B-cell receptor epitope recognition correlates with the clinical course of chronic lymphocytic leukemia

Authors

  • Mascha Binder MD,

    1. Department of Oncology and Hematology, Hubertus Wald Cancer Center, University Medical Center, Hamburg-Eppendorf, Hamburg, Germany
    Search for more papers by this author
  • Fabian Müller MD,

    1. Department of Oncology and Hematology, Hubertus Wald Cancer Center, University Medical Center, Hamburg-Eppendorf, Hamburg, Germany
    Search for more papers by this author
  • Antje Jackst,

    1. Department of Oncology and Hematology, Hubertus Wald Cancer Center, University Medical Center, Hamburg-Eppendorf, Hamburg, Germany
    Search for more papers by this author
  • Barbara Léchenne,

    1. Department of Oncology and Hematology, Hubertus Wald Cancer Center, University Medical Center, Hamburg-Eppendorf, Hamburg, Germany
    Search for more papers by this author
  • Milena Pantic PhD,

    1. Department of Hematology and Oncology, University of Freiburg Medical Center, Freiburg, Germany
    Search for more papers by this author
  • Ulrike Bacher MD, PhD,

    1. Department of Oncology and Hematology, Hubertus Wald Cancer Center, University Medical Center, Hamburg-Eppendorf, Hamburg, Germany
    Search for more papers by this author
  • Christine zu Eulenburg PhD,

    1. Medical Biometry and Epidemiology, University Medical Center Hamburg-Eppendorf, Hamburg, Germany
    Search for more papers by this author
  • Hendrik Veelken MD, PhD,

    1. Department of Hematology and Oncology, University of Freiburg Medical Center, Freiburg, Germany
    Search for more papers by this author
  • Roland Mertelsmann MD, PhD,

    1. Department of Hematology and Oncology, University of Freiburg Medical Center, Freiburg, Germany
    Search for more papers by this author
  • Renata Pasqualini PhD,

    1. David H. Koch Center, The University of Texas M. D. Anderson Cancer Center, Houston, Texas
    Search for more papers by this author
  • Wadih Arap MD, PhD,

    1. David H. Koch Center, The University of Texas M. D. Anderson Cancer Center, Houston, Texas
    Search for more papers by this author
  • Martin Trepel MD, PhD

    Corresponding author
    1. Department of Oncology and Hematology, Hubertus Wald Cancer Center, University Medical Center, Hamburg-Eppendorf, Hamburg, Germany
    2. Department of Hematology and Oncology, University of Freiburg Medical Center, Freiburg, Germany
    • Department of Oncology and Hematology, University Medical Center Hamburg-Eppendorf, Martinistrasse 52, D-20246 Hamburg, Germany
    Search for more papers by this author
    • Fax: (011) 49-40-7410-57187


Abstract

BACKGROUND:

B-cell receptors (BCRs) and their recognition of specific epitopes may play a pivotal role in the development and progression of chronic lymphocytic leukemia (CLL). In this study, the authors set up a model system to explore epitope reactivity and its clinical relevance in CLL.

METHODS:

Epitope-mimicking peptides were selected from phage display libraries on 6 CLL BCRs from randomly chosen patients. The binding of the 6 index epitope mimics was evaluated in a set of 100 unrelated CLL samples. Epitope recognition patterns were correlated with the clinical course of the disease.

RESULTS:

Surprisingly, all CLL samples recognized 1 or several index epitopes, and some revealed marked polyreactivity. Patients with CLL who expressed BCRs that reacted with ≥5 epitope mimics had a significantly worse clinical course than less polyreactive patients (median time to first treatment, 24 months vs 102 months). This effect was independent of otherwise known prognostic markers.

CONCLUSIONS:

The authors introduced a system with which to model epitope reactivity of CLL BCRs without previous knowledge of potential antigens. The findings indicated that a polyreactive epitope recognition pattern may be a determinant of an aggressive clinical course in this disease. This further emphasizes the functional and prognostic relevance of BCR epitope recognition in CLL. Cancer 2011. © 2010 American Cancer Society.

