Partial versus Productive Immunoglobulin Heavy Locus Rearrangements in Chronic Lymphocytic Leukemia: Implications for B-Cell Receptor Stereotypy

Abstract

The frequent occurrence of stereotyped heavy complementarity-determining region 3 (VH CDR3) sequences among unrelated cases with chronic lymphocytic leukemia (CLL) is widely taken as evidence for antigen selection. Stereotyped VH CDR3 sequences are often defined by the selective association of certain immunoglobulin heavy diversity (IGHD) genes in specific reading frames with certain immunoglobulin heavy joining (IGHJ ) genes. To gain insight into the mechanisms underlying VH CDR3 restrictions and also determine the developmental stage when restrictions in VH CDR3 are imposed, we analyzed partial IGHD-IGHJ rearrangements (D-J) in 829 CLL cases and compared the productively rearranged D-J joints (that is, in-frame junctions without junctional stop codons) to (a) the productive immunoglobulin heavy variable (IGHV )-IGHD-IGHJ rearrangements (V-D-J) from the same cases and (b) 174 D-J rearrangements from 160 precursor B-cell acute lymphoblastic leukemia cases (pre-B acute lymphoblastic leukemia [ALL]). Partial D-J rearrangements were detected in 272/829 CLL cases (32.8%). Sequence analysis was feasible in 238 of 272 D-J rearrangements; 198 of 238 (83.2%) were productively rearranged. The D-J joints in CLL did not differ significantly from those in pre-B ALL, except for higher frequency of the IGHD7-27 and IGHJ6 genes in the latter. Among CLL carrying productively rearranged D-J, comparison of the IGHD gene repertoire in productive V-D-J versus D-J revealed the following: (a) overuse of IGHD reading frames encoding hydrophilic peptides among V-D-J and (b) selection of the IGHD3-3 and IGHD6-19 genes in V-D-J junctions. These results document that the IGHD and IGHJ gene biases in the CLL expressed VH CDR3 repertoire are not stochastic but are directed by selection operating at the immunoglobulin protein level.

INTRODUCTION

A major mechanism for the generation of a broad repertoire of human antibody specificities is the rearrangement of one each from a cluster of multiple and distinct immunoglobulin (IG) heavy and light variable (V), diversity (D: for heavy chains only) and joining (J) genes (1). This combinatorial diversity is further increased by the junctional diversity that results from differential trimming of nucleotides from the ends of the rearranged genes (exonuclease activity) along with the addition of random nucleotides between them (terminal deoxynucleotidyl transferase [TdT] activity) (2,3). This step ensures that the somatically created complementarity-determining region 3 (CDR3) is the most diverse part of the variable domain.

The CDR3 of the V domain of the heavy chain (VH) plays a critical role in antigen recognition to the extent that the more similar the primary VH CDR3 sequences of two IGs, the more similar their folding and, likely, their specificities (4). Thanks to the accumulated effects of both combinatorial and junctional diversity, normally, the chances that two independent B-cell clones will carry identical VH CDR3 are considered as negligible.

Against this background, chronic lymphocytic leukemia (CLL) is uniquely characterized by the existence of subsets of cases with quasi-identical (stereotyped) VH CDR3 sequences within their B-cell receptors (BcRs) that collectively account for up to 30% of the cohort (5,6). Such striking BcR similarity in unrelated CLL cases has been widely interpreted as evidence for a role for a limited set of antigens or structurally related epitopes in leukemogenesis (5,7–10).

