Chronic lymphocytic leukemia specificity analyses unveil bacterial lipopolysaccharides as the cognate ligands of established stereotyped BCR subsets

Dear Editor,

Chronic lymphocytic leukemias (CLLs) collectively express a biased immunoglobulin heavy variable (IGHV) gene repertoire with subgroups expressing stereotypic IGHV-IGHD-IGHJ gene rearrangements with (near-)identical VH-complementarity determining regions 3 (CDR3) [1–3]. As yet, the European Research Initiative on CLL (ERIC) consortium has defined, among a cohort of 29,856 CLL IGHVs, 2,195 stereotyped B-cell receptor (BCR) subsets, together constituting more than 40% of all CLLs of the entire cohort [2].

The existence of groups of CLLs expressing highly similar BCRs points out that they recognize the same or highly similar epitopes. Indeed, antibodies (Abs) derived from a significant proportion of U-CLL stereotypic BCR subsets display subset-specific patterns of polyreactive binding to various self- and exo-antigens [4, 5]. By contrast, most M-CLL Abs are non-polyreactive in in vitro binding studies. At present, the immunological repertoire of M-CLLs, either or not belonging to BCR stereotyped subsets, is essentially unexplored [3]. Previously, we have defined two ‘immunological’ M-CLL subgroups, one expressing autoreactive high-affinity rheumatoid factors (RF), i.e., IgM specific for IgG [3, 6, 7], and a stereotyped IGHV M-CLL subset expressing BCRs selected for binding the fungus- and yeast cell wall- restricted sugar β-(1,6)-glucan [3, 8].

We analyzed a cohort of 172 CLLs enriched for M-CLLs, considering that M-CLL BCRs exhibit more restricted and specific antigen-binding properties as compared to the frequently polyreactive BCRs of U-CLL [3–5]. In our cohort, 140 of the 172 CLLs (81%) harbored >2% somatic mutations in the IGHV gene. We screened all VH-CDR3 amino acid sequences for assignment to one of the 2195 stereotyped CLL BCR subsets, defined by ERIC [2]. In total, 47 CLLs (27%) fitted with a defined CLL stereotyped BCR subset (Supplementary Table S1).

Of 169 CLL (98%), we successfully produced secreted CLL-derived Abs by inducing in vitro plasmacytic differentiation of primary CLL cells [9]. The CLL Ab panel, containing 116 M-IgMs, 31 U-IgMs, 21 M-IgGs, and 1 U-IgG, was tested for binding on a panel of 67 self/auto-, fungal-, viral-, and bacterial- antigens (listed in Supplementary Table S2). In total, 57 Abs (34%) exhibited polyreactivity, defined as binding to multiple antigens. All polyreactive Abs were IgM, comprising 18 U-IgM and 39 M-IgM Abs, representing 58% and 34% of all U-IgM and M-IgM Abs in the cohort, respectively. The relatively high proportion of polyreactive M-IgM Abs (34%) was not anticipated and not described previously. A notable distinction between our study and those of others is that we generated Abs directly from primary CLL cells, which predominantly express IgM, whereas most previous studies investigated recombinant IgG from CLL BCRs [4, 5]. Of note, none of our 21 M-IgG Abs tested displayed polyreactivity. Sixteen (33%) of the polyreactive Abs belonged to stereotyped BCR subsets. (Supplementary Table S3).

Interestingly, nine Abs were found to specifically bind to various LPS moieties, five of which (CLL128, CLL144, CLL194, CLL223, and CLL316) were cross-reactive with LPS variants derived from different bacterial species. CLL194 IgM showed particularly strong binding to LPS from Klebsiella pneumoniae. Three Abs exhibited more selective reactivity, i.e., CLL114 bound exclusively to LPS from Salmonella enterica serotype typhimurium, while CLL134 and CLL192 bound specifically to LPS from S. enterica serotype Minnesota. CLL183 Ab was exceptional as it reacted with various mutant LPS molecules all lacking both the O-antigen and core oligosaccharide (Fig. 1a). In parallel, IgMs of eight LPS-specific M-CLLs were produced recombinantly, all of which displayed the same LPS binding profiles as the Abs retrieved from the corresponding M-CLL cultures (Fig. 1a). We were unable to produce a recombinant IgM-λ of CLL134. The antigenic reactivity of this case was independently confirmed with IgM obtained by cultures of FACS-purified CD19⁺ CD5⁺ CLL cells.

