1. Immunology and Inflammation

Complex subsets but redundant clonality after B cells egress from spontaneous germinal centers

  1. Carlos Castrillon  Is a corresponding author
  2. Lea Simoni
  3. Theo van den Broek
  4. Cees van der Poel
  5. Elliot H Akama-Garren
  6. Minghe Ma
  7. Michael C Carroll  Is a corresponding author
  1. Program in Cellular and Molecular Medicine, Boston Children’s Hospital, United States
  2. Harvard-MIT Health Sciences and Technology, Harvard Medical School, United States
  3. Department of Pediatrics, Harvard Medical School, United States
https://doi.org/10.7554/eLife.81012
Share this article
Cite this article
  1. Carlos Castrillon
  2. Lea Simoni
  3. Theo van den Broek
  4. Cees van der Poel
  5. Elliot H Akama-Garren
  6. Minghe Ma
  7. Michael C Carroll
(2023)
Complex subsets but redundant clonality after B cells egress from spontaneous germinal centers
eLife 12:e81012.
https://doi.org/10.7554/eLife.81012
  2. Download BibTeX
  3. Download .RIS

Abstract

Affinity matured self-reactive antibodies are found in autoimmune diseases like systemic lupus erythematous. Here, we used fate-mapping reporter mice and single-cell transcriptomics coupled to antibody repertoire analysis to characterize the post-germinal center (GC) B cell compartment in a new mouse model of autoimmunity. Antibody-secreting cells (ASCs) and memory B cells (MemBs) from spontaneous GCs grouped into multiple subclusters. ASCs matured into two terminal clusters, with distinct secretion, antibody repertoire and metabolic profiles. MemBs contained FCRL5+ and CD23+ subsets, with different in vivo localization in the spleen. GC-derived FCRL5+ MemBs share transcriptomic and repertoire properties with atypical B cells found in aging and infection and localize to the marginal zone, suggesting a similar contribution to recall responses. While transcriptomically diverse, ASC and MemB subsets maintained an underlying clonal redundancy. Therefore, self-reactive clones could escape subset-targeting therapy by perpetuation of self-reactivity in distinct subsets.

Editor's evaluation

Understanding the heterogeneity of the B cell response induced in autoimmune individuals is important for the development of therapies designed to target the cells underlying disease progression. Here, the authors use a new model of autoimmunity to assess the heterogeneity of the B cell response using scRNA-seq and scBCR-seq and found that B cell responses are similar to those by exogenous protein immunization.

https://doi.org/10.7554/eLife.81012.sa0

Introduction

In autoimmune diseases like systemic lupus erythematosus (SLE), pathogenesis is caused by self-reactive antibody-secreting cells (ASCs) and sustained by memory B cells (MemBs). Numerous subsets of these B cells have been described in models of health and disease, but it is unknown whether and how these subsets change and relate to each other in autoimmune disease. Among MemBs, the major known subsets are conventional germinal center (GC)- and non-GC-derived MemBs (Taylor et al., 2012; Viant et al., 2021). In addition, atypical MemBs (AtMemBs) are present in models of acute and chronic infection in mice (Kim et al., 2019; Song et al., 2022), as well as in parasitic infection in humans (Portugal et al., 2017) and are similar to the age-associated B cells first described in ageing and autoimmune mouse models (Phalke and Marrack, 2018). DN2 (IgD−CD27−) cells resemble AtMemB cells and are an expanded subset of B cells in human SLE. They likely have a naive B cell origin, and data suggest they can be readily activated to mature into ASCs in the extrafollicular space (Jenks et al., 2018). Indeed, these subsets of B cells maintain similarities to marginal zone (MZ) B cells, including the capacity to differentiate into ASCs. However, it is unclear how diverse these distinct subsets are and how clonally related they might be.

ASCs can also segregate into different subsets with specific properties related to lifespan (short- and long-lived) (Nutt et al., 2015; Radbruch et al., 2006) and origin (GC- and EF-derived) (Elsner and Shlomchik, 2020), though data on the differences in the metabolic profile and clonal overlap between these populations are still lacking. Importantly, experiments with LPS-driven derivation of B cells into ASCs have demonstrated the existence of multiple clusters of ASCs as they mature into alternate terminal states (Scharer et al., 2020). Overall, the complexity of the ASC compartment remains underexplored.

In SLE, isotype-class switching and somatic hypermutations of autoreactive antibodies indicate antibody maturation in GCs (Cappione et al., 2005). Therefore, to investigate the post-GC subpopulations of B cells in SLE, we utilized a chimeric mouse model where self-reactive clones, which are derived from B cells with a wild-type (WT) B cell receptor (BCR) repertoire, compete in spontaneous GCs (Degn et al., 2017). We used fate mapping and single-cell RNA sequencing to gain insight into the post-GC populations of B cells in both autoimmune chimeras and immunized chimeras. We found a diversity of ASCs at different maturation stages with distinct terminal states and multiple MemB cell clusters, including GC-derived FCRL5+ and CD23+ MemB cells. These ASCs and MemB subsets showed specific transcriptomic profiles, antibody repertoire profiles, and localization, suggesting functional diversity yet maintaining redundant clonality. Together, our findings outline the underlying complexity of the MemB and ASC compartments and indicates that therapeutic efforts to target autoreactive B cells must consider the likely reservoirs of self-reactivity in other subsets.

Results

Fate-mapped B cells are distributed into four main clusters

To track WT B cells as they develop in an autoimmune environment, we used a mixed bone marrow (BM) chimera model (Akama-Garren et al., 2021; Degn et al., 2017; van der Poel et al., 2019). Irradiated host mice received a combination of BM donor cells from WT and 564 Igi mice (Berland et al., 2006), which bear transgenic BCRs with anti-RNP specificities. In this model, WT donor B cells expand, are selected into spontaneous GCs, mature into ASCs, and contribute to the circulation of self-reactive antibodies. To track activated B cells and their derived populations, Aicda-CreERT2-EYFP fate-mapping reporter mice (Dogan et al., 2009) were used as donors for the WT B cell repertoire (Figure 1A). Fate-mapped WT B cells were purified, sorted, and processed for droplet-based single-cell RNA-seq. Moreover, we compared the response to that of foreign antigen WT chimeras immunized with a primary and secondary dose of haptenated protein.

Figure 1 with 1 supplement see all
Download asset Open asset
B cells with a WT repertoire are found in all compartments of autoimmune chimeras.

(A) Experimental setup to produce the autoimmune and immunized chimeras and selection for single-cell sequencing. (B–H) Single-cell RNA-seq data from autoimmune (n = 5) and immunized (n = 3) chimeras. (B) Two-dimensional UMAP representation of single-cell transcriptomic data from all cells from autoimmune and immunized chimeras. (C) Gene expression of conventional markers was used to assign an identity to the clusters. Log normalized expression. (D) Proportion of cells belonging to the antibody-secreting cell (ASC), DZ, LZ, and memory B cell (MemB) clusters among all cells sequenced from each chimera. Each dot represents a mouse. (E) Phylogenetic tree of the most expanded clone in one of the autoimmune chimeras. Clonal members colored by assigned cluster. (F) Accumulated replacement mutations in each cluster by condition. Violin plot with embedded Tukey boxplot. (G) Isotype usage per cluster for each condition. (H) Clonal usage of selected V genes by condition. Each dot represents a mouse in a Tukey boxplot. IGHV 1–75 was only found in two immunized chimeras. (I) Z-score ranking of fluorescent detection for monoclonal pools Mab1 (left) and Mab2 (right) after HuProt microarray binding assay. The monoclonals originate from two autoimmune chimeras. Statistical values correspond to two-tailed Mann–Whitney test (F) and unpaired t-tests (H).

Figure 1—source data 1

Data for Figure 1D.

https://cdn.elifesciences.org/articles/81012/elife-81012-fig1-data1-v2.csv
Figure 1—source data 2

Data for Figure 1F.

https://cdn.elifesciences.org/articles/81012/elife-81012-fig1-data2-v2.csv
Figure 1—source data 3

Data for Figure 1H.

https://cdn.elifesciences.org/articles/81012/elife-81012-fig1-data3-v2.csv

A total of 12,839 cells (9535 from 5 autoimmune chimeras, 3304 from 3 immunized chimeras) were retained post quality control for further analysis. Expression of the EYFP reporter transcripts was confirmed. Using unsupervised clustering, the cells grouped into four clusters (Figure 1B), representing the major known B cell compartments: GC B cells (e.g. S1pr2), which are either dark zone (DZ, e.g. Mki67) or light zone (LZ, e.g. Cd83), ASCs (e.g. Xbp1), and MemBs (e.g. Sell) (Figure 1C, Figure 1—figure supplement 1A). All compartments, LZ, DZ, MemB, and ASC clusters were clearly represented in both conditions (Figure 1D).

Using paired single-cell BCR sequencing and repertoire analysis, we observed clones with clear expansion in both autoimmune and immunized chimeras (Figure 1—figure supplement 1B). Indeed, in both conditions, we were able to find clones that had members belonging to all four clusters, exemplified by the phylogenetics of the most expanded clone in one autoimmune chimeric mouse (Figure 1E). The GC mutational level was in a similar range between conditions, although autoimmune chimeras had more DZ and LZ replacement mutations. MemBs in autoimmune and immunized chimeras had a similar number of nucleotide replacement mutations. However, ASCs accumulated more mutations in immunized chimeras (Figure 1F). Nevertheless, ASCs in both conditions reached similar maximum levels of mutations (20 nt in autoimmune and 19 nt in immunized chimeras).

Cells from autoimmune chimeras showed more isotype diversification in all compartments (Figure 1G). Importantly, both preferential and shared clonal usage of heavy chain genes were observed among the most used V genes in each condition. Seven V genes were shared among the top 10 V genes in autoimmune (Figure 1—figure supplement 1C) and immunized chimeras (Figure 1—figure supplement 1D): IGHVs 1–26, 1–53, 1–64, 1–72, 3–6, 6–3, and 9–3. V genes traditionally associated with nitro-phenyl (NP) specificity, like IGHVs 1–72 (a.k.a. V168) 6–3 and 1–53 (Xue et al., 2019), were more expanded upon immunization. On the other hand, autoimmune chimeras preferred IGHVs 1–50, 1–75, and 8–8, whereas IGHV 1–64 was used at a similar proportion in both conditions (Figure 1H). Notably, commonly used V genes in both conditions had clonal usage frequency in a similar range to that observed in B cells from naïve mice (Figure 1—figure supplement 1E; Rettig et al., 2018). IGHV genes 1–72, 1–55, 1–64, 1–53, and 1–50 sequences are among the most closely related to that of the 564 IgI autoantibody (Figure 1—figure supplement 1F).