VH CDR3 formation starts at the early stages of B-cell development by the rearrangement of an immunoglobulin heavy constant delta (IGHD) to an immunoglobulin heavy joining group (IGHJ) gene, followed by the rearrangement of an immunoglobulin heavy variable group (IGHV) gene to the partially rearranged IGHD-IGHJ genes. Partial IGHD-IGHJ rearrangements (D-J) hallmark the differentiation of common lymphoid precursors into the B-cell lineage and are already present in CD34⁺/CD19⁻/CD10⁺ precursor cells (11,12). Partial IGHD-IGHJ rearrangements (even if productively rearranged) do not always recombine to an IGHV gene. In this case, the IGHD-IGHJ remains as a partial D-J rearrangement in the immunoglobulin heavy (IGH) locus. The biological mechanism, which determines whether a partial IGHD-IGHJ rearrangement will develop into an IGHV-IGHD-IGHJ rearrangement is still unknown.

Studies in the mouse have shown a biased usage of IGHD genes in reading frames (RFs) encoding for hydrophilic amino acids in IGHV-IGHD-IGHJ rearrangements (V-D-J) along with counter-selection of IGHD-IGHJ rearrangements with IGHD genes in RF encoding for hydrophobic amino acids or stop codons (13–18). Relevant studies have also shown that IGHD-IGHJ with IGHD genes rearranged in RF encoding for hydrophobic amino acids can be translated into a truncated μ protein, called Dμ (13,19). The Dμ protein was found to be expressed on the cell surface of murine pre-B cells assembled with surrogate light chains and the coreceptors CD79A (Igα) and CD79B (Igβ). This defective pre-B receptor acts as a signal transducing receptor and may interfere with B-cell development by inhibiting further IGHD-IGHJ replacements as well as IGHV-IGHD-IGHJ rearrangements (20).

Partial IGHD-IGHJ rearrangements are readily found in pre-B cells. They have also been reported in a proportion of both precursor B-cell acute lymphoblastic leukemia cases (21) and mature B-cell malignancies (22–24). Accumulating evidence from previous reports suggests that these rearrangements may offer a glimpse into the ontogenetic history of normal and malignant B cells as well as the mechanisms underlying the diverse IG repertoires observed at different developmental stages or different B-cell neoplasms.

With this in mind, we here analyzed the molecular features of D-J rearrangements in a series of 829 patients with CLL, which we subsequently compared with (a) the productive V-D-J from the same CLL clones and (b) the D-J obtained from a series of 160 patients with acute lymphoblastic leukemia (ALL), selected for comparison as representative of a malignancy of early B cells, thus contrasting CLL, a malignancy of mature B cells. Our aim was to gain insight into the mechanisms that shape the IG repertoire and determine the developmental stage when VH CDR3 restrictions are imposed in CLL.

MATERIALS AND METHODS

Patient Group

Blood samples were collected from 829 consecutive typical CLL cases from collaborating institutions in France and Greece who met the International Workshop on CLL/National Cancer Institute (iwCLL/NCI) diagnostic criteria (25). Blood and/or bone marrow samples were also collected from 160 cases with B-cell lineage ALL from Germany, diagnosed according to World Health Organization criteria and immunological marker analysis and selected on the basis of carrying partial IGHD-IGHJ rearrangements, amplified with BIOMED-2 protocols (26). The study was approved by the local ethics review committee of each institution. In all CLL cases, the tumor load was at least 80%.

Polymerase Chain Reaction Amplification and Sequence Analysis of IGHV-IGHD-IGHJ Gene Rearrangements

Mononuclear cells were separated from blood samples (and/or bone marrow) by centrifugation on a Ficoll-Hypaque gradient. Polymerase chain reaction (PCR) amplification of IGHV-IGHD-IGHJ gene rearrangements was performed on either genomic DNA (gDNA) or complementary DNA (cDNA), as previously described (27). Purified PCR amplicons were subjected to direct sequencing on both strands. Sequence data were analyzed using the IMGT^® databases and the IMGT/V-QUEST tool (http://www.imgt.org) (28,29).

