→Tyr
EU797178	VL-FR2 (98)	GCT→GCC	T*	Silent (Ala)
EU797179	VL-CDR1 (41)	ACT→ACC	T*	Silent (Thr)
	VL-CDR3 (259)	TGC→TAC	T	Cys→Tyr
	JL (30)	GTT→GAT	V	Val→Asp
EU797180	VL-CDR1 (58)	GGT→GCT	V	Gly→Ala
	VL-FR2 (103)	CCT→CCC	T*	Silent (Pro)
EU797181	VL-CDR3 (261)	AGT→GGT	T	Ser→Gly
EU797183	VL-FR2 (82)	CAG→CGG	T	Gln→Arg
	JL (11,12)	GGA→AAA	T,T	Gly→Lys
EU797184	VL-CDR2 (137)	TCT→TCC	T*	Silent (Ser)
EU797185	VL-FR3 (208)	CCT→CAT	V	Pro→His
	VL-FR3 (212)	GAA→GA	NA	Frameshift
	VL-FR3 (214)	GAT→GCT	V	Asp→Ala
EU797186	CL (122)	AAG→AAA	T	Silent (Lys)
EU821496	VL-CDR2 (124)	AGT→AAT	T	Ser→Asn
	VL-CDR3 (259)	TGC→TAC	T	Cys→Tyr
EU821497	VL-CDR3 (259)	TGC→TAC	T	Cys→Tyr
	JL (30)	GTT→GAT	V	Val→Asp
EU821498	VL-FR3 (218)	GCA→GCG	T*	Silent (Ala)
	VL-CDR3 (259)	TGC→TAC	T	Cys→Tyr
EU821500	VL-FR2 (97)	GCT→GTT	T	Ala→Val
	VL-CDR3 (259)	TGC→TAC	T	Cys→Tyr
EU821501	VL-CDR1 (71)	AGC→AGT	T	Silent (Ser)
EU821502	CL (78)	GTG→CTG	V	Val→Leu
EU821503	VL-CDR1 (64)	GAC→GGC	T	Asp→Gly
	VL-CDR2 (176)	CCT→CCC	T*	Silent (Pro)
EU821504	VL-CDR2 (172)	GAA→GGA	T	Gln→Gly
	VL-CDR3 (259)	TGC→TAC	T	Cys→Tyr
	JL (30)	GTT→GAT	V	Val→Asp
EU821505	VL-CDR1 (65)	GAC→GAT	T	Silent (Asp)
	VL-CDR3 (259)	TGC→TAC	T	Cys→Tyr
	CL (43)	GAT→GGT	T	Asp→Gly
EU821507	VL-FR1 (34)	AGC→ACC	V	Ser→Thr
	VL-CDR1 (56)	GTT→GTC	T*	Silent (Val)
	VL-CDR1 (71)	AGC→AGT	T	Silent (Ser)
EU821508	VL-CDR1 (44)	GGG→GGA	T*	Silent (Gly)
	VL-CDR3 (259)	TGC→TAC	T	Cys→Tyr
EU821510	VL-FR3 (180)	TTT→GTT	V	Phe→Val
EU821511	VL-FR1 (34)	AGC→ACC	V	Ser→Thr
	VL-CDR1 (71)	AGC→AGT	T	Silent (Ser)
EU821512	VL-CDR3 (259)	TGC→TAC	T	Cys→Tyr
	CL (78)	GTG→ATG	T	Val→Ser
EU821513	VL-CDR3 (235)	ATG→ACG	T	Met→Thr
	VL-CDR3 (259)	TGC→TAC	T	Cys→Tyr
	JL (30)	GTT→GAT	V	Val→Asp
EU821518	VL-FR1 (23)	TCT→TCC	T*	Silent (Ser)
	VL-FR2 (97)	GCT→GTT	T	Ala→Val
	VL-CDR2 (129)	CTT→TTT	T	Leu→Phe
	VL-CDR3 (259)	TGC→TAC	T	Cys→Tyr
EU821519	VL-FR1 (27)	TCT→CCT	T	Ser→Pro
	VL-CDR3 (259)	TGC→TAC	T	Cys→Tyr
	CL (28)	CTT→CGT	V	Leu→Arg
EU821520	VL-FR1 (32)	ATC→ATT	T	Silent (Ile)
	VL-FR3 (218)	GCA→GCG	T*	Silent (Ala)
	VL-FR3 (222)	GTT→TTT	V	Val→Phe
	VL-CDR3 (259)	TGC→TAC	T	Cys→Tyr
EU821521	VL-CDR1 (71)	AGC→AGT	T	Silent (Ser)
	VL-FR3 (181)	TTT→TCT	T	Phe→Ser
EU821522	VL-CDR3 (259)	TGC→TAC	T	Cys→Tyr
	JL (29)	GTT→ATT	T	Val→Ile
EU821523	VL-CDR2 (124)	AGT→AAT	T	Ser→Asn
	CL (95)	TTT→TT	NA	Frameshift
EU825202	CL (152)	T, insertion	NA	Frameshift
EU825203	VL-CDR2 (125)	AGT→AGC	T	Silent (Ser)
EU825204	VL-FR3 (196,197)	CGT→CAC	T, T	Arg→His
	VL-CDR3 (259)	TGC→TAC	T	Cys→Tyr
EU821516***

¹V_L, variable region; J_L, joining region; FR, frame work region; CDR, complementarity-determining region.

²Mutated nucleotides are underlined, and depicted within triplet bases of codons.

³Bases shaded correspond to targeting of the G and C nucleotides of DGYW/WRCH^** AID hotspot motifs.

⁴Transition (T) or transversion (V) base mutations.

⁵Amino acids resulting in a change in side-chain polarity are shown in italics.

