Somatic mutation of bcl-6 genes can occur in the absence of V_H mutations in chronic lymphocytic leukemia

V(D)J recombination in precursor B cells is the first step in a multistage process to generate antibody diversity.1 This process generates a unique signature CDR3 sequence that identifies derivative B-cell clones.1,2 After antigen encounter, the V(D)J transcriptional unit is targeted by a potent somatic hypermutation mechanism to generate high-affinity antibody, which occurs by the selection of mutated progeny.3 This process is thought to be largely restricted to the germinal center.1-3 Consequently, analysis of V gene somatic mutation in neoplastic B cells has enabled a characterization of the clonal history of the cell of origin. B-cell tumors have recently been classified in 3 categories: Those with unmutated V genes, in which the cell of origin does not enter the germinal center; those tumors with ongoing V gene mutations, such as follicle center lymphoma (FCL), which arise and remain in the germinal center environment; and tumors with mutated stable V genes, such as multiple myeloma, which have traversed the germinal center and exit, never to return.4 

These distinctions have also revealed a significant and clinically relevant division of chronic lymphocytic leukemia (CLL) into 2 subsets, one of which has V-gene sequences in germline configuration, and the other with somatically mutated V genes, a finding independently reported in several studies.5-8 The initial concept that all CLLs arose from a naive B cell clearly had to be revised. In our early study of 22 cases of CLL with the common phenotype of CD5⁺, CD23⁺, low-surface immunoglobulin (Ig), we found that unmutated V_H genes were associated with atypical morphology and trisomy 12, both features of a less favorable prognosis.8 Expansion of these analyses has confirmed that unmutated V genes, found in approximately 50% of CLLs, are highly associated with a poorer clinical outcome, even at stage A.5 These findings indicate that CLL is derived from B cells arrested at 2 distinct stages of differentiation, with the more mature cell that has encountered the somatic mutation mechanism, being less malignant. In contrast to FCL, however, the mutational pattern in CLL appears stable, with no intraclonal heterogeneity.4-9 

It is now known that somatic mutation also targets nonimmunoglobulin loci, to date characterized in the 5′-intronic region of thebcl-6 gene.10,BCL-6 is a proto-oncogene, encoding a transcriptionally active protein with a POZ/Zinc finger motif,11 and maps to chromosome 3q27.12 Its expression regulates germinal center formation and B-cell activity at this site, as well as Th2 T-cell responses.13,14 A high frequency of chromosomal translocations in non-Hodgkin lymphoma, particularly diffuse large-cell lymphoma (DLCL), cluster at band 3q27, specifically at the bcl-6promoter and first noncoding exon regions, and may lead to deregulated bcl-6 function with a potential role in lymphomagenesis.14-16 Mutations in bcl-6 at this locus also occur in the absence of translocation events, with a target region approximately 2 kilobase (kb) 3′ from the transcription initiation site.10 These mutations may be biallelic and are characterized by single point mutations, insertions and deletional events.10,17,18 Single base-point mutations predominate, favoring transitions over transversions,10,17-19 with evidence for strand specificity,18 and although the frequency of mutations in bcl-6 is notably higher than basal rates, in general, it appears to be lower than the frequency of V gene mutations.17 18 

Somatic mutation targets both the V_H and bcl-6 loci in normal B cells, and is found only in memory and not naive B cells.17-19 In a study at the single cell level, bcl-6mutation tended to occur in normal B cells containing mutations in their V_H genes.17 Although 80% to 100% of these cells displayed mutations in V_H, only approximately 30% of the cells had corresponding mutations in bcl-6, implicating differences in targeting of the 2 loci. Bcl-6mutations also occur in a range of B-cell tumors, and a correlation between mutational status in V_H and bcl-6 in B cell has been suggested. However, there were many exceptions, particularly among cases of Burkitt's lymphoma and multiple myeloma.17Overall, more than half the numbers of germinal centers (GCs) or post-GC cell-derived B-cell tumors studied, which would certainly be expected to harbor V_H mutations, did not show mutations in bcl-6.17 Furthermore, GC-resident lymphoma cells characteristically show high levels of intraclonal variation in tumor V_H gene sequences, which to date has not been described as a feature of bcl-6 mutations in these tumor cells. Interestingly, this intraclonal variation inbcl-6 sequence has been reported in normal tonsil GC B cells.19 The question that is raised from these findings is whether the same mutational pathway is invariably involved.