Chronic lymphocytic leukemia (CLL) most commonly derives from the B-cell lineage. CLL cells express membrane immunoglobulin, the B-cell receptor (BCR), which is specific for the leukemic clone of each individual patient.1 Antigen selection through the BCR may be important in the pathogenesis and/or progression of CLL2-8: Unlike normal B cells, CLL BCRs use a restricted set of immunoglobulin variable heavy-chain genes (IGHV).9-11 In addition, >25% of patients with CLL display stereotyped BCRs that belong to 1 of approximately 150 different subsets with homologous receptors.2, 4-6, 10, 12-16 Such findings support the hypothesis that defined clones of B lymphocytes in CLL are selected by a set of recurring, disease-specific antigenic epitopes. In fact, binding studies with CLL BCRs have revealed that some of these cross-react with autoantigens, microbial antigens, and antigens that are exposed during apoptosis.17-20 In fact, the activation of BCR signaling by surface immunoglobulin M (IgM) cross linking as a surrogate for antigen binding promotes CLL cell survival.21, 22 Taken together, these data suggest that the functional status of the immunoglobulin and its interaction with antigenic epitopes may play a role in CLL pathogenesis and progression.

Despite identification of the binding characteristics of exemplary CLL BCRs,17, 19, 23-25 the overall functional diversity of epitope recognition in CLL remains elusive. The objective of the current study was the systematic modeling of the epitope recognition of CLL BCRs without prior knowledge on the nature of potential antigens. This should allow us to address 1) how diverse epitope recognition in CLL really is, 2) whether there are general differences in BCR reactivity between stereotyped and nonstereotyped as well as mutated and unmutated receptor configurations, and 3) whether certain BCR reactivity profiles correlate with the clinical course of the disease.

Therefore, we selected phage-displayed peptides from combinatorial libraries25-33 from a small group of CLL-derived BCRs and investigated the binding of these epitope mimics to the leukemia cells in a large set of mostly unrelated CLL cases. Unexpectedly, all of the studied CLL samples recognized 1 or several of the selected epitope mimics and could be clustered into a small number of distinct binding patterns. Polyreactive epitope binding patterns clearly correlated with an aggressive clinical course of the disease. Our results indicate that epitope recognition in this very common leukemia is surprisingly uniform and that the epitope binding pattern of an individual patient with CLL may be functionally and clinically relevant.

MATERIALS AND METHODS

Patients and Sample Characterization

Blood samples from 107 random patients with CLL who visited outpatient clinics at the Freiburg and Hamburg medical centers were collected between 2003 and 2008 after informed consent, as approved by the institutional review board of the University of Freiburg and the Physician';s Association in Hamburg. Cells were purified by Ficoll separation. The CLL BCRs were characterized by flow cytometry, immunoglobulin sequencing, and determination of mutational status.34 Control cells were obtained from healthy donors and were affinity-purified on CD19-antibodies (Miltenyi Biotec, Bergisch Gladbach, Germany).

Recombinant Expression of CLL-BCR

Combinational DNA from CLL cells was prepared and fragment antigen-binding (Fab) fragments were produced commercially (CellGenix, Freiburg, Germany), as described previously.35, 36 Fab fragments were purified by His-tag affinity chromatography. The BCR of Patient 003 with an artificial heavy chain complementarity-determining region (CDR) 3 (HCDR3) region was cloned into the plasmid pBUD, expressed as IgG1 antibody in HEK293T cells, and purified by protein-A affinity chromatography.37

Random Phage Display Peptide Libraries

Five random phage display peptide libraries were used: 1 cyclic (CX7C), 3 linear (X7, X12, X18), and 1 beta-sheet (BS) conformation (X4CX6CX4; where C indicates cysteine, and X indicates a random amino acid). The X12 library was purchased from New England Biolabs (Ipswich, Mass) and processed as recommended by the manufacturer. All other libraries were generated as described previously.26, 38-43

Phage library selection

The libraries were screened on BCR Fab fragments as described previously27, 28 after a 2-fold negative selection on polyclonal human immunoglobulins (Octapharma, Lachen, Switzerland). Random clones from the third or fourth panning round were sequenced (GATC).

