Subsets with restricted immunoglobulin gene rearrangement features indicate a role for antigen selection in the development of chronic lymphocytic leukemia

The recombination of the immunoglobulin (Ig) genes during B-cell development is a key step in generating the repertoire of antibody diversity in the immune system,¹  and the large number of combinatorial possibilities of the different heavy-chain variable (V_H), diversity (D), and joining (J_H) gene segments in the Ig heavy-chain locus and light-chain variable (V_L) and joining (J_L) gene segments in the 2 light-chain loci constitute the base for this diversity.^2,3  During V(D)J recombination, additional diversity is provided by random insertion/deletion of nucleotides in the D-J_H, V_H-DJ_H, and V_L-J_L junctions and by the usage of 3 different reading frames for the D gene, hence creating the most hypervariable regions for antigen binding (ie, the heavy- and light-chain complementarity determining region 3 [CDR3]). Furthermore, the final affinity maturation of the Ig molecules occurs in the germinal center during an antigen response, which can generate high-affinity antibodies by introduction of somatic hypermutation in the Ig gene rearrangements. Thus, the rearrangements of the Ig genes will provide each B-cell with a highly unique B-cell receptor (BCR) and can therefore serve as tumor markers in B-cell malignancies.⁴ 

B-cell chronic lymphocytic leukemia (CLL) is the most common adult leukemia in Western countries and is characterized by a continuous expansion of CD5⁺/CD19⁺/CD23⁺ monoclonal B cells. In this clinically heterogeneous disease, analysis of the somatic hypermutation status of the V_H genes has proven to be a powerful prognostic predictor with mutated cases showing almost twice as long overall survival compared with patients with tumor cells lacking somatic hypermutation.^5-10  Analysis of V_H gene usage has revealed a biased repertoire with particular V_H genes preferentially used compared with normal B cells, where the most prevalent are the V_H1-69, V_H3-07, V_H3-21, and V_H4-34 genes.^6,11-18  The V_H1-69⁺ rearrangements in CLL were previously reported to display molecular peculiarities such as longer than average heavy-chain CDR3s (HCDR3s; ∼19 amino acids) with skewed D3-3 and J_H6 gene usage compared with normal B cells.^13,14,17,18  This finding of preferential V_H, D, and J_H gene usage as well as similar amino acid motifs in the HCDR3s in this subset of CLL led to the speculation of a possible antigen component involved in the pathogenesis of CLL with the hypothesis that BCRs encoded by specific V_HDJ_H combinations could generate Ig molecules with affinity to similar antigenic epitope(s).^13-15 

Recently, we demonstrated that CLL cases with tumor cells rearranging the V_H3 family gene, V_H3-21, showed poor overall survival,^16,18  which has also lately been confirmed by others.^19,20  The V_H3-21 subset (∼11% of our CLL cohort) displayed predominantly mutated V_H genes but still had an inferior clinical outcome and therefore did not fit into the prognostic classification scheme of mutated or unmutated CLL.¹⁸  In addition, the V_H3-21 cases showed certain molecular characteristics with shorter than average HCDR3s (∼10 amino acids) and almost identical HCDR3 sequences with a conserved motif (ala-arg-asp-ala-asn-gly-met-asp-val) and a highly restricted usage of the V_λ2-14 gene in most cases.^16,18  Considering that the CDRs of both the heavy- and light-chain V region comprise the antigen-binding site of the Ig molecule and that the HCDR3 is believed to contribute the most diversity in this respect,²¹  the finding of similar HCDR3s coupled with V_λ2-14 gene usage indicated a common antigen-binding site in the BCRs of V_H3-21⁺ CLL and suggested stimulatory influence of an unknown antigen in the disease development.

To further investigate signs of antigen selection in CLL, we studied the BCRs by analyzing the V_H, D, and J_H gene usage as well as performing multiple sequence alignment of the HCDR3 amino acid sequences obtained from 368 functional V_H gene rearrangements amplified in 346 CLL cases. Furthermore, we carried out V_L gene analysis in selected groups. Our data reveals further subsets of CLL with restricted HCDR3 features combined with preferential V_L gene usage, thus indicating that BCRs with similar structures of the antigen-binding site were selected, which further corroborates the possibility of antigen involvement in CLL development.

