The study was approved by the Institutional Review Board of Northwell Health. Bone marrow (BM) samples were collected as discarded bone segments from anonymized patients who had undergone joint replacement surgery. Persons with a history of any autoimmune disease or condition and of any cancer were excluded from the study. Peripheral blood and umbilical cord blood samples were similarly collected from anonymous healthy donors. Mononuclear cell (MC) fractions were separated by density gradient centrifugation (Ficoll, GE Healthcare), frozen (10% DMSO 45% FBS and 45% RPMI1640) and stored in liquid nitrogen until used.

BM samples were placed in a large Petri dish containing cold PBS with 2.5% BSA, and the tissue was gently dissociated using the plunger of a 60 ml sterile plastic syringe. Bone fragments were broken into small pieces using scissors and rinsed with the same buffer to extract cells from tissue niches. Cell suspensions were then passed through a 70uM cell strainer into a 50 ml tube. To optimize the yield, after processing Petri dishes were rinsed with buffer used for the dissociation, and the contents were added to the previously filtered suspension. BMMCs were separated by Ficoll density gradient centrifugation, frozen (10% DMSO 90%FBS), and stored in liquid nitrogen until used.

BM cell suspensions were incubated with V500 anti-CD19 and with PE-cy7 anti-CD10 mAbs (both BD Biosciences) for 20 minutes at 4°C, and after washing were sorted into CD10⁺ and CD10^- fractions. Non-B cells were excluded by using efluor-450 anti-CD3 and anti-CD16 mAbs, and dead cells were triaged by Sytox Blue (ThermoFisher) staining. The CD10⁺ fractions were then additionally stained with FITC anti-CD34 (BD Biosciences), PE anti-IgM (Ebioscience) and eflour-450 anti-CD27 (Ebioscience) to further prohibit contamination with mature B cells. Pro B cells (PRO, CD34⁺IgM^-), Pre B cells (PRE, CD34^-IgM^-) and immature B cells (IMM, CD34^-IgM⁺) were collected.

CD10^- fractions were also stained with PerCPcy5.5 anti-CD38 (BioLegend), FITC anti-IgD (ThermoFisher), APC anti-CD27 (BD Bioscience), and PE anti-CD24 (Bioscience) to discriminate naïve (NAÏVE, CD24⁺IgD⁺CD27^-) and memory (MEM, CD24⁺IgD^-CD27⁺) B cells and plasmablasts/plasma cells (PB/PC, CD24^-CD38⁺⁺).

PBMCs from normal blood donors were incubated with the following anti-human Abs for 20 minutes at 4°C: V500 anti-CD19 (BD Biosciences), PerCPcy5.5 anti-CD38 (BioLegend), PE-cy7 anti-CD24 (BioLegend), FITC anti-IgD (ThermoFisher), and allophycocyanin anti-CD27 (BD Bioscience), and then sorted to isolate Transitional (TRANS; IgD⁺CD27^-CD10⁺CD38⁺), NAIVE (CD27^-IgD⁺), recirculating Marginal Zone (rcMZ; IgD⁺CD27⁺ (24), MEM (IgD^-CD27⁺) and double negative (DN; IgD^-CD27^-) B cells.

Total CD19⁺IgD⁺CD27^- B cells were sorted from umbilical cord blood samples. In both cases, non-B cells were excluded with efluor-450 anti-CD3 and anti-CD16 mAbs, and dead cells barred with Sytox Blue (ThermoFisher) staining.

For all samples, B cells were sorted directly into 200µl PCR tubes containing 100µl Dynabeads Oligo(dT) (ThermoFisher) lysis buffer and stored at -80°C.