Fig. 1. Binding profiles and immunoglobulin characteristics of LPS-specific CLLs.

a LPS binding of CLL IgMs from culture and recombinantly produced. Red and orange respectively indicate strong (>5 times background ABS 450 nm) and moderate (3–5 times background ABS 450 nm) antigen binding of the CLL IgM at 500 ng/ml in ELISA. Green indicates no antigen binding, nd = not done. The E. coli EH100 mutant lacks the O-antigen chain, but retains a complete or near-complete core oligosaccharide structure. The E. coli J5 mutant, a UDP-galactose-4-epimerase-deficient strain, lacks the O-antigens (Rc), consisting only of the lipid A component and a core polysaccharide. The Salmonella enterica typhimurium SL1181 mutant has a mutation in the rfa gene cluster and lacks the core oligosaccharide and the O-antigen. The Salmonella enterica Minnesota Re595 mutant lacks the O-antigen and part of core polysaccharide. b IGHV and IGKV/IGLV rearrangements of nine LPS-specific M-CLLs. c Four LPS-specific M-CLLs fit in ERIC-defined CLL stereotyped BCR subsets. The VH-CDR3 amino acid sequence homology between LPS-specific M-CLL and CLL stereotyped BCR subsets is highlighted in red and blue for identical and similar amino acids, respectively. d Binding profiles of four additional CLL IgMs, two IgMs from subset #V1-2-LPS-1 and two IgMs from subset #V4-34-LPS-3. The colors indicating the binding strengths in ELISA are identical as described under (a).

Analyses of IGHV and IGKV/IGLV sequences revealed that four M-CLLs fitted in distinct ERIC-defined stereotyped BCR subsets, that is: CLL128, CLL134, CLL144, and CLL316 belong to subsets #156, #137, #29C, and #V1-3|J3.4.5.6|19|1, respectively. We designated these LPS-specific BCR subsets as #V1-2-LPS-1 (#156), #V3-23-LPS-2 (#137), #V4-34-LPS-3 (#29), and #V1-3-LPS-4 (#V1-3|J3.4.5.6|19|1) (Fig. 1b, c) (Supplementary Table S4).

To determine whether the LPS binding profiles were conserved across independent members of stereotyped BCR subsets, we examined additional CLLs. From the ERIC consortium, we obtained two #V1-2-LPS-1 subset members (CLL413 and CLL414) and two #V4-34-LPS-3 subset members (CLL402 and CLL-GN27). Notably, all three #V1-2-LPS-1 subset members (CLL128, CLL413 and CLL414) utilized IGKV1-6-encoded light chains, while the #V4-34-LPS-3 subset members (CLL144, CLL402 and CLL-GN27) shared IGKV3-11 expression (Supplementary Table S4).

The IgMs produced of CLL413 and CLL414 exhibited LPS reactivity profiles similar to that of CLL128, confirming subset-specific antigen recognition. Interestingly, CLL413 IgM also demonstrated low-level cross-reactivity with LPS mutants. The LPS binding profiles of CLL402 and CLL-GN27 IgMs mirrored that of our CLL144 IgM, reinforcing the notion of uniform epitope specificity within the #V4-34-LPS-3 subset (Fig. 1d).

To investigate whether the M-CLL Abs were affinity-selected, we generated variants of eight M-CLL Abs by reverting their somatic mutations to the closest corresponding germline IGHV sequences. In five of these CLLs (CLL144, CLL183, CLL192, CLL223, and CLL316), the reversion led to a complete loss of LPS binding, indicating that somatic hypermutation has been essential for the acquisition of LPS reactivity. Three IGHV germline-reverted IgMs (CLL114, CLL128, and CLL194) retained LPS binding. However, when additional reversion of the light chain somatic mutations (IGKV/IGLV) was performed, both CLL114 and CLL194 lost their LPS binding capacity. These findings indicate that LPS-binding M-CLL Abs result from antigen-driven affinity selection processes involving adjustments in both heavy and light chain IGV genes. Intriguingly, the fully reverted #V1-2-LPS-1 subset member CLL128, instead of LPS binding loss, acquired strong polyreactivity (Fig. 2a, b). Apparently, CLL128 originated from a self/poly-reactive BCR expressed by a B cell precursor, which, due to the introduction of somatic IGV mutations, had obtained specific affinity for LPS. The phenomenon that somatic hypermutation counteracts self/poly reactivity and redirects specificity has been reported for five M-CLLs and for IGHV4-34-expressing B cells, which have natural super-antigen binding to poly-N-acetyllactosamine epitopes [5, 10].