To investigate the potential targets, we produced six monoclonal antibodies using the VDJ sequences from two autoimmune chimeric mice. The VDJ sequences originated from members of four different clones, including the most expanded clone from one mouse. We tested the reactivity of these antibodies using a native protein HuProt array containing more than 20,000 recombinant proteins. We combined the monoclonal antibodies into two pools (Mab1 and Mab2), each containing two distinct clones. We identified multiple potential binders, as exemplified by the top 30 rank based on fluorescence Z-score for each pool (Figure 1I). The potential targets include the housekeeping genes ARIH1, ARHGEF39, and DNAJC7, as well as the poorly described C6orf165 (CFAP206). We also observed reactivity for the known autoantigens SCL-70 and Sm/RNP, and the single-stranded binding protein SSBP1.

Thus, using single-cell RNA-seq, we confirmed that in the 564Igi mixed BM chimeric autoimmune model, B cells with a WT BCR repertoire break tolerance, expand in spontaneous GCs and develop into MemBs and ASCs in a seemingly unrestricted manner, much like they do in response to foreign antigens.

ASCs have alternative terminal states

As ASCs progress in development, they downregulate Cd19 and upregulate Sdc1 (Syndecan-1, CD138). Indeed, clear changes in the expression level of these markers were observed across the seven ASC subclusters (Figure 2A, B, left), and this information was used to define cluster ASC_Early_1, which bears the highest Cd19 and lowest Sdc1 expression, as the start of pseudotime analysis with Slingshot (Street et al., 2018). Two pseudotime lineages were observed (Figure 2B, right), concluding in two different clusters: ASC_Late_1 and ASC_Late_2. While the seven clusters displayed a similar overall transcriptional profile (Figure 2—figure supplement 1A), they could be characterized by the expression of marker genes, among them: Fcmr, Cd52, Il21r, Cd74, and Ms4a1 for ASC_early1; Vim, Mki67, Ccna2, Ccnb2, and Top2a for ASC_early_2; Ssr3, Selenof, Selenok, Lgals1, and Kdelr2 for ASC_Mid_1; Slpi, Kdelr1, Bst2, Eif5a, Ctsb, and Ubb for ASC_Mid_2; Cxcr4, Ccnd2, Lars2, Itgal, and Trp53inp1 for ASC_Mid_3; Ptprc, Atf6, Hspa5, Slamf7, and Itga4 for ASC_Late_1; Bcl2, Ly6e, Il2rg, Tnfrsf17, Tnfrsf13b, and Lamp1 for ASC_Late_2 (Figure 2—figure supplement 1B). All seven subclusters were observed both in autoimmune and immunized chimeras (Figure 2—figure supplement 1C).

Figure 2 with 1 supplement see all
Download asset Open asset
Antibody-secreting cells (ASCs) mature into two distinct terminal clusters.

(A) Two-dimensional UMAP representation of ASCs and their assigned subcluster for all ASC cells from autoimmune and immunized chimeras. (B) Cluster normalized gene expression levels for Cd19 and Sdc1 (left) and Slingshot pseudotime-based trajectory of ASCs (right). Tukey boxplot for gene expression. (C) Normalized gene expression level of Xbp1, Atf6, and Hspa5 in ASC_Late_1 and ASC_Late_2. (D) Total counts of Ig transcripts for all ASC subclusters, including both autoimmune and immunized chimeras. (E) Normalized gene expression level of surface markers Ptprc, Itgb2, Itga4, and Cd28 in ASC_Late_1 and ASC_Late_2. (F) Single-cell level scoring for MYC and OXPHOS signature profiling of ASC subclusters with AUCell, split by condition. Statistical values correspond to one-way analysis of variance (ANOVA) with Tukey correction for multiple comparisons (F).

Figure 2—source data 1

Data for Figure 2F, antibody-secreting cell (ASC) OXPHOS autoimmune graph.

https://cdn.elifesciences.org/articles/81012/elife-81012-fig2-data1-v2.csv
Figure 2—source data 2

Data for Figure 2F, antibody-secreting cell (ASC) OXPHOS immunized graph.

https://cdn.elifesciences.org/articles/81012/elife-81012-fig2-data2-v2.csv

The two terminal clusters differed in key aspects. Xbp1 (X-box binding protein 1), a key modulator of the endoplasmic reticulum (ER) stress response and the unfolded protein response and a major regulator of the transition from B cell to ASC, had slightly more expression in ASC_Late_2 than in ASC_Late_1, whereas ASC_Late_1 had Atf6 and Hspa5 as markers (Figure 2C). These two genes encode Activation transcription factor 6 and Heat Shock Protein Family A (HSP70) Member 5, which are also involved in the regulation of ER stress. These results indicate that terminally differentiated ASCs might use different mechanisms to handle ER stress. Moreover, ASC_Late_2 had more reads mapped to Ig genes than ASC_Late_1 (Figure 2D). Similar observations regarding differential usage of ER stress modulation pathways have been made using LPS-stimulated B cells (Scharer et al., 2020). However, we are not aware of other reports identifying in vivo genetic and repertoire differences of these terminally differentiated ASCs in spontaneous GCs. Although few surface markers were unique to the two subclusters, higher expression of Ptprc (CD45, B220), Itga4 (CD49d), and Itgb2 (CD18) were observed in ASC_Late_1 cells and Cd28 in ASC_Late_2 (Figure 2E).

Further characterization of the subclusters using gene set activity analysis with AUCell (Aibar et al., 2017) identified similar levels of gene expression for the Myc pathway, as expected, with the exception of ASC_early_2, which was characterized by proliferation markers. Importantly, a higher score signature of oxidative phosphorylation (OXPHOS) in ASC_Late_2 relative to ASC_Late_1 was observed, suggesting specific metabolic requirements (Figure 2F).

We analyzed the antibody repertoire for the two terminal clusters, taking advantage of the paired Ig heavy and light chain sequence results. In our dataset, ASC_Late_2 consisted mainly of IgM ASCs, regardless of the immune status (Figure 3A), whereas ASC_Late_1 cells had a highly diverse isotype usage, with IgG2c being the most common in autoimmune mice. ASC_Late_1 cells accumulated more replacement mutations than ASC_Late_2 cells, and more replacement mutations were observed in immunized than in autoimmune chimeras (Figure 3B). Comparison of the clonal composition of the ASC clusters revealed extensive clonal overlap (Figure 3C) in both conditions. This indicates that although the terminal states have transcriptomic and repertoire differences, clones have members with a potential of expansion in any cluster. Moreover, analysis of clonal usage of V genes identified multiple instances of redundancy in the ASC_Late clusters, for example IGHV 1–26, 1–69, and 14–4, among others (Figure 3D).

Figure 3
Download asset Open asset
Terminal states of antibody-secreting cells (ASCs) have different repertoire characteristics but maintain clonal overlap.

(A) ASC subcluster isotype usage proportion by condition. (B) Replacement mutation accumulation in ASC_Late_1 and ASC_Late_2 by condition. Violin plot with Tukey boxplot overlay. (C) ASC subcluster clonal intersects in autoimmune (left) and immunized (right) chimeras, including all ASCs per condition. Connector ribbons’ width represents number of shared clones (see Supplementary file 1). (D) Clonal V-gene usage in ASC_Late_1 and ASC_Late_2 of autoimmune chimeras, per mouse. Only V genes used in two clones or more are displayed. Statistical values correspond to two-tailed Mann–Whitney test (B).

Figure 3—source data 1

Data for Figure 3B.

https://cdn.elifesciences.org/articles/81012/elife-81012-fig3-data1-v2.csv

In summary, we identified multiple ASC clusters as they differentiated in both autoimmune and immunized environments, with pseudotime analysis suggesting two terminal states with divergent transcriptomic and VDJ maturation profiles, as well as potentially distinct capacities for antibody secretion but that nonetheless share or contain similar clonal members.

MemBs are composed of diverse clusters with differences in their transcriptome and repertoire

An important factor in the pathology of autoimmune disease is the self-sustaining chronicity of autoreactive B cells, which suggests the presence of MemBs. However, it is unclear whether the MemB compartment is similar in autoimmune and immunized responses, and little is known about its internal complexity. To address this question, we reclustered B cells identified as MemBs (Figure 1B), identifying four subclusters (Figure 4A). Notably, all MemB subclusters were observed for both autoimmune and immunized mice (Figure 4B). Remarkably, clusters MemB_1 and MemB_3 accumulated more replacement mutations than clusters MemB_2 and MemB_4, though the latter two still contained cells with similar maximum mutations as MemB_1 and MemB_3 (Figure 4C). Examination of Ig isotype revealed that Cluster MemB_2 was mostly composed of IgM cells in both autoimmune and immunized chimeras. A hallmark of the inflammatory interferon-driven response common in lupus and viral infections is expression of IgG2c. While this isotype was nearly absent among the four subclusters of MemB in the immunized mice, all four subclusters of MemB in autoimmune chimeras included the inflammatory isotype (Figure 4D).

Figure 4 with 1 supplement see all
Download asset Open asset
Memory B cells (MemBs) are grouped in four subclusters with distinct transcriptomic and BCR profiles.

(A) UMAP distribution of all cells in the MemB compartment colored by subclusters. (B) Proportion of MemBs in each subcluster per mouse. Each dot represents a mouse. (C) Replacement mutation accumulation in each MemB subcluster by condition. Violin plots embedded with Tukey boxplots. (D) Proportion of isotype usage for each MemB subcluster by condition. (E) Normalized gene expression of markers that characterize each MemB subcluster. Combined expression from autoimmune and immunized chimeras. Statistical values correspond to two-tailed Mann–Whitney tests (C).