Single clone phage binding

Phage clones (107 transducing units [TU]) were incubated on BCRs. Bound phages were quantified based on Escherichia coli K91kan or ER2738 (for X12 phage) infection.27, 28, 38 For phage binding to BCRs on CLL cells, we incubated 106 cells with 107 TU phage and control phage (insertless fd-tet or control phage with random inserts: CX7C [CERIYHFVC], X7 [RQKMPLL], X12 [YMTPPLSSQQKS], X18 [RHLTLTTHSLGISAYMAS], and BS-library [PAGTCVGHLEDCYGER]). For binding to the large panel of CLL cells, phage input was increased to 108 TU to ensure detection of weak binding. Otherwise, the assays were carried out as described reviously.44 The HEK293T, HeLa, Raji, and Daudi cell lines and CD19-sorted peripheral lymphocytes from healthy donors were used as controls.

Glutathione S-transferase fusion proteins and competition assays

The oligonucleotide encoding the phage-derived peptide was digested with BamHI and EcoRI, and cloned into pGEX-2TK (Amersham Biosciences, Piscataway, NJ). Proteins were expressed in E. coli BL21 (Invitrogen, Carlsbad, Calif) and purified according to the manufacturer';s instructions. Competition experiments were done with glutathione S-transferase (GST) fusion proteins as described previously.27

Immunofluorescence

CLL cells were incubated for 5 minutes with a 10-fold excess of phage, washed with phosphate-buffered saline, sedimented on adhesion slides (Marienfeld, Lauda-Koenigshofen, Germany), fixed with 4% formaldehyde, permeabilized with 0.05% Triton X-100, and blocked with 3% bovine serum albumin. Bound phages were stained with a polyclonal anti-fd bacteriophage antibody (Sigma, St. Louis, Mo) and a secondary Alexa-Fluor 488-labeled goat antirabbit antibody (Invitrogen). Membranes and cytoplasm were counterstained with Alexa Fluor 594 phalloidin (Invitrogen), and nuclei were stained with 4′,6-diamidino-2-phenylindole (VectaShield; Vector Laboratories, Burlingame, Calif). Images were obtained by confocal microscopy (Leica TCS SP2; Leica Microsystems, Nussloch, Germany).

Biostatistics

Time-to-first-treatment (TTFT) curves were plotted using the Kaplan-Meier method. Univariate differences between categories concerning the distribution of the treatment events were studied with log-rank tests. For multivariate analysis, Cox regression analysis was performed to adjust for multiple independent factors. To analyze the correlation between high-risk receptors and patients in polyreactive binding category II, chi-square tests were used. All statistical tests were performed using SPSS software (version 15; SPSS, Inc., Chicago, Ill), and P values <.05 were considered statistically significant.

RESULTS

Identification of CLL BCR Binding Peptides

BCRs from 8 randomly chosen patients with CLL were expressed as Fab fragments, termed “Fab” followed by a patient-specific number. Three of the BCRs belonged to so-called “subsets” with stereotyped CDR3 regions5, 13, 14 (Table 1).

Table 1. Characteristics of Chronic Lymphocytic Leukemia B-Cell Receptors That Were Cloned as Fragment Antigen-Binding (Fab)
Patient CodeIsotypeVH GeneVL GeneHCDR3 SequenceLCDR3 SequenceMutation Status VH (%)a
  • VH indicates variable heavy chain; VL, variable light chain; HCDR3, heavy chain complementarity-determining region 3; Fab, fragment antigen-binding fragment; LCDR3, light chain complementarity-determining region 3; M, mutated; UM, unmutated.

  • a

    Pecentages in parentheses indicate the proportion of germline sequence.