PCR Amplification and Sequence Analysis of Partial IGHD-IGHJ Gene Rearrangements

PCR amplification of partial IGHD-IGHJ gene rearrangements was performed on gDNA using seven subgroup-specific IGHD primers in combination with a consensus IGHJ primer, following the BIOMED-2 experimental protocol (26). Primers were designed to anneal to the intron upstream of the IGHD gene in distinct positions relative to the recombination signal sequence (RSS) elements for each IGHD subgroup, resulting in PCR product sizes specific for each IGHD subgroup, ranging from 100 to 420 bp. Primer annealing to intron sequences precludes the possibility of amplification of the IGHD-IGHJ joint of the coexpressed IGHV-IGHD-IGHJ rearrangement (26). In all cases, the clonal amplification PCR product was easily discriminated from the polyclonal “background” (if present) obtained of residual normal B cells that was faint and did not hinder the detection of a predominant PCR band.

Purified PCR products were subjected to direct sequencing on both strands, and sequence analysis was performed by a multistep procedure using the Basic Local Alignment Search Tool (BLAST, http://blast.ncbi.nlm.nih.gov), the Expert Protein Analysis System (ExPASy, http://au.expasy.org) and databases, algorithms and tools from IMGT^®, the international ImMunoGeneTics information system (http://www.imgt.org).

As a first step, we identified the rearranged IGHD and IGHJ genes by aligning each sequence to the entire IGHD-IGHJ germline sequence (GenBank accession number EMB/X97051) using the “align2sequences” tool of “specialized BLAST.” The imprint of exonuclease and TdT activity in the D-J joint was determined by comparing each sequence to the corresponding germline IGHD and IGHJ gene according to the IMGT rules for junction analysis of the immunoglobulin genes (http://www.imgt.org/IMGT_jcta/share/textes/userGuide.html#VDJends). The patterns “m,” “m-“ and “‘mm—” were trimmed from the 3′ end of the D-REGION, the patterns “m,” “-m” and “—mm” were trimmed from the 5′ end of the J-REGION, where “m” indicates a mutation and “-” indicates an identical nucleotide by comparison with the corresponding germline sequences, ensuring that at least two nucleotides at each end of the IGHD and IGHJ genes are identical to germline. Following this rule, we then determined all nucleotide changes in these sequences that met the criteria for being considered as mutations.

Subsequently, each nucleotide sequence was translated in all RFs using the “translate tool” from ExPASy. Given that, in a productive rearrangement, the IGHJ gene must be read in the RF that encodes the motif WGXG (where “W” is tryptophan, “G” is glycine, “X” is any amino acid and “G” is glycine, respectively), each amino acid sequence carrying the WGXG motif was further compared with the amino acid sequences of the corresponding IGHD gene in all three RFs, to define the proper IGHD RF in each rearrangement. According to IMGT^®, IGHD RF1/RF2/RF3 was counted from the first/second/third nucleotide of the D-REGION (that follows the 5′D-HEPTAMER), respectively.

Rearrangements carrying a stop codon in the IGHD-IGHJ joint were considered as unproductive (following the IMGT definition of the functionality), since they could not lead to a productive IGHV-IGHD-IGHJ rearrangement. Productively rearranged D-J joints with a stop codon within the IGHD gene sequence were retained in the analysis, since they could eventually lead to a productive IGHV-IGHD-IGHJ rearrangement after removal of the stop codon during IGHV to IGHD-IGHJ recombination (for example, by exonuclease trimming of the IGHD gene, a feature often seen in productive rearrangements from normal, autoreactive or malignant B-cell clones, including CLL).

Statistical Analysis

Descriptive statistics for discrete parameters included counts and frequency distributions. For quantitative variables, statistical measures included means, medians, standard deviation and min–max values. Significance of bivariate relationships between variables was assessed using χ² and t tests. For all comparisons, a significance level of P at 0.05 was set, and all statistical analyses were performed with the statistical package SPSS version 17.0 (SPSS, Chicago, IL, USA).

All supplementary materials are available online at www.molmed.org.

