key: cord-0427505-h69ve17b
authors: Tull, Thomas J.; Pitcher, Michael J.; Guesdon, William; Siu, Jacqueline H.; Lebrero-Fernández, Cristina; Zhao, Yuan; Petrov, Nedyalko; Heck, Susanne; Ellis, Richard; Dhami, Pawan; Kadolsky, Ulrich D.; Kleeman, Michelle; Kamra, Yogesh; Fear, David J.; John, Susan; Jassem, Wayel; Groves, Richard W.; Sanderson, Jeremy D.; Robson, Michael D.; D’Cruz, David; Bemark, Mats; Spencer, Jo
title: Human marginal zone B cell development from early T2 progenitors
date: 2020-09-25
journal: bioRxiv
DOI: 10.1101/2020.09.24.311498
sha: 1d5361d0534786430c360d9943a83e67c78a4739
doc_id: 427505
cord_uid: h69ve17b

B cells emerge from the bone marrow as transitional (TS) B cells that differentiate through T1, T2 and T3 stages to become naïve B cells. We have identified a bifurcation of human B cell maturation from the T1 stage forming IgMhi and IgMlo developmental trajectories. IgMhi T2 cells have higher expression of α4β7 integrin and lower expression of IL4 receptor (IL4R) compared to the IgMlo branch and are selectively recruited into gut-associated lymphoid tissue. IgMhi T2 cells also share transcriptomic features with marginal zone B cells (MZB). Lineage progression from T1 cells to MZB via an IgMhi trajectory is identified by pseudotime analysis of scRNA-sequencing data. Reduced frequency of IgMhi gut homing T2 cells is observed in severe SLE and is associated with reduction of MZB and their putative IgMhi precursors. The collapse of the gut-associated MZB maturational axis in severe SLE affirms its existence and importance for maintaining health.

Transitional (TS) B cells are the immature B cells in human blood from which all mature B cells develop. Following emigration from bone marrow, TS B cells mature through T1, T2 and T3 phases, when autoreactive cells are depleted (Palanichamy et al., 2009; Suryani et al., 2010; Yurasov et al., 2005) .

In mice a B cell lineage split that is dependent on B cell receptor (BCR) engagement and the serine/threonine kinase Taok3 is initiated at the T1 phase (Hammad et al., 2017) . This directs B cells towards marginal zone B (MZB) cell fate, requiring subsequent Notch2 cleavage by a disintegrin and metalloproteinase-containing protein-10 (ADAM10).

MZB lineage progression in humans is not clearly understood, or indeed, universally accepted. A MZB precursor (MZP) population has been proposed that undergoes terminal differentiation to MZB following NOTCH2 ligation and can be discriminated from naïve B cells by expression of high levels of IgM (IgM hi ), CD24 and the glycosylation-dependent epitope CD45RB MEM55 (referred to here as CD45RB). An additional CD45RB hi IgM hi population that lacks the ABCB1 cotransporter has previously been referred to as T , although the relationship between this subset, MZB and MZP is unclear (Bemark et al., 2013; Descatoire et al., 2014; Koethe et al., 2011; Zhao et al., 2018) .

In humans, MZB develop over the first 2 years of life and are important for immunity against encapsulated bacteria (Weller et al., 2004) . They undergo a phase of clonal expansion and receptor diversification in the germinal centres (GC) of gut-associated lymphoid tissue (GALT) (Zhao et al., 2018) (Weill and Reynaud, 2019) . The shared expression of MAdCAM1 between the splenic marginal zone reticular cells and GALT high endothelial venules (HEV) creates the potential to recruit B cells to both sites mediated by 4 7 integrin binding (Kraal et al., 1995; Vossenkamper et al., 2013) . We have described the expression of 7 integrin (used here and previously as a surrogate for 4 7) by T2 B cells in humans and observed their selective recruitment into GALT where they become activated (Vossenkamper et al., 2013) . Therefore, exposure to the GALT microenvironment could be associated with multiple stages of MZB cell development from as early as the T2 stage.

The systemic autoimmune disease systemic lupus erythematosus (SLE), in particular the severe variant lupus nephritis (LN), has markedly distorted profiles of B cell subsets in blood.

T TS B CD I D DN B (Landolt-Marticorena et al., 2011; Wei et al., 2007) .