mRNA isolation from B-cell lysates was performed in 96-well plates using Dynabeads Oligo(dT) (ThermoFisher) according to the manufacture’s protocol. mRNA was used in its entirely for reverse transcription in 10 µl (50°C 1h, 72°C 10min) using SuperScript III Enzyme (ThermoFisher) in solid phase with Dynabeads Oligo(dT) as primer. After RNase H treatment, second-strand synthesis was performed (37°C 20 min, 98°C 30s, 62°C 2min and 72°C 10min) in solid phase in 10µl using Q5 Polymerase (NEB) and a mix of 13 primers covering all IGHV leader sequence segments reported in the IMGT database (25); primers contained a maximum of one mismatch, along with 13 to 16 random nt and partial Illumina adaptor sequences. Double-stranded cDNA was washed 3 times in 10mM tris-HCl to remove the remaining primers, and the entire sample was utilized as template for PCR amplification in 10 µl using Q5 Polymerase with universal FW primer and mix of reverse isotype specific primers (98°C 30s; 10 cycles at 98°C for 10s, 58°C for 15s, and 72°C for 1min; 72°C 10min). Two µl of the PCR product were used for a semi-nested PCR with inner RV primers for the constant region which also introduce partial Ilumina adaptors. This reaction was carried out in 20µl (98°C 30s; 15 cycles at 98°C for 10s, 58°C for 15s, and 72°C for 1min; 72°C 10min). The PCR product was purified with Ampure XP beads at a ratio of 1:1, and 1 - 10ng were used to add Illumina Indices with Nextera XT kit (Illumina). The MiSeq Illumina (v3 2 x 300 kit, Illumina MS-102-3003) was used to sequence the library. The library was loaded at 12pM with 10% PhiX [14].

The 9G4 rat anti-IGHV4-34 mAb (26) was labeled with Alexa Fluor ™ 488 according to the manufacturer’s recommendations (Alexa Fluor™ 488 Antibody Labeling Kit, ThermoFisher). Briefly, antibody solution was mix with1/10^th of 1M sodium bicarbonate and then incubated with Alexa Fluor™ 488 dye for 1h at room temperature. After the recommended period, the solution was placed in the provided purification column and labeled antibody was collected from the flow through.

To analyze the paired IG heavy and light chain IGHV-D-J and IGLVκJκ sequences of B cells expressing IGHV4-34, we used 10x methodology. After exposing normal Naïve, MZ, MEM, and DN B cell populations to the Alexa Fluor ™ 488-labeled 9G4 mAb, labeled cells were enriched by FACS ( Figure S6A ). For each sample, cells were washed and resuspended in 31.7 ul of PBS (0.04% BSA) immediately after sorting. Single-cell libraries were then generated using the Chromium Controller, Chromium Single Cell 5’ Library & Gel Bead Kit v2 and i7 Multiplex Kit (10x Genomics, Pleasanton, CA, USA), according to the manufacturer’s protocols. Target enrichment from cDNA was performed using the Chromium Single Cell V(D)J Enrichment Kit, Human B Cell (10x Genomics), followed by adaptor ligation. Enriched libraries were quantified on an Agilent Bioanalyzer High Sensitivity chip, and then sequenced on an Illumina Nextseq500 instrument (Illumina, San Diego, CA, USA) with the paired-end (2x150bp) mid output kit (300 cycles) according to manufacturer’s protocols.

For bulk VDJ-seq, processing of raw reads was performed using a custom workflow built with pRESTO (REpertoire Sequencing TOolkit) (27). The IGHV-D-J sequences obtained were submitted to IMGT/HighV-QUEST and analyzed using ChangeO and custom R scripts (23, 27).

Cellranger vdj pipeline was used to analyze sequencing data obtained from 10x Chromium V(D)J libraries.

IGHV-IGHD-IGHJ gene rearrangements were analyzed for similarity to stereotyped CLL BCRs using our established bioinformatics method (10). In more specific, the subsequent clustering criteria were applied: (i) utilization of IGHV genes belonging to the same phylogenetic clan, (ii) ≥ 50% amino acid identity and ≥ 70% similarity within the VH CDR3, (iii) equal VH CDR3 length and, (iv) identical offset of the common amino acid motif.

Satellite CLL-SLS, i.e. sequences with strong immunogenetic similarities with major CLL subsets, were identified using a purpose-built bioinformatics algorithm, which is based on a set of previously described parameters (10) (1): utilization of phylogenetically associated IGHV genes (2), maximum VH CDR3 length difference of 2 amino acids, and (3), presence of the “subset-specific” VH CDR3 sequence motifs with an offset of ± 2 amino acids. This analysis was performed individually for each major subset.

Statistical analyses and identification of outliers were performed in Graphpad Prism 9. Tests for statistical significance are described in figure legends for the relevant graphs.

First, we asked if and to what extent B lymphocytes expressing BCRs closely resembling CLL-stereotyped BCRs exist in healthy people. To do so, we collected samples from CB of neonates (n=5), BM (n=11) of elderly people (≥ 70 years of age) who had undergone hip replacement surgery, and peripheral blood mononuclear cells (PBMC) of adult volunteers (35-60 years of age; n=16) ( Figures S1A, B ).