Fig. 2. LPS-specific CLL are affinity selected and are responsive to in vitro stimulation with LPS.

a Binding profiles of IGV germline-reverted IgM variants of seven LPS-reactive CLL IgM. Red and orange indicate, respectively strong and medium antigen binding of the CLL IgMs in ELISA. Green indicates no antigen binding. b Titration curves of the CLL IgMs for specific binding to the indicated LPS moieties. c Primary cells of LPS-specific CLL128 and non-LPS-specific CLL308 were labeled with CFSE and cultured for 8 days in the presence of irradiated CD40L L-cells with either anti-IgM or LPS from E. coli O55:B5. Proliferation of CLL128 was induced by both anti-IgM and LPS, whereas only anti-IgM induced proliferation in CLL308.

Six out of the nine LPS-binding M-CLL-derived Abs demonstrated, without being polyreactive, cross-specificity for LPS of various gram-negative bacterial strains. This binding pattern resembles normal physiological B cell responses following bacterial infection or immunization. The resemblance of the LPS binding spectra of the M-CLLs with normal LPS-specific B cells suggests that the LPS-specific M-CLL cells originated from normal B cells responding to LPS (see Supplementary text).

Finally, we demonstrate that primary tumor cells of CLL128 (#V1-2-LPS-1) showed a specific proliferative response upon LPS stimulation, indicating that the LPS-specific BCRs are functional and potentially able to support tumor cell survival and growth.

The phenomenon of CLL stereotypic BCR subsets is strong evidence in favor of the communal idea that the genesis and growth of most B-cell lymphoma entities is BCR-driven. Several mechanisms of BCR-driven growth and survival may apply in CLL. First, CLL as a group exhibits a biased IGHV gene repertoire, with U-CLLs frequently expressing polyreactive BCRs. These CLLs may derive from progenitor B cells carrying initial transforming alterations, such as NOTCH1 and SF3B1 mutations, which may support survival of polyreactive clones that would normally be deleted in the bone marrow [11–13]. Second, an unknown proportion of CLLs harbor structural BCR mutations that, like IGLV3-21 with an R110 mutation, provoke homotypic BCR ligation and constitutive signaling [14]. Third, the identification of cognate BCR ligands of stereotyped M-CLL subsets, such as LPS and β-(1,6)-glucan, indicates that these CLLs have retained the antigenic specificity and affinity for which their precursor B cells were originally selected [3, 8].

It is unknown to what extent the immune repertoire of non-polyreactive CLLs mirrors that of healthy people's B-cell repertoires and, more importantly, up to which stage of tumor development natural BCR ligands may have a growth-sustaining role. It is to be expected that in advanced-stage CLL multiple oncogenic mechanisms prevail, i.e., genomic alterations combined with reinforcing BCR signals. Intriguingly, CLLs of some stereotyped BCR subsets have distinctive genomic alterations, growth characteristics, and/or clinical behavior [3]. It cannot be excluded that M-CLLs with retained natural immune specificity may simultaneously carry structural IG alterations provoking lasting homotypic BCR-BCR interaction. The findings, however, that CLLs, either specific for LPS, β-(1,6)-glucan, or IgG (rheumatoid factors), are all antigen-responsive in culture, seem to argue against sustained BCR signaling induced otherwise in these tumors [6, 8].

We thus describe a novel “immunological” M-CLL subset that specifically binds to a variety of lipopolysaccharides (LPS) derived from common gram-negative bacteria. Together, the four ERIC CLL stereotypic BCR subsets with LPS specificity identified comprise 133 CLLs, representing 0.45% of the 29,856 VH-CDR3 amino acid sequences of the cohort, and thus, according to the ERIC definition, are to be labeled as a “major” subset. In patients, LPS together with bacteria-derived CpG-DNA may support tumor survival and growth by dual BCR and TLR signaling [15].

Author contributions

Study conception and design: JJ, ZL, TF, RJB, and CJN. Experiments for acquisition of data: JJ, TAW, AB, HN, KPK, and RJB. Analysis and interpretation of data: JJ, ZL, TAW, AB, HN, CJN, and RJB. Patient materials and discussions: CES, APK, and AWL. Supervision of the studies: TF, RJB, and CJN. Writing of the manuscript: JJ, JEG, RJB, and CJN. All authors read and approved the final manuscript.

Funding

This research was supported by a grand from the Dutch Cancer Society (UVA2014-6824).

Data availability

The data and materials that support the findings in this study are available from the corresponding author upon reasonable request.

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

All methods were performed in accordance with the relevant guidelines and regulations. The study was approved by the medical ethics committee at the Amsterdam University Medical Center under the name B cell malignancies Biobank and number 2013_159. Written informed consent was obtained from all subjects in accordance with the Declaration of Helsinki.

Supplementary information

The online version contains supplementary material available at 10.1038/s41408-025-01405-7.