Disproportionate expansion of a population of DN cells lacking expression of CD21 and CXCR5 and with upregulated CD11c (DN2 cells) is a particular feature of LN (Jenks et al., 2018) . DN2 cells may be derived from activated naïve B cells (aNAV), driven by TLR7 engagement, resulting in the generation of self-reactive antibody producing plasma cells (Jenks et al., 2018; Tipton et al., 2015) . Interestingly, a recent study of a cohort of newly diagnosed patients with SLE demonstrated that MZB may be reduced in frequency (Zhu et al., 2018 ). Since we have previously shown that TS B cells in SLE may have significantly reduced expression of 7 integrin, we were interested to know if this may be associated with defective MZB development and the increase in aNAV and DN2 cells.

Here, we identify bifurcation in human B cell development from the T2 stage. Cells in one branch are IgM hi , express 7 integrin and are gut homing. Cells in the alternative IgM lo branch have high expression of IL4R, lower expression of 7 integrin and do not tend to enter the gut. Transcriptomically, IgM hi T2 cells share features with MZB. B cell development progresses from T1 to MZB via an IgM hi trajectory by pseudotime analysis. IgM hi T2 cells are stably IgM hi in culture and have a greater tendency to make IL10 than IgM lo cells. Markedly reduced frequency of IgM hi 7 hi T2 cells was seen in patients with severe SLE and this was associated with stark reduction in cell populations associated with MZB development. Our data link reduced access of IgM hi T2 cells to GALT with defects in all stages of MZB differentiation and enables the assimilation of these elements of human MZB differentiation into a model of human B cell development.

In mice, B cells commit to MZB differentiation soon after bone marrow emigration at the T1 stage. To seek evidence of this in humans, a deep phenotypic analysis of peripheral blood mononuclear cells (PBMC) from healthy control donors (HCD) was undertaken by mass cytometry (Fig. S1 A, B and C). SPADE on viSNE identified B cell subsets including TS B cells represented by CD27 -IgD + CD24 +++/++ CD38 +++/++ nodes that included CD10 + T1 and T2 cells as well as CD10 -T3 cells ( Fig. 1 A) (Qiu et al., 2011; Zhao et al., 2018) . T3 cells can only be definitively distinguished from naïve cells by their failure to extrude dyes such as rhodamine 123 (R123) due to lack of the ABCB1 cotransporter (Wirths and Lanzavecchia, 2005) (Fig. S1 D). Since mass cytometry cannot be used to detect dye extrusion, the boundary between T3

and naïve B cells was estimated to generate the TS SPADE bubble. As previously reported, CD27 + B cells included CD27 hi and CD27 lo cells ( Fig. 1 A) (Grimsholm et al., 2020) .

To perform a deep phenotypic analysis of TS B cells, events within the TS bubble identified in Fig. 1 A were exported and re-clustered by SPADE on viSNE using all expressed panel markers and then grouped according to gradients of loss of CD10, CD38 and CD24 and gain of CD21 corresponding to T1, T2 and T3 stages of differentiation ( Fig. 1 B) (Bemark, 2015) .

The SPADE trees branched, forming 2 chains of nodes that each extended through the T2 and T3 SPADE bubbles with no lateral connections between the branches. Branches differed most notably in their expression of IgM ( Fig. 1 C) . IgM hi T2 B cells also had lower expression of CCR7 but higher expression of 7 integrin than IgM lo T2 cells by mass cytometry. This was validated by qPCR ( Fig. 1 D and E ). In addition, IgM hi T3 cells had higher median expression of CD24 and CD45RB than IgM lo T3 cells ( Fig. 1 B and F (Fig. 1 H) . The major contributors to PCA1 and PCA2 in addition to IgM and IgD were mediators of cell traffic ( Fig. 1 I) .

Human B cells therefore segregate phenotypically as T1 cells enter into the T2 stage, forming two branches that differ in their expression of IgM and in markers of migratory potential. IgM hi T2 cells resemble IgM hi naïve cells more closely than they resemble IgM lo T2 cells with which they share markers of differentiation.