Total CD19⁺ B cells were FACS isolated from CB. Whereas from BM we isolated several B-cell subsets representative of the distinct B-cell developmental stages, using a combination of surface membrane markers ( Figure S1A ).

From each of the 3 different cell sources (CB, BM, PBMC), we were able to identify B cells bearing an IGHV-D-J rearrangement that corresponded to that of a known CLL stereotyped subset. For convenience, we refer to this type of rearrangement found in B cells from normal individuals as a “CLL stereotype-like sequence” (CLL-SLS). Table S1 summarizes the number of unique CLL-SLS obtained for each cell population sorted from the various sites. We identified a total of 123, 513, and 999 CLL-SLS in CB, BM, and PBMC, respectively. The average frequencies were relatively comparable among the three (CB: 0.044%; BM: 0.037%; PBMC: 0.051%) ( Figures 1A , S1C ).

We then compared the distribution in BM of CLL-SLS at the various stages of B-cell development and among mature B-cell subsets. The highest frequency values were found among B-cell populations representative of the first stages of maturation: PRO-B, PRE-B, IMM, and NAIVE B cells (average frequencies: 0.046%, 0.048%, 0.031%, and 0.037%, respectively). The average frequency of CLL-SLS in MEM (0.023%) was lower, although these values did not reach statistical significance despite the number of total sequences queried being similar ( Table S1 ). Interestingly, the PB/PC population harbored cells carrying CLL-SLS rearrangements at higher frequencies than immature/naïve stages (0.059%) ( Figures 1B , S1D ).

CLL-SLS were identified in each of the circulating B-cell subsets in PBMC: TRANS, NAIVE, rcMZ, MEM, and DN B cells (average frequencies: 0.047%, 0.047%, 0.076%, 0.032%, and 0.03%, respectively) ( Figures 1C , S1B ). CLL-SLSs were significantly higher in rcMZ compared to MEM and DN B-cell subsets (rcMZ vs MEM and DN, P ≤ 0.05) ( Figures 1C , S1E ). Lastly, we did not find a difference of CLL-SLS frequencies between NAÏVE from BM and NAÏVE from PBMC ( Figure 1D ).

In summary, normal B cells expressing standard CLL-SLS exist in the repertoires of healthy subjects and are present at similar frequencies from tissues that dramatically differs in age. Moreover, CLL-SLS are found at all stages of B-cell development, but, interestingly, at significantly higher frequencies in the PB/PC compartment in the BM and significantly higher frequencies in the rcMZ B-cell subset isolated from PBMCs.

Together, these findings suggest that CLL-SLS are not triaged during first stages of B cell development by clonal deletion. Since our data are generated from whole B-cell subpopulations, we cannot determine if receptor editing could have occurred in these dells, which also would allow them to transit through B-cell maturation. In addition, their presence in antigen-experienced B cell subsets, e.g., MZ and MEM, and among antibody-secreting B cells in the BM, could represent class switch recombination (CSR) and SHM along with positive selection by a particular antigen.

Next, we extended our analysis to IGHV-D-J sequences that resemble those referred to as “satellites” of known CLL stereotyped subsets (10). Satellite stereotyped sequences resemble the VH CDR3 motif of CLL stereotyped subsets but differ in certain amino acid residues at specific positions in the IGHV-D-J rearrangement or vary in VH CDR3 length. Although satellites are only in a minor component of the total number of stereotyped subsets found in CLL patients (10), we identified a ≥ 10-fold enrichment of “satellite” CLL-SLS compared to standard CLL-SLS in every tissue (CB: 3,545 vs 123; BM: 6,145 vs 513; PBMC: 11,707 vs 999; Tables S1, S2 ). Notably, the highest frequency of satellite CLL-SLS was measured in CB, where they reached an average frequency of 1.31% of the total IG sequence ( Figures 2A , S2A ).

When examining satellite CLL-SLS at the different stages of B-cell development, we were able to detect satellite sequences in every B-cell subset isolated from the BM (PRO-B: 0.27%; PRE-B: 0.42%; IMM: 0.40%; NAÏVE: 0.70%; MEM: 0.20%; PB/PC: 0.39%). Differently from the standard we observed a statistically significant enrichment in satellite CLL-SLS in BM NAÏVE compared to BM PRO-B (P = 0.0089) and NAÏVE to PBMC MEM (P = 0.0016) ( Figures 2B , S2B ).

