A new perspective: molecular motifs on oxidized LDL, apoptotic cells, and bacteria are targets for chronic lymphocytic leukemia antibodies

B-cell chronic lymphocytic leukemia (CLL) cells show gene expression profiles similar to memory B cells¹  as well as a surface membrane phenotype reminiscent of activated and antigen-experienced B cells.²  CLL can be divided into subgroups with either unmutated or mutated Ig heavy-chain variable (IGHV) genes, where unmutated IGHV (IGHV_UM) genes are associated with a more aggressive clinical course than mutated IGHV (IGHV_M) genes.^3,4  Biased use of certain IGHV genes, foremost IGHV1-69, IGHV3-07, IGHV3-21, and IGHV4-34, has been reported.^4-6  Recently, more than 110 subgroups of CLL have been identified that showed remarkably similar IGHV and Ig light-chain variable (IGLV) gene rearrangements with homologous complementarity determining region 3 (CDR3) amino acid sequences among members of the respective groups.^7,8  These findings of “stereotyped” B-cell antigen receptors (BCRs) in CLL subsets imply a role for specific antigens in leukemogenesis.^9,10 

Which antigens do the CLL-derived Abs then bind to? The fact that CLL patients often present with autoimmune phenomena including autoimmune hemolytic anemia and autoimmune thrombocytopenia indicates an immune dysregulation associated with the malignancy. However, antierythrocyte, antiplatelet, anti–double-stranded DNA, and anti-ia Abs found in these patients are polyclonal and, by definition, not produced by the malignant clone. Interestingly, recent studies have shown that several monoclonal recombinant CLL-derived Abs, each representing the malignant clone in subsets with stereotyped BCRs, bind to cytoplasmic and nuclear autoantigens.¹¹ 

Previous attempts to isolate Ig from malignant CLL clones have been hampered by minute amounts of Ig secreted in vitro, contaminations of a small number of normal nonmalignant B cells, and transient short survival times in vitro, impeding a correct verification of clonal origin. Recombinant Abs (often monomeric) from CLL clones have been successfully used,¹¹  but the difference between recombinant Ab and native pentameric IgM may present drawbacks regarding conclusion based on avidity/affinity of the mAb.

The aim of the present study was to characterize in detail the antigens in CLL, first using cell lines¹²  derived from the neoplastic CLL clone by Epstein-Barr virus (EBV) transformation,¹³  and second using primary ex vivo CLL cultures. Unexpectedly, the specificities of this panel of 28 CLL Abs showed limited target structure recognition on apoptotic cells and bacteria, and we identify here several previously unpublished (auto)antigens. Our findings may give new insight into the origin and clonal expansion of these leukemic B cells.

EBV-transformed CLL cell lines representing the malignant clone were previously established in our laboratory and by colleagues (Table 1). Their CLL origin was reconfirmed by one or several of the following methods: karyotyping, fluorescent in situ hybridization (FISH) analysis, fluorescence-activated cell sorting (FACS) phenotype, Ig gene sequencing, and microsatellite analysis, verifying that the cells were neoplastic clones (A.L.M, E.H., A.R. et al, unpublished data, November 2007). A recombinant monovalent IGHV3-21_M mAb (rIGHV3-21_M/subset-2) expressing IGH/IGL sequences¹⁴  of the stereotyped IGHV3-21 subset 2 was also included in this study (for details, see Document S1, available on the Blood website; see the Supplemental Materials link at the top of the online article). Primary CLL cultures were established from peripheral blood mononuclear cells (PBMCs) of CLL patients consecutively attending the Hematology Clinic, Linköping University Hospital, after informed consent, under the guidelines from the Linköping University Hospital ethics committee approval D.no.02-459, in compliance with the Declaration of Helsinki. The CLL cell lines and cultures, FM55_M2 human melanoma, HepG2 hepatoma (no. 85011430; ECACC, Salisbury, United Kingdom) and Jurkat T cell line (no. 88042803; ATCC, Manassas, VA) were cultivated in RPMI 1640 medium (Invitrogen, Paisley, United Kingdom) with 10% FCS (InVitrogen) with additives. Rat smooth muscle cells (SMCs) were kindly provided by Dr Hans Arnqvist (Linköping, Sweden). Human fibroblasts, Ag1523, were cultivated in MEM (Invitrogen) obtained from Coriell Institute for Medical Research (Camden, NJ). Monocyte derived macrophages were prepared from PBMCs from healthy volunteers by Ficoll-Hypaque (Pharmacia Biotech AB, Uppsala, Sweden). Plastic adherence-purified monocytes were cultured in DMEM (Invitrogen) with 20% human AB⁺ serum. The rIGHV3-21_M Ab was produced in CHO cells grown in 20% RPMI 1640 + 80% CHO-S-SFMII with 2% FBS (Invitrogen).