Human TS B cells can home to GALT where they become activated (Vossenkamper et al., 2013 (Bemark et al., 2013; Descatoire et al., 2014; Koethe et al., 2011; Zhao et al., 2018) . MZP and CD45RB hi T3 cells were defined by the phenotype CD27 -IgD -CD10 -CD45RB hi with expression of the ABCB1 cotransporter or not, respectively ( 

The shared surface properties of IgM hi TS with IgM hi naïve B cells ( Fig. 1 H) Data from single HCD were initially analysed individually. UMAP plots were used to visualize clusters and identify the B cell subsets they corresponded to by overlaying signal from lineage defining transcripts and CITE-Seq antibodies ( Fig. S3 and S4 ). TS B cells were identified as CD27 -IgD + clusters with high surface expression of CD38. Of the remaining CD27 -IgD + clusters that represented naïve cells, those with the top 30% of median IgM-ADT signal were designated IgM hi ( Fig. S3 and S4 ). Note that because identification of MZP and CD45RB hi T3 would require reagents that are incompatible with this method (Fig. 3 J) , they will be included in the IgM hi naïve cell groups in this analysis. CD27 + IgD + clusters that were enriched in CD1C transcripts were designated as MZB. CD27 + IgD -IgM + clusters were IFN induced genes as well as DAPP1 and FCRL5 are highly expressed by DN2 cells, although the relationship of this subset with MZB is not known (Jenks et al., 2018) . Pseudotime analysis of HCD PBMC therefore identified an IgM hi developmental trajectory from TS B cells to MZB.

We A subpopulation of human cells with a TS phenotype are regulatory, and murine IL10 producing B regulatory (Breg) cells are T2 marginal zone progenitor cells and the gut is important for their induction (Blair et al., 2010; Pillai et al., 2005; Rosser et al., 2014) . We therefore investigated the capacity of IgM hi TS B cells to produce IL10. Following 6 hours stimulation with PMA and ionomycin, IgM hi TS B cells produced significantly more IL10 than IgM lo cells ( Fig. 5 F) , inferring greater regulatory capacity of this subset.

We have previously observed reduced frequencies of circulating TS B cells expressing 7 integrin in a subset of SLE patients, implying reduced potential for TS B cells to access GALT in these cases. Data presented here implicates GALT as an important site for MZB differentiation and MZB depletion has been reported in SLE (Rodriguez-Bayona et al., 2010; Zhu et al., 2018) . Hence, we sought to determine whether our proposed MZB differentiation pathway was defective in SLE.

Flow cytometry was used to quantify B cell subsets in a cohort of 41 SLE patients and matched HCD (Table S1 ,2). Reduced MZB frequency was seen in patients with SLE compared to HCD ( We next asked if reduced MZB frequency in LN patients was intrinsic to the disease or secondary to patient demographics, immunosuppressants or disease severity. HCD, LN and OL cohorts were matched for age, gender and ethnicity (Table S1 ,2) and there was no difference in MZB frequency between Caucasian and African Caribbean healthy donors and SLE patients ( Fig. 6 K) . A subset of Caucasian SLE patients were seen to have higher MZB than African-Caribbean patients but this subset had mild disease ( Fig. 6 L) . There was no difference in immunosupression between SLE patients with high and low MZB frequencies and MZB depletion was not seen in patients with pemphigus vulgaris (PV) taking prednisolone and mycophenolate mofetil (MMF) and did not differ between SLE patients taking or not taking hydroxychloroquine (HCQ) (Fig. 6 M, N and O) . However, MZB, MZP and CD45RB hi T3 cell frequencies correlated with disease severity (Fig. 6 P) . The more marked reduction of MZB in the LN patients is therefore likely to be due to this patient group representing a severe disease cohort. As previously reported, CD27 -IgD -DN cells were more abundant in LN (Fig. 6 Q) . These were predominantly CD24 lo CD21 lo and therefore DN2 consistent with other studies (Fig. 6 R) (Jenks et al., 2018) .

MZB depletion in SLE is therefore associated with reduced frequency of MZP and T3

CD45RB hi cells. This consolidates the concept of these cells as being in a developmental continuum in health and suggests that aberrant transitional B cell maturation may result in failure of their genesis in SLE.

