BALB/cJ and C57BL/6J, as well as μMT and Rag2^−/− mice on the C57BL/6 background, were purchased from Jackson Laboratories and bred and maintained in animal facilities at the University of Alabama at Birmingham (UAB). Dntt^−/− (Tdt), Igll1^−/− (λ5), and Rag1^−/− mice on the BALB/c background were generously provided by Dr. Harry Schroeder Jr. at UAB. Eμ-TCL1 Tg mice were kindly provided by Dr. Carlo Croce at Ohio State University. Unless otherwise specified, 8–12 week-old female mice were used for these studies. Embryonic fetal livers (FL) were obtained from timed pregnancies. Vaginal plug formation after mating was counted as day 0. All studies and procedures were approved by the UAB Institutional Animal Care and Use Committee (IACUC).

Total RNA was extracted from mouse tissues and single cell suspensions using RNeasy Plus kits (Qiagen). cDNAs were generated using SuperScript II (Invitrogen). cDNAs were mixed with primer pairs and amplified using SYBR green master mix and the 7900HT Fast real-time PCR system (Applied Biosystems). Iμ, μ0, and Myc primers have been published (48–50). Fcrl6 qPCR primers were designed to hybridize with the first extracellular domain using primer express software (Applied Biosystems). Samples were normalized to Polr2a (RNA polymerase II) expression (51). Primer sequences were as follows:

Rat anti-mouse FCRL6-specific mAbs, 1C3 (IgG1κ) and 3C1 (IgG2aκ), were generated using Escherichia coli-derived His-tagged recombinant protein comprised of the two extracellular Ig domains of mouse FCRL6. Respective cDNA regions were PCR amplified and cloned into the pET24b vector (Novagen) for bacterial expression, as previously described (52). Fisher rats (Jackson Laboratory) were immunized at 3–4 day intervals over a 3 week period and popliteal nodes were fused with the mouse plasmacytoma Ag8.653 and plated for selection in 96-well-plates (41). At 10–14 days after fusion, hybridoma clones were screened for specificity and cross-reactivity by staining hemagglutinin (HA)-tagged FCRL1, FCRL5 (C57BL/6 and BALB/c alleles), and FCRL6 TM transductants generated as previously (41). Rabbit anti-FCRL6 polyclonal Abs were generated by hyperimmunizing New Zealand White rabbits (Charles River Laboratories) with Escherichia coli-derived His-tagged recombinant protein.

To analyze the molecular nature of FCRL6, HA-tagged BW5147 FCRL6 retroviral transductants (1 × 10⁷) were lysed in 1% NP-40 lysis buffer. Whole cell lysate proteins were quantitated using the BCA reagent (Pierce) and incubated at 4°C for 30 min with rat anti-mouse FCRL6 (3C1) or an isotype-matched control (rat IgG2aκ) mAb, followed by the addition of 30 μl of a 50% slurry of protein G beads (GE Healthcare), and incubation overnight at 4°C. The beads were washed five times with 1 ml of lysis buffer to reduce non-specific binding, resuspended in an equal volume of SDS sample buffer, and boiled. Proteins were resolved by SDS-PAGE and transferred to PDVF membranes (Millipore) before immunoblotting with anti-HA (12CA5, Roche) or rabbit anti-mouse FCRL6 polyclonal Abs, followed by rabbit anti-mouse or goat anti-rabbit HRP (Southern Biotech). Bound Abs were visualized using the ECL reagent (GE Healthcare) and detected by Biomax XAR film (Kodak).

Spleen, FL, and bone marrow (BM) cells were prepared as single cell suspensions after red blood cell lysis with ACK lysing buffer (Gibco). PeC cells were prepared by lavaging with complete 10% FCS RPMI-1640 medium. FL were obtained from timed pregnancies. Neonatal livers, spleens, and BM were harvested at specific days after birth. Cells were stained with antigen–specific or isotype-matched control Abs (Supplementary Table—Key Resources) after blocking with unlabeled anti-CD16/32 for 5 min.

AF647 and biotin-conjugated anti-mouse FCRL6 (1C3 and 3C1) were generated and labeled by our laboratory (Invitrogen and Thermo Scientific). Anti-V_H11 [3H7 (7); a gift from K. Hayakawa], anti-V_H12 [5C5 (53); a gift from K. Rajewsky], and anti-V_H7 [TC68 (54), a gift from J. Kearney] were biotinylated in our laboratory (Thermo). Cells were analyzed using FACSCalibur (BD) or LSRII (BD) flow cytometry instruments and plotted with FlowJo software (Treestar).

For cell sorting, leg bones were pooled from 5 to 10 adult male and female mice and 7–10 FL were pooled from E18 fetuses isolated from the wombs of pregnant mothers according to the timing of vaginal plug formation. Single cell suspensions from the BM of BALB/cJ mice were prepared by flushing the femur and tibia bones with a 29 G needle syringe. Single suspensions of FL cells were prepared by crushing the tissue with a 1 ml syringe plunger and passage through a 40 μm strainer (BD). Single cell suspensions were treated with ACK lysing buffer and re-suspended with PBS containing 2% FCS. After Fc blockade with unlabeled anti-CD16/32, cells were stained with fluorochrome-labeled mAbs [FITC-CD43, BV421-CD19, PE-B220, PECy7-AA4.1/CD93, and Alexa647 FCRL6 (1C3)], as well as biotin-labeled mAbs [CD3 (145-2C11), Mac1 (M1/70), Gr1 (RB6-8C5), Ly6C (HK1.4), Ter119, and IgM (RMM-1)] to exclude T, myeloid, and erythrocyte lineage cells and immature/mature B cells, and counterstained with SA-BV570. FCRL6⁺/FCRL6⁻ pro B (AA4.1⁺Lin⁻CD43⁺B220⁺CD19⁺IgM⁻) cells from the FL and BM were sorted using a FACSAria II sorter (BD).