In this study, we have investigated whether somatic mutation of thebcl-6 gene occurs conjointly with V_H mutations in the 2 CLL disease subsets. Previously, Pasqualucci et al17examined bcl-6 mutations in 33 cases of CLL by SSCP analysis, and a parallel presence of V_H and bcl-6 mutations was observed in 5 cases by sequence analysis. Our results from this study suggest distinct routes for V_H and bcl-6mutations in CLL, with ongoing somatic mutations in bcl-6,even in cases with no V_H mutations.

Patients with CLL who were analyzed in this report were from 2 cohorts extensively characterized and reported previously.5Two additional patients (B.P.S. and A.G.) were included in this study, and were assessed clinically and in the laboratory as reported.5 Relevant features and lymphocyte counts are shown in Tables 1 and2.

Standard cytogenetic preparations made from whole peripheral blood were analyzed as previously reported.5 

Whole peripheral blood taken at the time of diagnosis was either used for isolating DNA immediately or stored at −20°C before use. DNA was isolated from 300 μL blood using the Wizard Genomic DNA Purification Kit (Promega, Madison, WI) according to the manufacturer's instructions. Isolated genomic DNA was resuspended in 100 μL of 10 mmol/L Tris HCl, 1 mmol/L EDTA.

These methods were reported previously.5 

For β-globin (βG), polymerase chain reaction (PCR) primers were designed to amplify a 724-base pair (bp) fragment of intronic DNA, spanning exons 2 and 3 (GenBank accession no. V00499). The forward βG primer was 5′-AGG AAG GGG AGA AGT AAC and the reverse primer was 5′-AAT CCA GCC TTA TCC CAA, each primer used at a final concentration of 400 mmol/L. DNA (5 μL) isolated from patients A.G. and D.B. was amplified using the Advantage-HF2 PCR Kit (Clontech, Hants, UK) in a 50 μL volume and the following PCR conditions: 95°C for 5′ (1 cycle), followed by 95°C for 1′, 54°C for 1′, 72°C for 3′ (30 cycles), and 72°C for 10′ (1 cycle). Contamination was checked in control reactions with no added template, and stringent working conditions for PCR analysis were used as recommended.20 

For amplification of bcl-6 intronic DNA, primers E1.21C and E1.26 described by Migliazza et al10 were used to generate a 732-bp product reported by these authors as a hot-spot region. The same template volume, enzyme system, and PCR conditions used for amplification of βG were used.

Gel-purified PCR products of a predicted size were cloned into the pGEM-TA vector as described and used to transform JM109 competent cells.21 Randomly selected clones found to contain an insert of an appropriate size by restriction analysis of plasmid DNA were sequenced using M13 universal primers, primers used for PCR amplification, or internal primers (bcl-6 sense 5′-AGC AGA GAG GAC GAG ACA GTG CTT; antisense 5′-GAA AAA ACA CAG CCG CAC GAA TCC AGA). Mutations were confirmed by at least 2 separate primers by bidirectional sequencing. Both for V_H and bcl-6,a minimum of 4 clones were analyzed per patient.

Sequence analysis was carried out using the PCR dye-based dideoxy chain termination method, according to the manufacturer's instructions, and analyzed on an AB Prism 377 sequencer (Perkin-Elmer Applied Biosystems, Warrington, UK). Nucleotide sequences were aligned with GenBank database germline sequences (accession no. Z79581 for bcl-6intron) using MacVector 4.5.3 software (Scientific Imaging Systems, New Haven, CT).

