ABSTRACT
Hypermutation of immunoglobulin genes is a key process in antibody diversification. Little is known about the mechanism, but the availability of rapid facile assays for monitoring immunoglobulin hypermutation would greatly aid the development of culture systems for hypermutating B cells as well as the screening for individuals deficient in the process. Here we describe two such assays. The first exploits the non-randomness of hypermutation. The existence of a mutational hotspot in the Ser31 codon of a transgenic immunoglobulin V gene allowed us to use PCR to detect transgene hypermutation and identify cell populations in which this mutation had occurred. For animals that do not carry immunoglobulin transgenes, we exploited the fact that hypermutation extends into the region flanking the 3'-side of the rearranged J segments. We show that PCR amplification of the 3'-flank of VDJH rearrangements that involve members of the abundantly-used VHJ558 family provides a large database of mutations where the germline counterpart is unequivocally known. This assay was particularly useful for analysing endogenous immunoglobulin gene hypermutation in several mouse strains. As a rapid assay for monitoring mutation in the JH flanking region, we show that one can exploit the fact that, following denaturation/renaturation, the PCR amplified JH flanking region DNA from germinal centre B cells yields mismatched heteroduplexes which can be quantified in a filter binding assay using the bacterial mismatch repair protein MutS [Wagner et al. (1995) Nucleic Acids Res. 23, 3944-3948]. Such assays enabled us, by example, to show that antibody hypermutation proceeds in the absence of the p53 tumour suppressor gene product.
Localised hypermutation of immunoglobulin V genes plays a key role in antibody diversification, contributing to the affinity maturation of antibodies in mouse and man as well as to the generation of the primary antibody repertoire in sheep. The hypermutation is targeted to a 1-2 kb region of DNA that includes the V gene exon with nucleotide substitutions being introduced at a rate of ~10-3 mutations per base pair per cell cycle. The process is restricted to a narrow window of B cell development and takes place in germinal centres. Its mechanism is unknown, although transgenic studies have identified some of the cis-acting sequences necessary for its recruitment to the V gene (reviewed in 1 -4 ).
Hypermutation in antigen-specific responses has classically been analysed by cloning the expressed V genes from sorted (or immortalised) antigen-specific B cells and comparing the sequences of these V genes to those of their presumed germline counterparts. More recently, larger databases of mutations have been obtained by PCR cloning the expressed immunoglobulin V genes (or transgenes) from total sorted germinal centre B cells. Whilst such studies have provided valuable information on the extent and distribution of immunoglobulin hypermutation, the process of cloning and sequencing large numbers of V genes is laborious when simply used to quantify the extent of V gene hypermutation. Sensitive methods of quantification that allow a rapid throughput of a large number of samples could greatly facilitate the development of culture systems for B cells performing hypermutation. Such methods could also prove of use in screening for individuals deficient in antibody hypermutation.
Here, we describe methods that can be used in transgenic and non-transgenic mice to rapidly quantify immunoglobulin hypermutation and to enrich for hypermutated sequences.
Transgenic mice were maintained by breeding with (C57BL/6*CBA) F1s. The L[kappa]6 mice carry six copies of the L[kappa] transgene (Fig. 1 ), which comprises a functional mouse V[kappa]Ox1-J[kappa]5 rearrangement linked to rat C[kappa], and the [kappa] intronic and 3' enhancers which are both required for efficient hypermutation(5 ,6 ). This light chain is characteristic of antibodies produced by BALB/c mice in the primary response to the hapten 2-phenyl-5-oxazolone (7 ). The L[kappa][Delta]I, L[kappa][Delta]E and L[kappa]Tr mice carry two to five copies of transgenes derived from L[kappa] by deleting sequences around the intronic enhancer, in the intronic enhancer, or by transposing the intronic enhancer, respectively. All of these alterations cause the V gene to be targeted less efficiently for mutation than it is in L[kappa] transgenes and will be described in detail elsewhere (Goyenechea,B., Klix,N., Yelamos,J., Williams,G., Riddell,A., Neuberger,M.S. and Milstein,C., submitted; Klix,N., Jolly,C., Brüggemann,M., Williams,G. and Neuberger,M.S., in preparation). The p53- and CD22-deficient mice have been described elsewhere (8 ,9 ) and were kindly provided by Isabelle Lavenir and Theresa O'Keefe (MRC Laboratory of Molecular Biology, Cambridge), respectively.