Single cells were prepared from the FL and BM of BALB/cJ mice at 24 h after E17 pregnant dams or adult mice were injected i.p. with BrdU (twice at 12 h intervals). Single cell suspensions were stained with anti-mouse AA4.1, CD43, CD19, B220, IgM, and FCRL6 (1C3) for surface detection, then fixed and permeabilized, treated with DNase I, and stained with anti-BrdU and 7AAD to examine proliferation and cell cycle status. Stained cells were analyzed by FACS with an LSRII instrument and profiles were plotted with FlowJo software.

Single cells from FL and BM were stained for AA4.1, CD43, CD19, B220, IgM, and FCRL6 (3C1), and fixed with Cytofix (BD) for 15 min on ice, then permeabilized with Foxp3 Fixation/Permeabilization buffer (eBio) for 30 min on ice, and stained for either Ki-67, c-Myc, NFAT2, Ikaros, or Aiolos, along with F(ab′)₂ goat anti-mouse IgM for 1 h at room temperature. Cells were examined using an LSRII cytometer and plotted with FlowJo software.

FACS sorted FCRL6⁺ and FCRL6⁻ pro B (CD43⁺CD19⁺B220^hiIgM⁻) cells from adult BM were treated with the phosphatase inhibitor pervanadate (NaVO₄) for 10 min. Stimulated cells were fixed with prewarmed Phosflow Lyse/Fix buffer (BD) at 37°C for 10 min and permeablized with Phosflow Perm Buffer III (BD). After Fc blockade (CD16/32), cells were stained with anti-phospho ERK pT202/pY204, STAT5 Y694, or isotype control mAbs (BD) for 30 min at room temperature. Phosphorylation was analyzed using a FACSCalibur flow cytometer (BD) and plotted with FlowJo software. The fold induction change in phosphorylation for FCRL6⁻ and FCRL6⁺ pro B cells was calculated by comparing the MFI ratios of FCRL6⁻/FCRL6⁺ pro B cells with and without stimulation.

FCRL6⁺ and FCRL6⁻ pro Bcells (AA4.1⁺CD43⁺CD19⁺B220^hiIgM⁻) stained for cell surface markers, were fixed with Cytofix/Cytoperm buffer (BD) for 20 min on ice and stained with anti-pre-BCR (SL156) and/or F(ab′)₂ goat anti-mouse IgM for 1 h at room temperature. Cells were analyzed using an LSRII instrument and plotted with FlowJo software.

Single cells from the FL and BM were stained for cell surface markers then washed twice in Annexin V binding buffer, followed by staining with anti-Annexin V for 15 min at room temperature. After incubation with the Annexin V binding buffer, cells were analyzed using an LSRII instrument and plotted with FlowJo software.

FACS-sorted FCRL6⁺ and FCRL6⁻ pro Bcells (AA4.1⁺CD43⁺CD19⁺B220^hiLin⁻IgM⁻) from BM were washed twice with complete RPMI 1640 media containing 10% FCS. Pro B cells (10³ cells per well) were loaded into round-bottom 96-well-plates in triplicate, in the presence or absence of different concentrations of TSLP (0–100 ng/ml). Cells were incubated at 37°C in a CO₂ incubator for 4 days, total cell numbers from each well were counted, and live cells were enumerated by flow cytometry analysis using a FACScalibur instrument with dead cell exclusion by PI.

Stromal cell-dependent B cell progenitor differentiation assays were previously described by Montecino-Rodriguez et al. and performed as follows (19, 55). S17 stromal cells were freshly thawed and grown in complete RPMI with 10% FCS for 7–10 days. S17 cells were detached with 0.25% Trypsin-EDTA solution (Gibco), filtered with a 70 μm cell strainer (BD), and 2.5 × 10⁴ cells were added to 6-well-Biocoat Transwell insert plates (BD) for growth on inserts overnight. FACS-sorted FL or BM B-1P and pro B cells (0.5–1 x 10⁴ cells/well), gated as in Figure 1B, were added to the bottom chamber of the Transwell plate, separated from S17 stromal cells in the upper chamber, and cultured in αMEM media (Gibco) with 5% FCS and a cytokine cocktail including TSLP (10 ng/ml), IL-3 (5 ng/ml), IL-6 (10 ng/ml), SCF (10 ng/ml), and Flt3L (10 ng/ml). At day 4, 25% of the upper and lower culture media was replaced with fresh cytokine-containing media. At day 8, the cells were harvested and washed with PBS containing 2% FCS. Viability, proliferation, and differentiation into immature B220⁺IgM⁺ B cells were examined with an LSRII cytometer (BD) and analyzed with Flow-Jo software.