RESULTS

Molecular Features of Productive IGHV-IGHD-IGHJ Rearrangements in CLL

Productive IGHV-IGHD-IGHJ rearrangements (V-D-J) were successfully sequenced in all 829 CLL cases included in the study. A total of 12 of 829 CLL cases (1.4%) carried double productive IGHV-IGHD-IGHJ rearrangements. Using the 98% cutoff value for identity to the closest germline gene, 455 of 842 (54%) IGHV sequences were considered as “mutated,” whereas the remaining 387 of 842 (46%) were considered as “unmutated” (283 of 387 unmutated sequences had 100% identity to germline). In keeping with previous reports (30,31), the most frequent IGHV genes in the mutated subgroup were IGHV4-34, IGHV3-23 and IGHV3-7, whereas the IGHV1-69, IGHV1-2 and IGHV4-39 genes predominated in the unmutated subgroup (Supplementary Table 1). Following previously described criteria (6), overall, 215 of 829 cases (25.9%) were assigned to different subsets with stereotyped BcR.

IGHD3 subgroup genes predominated among V-D-J, followed by IGHD2 subgroup genes. At the individual gene level, IGHD3-3 was the most frequent gene (131/842 V-D-J, 15.6%), followed by the IGHD6-19 (92/842 V-D-J, 10.9%) and IGHD2-2 genes (76/842 V-D-J, 9%) (Supplementary Table 2) (Figure 1). Among the IGHJ genes, IGHJ4 and IGHJ6 were the most frequently used genes in V-D-J (365/842, 43.3%, and 244/842, 29%, respectively) (Supplementary Table 3).

Figure 1.

Differential usage of selected IGHD genes in D-J in CLL, D-J in ALL and productive V-D-J in CLL.

Molecular Features of Partial IGHD-IGHJ Rearrangements in CLL

Partial IGHD-IGHJ rearrangements (D-J) were detected in 272 of 829 patients with CLL (32.8%). No significant associations were identified regarding the frequency of D-J in relation to IGHV gene usage or mutational status or BcR stereotypy observed in the productive rearrangement from the other allele, thereby indicating no intrinsic differences between CLL cases with different BcR.

Sequence analysis was feasible in 238 of 272 D-J, of which 198 (83.2%) were productively rearranged (no stop codon at the IGHD-IGHJ joint). Five D-J carried readily identifiable IGHD-IGHD gene fusions (Supplementary Table 4), four of which were productively rearranged. Overall, 21 different IGHD genes were identified in 198 productively rearranged D-J.

IGHD2 subgroup genes predominated among productively rearranged D-J, with IGHD2-2 being the far most frequent gene (56 of 198 D-J, 28.3%), followed by IGHD2-15 (19 of 198 D-J, 9.6%) (Figure 1; Supplementary Table 5). The IGHJ4 and IGHJ6 genes were used at similar frequencies, collectively accounting for 65.6% of the series (130 of 198 cases; Supplemental Table 6). All three RFs (RF 1, 2 and 3) of the IGHD genes were equally distributed among productively rearranged D-J (Table 1).

Table 1.

IGHD gene RF usage and hydropathicity in productively rearranged IGHD-IGHJ (D-J) versus productive IGHV-IGHD-IGHJ (V-D-J) in CLL and ALL.

	D-J in CLL	V-D-J in CLLa	D-J in ALL	V-D-J in ALLa
IGHD RF
RF1	73 (36.1)	40 (20.8)	48 (34.5)	14 (28.6)
RF2	64 (31.7)	100 (52.1)b	43 (31.0)	17 (34.7)
RF3	65 (32.2)	52 (27.1)	48 (34.5)	18 (36.7)
IGHD RF
Hydrophilic RF	70 (34.6)	118 (61.5)b	51 (36.7)	18 (36.7)
Hydrophobic RF	69 (34.2)	51 (26.6)	45 (32.3)	22 (44.9)
RF with stop codon(s)	63 (31.2)	23 (11.9)	43 (31.0)	9 (18.4)

Data are n (%).