Figure 4—source data 1

Data for Figure 4B.

https://cdn.elifesciences.org/articles/81012/elife-81012-fig4-data1-v2.csv
Figure 4—source data 2

Data for Figure 4C, by cluster.

https://cdn.elifesciences.org/articles/81012/elife-81012-fig4-data2-v2.csv
Figure 4—source data 3

Data for Figure 4C, within cluster.

https://cdn.elifesciences.org/articles/81012/elife-81012-fig4-data3-v2.csv

Overall, although distinct, the clusters displayed extensive overlap in gene expression (Figure 4—figure supplement 1A). Comparing the transcriptomic profiles of the four clusters revealed distinct markers for each cluster (Figure 4E, Figure 4—figure supplement 1B). MemB_1 cells had the highest expression of Cd83, a common B cell activation marker typically observed in LZ B cells, and of Il4i1, plus they had higher expression of Apex1, Eif5a, Eif4a1, Slc25a5, C1qbp, and Mif. MemB_2 cells had the highest expression of Fcrl5 and Cd72, similar to DN2 cells in humans (Jenks et al., 2018) and atypical memory B cells in mice and humans (Kim et al., 2019), along with higher expression of Zeb2, Apoe, Cd38, Cd81, Itgb1, and Syk, among others. Markers of MemB_3 were S100a10, Vim, Lgals1, Ass1, Itgb7, Sec61b, Anxa2, and Stk38. Interestingly, Vimentin (Vim) is a cytoskeleton component important for the filament reorganization following BCR stimulation (Tsui et al., 2018). Although unique markers for MemB_4 were scarce, Fcer2a (encoding for CD23) and Icosl showed the highest expression. Other MemB_4 markers include Cd55, Il2rg, Ets1, Ltb, Lmo2, and Zfp36.

Altogether, our interpretation is that MemB_1 likely represents recent GC-derived B cells, as it shows the highest single-cell score for the MYC pathway (Figure 4—figure supplement 1C), MemB_2 corresponds to the AtMemBs described in chronic antigen exposure and aging and is analogous to DN2 cells in humans, MemB_3 corresponds to MemBs in an activated state, as it had the highest score for a signature of genes downregulated by Pten, a major regulator of B cell activation (Figure 4—figure supplement 1D), and MemB_4 corresponds to a memory compartment dependent on interacting with T cells for reactivation.

To further understand the relationship between MemB subclusters, we investigated the clones present in the four MemB clusters. Remarkably, we found that in both autoimmune and immunized mice (Figure 5A), all MemB clusters share clones with one another. This suggests that, regardless of their repertoire origin and specificities, MemBs can acquire any of the different cluster characteristics and that circulation among them might exist. Moreover, as observed in ASCs, regardless of their transcriptomic and repertoire differences, subclusters can have similar V gene usage for IGHV 1–15, 1–42, 1–53, 1–64, 1–75, 3–6, and 9–3, among others (Figure 5B).

Figure 5
Download asset Open asset
Clonal relations among and between memory B cell (MemB) and antibody-secreting cell (ASC) subclusters.

(A) Clonal intersects among MemB subclusters in autoimmune (left) and immunized (right) chimeras. Connector ribbons’ width represents number of shared clones (see Supplementary file 1). (B) Clonal V gene usage in MemB subclusters for the autoimmune chimeras, per mouse. Only V genes used in two clones or more. (C) Clonal intersects between ASC and MemB subclusters in autoimmune (left) and immunized (right) chimeras. Connector ribbons’ width represents number of shared clones (see Supplementary file 1).

To evaluate the relationship between potential reservoir (memory) and active response (secretion), we looked for clonal intersection between ASCs and MemBs. We found that clones across all the MemB clusters can seed all the different ASC clusters, regardless of the autoimmune or immunized context (Figure 5C). Cluster MemB_4 showed higher clonal contribution in immunized chimeras than in autoimmune chimeras, and in both conditions, most clones were shared between MemB_3 and all the ASC subclusters.

Validation of the GC origin of the MemB compartment

We generated a new set of autoimmune chimeras (n = 5, Figure 6A), which were maintained on a tamoxifen diet for a period of 8 weeks before removing tamoxifen for 4 weeks before analysis. Among the MemBs (AID-EYFP+ GL7− CD138−), we found that MemB_2 (FCRL5+) and MemB_4 (CD23+) cells were in similar proportions in all mice (Figure 6B, Figure 6—figure supplement 1A).

Figure 6 with 1 supplement see all
Download asset Open asset
FCRL5+ and CD23+ fate-mapped memory B (MemB) cells have different preferential localization.

(A) Experimental design to validate presence of fate-mapped FCRL5+ and CD23+ MemBs. (B) Identification of discrete populations of FCRL5+ and CD23+ MemB cells with Aicda-CreERT2-EYFP reporter mice by flow cytometry (n = 5 chimeras). (C) Experimental design for using S1pr2-CreERT2-tdTomato fate mapping in autoimmune chimeras. (D) Identification of discrete populations of FCRL5+ and CD23+ MemB cells with S1pr2-CreERT2-tdTomato fate mapping in autoimmune chimeras by flow cytometry (n = 3 chimeras). (E) Splenic localization of FCRL5+ S1PR2(Tomato)+ MemB cells by confocal microscopy. Overview of a spleen from autoimmune chimeras (Figure 3C), left, and selected area, center, delineates a bridging channel for a close-up examination of FCRL5+ S1pr2tomato+, right. Arrow points to FCRL5+ MemBs. Scale bars represent 100 µm (left), 50 µm (center), and 20 µm (right). Arrow heads point to FCRL5+ S1PR2(Tomato)+ cells. (F) Experimental design for in vivo marginal zone labeling. S1pr2-CreERT2-tdTomato:564Igi chimeras intravenously (i.v.) injected with 5 µg of anti-CD45-APC for 5 min before organ extraction. (G) Flow cytometry gate strategy to evaluate i.v. CD45 labeling between MemB FCRL5+ and MemB CD23+ cells. Statistical values correspond to one-tailed unpaired Student t-test (G).

Recent studies show that Aicda (AID) expression can precede the commitment and formation of GC B cells at least in immunized mice (Roco et al., 2019). This could suggest that the observations with AID reporter mice could also include non-GC-derived MemBs. Indeed, formation of extrafollicular MemBs cannot be ruled out from the AID-based model (Lee et al., 2011; Toyama et al., 2002). As an alternative solution for GC fate mapping, we used S1pr2, which is an established fate marker for GC-derived B cells (Shinnakasu et al., 2016). To confirm the observations made with our single-cell dataset and Aicda-CreERT2-EYFP reporter chimeras, a new set of chimeras using the S1pr2-CreERT2-tdTomato reporter system in combination with 564Igi BM were constructed (n = 3, Figure 6C). The donor S1pr2 reporter mice were also crossed with Prdm1-EYFP reporter mice to distinguish GC-derived ASCs.

As predicted, MemB cells (S1pr2tdTomato+ Prdm1EYFP−) were observed among the GL7− CD138− population based on flow cytometry. Further analysis of S1pr2tdTomato+ B cells identified CD23+, FCRL5+ MemBs, corroborating our previous observations (Figure 6D, Figure 6—figure supplement 1B).

To gain insight into the spatial distribution of Memory S1pr2tdTomato+ B cells, cryosections of splenic tissue were characterized by fluorescent scanning confocal microscopy. As expected, S1pr2tdTomato+ cells localized mostly to GCs in the B cell follicles. Interestingly, a substantial number of GC-derived FCRL5+ MemBs localized close to the MZ and particularly near the bridging channels (Figure 6E), delineated by CD169+ macrophages.

To confirm the MemB FCRL5+ localization, we injected anti-CD45 antibodies intravenously (i.v.) into autoimmune chimeras (S1pr2-CreERT2-tdTomato:564Igi) 5 min before sacrifice (Figure 6F). This labels cells exposed to blood circulation in the spleen, as those residing in the MZ would be (Cinamon et al., 2008; Song et al., 2022). We observed a significant difference in the i.v. labeling for MemB FCRL5+ in contrast to MemB CD23+ cells, confirming a preferential MZ localization (Figure 6G, Figure 6—figure supplement 1C). We confirmed the efficiency of MZ preferential labeling using a traditional gating strategy for MZ and follicular B cells in the same mice (Figure 6—figure supplement 1D). This privileged location would allow MemB FCRL5+ cells to quickly reactivate upon exposure to foreign antigen but also perpetuate a detrimental response when reacting to self-antigens.

Altogether, we found an underlying complexity of MemBs that, in autoimmune and immunized mice, can be further subdivided into groups that share clonality with ASCs and have transcriptomic and VDJ repertoire signatures suggesting distinct roles in the immune response. Indeed, Memb_2 cells are evidence of memory contribution to the MZ and that the GC-derived MemB compartment can contain DN2-like cells and have the capacity to contribute to pathology through the preservation of harmful self-reactive specificities.

Discussion

The study of antibody-driven autoimmunity often relies on the use of mice with knock-in BCRs specific for a self-antigen. This approach has been important in establishing many of the current concepts of B cell autoimmunity such as receptor editing, follicular exclusion, and clonal deletion. However, introducing a BCR with a certain pre-determined affinity can potentially bias the observations of the rules governing the activation of those cells. Moreover, conclusions drawn from monogenic BCR mice are unlikely to accurately predict the response to self-antigen when a full repertoire of cells are competing for self-antigen and T cell help. Self-reactivity exists as a spectrum of affinities controlled by sensitive mechanisms that might have different regulatory properties (Smith et al., 2019; Tan et al., 2019).

The mixed BM chimera model used in our study allows for the pre-defined self-reactive cells to kick-start autoimmunity. In this environment, naive self-reactive WT B cells that normally undergo negative selection can spontaneously become activated and enter self-reactive GCs. Although this approach still depends on a BCR knock-in transgene, it provides a model to study the development of WT B cells as they mature in an autoimmune environment (Degn et al., 2017).

We found that self-reactive WT B cells mature and reach pausi-clonality through GC selection and eventually enter all the major mature B cell compartments, similar to that of the immunized controls. Both differential and overlapping clonal usage of V genes were observed in the autoimmune and immunized chimeras. For example, among the most used V genes in both conditions were IGHV 1–53. This VH family member appears in many scenarios such as MZ B cells reacting to HIV gp120 immunization (Pujanauski et al., 2013) and self-reactive clones in FcgRIIB-deficient mice (Tiller et al., 2010; van der Poel et al., 2019). Intriguingly, 1–53 is among the most commonly used genes in the mouse response to the hapten group NP (Xue et al., 2019), most likely due to its similarity to IGHV 1–72. Thus, IGHV 1–53 can be considered part of the collection of baseline repertoire that gives rise to ‘polyreactive’ antibodies. Although polyreactive antibodies may be detrimental in autoimmunity, they are likely retained in the repertoire for use when a quick response to a foreign antigen is needed. Notably, mouse IGHV 1–53 is the mouse ortholog to human IGHV 1–69 (Ighv1-53, 2023). IGHV 1–69 is commonly observed in broadly neutralizing antibodies in the anti-viral response, but together with IGHV 4–34, it is also a feature of antibodies in autoimmune disease (Bashford-Rogers et al., 2018; Watson and Breden, 2012). Additionally, the ortholog to IGHV 4–34 in mice is IGHV 3–6 (Ighv3-6, 2023), which was the most clonally used gene in our autoimmune chimeras. This suggests that self-reactivity can be driven by predominant V genes that might initially be polyreactive but are further tailored in GCs. Altogether, these data suggest that there might be different ways to generate self-reactive pathogenic clones, starting with common V genes but leading to the usage of more specific ones. Distinguishing the contributions of originally polyreactive vs non-polyreactive clones will allow for better understanding of the mechanisms that lead to epitope spreading.