To identify early stages of aberrant marginal zone lineage development in LN we used mass cytometry to compare blood B cell subsets from LN patients and HCD in an undirected way Depletion of IgM hi T2 cells with high expression of 7 integrin is therefore associated with defective MZB maturation in LN patients. This affirms the association between MZB and IgM hi T2 cells in health and implicates reduced access of these cells to GALT in the breakdown of this developmental axis in patients with severe lupus (Fig. S5 D) .

We have identified branches of human B cell lineage maturation that are evident from the T2 stage. An IgM hi branch, that expresses higher levels of 7 integrin and lower levels of IL4R compared to the IgM lo branch, is gut homing. Confirmation of differentiation through IgM hi stages of differentiation from IgM hi T2, including IgM hi CD45RB hi T3 and naïve B cell variants to MZB is gained from pseudotime analysis coupled with the observed concerted reduction of the stages in this sequence in patients with severe SLE ( We have previously observed that human T2 cells are recruited into GALT where they are activated by intestinal microbes (Vossenkamper et al., 2013) . Here we demonstrate that specifically the IgM hi T2 subset of TS B cells is recruited into GALT, where they have a phenotype of activated cells including expression of CD69 and CD80. The IgM hi T2 subset is also enriched in ROR and LPS inducible genes, consistent with exposure to the microbiota.

We show that the TLR9 agonist CpG that upregulates IgM and NOTCH2 in human TS B cells (Capolunghi et al., 2008; Guerrier et al., 2012) also upregulates CD45RB on IgM hi TS and IgM hi naïve cells. PDL4, that is a lupus risk allele and the most highly upregulated gene in IgM hi compared to IgM lo TS B cells, is upregulated along the developmental pathway to MZB and limits responses to CpG (Gavin et al., 2018) . This suggests that PDL4 defects could contribute to SLE pathogenesis by impacting an aspect of the development or function of IgM hi TS B cells involving TLR9. Interestingly, PLD4 is also expressed in the splenic marginal zone in mice (Yoshikawa et al., 2010) and PLD4 knockout mice develop autoantibodies and immune complex mediated renal damage similar to SLE with LN (Gavin et al., 2018) . IgM hi TS B cells also show a transcriptomic signature indicative of retinoic acid regulation that is a feature of GALT microenvironment. Together, these data suggest that innate signals and the gut environment impact the origin, fate and function of IgM hi TS B cells.

Consistent with proposed developmental continuum from the IgM hi T2 stage through to MZB, GALT is involved in MZB development, including a stage of receptor diversification in GALT GC. However, supporting a relatively short-term transit coupled to differentiation, the frequencies of somatic mutations in MZB are lower than those of memory B cells or plasma cells in the gut (Zhao et al., 2018) . Together these data suggest that GALT transit and GC occupancy are important but transient phases in IgM hi T2 to MZB lineage progression.

Pseudotime analysis also identified a population of B cells that appear to develop from MZB and that are activated and more mature. It is possible that activation of MZB might generate a novel population of effector or memory cells.

MZB differentiation is associated with a distinctive gene expression changes and acquisition of the transcription factor ZEB2 (SIP1). ZEB2 has previously been identified as a component of a network including miR200 and TGF-1 that can regulate cell fate decisions (Gregory et al., 2008; Guan et al., 2018) . Activated TGF -1 is produced abundantly in the gut. It is possible that in addition to playing important roles in regulation of intestinal immunity as a switch factor for IgA and induction of regulatory T cells, it could also be involved in gutassociated MZB development by interactions with ZEB2 (Borsutzky et al., 2004; Chen et al., 2003) .

Collapse of the MZB developmental pathway in severe SLE was accompanied by expanded T3, aNAV and DN2 cell populations. Expansion of aNAV and DN2 populations is a product of excessive TLR7 and IFN-signalling. We were therefore interested in enrichment of IFN- (Laing et al., 2020; Woodruff et al., 2020) .

LN represents a severe lupus subtype associated with the worst clinical outcomes (Yap et al., 2012) . B regulatory (Breg) IL10 responses associated with expression of CD80 and CD86

are defective in SLE (Blair et al., 2010) , permitting aberrant T effector functions (Oleinika et al., 2019) . In mice Breg cells are IgM hi CD21 hi CD23 hi T2 MZP cells and interaction with the gut microbiome is essential for their induction (Evans et al., 2007; Rosser et al., 2014) . We have identified that IgM hi TS B cells express CD80 in GALT and represent the predominant IL10

producing TS B cell subset. Their depletion in LN may be synonymous with the loss of Breg IL10 responses and associated with the lack of T cell regulation in SLE. MZB confer immunity to encapsulated bacteria such as pneumococcus, thus their depletion in LN may confer increased risk of such infections SLE (Danza and Ruiz-Irastorza, 2013) . This also reinforces the importance of pneumococcal vaccination in this patient cohort.