Tumor samples were collected from 346 patients with CLL that had been identified from the archives of frozen tissue specimens at the University Hospitals of Uppsala (n = 164), Linköping (n = 79), Umeå (n = 51), and Huddinge (n = 19) in Sweden and at Tampere University Hospital (n = 33) in Finland, between 1981 and 2001. Frozen tumor material was obtained mainly from peripheral blood and bone marrow but in a few cases also from lymph nodes, spleen, and other sources. Morphologic classification was performed according to the World Health Organization (WHO) classification, and tumor cells typically expressed CD5 and CD23 and had weak expression of Ig.²²  The median age at diagnosis was 65 years (range, 36-88 years) for the entire cohort, with a male-to-female ratio of 2:1.

DNA was prepared from fresh-frozen tumor material using standard protocols including proteinase K treatment. V_H and V_L gene family-specific polymerase chain reaction (PCR) amplification was performed using consensus V_H/J_H primers and V_L/J_L primers as previously described.^23,24  The sequence reactions were performed using either the BigDye Terminator Cycle Sequencing Reaction kit (Applied Biosystems, Foster City, CA) or the DYEnamic ET Dye Terminator Kit (Amersham Biosciences, Piscataway, NJ) and analyzed with an automated DNA sequencer (ABI 377 or ABI3700, Applied Biosystems; and MegaBACE 500 DNA Analysis System, Amersham Biosciences). The sequences were aligned to Ig sequences from the GenBank,²⁵  V-BASE,²⁶  and IMGT²⁷  databases. V_H sequences deviating more than 2% from the corresponding germ line gene were defined as mutated. For D gene determination, a requirement of a minimum of 7 matching nucleotides was used. The length of the HCDR3 was calculated between codons 95 and 102 as previously described by Kabat.²⁸  All in-frame V_H rearrangements were converted into amino acid sequences and the HCDR3 sequences were aligned in the multiple sequence alignment software Clustal X (1.83) for Windows (UBC Bioinformatics Centre, Vancouver, BC, Canada). Selection of V_H groups with restricted HCDR3s was performed using the following criteria: (1) the group members had same V_H gene usage, (2) the V_H gene rearrangements showed at least 60% mean homology between the HCDR3s, and (3) the group contained at least 3 CLL cases. The numbering of cases in the figures and Table 3 refers to the order in which the V_L gene PCR was performed.

A total of 407 V_H gene rearrangements were PCR amplified and sequenced from 346 CLL cases, of which 294 cases displayed 1 rearrangement, 52 cases displayed 2 rearrangements, and 3 cases displayed 3 rearrangements. In total, 39 V_HDJ_H sequences were either out of frame or had a stop codon introduced in the HCDR3. The V_H family gene usage was as follows: V_H1 (28%); V_H2 (2%); V_H3 (47%); V_H4 (19.5%); and V_H5, V_H6, and V_H7 (< 2% each). In this CLL cohort, 7 V_H genes accounted for approximately 50% of the total V_H genes rearranged: V_H1-69 (14%), V_H3-21 (9%), V_H4-34 (7%), V_H3-30 (6%), V_H3-23 (5%), V_H4-39 (4%), and V_H1-2 (4%) (Table 1). Of the 51 functional V_H genes, 49 were represented in this material. Using the 2% mutation border, 146 (42%) cases were considered mutated with a mean mutation frequency of 5.8% (range, 2.1%-13.2%), and 200 (58%) cases were unmutated. In terms of mutated CLL cases, the most frequent V_H gene rearranged was V_H3-21 (24 cases), whereas V_H1-69 (54 cases) represented the most frequent V_H gene in unmutated cases (Table 1).