  • b

    Fab001 belongs to B-cell receptor (BCR) Subset 2 according to Stamatopoulos 20075 and Murray 2008.12

  • c

    Fab005 belongs to BCR Subset 30 according to Stamatopoulos 20075 and Murray 2008.12

  • d

    Fab006 belongs to BCR Subset 58 according to Murray 2008.12

Fab001bμ, λ3-213-21ALDRDGMDVQVWDSSSDHPWVM (97.1)
Fab002μ, λ3-483-21ARVGGSYYSDYQVWDSSSDHPWVM (97.6)
Fab003γ, κ3-301-33AGVVEMATIGGVFSGMDVQQYDNLPVTUM (100)
Fab004μ, κ4-44-1ARMYGGYAAYYYYGMDVQQYYSTPWTUM (100)
Fab005cμ, κ3-91-NL1AKDRYDFWSGYPNRSPFDYQQYYSTLGTUM (100)
Fab006dμ, κ1-582-28AAGYDFWSGYYPMQALQTPSTUM (100)
Fab007μ, λ4-613-21ARMGGGEVDYQVWDGSSDPWVUM (99)
Fab008μ, κ2-53-20AYAPVAVTAFHNFFNPQQFDDSPPMYTM (96.6)

We identified peptides binding to these recombinant CLL BCRs as epitope mimics by the selection of random phage display peptide libraries. Assuming that 1 type of peptide design structurally might mimic only a limited set of epitopes that were recognized by CLL BCRs, we used a variety of random libraries with different peptide lengths (X7, X12, X18) and constraints (CX7C, X4CX6CX4 = BS conformation). By using this broad approach, the selection was successful on 6 of 8 Fab fragments with at least 1 library. Table 2 lists the Fab fragments and the peptide sequences that were selected on each. All of the selected phages, but not the random insert control phage, bound specifically to the Fab fragment on which they were selected and not to control immunoglobulins (Fig. 1A).

Figure 1.

Phages selected on chronic lymphocytic leukemia (CLL) fragment antigen-binding (Fab) fragments bind their target specifically. (A) Phages selected on CLL Fab fragments specifically bind the idiotype on which they were selected. For example, the binding of phage CRWWFGQFC and random control phage CERIYHFVC to Fab fragments from Patient 003 (Fab003) and to control polyclonal immunoglobulin (IgG) is shown. Bound phages were quantified as the numbers of transducing units (TU) based on bacterial infection (values shown are the means from triplicate platings ± standard errors of the mean [SEM]). (B) The peptide RWWWG blocks binding of phage CRWWFGQFC to Fab003. Phage CRWWFGQFC selected on Fab003 was incubated on Fab003 in the presence of either glutathione S-transferase (GST)-RWWWG or GST protein alone. Bound phages were quantified as described for A. Data are shown as relative values compared with binding in the absence of GST and GST-RWWWG (values shown are the means from triplicate platings ± SEM). (C) Phages TPEKWHRLLTMS and RWWWPTR compete for their binding site on Fab003. Phage TPEKWHRLLTMS selected on Fab003 was incubated on Fab003 in the presence of phage RWWWPTR or insertless fd-tet phage. Bound phages were quantified as the numbers of TU based on ER2738 bacterial infection, which only detects TPEKWHRLLTMS phage but not the blocking phage RWWWPTR or fd-tet. Data are shown as relative values compared with binding in the absence of blocking phage (means from triplicate platings ± SEM).

Table 2. Inserts of Phages Selected on Chronic Lymphocytic Leukemia Fragment Antigen-Binding (Fab) Fragments Using 5 Libraries With Different Insert Lengths and Constraintsa
FabLibrary X7Library CX7CLibrary X12Library X18Library BS
  • Fab indicates fragment antigen-binding fragment.

  • a

    Libraries with 7-mer constrained (CX7C), 7-mer linear (X7), 12-mer linear (X12), 18-mer linear (X18), and 14-mer constrained beta sheet (BS) inserts were selected on chronic lymphocytic leukemia Fab fragments 001 through 007. The table lists the peptide insert sequences (single-letter amino acid code) of phage clones for which specific binding to the respective Fab fragment was validated in single clone-binding assays.