In summary, we identify an MZB maturation pathway that becomes evident at the T2 stage of B cell development and that is depleted in severe SLE. Traffic through GALT is a component of this pathway that is potentially linked to the induction of human IL10

producing Breg cells (Rosser et al., 2014) . Together, this affirms the importance of tissue microenvironments in shaping the B cell functional repertoire and maintaining health.

Understanding the regulators of early B cell fate will be a key to resolving the disturbances in B cell function in severe SLE.

We 

The authors declare no competing interests. P) The proportion of MZB, MZP and T3 CD45RB hi showed a negative correlation with disease activity as indicated by the SLEDAI score S ank coefficient). 

Information regarding reagents and resources should be directed to Professor Jo Spencer (jo.spencer@kcl.ac.uk).

The datasets generated during this study will be made available on acceptance of the manuscript.

All blood and tissue samples were obtained from adults with REC approval and informed consent. SLE patients were recruited using the following criteria; i) fulfilment of 4 or more revised ACR classification criteria; ii) ANA positive; iii) Biologic (Belimumab or rituximab) naïve; iv) Immunosuppressive regimen does not include azathioprine or cyclophosphamide within 6 months of sample collection due to the severe depletion of naïve B cells by these medications. All LN patients had diagnostic confirmation by renal biopsy. Blood was obtained from SLE patients and HCD (REC reference 11/LO/1433: Immune regulation in autoimmune rheumatic disease). Paired gut biopsies and blood were obtained from individuals undergoing colonoscopies in whom no mucosal abnormality was detected (REC reference 11/LO/1274: Immunology the intestine; features associated with autoimmunity).

Samples of draining the gut via the hepatic portal vein were obtained from liver perfusion prior to transplantation (REC reference 09/H0802/100: The role of innate immune system in hepatic allograft outcome). Patient demographic data can be found in Tables S1 and S2.

Blood samples were diluted 1:1 in RPMI-1640 containing 10% foetal calf serum (FCS), 100 U/ml penicillin and (RPMI-P/S). Diluted blood was then layered onto Ficoll and centrifuged for 25 minutes with brake and accelerator set to 0. The buffy coat layer was then removed and cells were washed in RPMI-P/S. PBMC isolated from patients undergoing colonoscopy was used fresh, whilst PBMC used for the analysis of HCD and patients with SLE, UC, GPA and PV was cryopreserved in FCS + 10% dimethyl sulfoxide (DMSO). Mononuclear cells from gut were obtained by the removal of epithelial cells with 1mM EDTA in HBSS containing 100 U/ml penicillin and for 30 minutes. Collagenase digest was then used to generate a cell suspension using collagenase D (1 mg/ml) and DNase (10 U/ml) in RPMI-P/S for 1 h.

3 mass cytometry panels were utilized, the staining protocols were as follows. Metal tagged antibodies used are listed in Fig S1 and The Key Resources Table. Cells were then washed twice in 1xPBS and fixed overnight in 16% paraformaldehyde. The following day cells were washed in 1xPBS and DNA was stained with 1 M Intercalatin in 500ul permeabilization buffer at room temperature for 20 minutes. Cells were then washed twice in 1xPBS and twice in Milli-Q water before being resuspended in Milli-Q water plus EQ beads to a concentration of 0.5 x 10 6 /ml and run on a Helios Mass cytometer.

Panel 2: As for panel 1 except cells were stained fresh, were not enriched and 2 x 10 6 cells were viability stained with 1ml rhodium intercalator diluted in 1:500 in PBS for 20 minutes at room temperature. Metal tagged antibodies used are listed in Fig S2 and The Key

Resources Table. Panel 3: As per panel 1 except cells were not enriched and IgG staining was not performed.