^*Designates mutation in the wobble position of a degenerate codon. For example, in the third position of glycine codons (GGA, GGC, GGG and GGT) all nucleotide substitutions are synonymous (do not change the amino acid).

^**DGYW (AGT/G/CT/AT); WRCH (AT/GA/C/TAC).

^***Clone with no mutations from germline gene segments.

Tests for genomic contamination

cDNA utilized to construct cDNA libraries was subjected to PCR-based amplification using primers corresponding to the zebrafish elongation factor 1α (EF1α) housekeeping gene (Table 1). These EF1α primers amplify segments separated by an intronic region, thus facilitating detection of contaminating genomic DNA in cDNA preparations through the presence of larger EF1α bands on agarose gels. In each of the cDNA preparations used to generate VJ-C cDNA libraries, DNA contamination was not detected.

Taq polymerase fidelity assay

To calculate potential error rates arising from amplification or sequencing, a set of experiments was undertaken to generate subclones from one of our resultant IgL rearranged VJ-C clones (accession no. EU795303). This isolated plasmid was nicked and used as the template in PCR using gene-specific primers identical to those used in the cDNA library construction. Products were run out on agarose gels, bands were purified, amplicons were cloned into pCR2·1 cloning vectors, and plasmids were transformed into TOP10 cells (Invitrogen). Twelve subclones were randomly selected for bi-directional sequencing and no mismatches were identified in 11 280 resultant bases. Thus, the PCR amplification error rate can be considered negligible and few, if any, of the base pair changes found in the VJ-C cDNA clones warrant being ascribed to PCR or sequencing errors.

Calculation of mutability indexes

Mutability indexes were calculated as described by Shapiro et al.¹⁴ Briefly, a mutability index is a measure of observed/expected mutations. The observed number of mutations within a mono-, di- or trinucleotide (the target) is divided by the number of times the specific base or group of adjacent bases would be expected to be mutated for a mechanism without target bias. A mutability index score of 1·0 is assumed to represent random (unbiased) mutation whereas higher scores imply that the specific bases are targeted by mutation or selected over other mutations during B-cell development. Mutability indexes were calculated for each of the mononucleotides (A,T,C and G), the 16 dinucleotide combinations, and 64 possible trinucleotides of the germline V_L region sequenced. The SMS DNA Software Suite¹⁵ and Microsoft Excel were used for calculations and database management. Initially, the number of times each target occurred in the germline V_L region was determined. For dinucleotide analysis, one extra nucleotide from adjacent germline sequence at the end of each region was included and two extra germline nucleotides were included for trinucleotides. The total number of each specific target present in the germline V_L was divided by the number of all targets to generate relative target frequencies. The use of relative target frequencies was necessary to normalize the data as each base or group of adjacent bases does not occur in equal proportions over the regions analysed. Finally, relative frequencies were multiplied by the total number of mutations to yield an expected mutation number. Observed numbers of mutations were divided by expected numbers for each target to obtain mutability index scores.

Statistical analyses

Chi-squared analyses of mono, di- and trinucleotide mutability indexes were carried out by contrasting observed mutational frequencies to their expected (non-biased) mutational frequencies. For mutability indexes, P values < 0·01 were considered statistically significant. In cases where mutations could be assigned to differentdi- or tri-nucleotide targets, Bonferroni corrections were applied. Chi-squared analyses were also performed to evaluate mutational frequencies and distributions of each WRCH/DGYW motif as reported within the results. Statistical analyses of antigen selection pressure on FR and CDR regions were carried out using the multinomial distribution model of Lossos et al.¹⁶ which is presently available as a JAVA applet at http://www-stat.stanford.edu/immunoglobulin. For multinomial distributions, an excess of CDR replacements or scarcity of FR replacements was judged significant at P<0·05.

Results

Somatic mutation occurs within V_L, J_L and C_L encoded regions of zebrafish IgL

Alignments of 207 682 VJ-C encoded nucleotides with concordant germline gene segments revealed 125 mutations over 55 unique VJ-C cDNA sequences (Table 2). The majority of mutations were found in V_L; however, the J_L and C_L regions were also found to have mutations. The percentage of total mutations in the V_L, J_L and C_L regions was 75, 8 and 17%, respectively. When weighted as the number of mutations per base sequenced in each gene segment region, the percentages became 0·65, 0·51 and 0·22. These results show that mutations are concentrated in the V_L regions in terms of both higher overall numbers and higher density when compared with the J_L and C_L regions.

Distribution of V_L, J_L and C_L mutations

The distribution of somatic mutations in V_L, J_L and C_L regions is depicted in Fig. 2. In this figure, mutation frequencies are plotted against position number, with position 1 being the initial base of the first framework region of the V7 gene segment. Mutation frequencies are illustrated for 20 base pair intervals along the regions sequenced. The relative locations of CDR regions are depicted by lines at the top of the diagram and the approximate locations of V_L, J_L and C_L are shown as boxes under the position number. Mutations were found in all regions, with the highest incidence being in the V_L CDR3 region. The J_L region, although small, exhibited a slightly higher overall mutation frequency rate than the collective region spanned by C_L. Within the C_L region, the mutation frequency decreased slightly with distance from V_L, with no mutations found in the last two segmental groupings of C_L.

Figure 2.