001  WNWPLPPVRQFS  
   WPWPLPPEPPLG  
   SWYWPLPPWRLG  
   SWPWYHPHIKSH  
003RWWWPTRCRWWSGQPCTPEKWHRLLTMS SYWPCHPGTRHCSNRV
 RWWWLPRCRWWWQDTCATPWSQWLDAPR  
 RWWWFPRCRWWMGVSC   
 KWWMASRCRWWNGSWC   
 RWWWDPFCRWWMSYTC   
 RWWWSFSCRWWRGQGC   
 VKWWWSACRWWFGQFC   
 IWKWWWKCRWWRGEVC   
  CRWWLGSAC   
  CRWWAGSTC   
  CRWWWVETC   
  CKWWGGRGC   
004FCSFCVL EMSVDWWSPISS PVLVCGPKWSNCSPAN
 FCSDCIL ASSVDWWPVRPP SPSLCWPWLEQCTEGI
 FCGDCVL FMWPDQNPRHSM GDQPCPIFDRECHKPT
     GIGPCELIDSECEASE
005    YHRWCVMDRRACFEAP
     NLSECAMPTRKCSRTA
006   QLLFPSSPRAPAPWTFTFLEYACNGPSQLCSYVR
     LPWACPFTGWFCDLIT
     TEKTCAGPGDLCLLVR
     AQRKCAGNWAICRLVY
007   YYCYFTEAPYSYWGNLVC 
    SQSPPRYWAWCAGYWCEL 

To confirm that the phage displayed peptides bound the respective BCR independent of other phage components, we exemplarily performed competition assays with a GST-fusion protein that contained the phage-derived peptide RWWWG, which inhibited the binding of phage CRWWFGQFC to Fab003 (Fig. 1B). Moreover, phages from different libraries selected on the same Fab competed with each other in binding to the paratope of that Fab, although they did not share a common sequence motif (Fig. 1C), suggesting that, structurally, they mimicked the same epitope.

We explored whether binding of the epitope mimics to the BCR is mediated by the HCDR3 or by other regions of the BCR. Thus, as proof of principle, we replaced the HCDR3 region of BCR from Patient 003 with an unrelated amino acid sequence of similar length. The epitope-mimicking phage that was selected on the natural BCR from Patient 003 bound specifically to the BCR of this patient but not to the BCR with a modified HCDR3 region, suggesting that the epitope mimic indeed binds to the HCDR3 region and not to other parts of the BCR. In addition, we cloned and expressed a BCR from another patient with CLL who belonged to the same subset with a stereotyped CDR3 region5, 13, 14 with IGHV/Ig variable light chain (IGLV) genes 3 through 21 (IGHV3-21) using Fab 001. In this instance, as expected, we enriched the very same peptide sequence motif, WNWPLPPXXXXG, by library screening, confirming the assumption that homologous BCRs would bind the same epitopes.

Epitope-Mimicking Phages Bind to the Corresponding BCR-Expressing CLL Cells

Epitope-mimicking phage clones that were selected on recombinant Fab fragments bound the corresponding parental CLL cells 30 to 100 times more than the binding observed with the control phage (exemplified in Fig. 2A). Specific phage binding also was observed by immunofluorescence (exemplified for CLL Patient 003 and the selected phage TPEKWHRLLTMS in Fig. 2B). Therefore, we concluded that phage-displayed peptides that were selected on recombinant BCR Fab fragments also bound equivalent immunoglobulin idiotypes on CLL cells.

Figure 2.

B-cell receptor (BCR) epitope mimics bind parental chronic lymphocytic leukemia (CLL) cells. (A) Phages selected on CLL fragment antigen-binding (Fab) fragments specifically bind native BCR from parental CLL cells. Phage RWWWPTR and control phage fd-tet were incubated with CLL cells from Patient 003, control CLL cells from Patient 026, and peripheral blood mononuclear cells (PBMC) from a healthy donor. Bound phages were quantified after differential centrifugation as described for Figure 1 (values shown are the means from triplicate platings ± standard errors of the mean [SEM]). (B) TPEKWHRLLTMS phage binding to CLL cells from Patient 003 by immunofluorescence is shown. CLL cells from Patient 003 or from control Patient 026 were incubated with phage TPEKWHRLLTMS and with random control YMTPPLSSQQKS phage. Phages were observed by immunofluorescence (fluorescein isothiocyanate; green). Cell membranes and cytoplasm were counterstained with Alexa Fluor 594 phalloidin (red), and nuclei were stained with 4′,6-diamidino-2-phenylindole (blue). Images are shown (Left) at low magnification and (Right) at high magnification.