Metal tagged antibodies used are listed in Fig S6 and The Key Resources Table. Analysis of mass cytometry data FCS files were normalised using Nolan lab software (v0.3, available online at https://github.com/nolanlab/beadnormalization/releases). Pre and post normalisation plots are shown in Fig. S1 B and F and S5 A for the respective datasets. Where files were concatenated the Cytobank FCS File Concatenation Tool was used (available online at https://support.cytobank.org/hc/en-us/articles/206336147-FCS-file-concatenation-tool).

Files were then loaded onto the Cytobank (https://mrc.cytobank.org/) and gated to identify live CD19 B cells and analysed as described in Fig. S1 C and G and S5 C.

For the analysis of HCD PBMC in Fig. 1 , viSNE was run on equal numbers of CD19 + events (n=35000) from each HCD (n=10). SPADE was then run on the viSNE coordinates and B cell subsets were identified by placing nodes into bubbles. The TS bubble was identified as CD27 -IgD + CD24 +++/++ CD38 +++/++ . Events within the TS bubble were exported and a further viSNE was run using equal events (n=3535) and all panel markers except CD45, CD3, CD14

and class switched isotypes IgA and IgG which are not expressed by TS B cells. CD45 was excluded due to homogenous expression and lack of contribution to clustering. SPADE was then run on the viSNE coordinates and TS populations were defined as demonstrated in Fig.   1 B.

For the analysis of PBMC and GALT derived B cells in Fig. 2 , equal numbers of CD19 + events (n=118,934) from concatenated PBMC (n=7) and GALT (n=7) samples were used to run a viSNE using all markers except for CD45, CD3 and CD14. SPADE was then run on the viSNE coordinates and TS B cells identified as CD27 -IgD + CD10 + nodes. Events within the TS bubble were then exported and equal numbers of events (n=4520) were clustered using FlowSOM.

CD10, CD24, CD38, IgM as clustering channels to allow the undirected visualization of markers of TS populations.

For the analysis of PBMC from HCD and SLE samples in Fig 7, CITRUS was run using equal numbers of CD19 + events (n=20000) from HCD (n=8) and SLE patients (n=8) and the following clustering channels : CD5, CD9 CD10, CD24, CD27, CD38, CD45RB, IgD, IgM, IgA.

Due to event sharing amongst CITRUS nodes, node 321672 identified in Fig. 7 A contains all CD27 -IgD + CD24 +++/++ CD38 +++/++ events and was therefore used for analysis of TS B cells. All events from this node were exported and FlowSOM was run using equal event sampling (n= 657) and using all marker channels except CD45, CD3, CD14 and IgA. 

Fresh PBMC was isolated from HCD and incubated for 6 hours at 37 degrees with phorbol 12-myristate 13-Acetate (PMA) 50ng/ml and Ionomycin 250ng/ml with Golgiplug at a dilution of 1:1000. Cells were then surface stained as above followed by fixation with Cytofast buffer (Biolegend). Cells were then washed twice and stained with resuspended conjugated antibodies in permeabilization / wash buffer (Biolegend) for 20 minutes at room temperature.

Sorted IgM hi and IgM lo TS and naïve (CD27 -IgD + CD10 -) B cell subsets were plated onto a 96 well plate seeded with 2 x 10 4 cells per well. Wells containing CD40L expressing HEK cells were also seeded with 2 x 10 4 irradiated HEK cells per well. Cells were then stimulated with CPG-ODN 2.5 g/ml or anti-IgM 10 g/ml. Proliferation assays were performed on cells stained with CellTrace violet as per the manufacturer s guidelines. Cells were then stained and analysed by flow cytometry as above.

Single cell RNA sequencing library preparation Sorted cell populations were loaded onto a 10x Genomics Chromium Controller expression, VDJ and ADT (for samples in Fig. 4) were prepared according to the manufacturers guidelines. Samples used in Fig. 3 were sequenced using an Illumina NextSeq 500 platform. Samples used in Fig. 4 were sequenced using an Illumina HiSeq 2500 High Output platform. The 10x Genomics Cellranger workflow was then used for transcript alignment and the generation of sparse matrices for downstream analysis.

Cryopreserved samples were thawed and sorted using the gating strategy in Fig. S3 A. Cells were then washed and stained in a CITE-Seq antibody cocktail at a concentration of 8 g/ml for 30 minutes on ice. Cells were then washed three times before loading onto the 10x

Chromium controller.