Distribution of mutation frequency and activation-induced cytidine deaminase (AID) WRCH/DGYW hotspot motifs in targeted zebrafish immunoglobulin L (IgL). The combined frequency of mutations (primary vertical axis) per base sequenced for 55 unique VJ-C cDNA sequences plotted against position number (horizontal axis) reveals variability in V_L, whereas mutations in C_L were infrequent. The additive number of WRCH/DGYW hotspot motifs (secondary vertical axis) in the targeted germline IgL when plotted against position number shows an apparent trend of increasing mutation at AID hotspots. Within mammals, WRCH/DGYW motifs have been found to be the principal hotspot for AID-induced G:U lesions in rearranged immunoglobulin genes during somatic hypermutation. The bi-directional arrows show locations of palindromic WRCH/DGYW hotspot motifs which coincide with the highest mutation frequencies of the zebrafish IgL. The lower diagram depicts locations of V_L, J_L and C_L regions with respect to overall distribution of the VJ-C sequence and the complementarity-determining regions (CDRs) are depicted in the graph plot area.

Absence of N and P addition

Interestingly, we did not find evidence of either N or P addition at the CDR3/J_L junctions in any of the cDNA sequences analysed. The assignment of nucleotides to the CDR3 and J_L regions at the CDR3/J_L junction was straightforward as, for each of the 55 sequences, bases at this junction could be assigned to the 3′ and 5′ ends of concordant germline sequences. We did, however, find a single clone (accession no. EU797180) for which three bases from the 3′ end of the germline V_L and three bases from the 5′ end of the germline J_L were not present. This six-base difference can probably be attributed to imprecision in rag-mediated double-strand breakage or exonuclease activity prior to joining. These six bases were therefore not counted in the somatic mutation frequency analysis as they would have been similarly absent in the original B-cell rearrangement and thus not subject to SHM. The lack of N and P addition and the almost uniform size of the coding junctions were surprising and suggest that recombination site diversity might be somewhat limited in zebrafish IgL.

Mutational bias towards single base transitions in zebrafish V_L

In total, 94 V_L mutations were identified among the 55 unique VJ-C sequences (Table 2). Of these, 93 were base substitutions and one was a single base deletion (EU797185). As insertions or deletions in non-triplet increments result in frameshifts, it is not surprising that their occurrence was rare in the VJ-C cDNA clones. B cells with non-productive V_L caused by frameshifts would inherently lack functional B cell receptor (BCR) and probably be selected against. Of the 93 V_L single base substitutions, base transitions (79; 85%) were more abundant than transversions (14; 15%). The transition to transversion ratio of V_L mutations was 5·64, which is considerably higher than the theoretical 1 : 2 ratio (P<0·01) if random bases were incorporated during substitutions. These results are consistent with analyses in mammalian models demonstrating a strong tendency for transitions by SHM.¹⁷ Notably, 24 of the 94 V_L mutations occurred in codon wobble positions (Table 2) and, of these, 96% (23/24) resulted in silent mutations. This ratio (23 : 1) is also higher than expected assuming random base substitution. These results indicate that transitions and silent mutation accumulation are prevalent in the zebrafish IgL analysed in this study.

Mutations are concentrated at WRCH/DGYW hotspot motifs

Overall, G-to-A exchanges were the most common mutation observed in the zebrafish V_L (41; 44%). In addition, C-to-T (11; 12%) exchanges were also present, albeit in lower numbers (Table 2). Either of these types of mutations could occur through fixation by AID-targeted deamination of deoxycytidine to deoxyuracil. For example, a C deaminated to a U on the coding strand could prompt incorporation of an A in the daughter strand at the complementary position during replication. In a subsequent round of DNA replication, if left unrepaired, the A in the daughter strand results in a fixed C-to-T mutation in the coding strand. Similarly, G-to-A exchanges can result from C deamination on the non-coding strand and two rounds of replication. As illustrated in Fig. 3, the majority of the cytidine mutations observed in V_L (8/14) were found at the widely accepted WRCY (refined WRCH¹⁸) AID mutation hotspot motif. Even more abundant was cytidine mutation (37/45) on the non-coding strand (reverse complement of the hotspot motif RGYW resulting in G mutations on the coding strand). These findings suggest that AID hotspot motifs are targets for mutations in zebrafish and resultant mutations appear biased to fixation in daughter strands by U to A base pairing during DNA replication.

Figure 3.

Base exchanges are biased to replacement (R) mutations and nucleotide targeting at activation-induced cytidine deaminase (AID) hotspot motifs (WRCH/DGYW). The majority of the base exchanges observed resulted in amino acid changes [replacements (R); black areas of bars]. Silent mutations (S; white areas), while found in all four bases, were proportionally higher outside of WRCH/DGYW hotspot motifs. These data suggest that neutral mutations may be more prone to accumulate in bases outside of hotspots whereas replacement mutations are favoured at AID hotspot motifs.

Uracil glycosylase DNA repair appears to be limited at AID hotspots

In total, 49% (46/93) of all V_L base exchanges and 76% (45/59) of all C/G mutations were found at WRCH/DGYW hotspot motifs. In addition, 89% (40/45) of the C/G mutations at these hotspots were G-to-A and C-to-T base exchanges. Both this hotspot targeting and the high frequency of these types of base exchanges suggest that the mutations observed at the WRCH/DGYW hotspot motifs positions may be attributable to AID activity. Moreover, given that AID induces C deamination to uracil in DNA, which, if left unrepaired, becomes fixed as G-to-A or C-to-T mutations following replication, it appears that uracil DNA glycosylase repair may be limited at AID hotspots in zebrafish. The remaining base exchanges (5/45) identified at these hotspots, however, suggest that base excision repair may not be entirely absent. For example, in these five cases (one G-to-T, three G-to-C, and one C-to-A), these mutations could have arisen if, after C deamination, uracil DNA glycosylase initially removed the corresponding deoxyuracil, thus creating an abasic site. Subsequent endonuclease activity at the lesion followed by error-prone DNA polymerase repair could in theory have been responsible for these G-to-T, G-to-C, and C-to-A mutations. Although deoxyuracil base excision repair could in theory generate a wider spectrum of base changes than C-to-T and G-to-A enabled by AID alone, it appears that diversification by this system is somewhat limited at WRCH/DGYW hotspot motifs in the zebrafish model.