RNA from FCRL6⁺ to FCRL6⁻ pro B cells sorted as in Figure 1B, from pooled E18 FL (n = 7) or adult BM samples (n = 8) from BALB/c mice, was isolated using TRIzol (Invitrogen) reagent and used to generate cDNA with the High-Capacity cDNA synthesis kit (Applied Biosystems). cDNAs were used as a template for PCR using the KOD high fidelity polymerase (Novagen) and a mixture of 8 V_H family and IgM-specific primers (listed below) containing 5′ adaptors to allow barcoding by the Illumina TruSeq multiplex PCR kit (Illumina). Resulting PCR products were subjected to 10 cycles of PCR using Illumina TruSeq indexing primers. The resulting amplicons were purified using AMPure (Beckman), submitted to the UAB Heflin Genomics core, and sequenced using Illumina's MiSeq Reagent Kit v3 2X300 kit in a Miseq next generation sequencer. The resulting FastQ files were filtered for quality (>Q30) and assembled using the vsearch suit of informatic tools (57). Fasta files containing the full length paired Igh sequences were submitted to the IMGT High-VQuest server for Ighv, Ighd, and Ighj identification, as well as Ighv junction and other bioinformatic analyses (58). IMGT files were visualized using R software (R Foundation for Statistical Computing) and the following software packages: vegan, plyr, plots, ggplots, ggrepel, ggseqlogo, pbmcapply, seqinr, and stringr. Some rarefaction curves were replicated with VDJtools. Sequences were deposited at NCBI accession: PRJNA547609.

Ighv-specific primer sequences:

An i5 linker (5′-CCCTACACGACGCTCTTCCGATCT-3′) was added to the 5′ end of the forward V_H primers and an i7 linker (5′-ATCTCGTATGCCGTCTTCTGCTTG-3′) to the 5′ end of the reverse Cμ primer.

CDR-H3 analysis was performed as in Khass et al. (32). Briefly, the CDR-H3 was defined as the sequence starting immediately after the cysteine at the end of FR3 at position 96 and then through, but not including, the tryptophan that begins FR4. At positions 99 to 103, we calculated the frequency of amino acids for individual B cell subsets and common Ighv segments. Logo plots were prepared with the ggseqlogo R package (59).

RNA sequencing (RNA-seq) was performed at the UAB Heflin Center for Genomic Science. E18 FL and adult BM FCRL6^+/− pro B cells were sorted in duplicate in four independent experiments from BALB/cJ mice by flow cytometry as detailed above and shown in Figure 1B. mRNA was isolated using RNeasy Plus kits (Qiagen) and sequencing was performed on a Illumina NextSeq 500 (paired-end sequencing 2 × 75 bp) according to the manufacturer's guidelines. Briefly, the quality of the total RNA was assessed using the Agilent 2100 Bioanalyzer (all samples had RIN values of >9.9) followed by two rounds of poly(A) selection and conversion to cDNA. TruSeq library generation kits were constructed by random fragmentation of the polyA mRNA, followed by cDNA production using random primers (Agilent Technologies). The ends of the cDNA were repaired, A-tailed, and adaptors were ligated for flow cell attachment, sequencing, and indexing as per the manufacturer's instructions (Agilent Technologies). cDNA libraries were quantitated using qPCR in a Roche LightCycler 480 with the Kapa Biosystems kit for library quantitation (Kapa Biosystems) prior to sequencing.

All samples contained a minimum of 32 million reads with an average number of 42.9 million reads across all biological replicates. The FASTQ files were uploaded to the UAB High Performance Computer cluster for bioinformatics analysis with the following custom pipeline built in the Snakemake workflow system (v4.8.0) (60): first, quality and control of the reads were assessed using FastQC, and trimming of the bases with quality scores of <20 was performed with Trim_Galore! (v0.4.5). Following trimming, the reads were aligned with STAR (61) (v2.5.2a, during “runMode genomeGenerate,” the option “sjdbOverhang” was set to 74) to the Ensembl mouse genome (mm10), which resulted in an average of 80.3% uniquely mapped reads. BAM file indexes were generated with SAMtools (v1.6) and gene-level counts were generated using the function “featureCounts” from the R package, Rsubread (v1.26.1; r-base v3.4.1), with the “Mus_musculus.GRCm38.90.gtf” file from Ensembl. The parameters used in “featureCounts” included: isGTFAnnotationFile = TRUE, useMetaFeatures = TRUE, allowMultiOverlap = FALSE, isPairedEnd = TRUE, requireBothEndsMapped = TRUE, strandSpecific = 2, and autosort = TRUE. Additional parameters were kept as default. Logs of reports were summarized and visualized using MultiQC (v1.5) (62).

Count normalization and differential expression analysis were conducted with R software with the DESeq2 package (63) (v1.16.1). Following count normalization, principal component analysis (PCA) was performed (Figure 3A) and genes were defined as differentially expressed genes (DEGs) if they passed a statistical cutoff containing an adjusted P < 0.05 [Benjamini-Hochberg False Discovery Rate (FDR) method] and if they contained an absolute log₂ fold change ≥1. Functional annotation enrichment analysis was performed in the NIH Database for Annotation, Visualization and Integrated Discovery (DAVID, v6.8) by submitting all DEGs identified. The Benjamini-Hochberg FDR correction was also applied to determine gene ontology (GO) terms with the cutoff of an adjusted P < 0.05.

RNA-seq files are available at the Gene Expression Omnibus under accession number GSE132438. Figures, including the heatmap, scatter plots, and volcano plot, were made with R software using the following packages: ggplot2 (v3.1.0) and ComplexHeatmap (v1.20.0).