Cell suspensions from mouse thymus, spleen and Peyer's patches were made by forcing through a cell sieve in culture medium. Germinal centre B cells (B220+PNAhi) and non-germinal centre B cells (B220+PNAlo) were purified by flow cytometry from Peyer's patch cell suspensions (10 ). DNA was extracted from cells that had been lysed by freeze-thawing in water (~10 [mu]l/105 cells) and incubated with proteinase K (100 [mu]g/ml; Boehringer Mannheim) in Pfu DNA polymerase PCR buffer (Stratagene) at 56oC for 1 h. The proteinase K was then inactivated by heating at 95oC for 30 min.
Transgenic V sequences were amplified using `touchdown' PCR (11 ), cloned and sequenced as previously described (10 ), except that Pfu DNA polymerase (Stratagene) was used. Clones were confirmed to contain transgenic (not endogenous) V[kappa]Ox1 sequences by oligonucleotide hybridisation (6 ). For monitoring mutation at the endogenous IgH locus, a 1.2 kb segment from the 3'-flank of rearranged endogenous VH genes was amplified using a primer complementary to a sequence common to FR3 of most VHJ558 family members (GGA ATT CGC CTG ACA TCT GAG GAC TCT GC) together with a primer complementary to a sequence located at the 3'-end of the IgH intronic enhancer (GAC TAG TCC TCT CCA GTT TCG GCT GAA TCC). M13 plaques were positively selected for JH4 rearrangements by hybridisation with oligonucleotide JH4 (TAT GCT ATG GAC TAC TGG) and negatively selected by hybridisation with oligonucleotide JH3 (CCT GGT TTG CTT ACT GG). Sequencing reactions were primed with oligonucleotide JH4 to facilitate alignment and mutation identification.
Reaction mixtures (45 [mu]l) contained Taq PCR buffer (Promega), 0.2 mM dNTPs, 0.2 [mu]M each of LkFOR (11 ) and either ASN31 (GCG GCA TGC CAG CTC AAG TGT AAA) or SER31 (GCG GCA TGC CAG CTC AAG TGT AAG) oligonucleotides and genomic DNA corresponding to ~104 cells. The last three 3' bases in the ASN31 and SER31 oligonucleotides were joined by phosphorothioate linkages to prevent degradation by 3'-exonucleases (12 ,13 ). Since the modification step that produces the phosphorothioate linkages is very efficient (12 and data not shown), the primers were used without purification. After overlaying the reaction mixtures with oil and heating to 90oC, Taq DNA polymerase (1.25 U; Promega) in 5 [mu]l 1* Taq PCR buffer was added (`hot start') and 45 (unless otherwise stated) thermal cycles performed [94oC for 30 s, 67oC (unless otherwise stated) for 30 s, 72oC for 40 s]. Reactions to control for transgene abundance used the LkFOR/VkOXBACK primers (10 ). Reaction products were detected, following electrophoresis, by Southern blotting with a V[kappa]Ox1 probe.
We found that Taq DNA polymerase works better than Pfu and Vent (New England Biolabs) DNA polymerases in discriminating between wild-type and mutant templates (data not shown). The reason for this is not clear but it was not necessarily due to 3'-exonuclease activity in the Pfu and Vent DNA polymerases because the primers used were exonuclease resistant (see above) and because engineered exo- Vent DNA polymerase (14 ) produced results no different from wild-type Vent (data not shown).