Epitope Recognition Pattern in a Large CLL Patient Cohort

CLL BCRs may react with several antigens.17, 19, 23-25 Yet, the overall functional diversity of epitope recognition in CLL remains elusive. Therefore, we systematically modeled the epitope recognition of 100 random CLL cases by using 6 of the index epitope mimics that were selected on 6 random CLL BCRs, as described above. The vast majority of CLL samples in this experiment were unrelated to the 6 samples that were used for selection of the epitope mimics. CLL Patients 001 and 003, however, were included in the panel as positive, proof-of-principle controls. The binding patterns subsequently were considered in relation to the CLL BCR antigen recognition sites of each sample (ie, IGHV gene use, status of somatic hypermutation, and CDR3 regions of each patient';s BCR).

Of the 100 patients who were included in this screening, approximately 66% expressed mutated IGHV genes, and 33% expressed unmutated IGHV genes. Approximately 23% of the patients with CLL were included in previously described subsets of stereotyped BCRs, consistent with the frequencies of stereotyped receptors reported recently.5, 12, 13

To assess the level of phage binding, we differentiated in a semiquantitative manner between strong phage binding (defined as consistently >5-fold higher than control phage binding), moderate binding (defined as consistently 2-fold to 5-fold higher than control phage binding), and no binding. Compared with the phage binding assays described above (eg, Fig. 2), the phage input for this large cohort screening assay was increased from 107 TU to 108 TU per 106 cells to ensure detection of moderate binding. To control for unspecific cross reactivity under these less stringent binding conditions, numerous control cell samples were included as specified below.

We reasoned that, if each CLL clone recognized a single and unique epitope, then we would expect to observe no epitope binding in most CLL samples in the panel, because almost none of the samples were used for combinatorial selection of the epitope mimics. Yet, unexpectedly, all CLL samples bound to at least 1 epitope mimic, and the majority of them exhibited some degree of multireactivity (Fig. 3). In contrast, the epitope mimics did not reveal binding to CD19-sorted mononuclear blood cells from healthy donors (2 samples), B-cell lymphoma cells expressing unrelated immunoglobulins (the Raji and Daudi cell lines), or to cell lines that did not express immunoglobulin (the 293T and HeLa cell lines), indicating specificity of the binding. Epitope-mimicking phage WNWPLPPVRQFS reacted to a various degree with all CLL samples irrespective of IGHV gene use or mutational status but did not bind to the control samples. All binding patterns were reproducible in independent experiments.

We observed that 90% of CLL samples could be assigned to 1 of 17 distinct categories of epitope binding, termed I through XVII (Fig. 3), which ranged from monoreactive patterns to oligoreactive and polyreactive patterns. CLL samples that belonged to so-called stereotyped subsets (see above) had similar or identical binding patterns, suggesting that homologous BCRs recognize the same epitopes (Fig. 3). However, for the vast majority of samples, the binding behavior could not be deduced from their IGHV gene use or their CDR3 amino acid sequence, because most of the epitope binding categories contained CLL samples with seemingly heterogeneous (but functionally obviously homogenous) BCRs.

Figure 3.