Tissue microarray slides (SuperBioChips Laboratories, Seoul, South Korea) containing paraffin-embedded human tissue samples (n = 59) were used for antigen-specificity screening according to the manufacturer's protocol. IgM from CLL cell lines was used with biotinylated goat anti–human IgM (Sigma-Aldrich, St Louis, MO) and streptavidin-HRP conjugate (Zymed Labs, San Francisco, CA). The slides were developed with DAB substrate and counterstained in Mayer hematoxylin solution (Histolab, Göteborg, Sweden). Formalin-fixed, paraffin-embedded sections from human tonsils were used for confirmatory immunohistochemical analysis.

CLL IgMs were used for detection of autoAb reactivity in 5-μm unfixed cryostat sections of rat liver, stomach, and kidney tissues. The binding was visualized by FITC–rabbit antihuman μ-chain Ab (Dako, Glostrup, Denmark). Extended screening was performed on FM55_M2, Jurkat, HepG2 (Immunoconcepts, Sacramento, CA), Ag1523, and rat SMCs. The cells were fixed in 4% paraformaldehyde and permeabilized in 0.1% saponin. CLL IgM binding was visualized using FITC–rabbit antihuman μ-chain Ab (Dako) or goat anti–human IgM Ab-Alexa594 (Molecular Probes, Eugene, OR). Anti-TNFα mAb (Innogenetics, Gent, Belgium), anti-CD3 mAb (Dako), and anti-SMA mAb (Sigma-Aldrich) were used as positive controls for FM55_M2, Jurkat, and rat SMCs, respectively. Human IgM antimyelin P₀ mAb¹⁵  was used as a negative control for all cell types. The cells were mounted in Gel/Mount medium with antifading agents (BioMeda, Foster City, CA) and examined in a Zeiss Axiovert 200M inverted fluorescence microscope (Carl Zeiss, Heidelberg, Germany), equipped with a 63×/1.2 NA Plan Apochromat water objective. Images were collected with an AxioCam MRm CCD camera (Carl Zeiss) using Zeiss AxioVision software (Carl Zeiss). Final presentation and mounting of the images were performed using Adobe Photoshop version 7.0 (Adobe Systems, San Jose, CA).

IgM/IgG mAbs from the CLL cell lines were used for further autoAb specificity screening. Total protein (30 μg) from FM55_M2 or HepG2 cell lysates was separated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) and analyzed in Western blots. AutoAb screening of CLL IgM mAbs was performed on RZPD GmbH (German Resource Center for Genome Research, Berlin, Germany) and produced high-density protein macroarray membranes representing E coli–expressed proteins (n = 38 016) from a human fetal brain cDNA library. Colonies were grown on the membranes and IPTG stimulated. Cells were lysed to release the recombinant protein, which was immobilized on the membrane as spots. Primary Abs were incubated overnight at + 4°C, and secondary Ab rabbit anti–human IgM/G/A-HRP (Dianova, Hamburg, Germany) was incubated for 2 hour at room temperature and developed by chemiluminescence (Attophos; Roche Diagnostics, Mannheim, Germany). Secondary Ab–alone controls were included in each assay and low-intensity binding patterns were regarded as negative. Binding quality/strength was graded ++, +, or +/−.

The I83 IgM-specific antigen(s) were isolated by Sieze X protein G immunoprecipitation (IP) kit (Pierce Biotechnology, Rockford, IL) according to the manufacturer's protocol, including anti-IgM/IgM cross-linking to protein G by disuccinimidyl suberate. Purified proteins were separated by SDS-PAGE. IP was not required in the case of HG3, since the HG3-specific antigen was adequately separated on the gel. For in-gel digestion, the Coomassie-stained proteins were excised from the gel, destained, and trypsin digested (Sequencing Grade Modified Trypsin, 2.5 μg/mL; Promega, Madison, WI) as previously described.¹⁶ 

Equal volumes of sample and saturated α-cyano-4-hydroxycinnamic acid in 70% (vol/vol) acetonitrile with 0.3% (vol/vol) trifluoroacetic acid were mixed and applied on the target plate. Matrix-assisted laser desorption/ionization time-of-flight mass spectometry (MALDI-TOF-MS) spectra were acquired on a Voyager-DE Pro (Applied Biosystems, Foster City, CA) and recorded using the instrument settings recommended by Applied Biosystems. The obtained monoisotopic peptide masses were used to search databases (ProteinProspector),¹⁷  allowing a peptide mass accuracy of 100 ppm and one partial cleavage. Carbamido-methylation modification of cysteine was considered.