In some cases, a prior amplification of transgenic sequences was performed using LkFOR and VkOxBACK primers for 20 cycles of PCR with Pfu DNA polymerase (94oC for 1 min, 60oC for 1 min, 72oC for 2 min). The products were purified using a Qiaquick PCR purification kit (Qiagen) into a volume of 100 [mu]l TE buffer (15 ). Ten [mu]l of the primary PCR products were then used as template for the secondary ASN31 or SER31 PCR reactions (25 cycles, unless otherwise indicated) which were essentially as in the previous paragraph.
The biotinylated DNA substrates for the MutS binding assays were generated by PCR. For analysis of transgenes, a 155 bp segment of V[kappa]Ox1/J[kappa]5 was amplified using `touchdown' PCR (11 ), Pfu DNA polymerase and a maximum of 30 PCR cycles, with primers JOL24 (biotin-CTT CTG CAG TCC CTG CTC GCT TCA) and JOL25 (biotin-GCT CAA GCT TGG TCC CAG CAC CGA).
For the analysis of endogenous genes, sequences in the JH1-JH2 spacer region were amplified under similar conditions using primers JOL34 (biotin-GGA ATT CTA CTG GTA CTT CGA TGT CTG G) and JOL35 (biotin-CAG GCA GCT AGC CTC TGA CTG). However with p53-/- mice, polymorphisms in this region interfered with the MutS assay (data not shown); therefore sequences 3' to JH4, in a region where we have not detected polymorphism, were amplified with primers JOL40 (biotin-GTG AAT TCT ATG CTA TGG ACT ACT GG) and JOL29 (AGA TGT GGA GAT AAT CTG TCC).
After purification using a Qiaquick kit (Qiagen), the PCR products in 10 mM Tris-HCl, pH 7.4, 0.1M NaCl were denatured/renatured by heating to 100oC for 3 min, then 75oC for 90 min before chilling to 4oC. For MutS assays (16 ), a sheet of nitrocellulose (0.45 [mu]m pore size, Schleicher and Schuell) and two sheets of underlying Whatman 3MM paper were pre-soaked in MutS buffer (16 ) for 5 min and assembled in a slot blotter (Stratagene). After removing excess liquid by suction, slots were incubated with MutS protein (20 [mu]l at 25 ng/[mu]l; Genecheck Inc., Fort Collins, CO) for 20 min, blocked with 200 [mu]l 3% BSA/MutS buffer for 30 min and the solution pulled through the slots by suction. Aliquots (20 [mu]l) of the heteroduplexed PCR products in 3% BSA/MutS buffer were then added to the slots and incubated with the immobilised MutS for 20 min. After five washes with 150 [mu]l MutS buffer each, horseradish peroxidase-conjugated streptavidin [100 [mu]l (Pierce) diluted 1:4000 in 3% BSA/MutS buffer] was added to the washed slots and left for 20 min. After washing as before, the nitrocellulose was removed from the slot apparatus, washed five more times with 50 ml MutS buffer and the MutS-bound DNA detected by enhanced chemiluminescence (ECL, Amersham).
For control blots, aliquots (10 [mu]l) of the heteroduplexed PCR products diluted in MutS buffer were added to 400 [mu]l 20* SSC (15 ), blotted directly onto nitrocellulose and fixed by cross-linking with 120 mJ of UV light in a Stratalinker (Stratagene). After blocking with 5% (w/v) skim milk powder in PBS for 1 h, and washing 4 * 5 min with PBS, the bound biotin-DNA was detected using horseradish peroxidase conjugated streptavidin and ECL.
Self-heteroduplexed DNA amplified from germinal centre B cells by 28 PCR cycles with JOL24 and JOL25 primers was used in an immobilised MutS assay. The nitrocellulose with the MutS-bound DNA was excised, immersed in 10 [mu]l 1* EcoRI digestion buffer (New England Biolabs) containing 100 [mu]g/ml proteinase K and overlaid with mineral oil (Sigma). After incubation at 55oC for 10 min, proteinase K was heat inactivated at 95oC and the DNA renatured at 75oC for 90 min before chilling to 4oC. Following centrifugation at 16 000 g for 10 min, eluted DNA (2 [mu]l aliquots of the supernatant) was further amplified using seven PCR cycles with JOL24/JOL25. The Qiaquick purified product was cloned into M13mp19 for sequencing. Sequencing of the same region of DNA directly amplified by 35 PCR cycles was used as a control.