Many characteristics have been attributed to ASCs: short- vs long-lived, plasmablasts vs plasma cells, and EF- vs GC-derived. We observed two terminal clusters of ASC development with clear distinctions in the number of Ig transcripts, metabolic requirements and ER-stress-related gene expression. ASC_Late_1 had lower Ig counts and score signature for oxidative phosphorylation, together suggesting a reduced rate of antibody secretion (Lam et al., 2018a). Intriguingly, ASC_Late_1 had more replacement mutations than ASC_Late_2 and a more diversified usage of isotype. Given that MemBs have accumulated mutations over-time in the GC and that our sequencing dataset was processed ~5 weeks after tamoxifen treatment while the lifespan of short-lived ASCs is estimated at ~2 weeks, reactivated MemBs are the likely origin of most ASC_Late_1 cells. Although pseudotime analysis leads us to classify ASC_Late_2 as one of the two terminal clusters, the immediately preceding cluster ASC_Mid_2 also possesses characteristics of mature ASCs as well, including prominent expression of Cxcr4 and Slpi, which is a marker for BM and long-lived ASCs (Lam and Bhattacharya, 2018b), suggesting it could represent another terminal state.

We identified four MemB subclusters. MemB_1 has characteristics of GC B cells and likely represents recently egressed cells. MemB_2 has similar markers as atMemB and DN2 cells. MemB_3 shares similar characteristics to recently activated B cells. MemB_4 shows patterns of interaction with T cells (and resembles DN4 in humans). We suspect that the different properties of MemB_2 and MemB_4 exist to allow the B cell compartment to respond adequately in a diversity of contexts. Indeed, we observed distribution of MemB_2 GC-derived cells in the MZ and in bridging channels – convenient locations for immediate antigen surveillance and response. MemB_4 might represent T-dependent bona fide long-lived MemBs, maintaining a record of previous exposure and requiring involvement of T cells for retriggering.

Finally, we found that, to varying degrees, all subclusters of ASCs and MemBs shared members of the same clones. MemB_3 showed striking predominance in clonal overlap with ASCs of all subclusters, indicating this cluster is an activated version of all the other MemB subpopulations. The clonal overlap between all these subclusters was surprising, although not completely unexpected after finding overlapping clonality among ASCs and MemBs when analyzed separately. This overlap likely provides a wider safety-net, an insurance of long-lived memory and a readiness to reactivate to deal with prospective pathogens. On the other hand, this implicates that the efforts to target any one cellular compartment for the sake of therapeutically ablating autoimmunity may overlook the presence of shared self-antigen specificity across different cellular niches.

Overall, using a WT BCR repertoire mouse model that maps spontaneous GC-derived cells, we observed a diversity of subgroups, reflecting the inner complexities of the ASC and MemB compartments, with specific transcriptomic and repertoire characteristics, but with underlying redundant clonality.

Materials and methods

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Strain, strain background (M. musculus)AicdaCreERT2C-A Reynaud, J-C Weill. (Institut Necker)
Strain, strain background (M. musculus)S1pr2CreERT2 BAC-transgenicT. Kurosaki (RIKEN-Yokohama)
Strain, strain background (M. musculus)B6.Cg-Tg(Prdm1-EYFP)1Mnz/JJackson Laboratories
Strain, strain background (M. musculus)C57BL/6JJackson Laboratories
Strain, strain background (M. musculus)B6.SJLJackson Laboratories
Strain, strain background (M. musculus)564IgiTheresa Imanishi-Kari (Tufts University)
Strain, strain background (M. musculus)564Igi.1This manuscript
Antibodyanti-CD45.2-APC (104) (mouse monoclonal)Biolegend1098141:300
Antibodyanti-CD23-BV510 (B3B4) (rat monoclonal)Biolegend1016231:300
Antibodyanti-B220-PacBlue (RA3-6B2) (rat monoclonal)Biolegend1032271:300
Antibodyanti-B220-PerCP/Cy5.5 (RA3-6B2) (rat monoclonal)Biolegend1032341:300
Antibodyanti-GL7-PacBlue (GL7)(rat monoclonal)Biolegend1446141:300
Antibodyanti-GL7-PerCP/Cy5.5 (GL7) (rat monoclonal)Biolegend1446101:300
Antibodyanti-CD138-Biotin (281-2) (rat monoclonal)Biolegend1425121:300
Antibodyanti-CD138-PE (281-2)(rat monoclonal)Biolegend1425041:300
Antibodyanti-CD45.1-APC (A20)(mouse monoclonal)Biolegend1107141:300
Peptide, recombinant proteinStreptavidinBiolegend405206
Antibodyanti-FCRL5-Alexa488 (sheep polyclonal)biotechneFAB6757G1:10
Antibodyanti-FCRL5-APC (sheep polyclonal)biotechneFAB6757A1:10
Commercial assay, kitFixable Viability Dye eFluor 780Thermo Fisher65-0865-141:1000
Commercial assay, kitPan B Cell Isolation Kit II, mouseMiltenyi130-104-443
Peptide, recombinant proteinNP-OVABiosearchN-5051-10
Chemical compound, drugTamoxifenSigma-AldrichT5648-5G
Chemical compound, drugImject AlumThermo Fisher Scientific77161
Software, algorithmFlowJoFlowJo LLC
Software, algorithmPrismGraphPad
Software, algorithmRR Foundation4.1.2
Software, algorithmCellranger10× Genomics5.0.1
Software, algorithmGC treehttps://github.com/matsengrp/gctree (DeWitt et al., 2018)
Commercial assay, kitHuProt v4.0 arrayCDI Labs
Commercial assay, kitSingle Cell Immune Profiling10× Genomics
Software, algorithmImmcantationhttps://immcantation.readthedocs.io/en/stable/

Study design

Request a detailed protocol

The purpose of this study was to characterize the post-GC populations of antibody-secreting and MemBs in the context of autoimmunity. We used GC fate-mapping and single-cell transcriptomics coupled to BCR repertoire analysis. The BM chimeric mouse model we used allows to track WT B cells as they develop and exit from spontaneous GCs. We contrasted the autoimmune chimeras with NP-OVA immunized chimeras.

Mice

C57BL/6J and B6.SJL (CD45.1), B6.Cg-Tg(Prdm1-EYFP)1Mnz/J (Blimp-EYFP) were obtained from Jackson Laboratories. Aicda-CreERT2 flox-stop-flox- EYFP mice (Dogan et al., 2009) were from Claude-Agnes Reynaud and Jean-Claude Weill (Institut Necker). S1pr2-CreERT2 BAC-transgenic mice (Shinnakasu et al., 2016) were generated and generously provided by T. Kurosaki (RIKEN-Yokohama). 564Igi mice (Berland et al., 2006) were originally provided by Theresa Imanishi-Kari (Tufts University) and were maintained in-house. 564.1 mice were generated by crossing 564Igi and B6.SJL (CD45.1) mice. 564Igi and 564.1Igi mice were genotyped by ddPCR using the primers: 564Igi_H_Fwd-cacagattcttagtttttcaa, 564Igi_H_Rev-tggagctatatcatcctcttt, 564Igi_K_Fwd-ccagtgcagattttcagcttc, 564Igi_K_Rev-cagcttggtcccagcaccgaa, mRPP30_Fwd-tgaccctatcagaggactgc, and mRPP30_Rev-ctctgcaatttgtggacacg. All mice were bred and maintained in the AAALAC-accredited facility at Harvard Medical School. Mice were specific pathogen-free and maintained under a 12-hr light/dark cycle with standard chow diet. Both male and female mice were used. All animal experiments were conducted in accordance with the guidelines of the Laboratory Animal Center of National Institutes of Health. The Institutional Animal Care and Use Committee of Harvard Medical School approved all animal protocols (protocol number IS111).

Immunization and antibody injection

Request a detailed protocol

For single-cell sequencing, non-autoimmune chimeric mice were i.p. immunized with 100 μg of 4-hydroxy-3-nitrophenylacetyl hapten conjugated to ovalbumin (NP-OVA, Biosearch) in 50 μl Hanks' Balanced Salt Solution (HBSS) precipitated in 50 μl of Imject Alum (Thermo Scientific) 6–8 weeks after irradiation and BM reconstitution. Four weeks after immunization, mice received an intraperitoneal booster immunization of 100  μg of NP-OVA in 100 μl HBSS. Mice were sacrificed 2 weeks after boost.

For in vivo labeling of MZ proximal cells, 5 min prior to euthanasia each mouse was injected retro-orbitally with 5 µg of anti-CD45.2-APC (Biolegend) diluted in 200 µl of phosphate-buffered saline (PBS).

Tissue processing

Request a detailed protocol

Mice were euthanized by cervical dislocation under isoflurane induced anesthesia. Spleens were extracted, dissected, and processed for immunofluorescence microscopy or flow cytometry analysis and cell sorting.

BM chimeras

Request a detailed protocol

BM chimeras were prepared using marrow from 564Igi mice and BCR WT donor as previously described (Degn et al., 2017). Mice were lethally irradiated at 1100 rad and kept on antibiotics (sulfamethoxazole/trimethoprim) through drinking water for 7 days after irradiation. Femurs and tibia from donor mice were cleaned from muscle tissue and subsequently rinsed with cell transfer buffer (HBSS supplemented with 10 mM N-2-hydroxyethylpiperazine-N-2-ethane sulfonic acid (HEPES), 1 mM Ethylenediaminetetraacetic acid (EDTA), and 2% heat inactivated fetal bovine serum). Marrow was extracted from bones by crushing them using mortar and pestle and the detached cells were resuspended in cell transfer buffer and passed through a 70-mM sterile filter. Cells were counted by using an erythrocyte lysed aliquot. All autoimmune chimeras were prepared at 2:1 ratio for 564Igi:WT-reporter, with WT-reporter and irradiated hosts genotype as specified for each experiment. BM recipients received 15–20 × 106 cells i.v. in 100 ml cell transfer buffer by retroorbital i.v. injection approximately 8 hr post irradiation.

