B cells were obtained from buffy coats of healthy adult donors older than 18 years from the Hemocentro Distrital, Bogotá, Colombia, that signed informed consent form approved by the ethics committee of the Pontificia Universidad Javeriana. PBMCs were separated by Ficoll-Hypaque and B cells were enriched by negative selection using the RosetteSep enrichment kit (Stem cell Technologies, Vancouver, BC, Canada#15024). Cells were then stained with LIVE/DEAD^TM Fixable Aqua Dead Cell Stain Kit (Invitrogen Molecular Probes, Waltham, MA, United States # L34957) and subsequently with the following previously tittered anti-human monoclonal antibodies (all from BD Bioscience, San Jose, CA, United States, unless otherwise specified): CD3 (V500, clone UCHT1), CD14 (V500, clone M5E2), CD19 (APC-H7, clone SJ25C1), CD27 (PE-CF594, clone M-T271), CD38 (AF-700, clone HIT2), IgD (VB421, clone IA6-2), and IgM (PerCPCy5.5 clone G20-127). Five subpopulations were obtained by means of cell sorting by fluorescence (FACS): naïve B cells (IgM⁺IgD⁺CD27^–), IgM only (IgM⁺IgD^–CD27⁺), and class-switched Bmem cells (IgM^–IgD^–CD27⁺) as biological controls, along with IgM^hi (IgM^{+ +}IgD⁺CD27⁺) and IgM^lo (IgM⁺IgD^{+ +}CD27⁺) B cells (Supplementary Figures 1A,B). Between 150.000 and 1.000.000 total cells were obtained per subpopulation. All cell subpopulations obtained had a post sorting purity greater than 95% (Supplementary Figure 3). Plasmablasts (CD19⁺CD38^{+ +}CD27^{+ +}) were excluded from the analysis. Samples were sorted on a FACS ARIA IIu (BD Bioscience, San Jose, CA, United States) (36).

PBMCs were separated by Ficoll-Hypaque from healthy adult donors that signed an informed consent form and stained with antibodies described in section “Samples and Cell Sorting.” Additionally, samples were stained with anti-human monoclonal antibodies (BD Bioscience, San Jose, CA, United States): CD40 (PeCy7, clone 5C3), TLR9 (PE, clone eB72-1665), CD95 (PE, clone DX2), CD73 (PE, clone AD2), CD21 (PE, clone Bly-4), IL-21-R (PE, clone 17A12), CD45RO (APC, clone uch1), CD11c (APC, clone S-HCL-3), BAFF-R (APC, clone 11C1), CD43 (APC, clone eBio84-3C1), PD-1 (PE, clone J105), IL2-Rβ-CD122 (PE, clone Mik-β3), CD24 (PE, clone ML5), CD1c (PE, clone L161), CD184 (APC, clone 1265), CD69 (PE-Cy7, clone FN50), CD23 (APC, clone EBVC5-5), CD11b (AF-488, clone ICRF44), CD5 (PE, clone UCHT2), and CD45RB (PE, clone MEM55) for 30 min on ice. For ABCB1 transporter expression experiments PBMC were incubated with MTG at 100 nM (Thermo Fisher Scientific, Waltham, MA, United States #M7514) for 30 min at 37°C (37). Phenotype analysis was done on five unsorted B cells subpopulations described before. For all markers analyzed, a fluorescence minus one (FMO) was included as a negative control during standardization of the staining protocol (36). Samples were acquired on a LSR Fortessa or FACs ARIA II (BD Bioscience) and analyzed using FlowJo v.9.3.2. The data were represented as an integrated mean fluorescent intensity (iMFI = percentage of positive cells × MFI), a metric measure that combines the percentage of cells expressing a marker) and the MFI of the same marker. Non-parametric tests were used for comparisons.

For some experiments, frequencies of RV Bmem cells were determined as previously reported (7). Green fluorescent protein (GFP) coupled to RV VP6/VP2 virus-like particles (VLP) were generated as previously described (38, 39). Total B cells were washed once with PBS, 2% fetal bovine serum, 0.02% sodium azide, and then incubated with 0.8 μg/test GFP-VLPs for 45 min at 4°C in the dark (7).