Peptide sequencing by electrospray ionization tandem MS (ESI-MS/MS) was performed on a hybrid mass spectrometer API Q-STAR Pulsar i (Applied Biosystems) equipped with a nanoelectrospray ion source (MDS Protana, Odense, Denmark). Collision-induced decomposition of selected peptides was performed using the instrument settings recommended by Applied Biosystems with manual operation of collision energy during the spectra acquisition.

I83 vimentin specificity was reconfirmed by Western blot analysis using purified human recombinant vimentin (no. 62115; Progen Biotechnik, Heidelberg, Germany) and by vimentin–enzyme-linked immunosorbent assay (ELISA). IgM in cell culture medium from several CLL patients was tested along with the I83 IgM in the ELISA assay. The plates were coated with 0.5 μg/mL recombinant vimentin in PBS overnight at 4°C and then blocked with 3% BSA in PBS for 4 hours at room temperature (RT). Primary Abs were incubated overnight at 4°C and secondary polyclonal rabbit anti–human IgM HRP-conjugated Ab (Dako) for 1 hour at room temperature (RT). The plate was washed 3 times with 0.9% NaCl with 0.05% Tween-20 between each step. OPD-substrate tablets (Dako) and H₂O₂ were used for development and detection of vimentin-binding IgM. The antivimentin clone V9 mAb (Dako) was used as positive control and the human IgM antimyelin P₀ mAb as negative control. The primary Abs were tested at 7.8 nM to 500 nM. We also used an anti-CCP ELISA (CCP2 RA scan; EuroDiagnostica, Arnhem, The Netherlands) for detection of I83 IgM reactivity with citrullinated (vimentin) epitopes. Samples were tested in triplicates in 3 separate experiments.

CLL mAbs were screened for S pneumoniae capsular polysaccharide binding using a multiplex bead assay including serotypes 4, 6B, 9V, 14, 18C, 19F, and 23F. Three separate experiments were performed. For detailed description, see Document S1.

Macrophages were cultured for 6 days in 20% human serum, followed by 2 days of serum deprivation. The cells were stained for vimentin, using the I83 mAb and V9 mAb. Human IgM or mouse IgG1 (Dako) was used as negative control. Rabbit anti–human IgM–FITC or goat anti–mouse IgG1–FITC (Dako) was used as secondary reagent. Analysis was performed in a FACS Calibur flow cytometer (BD Biosciences, Mountain View, CA).

For membrane vimentin analysis, CLL cells were incubated with V9 mAb, developed with anti–mouse Ig-Alexa594 conjugate (Molecular Probes). Differential interference contrast illumination was used for overlay immunofluorescence images.

Low-density lipoprotein (LDL) was isolated by sequential ultracentrifugation from pooled healthy donors and oxidized by exposure to copper ions (10 μM) at 37°C for 24 hours. MDA modification was performed as previously described (Document S1).

Ab binding to various antigens was performed as described in detail in Document S1 using antigen-coated MicroFluor plates (Dynatech Laboratories, Chantilly, VA), primary Ab (< 3 μg/mL), and alkaline phosphatase–labeled goat anti-IgM or anti-IgG (Sigma-Aldrich) and LumiPhos530 (Lumigen, Southfield, MI) as luminescent substrate. Five to 11 separate experiments were performed.

Apoptotic human Jurkat cells were prepared by exposure to UV light (51 mJ/cm²) after which they were further cultured 18 hours before use. The apoptotic cells were incubated with the primary human mAbs (0.1 to 1.0 μg/mL) for 45 minutes at 4°C in the presence or absence of 100 μg/mL MDA-LDL as specific competitor, followed by incubation with FITC-labeled anti–human-IgM (or IgG) secondary Ab for 30 minutes at 4°C. After staining, cells were double-stained with 7-AAD and analyzed using a FACSCalibur (BD Biosciences). Results were validated using camptothecin-exposed Jurkat cells.

The CLL cell lines used in this study represent EBV-transformed neoplastic clones as evidenced by IGH rearrangements, IGHV sequences, or phenotypic markers (Table 1; Table S1). The cell lines showed different IGHV gene use and mutational status; and some of them belonged to recently described subsets with stereotyped CDR3s. The subset numbering is according to Stamatopoulos et al⁷  and Murray et al.⁸  The CLL cells secreted mAbs in the concentration range of 0.1 to 3 μg/mL per 3 d/10⁷ cells.