The D gene segment rearranged was identified in 341 of 407 V_H gene rearrangements with 5 genes representing 57% of the rearranged D genes: D3-3, D6-19, D2-2, D3-22, and D3-10 (Figure 1). The identification of a D gene segment was possible in 94% of the unmutated rearrangements but in only 64% of the mutated rearrangements. The D genes D3-3, D2-2, and D6-19 represented 49% of the rearrangements in the unmutated V_H gene rearrangements, whereas a more diverse spread was seen in the mutated rearrangements. J_H4 and J_H6 were the most frequently rearranged J_H genes (71%, 287 of 365 identified J_H genes; Figure 1) in both the mutated and unmutated subsets.

In this material, there were 13 different V_H/D/J_H combinations that were present in 3 or more cases (data not shown). However, considering that the preferential usage of certain V_H/D/J_H combinations does not necessarily indicate that the HCDR3s will show similarities at the amino acid level, since the D segment can be used in all 3 reading frames and as the rearrangements can display different N-regions, we decided to perform cluster analysis of the HCDR3 amino acid sequences only to identify subsets with restricted HCDR3 structures using a particular V_H gene, as have been shown for the V_H3-21 group. Thus, the translated amino acid sequences of the HCDR3s from 368 functional V_H gene rearrangements were collected into one file and multiple sequence alignment analysis was carried out. From this analysis, we could identify 4 groups using the V_H3-21 (22 cases), V_H1-69 (2 groups, 4 and 3 cases, respectively), and V_H1-2 (5 cases) genes with high similarities (75%-95% mean homology) in their HCDR3s (designated as “Groups with highly restricted HCDR3s”; see Table 2). In addition, another 3 groups with V_H1-3 (9 cases), V_H1-18 (3 cases), and V_H4-39 (5 cases) gene usage showed restricted features but with less HCDR3 homology (60%-75%; designated as “Groups with moderate HCDR3 homology”; see Table 2). The largest group contained 22 V_H3-21 cases with highly restricted HCDR3s, 19 of which have been published previously and will therefore not be shown further.¹⁸  All amino acid sequences are illustrated in Figure 2 in their respective groups. None of the groups showed evidence of somatic hypermutation in their V_H regions. The nucleotide sequences from the V_H gene rearrangements are available in GenBank²⁵  at accession numbers AY486162-AY486231.

To investigate any association between V_L gene usage and HCDR3 composition as in the V_H3-21 group, V_L gene analysis was performed in these groups with high or moderate HCDR3s similarities. The nucleotide sequences from the V_L gene rearrangements are available at GenBank²⁵  accession numbers AY490826-AY490888. Interestingly, the same rearranged V_L gene was found in the majority of cases when coupled to a similar HCDR3 in the V_H gene rearrangement (Figure 2).

V_H1-69 (2 groups). The first V_H1-69 group (4 cases) used a D3-16 and a J_H3 gene (Table 2) and displayed only 1 to 2 amino acid differences in the 18-amino acid-long HCDR3 sequence (Figure 2A). The D gene was used in the same reading frame and the HCDR3 showed identical lengths in all rearrangements. The N-segment in the V_H/D junction consisted of 3 amino acids (except CLL27) and 2 amino acids were identical in all rearrangements, whereas 4 of 5 amino acids were the same in the D/J_H junction. In one position, 1 amino acid difference was found between all rearrangements, which contained either an Asn, Tyr, His, or Asp amino acid, 3 of which have polar/hydrophilic properties. All 4 cases rearranged a V_κA27 gene combined with a J_κ1 or J_κ3 gene (Figure 2A; Table 3). In addition to these cases, we identified from the public databases 4 other V_H1-69⁺ CLL cases and 1 anticardiolipin antibody with highly similar HCDR3s. The differences between the HCDR3s in the 8 CLL cases were restricted to 2 positions in this 18-amino acid-long region (Figure 2A).

The second V_H1-69 group (3 cases) rearranged a D3-3 and a J_H6 gene. The HCDR3 was 21 amino acids long and the same D gene reading frame was demonstrated. The D/J_H junction contained 1 amino acid that was identical in all sequences (Figure 2A), whereas the V_H/D junction displayed higher variability with varying amino acid properties between the sequences. Two of the 3 cases rearranged a V_λ2-6 gene combined with a J_λ2 gene (Figure 2A; Table 3), whereas the V_L sequence was not determined in the remaining case (CLL72).