Sort purified B cells subsets were labeled with carboxyfluorescein succinimidyl ester (CFSE) (40). B cells (0.4–1.0 × 10⁶ cells/ml) were washed twice with sterile PBS and stained with CFDA-SE (Cell-TraceTM CFSE Cell Proliferation Kit, Invitrogen Molecular Probes, Waltham, MA, United States # C34554) for 5 min at room temperature, protected from light. After being washed three times with 10 ml of PBS FBS 5%, naïve B cells subpopulations were stimulated with 2.5 μg/ml of CpG (ODN 2006; InvivoGen, San Diego, CA, United States), 10 ng/ml human recombinant IL–2, 10 ng/ml human recombinant IL–6, 15 ng/ml IL–10 (all cytokines from PreproTech, Rocky Hill, New Jersey), and NIH 3T3 murine fibroblasts (ATCC, Manassas, VA, United States). The NIH 3T3 feeder cells were irradiated with 3,000 rads (Radiotherapy unit, Centro Javeriano de Oncología, Bogotá) and then used at a concentration of 5.000 cells/well. B cells (20.000 in 200 μl of complete medium with 10% FBS) were cultured at 37°C with 5% CO₂ in flat bottom 48 well plates for different periods. At the end of the cultures, cells were washed twice with sterile PBS and stained on ice with LIVE/DEAD^TM Fixable Aqua Dead Cell Stain Kit (Invitrogen Molecular Probes, Waltham, MA, United States # L34957) and subsequently with antibodies (all from BD Bioscience, San Jose, CA, United States) against (labeled): CD19 (APC-H7), CD38 (PerCp-Cy5.5), and CD27 (PE-CF594). After washing, cells were resuspended in 100 μl PBS, 0.5% BSA (Sigma-Aldrich) plus 250 μl of cytofix/cytoperm (BD Pharmingen) and incubated for 20 min at 4°C. After washing twice with perm/wash (BD Pharmingen) cells were stained with antibodies against IgM APC (polyclonal serum, Jackson ImmunoResearch) or IgG APC (clone G18-145, BD Bioscience, San Jose, CA, United States) for 30 min on ice. Samples were acquired on an LSR Fortessa (BD Bioscience, San Jose, CA, United States) and analyzed using FlowJo v.9.3.2 (FlowJo, LLC, OR, United States). Differences among groups were determined using Kruskal-Wallis test followed by a Wilcoxon signed-rank test using GraphPad Prism software (36).

Sort purified B cells from each subpopulation (naïve B cells: 800.000–1.000.000, IgM^lo B cells: 325.000–380.000, IgM^hi B cells: 410.000–500.000, IgM only Bmem cells: 172.000–220.000, and class-switched Bmem cells: 700.000–1.000.000) were recovered in cold TRIZOL (Invitrogen Molecular Probes, Waltham, MA, United States # 15596026). RNA was extracted using the RNeasy micro kit (Qiagen GmbH, Hilden, Germany # 74004). Samples were re-suspended in RNase-free water and their purity and integrity were verified by spectrophotometry in nanodrop 2000/2000c (Thermo Fisher Scientific, Waltham, MA, United States # ND-2000) and electrophoresis using Bionalyzer 2100 (Agilent Technologies, Santa Clara, CA, United States # G2939BA) and RNA pico 6000 kit (Agilent Technologies, Santa Clara, CA, United States # 5067-1513). Only RNA samples with a RNA integrity number (RIN) greater than 7 were used.