Epitope recognition pattern in a panel of 100 chronic lymphocytic leukemia (CLL) cases. CLL cells were purified from the peripheral blood of 100 randomly chosen patients. Epitope-mimicking phage clones that were selected on fragment antigen-binding (Fab) fragments from Patient 001 (Fab001) (WNWPLPPVRQFS), Fab003 (RWWWPTR), Fab004 (FCSFCVL), Fab005 (YHRWCVMDRRACFEAP), Fab006 (LEYACNGPSQLCSYVR), Fab007 (SQSPPRYWAWCAGYWCEL), control phage with random insert (YMTPPLSSQQKS), and control phage were incubated on CLL samples and control cells (HEK293T, HeLa, Raji, and Daudi cells and CD19-sorted peripheral mononuclear cells [PBMC1 and PBMC2] from healthy donors). Bound phages were separated from unbound phages by differential centrifugation and were quantified based on bacterial infection. Dark green indicates strong phage binding; light green, moderate phage binding; red, no phage binding. Samples are grouped into 17 categories, indicated by Roman numerals I through XVII, according to their epitope-mimicking binding pattern. Mutational status (MS), IGHV gene usage, subset (SS) according to Stamatopoulos et al. (2007), Murray et al. (2008) and Bomben et al. (2009) if applicable; HCDR3 sequences are shown on the right side.

Correlation of BCR Reactivity With the Clinical Course of CLL

Sustained or repetitive BCR signaling promotes survival in CLL cells.21, 22 Because BCRs reacting with multiple epitopes may be more prone to sustained signaling, we analyzed the clinical course of patients with CLL who had such polyreactive BCRs.

We used the TTFT as a surrogate marker of disease progression, as is customary for patients with CLL. A global analysis of all statistically evaluable epitope binding categories (all categories that included >6 patients) revealed significant differences in TTFT. In a univariate analysis, patients with CLL in the most polyreactive binding category (Category II) had a shorter TTFT than patients in the control group with other epitope binding patterns (log-rank test; P = .014) (Fig. 4A). The median TTFT was 27 months in Category II compared with 87 months in the control group. This effect remained if we adjusted for the effect of mutation status (Cox regression; P = .035). This suggests that the more rapid course of patients in Category II was independent of the mutation status of their BCR, a parameter known to affect prognosis. It is noteworthy that the majority of patients with CLL who had “high-risk” receptor configurations (CDR3 subsets 1, 2, and 3 according to Stamatopoulos et al, Murray et al, and Bomben et al5, 12, 13; and IGHV3-21 gene use45) exhibited a polyreactive epitope binding pattern and, thus, accumulated in Binding Category II (chi-square test; P<.001). We reasoned that the over representation of “high-risk” receptors in Epitope Binding Category II could account for the short TTFT in this group. Consistent with previous publications, this subgroup of patients tended to have a shorter TTFT compared with patients in the control group, but this trend was not statistically significant. However, it is noteworthy that the shorter TTFT for patients in Category II could not be attributed to the over representation of patients with “high-risk” receptor configurations in this category. Therefore, short TTFT appears to be an independent feature of the epitope binding pattern. In a multivariate analysis, the TTFT of patients in Category II remained significantly shorter compared with the TTFT of control patients despite adjustment for the effects of mutational status and “high-risk” receptor configurations (Cox regression; P = .009) (Table 3).

Figure 4.

These Kaplan-Meier curves illustrate the clinical significance of the epitope binding pattern in patients with chronic lymphocytic leukemia (CLL). Kaplan-Meier curves of the time to first treatment (TTFT) as a surrogate marker for disease progression were calculated according to the assignment of CLL patient populations to (A) Polyreactive Epitope Binding Category II (dashed line) and the control group (ie, not Category II) or (B) multireactive categories that bound ≥5 epitopes (dashed line) and categories that bound ≤4 epitopes (solid line).

Table 3. Multivariate Analysis of the Time to First Treatment in Subgroups of Patients With Chronic Lymphocytic Leukemia
VariableHRP
  1. HR indicates hazard ratio; BCR, B-cell receptor.

Polyreactive BCR category II vs other categories4.317.009
Mutated vs unmutated BCR2.515.001
“High-risk” BCR configuration vs other configurations0.379.120

The correlation between polyreactivity of the BCR and an aggressive clinical disease course was even more pronounced when patients with CLL had samples that recognized ≥5 epitopes (“multireactive BCR categories”) were compared with patients with CLL who had samples that recognized only ≤4 epitopes. The median TTFT in the multireactive categories was 24 months, compared with 102 months in the remaining categories (log-rank test; P = .001) (Fig. 3B). Again, the negative effect of epitope multireactivity on TTFT could not be attributed to an over representation of samples with “high-risk” receptor configurations in this patient subgroup (Cox regression; P < .001).