In 18 of 19 patients studied in this report, data were available from karyotype analysis (Tables 1 and 2 and Hamblin et al5), and did not reveal any chromosome 3 abnormality. In our initial cohort of 84 patients with CLL,5 only 2 patients exhibited a karyotype with a chromosome 3 abnormality, del(3)(p21) and t(3;11)(q25:q25), respectively, suggesting a low incidence in the CLL disease.

Data showing a clear demarcation of the 2 CLL subsets in a large cohort of 84 patients have been reported previously,5 and are reproduced in part in Tables 1 and 2 for reference. V_Hgene use by 2 additional patients included in this study (B.P.S. and A.G.) was in germline configuration, and has been deposited in the GenBank/EMBL database (accession no. AJ272398/9). Of note, 4 cases (C.W., D.F., J.B., and N.O.) each showed identical V_H sequences in 4 tumor-derived clones.

In all patients analyzed, a 7-base difference (at positions 607, 769-771, 889, and 987) was observed from the reported germline database sequence (GenBank accession no. Z79581). The first nucleotide ofbcl-6 complementary DNA (cDNA) from this reported sequence was arbitrarily chosen as position +1,10 and the amplifiedbcl-6 5′ intronic region spanned +415 to +1140.

In the characterization of bcl-6 mutations in normal and malignant B cells, Pasqualucci et al17 used a 394-bp intron fragment of βG to assess Taq misincorporation rate, which was consistently found to be unmutated, indicating that this gene is not targeted for somatic mutation, and is suitable as an internal control to assess bcl-6 mutations. To match the target length ofbcl-6 under investigation in this study, a 724-bp fragment of intronic DNA spanning exons 2 and 3 of the βG gene was amplified. Subcloning of βG PCR DNA and subsequent sequence analysis enabled a quantitation of Taq error rate. For 2 patients (A.G. and D.B.), 15 and 12 clones respectively from the βG PCR products were completely sequenced, and 5 mutations were detected in 19 548-bp sequenced, giving a background error rate of 2.6 × 10⁻⁴ bp⁻¹ (less than 0.03%). The same PCR amplification conditions were used forbcl-6 analysis.

Solely to verify that particular sequence motifs in bcl-6 do not influence the Taq error rate, a bcl-6 clone identified as being unmutated was PCR amplified using the bcl-6primers and cloned. No mutations were detected in 8 clones sequenced, giving an error rate of less than 1 in 5792 bases (less than 0.02%).

Two nucleotide changes likely to represent polymorphic variants were observed, a G-C change at position +753 in 16 clones from 5 cases (also described by Migliazza et al10) and a single base deletion (ΔT) at position +875, which was observed in 15 clones from 4 cases (Tables 3, 4, and 5). Polymorphic changes permitted identification of biallelic mutations in bcl-6.

The overriding criterion in determining the significance of low levels of bcl-6 mutations is theTaq error rate. Of the CLL cases identified as carryingbcl-6 mutations (Tables 3 and 4), 2 patients, A.G. and D.B., both from the V_H unmutated CLL subset, were selected for simultaneous analysis of the frequency of mutations in bcl-6and β-G genes, using identical amounts of template genomic DNA, from the same preparations. It was apparent (Table 5) that in both cases the level of bcl-6 mutation detected was significant, approximately 4-fold higher than the level of Taq misincorporation as assessed by the reference β-G gene. This finding was based on a replicate PCR analysis of bcl-6 in these 2 cases (Table 5). As an internal control, replicate PCR analysis of patient A.W. confirmed an absence of mutations in bcl-6 (Table 4), negating any influence of the bcl-6 target gene on Taq error rate at levels present in the isolated genomic DNA in these samples. The analysis of plasmid DNA carrying an unmutated bcl-6 insert, discussed above, confirmed that localized motifs in bcl-6sequence do not interfere adversely with Taq misincorporation.