V_H1-2. The V_H1-2 group consisted of 5 cases using the D1-26 and J_H6 genes. These rearrangements displayed an identical D gene reading frame and identical HCDR3s lengths (14 amino acids; Figure 2A). The V_H/D and D/J_H junctions consisted of only 1 amino acid each; 1 position was identical between the rearrangements, whereas the other position contained the same amino acid in 3 cases and different amino acids in the remaining 2 cases, all differences resulting in a hydrophobic amino acid. Three cases rearranged a V_κB3 gene joined with a J_κ2 or J_κ4 gene, whereas 1 case rearranged a V_κA17 gene, and in 1 case the V_L usage was not determined (Figure 2A; Table 3).

V_H1-3. Nine cases using the V_H1-3/D6-19/J_H4 genes displayed a mean of 65% homology between their HCDR3s, with 5 of 9 showing 70% to 80% homology (Table 2; Figure 2A). The same D gene reading frame was used in all rearrangements and the HCDR3 had the same length in 7 of 9 cases. The V_H/D junction (1 amino acid) was identical in 7 of 9 cases; however, more variability was indicated in the D/J_H junction. Difference in the HCDR3s included amino acids with similar properties in 3 positions (1 in the V_H/D junction and 2 in the D/J_H junction), such as glu/asp/gln (all polar/hydrophilic), val/pro/gly/ala or ile/leu/met/phe/ala/pro (all non polar/hydrophobic), and 2 positions (both in the D/J_H junction) with varying properties. All V_L rearrangements used a V_κO2 gene together with a J_κ1 (4 cases), J_κ2 (4 cases), or J_κ4 (1 case) gene (Figure 2A; Table 3).

V_H1-18. Three cases were found to rearrange either a D6-19 (CLL57 and CLL56) or a D3-22 (CLL59) gene, all of which used a J_H4 gene (Table 2). Even though 2 different D gene segments were used, they resulted in a conserved 3-amino acid segment, Gln, Trp, and Leu, and the HCDR3s were of the same length in all 3 sequences (Figure 2B). The CDR3s showed 60% homology, but the N regions were less conserved and the different amino acids displayed varying properties. All 3 cases rearranged a functional V_κO2 with either a J_κ1, J_κ2, or J_κ3 gene (Figure 2B; Table 3).

V_H4-39. Homology in the V_H4-39/D6-13-D6-19/J_H5 group varied between 62% and 75% in 5 cases (Table 2). The D gene displayed the same reading frame, and the rearrangements had similar lengths of the HCDR3s despite differences in the V_H/D and D/J_H junctions (Figure 2B). In 3 positions (1 in the V_H/D and 2 in the D/J_H junction), 4 different amino acids were used in the separate sequences, with 2 positions (amino acid differences of Asp, Gly, Ser, or Ala and Thr, Asn, Gly, or Leu) showing differing properties and the third position (Arg, Ser, Thr, or Glu) having similar properties (polar/hydrophilic). Four of 5 cases rearranged a V_κO2 gene with either a J_κ2 or J_κ4 gene and the remaining case (CLL8) had a V_κA3/J_κ2 rearrangement (Figure 2B; Table 3).

In addition to the above groups, a number of HCDR3s were found in sets of 2 sequences, which, because of their low frequency, are detailed separately. Two V_H1-69/D3-16/J_H4, 2 V_H1-2/D3-9, D3-3/J_H4, and 2 V_H1-46/D3-22/J_H4 rearrangements as well as 2 V_H4-34/D-/J_H6 and 2 V_H4-34/D2-15/J_H6 rearrangements showed restricted HCDR3s with more than 60% homology (data not shown). The 2 V_H4-34 groups and 1 of the V_H1-2 sequences displayed evidence of somatic hypermutation. The V_H4-34/D-/J_H6 cases both rearranged aV_κA27 gene, whereas the 2 V_H4-34/D2-15/J_H6 cases rearranged a V_κA17 gene (Table 3). The V_L gene rearrangement was determined in 1 of the V_H1-69/D3-16/J_H4 cases, which rearranged a V_κO8 gene, and in 1 of the V_H1-46/D3-22/J_H4 cases rearranging a V_κL2 but in none of the 2 V_H1-2/D3-9, D3-3/J_H4 cases (Table 3).