The L[kappa] transgene efficiently hypermutates in germinal centre B cells with the second base of the serine-31 codon of the transgenic V region being a mutational hotspot (5 ,10 ,17 ). Thus, in ~10% of all mutated L[kappa] transgenes, a transition mutation has altered the Ser31 codon from AGT to AAT (Asn) (18 ). Since Taq DNA polymerase is sensitive to 3'-terminal A:G mismatches in primers (19 ) we reasoned that a PCR primer specific for the Asn31 mutation might be used to give a PCR product only when the transgene was mutated at this hotspot. Sense primers ending with either a Ser31 G (SER31) or Asn31 A (ASN31) were synthesized. The last three 3'-bases in these oligonucleotides were joined by phosphorothioate bonds to prevent their degradation by 3'-exonucleases (12 ,13 ) which could lead to priming from the wrong templates. These SER31 or ASN31 oligonucleotides were then paired with downstream antisense oligonucleotides and used to amplify the V[kappa]Ox1 gene from plasmid templates that included either the wild-type V[kappa]Ox1 or a mutated derivative that contained an Asn codon at position 31 (Fig. 1 A). At appropriately stringent annealing temperatures, the ASN31 oligonucleotide would only prime PCR amplification from the mutant and not the wild-type template; a complementary discrimination was achieved with the SER31 oligonucleotide (Fig. 1 B). A homology length of 18 bases was optimal for both SER31 and ASN31 primers. Shorter primers gave a much narrower range of annealing temperatures in which wild-type and mutant templates could be discriminated, whereas longer primers could not discriminate at annealing temperatures below 72oC (data not shown).
To test whether hotspot-specific PCR could correctly detect transgene mutation in genomic mouse DNA (as opposed to cloned plasmid samples), we prepared DNA from thymus, spleen and Peyer's patch of L[kappa] transgenic mice. For each DNA sample, separate PCR reactions were set up to amplify both mutated (ASN31 and L[kappa]FOR primers) and total (V[kappa]OxBACK and L[kappa]FOR primers) L[kappa] transgenes. Equal aliquots of the two reactions were then pooled, run together in agarose gels and the products detected by blotting (Fig. 2 A). The DNA samples from Peyer's patches (which contain many highly mutated germinal centre B cells, 10 ) gave a strong band in the ASN31 PCR reactions. Spleen, which contains relatively few mutated (memory) B cells gave weak and rather variable ASN31 signals and thymus, which is not expected to contain mutated B cells, gave barely detectable ASN31 products (Fig. 2 A). Germinal centre B cells stain brightly with peanut agglutinin and therefore, as predicted, ASN1 PCR bands were obtained from PNAhi (but not the PNAlo) B cell sub populations from Peyer's patches.
These results indicate that PCR with the ASN31 oligonucleotide should selectively amplify mutated L[kappa] transgenes. To confirm this, ASN31 and SER31 products were amplified in parallel from total Peyer's patch DNA of L[kappa]6 mice, cloned into M13 and sequenced. In an unfractionated Peyer's patch cell sample, only 10-15% of the total cell population (i.e. the PNAhiB220+ fraction) should carry mutated immunoglobulin genes (see ref. 10 ). Thus, the SER31 PCR clones carried few mutations, but the ASN31 clones were heavily mutated (Fig. 2 C). Even when an annealing temperature of 55oC was used with the ASN31 primer (which can cause some mispriming from Ser31 templates-see Fig. 1 B), the sequences amplified from L[kappa]6 total Peyer's patch DNA were mostly mutated (data not shown). Thus, the ASN31 primer was selective for mutated sequences.
