Type I interferons induce an epigenetically distinct memory B cell subset in chronic viral infection

SUMMARY

Memory B cells (MBCs) are key providers of long-lived immunity against infectious disease, yet in chronic viral infection they do not produce effective protection. How chronic viral infection disrupts MBC development, and whether such changes are reversible, remains unknown. Through single-cell (sc)ATAC-sequencing and scRNA-sequencing during acute versus chronic lymphocytic choriomeningitis viral infection, we identified a memory subset enriched for interferon (IFN)-stimulated genes (ISGs) during chronic infection that was distinct from the T-bet⁺ subset normally associated with chronic infection. Blockade of IFNAR-1 early in infection transformed the chromatin landscape of chronic MBC, decreasing accessibility at ISG-inducing transcription factor binding motifs and inducing phenotypic changes in the dominating MBC subset, with a decrease in the ISG subset and an increase in CD11c⁺CD80⁺ cells. However, timing was critical, with MBC resistant to intervention at 4 weeks post-infection. Together, our research identifies a key mechanism to instruct MBC identity during viral infection.

One Sentence Summary:

IFN dynamics in chronic versus acute viral infection determines memory B cell development.

ETOC BLURB

Chronic viral infection disrupts the ability to form long-lived B cell-mediated protective immunity, but the mechanisms are unclear. Cooper et al. demonstrate a key role of type I interferon (IFN-I) in governing memory B cell identity and show that the chromatin landscape and phenotype of memory B cells are set during a critical early window of IFN-I exposure.

Graphical Abstract

INTRODUCTION

A fundamental pillar of immunity is immune memory, consisting of antigen-experienced cells with increased capabilities to persist and rapidly reactivate upon reinfection. Memory B cells (MBCs) drive superior antibody-mediated responses upon re-infection compared to the primary response^1–3, enabling the rapid clearance of an infection before it can cause severe disease. However, chronic viral infections such as HIV, hepatitis C and cytomegalovirus can disrupt MBC development and antibody production, leading to incomplete and ineffective immune protection^4,5. The key molecular and microenvironmental determinants of chronic MBC development are still mostly unknown, thus limiting our ability to therapeutically redirect MBC development to sought after functional subsets.

The microenvironment plays a crucial role in shaping the immune response against viral infections. The interplay between viral replication and inflammatory mediators, such as interferons (IFNs), not only charts the course for whether an infection will be cleared effectively but also whether the immune response becomes dysfunctional^6–11. To understand how humoral memory is disrupted by chronic infection, much research has focused on understanding the shift in the phenotype of MBCs that is common across several severe infectious and inflammatory diseases^5,12. Antigen-experienced cells that have downregulated CD21 and CD27 and may upregulate Fc receptor-like 5 (FCRL5), the transcription factor T-bet and/or CD11c are detected at elevated frequency in the peripheral blood of HIV, hepatitis C, cytomegalovirus, systemic lupus erythematosus (SLE) and Long COVID patients^5,9,13–18. Cells of this phenotype can also be generated during acute viral infection^19,20, and thus it is still unclear how chronic viral infection fundamentally alters the MBC population to cause potential dysfunction, and why these changes may not be completely reverted by treatment. Antiretroviral therapy in HIV patients, for example, restores the prominence of the conventional MBC phenotype, but if viral loads rebound, MBCs revert to a disease state²¹. We do not know when, or if, the point of no return for recovery of MBC function occurs in chronic viral infection²². It is therefore vital that we better understand the fundamental changes to MBCs that continual pathogenic pressure imparts.

During an immune response, antigen-activated B cells react to signals within their pathogen-induced microenvironment and enact large-scale gene expression changes to induce differentiation into antibody-secreting cells, germinal center B cells (GCB) or MBCs. Epigenetic regulation facilitates the integration of these signals by enabling new gene expression programs^4,23,24. As a result of chronic viral pathogens and/or chronic antigen exposure, memory T cells become functionally exhausted, which has been linked to changes in the epigenome^23,25–29. In contrast, little is known about (i) how MBCs are fundamentally altered by epigenetic changes during chronic infection, (ii) whether there are key microenvironmental differences between acute and chronic infection that determine the type of MBC produced, and (iii) if chronic MBCs can be epigenetically altered to resemble conventional MBCs upon administration of therapeutics or whether they are resistant to recovery. Thus, improved resolution of the molecular architecture and regulation of distinct MBC subsets is key to enabling therapeutic targeting to drive effective humoral memory.

We set out to determine how and when the microenvironment during chronic viral infection codifies intrinsic changes within MBCs. We established a system in which MBCs formed in chronic infection, compared to acute infection of the same virus, mimicked CD21 and CD11c phenotypic changes observed in patients with ongoing severe viral infection. Single-cell ATAC-sequencing (scATAC-seq) and scRNA-seq defined the discrete subsets that arose during chronic or acute lymphocytic choriomeningitis virus (LCMV) infection. MBCs which substantially expand in the spleen during chronic infection were enriched for IFN-stimulated genes (ISGs) and were independent of the T-bet⁺ subset normally associated with chronic infection. The chronic MBC chromatin landscape was amenable to remodelling but relied on the timing of therapeutic intervention. Blocking type I IFN signaling early in the response both increased cell numbers and changed the ascendant epigenetic signature of MBCs. In contrast, modulation of viral load or administration of IFN blocking antibodies at 4 weeks post-infection did not convert the profile of chronic MBCs, suggesting that molecular imprinting of chronic MBCs occurs early in the response, and was not due solely to persistent viral load. Collectively, our study reveals the consequences of delayed IFN induction and magnitude on MBC identity in chronic infection.

RESULTS:

LCMV-Docile infection alters the formation of antigen-specific MBCs

We set out to investigate how the MBC epigenetic landscape is remodelled by a chronic viral infection-induced microenvironment. To do this, we established a system in which MBC formation was assessable during a polyclonal response to acute versus chronic viral infection. We used the comparative LCMV modelusing two strains of the same virus: one that induces an acute infection resolved by two weeks (LCMV-WE), or a chronic, persisting infection (LCMV-Docile)^4,30,31. The strength of this model is that B cell-differentiation in response to the same virus can be assessed in either an acute or chronic setting, overcoming potential issues that arise by comparing MBCs across different antigenic settings^5,32. In previous studies, adoptive transfer of clonal BCR transgenic B cells was used to assess LCMV-specific B cells. However, in these models, most if not all transferred cells were eventually depleted^33–35, thus prohibiting tracking of MBCs over time. To bypass these obstacles, we established a LCMV-nucleoprotein (NP)-specific antigen tetramer to identify low-frequency, endogenous, polyclonal antigen-specific B cells (Figure 1A). The NP protein sequence is almost identical between LCMV-WE and LCMV-Docile strains (data not shown), and unlike the glycoprotein does not decline over time in infected cells³⁶.