Consequently, and as a conservative measure, bcl-6 mutations in each case were accepted as significant only when the frequency of mutation was more than 2-fold above the background Taq error rate (Table 5). Using this cutoff value, the incidence of bcl-6mutations appeared comparable in the 2 disease subsets, with 4 of 10 cases with unmutated V_H genes, CLL(UM), and 4 of 9 cases with mutated V_H genes, CLL(M), exhibiting bcl-6mutations (Tables 3 and 4). Overall, 11 of 19 CLL cases showed no mutations in bcl-6 in multiple clones, and this underlines the lack of false positives generated by the PCR cloning and sequencing procedure, as does the detection of polymorphic alleles bearing no other mutations. As in other B-cell tumors, bcl-6 is not targeted for mutation in all CLL cases.

The frequency of mutations in the CLL(UM) subset in this study, calculated in those cases displaying bcl-6 mutations, was in the range 6.8 to 12.0 × 10⁻⁴bp⁻¹ with a mean of 9.7 × 10⁻⁴ bp⁻¹appproximately 4-fold higher than the Taq error rate. In the CLL(M) category, in the bcl-6 mutated cases, the frequency was in the range 6.8 to 23.9 × 10⁻⁴bp⁻¹ with a mean of 13.7 × 10⁻⁴ bp⁻¹approximately 6-fold higher than the basal Taq error rate.

With respect to the CLL(M) cases, the frequency was comparable to that of bcl-6 mutations (approximately 8 to 20 × 10⁻⁴ bp⁻¹) reported in 5 mutated V_H CLL cases by Pasqualucci et al.17By contrast, the frequency of V_H gene mutations in the 4 CLL(M) cases in this investigation was 6.7 × 10⁻² bp⁻¹ about 50-fold higher than the bcl-6 rate of mutation.

The number of bcl-6 mutations per mutated clone from both subsets was in the range 1 to 6 mutations, with the bulk of clones displaying 1 to 3 substitutions (Tables 3, 4, and 5). Overall, a total of 45 distinct mutations were observed, each unique in any given allele. Almost exclusively, these were point mutations (42 of 43). However, mutations included a nucleotide insertion event +G (757) in clone JB1 (Table 4) and a deletion, ΔT (1023) in clone AG25R/2 (Table3).

In 1 CLL(UM) case, 1 mutation was repeated in more than 1 clone: in A.G., clones AG8/1 and AG/25R2 displayed the G-A (756) transition (Table 3). In 2 CLL(M) cases, repeat mutations were also found in separate clones: in C.W., 3 of 4 clones displayed an identical G-A (733) mutation, and in N.O., 2 of 6 clones displayed the same G-C (786) mutation (Table 4).

Single nucleotide substitutions predominated as the type of mutation detected. The overall characteristics of mutational changes are shown in Table 6. There was no apparent difference between the frequency of mutations affecting each base in the 2 subsets (data not shown). The low number of bcl-6mutations in this study excluded statistical analysis. However, transitions (86%) predominated over transversions (14%), with a preferential bias for T:N over A:N target bases. Each mutation was assessed for influence of flanking bases, and only 2 mutations, G-A (733) in clone CW1 and G-A (721) in clone AG2, occurred in G within an RGYW motif.

To confirm that cloning was a prerequisite to identify low levels of heterogeneous bcl-6 mutations in tumor-derived clones, a mixing experiment was undertaken: 8 clones showing no bcl-6 mutations were mixed with 2 clones showing 2 mutations each in the bcl-6intron, and the mix PCR amplified. PCR DNA was eluted, and directly sequenced. None of the mutations expected from the 2 parental clones were visible in the resultant chromatogram, confirming the necessity of cloning before bcl-6 sequence analysis.