Similarly, in PBMC samples, satellite CLL-SLS were present at discrete frequencies inn all the BB cell subsets analyzed but we only observed a significant difference in satellite CLL-SLS frequencies when comparing the average of NAÏVE to MEM and DN cells (NAÏVE: 0.798 vs. MEM: 0.37%, P = 0.0007; NAÏVE vs DN 0.61% vs MEM: 0.37%, P = 0.0011; Figures 2C , S2C ).

Like what we observed in the case of standard CLL-SLS, we did not detect any differences in the frequencies of satellite CLL-SLS when comparing NAÏVE from BM and PBMC samples ( Figure 2D ).

Thus, we identified satellite CLL-SLS in the B-cell repertoires of healthy donors. However, unlike CLL patients, where satellite subsets comprised only 3% of the total cohort versus 13.5% assigned to major subsets (10), satellite CLL-SLS in normal individuals were found at higher frequencies and numbers than standard CLL-SLS. This difference might be the result of their immunogenetic properties together with the selection forces shaping the normal B cell repertoire.

Finally, the highest frequency of satellite CLL-SLS was found in NAÏVE B cells coming from BM and PBMC, in contrast to what we observed in the context of standard CLL-SLS where the highest frequencies were found in PB/PC and rcMZ. Thus, similarly to what observed for standard CLL-SLS, the presence of satellite CLL-SLS distributed along all B cell differentiation axis further strengthens the idea that CLL-like BCRs are not subjected to elimination by central tolerance mechanisms and are present in the normal B-cell repertoire.

The second tier of analysis was directed at understanding to which specific standard stereotyped CLL subset the CLL-SLS belong, and at determining if the distribution of the CLL-SLS differs from the standard stereotyped sequences observed in cohorts of CLL patient ( Figures 3A–D ). When comparing the distribution of standard CLL-SLS resembling the 29 most prominent subsets identified in patients with CLL across the 3 different cell sources ( Figures 3B–D ), we found at least one sequence belonging to each of the major CLL stereotyped subset from each site apart from subsets #16 and #7D3 ( Table S3 ). When examining CB B cells, 39.8% of the CLL-SLSs were members of subset #14 and 13.01% of subset #73 ( Figure 3B ). Together these two subsets made up ~50% of all the CLL-SLS in the CB. Interestingly, these subsets were not the most prevalent ones among the major CLL subsets ( Figure 3A ). Finding these at increased frequencies in the CB, where most B cells have not encountered foreign antigen and have not undergone somatic hypermutation (SHM), was also surprising because, among CLL patients, these two subsets display mutated IGHVs. However, for subsets # 73 and 14, most of the CLL-SLS sequences were IGHV-unmutated (75% and 84%, respectively).

In the BM, the most recurrent subsets were # 148B, #14, #28A, #12 and #1 (15.8%, 12.3%, 12.1%, 7.4% and 5.1 respectively; Figure 3C ). Like the CB, these abundant subsets were not the most frequently found in patients ( Figure 3A ). CLL B cells belonging to subsets #1 and #28A are part of the IGHV-unmutated (U-CLL) group and are encoded by IGHV1-69 (subset #1) and IGHV1-2 (subset #28A), respectively. Conversely, CLLs falling into subsets #148B and #14 express IGHV2-5 and IGHV4-4, respectively, and belong to the IGHV-mutated (M-CLL) group. For those CLL-SLS identified among PBMCs, again subsets # 148b and #14 were highly represented (25.8% and 13%), followed by subsets # 73, #28A and #7C2 (10.9%, 6.0, and 5.8% of the total CLL-SLS) ( Figure 3D ).

Finally, when comparing among the three tissues, we found that BM and PBMC display a very similar distribution of CLL-SLS with a statistically significant difference in the frequencies only for subset #148b ( Figure S3A ).

In conclusion, the distribution of certain standard CLL-SLS found in normal individuals can be dramatically different from the distribution found in CLL patients. However, since little is known about the immunogenetic properties of most of the subsets found frequently in the normal repertoire, it is difficult to speculate about the underlying biological process or force that preferentially selected for a B cell expressing a particular stereotyped sequence. In addition, we did not observe major differences in the distribution of CLL-SLS subsets when comparing BM and PMBC samples. Nevertheless, we noticed a discrete difference when comparing these to CB sample, where 50% of CLL-SLS are attributed to two subsets. Finding such high-level restriction in the CB is not necessarily paradoxical, given the absence of terminal deoxynucleotidyl transferase expression with the consequent lack of non-templated additions during the neonatal period, which often significantly limits diversity in the VH CDR3.