Prior to proteomic analysis, we analyzed the mAbs for possible autoantigen binding in a cell panel including vascular smooth muscle cells, FM55_M2 melanoma cells, HepG2 hepatocellular carcinoma, Ag1523 fibroblasts, and Jurkat T-cell leukemia. Autoreactivity with various binding profiles was observed for the CLL mAbs. The binding pattern of the I83/IGHV3-30_M mAb is illustrated in the color micrograph images: human lymph node blood vessel lining endothelial cells (Figure 1A), human malignant melanoma FM55_M2 cells (Figure 1B), rat smooth muscle cells (Figure 1C), and Jurkat human T-cell leukemia cells (Figure 1D). The intracellular cytoskeletal pattern (Figure 1B-D) resembles intermediate filaments. The 2 HG3/IGHV1-2_UM and Wa/IGHV3-30.3_UM/subset-32 mAbs revealed cytoskeletal and cytoplasmic patterns, respectively, in HepG2 cells (Figure 1E,F), whereas AIII/IGHV4-59_M showed a nuclear dotted pattern in Jurkat cells (Figure 1G).

In tissue microarrays based on thin sections of paraffin-embedded human tissues (n = 59), Wa/IGHV3-30.3_UM/subset-32 IgM revealed multiple binding patterns (not shown), thus classified as polyreactive. The rIGHV3-21_M/subset-2 mAb reacted with stomach chief cells (Figure 1H) and pancreatic exocrine glands, but showed no reactivity with the cell lines represented in Figure 1A-G. The CII/IGHV1-69_UM/subset-5 mAb also bound stomach chief cells, whereas no cell line reactivity was found.

The mAbs I83/IGHV3-30_M and HG3/IGHV1-2_UM that showed distinct cytoskeletal patterns (Figure 1B-E) were analyzed in detail by mass spectrometry (MS). These 2 IgM clones showed single bands of 54 kDa and 277 kDa MW in Western blots (Figure 2A,B). Affinity-chromatography isolation of I83/IGHV3-30_M–reacting proteins from melanoma cell extracts on protein G anti-IgM/I83 IgM matrix revealed the 54-kDa immunoreactive moiety, and additional 39-kDa and 17-kDa moieties (Figure 2A left panel). Trypsin digest peptides from gel-excised bands were analyzed by matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) MS and electrospray ionization tandem MS (ESI-MS/MS). The MS peptide profile of the immunoreactive 54-kDa protein is shown in Figure 2A right panel. The HG3/IGHV1-2_UM IgM immunoreactive protein of 277 kDa was excised from SDS-PAGE–separated total cell extract of HepG2 cells, and analyzed in a similar fashion (Figure 2B).

Sixteen of 17 peptides from the MALDI-TOF-MS analysis of the I83/IGHV3-30_M–reactive 54-kDa protein band matched vimentin. In the case of the 39-kDa protein (Figure 2A), 8 of 15 peptides matched aldolase A (Table S3). The 17-kDa protein (Figure 2A) did not generate any peptide signals. The HG3/IGHV1-2_UM–specific protein of 277 kDa generated 20 major peptides, of which 15 matched filamin B (Table S2). The matched peptides covered 47% and 10% of vimentin and filamin B, respectively. The proteins identified by MALDI-TOF-MS were confirmed by sequencing the aa 411-424 peptide of the 54 kDa protein, and the aa 537-551, 658-673, 823-836, 932-946, 2401-2415, and 2465-2484 peptides of the 277-kDa protein (Table S2). Hence, the I83/IGHV3-30_M–specific antigen of 54 kDa was identified as vimentin, and the HG3/IGHV1-2_UM–specific protein of 277 kDa as filamin B.

The specificities of I83/IGHV3-30_M and HG3/IGHV1-2_UM were verified with recombinant proteins or specific mouse mAbs. Furthermore, mAbs from 19 primary ex vivo CLL cultures (Table 2) were obtained by PMA/ionophore stimulation for 48 hours (a mean value of 0.2 μg/mL Ig was released) and analyzed by vimentin-ELISA. I83/IGHV3-30_M showed a robust dose-response binding to vimentin, whereas none of the ex vivo CLL Abs were positive (Figure 2C), which was expected since none of the primary CLL cultures expressed the same IGHV gene as the I83/IGHV3-30_M clone did. The V9 mouse antivimentin mAb served as an external assay control. The antivimentin mAb I83/IGHV3-30_M did not react with citrullinated vimentin epitope as analyzed in a specific immunoassay (data not shown).