^aIn 6 of 198 and 4 of 53 V-D-J in CLL and ALL, respectively, the IGHD gene was not determined.

^bStatistically significant higher frequency for RF2 and hydrophilic RF was observed in productive V-D-J junctions versus productively rearranged D-J joints in CLL (P < 0.001 for both comparisons).

Additionally, no difference was identified regarding the usage of hydrophobic RF versus hydrophilic RF versus RF with stop codon(s) (Table 1). A total of 10 of 63 cases with the IGHD gene read in an RF with stop codon(s) had lost the germline-encoded stop codon during IGHD-IGHJ rearrangement. Evidence for TdT, 5′- and 3′-exonuclease activities was identified in most cases (Table 2).

Table 2.

TdT and exonuclease activity in productively rearranged IGHD-IGHJ (D-J) versus IGHV-IGHD-IGHJ (V-D-J) in CLL and ALL.

	D-J in CLL	V-D-J in CLL	D-J in ALL	V-D-J in ALL	D-J versus V-D-J in CLL (P)	D-J versus V-D-J in ALL (P)	D-J in CLL versus ALL (P)
TdT activity
Positivity (%)	96.9	92.0	90.0	89.8	0.03	0.92	0.01
Median number of nucleotides (range)	9 (1–76)	7 (1–39)	8 (1–34)	6 (1–81)	0.01	0.96	0.13
3′ IGHD exonuclease activity
Positivity (%)	87.6	81.8	88.5	85.7	0.06	0.98	0.52
Median number of nucleotides (range)	4 (1–24)	4 (1–19)	4 (1–25)	5 (1–16)	0.74	0.12	0.26
5′ IGHJ exonuclease activity
Positivity (%)	86.6	86.6	85.4	77.6	0.99	0.08	0.84
Median number of nucleotides (range)	5 (1–27)	5 (1–29)	6 (1–25)	5 (1–16)	0.39	0.06	0.06

Sequence changes in the form of point mutations were detected in 24 of 198 D-J (12.1%): in particular, 22 cases carried a single mutation, whereas the remaining two cases carried two mutations. Of 24 cases with mutations in D-J, 20 carried somatically hypermutated IGHV genes in the corresponding V-D-J, whereas the remaining four cases carried coding IGHV genes with 100% identity to germline.

Comparative Assessment of Partial IGHD-IGHJ versus Productive IGHV-IGHD-IGHJ Rearrangements in CLL

For all subsequent analyses, we focused on the subgroup of 198 cases carrying a productively rearranged D-J on one allele along with a productive V-D-J on the other allele and performed a detailed comparative assessment of the D-J joints and V-D-J junctions. Regarding the molecular features of rearranged IGHD genes in V-D-J, a preferential usage of IGHD genes in RF2 versus RF1 or RF3 and hydrophilic RF versus hydrophobic RF or RF with stop codon(s) was recorded. Evidence for TdT, 5′- and 3′-exonuclease activities was found in most cases (Table 2). Comparison of the productively rearranged D-J joints versus productive V-D-J junctions revealed that, except for exonuclease activities, which were similar in both D-J and V-D-J, the D-J and V-D-J differed significantly.

In particular, TdT activity was recorded in a significantly higher proportion of D-J versus V-D-J (P = 0.03); additionally, a higher number of inserted nucleotides was found in D-J versus V-D-J (P = 0.01) (Table 2). Furthermore, certain genes were overrepresented (IGHD3-3, IGHD6-19, IGHD3-10) or underrepresented (IGHD2-2, IGHD2-15, IGHD1-26, IGHD4-23) in V-D-J versus D-J (P < 0.01 for all comparisons) (Figure 1). All genes overrepresented in V-D-J versus D-J exhibited biased usage of a single RF (IGHD3-3: RF2; IGHD3-10: RF2; IGHD6-19: RF3), whereas no such bias was observed in the (few) D-J using the same genes (P < 0.001 for all comparisons). Additionally, a significantly higher frequency (P < 0.001) of hydrophilic RF was recorded among V-D-J (Table 1).