In both categories of CLL, there was evidence for ongoing mutation inbcl-6, clear in all 4 cases (A.G., B.St., D.B., and R.D.) in the CLL(UM) subset, and in 2 of 4 cases (J.B. and N.O.) in CLL(M). In contrast, V_H analysis revealed no intraclonal variation in any case, including J.B. and N.O. In all cases, contamination by normal B cells was most likely excluded by high tumor lymphocyte counts (Tables 1 and 2). Indeed, the highest level of variation was observed in the patient (A.G.) with the highest lymphocyte count. To confirm intraclonal heterogeneity, bcl-6 analysis was repeated in 2 cases (A.G. and D.B.), and again, variation was observed in both cases (data not shown), with the higher level again found in patient A.G. Overall, single mutations accounted for most of the mutational differences among multiple tumor-derived clones, with rarer differences of up to 3 mutations, a single base insertion or a doublet mutation also observed. Intraclonal variation in bcl-6 sequence was apparent on both alleles in a number of cases, with each polymorphic allele displaying a variable level of mutation within each individual case.

The price of somatic mutation in immunoglobulin V genes of B cells, so essential for the development of high-affinity antibodies, is the danger of exposing the genome to potentially damaging mutation in other genes. Generally, it appears that control of the mechanism is tight and that extraneous mutation is rare. However, recent findings on thebcl-6 gene have revealed nucleotide changes, and it has been suggested that this gene may be a target for the same mutational process. This has been supported by the finding that mutations inbcl-6 tend to be more common in B cells, both normal and neoplastic, which have encountered the germinal center environment, where somatic mutation in V genes occurs.1-3 

Data on the incidence and distribution of somatic mutations in candidate genes clearly depend on the material used and the technology applied. For a gene such as bcl-6, the level of mutation described can vary, with frequencies in the range of 7 × 10⁻⁴ bp⁻¹ to 1.6 × 10⁻²bp⁻¹.17 The preliminary indication was that in all 5 CLL(M) displaying bcl-6 mutations, the frequency is at the lowest end of this range, approximately 8 to 20 × 10⁻⁴bp⁻¹.17 These mutations are found only in a small number of CLL cases, with over 70% showing no mutations inbcl-6.17,22 A paramount requirement therefore, given the expected low level of bcl-6 mutations in CLL, is to distinguish such mutations from sequencing error. For our study, this error rate was stringently documented using a relevant control genewithin each selected sample: two cases of CLL(UM), A.G. and D.B., which unexpectedly showed bcl-6 mutations, did so in a background of low Taq error as judged by the frequency of βG mutations in an identical amount of template genomic DNA. A number of cases, using this stringency of assay, were not mutated inbcl-6, and this was verified by repeat PCR in 1 case, showing that Taq catalysis did not introduce extraneous mutations given the level of target bcl-6 gene present in these DNA preparations. Any possible bias introduced by localized bcl-6 motifs was also excluded. The second point to recognize is polymorphic changes in the bcl-6 locus, some, but possibly not all of which have been described.10 17 Thirdly, it is highly desirable to have a homogeneous tumor population to avoid contamination with other B cells that carry mutations.

For CLL, the material consists of tumor cells at high count obtained from blood. Because the cases are generally more than 98% tumor cells, the possibility of contamination with normal B cells is negligible. We chose to use PCR amplification and cloning to investigate bcl-6mutations, partly because of the higher sensitivity, but also because this allows analysis of intraclonal variation, as verified by the mixing experiment. This phenomenon is a feature of V_Hgenes in B-cell tumors of the germinal center,4 and has also been suggested in bcl-6 genes of normal B cells.19 Intraclonal variation would be difficult to detect to a comparable degree by SSCP analysis.

We were also able to exclude in this study any association between the translocation status of the bcl-6 gene and the incidence ofbcl-6 mutations, as all the cases studied were free of this specific chromosomal abnormality. This observation in CLL extends findings in FCL and DLCL,10 where somatic mutation ofbcl-6 genes has been shown to occur independently of any chromosomal rearrangement.