Fate-mapping tamoxifen induction

Request a detailed protocol

Mice were exposed to tamoxifen in two different ways, as specified in each experimental design figure. Mice were gavaged with 10 mg of tamoxifen (Sigma) dissolved in Corn Oil at 50 mg/ml twice, at days 4 and 7 post primary immunization in the case of immunized chimeras and at the same time for autoimmune chimeras. For validation experiments, BM chimeric mice were maintained in a tamoxifen enriched diet for the specified timeframes (Envigo) (Song et al., 2022).

Flow cytometry

Request a detailed protocol

Spleen fragments were harvested into ice-cold Magnetic-activated cell sorting (MACS) buffer (PBS pH 7.2, 0.5% bovine serum albumin (BSA), and 2 mM EDTA) and mechanically dissociated using pestles in 1.5 ml Eppendorf tubes. Samples spun down at 300 × g for 5 min and resuspended in RBC lysis buffer (155 mM NH4Cl, 12 mM NaHCO3, 0.1 mM EDTA). Samples were washed with MACS buffer and spun down at 300 × g for 5 min. Finally, samples were resuspended in MACS buffer and filtered through 70 µm cell strainers (Corning). Samples were added to 96-well round-bottom plates, spun down and resuspended in 50 µl staining mix and stained for 30 min on ice. Cells were washed with 150 µl of MACS buffer twice by spinning down at 300 × g for 5 min. Finally, cells were resuspended in 200 µl of MACS buffer and transferred to 5 ml fluorescence activated cell sorting (FACS) tubes. The following antibodies and proteins were used: anti-CD23-BV510 (B3B4), anti-B220-PacBlue (RA3-6B2), anti-B220-PerCP/Cy5.5 (RA3-6B2), anti-GL7-PacBlue (GL7), anti-GL7-PerCP/Cy5.5 (GL7), anti-CD138-Biotin (281-2), anti-CD138-PE (281-2), and PE/Cy7-Streptavidin from Biolegend, and anti-FCRL5-Alexa488 and anti-FCRL5-APC (Polyclonal) from biotechne. Viability was determined with Fixable live/dead stain Efluor780 from Thermo Fisher Scientific. A standard 3 lasers configuration (405, 488, and 633 nm) FASCSCanto2, with 8-color and 10 parameter analytical capabilities was used for acquisition. Data were analyzed with FlowJo 10.

Cell sorting for single-cell sequencing

Request a detailed protocol

Spleens from autoimmune and immunized chimeras were processed following viability recommendations for droplet-based Chromium single-cell RNA-seq gene expression (10×). The full organs were dissociated in MACS buffer (PBS 1×, 0.5% BSA, 2 nM EDTA) using syringe plungers and 70 µm cell strainers, spun down at 1000 rpm for 5 min, resuspended in RBC lysis buffer and incubated on ice for 5 min. Samples were washed with MACS buffer and spun down 1000 rpm for min. Cells were enriched by negative selection with a Pan II B cell enrichment kit (Miltenyi) according to the provider specifications. During enrichment incubation cells were also stained for flow sorting. The antibodies used for staining were anti-GL7-PacBlue, ant-CD138-PE, anti-B220-PerCP/Cy5.5, anti-CD38-PECy7, and anti-CD45.1-APC from Biolegend. eFluor780 from Thermo Fisher was used for viability. After enrichment and antibody staining cells were resuspended in resuspension buffer (PBS 1×, 0.04% BSA). Reporter cells were sorted based on EYFP expression (Efluor 780- EYFP+ CD45.1+) at two-way purity sort with a FACSARIA II Special Order system (BD Biosciences) with 355, 405, 488, 640, and 592  nm lasers. Sorted cells were spun down and resuspended in resuspension buffer prior to single-cell encapsulation.

Droplet-based single-cell sequencing and data processing

Request a detailed protocol

The sorted cells were encapsulated with barcoded hydrogels using the Chromium system for Single Cell Immune Profiling, that allows coupling of transcriptomic and VDJ information per cell. cDNA libraries were prepared according to the manufacturer’s recommendations. Library quality control and sequencing (NextSeq 500, Illumina) were performed by the HMS Biopolymers Facility. Cellranger (5.0.1) was used to generate the count matrices for gene expression and the VDJ contigs per cell using the multi function for each mouse. Reads were aligned to a custom mm10 reference genome incorporating the transcript sequences for EYFP and Cre Recombinase. R (4.1.2) was used for further processing the Cellranger gene expression count matrix output using the OSCA workflow (https://bioconductor.org/books/release/OSCA/) as template and using the Single Cell Experiment (SCE) format. Scater and Scran were used for data QC. Cells were subset for less than 5% mitochondrial and 40% ribosomal content. Ig genes were excluded from clustering and posterior expression analysis. Correction was done with fastMNN function from batchelor. Clustering was performed using the Leiden algorithm. Markers for each cluster were identified using the scoreMarkers function from Scran. Pre-defined genesets were used for Oxidative Phosphorylation (GO:0006119), Myc upregulated (Wang et al., 2020; Chen et al., 2021) and PTEN_DN.V1_DN (GSEA) were used for signature analysis with AUCell. Data visualization was done with dittoSeq.

BCR repertoire analysis

Request a detailed protocol

The VDJ output from Cellranger multifunction was further processed following the Immcantation (https://immcantation.readthedocs.io/en/stable/) recommendations for 10× derived single-cell data. Genes were assigned using IgBlast and the IMGT reference sequence database. Clones were set using the DefineClones function from Change-O, with a 0.1691791 threshold defined using the findtThreshold function from Shazam. countClones (Alakazam) and observedMutations (Shazam) functions were used for quantification.

Phylogenetic trees were generated with GCTree (DeWitt et al., 2018) using all VDJ sequences from a given clone and rooted in the germline obtained with CreateGermlines function from Change-O.

Antibody production and protein array profile

Request a detailed protocol

VDJ sequences were synthesized (IDT) for sequences from mice M13 and M21 with barcodes: GGGACCTGTAGCTGCC_B06M13, GATGAGGCATCGGGTC_B10M21, TGTTCCGCAATGGACG_B10M21, CGTGTCTCAAACCCAT_B06M13, GCGCAACCAATCTACG_B06M13, and AAATGCCGTACACCGC_B06M13. Geneblocks were cloned into a modified pVRC8400 vector between a tissue plasminogen activation (TPA) signal sequence and the constant domains of the mouse IgG1 CH1-CH3 and CL (Kuraoka et al., 2016). Monoclonal IgG1,k antibodies were produced in suspension by transient transfection of 293F cells, using polyethylenimine (Polysciences). The supernatant was harvested 5 days after transfection. Antibody was recovered from culture supernatant after centrifugation at 4200 rpm for 20 min and clearing with 45 µm filters. IgGs were purified using Protein G agarose (Thermo) and dialyzed in PBS. Purified IgGs were concentrated and stored at 4°C.

For target detection, the monoclonal antibodies were pooled in two different samples and processed for reactivity by CDI Labs with a HuProt v4.0 array, containing 21,000 human proteins. Monoclonal antibodies at 1 µg/ml were diluted in a final volume of 3 ml and were probed with the arrays for 2 hr at room temperature (RT). The arrays were washed according to the company protocol and were probed with the secondary antibody (Alexa-647-goat-anti-mouse IgG gamma-specific) under conditions optimized by CDI Labs for signal detection with GenePix software. Data were processed with CDI’s proprietary data analysis software (Z-score analysis).

Immunofluorescence and confocal microscopy

Request a detailed protocol

Following 4% Paraformaldehyde (PFA) fixation on ice for 3 hr, spleens were embedded in OCT (Fisher Healthcare), frozen in dry ice and stored at −80°C. Spleens were cut into 10 µm sections, blocked for 1 hr at RT with Blocking buffer (PBS, 0.01% Tween20, 2% BSA, and 5% Fetal Bovine Serum (FBS)) and stained O.N. at 4°C with antibodies diluted in Blocking buffer. Anti-CD169-BV510 (Biolegend) and anti-FCRL5-APC (biotechne) were used for staining. Slides were washed three times in PBS, 0.01% Tween20 and mounted with FluoroGel (Electron Microscopy Sciences) prior to image acquisition. Images were acquired with an OLYMPUS FV3000R resonant scanning confocal microscope equipped with four laser lines (405, 488, 514, and 633 nm), hybrid galvo and fast resonant scanning capabilities, ultra-sensitive GaAsP detectors with full spectral imaging and motorized XYZ stage for tiling. The images were processed in Fiji.

Statistical analysis

Request a detailed protocol

Two-tailed Mann–Whitney, one-way analysis of variance with Tukey correction and one-tailed unpaired Student t-tests were performed with Prism 9 (GraphPad).

Data availability

The sequencing data presented in this study have been submitted to the Gene Expression Omnibus under accession number GSE203132. Code is available on GitHub (copy archived at Castrillon, 2023).

The following data sets were generated
    1. Castrillon C
    2. Carroll M
    (2023) NCBI Gene Expression Omnibus
    ID GSE203132. Transcriptomic diversity and overlapping clonality across subsets of antibody-secreting and memory B cells from spontaneous germinal centers.
    https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE203132

References

Decision letter

  1. Tomohiro Kurosaki
    Reviewing Editor; Osaka University, Japan
  2. Betty Diamond
    Senior Editor; The Feinstein Institute for Medical Research, United States

Our editorial process produces two outputs: (i) public reviews designed to be posted alongside the preprint for the benefit of readers; (ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "Complex subsets but redundant clonality after B cells egress from spontaneous germinal centers" for consideration by eLife. Your article has been reviewed by 2 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Betty Diamond as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

Both of the reviewers recognize the potential interest of this manuscript but have raised several concerns.

1) The authors should perform additional experiments to validate whether the memory B cell subsets identified using Fcrl5 and CD23 are spatially distinct. These experiments should include quantification of the imaging data and a revised flow cytometry gating strategy to identify memory B cell subsets. The imaging data will likely require an improvement in the quality of the staining to better identify the Fcrl5+ cells.

2) The authors should add additional discussion regarding how their memory B cell subsets related to subsets that have been previously identified in mice using markers such as CD80 and PDl2.