The evident selection for the IGHD3-3 gene in productive V-D-J in CLL prompts a closer examination of its properties. As previously shown for productive rearrangements in CLL (32,33), this gene is preferentially used in RF2, thus encoding for a relatively long peptide sequence (10 amino acids: YYDFWSGYYT) with a predicted acidic isoelectric point value (3.8). Additionally, rearrangements carrying the IGHD3-3 gene are biased to the usage of the IGHJ6 gene (YYYYYGMDV) (32,33), leading to the formation of long VH CDR3 enriched in tyrosine residues (Figure 2). In principle, increased usage of any single amino acid restricts the overall VH CDR3 variability (34). However, tyrosine may be considered as an exception in that it allows for many forms of inter- and intramolecular bonding that enable aromatic stacking interactions and, thus, increase VH CDR3 flexibility and, likely, the range of antigenic epitopes that can be “accommodated” in the antigen-binding loop (34). Interestingly, usage of IGHD3-3 in RF2 among V-D-J, especially those using the IGHV1-69 gene, has been reported as an adverse prognostic indicator in CLL (32).

Figure 2.

Graphical representation of IGHD and IGHJ gene associations in CLL and ALL. (A) Partial IGHD-IGHJ (D-J) rearrangements in CLL. (B) Partial IGHD-IGHJ (D-J) rearrangements in ALL. (C) Productive IGHV-IGHD-IGHJ (V-D-J) rearrangements in CLL. The graphs were prepared by using the Circos software package (http://mkweb.bcgsc.ca/circos). Strong biases are evident in the case of (a) the IGHD2-2 gene, which is very frequently recombined to the IGHJ6 gene in D-J; (b) the IGHD3-3 gene, which is recurrently recombined to the IGHJ6 gene in V-D-J; and (c) the increased usage of the IGHD7-27 and IGHJ6 genes in D-J in ALL.

Molecular Configuration of Partial IGHD-IGHJ Rearrangements in ALL and Comparison to CLL

A total of 160 cases with B-cell lineage ALL were selected for analysis on the basis of carrying a partial IGHD-IGHJ rearrangement. Overall, within this cohort, 174 D-J and 74 V-D-J were successfully analyzed. Fourteen of 160 cases carried double D-J, whereas 9 cases carried double V-D-J. Fifteen D-J bore multiple IGHD genes in the form of IGHD-IGHD gene fusions. In 13 of 15 such junctions, two IGHD genes were identified, whereas the remaining two junctions bore three IGHD genes (Supplementary Table 7).

In 10 ALL cases, the same IGHD-IGHJ joint was identified in both their D-J and V-D-J, indicative of clonal evolution (Supplementary Table 8); four of these cases carried IGHD-IGHD gene fusions. Cases with evidence of clonal evolution were excluded from further repertoire analysis.

A total of 30 D-J carried a stop codon in the IGHD-IGHJ joint (including seven of the remaining 11 cases with IGHD-IGHD gene fusions). Among the remaining 134 productively rearranged D-J, IGHD2 subgroup genes predominated, with IGHD2-2 being the most frequent gene (36 of 134 D-J, 26.9%) followed by IGHD2-15 (16 of 134 D-J, 11.9%) (Figure 1; Supplementary Table 9). IGHD7-27, the most IGHJ-proximal gene (that is, of all IGHD genes the most proximal to the IGHJ gene cluster), was identified in 6 of 134 (4.5%) cases. IGHJ6 was by far the predominant gene (73 of 134 D-J, 54.5%); collectively, IGHJ4, IGHJ5 and IGHJ6 were found in 118 of 134 D-J (88.1%) (Supplementary Table 10). All three RFs of the IGHD genes were equally distributed among the productively rearranged D-J, and there was no difference in the usage of hydrophilic RF versus hydrophobic RF versus RF with stop codon(s) (Table 1). Five of 43 cases with the IGHD gene read in an RF with stop codon(s) had lost the germline-encoded stop codon during IGHD-IGHJ rearrangement. Evidence for TdT, 5′- and 3′-exonuclease activities was identified in most cases (Table 2). Single sequence changes in the form of point mutations were detected in 9 of 134 D-J (6.7%); in all cases, the mutation was located within the IGHD gene sequence.