Figure 1. MBCs in LCMV-Docile infection adopt phenotypic changes.

(A) Schematic of acute (LCMV-WE) or chronic (LCMV-Docile) LCMV infection in C57BL/6 mice. Antigen-specific B cells assessed at various timepoints post-infection.

(B) Flow cytometric analyses of splenic MBCs (CD38⁺ % of B220⁺IgD⁻Tetramer⁺Decoy⁻) in mice at day 14 (n = 6 per strain), day 28 (WE: n = 13, Docile: n = 15), or day 56 (WE: n = 10, Docile: n = 9) post-LCMV-WE or LCMV-Docile infection. Combined data from 7 independent experiments. Data represent mean ± SEM.

(C) Flow cytometric analyses of splenic GCB (CD38⁻CD95^hi % of B220⁺IgD⁻Tetramer⁺Decoy⁻) at day 7 (n = 7 per condition), day 14 (n = 6 per condition), day 28 (n = 10 per condition), or day 56 (WE: n = 10, Docile: n = 9) post-LCMV-WE or LCMV-Docile infection. Combined data from 8 independent experiments. Data represent mean ± SEM.

(D and E) Flow cytometric analyses of (D) splenic PD-L2⁻CD80⁻ and (E) PD-L2⁺CD80⁺ within the MBC population (B220⁺ IgD⁻ Tetramer⁺ Decoy⁻ CD38⁺) at day 7 (n = 4 per condition), day 14 (WE: n = 11, Docile: n = 9), day 28 (WE: n = 9, Docile: n = 8), or day 56 (WE: n = 10, Docile: n = 8) post-infection. Combined data from 7 independent experiments. Data represent mean ± SEM.

(F and G) Flow cytometric analyses of (F) splenic CD21⁺ and (G) CD11c⁺ within the MBC population at day 28 (n = 10 per condition), or day 56 (WE: n = 10, Docile: n = 9) post-LCMV-WE or LCMV-Docile infection. Combined data from 4 independent experiments. Data represents mean ± SEM.

Related to Figure S1.

C57Bl/6 mice were infected with either LCMV-WE or LCMV-Docile and antigen-specific MBCs were investigated over time. Tetramer-binding B220⁺IgD⁻ cells that did not bind the negative control (decoy) tetramer (Figure S1A) were isolated by magnetic enrichment and analysed for quantitative and qualitative changes between acute and chronic infection. Splenic MBCs steadily increased over time in response to acute infection, up to d56 of assessment (Figure 1B). In comparison, in chronic infection there was a significant reduction in frequency of MBCs observed at d28 and d56 post-infection compared to acute (Figure 1B). As such, the affinity of NP-specific MBCs produced in chronic infection was decreased compared to those arising from acute infection (Figure S1B)³⁷. However, antigen-specific GCB were either comparable between conditions or increased following chronic compared to acute infection (Figure 1C, S1C), similar to previous observations of total GCB^4,38. While MBC numbers were comparable by d56 (Figure S1C) due to the increased total splenic cellularity in chronic infection⁴, by ~4 months post-infection MBCs were significantly reduced in chronic, compared to acute, infection (Figure S1C). Taken together, it appeared that chronic LCMV-Docile infection preferentially induces prolonged GCs and hypergammaglobulinemia^4,39 but MBCs are restrained over time compared to LCMV-WE.

MBCs in chronic LCMV infection undergo phenotypic changes

Phenotypically defined subsets within the MBC population have been linked to functional capacity^40–43. For example, conventional MBCs in mice which express the B7 family members PD-L2 and CD80 are primed to differentiate into antibody-secreting cells quickly during a recall response⁴¹. In comparison, “naïve-like” MBCs that lack expression of both PD-L2 and CD80 have fewer V gene mutations and are more prone to re-populate secondary GCs^41,44. In chronic LCMV infection, we found that there was a significant shift between these subsets: the naïve-like subset was increased by LCMV-Docile (Figure 1D), while PD-L2⁺CD80⁺ cells were decreased (Figure 1E). Similarly, other canonical MBC markers were also decreased (Figure S1D). Another layer of MBC complexity is observed in human patients across multiple chronic antigen disease contexts, with the emergence of MBCs that lack CD21 and upregulate CD11c in the peripheral blood⁵. This population was also increased following LCMV-Docile infection in vivo, with MBCs at d28 and d56 downregulating CD21 in both spleen and peripheral blood (Figure 1F Figure S1D–E) and upregulating CD11c (approximately 3-fold increase at d56; Figure 1G), compared to MBCs induced by LCMV-WE, although by ~4 months post-infection this difference had diminished (Figure S1F). Thus, MBCs formed in response to LCMV-Docile adopt phenotypic changes commensurate with chronic disease contexts in human patients.

LCMV-Docile infection diminishes MBCs expressing canonical transcriptomic signatures

Having established that LCMV-Docile induced these phenotypic changes, we sought to investigate the extent to which the transcriptome and chromatin landscape was also disrupted by chronic viral infection. Antigen-specific B cells were sort-purified at 4 weeks post-infection, a timepoint at which LCMV-WE has been cleared while LCMV-Docile infection persists, and profiled the transcriptome and immunoglobulin diversity via scRNA-seq and chromatin accessibility via scATAC-seq (Figure 2A). We used clustering analyses of the scRNA-seq (Figure 2B) and scATAC-seq (Figure 2C) to determine the key transcriptomic or epigenetic changes that ingrain different MBC fates during acute versus chronic infection. Specifically, this analysis would illuminate whether the heterogeneity seen at the phenotypic level was imprinted by the epigenome. Indeed, 12 clusters by scRNA-seq (Figures 2B, S2A, S2B) and 7 unique clusters were identified by scATAC-seq (Figure 2C). We first performed cell annotation using Immunological Genome Project (ImmGen) references to determine how these groups related to B cell differentiation states. The smaller, horseshoe-shaped group of scRNA-seq clusters (R3, R9 and R10) were identified as being GCB, confirmed by Fas transcript expression and the lack of Cd38 (Figures 2D, S2B–S2D). Furthermore, the gene encoding the proliferation marker Ki67 (Mki67) was mainly localised to GC clusters (Figure S2D). Conversely, MBCs, defined as antigen-specific B cells expressing Cd38 and lacking Ighd, were localised to the larger set of clusters (Figure S2C). Our dataset excluded plasma cells, as cells expressing the genes encoding the plasma cell markers CD138 and Blimp-1, Sdc1 and Prdm1 respectively, were scarce (Figure S2E). While our MBC definition would preclude analysis of IgD⁺ MBCs⁴⁵, it was necessary to ensure exclusion of any antigen-specific naïve B cells. To assess fidelity between datasets, we performed a label transfer whereby scRNA-seq annotations were transferred to the scATAC-seq dataset (Figure S3A, S3B). Analysis of merged data revealed that the GC scRNA-seq clusters overlaid with scATAC-seq cluster A5, and in confirmation, A5 was characterized by key loci of genes upregulated by GCB, such as Aicda, S1pr2 and Fas (Figures 2E and S3C).