In order to study whether the restricted V_L gene usage was linked to the HCDR3 composition or to the particular V_H gene used, an additional 33 cases were analyzed by V_L gene amplification within these groups using the same V_H gene (V_H1-69, V_H4-39, V_H1-2, V_H1-3, and V_H1-18; Table 3). As illustrated in Figure 3, a diverse usage of V_L genes was found in 20 V_H1-69 using cases, and V_L gene restriction was only demonstrated in cases with similar HCDR3s; the 4 V_H1-69/D3-16/J_H2 cases and the 2 V_H1-69/D3-3/J_H6 cases used a V_κA27 gene or a V_λ2-6 gene, respectively. These 2 V_L genes were found in only 2 of the 14 other V_H1-69 cases without restricted HCDR3s and instead 7 different V_κ and 3 different V_λ genes were used. In the V_H4-39 cases, 12 of 16 were analyzed for V_L gene usage, 4 of 5 cases with similar HCDR3s rearranged the V_κO2, whereas 7 other cases without HCDR3 homology rearranged 6 different V_L genes (Table 3). In the V_H1-2 cases, the 5 with restricted HCDR3s showed preferential V_L gene use, whereas the 4 nonrestricted cases rearranged different V_L genes: V_κO2, V_κL2, V_κL10, and V_λ1-17 (Table 3). In the V_H1-3-using cases, 11 were analyzed with 9 showing restricted HCDR3s and a V_κO2 rearrangement, whereas the remaining 2 cases without HCDR3 homology used a V_κB3 and a V_κL12 gene (Table 3). Six V_H1-18 cases were analyzed and 3 of these had restricted V_κO2 usage and similar HCDR3s, whereas the remaining 3 cases showed no bias in V_L gene use (Table 3).

Alignment analyses of the HCDR3s from 368 functional V_H rearrangements revealed 3 subsets with highly similar HCDR3s (> 75% homology) as well as 3 subsets with moderate HCDR3 homology (60%-75%) in addition to the previously described V_H3-21⁺ group. The first 3 groups (2 V_H1-69⁺ groups and 1 V_H1-2⁺ group) displayed identical V_H, D, and J_H gene use; identical CDR3s lengths; and shared N-regions; whereas the latter 3 groups (V_H1-3, V_H1-18, and V_H4-39) showed restricted V_H, D, and J_H gene usage; similar lengths of the CDR3s; and similar D reading frame but less conserved N-region composition. In parallel with the V_H3-21⁺ subset, V_L gene rearrangement analysis showed a strikingly biased V_L gene usage for both groups showing high and moderate HCDR3 homology (Figure 2). For instance, in the V_H1-69/D3-16/J_H3 group, all 4 cases with similar HCDR3s had the same V_κA27 gene usage, whereas 14 other cases without HCDR3 restriction used 10 different V_L genes (Figure 3). Additionally, the V_H4-39 group with HCDR3 restriction showed preferential V_L gene use, whereas the remaining V_H4-39 cases used a variety of V_L genes. Notably, despite having lower homology in the N-regions in the V_H1-3, V_H1-18, and V_H4-39 groups, they all displayed V_L gene rearrangements with almost identical V_L use (V_κ02), which is in contrast to the non-HCDR3-restricted cases using the same V_H genes. This latter finding strengthens the association between restricted HCDR3 features and biased V_L use, although the N-regions do not show identical amino acid sequences.