A total of 53 of 134 cases (39.5%) carrying D-J on one IGH allele also carried a V-D-J on the second IGH allele; only 19 of 53 V-D-J (36%) were in-frame. In keeping with the literature (21), IGHV6-1, the most IGHJ-proximal IGHV gene (that is, of all IGHV genes the most proximal to the IGHJ gene cluster), was overrepresented, being used in 11 of 53 V-D-J (20.7%). Detailed information on IGHV, IGHD and IGHJ gene usage is given in Supplementary Tables 11–13. Mutations (1–3/case) were present within the IGHV gene in 8 of 53 V-D-J (15.1%), of which four carried out-of-frame junctions.

Comparison of V-D-J versus D-J in regard to the molecular characteristics of the junctions (preferred IGHD RF, TdT or exonuclease activities) revealed no significant differences. In contrast, IGHD3 subgroup genes predominated in V-D-J over D-J (18 versus 3 of 53 cases, P = 0.02, and 7 versus 2 of 19 cases with productive V-D-J, P = 0.05), whereas the opposite was observed for IGHD2 subgroup genes (18 versus 29 of 53 cases, P = 0.03, and 5 versus 12 of 19 cases with productive V-D-J, P = 0.03).

Comparison of the productively rearranged D-J in B-cell lineage ALL versus CLL demonstrated few significant differences: (a) preferential usage of the IGHD7-27 and IGHJ6 genes in ALL (P < 0.001) and (b) lower frequency of cases with TdT activity in D-J in ALL (P = 0.01) (Table 2).

DISCUSSION

CLL is distinctive for the existence of subsets of cases with identical or highly homologous (stereotyped) VH CDR3 sequences that account for up to 30% of the cohort (5,6). This finding strongly indicates a role for selection by antigen in shaping the expressed CLL IG repertoire (5,9). Although this is a reasonable conclusion, also supported by several lines of recent research (7–9), the possibility that CLL progenitors might exhibit inherently biased IGHD-IGHJ rearrangements, thus eventually accounting for restricted VH CDR3 motifs, has not been formally tested.

We addressed this issue from a novel perspective by assessing in parallel the molecular configuration of partial D-J versus productive V-D-J rearrangements in a cohort of 829 CLL cases. To gain better insight into the mechanisms that shape the IG repertoire and determine more accurately the developmental stage when VH CDR3 restrictions are imposed in CLL, we compared the CLL profiles to those observed in ALL, taking advantage of a large series of ALL cases carrying D-J rearrangements. This approach uniquely enabled to document significant differences between productive V-D-J and productively rearranged D-J in CLL, which, together with the few observed differences among D-J in CLL versus ALL, indicate that the expressed IG repertoire in CLL is not stochastically shaped, but rather seems to be directed by selection operating at the IG protein level.

Previous reports have demonstrated the existence of D-J in various B-cell malignancies (namely, ALL, multiple myeloma, hairy cell leukemia) at frequencies ranging from 22% to 66% (21–24). In all reported series, the number of D-J rearrangements eventually analyzed was relatively small (36–50 cases), precluding definitive conclusions, especially with regard to gene usage among different entities. Here, we document that partial IGHD-IGHJ rearrangements can be detected in roughly one-third of CLL cases. In addition, we provide a more comprehensive view on the molecular profile and gene repertoires of such rearrangements in B-cell malignancies by reporting results from the analysis of a combined dataset of 412 such rearrangements from CLL and ALL.