After careful assessment of the Taq error rate and elimination of recurrent changes, considered to be polymorphic variation, from the analysis, we were able to identify somatic mutations in bcl-6in 8 of 19 cases of CLL. Surprisingly, these were not only in the cases with mutated V_H genes, but were also found in 4 of 10 of those cases with no mutations in V_H. The other unexpected feature was that intraclonal variation in bcl-6 was evident in 6 of the 8 mutated cases. The heterogeneity appeared to be quite wide, with only 3 cases (C.W., N.O., and A.G.) showing the same mutation in more than 1 clone. One of these (A.G.) was a CLL(UM) case. Clearly, these repeated mutations in separate clones again argue against origins by Taq error, and point to subdominant clonal populations. Not all clones in the CLL(UM) cases exhibit bcl-6mutations, and this might explain the low number of cases of CLL with mutations detected by SSCP analysis.17 In those cases in which no mutations were detected in our study, it remains possible that analysis of more than 4 clones is required to detect bcl-6changes, and to establish whether these cases are truly negative forbcl-6 mutations. It is known that intraclonal heterogeneity in V_H is extremely rare in CLL, as this is a tumor that does not reside in the germinal center.4-9 Because intraclonal heterogeneity was not found in the V_H genes of the cases with heterogeneity in bcl-6, it suggests that the mutation mechanism may operate differently on the 2 loci. Indeed, it may prove to be the case that the dislocation of somatic mutation at the V_H and bcl-6 loci is a feature peculiar to CLL. It is noteworthy here that somatic mutation of V_Loccurs in parallel to V_H in CLL tumor cells.9 

These observations raise important questions regarding current concepts of bcl-6 mutations in normal and malignant B cells. First, where do the bcl-6 mutations observed at a significant level in CLL(UM) cases, originating from naive B-cell precursors, arise, and is the same mechanism involved? Second, our data indicate that CLL cases are continually undergoing such mutations after transformation, some of which are detected as clonal mutations. These findings may not exclude a coupled mechanism targeting V_H and bcl-6mutations in CLL(M) cases, but suggest ongoing mutations at thebcl-6 locus alone. To date, analysis of bcl-6 mutations by PCR and cloning in DLCL suggest that there is no significant ongoing mutations in these cases.10 The inference is that the bcl-6 mutations in DLCL arose before neoplastic transformation, and that this locus is not targeted in tumor cells. Our observations suggest that other features of bcl-6 mutations, in addition to those currently described, may exist. Interestingly, preliminary observations of intraclonal variation in bcl-6mutations have recently been reported, both in non-Hodgkin lymphoma,23 and in highly purified multiple myeloma plasma cells.24 In addition, intraclonal sequence variation of thebcl-6 gene has been observed in normal tonsil GC B cells.19 

The frequency of bcl-6 mutations in CLL, and the predominance of point mutations favoring transitions, together with an apparent lack of an A-N strand bias are features comparable to other diverse lymphoid malignancies.17 19 Differences exist in that a T-N bias was detected, and a lack of an RGYW motif dependence of bcl-6mutations; however, the number of cases analyzed preclude any statistical evaluation.

Mutations at the V_H locus, or their absence, are a significant prognostic feature in the CLL disease, but bcl-6mutations in the 2 subsets do not apparently correlate with the prognosis. This is in contrast to posttransplantation lymphoproliferative disorders in which bcl-6 mutations have been found to have a predictive relevance to disease outcome.25 Analysis of bcl-6 rearrangements in DLCL have proved a useful prognostic marker,26 whereasbcl-6 mutational status did not correlate with prognostic significance in DLCL of primary gastric origin.27 It has been proposed that intronic mutations in the bcl-6 gene may have a role in lymphomagenesis.17,28 There are preliminary indications of these mutations leading to altered binding of regulatory proteins,29 and of defective Bcl-6 protein activity,28 but clearly the nature of the mutations and their affect on gene expression could be crucial. For CLL, Bcl-6 protein expression has not been detected,30 31 and the role of the mutations in the noncoding region are so far unknown. It is apparent that in a large number of malignant B-cell cases, including CLL, the bcl-6 gene remains unchanged. Evidently mutations inbcl-6 are not a requirement for development of CLL, but it will be of interest to probe further into the involvement, if any, of this gene in the pathology of CLL.

We would like to thank Dr Gianluca Gaidano, Amedeo Avogadro University of Eastern Piedmont, Novara, Italy, for useful discussions.