Reports on vimentin cross-reactivity of Ab to streptococcal antigens¹⁸  prompted us to test our CLL Abs for binding to S pneumoniae capsular polysaccharides (serotypes 4, 6B, 9V, 14, 18C, 19F, and 23F) by both solid- and fluid-based assays (ELISA and luminex assay). Non–serotype-restricted binding was demonstrated by Wa/IGHV3-30.3_UM/subset-32 and ID-6/IGHV1-69_UM. ID-20/IGHV4-34_M showed a type 14–restricted reactivity. The Wa/IGHV3-30.3_UM/subset-32 mAb dose-response binding to multiple polysaccharides is shown in Figure 2D. The remaining CLL mAbs included in Table 1 were negative (not shown).

CLL Abs have previously been described as “natural Abs.”^19-21  Interestingly, natural Abs in mice that bind S pneumoniae capsular polysaccharides were found to cross-react with moieties on oxidized LDL (oxLDL).²²  Therefore, we wanted to further characterize the binding specificities of our CLL mAbs to the 2 most commonly used models of oxLDL (LDL oxidized by copper ions: CuOx-LDL; and LDL modified by malondialdhyde: MDA-LDL). Figure 3 shows results from a direct Ab-binding assay yielding both positive and negative CLL mAbs binding to MDA-LDL, CuOx-LDL, native-LDL, pneumococcal cell wall polysaccharides (CPSs), and phosphorylcholine-modified BSA (PC-BSA). Each of these 6 CLL IgM clones positive for binding to oxLDL had their own unique binding pattern: Wa/IGHV3-30.3_UM/subset-32 and p241/IGHV3-21_UM showed broad reactivity including both CuOx-LDL and MDA-LDL, and also native LDL; I83/IGHV3-30_M reacted only with CuOxLDL and MDA-LDL, but not with native LDL. Interestingly, 2 of the patients, p457/IGHV1-18_UM/subset-1 and p497/IGHV1-2_UM/subset-1, belonging to the recently described HCDR3 homology subset 1 including IGHV genes of the same clan (IGHV1-2/IGHV1-3/IGHV1-18, IGHV5-a, IGHV7-4-1),^8,14  reacted with MDA-LDL only. The specificity of the binding was verified in competition assays (Figure 3), which demonstrate that only MDA-LDL was able to specifically compete the binding of the clones. Recombinant vimentin was also included in the assays as a positive control and it demonstrated high specificity in competing for the I83/IGHV3-30_M and Wa/IGHV3-30.3_UM/subset-32 IgM clones (Figure 3). The oxLDL experiments were confirmed in 5 to 11 separate experiments.

Natural Abs to oxidized cardiolipin binds to oxLDL, apoptotic cells, and atherosclerotic lesions.²³  Cardiolipin is found in membranes of bacteria, in the inner membrane of mitochondria and in plasma LDL. It was of interest, therefore, to analyze the CLL mAbs for cardiolipin. The results revealed that one of the clones (I83/IGHV3-30_M) had a dose-response binding, whereas the others were negative (data not shown).

Oxidation moieties for natural Abs are not only found in oxLDL, but also in apoptotic cells.²⁴  Therefore, the binding of CLL IgM to apoptotic cells was investigated. Figure 4A shows that the 4 mAbs binding to oxLDL (Wa/IGHV3-30.3_UM/subset-32, I83/IGHV3-30_M, HG3/IGHV1-2_UM, p241/IGHV3-21_UM) also bound to apoptotic Jurkat cells (apoptosis induced by UV light exposure). The specificity was verified by addition of 100 μg/mL oxLDL (MDA-LDL) into the incubation medium, which almost totally abolished the mAb binding to the apoptotic cells (Figure 4A). In addition to UV light, camptothecin, a topoisomerase I inhibitor, was also used for apoptosis induction in experiments confirming UV-induced apoptosis results (data not shown). In these experiments, annexin V–phycoerythrin (PE) or 7-amino-actinomycin D (7-ADD) was used to separate apoptotic from nonapoptotic cells during flow cytometry.

Human red blood cells (rbcs) were investigated for CLL mAb binding before and after UV light–induced apoptosis. None of the CLL mAbs reacted with viable or apoptotic rbcs (data not shown).

Based on previous reports that activated macrophages secrete vimentin, and expose it on the extracellular side of the plasma membrane,²⁵  we further analyzed I83/IGHV3-30_M IgM binding to serum-starved macrophages. Flow cytometry analysis of I83/IGHV3-30_M IgM revealed binding to macrophage-plasma membrane (Figure 4B) similar in strength to the V9 antivimentin mouse mAb that was used as a positive control. The surface membrane vimentin expression was also investigated on 14 freshly thawed primary CLL cells. All CLL cells revealed surface membrane vimentin, exemplified with CLL I83 cells stained with antivimentin V9 mAb (Figure 4C).