Figure 2. Acute and chronic LCMV infection drive (epi)genetically distinct MBC subsets.

(A) Schematic of scRNA-seq and scATAC-seq set-ups.

(B) Unsupervised clustering of splenic antigen-specific B cells from 2 mice per condition visualized using UMAP, split by infection type. Each cell is represented by a point and colored by cluster. Graphical representation of percentage of total cells per cluster, split by infection type.

(C) Unsupervised clustering of nuclei from splenic antigen-specific B cells from 2 mice per group visualized using UMAP. Each nuclei is represented by a point and colored by cluster. Graphical representation of percentage of total barcodes per cluster, split by infection type.

(D) UMAPs of scRNA-seq (left) and scATAC-seq (right) data, with GCB (pink) and non-GCB (aqua).

(E) Coverage plot showing chromatin accessibility at the Aicda gene region, split by cluster.

(F) UMAP of scRNA-seq data, with Cd38⁺Ighd⁻ non-GC cells split by condition shown as purple dots.

(G and H) Violin plots representing the log2 fold-change expression of a panel of genes associated with (G) classical murine MBCs (MBC: Pdcd1lg2, Cd80, Cr2, Sell, Cxcr5, Ccr7 and Cxcr4) or (H) PD-L2⁻CD80⁻ murine MBCs, in Cd38⁺Ighd⁻ non-GC antigen-specific MBCs isolated from LCMV-WE vs. LCMV-Docile-infected mice. Data represent mean ± SEM.

(I) Heatmap of the top 25 DEGs in cluster R0 compared to all other clusters (excluding Ig genes). Also shown in Figure S2.

(J) Dendrogram showing the relationships in chromatin accessibility between the scATAC-seq clusters.

(K) UMAP of Tbx21 motif enrichment, with Z-score representing the sums of cut sites per cell which fall within all the peaks associated with the Tbx21 motif, split by condition.

(L) UMAP plot with cluster A6 highlighted. Heatmap of the DARs between acute and chronic condition in cluster A6.

Related to Figure S2, Figure S3 and Figure S4.

To understand whether changes in gene expression correlated to the decrease in conventional MBC phenotype observed in Figure 1, we interrogated our scRNA-seq dataset. Comparison of genes expressed by MBCs in our scRNA-seq data to those from other published murine RNA-seq datasets demonstrated that conventional markers of memory^46,47 or genes expressed in MBCs or MBC precursors^48,49 were decreased, while genes upregulated in PDL2⁻CD80⁻ MBCs were increased (Figures 2F–2H, S3D, Table S1). Thus, MBCs generated in response to LCMV-Docile infection showed significant changes in gene expression compared to the previously established conventional MBC transcriptome.

In particular, cluster R0 dominated the acute MBC population, constituting approximately half of all tetramer-specific cells in the acute-infected mice, whilst R7 was barely detectable. However, this distribution was notably reversed in chronic infection, with cluster R0 substantially diminished (Figure 2B). Cluster R0 had increased expression of Cd55, which serves to protect host cells from complement-mediated damage, Fcer2a which encodes CD23, and the transcription factor Klf2, which regulates cell trafficking via upregulating CD62L⁵⁰, the gene for which (Sell) was also upregulated in R0 (Figure 2I, Table S1). Several transcription factors previously associated with MBC formation were also decreased in chronic infection (Figure S3E–F, Table S1). Given the shift away from a conventional MBC identity following LCMV-Docile infection, we assessed isotype distribution and somatic hypermutation (SHM) frequency across clusters by scVDJ-seq (Figures S4A and S4B). While there was little change in isotype distribution in antigen-specific MBCs in acute and chronic infection (Figure S4A), there was a large reduction in unmutated R0 cells (Figure S4B). Taken together, clustering analysis confirmed that chronic viral infection imparted discernible, sustained changes to MBCs, with the emergence of distinct MBC subsets and the diminution of MBCs with a conventional transcriptomic identity. Collectively, these data also demonstrate the power of this tetramer-based LCMV system to delineate MBC fate in vivo in chronic viral infectious disease.

Distinct epigenetic signatures define MBC subsets

We next assessed the chromatin landscape shift in clusters in chronic, compared to acute, LCMV infection. All clusters were detectable in both acute and chronic conditions (Figure 2C). However, there were clear changes in the distribution of clusters within each infectious condition. In the scATAC-seq dataset, cluster A1 was greatly expanded in chronic infection, whereas cluster A3 had become barely detectable (Figure 2C). Given that one cluster expanded in chronic infection whilst another diminished, we asked whether these clusters were closely related and may in fact just be reflective of a small shift in the chromatin landscape between infections. We performed a pseudo-bulk analysis of each cluster and calculated a distance matrix to visualize the hierarchical relationship between clusters. Despite the paired change in these clusters, the chronic (A1) and acute (A3) expanded clusters were highly epigenetically distinct from one another (Figure 2J).