Replacement mutations are favoured within and outside AID hotspot motifs

Overall, 74% (68/93) of the V_L base exchanges resulted in replacement mutations. Replacement (R) over silent (S) mutations were favoured in both the FR (26R; 11S) and CDR (42R; 14S) regions. When viewed in the context of WRCH/DGYW motifs, 89% (41/46) of the mutations at AID hotspots were replacements. To discern whether C/G targets of the AID motifs were more likely to be situated in either the first or the second codon position and hence by location be predisposed towards replacements, WRCH/DGYW positioning within the open reading frame of the targeted germline V_L was determined. This analysis showed that 89% (41/46) of the germline C/G targets in WRCH/DGYW motifs were in the first or second codon position. Given that this percentage is higher than the 66·6% expected if WRCH/DGYW targets were distributed equally among the three codon positions, it appears that there is a propensity for replacement mutations at these motifs. There was, however, no statistically significant difference between FR and CDR regions with respect to AID targets at wobble positions (P<0·62), meaning that, unless they are subject to other selective pressures, the mutations induced at AID hotspots in both FR and CDR regions are more prone to be replacements. Of the 46 mutations in WRCH/DGYW hotspots, 89% (41/46) occurred at non-wobble positions and, of these, 100% (41/41) resulted in replacements. These numbers are considerably higher than the 57% (27/47) replacement rate outside AID hotspot motifs (Fig. 3). Thus, the data suggest that non-random mutations are favoured, with a strong bias towards replacements in WRCH/DGYW motifs.

WRCH/DGYW mutations are highly prevalent in CDR3 regions

Using a pattern search algorithm (SMS), a total of 37 WRCH/RGYW motifs were identified in the V7 germline DNA (Fig. 4a). This number of AID hotspots is an overrepresentation (P<0·01) of this motif from a distribution of the motif expected in a random DNA sequence of this length. As shown in Fig. 4b, the proportion of bases present within WRCH/DGYW motifs in the different FRs and CDRs of V_L ranged from 39% (CDR3) to 48% (FR3). Despite this relatively narrow and consistent distribution range, the majority of the mutations observed (38/46) at WRCH/DGYW motifs were found in CDR regions (Fig. 4c), most notably CDR3 (31/46). These findings suggest the possibility that either WRCH/DGYW hotspot motifs in CDR1 and CDR3 regions are preferentially targeted or that mutations in these CDRs are disproportionately selected during B-cell maturation in zebrafish.

Figure 4.

Mutations at activation-induced cytidine deaminase (AID) hotspots are disproportionately concentrated at complementarity-determining region 3 (CDR3). (a) AID hotspot motifs (n = 37) are distributed across framework regions (FRs) and complementarity-determining regions (CDRs) of the targeted germline V_L. (b) The proportion of bases contained in WRCH/DGYW motifs varies slightly between FRs and CDRs. (c) Despite different densities of WRCH/DGYW motifs (a, b), mutations within WRCH/DGYW motifs are highly biased to occurrence in CDR3.

Palindromic WRCH/DGYW motifs appear to be disproportionately targeted for mutation

To discern if specific WRCH/DGYW motifs were preferentially targeted, the distribution of mutations at each WRCH/DGYW motif was determined (Fig. 5). In total, 37 WRCH/DGYW sequence motifs were present in the targeted V_L and, notably, at least one representative of each possible sequence combination was present (the number of each is shown in parentheses in Fig. 5). In addition, 12 of the motifs in this V_L were AGCT or TGCA palindromes. Both AGCT and TGCA conform to WRCY and DGYW in both directions. Although WRCH/WRCH palindromes represented only 32% (12/37) of the AID hotspot motifs in the targeted V_L, they accounted for 89% (41/46) of the mutations observed across all WRCH/DGYW motifs. AGCT/AGCT palindromes were found in FR1, CDR1 and FR2 (each region had one palindromic motif) with the number of mutations at each motif being 2, 4 and 2, respectively. The TGCA/TGCA palindromes were present in FR2, FR3 and CDR3 (also each having one palindromic motif) with respective mutation numbers at these motifs being 1, 1 and 31. These results suggest either preferential targeting of somatic mutation to palindromic AID hotspot motifs or an increased selection for resultant mutations at CDRs.

Figure 5.

Somatic mutations are highly concentrated in palindromic activation-induced cytidine deaminase (AID) hotspot motifs. Within mammals, WRCH/DGYW motifs are considered the principal hotspot for AID-induced cytidine deamination in rearranged immunoglobulin during somatic hypermutation (SHM) whereas palindromic WRCY/WRCY are often targeted during class switch recombination (CSR). In the zebrafish V_L targeted, every WRCH/DGYW sequence was present (the number of occurrences is listed in parentheses and the mutation number within each is depicted in bold). Overall, V_L mutations were disproportionately concentrated within palindromic AGCT/AGCT and TGCA/TGCA AID hotspot motifs.

WRCH/DGYW mutational targeting occurs on both DNA strands

Analyses of strand bias at AID hotspot motifs revealed that the majority (38/46) of the mutations at WRCH/DGYW motifs could be explained by C targeting in the template strand. The remaining (8/46) C mutations at WRCH/DGYW were within the coding strand. Although it cannot be determined if each mutation was an independent mutational event or a product of selection and clonal expansions of B-cell clones harbouring the mutation, it can be concluded that within the AID hotspot motifs both strands are targeted for mutation. Chi-squared analyses of the WRCH/DGYW mutations in both strands revealed that only TGCA and AGCT motifs were significant for G mutations in DGYW, and only TGCA and AGCT were significant for WRCH motifs (P<0·01). These results suggest that, when compared with all AID hotspots, either palindromic AID hotspot motifs are disproportionately targeted for mutation or the initial mutations at these motifs undergo positive selection.