Protein macroarrays including 38 016 E coli–expressed proteins (in duplicates) derived from a human fetal brain cDNA library (complete or partial ORF clones) were developed with CII/IGHV1-69_UM/subset-5, 232B4/IGHV3-48_M, and rIGHV3–21_M/subset-2 CLL mAbs (Figure 5). Secondary Ab–alone controls were included in each assay and regarded as negative. CII/IGHV1-69_UM/subset-5 bound to 5/38 016 proteins; 232B4/IGHV3-48_M bound to 4/38 016 proteins; and rIGHV3-21_M/subset-2 bound to 4/38 016 proteins. Table 3 shows the detailed macroarray data including low-intensity binding reactions. The CII/IGHV1-69_UM/subset-5 mAb reacted with 6 different clones, of which 2 reacted with PRAP-1, and one with Ig lambda–like polypeptide 1 isoform b precursor at medium intensity. The mAbs 232B4/IGHV3-48_M and rIGHV3-21_M/subset-2 bound to cofilin-1 with strong intensity, and medium intensity binding was observed for cleavage and polyadenylation specific factor 5. Figure 5 shows the digital image of a macroarray filter with the CII/IGHV1-69_UM/subset-5, 232B4/IGHV3-48_M, and rIGHV3-21_M/subset-2 mAb patterns. Macroarray data were validated with purified His-tagged E coli–produced PRAP-1 or cofilin-1. CII/IGHV1-69_UM/subset-5 showed specific binding to PRAP-1 in ELISA. The binding was dose-response inhibited by preincubation with recombinant PRAP-1. 232B4/IGHV3-48_M and rIGHV3-21_M/subset-2 mAbs showed binding to recombinant cofilin-1 only in complex with E coli extracts (IPTG-induced for cofilin-1 expression); there was no binding to E coli extracts alone or other E coli–expressing proteins (data not shown).

Figure 6 summarizes the CLL mAb specificities found in this study. Each CLL mAb-binding specificity is allocated to autoantigen/molecular motifs found on apoptotic cells or allocated to molecular motifs present on bacteria.

We assessed whether specific antigen exposure had any consequences for CLL viability in growth factor withdrawal–induced apoptosis. I83 cells, with an IGHV3-30_M antivimentin BCR, were exposed to solid-phase protein (0.5 μg vimentin/100 μL adsorbed to Nunc Maxisorb microplates [Rochester, NY]) for 48 hours in RPMI 1640 with 0.5% FCS. Flow analysis for annexin V–FITC and PI showed that control unstimulated cells incubated in medium alone had 59.0% viability after 48 hours, whereas in vimentin-exposed cultures 66.3% were viable. Anti-IgM–exposed cells (1.0 μg/100 μL) showed 66.9% viability. The experiment was repeated twice, also showing increased viability in vimentin-stimulated cultures. The effect was not seen in cultures of p498/IGHV3-49_UM cells.

The novel and previously unpublished finding of this study is that several CLL Ab clones react with molecular motifs present on apoptotic cells and bacteria. The molecular structures identified were vimentin, filamin B, cofilin-1, PRAP-1, phosphorylcholine, cardiolipin, oxLDL, and S pneumoniae polysaccharides. The CLL mAb-specificity patterns, in particular the observed antibacterial and anti-oxLDL antigens, are reminiscent of mouse natural Ab reactivities described in autoimmune diseases,²⁶  and in clearance of senescent cells and pathogenic bacteria.^22,23,27  Notably, microbial associations have been reported in several B-cell lymphomas,^28-30  and recently it was suggested that respiratory tract infections increase the risk of CLL.³¹ 

CD5⁺ CLL B lymphocytes have previously been reported to produce natural autoAbs.^11,19,32,33  Although less well defined in humans compared with mice, CD5⁺ B cells are considered to constitute a self-replenishing B-cell subset producing “natural IgM Abs” characterized by lack of somatic hypermutation, low affinity for microbes (eg, S pneumoniae), and autoreactivity. The unmutated CLL mAbs found in this study are polyreactive, thus resembling natural Abs. The mutated CLL clones tended to have a more restricted binding profile (eg, type 14 S pneumoniae polysaccharides and cofilin-1; Figure 6), confirming previous findings by Martin et al^20,21  and recent studies by Herve et al¹¹  showing that when in vitro–produced recombinant mutated nonautoreactive CLL mAbs were reverted to their original unmutated counterpart, they encoded polyreactive and autoreactive Abs.