Statistical analysis was performed using GraphPad Prism 7 software. A paired Student's t-test was used for comparing two experimental groups (FCRL6⁺ and FCRL6⁻ progenitors) for apoptosis, intracellular and surface staining, RQ-PCR, and differentiation assays. Unpaired Student's t-test was used for comparing cell populations in mutant mouse studies. Statistical analyses used in RNA-seq studies are detailed in the indicated section.

Following the discovery of human FCRL1-6 (38, 64), we identified a mouse Fcrl6 counterpart encoding putative type I transmembrane (TM), glycosylphosphatidylinositol (GPI)-linked, and secreted isoforms with two extracellular Ig-like domains (45). By RQ-PCR, we detected Fcrl6 transcripts in primary lymphopoietic tissues, including embryonic day 18 (E18) FL, day 3 postnatal liver, and adult BM (Figure 1A). The generation of monoclonal Abs (mAbs) yielded two FCRL6-specific subclones (1C3 and 3C1) and identified that the TM isoform had a molecular weight of ~42 kD (Supplementary Figures 1A,B). Although FCRL6 is confined to natural killer (NK) and T cells in humans (47), by flow cytometry analysis from adult BALB/cJ BM, FCRL6 was restricted to early stage B cells (Supplementary Figure 1C and data not shown). Expression was not detected by other lineages or in any other hematopoietic tissue at homeostasis. FCRL6 shared a similar pattern of expression by FL and BM B cells bearing CD19 and B220 as well as AA4.1 and CD43, but little if any surface IgM (Supplementary Figures 2A,B). These observations indicated that 11–12% of FL and 1–2% of BM CD19⁺ B cells express FCRL6.

Because its frequency was about 10 times greater in the FL, we examined FCRL6 using a differentiation scheme that segregates precursors capable of reconstituting the B-1 or B-2 lineage (19). FCRL6 marked subsets of CD19^hiB220^hi pro B cells in the FL and BM, as well as some CD19^hiB220^lo B-1 precursors (B-1P), which are more frequent in the FL, but not CD19^loB220⁺ B-2 precursors (B-2P) that chiefly populate the BM (Figure 1B). In day 3 and 7 neonatal tissues, FCRL6⁺ B cells variably expanded in frequency with time in the liver, BM, and spleen (Supplementary Figure 2C). This pattern of tissue expression demonstrated that FCRL6 segregates subsets of progenitor B cells conserved throughout life.

We then examined adult BM according to Hardy's definition (56). By this scheme, only a minor subset of B cell progenitors, estimated at 7–15% of the total A-C′ fraction (Fr), expressed FCRL6 (Figure 1C). Further separation based on CD24/HSA and BP-1 disclosed hardly any FCRL6⁺ Fr. A cells, but the frequency increased in Fr. B, peaked in Fr. C, declined in Fr. C′, and very few cells were found beyond the Fr. D stage. Thus, FCRL6 expression was tightly regulated and predominantly restricted to Fr. B and C pro and pre B cells, developmental stages during which V(D)J recombination and pre-BCR selection occur.

We next analyzed other phenotypic features in the FL and BM. By forward scatter, FCRL6⁺ pro B cells were smaller than FCRL6⁻ cells, suggesting they might proliferate more slowly (Figure 2A). The relatively lower density of CD127/IL-7Rα on FCRL6⁺ cells in both tissues indicated they might also be less responsive to IL-7. The early differentiation marker CD117/c-kit was uniformly upregulated by the FCRL6⁺ subset in FL, but not in the BM. BP-1 was also less abundant on FL FCRL6⁺ cells. Finally, MHCII and CD138, which can distinguish B-2 progenitors (65), were slightly higher on FCRL6⁻ pro B cells in the BM than FL. Together these results indicated that FCRL6 distinguished less differentiated pro B cells.

We then compared different mouse strains and models with defects in early B cell development. We first observed that the frequency of CD19⁺FCRL6⁺ cells in the BM was about 4-fold higher in WT BALB/c than C57BL/6 mice (Figure 2B). In Rag1^−/−, Rag2^−/−, and Igll1 (λ5)^−/− mutants, B cell development is blocked at Fr. C and C′ resulting in the enrichment of Fr. B and C pro B cells. BM cells from Rag1^−/−, Rag2^−/−, Igll1^−/−, and μMT mice of respective strains all had expanded frequencies of FCRL6⁺ B cells. In these models, CD19⁺FCRL6⁺ B cells were ~10 to 15-fold greater compared to WT mice. However, no difference was evident in Tdt^−/− mice. Thus, FCRL6⁺ cells emerged prior to RAG1/2 expression, and moreover V(D)J recombination, segregated a subset of pro B cells that varied by genetic background, and expanded following disruption of Igh rearrangement or pre-BCR/BCR assembly.

Because FCRL6⁺ pro B cells were relatively smaller and less differentiated, it was possible they represented defective cells destined for elimination. However, no differences in Annexin V reactivity, and thus survival status, were found (Figure 2C). We then analyzed their capacities for generating cytoplasmic μHC. Compared to FCRL6⁻ progenitors, frequencies of μHC⁺FCRL6⁺ cells were nearly 4-fold lower in the FL and 2-fold lower in the BM (Figure 2D). These results indicated marked mechanistic differences in μHC generation according to FCRL6 status.