We next analyzed the distribution of satellite CLL-SLS as above. This identified several differences from that found for the standard CLL-SLS. For example, in CB, 50% of CLL-SLS were satellites of subsets # 77, #73, #31, #12, #2, and #169 ( Figure 4A and Table S3 ). Within the BM, we also found that the most recurrent subsets were # 2, #169, #12, and #73 (17.5%, 12.9%, 8.8%, 9.1%, respectively; Figure 4B ). Similarly, for the peripheral blood samples, subsets # 2 and #169 were among the most frequent (19.1% and 21.1%, respectively), making up to 40% of the total sequences ( Figures 4C , S4A ). It is noteworthy that the satellite sequence for subset #169 was found among the top-ranking group for each of the three sites of collection.

Furthermore, when we compared the frequencies of standard and satellites subsets, we observed two different scenarios. For some specific subsets we found a significant increase in satellite CLL-SLS compared to the standard ones. For example, #169 and its companion #2 comprise 30 to 40% of all the sequences in PBCM and BM whereas only a minor fraction of standard CLL-SLS were attributed to subset #2 and #169 in both tissues examined (BM: #2, 2.9% and #169, 2.5%; PBMC: #2, 4.8% and #169, 2.3%) ( Figures 4D, E ). However, we also found an opposite behavior for some of most frequent standard subsets that instead were underrepresented in the satellites, e.g., # 148b, 28A, and 14 ( Figures 4D, E ).

Overall, the satellite CLL-SLS subsets enriched in specific B-cell subpopulations can differ significantly from the standard CLL-SLS, especially in PBMC and BM. Among individual satellite subsets, # 2 and #169 seem to be over-represented in both PBCM and BM, and they both significantly increase in frequencies when compared to their standard subset counterparts. Alternatively, some of the most frequent subsets in the standard CLL-SLS analysis were underrepresented in the satellite analysis. This last observation is in accord to what is observed in CLL patients, where satellite represent only a minor fraction of the total subset sequences. Instead, the significant increase in satellite CLL-SLS attributed to #2 and #169 suggest a selection pressure of these type of sequences compared to the standard counterpart that only occurs in the normal repertoire.

Like CLL clones (4), stereotyped subsets can be segregated based on IGHV-mutation status (10,28). Therefore, we next determined if the U-CLL CLL-SLS were enriched in the earlier stages of B-cell development and if the M-CLL CLL-SLS were enriched in the later stages of B-cell maturation ( Figures 5B, C ). Since the incidence of standard CLL-SLS differed significantly from satellite CLL-SLS, for the following analysis, we only analyzed the former ( Figure 5A ).

Since we did not find proof for negative selection of any individual CLL-SLS at the early maturation stages in the BM and at the transitional stage in the blood ( Figure 5B ), we checked for such evidence at later stages of maturation, by examining specific patterns of distribution of individual CLL-SLS. This analysis revealed that CLL-SLS belonging to the IGHV-mutated subsets were found at different frequencies than the IGHV-unmutated among the various antigen-experienced B cell subsets ( Figures 5B, C ).

Using subsets #2 and #169 as examples of IGHV-mutated standard subsets, for subset #2, there was a statistically significant enrichment in NAIVE and rcMZ compared to B cells at the earlier stages of development; a significant level of different was not found with B cells at the later stages of maturation, although there was a trend in this regard. The principal was the same for subset #169, except there were also statistically significant differences for the MEM and DN. There was a similar trend for enrichment in PC/PB, but this did not reach statistical significance. Since subsets # 2 and #169 are members of the IGHV-mutated subtype, this pattern of distribution in the mature stages of B-cell development seems consistent. Similar trends were found for M-CLL subsets # 73, #77, and #188 ( Figure S5A ).

In contrast, those CLL-SLS expressing unmutated IGHVs were detected in B-cell subpopulations at the early stages of development and in TRANS and NAIVE; they were virtually absent from the IGHV-mutated MEM and rcMZ populations. Examples are CLL-SLS belonging to U-CLL subsets # 12 and 31 (among the most frequent ones, Figure 5D ); a similar trend was observed for subsets # 8, 8B, and 59 ( Figure S5B ). Surprisingly, however, CLL-SLS attributed to U-CLL subsets were significantly enriched in DN B cells, an Ag-experienced, usually IGHV-mutated B cell subpopulation that plays a role in auto-reactive conditions and infectious disease (29). The reason for the abundance in this IGHV-mutated B-cell subpopulation is not obvious.