Mutability indexes (MIs) reveal additional non-random nucleotide targeting

Mononucleotides

To determine whether additional nucleotides or combinations of adjacent nucleotides in the V_L coding region were preferentially targeted for mutation, MIs were determined. Each index is a normalized measure that takes into account the fact that each base or group of bases does not occur at the same frequency over the V_L region analysed. The MIs reported for mononucleotides (Table 3) show by index score that the nucleotides are preferentially mutated in the order G>C>T>A. Transversion to transition ratios at G, C, T and A were 0·10, 0·16, 0·23 and 0·44, indicating a strong preference for base exchanges resulting in transition mutations. Each of the four ratios is significantly different (P<0·01) from the theoretical 2 : 1 transversion to transition ratio that would be expected if base exchanges were random and nondiscriminatory. Thus, mononucleotide analyses suggest that neither the nucleotides targeted nor the resultant patterns of substitution can be explained by assuming that somatic mutation events occur randomly in the zebrafish V_L.

Table 3.

Substitutions and mononucleotide mutability indexes in V_L regions of zebrafish immunoglobulin L (IgL)

	Substitution
	A	C	G	T	Observed	Expected	Mutability index1 (observed/expected)
From
A		2 (15·4)2	9 (69·2)	2 (15·4)	13 (100)	24	0·54
C	2 (14·3)		0 (0)	12 (85·7)	14 (100)	16	0·87
G	41 (91·1)	3 (6·7)		1 (2·2)	45 (100)	26	1·73**
T	1 (4·8)	17 (81·0)	3 (14·2)		21 (100)	27	0·76*
Total	45	22	12	15	93	93

¹The mutability index is the observed number of mutations in a specific nucleotide divided by the expected number of mutations given a mechanism without bias. Expected numbers were derived by multiplying the frequency of the nucleotide over the region sequenced by the total number of mutations observed. A mutability index score of 1·0 would be assumed to represent non-biased sequence insensitive mutations. The number of V_L nucleotides over the 55 VJ-C sequences was 14 355 (A = 3685; C = 2473; G = 4015; T = 4180) with 93 V_L mutations.

²Numbers in parentheses are substitution percentages for the indicated nucleotides.

Chi-squared contrasts of observed and expected mutations with significant differences.

^*P= 0·005;

^**P= 0·01.

Dinucleotides

Dinucleotide mutability indexes (Table 4) were calculated to determine if mutations were concentrated to specific pairs of adjacent bases. Overall, dinucleotide MIs varied considerably from the observed/expected score of 1·0 which is assumed to represent random sequence insensitive mutation. Index scores greater than 1·0 are assumed to represent preferential targeting whereas scores less than 1·0 imply either an avoided target for mutation (a cold spot) or that mutations at these positions are selected against. Chi-squared analyses of mutations at dinucleotides showed statistically significant targeting at GC and TG combinations. GC dinucleotides have also been found to be significant targets of mutation in human¹⁹ and catfish²⁰ V_H regions. In the present study, 90% (9/10) of the GC dinucleotides in the zebrafish germline V_L were encoded within WRCH/DGYW motifs. Moreover, 98% (40/41) of the mutations at GC positions were in the target C positions of AID hotspot motifs. In contrast to the high mutability score for GC (MI = 5·84), the MI for TA was 0·09 (Table 4), which would suggest that TA dinucleotides are avoided targets for mutation (cold spots) or mutations at TA are selected against. The presence of both statistically significant mutational hotspots and cold spots is also indicative of non-random nucleotide targeting in the zebrafish model.

Table 4.

Dinucleotide mutability indexes in zebrafish V_L

Dinucleotide	VL germline1	No. of mutations	Expected	Mutability index2 (observed/expected)
AA	17	6	12	0·50
AC	14	6	10	0·61
AG	19	10	13	0·75
AT	17	7	12	0·59
CA	12	11	8	1·31
CC	9	2	6	0·32
CG	2	2	1	1·43
CT	23	17	16	1·05
GA	22	7	15	0·45
GC	10	41	7	5·84** H
GG	22	6	15	0·39
GT	19	9	13	0·67
TA	16	1	11	0·09** C
TC	13	8	9	0·88
TG	30	42	21	1·99** H
TT	17	9	12	0·76

H, favoured somatic hypermutation (SHM) target or hotspot; C, avoided target or cold spot.

¹Number of times the dinucleotide is present in the targeted germline V_L.

²Mutability indexes were calculated as described in the legend of Table 3.

^**Statistically significant by chi-squared test at P = 0·01.

Trinucleotides

When expanded to trinucleotides, MIs (Table 5) for GC targeting were highest and statistically significant in GCA and TGC combinations whereas reduced targeting at TA remained consistent in TAC and TAA trinucleotides. In addition, GTG was also found to be a statistically significant base combination for mutation and ATG was identified as a coldspot. The GCA and TGC trinucleotides are both contained within DGYW/WRCH motifs and one of the highly mutated DGYW hotspots was immediately preceded by a G which accounts for GTG being statistically significant. Thus, GCA, TGC, and GTG appear to be targets for somatic mutation because of their inclusion within the larger WRCH/DGYW hotspot motifs, whereas ATC and ATG appear to be mutational cold spots or trinucleotides in which mutations are selected against.

Table 5.