We next performed high-throughput RNA sequencing (RNA-seq) to compare the gene expression profiles of FL and BM FCRL6⁺ to FCRL6⁻ pro B cells. By principle component analysis (PCA), duplicate samples clustered closely and segregated into four quadrants indicating distinct gene signatures for the four subsets (Figure 3A). This analysis identified a total of 1,276 differentially expressed genes (DEGs) in the FL and 384 in the BM [DEG criteria: a change in expression of 1-fold (log₂ value); false discovery rate (FDR), < 0.05] (Figure 3B and Supplementary Figures 3A,B). By gene ontology (GO) analysis, cytokine production, adhesion, innate defense, and migration pathways were upregulated by FL FCRL6⁺ pro B cells (Figure 3C). Because these features were evocative of innate-like B cells, we compared DEGs from FL pro B cells with those from PeC B-1a and B-1b cells or splenic B-1a and MZ B cells extracted from the ImmGen database. Greater gene overlap (~1.5-fold) was observed with FCRL6⁺ upregulated DEGs (Supplementary Figure 3C). Distinct biologic differences related to FCRL6 expression were further implied by 265 DEGs that overlapped between the FL and BM (Figures 3B,D). Dissimilar proliferation kinetics in FCRL6⁺ cells were indicated by the strong repression of mitotic cell cycle genes, including transcripts for Mki67 (encodes Ki-67). Intracellular staining confirmed higher frequencies of Ki-67⁺ FCRL6⁻ pro B cells (Figure 3E). Furthermore, an analysis of BrdU and 7-AAD status showed that most FCRL6⁺ cells had not entered the S/G2+M phase and were primarily in G0/G1 (Figure 3F). G2/M frequencies were 6.5 and 5.7-fold higher for FCRL6⁻ compared to FCRL6⁺ pro B cells in the FL and BM. These results demonstrated significantly lower proliferation and cell cycle activity for FCRL6⁺ cells in vivo regardless of their tissue of origin. By GO pathway analysis, genes for signal transduction regulation were also induced. Among DEGs encoding phosphatases, 14/20 in the FL and 4/5 in the BM were upregulated in FCRL6⁺ cells (Supplementary Figure 3D). We thus compared signaling features of sorted BM pro B cells by ex vivo treatment with the phosphatase inhibitor pervanadate. FCRL6⁻ cells had higher STAT5 and ERK phosphorylation (Figure 3G), indicating the diminished activation potential of FCRL6⁺ cells. Thus, in addition to other properties, FCRL6 defined pro B subsets with specific regulatory differences.

As expected, Fcrl6 was the most upregulated overlapping DEG between tissues. Bhlhe41, a transcriptional repressor critical for B-1a development and Ab repertoire formation (28), proved the most downregulated (Figure 3D). Furthermore, GO pathway components associated with nervous system development were upregulated. These included overlapping DEGs (Heyl, Dlx1, Aph1b) and pathways relevant for Notch, Wnt, stem cell, EMT biology, and TCF/LEF (Supplementary Figure 3E). By flow cytometry, intracellular LEF1 expression was indeed elevated in FCRL6⁺ compared to FCRL6⁻ progenitors in day 3 neonatal tissues (data not shown). These findings collectively demonstrated divergent developmental and regulatory features for FCRL6⁺ progenitors.

Because V(D)J recombination and DNA damage repair are coupled to the cell cycle (66, 67), enrichment in the G0/G1 phase indicated that FCRL6⁺ pro B cells might be more actively undergoing HC rearrangement. Indeed, the RNA-seq GO pathway analysis revealed the downregulation of DNA damage repair programs (Figure 3C). To assess the spectrum of transcription, we grouped the Ighv locus into four domains according to the locations of CTCF binding sites that organize locus contraction and transcript accessibility during V(D)J recombination [(68); Figure 4A]. PCA of normalized Ighv segments showed partitioning of duplicate samples into quadrants, but the variance between BM sets was smaller than that of FL samples (Supplementary Figure 4A). By unsupervised Euclidian clustering, we found that the Ighv signature of FL FCRL6⁺ pro B cells segregated independently, suggesting that this subset generated a distinct complement of V_H segments (Supplementary Figure 4B). FCRL6⁺ cells primarily upregulated distal Ighv1/J558 segments from domain 4, but downregulated proximal domain 1 and 2 Ighv genes (Figures 4A–C). Indeed, multiple V_H segments from these domains were among FL DEGs (Figure 4D). A similar trend of differential locus accessibility was evident in the BM, but did not reach the same threshold of significance (data not shown). Notably, no differences were evident for Ighv5-2/V_H81X or segments encoding PtC-reactive natural Abs, Ighv12-3 and Ighv11-2. Of two phosphorylcholine (PC)-reactive HCs, Ighv7-1 was downregulated in FCRL6⁺ cells, but Ighv7-3 was not.