The relative expression of HOPX and SOX7 was determined from sorted B cells. qRT-PCR for the human transcripts was performed with specific TaqMan gene expression assays HOPX (Hs041888695_m1), SOX7 (Hs00846731_s1), and B2M (Hs00984230_m1) as a housekeeping gene, all by designed by Applied Biosystems (CA, United States) (26). Taqman gene expression assays were used at a concentration of 0.6X in a final volume of 10 μl with the Lightcycler 480 RNA Master Hydrolysis probes kit (Roche Basel, Switzerland # 04991885001). The reaction was conducted with an annealing temperature of 60°C in 45 cycles using a LightCycler^® Nano instrument (Roche Basel, Switzerland) real time-PCR system. The PCR products were quantified and purified with a Wizard^® SV Gel and PCR Clean-Up System (Promega, Madison, WI, United States) and used to generate a calibration curve. Four serial log 10 dilutions ranging from 1 × 10⁹ to 1 × 10³ copies/l were employed, obtaining an efficiency of 100%. Each sample was analyzed in duplicate, and a negative control with water, instead of cDNA, was included in each analysis. Standard curve based method for relative expression was utilized (41). Differences among groups were determined with a p-value < 0.05 using Kruskal-Wallis test followed by a Wilcoxon signed-rank test in GraphPad Prism software.

A total of 50 ng of RNA from each B cell subpopulation from the same individuals (3 donors in total) was used for expression analysis of transcripts with the human Clariom D array (Affymetrix, Santa Clara, CA. United States # 902923). Affymetrix Clariom D CEL files were normalized to produce probeset-level expression values using Expression Console (version 1.4.1.46) with the Robust Multiarray Average (RMA) algorithm (42) and the default probesets defined by Affymetrix. Clariom D CEL files were re-normalized to produce Entrez-Gene-specific expression values using the implementation of the Robust Multiarray Average (RMA) [1] in the affy package (version 1.36.1) (43) included in the Bioconductor software suite (version 2.12) (43) and an Entrez Gene-specific probeset mapping (version 17.0.0) from the Molecular and Behavioral Neuroscience Institute (Brainarray) at the University of Michigan (44)^cpsbib1. An unsupervised Principal Component Analysis (PCA) was performed using the prcomp R function with expression values that had been z-normalized across all samples (i.e., set to a mean of zero and a standard deviation of one). Before z-normalization, the expression values were adjusted by donor by taking the residuals of linear models created using the lmFit function in the limma package (version 3.14.4), treating donor as a fixed effect. Linear mixed-effects modeling and the associated analysis of variance were performed using the anova.lme function in the nlme package (version 3.1–131). Pairwise differential expression between groups was assessed using Student’s paired two-sample t-test, i.e., performing Student’s t-test on the coefficients of a linear model created using the lmFit function in the limma package, treating donor as a fixed effect. Correction for multiple hypothesis testing was accomplished using the Benjamini-Hochberg false discovery rate (FDR) (45). All analyses performed using the Affymetrix-normalized data were carried out using the R environment for statistical computing (version 2.15.1).

GSEA (version 2.2.1) (46) was used to identify biological terms, pathways, and processes that are coordinately up- or down-regulated within each pairwise comparison. The Entrez Gene identifiers of the genes interrogated by the array (as obtained from the Entrez-Gene-specific re-normalization) were ranked according to the paired Student t statistic computed between IgM^hi and IgM^lo B cells. This ranked list was then used to perform pre-ranked GSEA analyses (default parameters with random seed 1234) using the Entrez Gene versions of the Hallmark, Biocarta, KEGG, Reactome, Gene Ontology (GO), transcription factor and microRNA motif, and immunologic signature gene sets obtained from the Molecular Signatures Database (MSigDB), version 6.0 (47).

All analyses performed using the Brainarray-normalized data were carried out using the R environment for statistical computing (version 3.2.3).