T-bet-expressing memory B cells undergo few molecular changes in chronic LCMV infection

One of the defining features of MBC changes during chronic infectious disease has been the increased prominence of a subset expressing T-bet, CD11c (encoded by the gene Itgax) and the chemokine receptor CXCR3 in human peripheral blood⁵. However, T-bet-expressing MBCs are also induced in acute infection^{15,16,19,20,51}. One key question that has remained unanswered is whether the T-bet-associated MBC subset which arises in chronic disease is distinct from T-bet-expressing MBCs produced in acute immune responses. To answer this question, we first asked whether T-bet-expressing MBC could be uniquely resolved as a distinct population in our multiomic dataset. ATAC cluster A6 showed enrichment of Tbx21 motif (Figure 2K) and had increased chromatin accessibility within genes associated with T-bet-expressing MBCs, including Tbx21, Itgax and Zeb2 (Figures S4C–S4E). Notably, A6 overlapped with cluster R5 in the merged dataset (Figure S3B), which correspondingly showed increased expression of T-bet-associated genes (Tbx21, Cxcr3, Itgax, Cd80, Zeb2), whereas other memory markers such as Cr2, Hhex and Pdcd1lg2 had a wider distribution (Figures S4F and S4G, Table S2). The relationship between A5 and A6 (Figures 2I, S4C), as well as the SHM frequency (Figure S4B), suggested that T-bet-expressing cells at week 4 post- infection may be derived from GC-dependent in addition to GC-independent pathways^19,52. In summary, both acute and chronic LCMV infection were able to induce a Tbx21-expressing MBC subset, with shared transcriptomic profiles.

To determine whether chronic infection imparts any further distinct chromatin landscape changes within this population, we compared the epigenetic signature between acute versus chronic barcodes in cluster A6 (Figure 2L). Only 6 differentially accessible regions (DARs) were detectable between conditions, indicating that Tbx21-expressing cells had a similar epigenetic profile regardless of acute or chronic infectious context. Thus, LCMV-Docile infection did not impart distinct changes to the T-bet⁺ population that may alter its characteristics during chronic infection.

Chronic LCMV infection induces an ISG-enriched MBC population

We then turned our attention to the unresolved question: what are the defining molecular features of MBCs that are expanded during chronic viral infection? To investigate how chronic infection remodels the MBC population, we examined the attributes of the scATAC-seq and scRNA-seq clusters that were expanded in chronic, compared to acute, viral infection. There was a clear expansion of cluster A1 in chronic infection (Figure 2C). To identify distinct properties of the chronic-emergent A1 cluster, we compared its chromatin landscape to all other non-GC clusters. This analysis identified 39 distinct chromatin regions that were open in A1 (Figures 3A and 3B, Table S3). These regions were enriched for binding motifs of transcription factors important for B cell biology, as assessed by HOMER⁵³ (which tests for motif enrichment in differentially accessible regions; Figure 3C, Table S3) or chromVAR⁵⁴ (which tests for significant differential motif activity between clusters; Figure S5A, Table S4). For example, several with known roles in regulating B cell transcriptional programs (e.g., E2A, PU.1, Foxo1) and negatively regulating GCB (Bhlhe40; corresponding increased transcriptional expression shown in Figure S5B)⁵⁵ were enriched in areas of increased chromatin accessibility in chronic MBCs.

Figure 3. Chronic infectionexpands an ISG-associated subset.

(A) UMAP plot with cluster A1 highlighted (excluding cluster A5).

(B) Heatmap of the DARs in cluster A1 compared to all other clusters (excluding cluster A5).

(C) Representative selection of transcription factor binding motifs enriched in cluster A1 vs. all other clusters (excluding A5), generated using HOMER.

(D and E) Heatmaps of the top 25 DEGs in (D) cluster R7 or (E) cluster R2 (excluding Ig genes), compared to all other clusters. Also shown in Figure S2.

(F and G) Coverage plots showing chromatin accessibility at the Rtp4 gene region, split by (F) condition, and (G) split by cluster.

(H) UMAPs of gene expression of Rtp4, Ifi44 and Oasl2 split by condition, with positive expression represented by pink dots.

Related to Figure S5.

Next, we interrogated the chronic-emergent clusters from the scRNA-seq dataset, clusters R7 and R2 (Figures 3D and 3E). Cluster R7 was almost entirely exclusive to mice responding to chronic infection, with only ~0.2% of antigen-specific B cells in acute infection populating this cluster. 20 out of the top 25 upregulated differentially expressed genes (DEGs) in cluster R7 were ISGs: Ifit3, Isg15, Ifitm3, Isg20, Usp18, Slfn5, Irf7, Ifi213, Ifi208, Trim30a, Irgm1, Ifi47, Bst2, Tor3a, Ly6c2, Phf11b, Lgals9, Ifi203, Ifi209, Ifi27l2a (Figure 3D, Table S3). Cluster R2 shared many of the top DEGs from R7, including Lgals9, Ifi27l2a, Ifi47, Ifi203, Ifi208 and Rtp4 (Figure 3E, Table S3). Both of these scRNA-seq subsets overlapped with A1 from the scATAC-seq data, although Cluster R7 had a broader distribution (Figure S3B). The high expression of Rtp4 in both R2 and R7 was particularly of note: Rtp4 is IFN-I-inducible and is a regulator of type I IFN responses⁵⁶. Chromatin accessibility was increased at the Rtp4 promoter region in antigen-specific B cells in chronic compared to acute infection (Figure 3F), particularly in the chronic-expanded cluster A1 (Figure 3G). Furthermore, Rtp4-regulated genes were mostly upregulated in MBCs during chronic infection⁵⁶ (Figure 3H and S5C).

We analyzed the dataset for enrichment of molecular signatures to better understand which biological pathways were enriched in MBCs during chronic infection. Ingenuity Pathway Analysis (IPA; Figure 4A) confirmed that chronic MBCs were enriched for an IFN gene set. This ISG signature appeared to be concentrated in clusters that were expanded in chronic infection (Ifi44 and Oasl2, Figure 3H; Irf7, Figure S5D), although some genes did have an overall general increase in expression (e.g. the IFN-inducible Ifi2712a, Figure S5D). Moreover, gene set enrichment analysis (GSEA) revealed that IFNα response and IFNγ response hallmark gene sets were enriched in the chronic-emergent subset, compared to other subsets (Figure 4B). Of the top 50 genes enriched in this cluster, Ly6c2, Irf7 and Lgals9 (Figure 4C) had commercially available antibodies available to assess protein by flow cytometry. While technical limitations prohibited confirmation of IRF7, we confirmed that Ly6c and Galectin-9 were increased in MBCs formed during LCMV-Docile, compared to LCMV-WE infection (Figure 4D). We proceeded to use Ly6c, which had distinct positive and negative populations, to sort-purify MBC subsets for functional analysis. Mice were infected with LCMV-Docile, and at ~4 weeks post-infection, Ly6c⁺ and Ly6c⁻ MBCs were stimulated in vitro with CD40L, anti-Ig and interleukin(IL)-21 (Figure 4E). Assessment after 4 days revealed that Ly6c⁺ MBCs had a reduced capacity to produce plasmablasts compared to Ly6c⁻ MBCs (Figure 4F), thereby suggesting potential functional differences between chronic-emergent MBCs compared to canonical MBCs.