In summary, when examining individual CLL-SLS subsets, CLL-SLS bearing IGHV-mutated IGs are most frequent in naïve and Ag-experienced B cells, whereas CLL-SLS attributed to IGHV-unmutated subsets in general are restricted to the early stages of B-cell maturation and to naive B cells, with DN B cells being the exception.

Next, we examined if the CLL-SLS enriched at different B-cell maturation stages in PBMCs exhibited the same IGHV-mutation status as that found in standard CLL stereotyped subsets ( Figures 6A, B ).

First, we analyzed CLL-SLS attributed to U-CLL subsets, observing that ~94% of CLL-SLS found in TRANS and NAIVE B cells matched the IGHV-mutation status of the stereotyped subsets in CLL patients ( Figure 6A ). This situation changed when examining the more mature B-cell subsets in PBMC. In the case of MZ and DN, there was a fall in the frequencies of IGHV-unmutated CLL-SLS matching the IGHV-mutation status of the patient-defined CLL subsets (61% and 72%, Figure 6A ). This drop was most evident for CLL-SLS in MEM B cells, where 72% displayed SHM and only 28% matched the original SHM status ( Figure 6A ).

When examining CLL-SLS belonging to M-CLL subsets ( Figure 6B ), we found the opposite: only a minor fraction present in the NAIVE and TRANS B-cell compartments displays SHMs and hence matches the IGHV-mutation status in CLL (8% and 5.7%, respectively). The majority in MEM (88.4%), MZ (68%), and DN (54.5%) were mutated and thus in agreement with CLL SHM status.

Thus, CLL-SLS that are IGHV-unmutated in patients are found more often in the normal repertoire among B cells at the earlier stages of B-cell maturation, which have usually not interacted with foreign antigens and therefore have not undergone SHM and developed IGV mutations. In contrast, CLL-SLS that are IGHV-mutated in patients are found more often in the normal repertoire among B-cell subpopulations at the later stages of B-cell maturation when SHM is common. So, in general, these results are consistent with U-CLL clones originating from and TRANS, NAÏVE, and MZ B-cell populations, and M-CLL clones coming from more antigen-experienced subsets such as MZ, MEM, and DN. However, the fact that some U-CLL-associated CLL-SLS can bear somatic IGHV mutations and can be enriched in antigen-experienced B cells suggests a positive selective driving away from the corresponding CLL-associated IGHV-mutation phenotype for these CLL-SLS.

We next expanded this type of analysis to individual CLL-SLS from the major CLL stereotyped subsets, examining the distribution patterns of U-CLL-like and M-CLL-like SLS sequences among the different B cell subpopulations ( Figures 6C–F ). This revealed 7 distinct patterns ( Figures 6C–F ). In the case of U-CLL subsets exhibiting pattern 1, the vast majority of IGHV-unmutated CLL-SLS belonging to subsets # 6, #7C2, #8, and #8B match the IGHV-mutation status of the CLL patients, independent of the B-cell subset in which they were found ( Figures 6C, D ). Regarding the other patterns, most were still characterized by high frequencies of CLL-SLS matching the SHM pattern of CLL patients; these were mainly restricted to the NAIVE and TRANS B cell stages. However, there were obvious differences in the other B-cell subsets. For example, in the case of subsets belonging to pattern 3 (subsets #202, #28A, #59, and #99), only a median of 65% DN, 32.9% MZ, and 7.7% MEM CLL-SLS matched the original CLL IGHV mutation status. Similarly, by looking at CLL-SLS attributed to M-CLL subsets, we observed that the most frequent ones found in our analysis, such as # 2, #73, and #169, group together in pattern 6 ( Figures 6E, F ). In this case, MEM display the highest median frequencies of CLL-SLS matching the original CLL IGHV mutation, whereas there was a progressive decline in both MZ and DN populations. In other cases, such as those subsets belonging to pattern 5, both MZ and MEM were similarly enriched in CLL-SLS matching the IGHV mutation status of the original CLL subsets ( Figures 6E, F ).