The hotspot-specific PCR worked for the detection of transgene mutation but, for more general use, we wished to devise an assay that did not depend on the existence of a dominant mutational hotspot. To this end, we exploited the ability of the bacterial MutS protein to recognise DNA mismatches (16 ). A 155 bp sequence of the transgenic V[kappa]Ox gene (extending across CDR3 and J[kappa]5) was amplified from tissues using Pfu DNA polymerase and 5'-biotinylated primers (JOL24 and JOL25, Fig. 1 A). The purified PCR product was denatured and renatured such that mutations in the original DNA template would lead to the creation of mismatched heteroduplexes. The amount of such duplexes was then measured by binding them to filter-immobilised MutS using the assay of Wagner et al. (16 ). The bound DNA was quantified using streptavidin-conjugated horseradish peroxidase and enhanced chemiluminescence (ECL). ECL detection of aliquots of the heteroduplex preparation that had been directly bound to filters allowed variation in DNA concentration to be controlled (Fig. 3 A).
The DNA bound to the filter-immobilised MutS should be enriched in mismatched duplexes and therefore in mutated V gene sequences. To confirm this, transgene DNA that had been PCR amplified from germinal centre B cells of the weakly mutating L[kappa]Tr mouse was eluted from nitrocellulose-immobilised MutS and subjected to an extra seven cycles of PCR to convert it to homoduplexes before cloning into M13. Sequencing of these clones revealed that they were indeed much more mutated than comparable L[kappa]Tr sequences cloned without MutS enrichment (Fig. 3 C).
We felt it should be possible to extrapolate the MutS assay to endogenous immunoglobulin sequences. However, whereas the V gene itself is the natural target of hypermutation, different B cells will express different rearranged V genes. This heterogeneity would greatly complicate the analysis. We thought we might be able to exploit the fact that the hypermutation domain extends into the J-C intron which flanks the 3'-border of all rearranged V segments (10 ,20 ,21 ) To this end, rearranged VH sequences were PCR amplified from sorted germinal centre B cells using a primer homologous to a conserved sequence in framework three of the VHJ558 family and an antisense primer complementary to the IgH intron enhancer (Fig. 4 A). Clones that contained V to JH4 rearrangements were selected by oligonucleotide hybridisation and sequence data from immediately 3' of JH4 obtained. This region was generally heavily mutated in germinal centre B cells isolated from several mouse lines (Fig. 4 B).
CD22-deficiency is known to affect antigen receptor signalling and p53-deficiency affects DNA repair pathways, but neither has much effect upon the number of PNAhi B cells in Peyer's patches (8 ,9 ,22 and unpublished data). Thus it was possible to use the above sequencing strategy to determine the effect of these gene deficiencies upon hypermutation. The results showed that immunoglobulin genes in CD22- and p53-deficient mice undergo hypermutation (Fig. 5 A), and the pattern of mutations is apparently unaffected (not shown).
We have here described two rapid assays for monitoring immunoglobulin hypermutation that do not entail DNA cloning and sequencing.One assay exploits the presence of a well characterised mutational hotspot in the V[kappa]Ox-1 gene by using PCR with the most 3'-base of one of the primers being complementary to the mutation hotspot. This assay rests on the assumption that DNA polymerases will only poorly extend primers with a 3'-mismatch. The other protocol involves the denaturation/renaturation of PCR-amplified immunoglobulin V gene sequences and exploits the ability of immobilised MutS protein to specifically bind mismatched DNA heteroduplexes (16 ).
How does the performance of the two assays compare? The hotspot PCR protocol has the advantage of being very easy to carry out, but obviously requires the presence of a major mutational hotspot in the DNA segment amplified. Furthermore, the PCR conditions are critical. `Hot-start' and an annealing temperature above a critical threshold must be used. The hot start can be greatly simplified by the use of Amplitaq GoldTM (not shown), which is inactive until heated at 95oC for ~15 min, making the procedure suitable for a 96-well assay format from DNA extraction up to gel loading.