Total RNA obtained from circulating purified B cells subpopulations (24–156 ng) of three adult donors was amplified using the 5’-RACE-PCR (48, 49). Briefly, an oligodT is added to the initial RNA at the 3’ end and then reverse transcription is performed using the SuperScript III reverse transcriptase (200 U/μl) (Invitrogen Molecular Probes, Waltham, MA, United States # 18080044) in the presence of the primer illuRACE (5’-TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGmGmGrG-3’, 12 pmol/μl). This is a hybrid primer where m means 2’O methyl and rG is a guanine with sugar ribose. Subsequently, the sample is subjected to amplification with a universal primer that binds 5’ of the FpAmpl cDNA (5’-TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGGGG-3’, 5 or 2.5 pmol/μl) and another hybrid containing the sequence of the constant region of IgM (5’-GTCTCGTGGGC TCGGAGATGTGTATAAGAGACAGAAAGGGTTGGGGCGG ATGCACT-3’, 5 pmol/μl) or IgG (5’-GTCTCGTGGGCTCGG AGATGTGTATAAGAGACAGACCGATGGGCCCTTGGTG-3’, 5 pmol/μl). Products obtained from this amplification were purified from gel using the Qiaquick 250 kit (Qiagen GmbH, Hilden, Germany # 28706) and subjected to a second amplification using specific sequence primers for IgM (5’-AAAGGGTTGGGGCGGATGCACT-3’, 5 pmol/μl) or IgG (5’-ACCGATGGGCCCTTGGTG-3’, 5 pmol/μl) in a nested PCR procedure. The products of the second amplification were purified using the AMPure XP PCR purification kit (Beckman coulter, Brea, CA # A63881). For Kappa light chain, a first amplification was used with sequence-specific primers (5’-ACAGATGGTGCAGCCAC-3’, 2.5 pmol/μl) and a second amplification with a first hybrid (5’-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGACA GATGGTGCAGCCAC-3’, 2.5 pmol/μl). The enzyme Accuzyme DNA polymerase high fidelity (Bioline, London, United Kingdom # BIO-21051) was used.

The amplification cycles used were the following: first PCR: 95°C 3 min, 20 cycles of 95°C 30 s, 60°C 30 s, 68°C 30 s, and 68°C 5 min. Second PCR: 95°C 3 min, 20 cycles of 95°C 30 s, 56°C 30 s, 68°C 30 s, and 68°C 5 min. In all cases water was used as a negative control.

The PCR products were quantified by fluorometry using the Qubit dsDNA HS Assay kit (Invitrogen Molecular Probes U.S MA Waltham # Q32851) before joining adapter Nextera XT index 24 indices -96 samples (Illumina, San Diego, CA, United States # 15032353). Kappa light (IGK) and heavy chain (IGH) products were joined to different adapters for sequencing two chains at the same time per subset. Adapters were joined using the following PCR protocol: 72°C 3 min, 95°C 30 s, 8 cycles of 95°C 10 s, 55°C 30 s, 72°C 30 s, and 72°C 5 min.

The final sample, at 6 nM, was treated with the MiSeq Reagent Kit v3 (Illumina, San Diego, CA, United States # 15043894) and finally ran on the sequencer MiSeq (Illumina, San Diego, CA, United States # SY-410-1003) in a paired manner.

Raw sequencing reads were subjected to a quality control process using the FastQC program². The Ig gene repertoire of each subpopulation was subsequently reconstructed with IgRec using single reads as input (50). Using the VJ FINDER tool, IgRec filters out contaminating reads by alignment with the Ig germline IMGT database. Also, VJ FINDER filters out reads that do not fully cover V(D)J region since their error correction can result in false inferences. Large clusters (sequences with more than 5 reads) were used as proxy of lineages and were analyzed with the IMGT/HighVQuest tool³ (51, 52), to obtain the following metrics: (1) # of productive IGH and IGK chain sequences; (2) % of mutations in the IGHV and IGKV regions (corresponding to FR1, CDR1, FR2, CDR2, FR3, and part of the CDR3, according to IMGT/HighVQuest); (3) % of IGHV and IGKV gene and family usage; (4) % IGHJ and IGKJ gene usage. The HCDR3 length, the relative frequency (number of reads per lineage/total reads in lineages ≥ 5 reads), N1 and N2 length, P (P3’V, P5’D, P3’D, and P5’J) length, and % identity of IGHV-IGHJ with the corresponding germline gene were used as input to perform a Principal Component Analysis (PCA) to compare the different subpopulations using the Orange3 v3.24.1 software (53). Additionally, the length and other properties of HCDR3 and KCDR3 were determined using the Peptides package in R⁴.