Thus, when examining the SHM status of individual CLL-SLS, different patterns can be identified. These patterns suggest that specific CLL stereotyped subsets might originate from particular subpopulations in the normal B-cell repertoire.

The preceding data indicate that CLL-SLS are present in normal, healthy people, and that the frequencies at which these exist in the normal B-cell repertoire do not appear to decrease when progressing from developing to mature B cells. Since CLL B cells appear to derive from autoreactive precursors (13) and CLL IGs are often autoreactive (13–17), finding CLL-SLS in healthy individuals at all stages of maturation is not consistent with the elimination of autoreactive BCRs/IGs from the normal B-lymphocyte repertoire (18). Alternatively, the IGK/LV-J rearrangements paired with the CLL-SLS IGHV-D-J could differ from those in CLL cells, thereby neutralizing or preventing autoreactivity. Since several of the major stereotyped subsets display IGK/LV gene restrictions, we isolated cells based on membrane L chain expression (κ or λ) and defined the frequencies at which certain specific CLL-SLS IGHV-D-J rearrangements were found in the κ and λ chain populations.

As representatives of those subsets that display IGK/LV L chain restriction, we examined subset #2 and its companion subset #169, both of which always express lambda light chains encoded by IGLV3-21. Notably, when sorting NAIVE, MZ, and MEM cells from a normal, healthy person based on L chain isotype, we observed a similar distribution of CLL-SLS subsets # 2 and 169 within the κ and λ chain expressing B-cell populations. Thus, for these subsets there was a lack of skewing toward λ light chain use ( Figure 7A ).

We then looked at CLL-SLS subsets characterized by the expression of IGHV4-34, such as subsets #201, #77, and #4 ( Figure 7B ). For subset #201, which always uses IGLVλ1-44, we found CLL-SLS only in the λ-expressing fraction of MEM B cells. Whereas for subsets # 77 and 4, whose IGHVs always pair with IGLV10-54 or IGKV2-30, respectively, we did not observe a particular bias toward the usage of a specific light chain type ( Figure 7A ). Thus, we found the CLL-SLS in the appropriate light chain population depending on the specific subset. However, by taking this approach, we could not determine the specific IG light chain expressed by CLL-SLS.

To overcome this challenge, we performed single cell IGHV-D-J sequencing of B cells in PBMCs that express IGHV4-34 by sorting using the 9G4 mAb which reacts specifically with IGs bearing this gene (30). This strategy allowed us to enrich for subsets that use IGHV4-34 and, at same time, to identify the matching L chain and its DNA sequence. After sorting 20,000 NAIVE, 4,274 MZ, 9,547 MEM, and 340 DN B cells ( Figure S6A ), we identified IGHV-D-J sequences from 5,406 NAIVE, 1,205 MZ, 2,386 MEM, and 122 DN cells. A large majority of clonotypes identified expressed IGHV4-34, as depicted by V to J heatmap ( Figure S6B ). In this way, we identified CLL-SLS representative of several stereotyped subsets associated with IGHV4-34 ( Figure 7C ) in 110 Naïve, 2 rcMZ, 7 MEM, and 5 DN cells.

When looking specifically at CLL-SLS sequences resembling subset # 201, we identified 6 such sequences, all in naïve B cells. However, none were paired with a λ light chain ( Figure 7D ). However, although subset #201 CLL clones show a strong restriction for λ light chain use, there are a few identified instances where subset #201 stereotyped sequences were paired with the κ light chain gene IGKV4-1. Notably, one of our CLL-SLS sequences attributed to subset #201 had an IGKV4-1 gene partner. Moreover, the VK CDR3 sequence of that cell was remarkably like the CLL stereotype and the CLL-SLS ( Figure 7D ). Thus, this apparently normal B cell could be a precursor to a subset #201 CLL clone.

Similarly, when examining CLL-SLS resembling subset #4, we found an IGHV-D-J gene rearrangement paired with IGKV 2-30, the gene most often co-expressed in this leukemic subset ( Figures 7E, F ). However, this CLL-SLS did not bear IGHV mutations, which all subset #4 rearrangements have. Nor did it carry a characteristic amino acid at a specific position in the IGKV-J rearrangement corresponding to the standard stereotyped CLL BCRs, i.e., an aspartic acid at position 66 in the VK FR3 that is introduced by SHM in CLL cells. Thus, this apparent precursor of standard CLL stereotyped subsets # 4 does exhibit the complete subset #4 CLL sequence.