Single cell sequencing analysis

The Seurat R package (vs 3.1.1) was used to filter data to remove cells with low numbers of RNA transcripts, doublets and cells with high levels of mitochondrial transcripts indicative of cell death. Immunoglobulin variable genes were then removed from the dataset as well as cells with low expression of B cell genes CD79A, CD79B, CD19 or MS4A1. Data from IgM hi and IgM lo TS B cells were merged and the data was transformed in accordance with the SCTransform workflow before UMAP based reduction of dimensionality and PCA-based clustering to identify populations (Hafmeister, 2020) . Heatmaps were then created using select genes from the top 60 differentially expressed genes in each sample, and dot plots and violin plots on selected genes. Data from sorted CD19 + cells from HCD used for Fig. 4 and Fig. S3 and 4 was initially analysed individually followed by an integrated analysis.

Individual analysis was performed using the quality control (QC) steps as well as the removal of IGHV genes and non-B cells as described above. Data was then normalized and scaled and UMAP run on a PCA generated using 2000 variable genes. Overlay of ADT and gene signal, violin plots and median expression of markers by UMAP clusters was used to identify which B cell subsets they corresponded to. For the integrated data analysis, data from 3 HCD was filtered using the QC steps as well as the removal of IGHV genes and non-B cells as described above. Data was then normalized using the SCTransform wrapper in Seurat followed by integration using The Satija Laboratory Integration and Label Transfer protocol (Butler et al., 2018) , using 3000 integration features. The 2000 most variable genes were then used to perform PCA and a 3-dimensional UMAP was obtained from this.

Clusters were obtained using the FindNeighbours and FindClusters functions within Seurat, using default parameters. The UMAP coordinates and cluster allocations were then used to run Slingshot (Street et al., 2018) . Randomised S1 . Gating and analysis of mass cytometry data used for Fig. 1 and 2. A) Mass cytometry panel used for analysis in Figure 1 . B) Pre-and post normalization plots of mass cytometry data used for Fig.1 . C) Gating strategy of mass cytometry data to identify live CD19 + B cells. D) Flow cytometry plots demonstrating identification of T3 as CD27 -IgD + CD10 -R123 hi and naïve (N) B cells as CD27 -IgD + CD10 -R123 lo . E) Mass cytometry panel used for analysis in Figure 2 . F) Pre-and post normalization plots of mass cytometry data used for Fig. 2 Fig. S4 . Identification of B cell subsets represented by UMAP clusters in 10x HCD 2 and 3.

A) UMAP plot generated from a PCA run on 2000 differentially expressed genes from 10x HCD 2. Clusters were merged and pseudocoloured to represent B cell subsets defined by ADT and gene signal of lineage defining targets as described in Figure S4D Table S1 . SLE patient demographic data used for Figure 6 and 7. C = Caucasian, AC = African Caribbean, SEA = South East Asian, MMF = mycophenolate mofetil, PRED = prednisolone, HCQ = hydroxychloroquine. Figure 5 and 6. C = Caucasian, AC = African Caribbean, SEA = South East Asian, MMF = mycophenolate mofetil, PRED = prednisolone, HCQ = hydroxychloroquine. Table S2 . HCD demographic data used in Fig. 6 . SEA = south east Asian, IA = Indian Asian. Table S3 . Demographic data of patients with other autoimmune diseases. GPA = granulomatosis with polyangiitis, PV = pemphigus vulgaris, UC = ulcerative colitis, AZA = azathioprine, MTX = methotrexate, MES = mesalazine.