To cluster lineages into clonotypes, we used ImmunediveRsity software (54) to create the IGH clonotypes at 100% nucleotide identity in the HCDR3. Shared IGH clonotypes within B cell subpopulations were searched at 100% of identity with R home scripts.

Differences between subpopulations and donors were assessed with the non-parametric Fisher, Wilcoxon, and Kruskal-Wallis tests. Only statistically significant differences present independently in 3 of 3 donors were considered for each of the parameters evaluated.

We have previously used GFP-VLP to identify and study human RV-B cells (4, 7, 8). We found that RV-IgM⁺IgD⁺CD27⁺ B cells can be grouped into two subpopulations: one with high levels of IgM and low levels of IgD (IgM^hi) and another with low expression of IgM and high expression of IgD (IgM^lo). Here, using an improved staining combination of antibodies (Supplementary Figures 1A,B), we found that although RV-B cells are enriched in both subpopulations, IgM^hi and IgM^lo B cells, most RV-IgM B cells are IgM^hi (twofold change) (Supplementary Figure 1C). Also, IgM^hi B cells made up roughly two-thirds and IgM^lo B cells one-third of total IgM⁺IgD⁺CD27⁺ B cells (Supplementary Figure 1C). These results confirmed our previous findings (7, 8) and led us to hypothesize that RV specificity discriminates two human circulating IgM B cells in total IgM⁺IgD⁺CD27⁺ B cells of healthy donors. After excluding plamablasts, a median of 3.9 and 6.62% of CD19 B cells are IgM^lo and IgM^hi B cells, respectively (Supplementary Figure 1D). Moreover, we noted these B cell subpopulations were notoriously distinct in a patient with a 22q11.2 distal deletion syndrome associated with a p. Ala7dup variant in the MAPK1 gene (Supplementary Figure 2) (55). Here, we sought to further characterize these two subpopulations of IgM B cells.

Employing a similar but not identical staining strategy, other investigators determined that IgM^lo B cells differ from IgM^hi B cells by the high expression of several markers (56). To compare the phenotype of IgM^hi and IgM^lo B cells, we evaluated 26 markers previously evaluated by us (36) and others (27, 37, 56) in PBMCs, and used naïve B cells, IgM only Bmem cells, and class-switched Bmem cells for comparison. Eleven markers (excluding IgM and IgD) showed statistically significant differences between IgM^hi and IgM^lo subsets (Supplementary Table 1 and Figures 1A,B). Comparatively higher expression of CD27, CD69, CD1c, CD45RB (MEM55), and MTG was observed in IgM^hi B cells, and higher expression of CD38, IL21R, BAFF-R, CD23, CD184, and CD5 in IgM^lo B cells (Supplementary Table 1 and Figure 1A). An unsupervised clustering analysis including the iMFI of the 11 markers that differ between both subpopulations separated naïve B cells from all other subsets and group all the IgM expressing B cells separately from class-switched Bmem cells (Figure 1B). However, IgM^hi B cells and IgM only Bmem cells formed a separate subgroup from IgM^lo B cells (Figure 1B). It is worth noting that IgM^lo B cells expressed lower levels of CD27, CD45RB, and MTG (Figure 1B). In addition, IgM^hi B cells were enriched in CD45RB⁺MTG⁺ B cells, while IgM^lo B cells presented higher frequencies of CD45RB⁺MTG^–, CD45RB^–MTG⁺, and CD45RB^–MTG^– cells (Kruskal-Wallis p < 0.0001 to p < 0.0008, Wilcoxon test p = 0.01 in all cases) (Figure 1C and data not shown). Thus, IgM^lo and IgM^hi B cells differ phenotypically, with IgM^lo being a less differentiated B cells subset.