We next examined how Ighv accessibility was differentially regulated. First, recombinatorial activity was elevated. FCRL6⁺ pro B cells demonstrated upregulation of both Rag1 and Rag2 transcripts by RNA-seq (Figure 4E). Second, non-coding RNAs (ncRNA) critical for Igh locus contraction were investigated (69). Among the 14 PAX5-associated intergenic repeat (PAIR) elements that serve as CTCF, PAX5, and E2A binding sites (70), ncRNA transcripts for PAIR4 and PAIR6, which are distally positioned in domain 4, were among upregulated, overlapping DEGs in FCRL6⁺ cells (Figures 3D, 4E). Within the proximal J_H-Cμ region, Iμ transcripts initiate from the Eμ intronic enhancer (71) and form long-range DNA loops that tether proximal and distal ends of the Igh locus via direct binding to both PAIR4 and PAIR6 in a YY1-dependent fashion (72, 73). μ0 transcripts arise just upstream of the most 3′ D_H gene (DQ52) (74). RQ-PCR indicated higher Iμ, but lower μ0 transcripts in FCRL6⁺ FL pro B cells (Figure 4F). With respect to accessibility factors, transcripts for Spi1 (PU.1), Tcf3 (E2A), Ikzf1 (Ikaros), Ikzf3 (Aiolos), Ctcf , Rad21, and Ezh2 were downregulated, whereas Ebf1 was upregulated, but Pax5 and Yy1 did not differ (Supplementary Figure 4C). Collectively, these data indicated that FCRL6 expression marked disparate regulation of Ighv accessibility that may relate to Eμ-dependent (long-range) vs. independent (local) mechanisms of locus contraction (69, 73). Thus, FCRL6 status indicated heterogeneity of V_H gene segment usage among progenitor B cell subsets.

To investigate Ighv sequences from these pro B cells, V(D)J rearrangements were amplified from the four sorted populations. After annotating the amplicons using the IMGT database, we identified 317,702 unique dereplicated sequences that included representation of 8/15 V_H families (Supplementary Figures 5A–C). The most striking finding was the low frequency of productive rearrangements among unique dereplicated Ighv sequences (both non-productive and productive) generated from FCRL6⁺ vs. FCRL6⁻ pro B cells in the FL (31 vs. 79%) (Figure 5A). In the BM, where Tdt is operative (10), productivity was higher for cells marked by FCRL6, but was still lower than FCRL6⁻ pro B cells. Notably, Dntt (Tdt) expression did not differ according to FCRL6 status by RNA-seq. By dividing the number of unique productive sequences from unique dereplicated sequences among the eight amplified V_H families, we found that Ighv1/J558 was the most prolific in FCRL6⁺ cells (42.6% in FL and 51.1% in BM) (Supplementary Figure 5C). However, relative to all unique dereplicated sequences, the frequency of productive Ighv1/J558 rearrangements in FCRL6⁺ cells was very low in FL, but increased in BM (Supplementary Figure 5B, top). Surprisingly, the most productive FL FCRL6⁺ cell rearrangements derived from the small Ighv11 family (49.5%). However, Ighv11-2 productivity in the BM was similar for both subsets. Ighv5-2/V_H81X was the most proficient segment among FL FCRL6⁺ cells (70.8%), but was similar in FCRL6⁻ pro B cells (70.3%). However, V_H81X productivity globally dropped in the BM. Among D and J segments, productivity for FCRL6⁺ cells was higher for D-less joins in the FL and Ighd1/DFL16.1 in the BM, whereas Ighd2/DSP segments were generally less favored (Supplementary Figures 5D,E). Ighj1-4 segment usage did not markedly differ by FCRL6 status. The CDR-H3 diversity of rearrangements in FCRL6⁺ cells was relatively lower in both tissues (Figure 5B). This was coincident with generally decreased CDR-H3 length and increased J trimming in the FL (Supplementary Figure 5B). Hence, the generally lower productivity and diversity of V(D)J rearrangements in FCRL6⁺ cells indicated significant mechanistic differences compared to their FCRL6⁻ counterparts.

At the pre-BCR checkpoint, the SLC proteins, λ5 and VpreB, make extensive contacts with the CDR-H3 of the nascent μHC by forming a “sensing site” to survey its biochemical properties (31). Recent work has demonstrated that tyrosine enrichment at position 101 of the CDR-H3, which interacts with three residues in VpreB, favors positive selection and checkpoint passage (32). We thus compared unique productive CDR-H3 sequences from the four subsets by determining amino acid usage at positions 99–103 relative to the cysteine at position 96 (PDB numbering). While variability was evident between tissues, tyrosine enrichment at these five positions was highest for FCRL6⁻ pro B cells (Figure 5C and Supplementary Figure 6A). In contrast, tyrosine content at residue 101 (Y101) was about 1/3 lower for FCRL6⁺ CDR-H3 sequences.

We then inspected the CDR-H3 Y101 frequency by comparing shared Ighv segments between FCRL6⁺ and FCRL6⁻ pro B cells with at least 50 productive sequences each (Supplementary Table 1). Marked differences were identified in the range of Y101 usage. Among 60 common FL Ighv genes, a comparably narrower spectrum of Y101 composition was evident for FCRL6⁻ pro B cell rearrangements (16.8–48.9%) vs. FCRL6⁺ cells (0.6–54.9%) (Figure 5D). This trend was also apparent for 74 Ighv genes shared for the BM (Figure 5E). We then analyzed common sets of Ighv genes to assess amino acid predominance at the CDR-H3 101 position. This breakdown identified 19/60 Ighv genes in the FL that preferentially utilized non-Y101 amino acids (Figure 5D, right). Surprisingly, while only two FCRL6⁻ Ighv sequences exhibited this feature (Ighv1-62 and Ighv11-1), non-Y101 usage was evident for 18 FCRL6⁺ Ighv genes. In the BM, 30/74 common Ighv genes were predominantly non-Y101 (Figure 5E, right). Although five of these were FCRL6⁻ derived, 27 came from FCRL6⁺ cells. Additionally, six non-Y101 enriched Ighv genes were shared between FL and BM FCRL6⁺ cells. These findings indicated that the CDR-H3 features of FCRL6⁺ pro B cells were disproportionately hydrophobic, charged, and less tyrosine enriched. Hence, FCRL6 marked progenitors predisposed to generating rearrangements with altered categories of diversity akin to the characteristics of Abs enriched in innate-like B cells (11).