Trinucleotide mutability indexes in zebrafish V_L regions

Trinucleotide	VL germline	No. of mutations	Mutability index1 (observed/expected)
AAA	6	2	0·32
AAC	4	1	0·24
AAG	5	6	1·14
AAT	2	2	0·95
ACA	4	4	0·95
ACC	4	2	0·47
ACG	1	2	1·90
ACT	5	2	0·38
AGA	3	3	0·95
AGC	5	10	1·90* H
AGG	3	2	0·63
AGT	8	5	0·59
ATA	1	0	0
ATC	5	0	0* C
ATG	7	0	0** C
ATT	4	2	0·47
CAA	1	3	2·85* H
CAC	2	2	0·95
CAG	8	8	0·95
CAT	2	3	1·42
CCA	2	0	0
CCC	0	0	NA
CCG	0	0	NA
CCT	7	8	1·08
CGA	1	0	0
CGC	0	0	NA
CGG	0	0	NA
CGT	1	3	2·85* H
CTA	4	1	0·24
CTC	4	5	1·19
CTG	13	12	0·88
CTT	2	1	0·47
GAA	2	3	1·42
GAC	2	4	1·90
GAG	6	1	0·16
GAT	6	2	0·32
GCA	4	39	9·25** H
GCC	2	1	0·47
GCG	0	0	NA
GCT	4	9	2·13* H
GGA	11	5	0·43
GGC	0	0	NA
GGG	5	2	0·38
GGT	6	1	0·16
GTA	5	0	0·00
GTC	2	0	0·00
GTG	6	35	5·53** H
GTT	6	6	0·95
TAA	3	0	0·00
TAC	6	1	0·16
TAG	0	0	NA
TAT	7	3	0·41
TCA	3	3	0·95
TCC	3	1	0·32
TCG	1	0	0·00
TCT	7	6	0·81
TGA	7	3	0·41
TGC	5	34	6·45** H
TGG	14	11	0·75
TGT	4	0	0·00
TTA	6	3	0·47
TTC	3	1	0·32
TTG	3	2	0·63
TTT	5	6	1·14

H, favoured somatic hypermutation (SHM) target or hotspot; C, avoided target or cold spot.

¹Mutability indexes were calculated as described in the legend of Table 3.

Statistically significant by chi-squared test at

^*P = 0·05;

^**P = 0·01.

NA, not applicable.

The impact of selection on somatic mutations

Regional accumulation of V mutations is a defining feature of antibody genes. In mammals, this tendency is believed to be attributable in large part to antigenic selection and clonal expansion. Mutations in FRs typically appear to be less tolerated, whereas mutations in CDRs provide the basis for the amino acid changes favoured during antigenic selection and affinity maturation. To address the possibility of antigenic selection, the distribution of R and S mutations within productive VJ-C cDNA sequences were compared. If selection were widespread, it is predicted that S mutations would be more abundant in FRs and R mutations would be favoured in CDRs. Of the 80 V_L mutations in productive VJ-C sequences, 31 were in FRs (21R; 10S) and 49 were in CDRs (35R; 14S). When weighted as the number of R and S mutations per base sequenced, the percentages become 0·41% for FRs and 0·72% for CDRs. These percentages would appear to suggest that either FRs are less targeted or mutations within FRs are selected against. When ratios of R to S mutations in FRs and CDRs were compared using chi-squared analyses, however, no significant differences were found (P=0·45). Multinomial distribution analyses of R and S mutations in productive VJ-C clones also did not reveal significant P values for a scarcity of FR replacements or an excess of CDR replacements. These results could be interpreted to mean that selection may be somewhat restricted in the zebrafish used in this study. Recent work by others²⁰ has shown that, in addition to collective assessments of R and S mutations, analyses of productive rearrangements in the context of clonal lineages and subsequent radiations may be more relevant to understanding the potential impact of selection on somatic mutation.

Mutations at consensus positions: potential founders in lineage radiation

Alignment of productive VJ-C sequences with both each other and the concordant targeted germline V_L revealed that the 80 V_L mutations localized to 44 positions of the germline sequence. This finding can be attributed to the fact that certain mutations, while present in a unique overall sequence, were common to more than one VJ-C clone. For example, the V_L(259) G→A transition within the DGYW motif in CDR3 appeared in several different unique VJ-C clones (Table 2). In total, 10 germline positions were found to be consensus positions for 60% (48/80) of the mutations among productive VJ-C clones. Moreover, 83% (40/48) of these mutations were at AID hotspot motifs with 92% (37/40) in CDRs. Collectively, these data suggest either repeated nucleotide targeting or the occurrence of clonal expansions in zebrafish B cells, with the majority of the progenitor clones initially acquiring founder mutations at AID hotspot motifs in CDRs.

Clonal lineages and affinity maturation

Shared mutations in productive VJ-C rearrangements were used as consensus positions to construct potential lineages of clonal radiation. As illustrated in Fig. 5, several potential lineages could be discerned for the VJ-C cDNA clones. Some clonal sets were found to harbour two or more sequentially additive mutations (Fig. 6a), while others appeared to represent radiations from a common founder mutation in CDR3 (Fig. 6b). When mutations within each of the lineages were counted, R mutations (45/65) were more prevalent than S mutations (20/65). In addition, it was evident that, in at least two cases, one or more depicted paths could have given rise to a particular clone. For example, the single mutations harboured in clones EU821523 and EU797183 (Fig. 6a) are intermediates to successive additive mutations in the EU821496 and EU795319 clones, respectively. The EU821496 and EU795319 clones could have also descended from EU795306. In several clones, mutations were not restricted to CDRs, suggesting a tolerance of R mutations in some FRs. For several VJ-C clones, R mutations in FRs appear to be early in the mutational lineage. The presence of founder mutations carried in the clonal descendants implies that, although the mutation may change the affinity for antigen, it does not ablate the structural integrity of the B-cell receptor. Identification of lineages in the zebrafish VJ-C clones is suggestive of both sustained and incremental mutational events and selection, both characteristics of affinity maturation responses seen in higher vertebrates.