3) One of the immunization models used in this study relies on a secondary dose of NP-OVA to activate previously labeled GC-derived B cells. This acute system differs from the chronic B cell activation observed in the autoimmune disease model, in which labeled B cells can be constantly selected into continually ongoing and spontaneous GC and differentiate into antibody-secreting cells. The authors should explain in some lines the differences between these two models and why they decide to compare an acute NP-immunization model with a chronic autoimmune disease model?

4) The model of tamoxifen administration used in figures 1 to 5 (two doses of tamoxifen on days 4 and 7 after initial immunization) and then in figure 6 (tamoxifen diet for 12 weeks) are different. This might result in a different labeling pattern for the cells expressing AID or S1pr2-Cre. Have the authors made sure that both tamoxifen administration schedules are comparable in terms of cell labeling output?

5) In Figure 2, the authors identified two terminally differentiated clusters, ASC_Late_1 and ASC_Late_2, which display different transcriptomics. Using differentially expressed markers, can the authors identify these two ASC populations by flow cytometry as performed in Figure 6 for memory B cells?

6) In figure 3A, ASC_Late_1 has a very diverse isotype usage in the auto-immune mice but not in the immunized ones. On the contrary, ASC_Late_2 in both conditions is overwhelmingly dominated by IgM. Are authors convinced that both ASC populations arise from the autoimmune disease? Are the antibodies generated by both ASC subsets specific for auto-antigens?

7) In Figure 3A, isotype usage for each ASC cluster and in both conditions is shown. According to previous pseudo-time analysis, ASC_Late_2, which mainly displays an IgM BCR, is originated from ASC_Early_1 to ASC_Early_2 to ASC_mild_1 to ASC_mild_2 according to the pseudo-time analysis (figure 2B). However, all these progenitors' clusters display a much lower IgM expression, especially in the auto-immune model, indicating that they have already class-switched and wouldn't be able to give rise to IgM. How do the authors reconcile these two findings?

8) In lines 275 to 281, the authors describe and define the different Memory B cell clusters. According to these lines, MemB_1 is defined as recent GC-derived and MemB_3 corresponds to MemB as activated cells. However, in Figure 6, the authors focused on MemB_2 defined as Fcrl5+ and MemB_4 as CD23+, which are defined as atypical Bmem and Bmem depending on T cell for their activation respectively. Why did the authors choose to focus only on these two clusters? Did the authors confirm that MemB_1 and MemB_3 have a GC origin?

Reviewer #1 (Recommendations for the authors):

Specific points

1) The scheme in Figure 1A does not include the exact time points in which mice received tamoxifen and are then sacrificed. Could the authors add this information to the figure for clarity?

2) Figure 1D is missing statistical analysis to support the claims of the authors in lines 87-88-113.

3) Figure 1E displays the phylogenetic tree of the most expanded clone in one of the autoimmune chimeras. The authors claim that expanded clones can be found in all 4 clusters, but looking closely at this figure the MemB cluster is scarcely represented. Previous work (Viant et al. 2020) showed that the memory B cell compartment is largely populated by single cell clones rather than expanded clones. Can the authors quantify the clonal expansion of memory B cells in their model (how many single-cell, 2-cell, 3-cell clones?).

4) In figure 1I, the authors show that two monoclonal antibody pools originating from autoimmune chimeras could bind to some auto-antigens from a human recombinant protein array. Why did the authors check the binding to human autoantigens and not mouse autoantigens? Could they be missing some targets that do not cross-react between mouse and humans?

5) In Figure 2, the authors identified two terminally differentiated clusters, ASC_Late_1 and ASC_Late_2, which display different transcriptomics. Using differentially expressed markers, can the authors identify these two ASC populations by flow cytometry as performed in Figure 6 for memory B cells?

6) In figure 3A, ASC_Late_1 has a very diverse isotype usage in the auto-immune mice but not in the immunized ones. On the contrary, ASC_Late_2 in both conditions is overwhelmingly dominated by IgM. Are authors convinced that both ASC populations arise from the autoimmune disease? Are the antibodies generated by both ASC subsets specific for auto-antigens?

7) In Figure 3A, isotype usage for each ASC cluster and in both conditions is shown. According to previous pseudo time analysis, ASC_Late_2, which mainly displays an IgM BCR, is originated from ASC_Early_1 to ASC_Early_2 to ASC_mild_1 to ASC_mild_2 according to the pseudo-time analysis (figure 2B). However, all these progenitors' clusters display a much lower IgM expression, especially in the auto-immune model, indicating that they have already class-switched and wouldn't be able to give rise to IgM. How do the authors reconcile these two findings?

8) Figure 3C, it would be very informative to show the frequency of clones that are clonally expanded within each ASC cluster, especially in ASC_Late_1 and ASC_Late_2.

9) In Figure 4B, please include appropriate statistical analysis to support the claims in lines 240-241.

10) In figure 5, please show the frequency of clones clonally expanded within each MemB sub-cluster.

11) In Figure 6, the authors elegantly used a S1pr2-creERT2-TdTomato-Prdm1-EYFP transgenic mouse to confirm that the B cell memory is indeed GC-derived. It would be very insightful to also confirm the GC origin or not of the clusters ASC_Late_1 and ASC_Late_2 using this model. As ASC_Late_2 is mainly IgM, it could be that this population is not arising from GC reactions.

12) In lines 275 to 281, the authors describe and define the different Memory B cell clusters. According to these lines, MemB_1 is defined as recent GC-derived and MemB_3 corresponds to MemB as activated cells. However, in Figure 6, the authors focused on MemB_2 defined as Fcrl5+ and MemB_4 as CD23+, which are defined as atypical Bmem and Bmem depending on T cell for their activation respectively. Why did the authors choose to focus only on these two clusters? Did the authors confirm that MemB_1 and MemB_3 have a GC origin?

13) In Figure 6E, can authors include GL7 staining to exclude that labeled S1pr2+ and Fcrl5+ cells don't belong to the GC compartment. Also, quantification of the confocal images is required.

https://doi.org/10.7554/eLife.81012.sa1

Author response

Essential revisions:

Both of the reviewers recognize the potential interest of this manuscript but have raised several concerns.

1) The authors should perform additional experiments to validate whether the memory B cell subsets identified using Fcrl5 and CD23 are spatially distinct. These experiments should include quantification of the imaging data and a revised flow cytometry gating strategy to identify memory B cell subsets. The imaging data will likely require an improvement in the quality of the staining to better identify the Fcrl5+ cells.

We agree with the reviewers that a thorough confirmation of spatial localization through quantifiable microscopy would be ideal for clarifying the distribution of the different Memory B cell subsets. Unfortunately, the sparsity of these cells, in addition to the suboptimal performance of the anti-FCRL5 antibody in microscopy limited our capacity to do so. However, it was the observation of fate mapped cells in the Marginal zone area that led us to perform the short-term anti-CD45 i.v. injection labelling as a high throughput method for validation. Rapid staining with i.v. injected anti-CD45 antibody before organ extraction is a well-documented method to contrast localization between Marginal and Follicular zone B cells through flow cytometry, based on the fluorescent labelling of CD45(Cinamon et al., 2008). Indeed, Song et al. (2022) recently used this approach in support to the localization of a B cell population akin to our MemB_2 with a Tbet-reporter mouse after infection and found similar results with flow cytometry and imaging.

To support this approach, we are including flow cytometry data from the same samples using a traditional CD21 vs CD23 gating strategy to distinguish Marginal Zone and Follicular B cells, demonstrating the capacity to distinguish cells in these compartments. Importantly, using this traditional gating strategy we observe S1PR2+ fate mapped cells in the Marginal zone compartment; the concept of memory in the Marginal Zone remains largely unexplored and we hope our publication will support this idea.

We are including now all the FMOs for Figure 6 and the confirmation of i.v. staining as supplementary Figure 6- supplement 1 in the revised manuscript.

2) The authors should add additional discussion regarding how their memory B cell subsets related to subsets that have been previously identified in mice using markers such as CD80 and PDl2.

We are aware of previous studies using CD80 and PDL2 for characterizing the Memory B cell compartment. Interestingly, we didn’t observe much expression of those genes in the clusters that we decided to focus on, MemB_2 and MemB_4, but did find them expressed for clusters MemB_1 and MemB_3, especially in the immunized mice (Author response image 1). These markers were first defined in the context of immunization and by specific staining with fluorophore labelled antigen and IgG1 positivity. We hypothesize that the distinct conditions of our model and the fact that we took the full Memory B cell compartment in consideration might account for the difference in expression pattern.

Author response image 1
Download asset Open asset
Normalized gene expression of Cd80 (left) and Pdcd1lg2(coding for PD-L2, right), split by MemB cluster and condition.

3) One of the immunization models used in this study relies on a secondary dose of NP-OVA to activate previously labeled GC-derived B cells. This acute system differs from the chronic B cell activation observed in the autoimmune disease model, in which labeled B cells can be constantly selected into continually ongoing and spontaneous GC and differentiate into antibody-secreting cells. The authors should explain in some lines the differences between these two models and why they decide to compare an acute NP-immunization model with a chronic autoimmune disease model?

Our focus going into these experiments and in the development of this manuscript was to explore the diversity of wild type B cells developing in the autoimmune environment and to understand their clonal intersects, much of which was still unknown or understudied in the field. It was of course important for us to “anchor” our observations into a more traditional scheme, so that our data could be of broader use. We decided to use a traditional immunization scheme with the widely used 4-Hydroxy-3-nitrophenylacetyl (NP) hapten coupled to Ovalbumin antigen as reference. We hope it can be appreciated that because we understand the major differences between these models, we refrain ourselves from major comparisons like differential expression analysis between conditions. Though we have explored those analysis we believed them to be outside the scope of this article. We propose that a more complete exploration of the system would likely require the usage of both acute and chronic foreign antigen presence as contrast. Indeed, recent analysis of chronic or severe viral infection are showing overlaps with self-reactive immune response characteristics (Woodruff et al., 2020, 2022).

4) The model of tamoxifen administration used in figures 1 to 5 (two doses of tamoxifen on days 4 and 7 after initial immunization) and then in figure 6 (tamoxifen diet for 12 weeks) are different. This might result in a different labeling pattern for the cells expressing AID or S1pr2-Cre. Have the authors made sure that both tamoxifen administration schedules are comparable in terms of cell labeling output?