A Reassessment of IgM Memory Subsets in Humans

Dimensionality reduction for visualizing single-cell data using UMAP

Translating transitions -how to decipher peripheral human B cell development

A glycosylationdependent CD45RB epitope defines previously unacknowledged CD27(-)IgM(high) B cell subpopulations enriched in young children and after hematopoietic stem cell transplantation

CD19(+)CD24(hi)CD38(hi) B cells exhibit regulatory capacity in healthy individuals but are functionally impaired in systemic Lupus Erythematosus patients

TGF-beta receptor signaling is critical for mucosal IgA responses

Integrating single-cell transcriptomic data across different conditions, technologies, and species

Clinical and diagnostic relevance of NOTCH2-and KLF2-mutations in splenic marginal zone lymphoma

CpG drives human transitional B cells to terminal differentiation and production of natural antibodies

Conversion of peripheral CD4+CD25-naive T cells to CD4+CD25+ regulatory T cells by TGF-beta induction of transcription factor Foxp3

Infection risk in systemic lupus erythematosus patients: susceptibility factors and preventive strategies

Identification of a human splenic marginal zone B cell precursor with NOTCH2-dependent differentiation properties

Novel suppressive function of transitional 2 B cells in experimental arthritis

PLD3 and PLD4 are single-stranded acid exonucleases that regulate endosomal nucleic-acid sensing

The miR-200 family and miR-205 regulate epithelial to mesenchymal transition by targeting ZEB1 and SIP1

The Interplay between CD27(dull) and CD27(bright) B Cells Ensures the Flexibility, Stability, and Resilience of Human B Cell Memory

ZEB1, ZEB2, and the miR-200 family form a counterregulatory network to regulate CD8(+) T cell fates

TLR9 drives the development of transitional B cells towards the marginal zone pathway and promotes autoimmunity

Normalization and variance stabilization of single-cell RNA-seq data using regularized negative binomial regression

Transitional B cells commit to marginal zone B cell fate by Taok3-mediated surface expression of ADAM10

Kruppel-like factor 2 (KLF2) regulates B-cell reactivity, subset differentiation, and trafficking molecule expression

Distinct Effector B Cells Induced by Unregulated Toll-like Receptor 7 Contribute to Pathogenic Responses in Systemic Lupus Erythematosus

Pivotal advance: CD45RB glycosylation is specifically regulated during human peripheral B cell differentiation

Expression of the mucosal vascular addressin, MAdCAM-1, on sinus-lining cells in the spleen

A dynamic COVID-19 immune signature includes associations with poor prognosis

Increased expression of B cell activation factor supports the abnormal expansion of transitional B cells in systemic lupus erythematosus

Effector and regulatory B cells in immunemediated kidney disease

Novel human transitional B cell populations revealed by B cell depletion therapy

Marginal zone B cells

Extracting a cellular hierarchy from highdimensional cytometry data with SPADE

Primary Sjogren's syndrome is characterized by distinct phenotypic and transcriptional profiles of IgD+ unswitched memory B cells

Decreased frequency and activated phenotype of blood CD27 IgD IgM B lymphocytes is a permanent abnormality in systemic lupus erythematosus patients

Regulatory B cells are induced by gut microbiota-driven interleukin-1beta and interleukin-6 production

Slingshot: cell lineage and pseudotime inference for single-cell transcriptomics

Differential expression of CD21 identifies developmentally and functionally distinct subsets of human transitional B cells

Diversity, cellular origin and autoreactivity of antibody-secreting cell population expansions in acute systemic lupus erythematosus

FlowSOM: Using self-organizing maps for visualization and interpretation of cytometry data

A role for gut-associated lymphoid tissue in shaping the human B cell repertoire

A new population of cells lacking expression of CD27 represents a notable component of the B cell memory compartment in systemic lupus erythematosus

IgM memory B cells: specific effectors of innate-like and adaptive responses

Human blood IgM "memory" B cells are circulating splenic marginal zone B cells harboring a prediversified immunoglobulin repertoire

ABCB1 transporter discriminates human resting naive B cells from cycling transitional and memory B cells

Critically ill SARS-CoV-2 patients display lupus-like hallmarks of extrafollicular B cell activation

Survival analysis and causes of mortality in patients with lupus nephritis

Phospholipase D family member 4, a transmembrane glycoprotein with no phospholipase D activity, expression in spleen and early postnatal microglia

Defective B cell tolerance checkpoints in systemic lupus erythematosus

Spatiotemporal segregation of human marginal zone and memory B cell populations in lymphoid tissue

Altered frequencies of memory B cells in new-onset systemic lupus erythematosus patients

Target Clone 141Pr Figure 3J share high IgM expression (MFI mean +/-SD, paired t test). N) T3 and naïve CD45RB hi subsets share similar high surface expression of CD24 (MFI mean +/-SD, paired t test).