We next considered Ighv sequence data from B-1a cells recently analyzed by the Herzenberg group (12) and CLL clones from different mouse models and strains (Supplementary Figure 6A and Supplementary Tables 2, 3). Among 150 B-1a and 291 CLL Ighv sequences, Ighv1/J558 segments were the most common and mainly domain 4 derived (Supplementary Figures 6B,C). Ighv11 was the second most favored among B-1a cells and fourth among CLL clones. CDR-H3 analyses disclosed that, like FCRL6⁺ pro B cells, tyrosine content was generally low and particularly poor at position 101 for both B-1a (42%) and CLL (35%) sequences (Supplementary Figures 6A,D). Collectively, these results demonstrated a progressively lower gradient of Y101 frequencies associated with FCRL6 expression and tissue of origin compared to B-1a and CLL cells (Supplementary Figure 6E). We concluded that the emerging μHC repertoire in FCRL6⁺ pro B cells, which was inefficient, less diverse, and autoreactive, would disfavor pre-BCR formation and positive selection.

Given the unconventional CDR-H3 features in FCRL6⁺ pro B cells and the possibility that alternative selection processes may regulate early B-1 cell development (37), we focused on the pre-BCR. By RNA-seq, transcripts for the SLC encoding genes Igll1 and VpreB1, were upregulated in FCRL6⁺ cells, whereas Ikzf1 and Ikzf3, which encode factors (Ikaros and Aiolos) that repress SLC genes (75, 76), were downregulated (Figure 3D and Supplementary Figures 3A, 4C). Intracellular and surface staining validated these relationships, which were variably more pronounced in the FL than BM (Figure 6A). Aiolos also induces exit from the cell cycle to promote LC rearrangement (75, 76). Recent studies proposed that premature LC rearrangement by FL pro B cells may obviate pre-BCR formation, leading to direct BCR generation of potentially autoreactive rearrangements, including V_H12 (37). Notably, Igkc transcripts were among downregulated overlapping DEGs (Figure 3D). Accordingly, the frequencies of surface κLC were 3-fold higher among μHC⁺ FL FCRL6⁻ pro B cells (Figure 6B). This finding indicated differences in the selection of progenitor B cells that correlated with FCRL6 expression.

We then examined pre-BCR formation. By co-staining for μHC, we detected 2 to 3-fold higher pre-BCR levels in FCRL6⁺ compared to FCRL6⁻ pro cells (Figure 6C). Given their unfavorable CDR-H3/Y101 biochemical features, this finding was unexpected. Because Ighv11 rearrangements were among the most productive in FL FCRL6⁺ cells (Supplementary Figures 5B,C), we employed anti-idiotype (Id) mAbs to detect the fraction of V_H11 or V_H7 μHCs that formed BCRs with emerging PtC or PC reactivity. Although V_H11 detection was generally rare within the μHC pool (Figure 6D), frequencies were indeed higher among FCRL6⁺ vs. FCRL6⁻ cells. V_H12 staining was extremely rare among progenitors (data not shown). V_H7 was more readily detected and while it did not markedly differ between FL subsets, its expression by BM FCRL6⁺ pro B cells was comparatively higher. Thus, associated with their disadvantageous CDR-H3 composition, FCRL6⁺ progenitors exhibited higher SLC, V_H11, and pre-BCR formation, but lower κLC production than their FCRL6⁻ counterparts. These findings suggested the existence of pre-BCR dependent and independent selection processes that varied according to FCRL6 status.

Although FCRL6⁺ cells were mitotically repressed (Figures 3C–F), we were surprised to find that Myc transcripts were upregulated in FL FCRL6⁺ cells by RNA-seq and RQ-PCR (Supplementary Figure 7A). Myc has important roles in stimulating early B cell development and proliferation at the pro to pre B cell transition (77). Notably, Myc is repressed by Ikaros and Aiolos, promotes both B-1 and B-2 lineage development, and is induced in CLL (15, 76, 77). We confirmed that intracellular Myc expression was elevated in splenic WT B-1a cells and CLL expansions in Eμ-TCL1 Tg mice (Supplementary Figure 7B). We then investigated Myc as a function of μHC status. While Myc was not appreciably induced in μHC⁻ progenitors, it was upregulated by μHC⁺ FCRL6⁺ B-1P and pro B cells (Supplementary Figure 7C). These findings were common to both tissues, but were more robust in the FL. To determine if Myc activation was pre-BCR dependent, we analyzed Igll1 (λ5)^−/− mice. These results confirmed that λ5, and thus pre-BCR assembly, was required for Myc activation by FCRL6⁺ μHC⁺ pro B cells (Figures 7A,B). Myc also amplifies calcium signaling in early B cells to promote NFAT activation (77), which is required for B-1a development and the generation of PtC-specific reactivity (78). NFAT2 was generally higher in FL FCRL6⁺ cells, but levels peaked in FCRL6⁺ μHC⁺ cells (Supplementary Figure 7D). In the BM, NFAT2 levels were globally lower and differences according to FCRL6 were not as pronounced. However, in contrast to Myc, NFAT2 induction in the FL and BM was independent of pre-BCR formation (Figures 7C,D). Finally, we considered the influence of the pre-BCR on B-1a cell V_H11 repertoire development. In the C57BL/6 strain, V_H12 usage between WT and Igll1 (λ5)^−/− B-1a cells did not differ (37). We confirmed these findings for V_H12, but found that V_H11 reactivity was dramatically lower in both frequency and absolute number in λ5-deficient mice (Figures 7E,F and Supplementary Figure 7E). In summary, these findings indicate that Myc-induction in FCRL6⁺ progenitor B cells and the development of V_H11⁺ B-1a cells is pre-BCR dependent.