Oxidation of LDL generates a wide variety of neoself determinants that have been shown to lead to a variety of cellular and humoral immune responses, which are important steps in early events of atherosclerosis.³⁴  Although these neoself structures occur in oxLDL particles, such oxidation-specific epitopes can also be found in other proteins and phospholipids. Furthermore, oxLDL can induce humoral immune responses in vivo, and both IgG and IgM autoantibodies specifically binding to these oxidation-specific epitopes are found in plasma of animals and humans.³⁴  Of particular relevance is the recent discovery that mice have “natural” germline Abs binding to oxidation-specific epitopes present on oxLDL, on apoptotic cells, as well as on certain microbial antigens.^24,35  These earlier observations on mice together with the new findings of the present paper on human CLL clones further validate the concept that natural Abs, in both mice and humans, exhibit a remarkably conserved repertoire, and that they are typically regarded as polyreactive so that they bind to a number of self- or foreign antigens. This was evident especially among those CLL clones demonstrating binding to oxLDL. This broad reactivity pattern of innate Abs can be considered a requirement for the rapid and immediate recognition and protection against a variety different types of invading pathogens.²²  On the other hand, the binding property to apoptotic cells may also represent physiological “housekeeping” roles in the recognition and removal of senescent cells, cell debris, and other self-antigens of the natural Abs.

It is intriguing that several of the autoantigens found in this study have functional association with microbial infections and/or apoptotic cell removal. Vimentin is a type III cytoskeletal protein and intermediate filament that recently was found to have functions on the extrafacial side of the plasma membrane,²⁵  and was secreted as a stress response after bacterial insult.^25,36  Vimentin is exposed on apoptotic primary T cells, oxLDL-binding macrophages, neutrophils, and platelets.^37,38  Cell membrane–exposed vimentin was recently shown to be a port of entry for group A streptococci.³⁹  Modified citrullinated vimentin (the “Sa” antigen), is an “eat me” signal for phagocytes, but can also generate autoAbs as demonstrated in patients with rheumatoid arthritis (RA).⁴⁰  In this study, we found that the I83/IGHV3-30_M and Wa/IGHV3-30.3_UM/subset-32 mAbs reacted with vimentin (Figure 3) and showed binding to apoptotic Jurkat cells (Figure 4A). The I83/IGHV3-30_M mAb was also found to bind to vimentin epitopes on serum-starved macrophages (Figure 4B), but lacked binding to the citrullinated vimentin epitope (data not shown). Antiphosphorylcholine Abs cross-react with vimentin,⁴¹  indicating that natural Abs may have a dual recognition for effective opsonization and removal of vimentin-coated microbes.

Filamin B was detected in Western blots, isolated, and identified in MS after SDS-PAGE separation (Figure 2B). Filamin A and filamin B are actin cross-linking filamentous proteins that are engaged in more than 40 molecular interactions such as with TRAF2, androgen receptors, dopamine receptors, and Smad.⁴²  Filamin B is also exposed on the outer surface membrane and bind human natural IgM Abs that induce apoptosis in human neuroblastoma cells.⁴³ Salmonella type III secretion effector proteins target filamin for successful invasion of epithelial cells.⁴⁴  In this study, the HG3/IGHV1-2_UM anti–filamin B mAb reacted with oxLDL epitopes and apoptotic Jurkat cells. The apoptotic cell binding was competed by oxLDL (MDA-LDL) in a dose-dependent fashion (Figure 4A).

Proline-rich acidic protein-1 (PRAP-1) is an epigenetically regulated protein preferentially found in epithelial cells of the gastrointestinal tract and uterus.⁴⁵  It is up-regulated 30-fold and secreted during induction of terminal differentiation/apoptosis as a stress response protein.⁴⁵  The biologic function is not known, but parallel studies reveal binding to bacterial surfaces of PRAP-1 (A.R. et al, unpublished data, November 2007), analogous to proline-rich antibacterial peptides (binding Gram-negative bacteria).⁴⁶ 

Cofilin-1 is an actin-binding and depolarizing protein that colocalizes with Arp2/3 and the Wiskott-Aldrich syndrome protein (WASp) complex in apoptotic blebs during programmed cell death.⁴⁷  WASp is required for efficient phagocytosis of apoptotic cells. During infection of epithelial cells by Gram-negative bacteria (eg, Salmonella, Chlamydia, Listeria), cofilin is being “hijacked” for a successful takeover of the cytoskeletal rearrangements.⁴⁸  CLL clones 232B4/IGHV3-48_M and rVH3-21_M/subset-2 showed binding to cofilin-1 (Figure 5), but revealed no binding to oxLDL epitopes or apoptotic Jurkat cells (Figures 3,4). This observed lack of binding may have several explanations, including posttranslational modifications such as phosphorylation/dephosphorylation of cofilin-1 at Ser3 (epitope conformational changes), restricted cofilin-1 expression to specific time intervals, cell type–dependent expression, cryptic epitopes, low avidity, and others. The CLL anti–cofilin-1–binding pattern renders further studies on the epitope structure, expression in multiple cell types, and time-kinetic studies necessary.