Figure 4. Chronic-emergent MBC have a reduced capacity to respond in vitro.

(A) IPA analysis of differentially expressed genes within the IFN pathway upregulated in MBCs (Cd38⁺Ighd⁻ non-GC) from LCMV-Docile-infected mice.

(B) GSEA of hallmark IFNα response (left) and IFNγ response (right) gene sets in cluster R7 compared to other clusters.

(C) UMAP of combined gene expression of Ly6c2, Irf7 and Lgals9, split by condition, with positive expression represented by pink dots.

(D) Flow cytometric assessment of Ly6c and Galectin-9 in mice infected with LCMV-WE or LCMV-Docile (n = 4 per condition) 15 days prior.

(E) Schematic of setup: Sort-purified Ly6c^pos and Ly6c^neg MBCs were stimulated in vitro with CD40L, anti-Ig and IL-21.

(F) Number of plasmablasts formed 4 days post-stimulation of Ly6c^pos and Ly6c^neg MBCs (n = 5 per subset). Combined from 2 independent experiments.

Related to Figure S5.

Our results thus far revealed a role for IFN-I in MBC phenotype and epigenetic landscape during chronic viral infection. Given that both direct^34,57 and indirect^33–35 effects of IFN-I have previously been shown to regulate B cells during LCMV infection, we set out to determine whether B cell-intrinsic IFNAR signaling directly modulated the phenotype of MBCs in chronic infection. Bone marrow from B cell-deficient mice (μMT) and either wild-type or Ifnar^−/− mice were mixed in a 80:20 ratio and injected into irradiated recipients. After at least 6 weeks post-reconstitution, mice were infected with LCMV-Docile and assessed 14 days post-infection (Figure S5E). In the absence of B cell-specific IFNAR, Ly6c-expressing cells were significantly reduced (Figures S5F and S5G). While Ly6c and Galectin-9 were both increased on MBCs during chronic, compared to acute, infection (Figure 4D), there was an increase in Galectin-9 expression such that Ly6cGalectin-9⁺ MBCs were significantly increased, in direct contrast to Ly6c⁺Galectin-9⁻ MBCs (Figures S5G and S5H). Given that Galectin-9 is the ligand for Tim3⁵⁸, the exhausted CD8⁺ T cell marker⁵⁹, this may suggest a distinct role for Galectin-9 in mediating B cell interactions with CD8⁺ T cells. Indeed, when CD8⁺ T cells were depleted during the first week of LCMV-Docile infection, Ly6c⁻Galectin-9⁺ MBCs were decreased while Ly6c was unchanged (data not shown), demonstrating that there was both a B cell-intrinsic and B cell-extrinsic impact of IFNAR signaling. Nevertheless, our data demonstrates that MBC phenotype was modulated directly via IFNAR signaling during chronic LCMV infection.

Type I IFN is a key driver of phenotypic and epigenetic changes to MBCs during chronic viral infection

The timing and magnitude of IFN is an important determinant for whether viral infections will progress to severe disease⁶⁰. A rapid but transient peak of IFN-I in the first days of infection enables the cellular response to effectively clear the virus. In contrast, low and/or delayed IFN-I induction leads to persisting viral loads, an ineffective and eventual dysregulated immune response, and increased morbidity^6,60–62. Given the ISG signature in the chronic-emergent subsets at 4 weeks post-infection in both the scRNA-seq and scATAC-seq analysis (Figures 3, 4), we hypothesized that IFN may be instructing identity and/or driving maintenance of the chronic MBC subset. Thus, we assessed whether either type I or type II IFN was a key driver of chronic MBC identity. Wild-type mice infected with LCMV were administered a blocking antibody to either IFNAR-1 or IFNγ every 3 days from d2 to d14 post-infection (Figure 5A). IFNAR-1-blocking induced qualitative and quantitative changes to the MBC population, while IFNγ-blocking did not (Figures 5B–5D). While the frequency of antigen-specific IgD⁻ B cells was similar with or without IFN blocking treatment (Figure S6A), an increased frequency of MBCs was dominated by a significant increase in CD11c⁺ and CD80⁺ MBCs (Figures 5B–5D, Figures S6B–S6D). Overall, activated B cell numbers were also increased (Figure S6E), which may be due to inhibition of IFN-mediated B cell-depleting mechanisms described in previous adoptive transfer studies^33–35. We confirmed that Ly6c, encoded by a gene highly expressed in chronic-emergent MBCs (Figures 3, 4), was decreased upon IFNAR-1 blocking treatment (Figure 5E).

Figure 5. Type I IFN is a key driver of phenotypic and epigenetic changes to MBCs during chronic viral infection.

(A) Schematic of early type I or II IFN blocking treatment cohort and timepoint.

(B, C and D) Flow cytometric analyses at day 15 post-infection of (B) antigen-specific MBCs (B220⁺IgD⁻Tetramer⁺Decoy⁻CD38⁺), (C) CD11c⁺ frequency within MBCs, and (D) PD-L2 and CD80 MBC subsets in mice treated with anti-IFNAR1 (MAR1–5A3) (n = 8), anti-IFNγ (XMG1.2) (n = 5) or IgG control (n = 8) at day 15 post-infection in mice infected with LCMV-Docile. Combined data from 2 independent experiments. Data represents mean ± SEM.

(E) Expression of Ly6c in antigen-specific MBCs in mice infected with LCMV-WE and treated with an IgG control, LCMV-Docile mice treated with anti-IFNAR1 or IgG control (n = 6 per condition) at day 15 post-infection.

(F) Volcano plot of DARs in MBCs in IgG control (red) vs. anti-IFNAR1 blocking treatment (blue) in LCMV-Docile infection at day 15 post-infection.

(G) Heatmap showing DARs in MBCs in LCMV-WE and LCMV-Docile infection, with IgG control, anti-IFNAR1 or anti-IFNγ blocking treatment at day 15 post-infection.

(H) Representative selection of transcription factor binding motifs with increased (left) or decreased (right) chromatin accessibility in MBCs following early IFNAR1-blocking treatment generated using HOMER.

Related to Figure S6.