Figure 6.

Lineage relationships of VJ-C cDNA clones are consistent with the possibility of clonal expansion and affinity maturation in zebrafish B cells. Individual VJ-C cDNA clones are depicted as circles by the accession number. A potential progenitor clone is depicted at the apex of each diagram. (a) Clones harbouring two or more additive mutations. (b) Clones radiating from a founder mutation in complementarity-determining region 3 (CDR3). In both diagrams, the progenitor VJ-C clone (accession no. EU821516) is identical in sequence to unmutated germline immunoglobulin L (IgL). Mutations listed for each VJ-C clone designate the location in framework regions (FRs) or CDRs and numbers in parentheses indicate the concordant germline position. Replacement mutations are indicated with an R while silent mutations are indicated with an S. Listings in italics are mutations at the C/G positions of WRCH/DGYW hotspot motifs. The directionality of arrow segments depicts sequential mutation accumulation in the radiation of clonal descendants.

Discussion

Zebrafish are rapidly emerging as an important immunological model for biomedical research. In contrast to mice and human models, far less is known concerning potential nucleotide targets and mutational hotspots that may underpin immunoglobulin affinity maturation in this species.

Previously, by aligning cDNA and expressed sequence tag (EST) sequences with genomic reference sequences, our laboratory had shown that somatic mutation can occur in the immunological light chains in zebrafish.⁷ In the present study, by focusing on a single V_L locus, we were able to obtain a sufficient quantity of VJ-C cDNA transcripts to identify specific nucleotide targets and patterns of SHM in this animal model.

In mammals, SHM is typically characterized by the presence of single base substitutions, with few insertions or deletions. Single base substitutions and limited insertion or deletion were also defining characteristics in the zebrafish VJ-C cDNA clones (Table 2). Given that insertions or deletions in non-triplet increments result in frameshifts, it is expected that B cells harbouring such mutations would be selected against, for without functional light chains the integrity of the BCR would be lost. Also similar to findings in mice and humans,²¹ transition (T) mutations were more prevalent than transversions (V) (T:V ratio in zebrafish V_L = 5·64). The transition preference in zebrafish is even more pronounced if only mutations in wobble positions of degenerate codons are considered (T:V ratio = 10; data in Table 2). Mutations at degenerate wobble positions may escape selection at the protein level as, regardless of the base exchanged, an identical amino acid would be encoded. Thus, mutations at wobble positions might offer a window into targeting and base change preferences of a mutational mechanism active in zebrafish. However, selection based on nucleotides adjacent to wobble positions cannot be excluded. Of the 11 mutations in wobble positions of degenerate codons, base exchanges were from T (seven), A (three), or G (one) and none of these 11 exchanges was in WRCH/DGYW motifs. Similarly, if all non-WRCH/DGYW V_L mutations are included, A:T mutations represent 73% (34/47; Table 2) of mutations outside of AID hotspot motifs. Collectively, these findings indicate that in zebrafish a substantial number of mutations outside of WRCH/DGYW motifs in V_L may be attributable to mutational mechanisms that target A:T base pairs.

AID-dependent deamination of cytidine to uracil, in addition to producing mutations at C/G nucleotides, has also been shown to activate mismatch repair at U:G mismatches in mouse models. The mismatch repair proteins MSH2–MSH6 have been found to bind U:G mismatches and in doing so can recruit a low-fidelity DNA polymerase called polymerase eta (η).²² Upon binding the MSH2–MSH6 heterodimer, the catalytic activity of η is stimulated, allowing the polymerase to move more rapidly along the template DNA. Being a low-fidelity polymerase, η is prone to incorporate base substitutions preferentially at A:T positions downstream of the original U:G lesion.^23,24 Thus, in theory, the G:C and A:T mutations observed in zebrafish V_L could be largely dependent on the combined outcome of uracil-DNA glycosylase (UNG) and mismatch (MSH) repair pathways. Somatic mutations in mammalian immunoglobulin¹⁸ and the nurse shark antigen receptor, (NAR)²⁵ are also proportionally distributed among G:C and A:T base pairs. However, in other vertebrates, including frogs,²⁶ and the V_H of shark IgM,²⁷ G:C mutations are favoured. These findings indicate mutational targeting of G:C and A:T pairs and subsequent repair strategies therein may occur at different capacities in different organisms.

When individual zebrafish VJ-C sequences are considered, 18 of the 55 clones contained mutations within a WRCH/DGYW and one or more mutations either upstream or downstream of the targeted AID hotspot. In total, the 18 clones harboured 29 V_L mutations outside of a targeted WRCH/DGYW (data in Table 2). Of these, 72% (21/29) were at A:T positions and 28% (8/29) were at G:C base pairs. Most intriguing, however, is that 93% (27/29) of these mutations were downstream of the targeted WRCH. Based on these findings, it is tempting to speculate that, similar to mammalian models, AID might target C:G at WRCH/DGYW in zebrafish which in turn may activate orthologous MSH2–MSH6 proteins to resultant U:G mismatches. If recruitment of a low-fidelity polymerase similar to mouse polymerase η were also to ensue, this could in theory account for the increased propensity for A:T mutations downstream of targeted AID hotspots.