Our novel data are indicative of antigen selection in subsets of CLL, especially in the groups with highly similar HCDR3 features, but we also consider this possibility for the subgroups showing moderate HCDR3 homology. This interpretation is drawn from the following facts. Considering the probability that 2 V_HDJ_H rearrangements occur by chance is low (1/51 × 1/27 × 1/6 = 1/8262), and in combination with a particular V_L gene rearrangement very low (in combination with 1 V_κJ_κ, 1/8262 × (1/40 × 1/5) = 1/1.6 million), these findings are unlikely to be a random phenomenon. Also, if the junctional diversity and the use of D reading frame is taken into account, the probability will diminish to less than 10^-12. Furthermore, using a similar approach to analyze 227 HCDR3s from normal CD5⁺ B cells available from the public databases (Genbank²⁵ ), similar rearrangements were demonstrated between only 2 sequences, indicating that our findings in CLL seem not to be a reflection of the normal B-cell repertoire (data not shown).^29,30  This is in parallel with the finding of no duplicate Ig rearrangements in about 10 000 Ig sequences obtained from normal tonsils.³¹  Further support to our data are also given by the previous reports showing CLL cases using the V_H1-69/D3-3/J_H6 genes or V_H4-39/D6-13/J_H5 combinations with similar restriction of amino acid motifs within the HCDR3 as in our groups.^13,32  We could also identify CLL rearrangements in the GenBank database²⁵  with similar sequences such as the 4 V_H1-69/D3-16/J_H3 rearrangements (db1-4; Figure 2A), thus strengthening the finding of this particular subgroup. In parallel with our report, a recent abstract also showed combined V_H/V_L gene usage in V_H1-69, V_H3-21, V_H4-34, V_H1-2, and V_H1-3-using CLL with high HCDR3 homology.³³  Our study thus revealed indirect signs of antigen selection in subsets of CLL, but we cannot exclude that other CLL subgroups could exist that we have not been able to detect due to low frequencies in this material. Indeed, a number of small groups occurring in low frequency were identified using the V_H1-69, V_H4-34, and the V_H1-2 genes with restricted HCDR3 structures and also with similar light-chain rearrangements. A large-scale worldwide study of V_H/V_L gene rearrangements in CLL is now necessary in order to elucidate all different subgroups in detail and possible geographical differences. For instance, the V_H3-21 group has been shown in a higher frequency in Sweden compared with other large V_H gene studies, which may indicate that regional difference between populations may be of importance.^34-36 

Our data suggests a limited number of antigens involved in the selection of these BCRs; however, the identity of such antigens is currently unknown. These antigens could either be self- or non-self-antigens that may trigger and promote proliferation of B cells expressing certain BCRs, leading to clonal expansion and increased risk of transformation. The groups with high HCDR3 restriction, such as the V_H1-69 and V_H1-2 subgroups, showed limited variation between their HCDR3s with a higher likelihood that they could bind similar antigens, whereas this is less obvious in the moderate-homology groups that had further amino acid differences, although their restricted V_H/D/J_H and V_L rearrangement may indicate antigen selection also within these groups. Interestingly, the recently described V_H4-39/V_κO2-expressing IgG⁺ CLL subset, with BCR structures similar to our V_H4-39 group, displayed comparable antigen-binding sites as monoclonal antibodies reacting toward bacterial carbohydrates as well as different autoantigens.³²  Furthermore, a V_H1-69/V_κA27 antibody obtained from a CLL patient showed low-affinity RF activity and bound to myoglobulin, thyroglobulin, actin, and ssDNA.³⁷  An anticardiolipin antibody encoded by a rearrangement similar to the V_H1-69/D3-16/J_H3 group in this study has also been indicated (GenBank²⁵  accession no. AAL67508), but no data were available concerning the V_L gene usage. Moreover, the restriction of the HCDR3s found in our subgroups was mainly confined to the unmutated cases with poor prognosis. Recent data has shown that unmutated cases display a higher capacity to signal through their BCRs as measured by syk phosphorylation.³⁸  Signaling through the BCR may therefore play a vital role in the development and, perhaps, progression of the malignant clone in unmutated CLL. The restricted BCRs shown in groups with unmutated V_H genes would hence indicate that a selected number of antigens could play an important role in the progression of the clonal expansion in the unmutated CLL subsets. Further studies of the binding sites of the restricted BCRs and putative antigens are now necessary to elucidate the possibility of antigen involvement in CLL development.

The authors are grateful to Inger Eriksson and Anita Lindström for skilful technical assistance and to Elisabeth Grönlund and Magnus Hultdin for assistance with the diagnostics of CLL.