One of the low-affinity reactions found in the protein macroarray (Table 3) is worth mentioning: cleavage and polyadenylation specific factor 5 (CPSF5) reacted with CLL mAb 232B4/IGHV3-48_M. CPSF5 interacts directly with U2 snRNP,⁴⁹  a nuclear complex structure, which is known to generate autoAbs in systemic lupus erythematosus (SLE) patients (anti-Sm, etc).⁵⁰  Clearance of apoptotic cells is of outstanding importance for a balanced cellular homeostasis and involves a complex network of scavenger, recognition, decorating, and bridging molecules.^51,52  In CLL, the CD5⁺ CLL cells are considered to be self-renewing, and our data revealing BCR–to–self-antigen recognition (on apoptotic cells) may have serious consequences if dysregulated. For example, overexpressed apoptotic molecular motifs (ie, membrane-exposed vimentin, as shown in Figure 4C and reported by others^53,54 ) may generate uncontrolled (BCR)–proliferative signaling. This hypothesis, however, is not proven by our data and requires further studies. We suggest that these studies should include experiments that consider activation and proliferation signals generated by specific antigens alone or antigens exposed on apoptotic cells and bacteria, as well as signaling via CD38⁵⁵  and its interaction with CD31, involved in “tethering” apoptotic cells.⁵⁶  In addition, antiapoptotic signals in the proliferative environment such as stromal cell–derived thioredoxin,⁵⁷  Bcl-2 overexpression,⁵⁸  or DAPK1 down-regulation⁵⁹  may contribute to a vicious circle.

In conclusion, the most important finding of this study is that molecular patterns on apoptotic cells were identified as neoautoantigen, specifically binding to CLL Igs encoded by the various IGHV genes. These results constitute a proof-of-principle that natural Abs previously described only in mice are also present in humans. The dual-recognition of the CLL mAbs is of interest in that they react with both bacterial components (S pneumoniae polysaccharides and phosphorylcholine) or self-antigens that have been reported to interact with microbes (vimentin, filamin B, cofilin-1) on one hand and motifs exposed on apoptotic blebs on the other hand (vimentin, PRAP-1, oxLDL, cardiolipin). Together with recent findings that pneumonial infections increase the risk of CLL,³¹  and preliminary observations on antibacterial CLL Abs,⁶⁰  our data point at the possibility that bacterial infections in synergy with neo–self-antigen/apoptotic cells may drive promotion of CLL by continuously triggering the BCR, thus making the B-cell population vulnerable to tumor-transforming events.

The authors express their sincere thanks to Drs Kenneth Nilsson and Eva Klein for valuable discussion on the results of this study. Many thanks to Dr Shing Chuan Hooi, Singapore, for the generous gift of anti–PRAP-1 and recombinant PRAP-1, and also to Dr Johan Akter Hossain for cell culture work, Ulrik Lindquist for ELISA determination, Drs Alexander Vener and Maria Hansson for advice with MS, Dr Per Hultman for advice on immunohistologic evaluations, Drs Gunnar Juliusson and Karin Karlsson for help with recruiting patients, and to Marina Johnson for pneumococcal capsular polysaccharide–binding studies.

This work was supported by the Academy of Finland, Finnish Foundation for Cardiovascular Research and Sigrid Juselius foundation (S.H.); Special Trustees of Great Ormond St National Health Service Trust (H.B.); the Swedish Research Council, Swedish East Gothia Cancer Foundation, Linköping University Hospital Funds, and Swedish Cancer Association no. 3171-B04-16XBB.GSD.

Contribution: A.L.M. and E.H. planned and performed research, analyzed data, and wrote the paper; E.S., A.S., H.B., C.D., G.T., E.B., and O.S. performed research and analyzed data; K.W. performed research; R.R. analyzed data and wrote the paper; S.H. performed research, analyzed data, and wrote the paper; A.R. designed and supervised the research, analyzed data, and wrote the paper.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Anders Rosén, Department of Clinical and Experimental Medicine, Division of Cell Biology, Linköping University, SE-58185 Linköping, Sweden; e-mail: anders.rosen@ibk.liu.se.