We next tested whether IFN-I blockade could induce extensive remodelling of the chromatin landscape regulating MBCs. Antigen-specific CD38⁺IgD⁻ MBCs were sort-purified at day 15 post-infection after blocking antibody treatment, and ATAC-seq was performed. The chromatin landscape of MBCs was largely remodelled in chronic mice treated with IFNAR-1 blocking antibody, compared to mice treated with control antibody, with 616 DARs identified (Figure 5F and 5G, Table S5). In contrast, IFNAR-1 blockade during acute LCMV infection induced far fewer chromatin landscape changes (Figure 5G). In chronic mice, gains in accessibility in MBCs were associated with access to transcription factor motifs such as Pu.1, Ebf1 and Tbx21 (Figure 5H), the latter likely responsible for the increased in CD11c expression (Figure 5C). While there were fewer changes associated with decreased accessibility, the transcription factor binding motifs in these decreased DARs were concordant with those that had been identified in the ISG chronic MBC subset (Figures 3C and S5A). In line with IFN response regulation, these included IFN regulatory factors (Figure 5H, Table S5); in addition, Tcf3, Tcf12, Tcf21, Ptf1a and POU factors (Figures 5H, 3C and S5A). Kruppel-like factors, which can play a role in increasing accessibility to cofactors⁶³, were also detected in these DARs. Thus, IFNAR-1 blockade led to a remodelling of MBC subsets away from the ISG molecular program and instead promoted the Tbx21, CD11c⁺ subset. Together, these data suggest that IFN-I signaling is a critical factor governing MBC identity at the epigenetic and phenotypic level in chronic LCMV infection.

Viral load is not a sole driver of chronic MBC phenotype

The ongoing persistence of virus and viral load has been linked with changes to MBC phenotype and incomplete immune protection in HIV patients^21,64. While many MBCs are formed within the first week post-immunization or infection, they can also continually form throughout an ongoing response^{39,45,49,65–67}. We therefore expected that the ongoing presence of virus was an important determinant in promoting chronic MBC attributes, in addition to the role for early microenvironment factors observed in Figure 5. Yet, patients on anti-retroviral therapy do not completely restore immune memory populations^68–71, suggesting that there may be a time-based limitation on the establishment of long-lived MBC identity during chronic viral infection. We therefore tested the timing of MBC formation and the extent to which viral load shaped MBC development and maintenance in chronic infection. We first investigated the in vivo kinetics of MBC formation and longevity in acute versus chronic viral infection with 5-bromo-2’-deoxyuridine (BrdU) labelling experiments. A defining characteristic of long-lived MBCs is that they become quiescent⁷ and thus should retain BrdU if formed during the labelling period. Infected mice were administered BrdU over 3-day windows (Figure 6A). BrdU uptake and long-term retention was examined across the course of infection (Figure 6B–D). Our data indicated that the majority of BrdU⁺ MBCs detected at d56 post-infection were generated early during infection (BrdU-labelled D5–7; Figure 6B). LCMV-WE-induced MBCs generated in this early time frame were stable over time, while BrdU⁺ chronic MBCs progressively declined (Figure 6B). This was likely due to continual viral presence-induced proliferation, as there was an increase of BrdU⁺ MBCs in chronic infection when BrdU was administered in the days immediately prior to d56 (Figure S6F).

Figure 6. Chronic LCMV infection induces delayed but sustained IFN exposure.

(A) Schematic of BrdU treatment groups and timepoints following LCMV-WE or LCMV-Docile infection.

(B, C and D) Flow cytometric analyses of BrdU⁺ frequency within MBCs (B220⁺IgD⁻Tetramer⁺Decoy⁻CD38⁺) at various timepoints post-infection in mice infected with LCMV-WE or LCMV-Docile and treated with BrdU at (B) days 5–7 post-infection, (C) days 12–14 post-infection or (D) days 25–27 post-infection. Combined data from 3 independent experiments. Data represents mean ± SEM.

(E and F) LegendPlex assay results of sera (E) IFNα or (F) IFNγ at various timepoints post-infection with LCMV-WE or LCMV-Docile (n = 4–8 per condition). Combined data from 3 independent experiments. Data represents mean ± SEM.

Related to Figure S6.

IFNAR-1 blockade increased viremia (Figure S6G), in line with the essential role of IFN-I signaling for viral control early in the course of infection^72–74. Given this, we asked whether elevated viremia, upstream of IFN production, may drive changes to MBC identity during chronic LCMV infection, rather than IFN itself. To assess this possibility, we used the antiviral pyrazinecarboxamide-derivative Favipiravir, which is broadly active against a wide variety of RNA viruses including LCMV⁷⁵. We administered Favipiravir daily from d2–8 post-infection and culled mice at d12 to investigate whether MBC subsets were reliant on the presence of elevated viremia (Figure S6H). However, although Favipiravir treatment lowered viral titers (Figure S6I), the distribution of MBC subsets did not fundamentally change (Figure S6J–S6L). This was true also when treatment was performed at 4 weeks post-infection (data not shown). Thus, the chronic MBC phenotype was not solely driven by sustained high viremia.

There is a critical window in which IFN-I exposure directs establishment of the MBC chromatin landscape during chronic viral infection

Our results thus far showed that the MBC phenotype and chromatin landscape could be shaped by IFN during the first two weeks of infection. Yet, IFN blocking administration had a limited effect on the chromatin landscape of MBCs during acute infection. Furthermore, we hypothesized that IFN may be elevated in chronic infection later in the response, given the ISG signature in MBCs was prominent at d28 post-infection. We therefore compared the kinetics of IFN production in LCMV-WE vs. LCMV-Docile infection. While previous studies have compared IFN induction in LCMV-Armstrong versus LCMV-Clone13 in fine detail^57,60,61, direct comparison of WE and Docile beyond the first two days post-infection⁷² was lacking. As previously described, LCMV-WE induced a high wave of IFNα by 24hr post-infection, which is critical to mount an efficient immune response⁶. In comparison, IFNα in mice infected with LCMV-Docile was delayed and did not reach the same peak as LCMV-WE (Figure 6E) and is substantially lower compared to the high wave seen in LCMV-Clone13^57,60,61. Similar differences between WE and Docile strains were observed with IFNγ kinetics (Figure 6F). While IFN induction was delayed in chronic infection, the amount of IFN in LCMV-Docile mice at d15 and d30 post-infection was increased over LCMV-WE and at a similar concentration to that observed at d4, suggesting a low concentration of sustained IFN in chronic infected mice (Figures 6E and 6F).