The immobilised MutS assay, whilst slightly more laborious than the hotspot assay, does not depend upon the presence of a mutational hotspot in the amplified DNA segment and, requiring fewer PCR cycles, is less prone to problems of PCR artefact or contamination. Care must, however, be taken to ensure that germline variation in the target sequence (allelic polymorphism) does not produce false positive results. Amplification of DNA from, for instance, thymus or PNAlo B cells is an essential control for germline or PCR-generated diversity.
When the MutS assay indicates mutation in a PCR product, those mutations can be characterised by eluting the DNA and cloning it (Fig. 3 C). This enrichment can be useful for analysing the pattern of mutation in poorly mutating sequences, but caution needs to be exercised in this analysis. We found that C to T transitions can be over represented by this enrichment (not shown), presumably because MutS has highest affinity for T:G mismatches (see ref. 16 ).
In order to analyse mutation of endogenous (as opposed to transgenic) immunoglobulin sequences, we used the region flanking the 3'-border of rearranged JH segments, reasoning that this would allow mutation to be assayed without focusing on an individual V gene segment. Sequences immediately 3' to rearranged J elements mutate as frequently as the rearranged V elements themselves (20 ,21 ,23 ). JH4 segments rearranged to the large VHJ558 family were readily amplified from germinal centre B cells, and thus provided a good target sequence for measuring and analysing hypermutation (Fig. 4 ).
Sequence analysis of mutated JH4 flanks has not identified a strong hypermutation hotspot yet (Fig. 4 B). This suggested that for rapid analyses, JH flanks might constitute more suitable substrates for a MutS-based, rather than the hotspot protocol. Indeed, fragments amplified from JH1 or JH4 3'-flanks without any selection for rearranged alleles were useful substrates for rapid MutS-based mutation screening (Fig. 5 B). JH-flanks alone were amplified, because denaturing/renaturing PCR products that included the V(D)J sequences themselveswould have produced spurious mismatches as a result of the imprecision of V(D)J recombination. If selectivity in the MutS-based assay for rearranged alleles (and thus increased assay sensitivity) were required it could easily be achieved by a primary amplification based on rearranged V genes (e.g., as in Fig. 4 A), followed by a nested amplification of the JH-flank alone.
To demonstrate the usefulness of mutation assays based on JH-flanks, we showed that CD22- and p53-deficient mice hypermutate their immunoglobulin genes (Fig. 5 ). Whereas CD22 affects signalling in B cells (9 ,22 ,24 ), we show here that its absence does not affect hypermutation. With regard to p53, it has long been thought that hypermutation may be initiated by DNA strand breaks (25 ), and it is known that p53 is required for the normal response of cells to such breaks (reviewed in ref. 26 ). Furthermore, a DNA repair abnormality specific for rearranged V genes in germinal centre B cells has been reported (27 ). However, our data suggests that if DNA strand breaks are required for hypermutation, they can be resolved without a p53-induced G1 checkpoint.
A major goal of hypermutation research is the identification of the proteins involved. This goal may be advanced by the development of culture systems in which hypermutation can be induced, or by the identification of individuals or gene-targeted mice in which germinal centres are formed, but where hypermutation is inactivated. The MutS-based assay described here is sufficiently sensitive for transgenes that mutate at very low rates. This sensitivity may be enough to detect mutation in in vitro culture systems and the assay is simple enough to enable screening of many humans or gene-targeted mice. Since the number of mouse gene-knockouts with obvious potential to directly affect immunoglobulin hypermutation is currently small, they are probably better analysed by the sequencing of rearranged JH4-flanks than by the rapid, but less informative MutS-based assay.
We thank Gareth Williams, Sarah Davies, John Jarvis and Theresa Langford for the generation and maintenance of transgenic mouse lines, Andrew Riddell for cell sorting, Richard Grenfell and Jan Fogg for synthesis of oligonucleotides and César Milstein for many useful discussions. C.J. was supported by grants to M.N. from the Leukaemia Research Fund and the Howard Hughes Medical Institute and N.K. is supported by a Research Training Grant from the European Commission.
REFERENCES