Other investigators have reported the transcriptomic profile of circulating IgM⁺IgD⁺CD27⁺ B cells purified, in most cases, with a similar sorting strategy used in this work (26, 33, 57). To compare the transcriptome of IgM^hi and IgM^lo B cells, we sort purified these B cells subsets from the same individual, three blood donors in total, and the RNA extracted from the B cells was analyzed with Clariom D microarrays, as described in section “Materials and Methods.” RNA extracted from sort purified naïve B cells, IgM only Bmem cells, and class-switched Bmem cells were run in parallel for comparison. The expression profiles of IgM^hi and IgM^lo B cells were similar, and only 78 genes differed between them with an FDRq LME-ANOVA < 0.001 and a p < 0.01 (Figure 2A and Supplementary Table 2). A clustering analysis considering the expression of these 78 genes showed that IgM^hi B cells were similar to IgM only and class-switched Bmem cells (B cells that have passed through germinal centers), while the gene profile of IgM^lo B cells differed from those of these three subsets and was somewhat closer to the gene profile of naïve B cells (Figure 2A). A PCA of 39,337 Brainarray renormalized genes showed that IgM^hi and IgM^lo B cells clustered with IgM only Bmem cells and diverged from naïve B cells and class-switched Bmem cells (Figure 2B). However, IgM^hi B cells tended to be closer to class-switched Bmem cells than IgM^lo B cells (Figure 2B).

A gene set enrichment analysis (GSEA) revealed 159 biological processes enriched in IgM^hi B cells with respect to IgM^lo B cells with an FDR < 0.25, 10 of which corresponded to B cells related processes (data not shown). Three of these processes suggest that IgM^hi B cells, relative to IgM^lo B cells, resemble Bmem cells, because they expressed genes downregulated in naïve B cells with respect to Bmem cells (GSE12366 NAIVE VS MEMORY B CELL DN). Moreover, IgM^hi B cells appear to be more activated, in comparison with IgM^lo B cells, given that they are related to sets of genes associated with the negative regulation of B cells activation (GO NEGATIVE REGULATION OF B CELL ACTIVATION) and the negative regulation of antigen receptor signaling (GO NEGATIVE REGULATION OF ANTIGEN RECEPTOR MEDIATED SIGNALING PATHWAY).

Previously, a study reveled that MZB cells from spleen and IgM⁺IgD⁺CD27⁺ B cells from blood have a marked positive regulation of the SOX7 transcription factor, and negative regulation of the COCH gene and the TOX and HOPX transcription factors, with respect to class-switched Bmem cells (26). Also, in samples from spleens of children, SOX7 was expressed at very low levels in naïve B cells and at low levels in MZPc, while HOPX was expressed at intermediate levels in naïve B cells, compared with class-switched Bmem cells, and expressed at low levels in MZPc (26). We performed qRT-PCR for these genes with purified B cells of five different blood donors. As expected, class-switched Bmem cells (26), but also IgM only Bmem cells, IgM^lo B cells, and naïve B cells expressed more HOPX than IgM^hi B cells (Figure 2C). In contrast, the levels of SOX7 on IgM^hi B cells were higher than those on naïve B cells and also on class-switched Bmem cells (although only statistically significant in the former case) (Figure 2C). Thus, the patterns of expression of HOPX and SOX7 genes of IgM^hi B cells are analogous to those reported for MZB cells, while the patterns of expression of IgM^lo B cells were intermediate between those of naïve and IgM^hi B cells (26).

Various investigators, including us, have previously determined that ex vivo stimulation with CpG, cytokines (IL-2, IL-6, and IL-10), and irradiated NIH 3T3 fibroblasts (as feeder cells) is optimal to induce proliferation and differentiation to ASC of purified B cells (4, 58). Moreover, using this stimulus, we have shown that only a subset of sort purified IgM⁺IgD⁺CD27⁺ B cells proliferate, change isotype, and differentiate to ASC (36).