Given the sustained presence of IFN in chronic infection, we asked whether continual low exposure to IFN-I was critical for the maintenance of the chronic MBC identity. To investigate, we treated mice with IFN-I blocking antibody late in the course of an established infection, from d28 to d40 post-infection (Figure 7A). We also administered an IFNγ blocking antibody given that there was also a sustained increase in IFNγ over time (Figure 6F), and the possibility that IFNγ could be inducing the ISG signature detected in our chronic MBCs as has been shown in CD4⁺ T cells⁷⁶ (Figure 7A). However, the use of these blocking antibodies did not substantially alter MBC frequency or phenotype (Figures 7B–7E). Therefore, MBC phenotype was impervious to phenotypic change upon IFN blocking antibody administration late in the primary response to LCMV-Docile. In agreement with the phenotypic data, we identified only 14 DARs between mice that received the control treatment and the IFNAR-1 blockade at day 41 post-infection (Figure 7F–7I, Figure S6M). However, it was possible that effects of IFNAR-1 blockade were undetected because GC-derived MBCs formed later in the response constitute only a small fraction of the total MBC population. To determine whether GC-derived MBCs formed during the late administration period of IFNAR-1 blocking antibody could undergo phenotypic change, we utilized the S1pr2-ERT2cre-tdTomato reporter model⁷⁷ to inducibly fate map GC-derived MBCs (Figure S7A). At approximately 3 weeks post-infection with LCMV-Docile, tamoxifen was administered to induce tdTomato expression in S1pr2-expressing GCB, followed by IFNAR blockade (Figure S7A). Assessment of tdTomato⁺ MBCs confirmed that GC-derived MBCs, similar to non-fate mapped MBCs that had formed earlier in the response, did not undergo significant phenotypic changes upon late IFN blocking (Figure S7B–S7G). Therefore, there is a critical window early in infection in which blocking IFN-I can govern MBC identity. Collectively, these experiments provide critical insights into how changes in IFN dynamics during early viral infection can redirect MBC development in addition to IFN-mediated effects on viral persistence and severity of disease.

Figure 7. IFN blocking antibody administration late in the primary response to LCMV-Docile infection did not induce changes in the phenotype and epigenetic profile of MBCs.

(A) Schematic of late type I or II IFN blocking treatment cohort and timepoint.

(B, C and D) Flow cytometric analyses of (B) antigen-specific MBCs (B220⁺IgD⁻Tetramer⁺Decoy⁻CD38⁺), (C) CD11c⁺ frequency within MBCs, and (D) CD21⁺ frequency within MBCs in mice treated with anti-IFNAR1 (MAR1–5A3; n = 7), anti-IFNγ (XMG1.2; n = 6) or IgG control (n = 5) at day 41 post-infection. Combined data from 2 independent experiments. Data represents mean ± SEM.

(E) Flow cytometric analyses of PD-L2 and CD80 MBC subsets in mice treated with anti-IFNAR1 (MAR1–5A3; n = 3), anti-IFNγ (XMG1.2; n = 3) or IgG control (n = 2) at day 41 post-infection. Combined data from 2 independent experiments. Data represents mean ± SEM.

(F) Volcano plot of DARs in MBCs in IgG control (red) vs. anti-IFNAR1 blocking treatment (blue) in LCMV-Docile infection at day 41 post-infection.

(G) Graph representing number of DARs in MBCs in treatment groups versus IgG control at day 15 and day 41 post-infection.

(H) Heatmap showing DARs in MBCs in IgG control, anti-IFNAR1 or anti-IFNγ blocking treatment in LCMV-Docile infection at day 41 post-infection.

(I) Coverage plots showing chromatin accessibility at the Cd28 (left), Adgrg5 (center) and Pou2af1 (right) gene regions, split by treatment group.

Related to Figure S6 and Figure S7.

DISCUSSION:

Here, we set out key findings which define how the chronic MBC identity diverges from MBCs responding to acute viral infection. Integration of scATAC-seq and scRNA-seq revealed that the diversity in the MBC population is wired by the epigenome. While most clusters were present in both acute and chronic LCMV infection, a clear expansion of a cluster characterised by ISGs was identified in chronic infection that was both independent of the Tbx21-expressing cluster and seemingly at the expense of a canonical MBC subset. The chromatin landscape of chronic MBCs was established early in the primary response, with IFNAR-1 signaling a key regulator of MBC subset identity. After this period, modulation of viral load or IFN blockade had little effect on chromatin accessibility or phenotypic distribution of subsets. Thus, plasticity was not inherent within the established MBC pool in response to these therapeutics, even with continual proliferation induced by persisting virus, limiting the potential to therapeutically convert MBC subsets. Collectively, our research illuminates that the early microenvironment not only charts a course for the severity of the disease, but also determines the trajectory of chronic MBC subset development.

Our data therefore suggests a limited window of opportunity to enact large-scale changes to the MBC chromatin landscape. This is congruent with previous studies demonstrating that the bulk of MBCs forms early after immunization of mice with a model antigen⁶⁵. However, given that fluctuations in viral load have been associated with similar fluctuations of atypical MBCs in HIV patients, we expected that the ongoing presence of virus would enable remodelling of the MBC population by administration of an antiviral or cytokine blocking experiments. Yet, despite the ongoing viral load and MBC proliferation, our data showed that MBCs were epigenetically stable when exposed to IFN-blocking antibodies at the first month of chronic viral infection. This may explain why HIV patients on anti-retroviral therapies are less likely to completely recover their ability to generate effective immune protection, even if MBC phenotype has been partly restored to resemble healthy patients^{22,64,68–71,78}. Thus, MBC identity may be imprinted in a ‘hit and run’ manner, in which the distinct epigenetic signatures of each cluster are established in the initial phase of the response.