We apologize for the confusion, as for the validation experiments demonstrated in Figure 6 in the manuscript, we used a similar strategy (diet) for tamoxifen induction in both Aicda- and S1pr2-Cre mice. We had a typo in the original manuscript, with the Aicda reporter validation done with tam diet for 8 weeks and left out of diet for 4 weeks, and all S1pr2 reporter experiments done with a tam diet of 12 weeks. This is corrected in line 247 from the revised manuscript and displayed in the figures as well. Our goal with the tamoxifen diet was to maximize the number of cells that we could fate map since as soon as possible after the chimeras are produced, in contrast to acute labeling by injection of gavage. We have been able to observe the MemB populations in all scenarios, tamoxifen diet and injection with both Cre reporters.

5) In Figure 2, the authors identified two terminally differentiated clusters, ASC_Late_1 and ASC_Late_2, which display different transcriptomics. Using differentially expressed markers, can the authors identify these two ASC populations by flow cytometry as performed in Figure 6 for memory B cells?

We agree that further validation of these populations is granted, unfortunately the markers we pursued didn’t allow complete cluster discrimination through flow cytometry. We hypothesize this happened because most of the cluster gene markers for ASCs didn’t appear as positive or negative but represented different degrees of expression.

We show here Itga4 (CD49d) as an example (Author response image 2), where we explored the staining for CD49d in our 564:S1pr2 chimeras, within the S1PR2+ compartment, by conventional markers for B cells (B220+ CD138-), Plasmablasts (B220+ CD138+) and Plasma cells (B220- CD138+), confirming we can identify CD49d lo and hi populations in the Plasma cells as suggested by the single cell data, but these populations were also found in the B220+CD138+ gate, traditionally considered a less mature phenotype, complicating the marker usage interpretation.

Author response image 2
Download asset Open asset
Flow cytometry profile for CD49d expression for a representative 564:S1pr2 chimera, showing the CD49d lo an CD49 hi populations within the B cell, Plasmablast and Plasma cell conventional gates.

6) In figure 3A, ASC_Late_1 has a very diverse isotype usage in the auto-immune mice but not in the immunized ones. On the contrary, ASC_Late_2 in both conditions is overwhelmingly dominated by IgM. Are authors convinced that both ASC populations arise from the autoimmune disease? Are the antibodies generated by both ASC subsets specific for auto-antigens?

We hypothesize that the discrepancy in isotype diversity between conditions (Immunized vs Autoimmune) more likely arises from the distinct antigenic exposure in the model. In the autoimmune mice there is likely recurrent exposure and ongoing response to a diversity of antigens, which leads to a larger engagement of cells into isotype switch, whereas the immunization strategy engages a more limited number of cells, most of which don’t seem to have switched yet, because of its acute and highly specific nature, and likely lesser degree of antigenicity and re-engagement of cells from the primary response.

7) In Figure 3A, isotype usage for each ASC cluster and in both conditions is shown. According to previous pseudo-time analysis, ASC_Late_2, which mainly displays an IgM BCR, is originated from ASC_Early_1 to ASC_Early_2 to ASC_mild_1 to ASC_mild_2 according to the pseudo-time analysis (figure 2B). However, all these progenitors' clusters display a much lower IgM expression, especially in the auto-immune model, indicating that they have already class-switched and wouldn't be able to give rise to IgM. How do the authors reconcile these two findings?

Our intention with the Pseudotime analysis was to use an unsupervised method to find likely terminal stages in ASC development and support our assignment, which was originally driven by already known markers. We would like to note that the interpretation of Pseudotime analysis can be tricky and should be taken with caution. For example, the ‘time’ transitions are unfortunately not as clear cut and a few Mid_3 cells will be represented in Pseudotime_2 leading to ASC_2, and even a few cells from Late_2 can show in Pseudotime_1. Nevertheless, a useful way to understand Pseudotime trajectories is to think about it as a transportation system where every cell cluster is a stop. It is possible then that IgM individuals follow along the system but get out mostly at the cluster ASC_Late_2 stop. Indeed, we observe an interesting distribution in isotype usage when splitting the cells into those that were considered only for Pseudotime_1 or Pseudotime_2, with an enrichment of IgM cells already in ASC_Mid_3 stop (Author response 3).

Author response image 3
Download asset Open asset
Isotype usage per ASC cluster, showed only for cells assigned to Pseudotime_1 (left) or Pseudotime_2 (right).

8) In lines 275 to 281, the authors describe and define the different Memory B cell clusters. According to these lines, MemB_1 is defined as recent GC-derived and MemB_3 corresponds to MemB as activated cells. However, in Figure 6, the authors focused on MemB_2 defined as Fcrl5+ and MemB_4 as CD23+, which are defined as atypical Bmem and Bmem depending on T cell for their activation respectively. Why did the authors choose to focus only on these two clusters? Did the authors confirm that MemB_1 and MemB_3 have a GC origin?

We decided to follow up on MemB_2 and MemB_4 because Fcrl5 and Cd23, two surface markers that characterize those clusters, are easily identifiable by flow cytometry and have traditional spatial distribution connotations in the context of B cell biology: Cd23 being a conventional marker for follicular B cells and Fcrl5 having been linked to the Marginal Zone compartment. It was interesting to us that such markers distinguished two populations of Memory B cells after leaving the GC, as distinct localization likely suggests distinct function. Moreover, FCRL5 allows us to us to link our interpretations to the work of other groups investigating the nature of “atypical” B cells in mice and human (DN2-like) and as a demonstration of that phenotype presence in our model. Importantly, as mentioned in the original manuscript, we believe this to be a confirmation of a DN2-like population within the GC-derived memory compartment. We expect this finding to be of high relevance to ongoing discussions on the nature of these cells within the autoimmune and B cell field at large as so far this phenotype has mostly been linked to recent activation from naïve B cells outside the GC.

We are of course interested in the other clusters as well but decided to constrain ourselves to two of them to maintain our research focused, as we learned that single cell rna-seq validation is on itself a daunting task to fit within a single publication.

Reviewer #1 (Recommendations for the authors):

Specific points

1) The scheme in Figure 1A does not include the exact time points in which mice received tamoxifen and are then sacrificed. Could the authors add this information to the figure for clarity?

Yes, we have now included that information.

2) Figure 1D is missing statistical analysis to support the claims of the authors in lines 87-88-113.

We originally decided not to include statistical support given the difficulties to do proper differential abundance in single cell rna-seq data, we have modified the text to reflect this. Though we have observed it being used in other articles, the approach to differential abundance has not yet matured as much as the differential expression in single cell rna-seq data.

3) Figure 1E displays the phylogenetic tree of the most expanded clone in one of the autoimmune chimeras. The authors claim that expanded clones can be found in all 4 clusters, but looking closely at this figure the MemB cluster is scarcely represented. Previous work (Viant et al. 2020) showed that the memory B cell compartment is largely populated by single cell clones rather than expanded clones. Can the authors quantify the clonal expansion of memory B cells in their model (how many single-cell, 2-cell, 3-cell clones?).

To clarify, our intention with that expression and figure was to show that clonal members can be present in all compartments, but this does not mean that these represent highly expanded clones in each compartment. We have now tried to make this evident in the text. Indeed, we do find many singletons when analyzing each compartment separately. We would like to point that negative results (e.g. absence or limited clonal members) have a restricted interpretation in single cell repertoire given the limited number of cells that one is capable to query so far, even when droplet-based methods allow to sequence more cells that plate sorted methods do; and because of the dominance of cells with high immunoglobulin transcript count (ASCs) in the data tend to dominate over MemB and GC cells when being part of the same sequencing library.

4) In figure 1I, the authors show that two monoclonal antibody pools originating from autoimmune chimeras could bind to some auto-antigens from a human recombinant protein array. Why did the authors check the binding to human autoantigens and not mouse autoantigens? Could they be missing some targets that do not cross-react between mouse and humans?

We agree with this observation and whole-heartedly wish we could have done otherwise. It is however not really a decision in the autoimmune B cell field to work with human proteins, as mouse reagent counterparts in that scale don’t exist. This is still a major limitation in the field.

5) In Figure 2, the authors identified two terminally differentiated clusters, ASC_Late_1 and ASC_Late_2, which display different transcriptomics. Using differentially expressed markers, can the authors identify these two ASC populations by flow cytometry as performed in Figure 6 for memory B cells?

Please see Essential revisions question 5.

6) In figure 3A, ASC_Late_1 has a very diverse isotype usage in the auto-immune mice but not in the immunized ones. On the contrary, ASC_Late_2 in both conditions is overwhelmingly dominated by IgM. Are authors convinced that both ASC populations arise from the autoimmune disease? Are the antibodies generated by both ASC subsets specific for auto-antigens?

Please see Essential revisions question 6.

7) In Figure 3A, isotype usage for each ASC cluster and in both conditions is shown. According to previous pseudo time analysis, ASC_Late_2, which mainly displays an IgM BCR, is originated from ASC_Early_1 to ASC_Early_2 to ASC_mild_1 to ASC_mild_2 according to the pseudo-time analysis (figure 2B). However, all these progenitors' clusters display a much lower IgM expression, especially in the auto-immune model, indicating that they have already class-switched and wouldn't be able to give rise to IgM. How do the authors reconcile these two findings?

Please see Essential revisions question 7.

8) Figure 3C, it would be very informative to show the frequency of clones that are clonally expanded within each ASC cluster, especially in ASC_Late_1 and ASC_Late_2.

We agree on the importance of this information but were not able to find a better way to visualize the intersects while adding properly sized numerical values. The data for these circos plots is of course provided in the Supplementary Tables.

9) In Figure 4B, please include appropriate statistical analysis to support the claims in lines 240-241.

Please see response to question 2 on this section.

10) In figure 5, please show the frequency of clones clonally expanded within each MemB sub-cluster.

Please see response to question 8 on this section.

11) In Figure 6, the authors elegantly used a S1pr2-creERT2-TdTomato-Prdm1-EYFP transgenic mouse to confirm that the B cell memory is indeed GC-derived. It would be very insightful to also confirm the GC origin or not of the clusters ASC_Late_1 and ASC_Late_2 using this model. As ASC_Late_2 is mainly IgM, it could be that this population is not arising from GC reactions.

We appreciate the reviewer interest on this aspect, as it represents a fundamental question to the ASC field we are actively pursuing. Indeed, one of our long-term goals is the distinction of GC and Extrafollicular contributions to the ASC compartment that we think the reviewer is alluding to. Unfortunately, we cannot yet answer that question with this data and system. For example, let’s say that a cell emerges from the GC as a fate mapped MemB, but the transition to an ASC happens directly (MemB-to-ASC) in the Extrafollicular space, without returning to the GC; that ASC is by localization non-GC derived but would be GC-derived by lineage, this distinction is clouded in the long-term fate mapping approach and requires more detailed kinetics and spatial studies. We are working on a modification on the experimental design to address this precisely.