DISCUSSION

Epitope recognition may play a crucial role in the pathogenesis of lymphatic malignancies like CLL. In this report, we introduce a practical tool with which to study CLL BCR epitope reactivity that does not require previous knowledge about the nature of potential antigens. A similar technical approach was used previously, suggesting that epitope mimics that bind to mutated CLL immunoglobulins differ from mimics that bind to unmutated immunoglobulins.25 We demonstrated that BCR epitope interaction can be profiled systematically and follows distinct monoreactive, oligoreactive, and polyreactive patterns. This finding was unexpected, because lymphatic malignancies long have been perceived as derivatives of a B-cell clone that randomly has acquired transforming genetic aberrations. Therefore, it seemed reasonable not to expect any interindividual functional similarity of BCRs among patients with CLL. To some extent, this view has changed, because immunoglobulin gene use in CLL is biased, and several highly similar CDR3 regions are expressed.2, 5, 6, 10, 12, 16 Almost 27% of CLLs express stereotyped BCRs that belong to 1 of almost 150 different subsets with homologous BCRs.2, 5, 6, 10, 12, 14-16 Nevertheless, based on these numbers, the likelihood that >2 patients will exhibit the same epitope recognition pattern remains considerably below 5%. Thus, our functional profiling of CLL BCR epitope recognition reveals a surprising functional homogeneity that was not predictable from the amino acid sequence of BCR CDR3 regions.

The 6 randomly chosen patients whose epitope mimics were used for screening the cohort of 100 patients certainly cannot be representative of the overall population of patients with CLL. The small sample size leads to biases in the distribution of stereotypy and mutational status of the receptors from these 6 CLL patients. However, this apparent limitation does not influence the remarkable finding of the limited number of reactivity profiles, because it is related to the sample size of 100 and not to the sample size of 6.

Exploring the clinical significance of epitope binding patterns, we observed that the polyreactive binding pattern of Category II or of any category that bound to at least 5 of 6 investigated epitopes was correlated with rapid disease progression. It is interesting to note that the effect of the binding pattern was statistically independent of the effect of established prognostic parameters, such as mutation status and certain BCR configurations. Therefore, epitope binding patterns as described here could be evaluated as independent prognostic parameters in CLL in larger patient cohorts.

In a functional context, the correlation between epitope recognition and clinical disease course provides an additional basis for recent concepts in CLL biology. Sustained or repetitive BCR signaling promotes survival in CLL cells.21, 22 The signaling capacity of BCRs may be determined by functional properties of the receptor itself46-49 and by the availability of external antigenic stimulation. Consequently, BCRs that react with various epitopes may be more prone to sustained signaling. This could explain our finding that CLLs expressing multireactive BCRs have a more aggressive course than CLLs expressing less reactive BCRs. The demonstration that some CLL cells are anergic to BCR binding or cross linking (see above) does not necessarily weaken this point, because the cells may have become independent of BCR signaling during the course of the disease despite its potentially important role in earlier stages of the disease.

In summary, our current results demonstrate that the epitope recognition of CLL BCRs is very homogenous and is insufficiently predictable using the amino acid sequence of their paratopes, because interactions on a structural basis apparently account for many antibody-epitope contacts. Correlations with clinical data highly suggest that a polyreactive epitope binding pattern could determine an aggressive clinical disease course in patients with CLL.

Acknowledgements

We thank Dr. George Smith for the fUSE5 plasmid and Gerald Illerhaus for helpful discussions.

CONFLICT OF INTEREST DISCLOSURES

This work was funded by the Josea Carreras Leukemia-Foundation (M.T.), German Cancer Aid (Deutsche Krebshilfe; M.T.), and the University of Texas M. D. Anderson Cancer Center Specialized Program of Research Excellence (SPORE) in Leukemia (W.A. and R.P.).

Ancillary