Most experimental evidence in mice and humans suggests that AID initiates deamination of cytidines in actively transcribed immunoglobulin genes.^28,29 Recently, it was shown that in vitro nucleosomes prevent AID access, unless the immunoglobulin segment is being transcribed.³⁰ Transcription is required for SHM in vivo, presumably in part to loosen the contact of nucleosomes with the DNA.³¹ During transcription, the single-stranded DNA (ssDNA) is prone to AID-mediated C to U conversion, producing U:G mismatches in the DNA. U:G mismatches cause modest distortions in the DNA which may in turn activate a suite of DNA repair mechanisms involving DNA glycosylases, general mismatch repair factors, and a variety of error-prone polymerases.³² Alternatively, if left unrepaired, the U:G mismatches become fixed as C→T mutations in replicated DNA as a result of the lack of discrimination by DNA polymerases between U and T in the template strand during DNA replication.

In the zebrafish V_L, the majority (41/46) of the cytidine mutations found at WRCH/DGYW hotspot motifs were C→T transitions. This finding suggests that uracil glycosylase-mediated DNA repair may be somewhat limited in the zebrafish V_L regions. For example, if uracil residues in U:G mismatches were substrates for base excision repair, it would seem likely that hydrolysis of the glycosidic bond between U and deoxyribose and subsequent endonucleolytic cleavage of the sugar would result in an abasic site. Polymerases involved in base excision DNA repair are generally more error-prone and their activity over an abasic lesion brought about by U removal from U:G mismatches would be predicted to result in C→A/G/T mutations. The precise preference for each type of substitution would in large part be reliant on the polymerase involved. Given that two of the three possible substitutions are transversions, it seems that a predominance of transversions would be apparent if base excision repair was extensive at U:G mismatches created at AID motifs.³³ The statistically significant preference (P<0·01) for transition over transversion mutations both within and outside AID motifs in the zebrafish V_L implies a tendency either for mutations at AID hotspots to escape DNA repair or for selective pressures to maintain transition mutations once they become fixed in the B-cell genome.

In mammals, it has been suggested that AID-induced mutagenesis saturates the overall repair capacity of B cells.³⁴ If AID mutagenesis were to saturate uracil glycosylase capacities in zebrafish B cells, this might in part explain why the majority of mutations at AID hotspots in zebrafish V_L were C→T. Conversely, if mismatch repair pathways were not as saturated, this could also explain the increased capacity for A:T mutations downstream of targeted at AID hotspots. Although the balance between saturation of repair mechanisms and toleration of mutation remains largely unknown, it appears that flexibility in this balance would result in an increased capacity to generate mutational diversity within immunoglobulin gene segments. Uracil glycosylase base excision and mismatch repair systems are evolutionarily ancient mechanisms for DNA repair thought to exist in all prokaryotes and eukaryotes. The utilization of these repair mechanisms in vertebrates to generate additional diversity within immunoglobulin gene segments is an area of research that has only recently begun to be explored.

The discovery just over a decade ago³⁵ that AID is responsible for both SHM and class switch recombination (CSR) dramatically enhanced the possibility for obtaining an in-depth understanding of the mechanistic processes underlying adaptive immunity in vertebrates. It had long been thought that the uracil in DNA was an adverse condition arising from inappropriate incorporation of dUTP during replication or spontaneous deamination of cytosine.³⁶ It is becoming increasingly apparent, however, that nature incorporates uracil into DNA as a central mediator of adaptive immunity and as a strategy against certain viruses during innate responses.^37,38 Thus, uracil incorporation, once thought to be solely a mutagenic burden, has been revealed as a mechanism to modify immunoglobulin DNA in B cells for diversity or even non-self DNA for degradation. An orthologue of AID has been identified in zebrafish and its expression in mammalian cells in vitro has been shown to induce both CSR and SHM.^39,40 In the present study, the patterns revealed for in vivo mutations in zebrafish V_L strongly suggest that AID and uracil incorporation are utilized as a means to diversify immunoglobulin diversity in the zebrafish model.

Despite drastically different outcomes for SHM and CSR (point mutations versus large-scale deletions) and functionally distinct target sequences (V_H/L exons versus switch regions), SHM and CSR are both contingent upon the B-cell specific AID enzyme and single-strand templates brought about by transcription. Point mutations similar to those at V_H WRCH sequences have also been found at the WRCH within switch regions in mice, suggesting a common AID targeting method for both SHM and CSR.^41,42 Given that SHM has been found in all vertebrates including fish, whereas CSR appears limited to amphibians, birds and mammals^43–47, it appears that SHM evolved earlier than CSR. In mammals, immunoglobulin switch regions appear to have further diverged to incorporate additional features, such as R-loop forming ability and cis-acting regulatory regions, which may serve to increase the overall efficiency of CSR.⁴⁸ It is also plausible that gene conversion (GCV) could be implicated in the non-WRCH/DGYW mutations found in the zebrafish IgL. Given that diversification of immunoglobulin by gene conversion has only been shown to exist in chickens^49,50 and rabbits,⁵¹ a comparative and evolutionary approach may prove necessary to elucidate additional currently unresolved aspects of SHM, CSR and GCV in vertebrates.

The present study demonstrates a strong dependence of SHM on cytidine targeting in zebrafish and reveals several important parameters, including strand biases and the direction in which mutations spread from a C to downstream A:T base pairs. In addition, the results reveal asymmetry in the AID hotspot motifs targeted with preferential targeting at palindromic WRCH sequences. These results combined with findings of distinct mutational hotspots, cold spots, and additive mutations are indications that SHM is not a random process in zebrafish. The presence of clonal lineages is also a strong indicator that AID deamination, mutation repair and affinity maturation may be crucial in shaping the somatic diversification of IgL in zebrafish. Based on these conserved and unique features it is probable that the zebrafish will prove to be an especially useful emerging new vertebrate model for understanding the role of immunoglobulin diversity in immune system development, function and disease.

Disclosures

The authors have no conflicts of interest to disclose.