12) In lines 275 to 281, the authors describe and define the different Memory B cell clusters. According to these lines, MemB_1 is defined as recent GC-derived and MemB_3 corresponds to MemB as activated cells. However, in Figure 6, the authors focused on MemB_2 defined as Fcrl5+ and MemB_4 as CD23+, which are defined as atypical Bmem and Bmem depending on T cell for their activation respectively. Why did the authors choose to focus only on these two clusters? Did the authors confirm that MemB_1 and MemB_3 have a GC origin?

Please see Essential revisions question 8.

13) In Figure 6E, can authors include GL7 staining to exclude that labeled S1pr2+ and Fcrl5+ cells don't belong to the GC compartment. Also, quantification of the confocal images is required.

We used CD169, delineating the Marginal Zone through macrophage staining, as a reference to localization. As shown in the images provided, this allows us to clearly distinguish the boundaries of the follicles and can be confident these don’t represent GC B cells (which can be seen in red clusters within the follicle).

https://doi.org/10.7554/eLife.81012.sa2

Article and author information

Author details

  1. Carlos Castrillon

    Program in Cellular and Molecular Medicine, Boston Children’s Hospital, Boston, United States
    Present address
    Department of Medicine, Division of Rheumatology, Emory University, Atlanta, United States
    Contribution
    Conceptualization, Data curation, Software, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing – original draft, Writing – review and editing
    For correspondence
    castrilloncarlos@gmail.com
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-2909-7371

  2. Lea Simoni

    Program in Cellular and Molecular Medicine, Boston Children’s Hospital, Boston, United States
    Contribution
    Conceptualization, Investigation, Methodology, Writing – review and editing
    Contributed equally with
    Theo van den Broek and Cees van der Poel
    Competing interests
    No competing interests declared

  3. Theo van den Broek

    Program in Cellular and Molecular Medicine, Boston Children’s Hospital, Boston, United States
    Contribution
    Conceptualization, Investigation, Methodology, Writing – review and editing
    Contributed equally with
    Lea Simoni and Cees van der Poel
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-2781-5731

  4. Cees van der Poel

    Program in Cellular and Molecular Medicine, Boston Children’s Hospital, Boston, United States
    Contribution
    Conceptualization, Investigation, Methodology, Writing – review and editing
    Contributed equally with
    Lea Simoni and Theo van den Broek
    Competing interests
    No competing interests declared

  5. Elliot H Akama-Garren

    1. Program in Cellular and Molecular Medicine, Boston Children’s Hospital, Boston, United States
    2. Harvard-MIT Health Sciences and Technology, Harvard Medical School, Boston, United States
    Contribution
    Investigation, Methodology, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-1690-2055

  6. Minghe Ma

    Program in Cellular and Molecular Medicine, Boston Children’s Hospital, Boston, United States
    Contribution
    Investigation, Methodology
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-4496-6240

  7. Michael C Carroll

    1. Program in Cellular and Molecular Medicine, Boston Children’s Hospital, Boston, United States
    2. Department of Pediatrics, Harvard Medical School, Boston, United States
    Contribution
    Conceptualization, Supervision, Funding acquisition, Writing – review and editing
    For correspondence
    michael.carroll@childrens.harvard.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-3213-4295

Funding

National Institutes of Health (R01AI130307)

  • Michael C Carroll

National Institutes of Health (R01AR074105)

  • Michael C Carroll

National Institutes of Health (T32GM007753)

  • Elliot H Akama-Garren

National Institutes of Health (T32AI007529)

  • Elliot H Akama-Garren

National Institutes of Health (F30AI160909)

  • Elliot H Akama-Garren

H2020 Marie Skłodowska-Curie Actions (BEAT (No. 796988))

  • Theo van den Broek

Koninklijke Nederlandse Akademie van Wetenschappen (Academy Ter Meulen Fund TMB/16/285)

  • Theo van den Broek

The funders had no role in study design, data collection, and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank J Moore of the Flow and Imaging Cytometry Resource at the BCH PCMM, the BCH Cell Function and Imaging Core for technical assistance, and all the members of the Carroll lab for their feedback and support. This work was funded by NIH grants R01AI130307 and R01AR074105 (MCC). EHA-G. was supported by NIH grants T32GM007753, T32AI007529, and F30AI160909. TvdB was supported by the H2020-MSCA-IF-GF project BEAT (No. 796988) and the Academy Ter Meulen Fund (TMB/16/285). Authors declare that they have no competing interests.

Ethics

All mice were bred and maintained in the AAALAC-accredited facility at Harvard Medical School. Mice were specific pathogen-free (SPF) and maintained under a 12-hr light/dark cycle with standard chow diet. Both male and female mice were used. All animal experiments were conducted in accordance with the guidelines of the Laboratory Animal Center of National Institutes of Health. The Institutional Animal Care and Use Committee of Harvard Medical School approved all animal protocols (protocol number IS111).

Senior Editor

  1. Betty Diamond, The Feinstein Institute for Medical Research, United States

Reviewing Editor

  1. Tomohiro Kurosaki, Osaka University, Japan

Version history

  1. Received: June 13, 2022
  2. Preprint posted: June 23, 2022 (view preprint)
  3. Accepted: June 20, 2023
  4. Accepted Manuscript published: June 21, 2023 (version 1)
  5. Version of Record published: July 17, 2023 (version 2)

Copyright

© 2023, Castrillon et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 432
    Page views
  • 72
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Carlos Castrillon
  2. Lea Simoni
  3. Theo van den Broek
  4. Cees van der Poel
  5. Elliot H Akama-Garren
  6. Minghe Ma
  7. Michael C Carroll
(2023)
Complex subsets but redundant clonality after B cells egress from spontaneous germinal centers
eLife 12:e81012.
https://doi.org/10.7554/eLife.81012

Categories and tags

Research organism

Further reading

    1. Immunology and Inflammation

    Selective loss of CD107a TIGIT+ memory HIV-1-specific CD8+ T cells in PLWH over a decade of ART

    Oscar Blanch-Lombarte, Dan Ouchi ... Julia G Prado
    Research Article

    The co-expression of inhibitory receptors (IRs) is a hallmark of CD8+ T-cell exhaustion (Tex) in people living with HIV-1 (PLWH). Understanding alterations of IRs expression in PLWH on long-term antiretroviral treatment (ART) remains elusive but is critical to overcoming CD8+ Tex and designing novel HIV-1 cure immunotherapies. To address this, we combine high-dimensional supervised and unsupervised analysis of IRs concomitant with functional markers across the CD8+ T-cell landscape on 24 PLWH over a decade on ART. We define irreversible alterations of IRs co-expression patterns in CD8+ T cells not mitigated by ART and identify negative associations between the frequency of TIGIT+ and TIGIT+ TIM-3+ and CD4+ T-cell levels. Moreover, changes in total, SEB-activated, and HIV-1-specific CD8+ T cells delineate a complex reshaping of memory and effector-like cellular clusters on ART. Indeed, we identify a selective reduction of HIV-1 specific-CD8+ T-cell memory-like clusters sharing TIGIT expression and low CD107a that can be recovered by mAb TIGIT blockade independently of IFNγ and IL-2. Collectively, these data characterize with unprecedented detail the patterns of IRs expression and functions across the CD8+ T-cell landscape and indicate the potential of TIGIT as a target for Tex precision immunotherapies in PLWH at all ART stages.

    1. Genetics and Genomics
    2. Immunology and Inflammation

    Epigenetic signature of human immune aging in the GESTALT study

    Roshni Roy, Pei-Lun Kuo ... Luigi Ferrucci
    Research Article Updated

    Age-associated DNA methylation in blood cells convey information on health status. However, the mechanisms that drive these changes in circulating cells and their relationships to gene regulation are unknown. We identified age-associated DNA methylation sites in six purified blood-borne immune cell types (naive B, naive CD4+ and CD8+ T cells, granulocytes, monocytes, and NK cells) collected from healthy individuals interspersed over a wide age range. Of the thousands of age-associated sites, only 350 sites were differentially methylated in the same direction in all cell types and validated in an independent longitudinal cohort. Genes close to age-associated hypomethylated sites were enriched for collagen biosynthesis and complement cascade pathways, while genes close to hypermethylated sites mapped to neuronal pathways. In silico analyses showed that in most cell types, the age-associated hypo- and hypermethylated sites were enriched for ARNT (HIF1β) and REST transcription factor (TF) motifs, respectively, which are both master regulators of hypoxia response. To conclude, despite spatial heterogeneity, there is a commonality in the putative regulatory role with respect to TF motifs and histone modifications at and around these sites. These features suggest that DNA methylation changes in healthy aging may be adaptive responses to fluctuations of oxygen availability.

    1. Immunology and Inflammation
    2. Neuroscience

    Molecular consequences of peripheral Influenza A infection on cell populations in the murine hypothalamus

    René Lemcke, Christine Egebjerg ... Birgitte R Kornum
    Research Article

    Infection with Influenza A virus (IAV) causes the well-known symptoms of the flu, including fever, loss of appetite, and excessive sleepiness. These responses, mediated by the brain, will normally disappear once the virus is cleared from the system, but a severe respiratory virus infection may cause long-lasting neurological disturbances. These include encephalitis lethargica and narcolepsy. The mechanisms behind such long lasting changes are unknown. The hypothalamus is a central regulator of the homeostatic response during a viral challenge. To gain insight into the neuronal and non-neuronal molecular changes during an IAV infection, we intranasally infected mice with an H1N1 virus and extracted the brain at different time points. Using single-nucleus RNA sequencing (snRNA-seq) of the hypothalamus, we identify transcriptional effects in all identified cell populations. The snRNA-seq data showed the most pronounced transcriptional response at 3 days past infection, with a strong downregulation of genes across all cell types. General immune processes were mainly impacted in microglia, the brain resident immune cells, where we found increased numbers of cells expressing pro-inflammatory gene networks. In addition, we found that most neuronal cell populations downregulated genes contributing to the energy homeostasis in mitochondria and protein translation in the cytosol, indicating potential reduced cellular and neuronal activity. This might be a preventive mechanism in neuronal cells to avoid intracellular viral replication and attack by phagocytosing cells. The change of microglia gene activity suggest that this is complemented by a shift in microglia activity to provide increased surveillance of their surroundings.