To evaluate the capacity of IgM^hi and IgM^lo B cells to proliferate, differentiate to ASC, and switch isotype, we stimulated sort purified B cells with CpG, cytokines (IL-2, IL-6, and IL-10), and irradiated NIH 3T3 fibroblasts. As for the gene expression experiments, we used purified naïve B cells, IgM only Bmem cells, and class-switched Bmem cells for comparison. After stimulation, higher frequencies of IgM^hi B cells proliferated (Figures 3A,D) and differentiated to ASCc (Figures 3B,E) compared with IgM^lo B cells, naïve B cells, IgM only Bmem cells, and class-switched Bmem cells. IgM^lo B cells proliferated (Figures 3A,D) and differentiated to ASC (Figures 3B,E) at frequencies intermediate between those of naïve B cells and IgM^hi B cells, and similar to class-switched and IgM only Bmem cells. Additionally, higher frequencies of IgM^lo B cells performed IgG class-switching in comparison with IgM^hi B cells; however, the observed frequencies were relatively low (Figures 3C,F) probably related to the short (3 day) stimulation time used. As expected, naïve B cells responses were undetectable. These results indicate that the IgM^hi and IgM^lo subsets proliferate, differentiate to ASC, and switch isotype differentially.

Characterization of the Ig gene repertoire of B cells has contributed to understanding the biology of B cells. To contrast the Ig gene repertoire of the five B cells subpopulations of interest, we sequenced the heavy (Figures 4, 5) and light (Figure 6) chain genes of purified B cells subpopulations of three adult donors. B cells subsets were characterized by usage of IGH and IGK sequences, HCDR3 structure (length, N lengths, and P lengths), mutation frequencies, and shared clonotypes.

Consistent with some previous observations (17, 22), but not others (34), IGHV family and IGHV/IGHJ gene usage frequencies were highly similar among B cells, with few exceptions: IgM^hi B cells had lower frequencies of the IGHJ6 genes (p < 0.0001) compared with IgM^lo B cells (Figure 4B), and class-switched Bmem cells had higher expression of IGHV1 family than IgM^hi B cells (Figure 4A). Other differences appeared when contrasting IgM B cells subpopulations with naïve B cells: IgM^hi and IgM^lo B cells had lower levels of IGHV1 family usage (Figure 4A), and IgM^hi B cells had greater usage of IGHV3 and IGHV6 families (Figure 4A). Naïve B cells expressed higher frequencies of IGHJ6 genes than IgM^hi and IgM only subsets, but lower frequencies of IGHJ4 than IgM^hi (Figure 4B). Also compared with naïve B cells, IgM^lo and IgM^hi B cells had lower usage of IGHV1-69 genes (Figure 4C), but IgM^hi had increased usage of IGHV3-7, IGHV3-74, and IGHV6-1 genes (Figure 4C). Finally, IGHV3-7 gene usage was significantly higher in IgM only Bmem cells compared with naïve B cells. However, a higher percentage of mutations was detected in IGHV region of IgM^hi B cells compared with IgM^lo B cells (p < 0.0007) (Figure 4D) and the fraction of rearrangements with 0–2% mutations for IgM^lo B cells was higher than IgM^hi B cells, while a fraction of rearrangements with more than 2% of mutations was higher in IgM^hi B cells (Figure 4E).As expected, naïve B cells presented the lowest frequency of mutations of the five subpopulations (Figure 4D) and 96.14% of their clones presented between 0 and 2% of mutation Figure 4E). IgM^hi B cells (like IgM⁺IgD⁺CD27⁺ B cells) exhibited a lower frequency of mutations than class-switched Bmem cells (Figure 4D) (17) and a higher percentage of their clones (median = 47.03%) had a range 0–2% of mutations in comparison with class-switched Bmem cells (median = 28.10%) (Figure 4E).