As in previous reports (21–24), we found that IGHD2 subgroup genes predominated in D-J in both CLL and ALL. At the individual IGHD gene level, we observed increased rearrangement frequency between 5′ genes of the IGHD cluster, for example, IGHD2-2, and 3′ genes of the IGHJ cluster, for example, IGHJ4, IGHJ5 and IGHJ6, suggestive of secondary rearrangements on the same allele (Figure 2). Comparative assessment of productively rearranged D-J in CLL versus ALL revealed an overall similar IGHD and IGHJ gene usage profile (with very few exceptions), thus indicating no significant disease-related biases regarding molecular events early in B-cell ontogeny. The presence of mutations in a minority of D-J could be plausibly attributed to a “bystander mutagenesis” effect, where nonexpressed rearrangements are mutated without selection for expression of a functional antigen receptor, as previously reported for either normal or neoplastic B cells (35–38).

Of great importance, the comparison of productively rearranged D-J joints versus productive V-D-J junctions in CLL revealed significantly different patterns regarding gene repertoires, IGHD RF use and hydropathicity profiles. Indeed, we documented a significant overrepresentation of IGHD3 subgroup genes in general and of the IGHD3-3 gene in particular among V-D-J versus D-J, alluding to selection at the amino acid level in the paratope. In contrast, IGHD2 subgroup genes, although frequent in both V-D-J and D-J, regardless their origin (that is, from either CLL or ALL), were still overrepresented in D-J. Given that the recombination signal (RS) sequences of IGHD2 and IGHD3 subgroup genes do not differ to a significant extent such that might account for their observed frequencies in D-J, alternative explanations must be sought for, for example, differential accessibility of different parts of the IGHD cluster to the VDJ recombination machinery, similar to what has been reported in the mouse (39).

Biased usage of particular IGHD genes in certain RF underlies the creation of highly restricted antigen-binding motifs in many stereotyped V-D-J in CLL (27,40). This bias evidently reflects selection, as also attested by our finding of significantly increased usage of RF encoding for hydrophilic motifs in V-D-J junctions versus D-J joints in CLL, alluding to solvent-exposed interactions with antigenic epitopes.

In both mice and men, the usage of hydrophilic VH CDR3 sequences seems to be promoted by positive selection, whereas a consecutive stretch of hydrophobic amino acids may be negatively selected on the basis of structural limitations imposed by the antigen- binding site (41,42). Therefore, not unexpectedly, RF usage in V-D-J of both species reflects this general principle.

A confounding factor when attempting to draw further analogies as well as when assigning RF within a locus arises from the different approaches adopted for defining RF in mice or men. In most previous mouse studies, the preferred (that is, hydrophilic) RF is designated as RF1, whereas a far more sensible procedure is followed in the human, first proposed by Kabat’s group, and adopted by IMGT, with RF1 counting from the first nucleotide of the D-REGION in germline configuration (that follows the 5′-HEPTAMER), thus defining three consecutively numbered “logical” RF (1,16,43). Hence, every IGHD gene has only one preferred RF, and this RF is not the same for all IGHD genes. Concerning hydrophilicity, the human IGHD genes encode for hydrophilic amino acids in different RFs, that is, RF2 for IGHD2, IGHD3 and IGHD4 subgroup genes; RF3 for IGHD1, IGHD5 and IGHD7 genes; and RF3 for IGHD6 subgroup genes. Our findings among V-D-J from CLL cases evaluated in this study are in keeping with this general trend; however, this is clearly not the case for D-J.

Taken together, the results presented here strongly indicate that sequence restrictions leading to VH CDR3 stereotypy in CLL reflect selective pressures exerted on the BcR rather than intrinsic features of the IGHD-IGHJ gene rearrangement process during the early ontogenetic stages in the life history of CLL progenitor cells.