Resemblance of the FCRL6⁺ pro B cell μHC repertoire and transcriptome to innate-like B cells was further underscored by the induction of gene programs encoding components of innate defense, including Toll-like (Tlr1, Tlr2, Tlr4, Tlr5), DNA (Aim2, Ddx58), and NOD-like (Nlrc3) pathogen recognition receptors, as well as cytokine production (Maf , Stat4, Ccl3, Ccl4) (Figures 3C,D and Supplementary Figures 3A,C). Migration and adhesion programs were also upregulated. DEGs encoding the LFA-1 (Itgb2) and VLA-4 (Itga4, Itgb1) integrins, which direct migration and adhesion to follicular dendritic (FDC) and endothelial cells (79), were elevated in FL and BM FCRL6⁺ cells (Supplementary Figure 8A). CD69 and Slamf1/CD150 were also among overlapping DEGs with higher expression by FCRL6⁺ cells. In contrast, we noted that Ccr7, which encodes a chemokine that mediates trafficking to T-cell enriched zones (80), was downregulated in FL FCRL6⁺ cells and elevated on FL FCRL6⁻ pro B cells (Supplementary Figures 3A, 8B). These findings indicated different migration properties for these two FL subsets.

B-1a cells express CXCR5 and home to the PeC and body cavities in a CXCL13-dependent fashion (81). Although Cxcr5 was among upregulated overlapping DEGs, it was not detected on FCRL6⁺ pro B cells (Figure 3D and Supplementary Figure 8B). However, CXCR5⁺ transitional B cells entering the spleen 18 h before birth acquire postnatal CXCL13 responsiveness to initiate white pulp formation (82). Thus, we postulated that CXCR5 might be dynamically induced to direct migration after parturition. We therefore examined CXCR5 in the liver, BM, and spleen from day 3 and 7 neonates. CXCR5 surface density and cell frequency was uniformly higher on FCRL6⁺ progenitors compared to FCRL6⁻ cells among day 3 tissues (Supplementary Figure 8C). Although FCRL6⁻ pro B cells had lower CXCR5, it escalated on subpopulations in the spleen > BM > liver. At day 7, largely homogenous CXCR5 expression was still evident for liver FCRL6⁺ pro B cells, but it bifurcated in the BM and was nearly lost in the spleen. Subpopulations of CXCR5⁺ FCRL6⁻ pro B cells were detectable at day 7 in liver and BM, but a distinct subset emerged in the spleen. Given the impact of pre-BCR formation on Myc induction in FCRL6⁺ cells, we examined CXCR5 as a function of μHC status. At day 3, CXCR5 levels on FCRL6⁻ pro B cells were generally low regardless of μHC expression (Figure 8A). In contrast, elevated CXCR5 expression by FCRL6⁺ μHC⁻ cells markedly declined in μHC⁺ cells. These results indicated that CXCR5 and μHC production were linked in FCRL6⁺ cells, and that migrating μHC⁻ cells repress CXCR5 upon becoming μHC⁺.

We then considered the impact of TSLP, a trophic factor capable of driving differentiation of fetal B cell progenitors to surface IgM⁺ cells (83). FCRL6⁻ pro B cells demonstrated higher IL-7Rα expression, but TSLPR levels did not markedly differ (Figure 2A and data not shown). TSLPR expression by FCRL6⁺ and FCRL6⁻ B cell progenitors was similar among day 3 and 7 postnatal tissues (Supplementary Figure 8D). However, with respect to μHC status, μHC⁺ FCRL6⁻ cells strongly downmodulated the TSLPR (Figure 8B). In contrast, sustained TSLPR expression by μHC⁺ FCRL6⁺ cells indicated that responses to this cytokine differed between these subpopulations. To investigate this, we first sorted BM FCRL6⁺ and FCRL6⁻ pro B cells and performed in vitro growth assays in the presence of TSLP. Survival of both subsets increased in a dose-dependent manner; however, FCRL6⁺ pro B cells exhibited greater TSLP responsivity (Figure 8C). We next examined differentiation by adopting an in vitro co-culture system established by the Dorshkind group that includes TSLP (19). In initial studies, we detected few IgM⁺ B cells at day 4 and by day 12, most cells were dead (data not shown). At day 8, however, sorted FCRL6⁺ FL B-1P and FL and BM pro B cells were consistently more proficient than FCRL6⁻ cells in differentiating to IgM⁺ cells (Figure 8D). Thus, FCRL6⁺ progenitors exhibited preferential CXCR5- and TSLPR-mediated homing and differentiation potential.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.