Additionally, in line with previous reports (17), the lengths of HCDR3 were similar among B cells subpopulations, including IgM^hi and IgM^lo B cells, but different from naïve B cells (Figure 5A). However, a PCA with the properties of HCDR3, including HCDR3 length, the lineage relative frequency, N1, and N2 length, P (P3’V, P5’D, P3’D, and P5’J) length, and percentage of identity of IGHV-IGHJ with the corresponding germline gene, grouped IgM^lo and IgM^hi B cells in two different clusters (Figure 5B). The two-component, PC1 and PC2, had a cumulative explained variance of 24.4% (13.9 and 10.5%, respectively). The differential clustering of IgM^hi and IgM^lo B cells was comparable in the 3 volunteers studied (Supplementary Figure 4).

With few exceptions, IGKV family and IGKV/IGKJ gene usage frequencies were highly similar among the five study populations, with frequencies corresponding to those previously reported for naïve B cells (59, 60). The IGKV gene families and IGKJ genes used by IgM^hi and IgM^lo B cells were similar (Figures 6A,B). However, compared with naïve B cells, IgM^hi B cells used the IGKV3 family of genes at higher frequencies and the IGKV1-8 and IGKV1-33 genes at lower frequencies (Figures 6A,C). Compared with class-switched Bmem cells, IgM^hi B cells used IGKV1-33 genes at lower frequencies (Figure 6C). As expected, the percentage of mutations in IGKV region were lower than in the IGHV region, but, similarly to what we observed for the IGHV region, a higher percentage of mutations was found in IGKV region of IgM^hi B cells compared with IgM^lo B cells (Figure 6D). Also analogous to the IGHV region, naïve B cells presented the lowest frequency of mutations of the five subpopulations in the IGKV region (Figure 6D) and 97.6% of their clones presented between 0 and 2% of mutations (Figure 6E). IgM^hi B cells exhibited a lower frequency of mutations than IgM only and class-switched Bmem cells and, like in IGH, higher percentage of their clones (median = 73.09%) had a range 0–2% of mutations in comparison with class-switched Bmem cells (median = 28.10%) (Figures 6D,E). The lengths of KCDR3 were similar among all populations of B cells (Figure 6F).

In summary, assessment of the Ig gene repertoire of IgM^hi and IgM^lo B cells showed that IgM^hi B cells had higher frequencies of mutations in IGHV/IGKV regions, used IGHJ6 genes at lower frequencies, and differ in their HCDR3 properties.

To determine if IgM^hi and IgM^lo B cells were clonally related, we compared the frequency of B cells clonotypes they shared between themselves or with other B cells subpopulations, except for class-switched Bmem cells since the low number of clonotypes identified in this subset (164, 101, and 137 in donors 1, 2, and 3, respectively) limited this analysis (Table 1). A clonotype was defined as group of sequences that shared 100% of identity in nucleotides of their HCDR3 and the same assignment of IGHV and IGHJ genes reported by IMGT/HighVQuest. Considering clonotypes defined by IGH, the four subsets of B cells shared few (less than 6%) of clonotypes analyzed separately by donor (Supplementary Table 3) and when B cells from the three donors were grouped (Table 1). IgM only Bmem cells had the highest frequency of shared clones (5.78%) followed by IgM^hi (2.39%), IgM^lo (1.76%), and naïve B cells (0.04%) (Table 1). IgM^lo B cells shared with IgM^hi B cells 1.49% of clonotypes, while IgM^hi B cells shared 1.65% of clonotypes with IgM only Bmem cells. Interestingly, the great majority of the shared clonotypes detected in IgM only Bmem cells were almost exclusively shared with the IgM^hi B cells (5.39%) (Table 1 and Supplementary Table 3). No common clonotypes were identified between naïve B cells and any of the IgM B cells, except for donor number three who presented one shared clonotype between naïve B cells and IgM^hi B cells (0.04%) (Table 1 and Supplementary Table 3). Similar results were obtained using 97% of identity in the sequence of nucleotides of HCDR3 to define a clonotype (data not shown). In summary, IgM^hi B cells shared most of their shared IGH defined clonotypes with IgM only Bmem cells, and IgM^lo B cells shared most of their shared IGH clonotypes with IgM^hi B cells.