As such, chronic infection did not largely change the core epigenetic identity of many of the MBC subsets formed, but instead expanded or contracted their overall representation. However, by avoiding a priori assumptions of which phenotypic markers chronic MBCs should possess, our data revealed that there was a clear expansion of a distinct subset enriched with an ISG signature in chronic infection. Similar enrichment of ISGs has been identified in analyses of B cells in HIV patients^79,80 and patients with severe SARS-CoV-2 infection^81–83. Notably, this chronic-emergent subset was independent of the Tbx21-associated, CD11c-expressing subset that dominates descriptions of ‘atypical’ or ‘age-associated’ B cells in humans and mice^18,84–86. In fact, it is possible that the ISG signature is highly expressed in the chronic MBC subset in part due to the absence of T-bet expression⁷⁶. In contrast to the chronic-emergent subset, we corroborate that Tbet⁺CD11c⁺ cells are not uniquely reliant on a chronic viral infection to be formed, nor does chronic infection impart major changes on their chromatin landscape. In addition to previously established positive regulators IFNγ, toll like receptor 7 and IL-21^9,18,87–89; we also showed that this subset can be promoted by IFNAR-1 blockade. Its function appears to be context-dependent, ranging from protective^90,91 to unnecessary for long-lived immunity¹⁸ to detrimental in human patients responding to different types of disease⁹². Thus, depending on whether this subset is functional (e.g. in malaria) or dysfunctional (e.g. in HIV or SLE), our data suggests a therapeutic strategy for modulation.

The emergence of a distinct ISG-enriched chronic memory subset has also been identified in single-cell RNA-sequencing datasets of SLE and malaria patients^93,94. A recent study identified the emergence of cells with high ISG expression in SLE patients compared to healthy donors⁹³, and another, separate subset which exclusively expressed a T-bet (Tbx21)-associated gene signature⁹³. The ISG-enriched cluster (R7) in our data correlates with the ISG-expressing subset expanded in SLE patients, with expression of genes such as Ifit3, Ifitm3, Isg15, Usp18, Irf7, Zbp1, Bst2, Lgals9, Epsti1, Ifi35, Stat1, Shisa5, Sp110. Moreover, Reyes et al. (2022) recently identified three subsets of atypical MBCs expanded in individuals with chronic Plasmodium falciparum exposure⁹⁴. In line with our data, one subset upregulated ISGs and another had high expression of Tbx21. While our data suggests that the ISG-enriched subset may be ineffective at generating plasmablasts, it is not yet determined whether it has a separate function, and whether its functional attributes will be more generalizable across antigenic contexts then T-bet⁺ MBCs. Moreover, it is likely that individual genes, such as Lgals9 (encoding Galectin-9), may have distinct functions such as mediating interactions with CD8⁺ T cells⁵⁸. Thus, future studies using new conditional deletion and lineage tracing models will enable dissection of the precise function of these identified genes upregulated in chronic-emergent memory B cells.

The kinetics and magnitude of IFN-I responses during viral infection has an important role in disease pathogenesis across various contexts of viral infection^73,95,96. Accumulating evidence points to delayed induction of a low amount of IFN-I being a key determinant in the severity and persistence of viral infection. This has particularly come to the fore during COVID-19^6,62 and the postulated mechanistic relationship between IFN and severity of disease in response to SARS-CoV-2 infection and Long COVID^8,97,98. Our study gives critical insight into how MBCs may be altered by this difference between acute (high, quick load of IFN-I) versus chronic (delayed, low but sustained IFN-I). The IFN-I-induced shaping of the MBC compartment may explain why patients with autoantibodies to IFN can form immune memory to vaccination but do not generate adequate memory protection during SARS-CoV-2 infection^99,100. It also suggests that IFN-I therapy^73,98 may have a significant impact on MBC development. It is possible that these differences also lead to the induction of CD21⁻CD27⁻ ‘double negative’ cells in Long COVID¹⁰¹, thus acting as a biomarker of the early microenvironment during the nascent immune response. Thus, studying the chromatin landscape at single-cell resolution in vivo can enable the identification of biomarkers of the critical early immunological events that govern the effectiveness of the host response and thus disease severity.

Beyond vaccines, few strategies exist to exert influence over the development of MBCs. This is a significant gap in our capabilities not only for treating non-vaccine-preventable infections, but also for immunodeficiencies in which patients cannot form adequate humoral memory. Our research demonstrates that antigen-specific B cell subset distribution during infection is underpinned by distinct epigenetic states, and that it is possible to remodel MBCs to promote or repress certain subsets. In addition, we have previously shown that the deletion of an epigenetic regulator can alter MBC subset distribution and hence their function during a recall response¹⁰². Given that epigenetic factors can be targeted in vivo, such as with small molecule inhibitors⁴, these findings portend the development of ways to therapeutically target MBCs. Development of this strategy and identification of druggable targets will be further enhanced by illuminating the epigenome – that is, the repressive or activation marks that regulate chromatin landscapes – that may enable longevity and rapid responsiveness of MBCs in productive immune responses or promote dysfunction during disease.

In summary, single-cell resolution has illuminated the shared and divergent splenic MBC populations during chronic versus acute viral infection. Furthermore, our research shows how chromatin landscapes underpinning MBC subsets can be reshaped through targeting IFN-I, thus initiating a therapeutic scope for promoting sought-after efficacious MBC subsets or repressing the formation of dysfunctional cells in disease.

LIMITATIONS OF THE STUDY:

Our study showed the Ly6c^pos MBCs had a reduced capacity of forming plasmablasts when stimulated in vitro. However, it did not rule out the possibility that ISG-enriched MBCs have a function(s) independent of antibody production, which warrants investigation in future studies. This study did not assess in vivo function of specific MBC subsets due to the rarity of these cells. It is possible that Ly6c or Galectin-9 expression may reflect transient MBC states that may not persist as long as canonical MBC. Future studies could test the longevity of these subsets in severe infections that resolve over time. Lastly, while we did not perform (epi)genetic analyses of MBCs from B cell-intrinsic Ifnar^−/− BM chimeras, previous work has demonstrated that IFN directly regulates ISG gene expression and chromatin landscape in B cells¹⁰³ and a suite of genes routinely used as predictive indicators of Ifnar signaling were upregulated in our dataset.

STAR METHODS:

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact Kim Good-Jacobson (kim.jacobson@monash.edu).

Materials availability

Biotinylated proteins generated in this study are available from the lead contact.

Data and code availability

• The sequencing datasets have been deposited at GEO and are publicly available as of the date of publication. Accession numbers are listed in the key resources table.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-mouse/human B220 APC-Cy7	BD Biosciences	Cat # 552094; RRID: AB_394335
Anti-mouse/human B220 APC-Cy7	BioLegend	Cat # 103224; RRID: AB_313007
Anti-mouse/human B220 BV711	BioLegend	Cat # 103255; RRID: AB_2563491
Anti-mouse IgD BV605	BD Biosciences	Cat # 563003; RRID: AB_2737944
Anti-mouse CD38 AF488	BioLegend	Cat # 102714; RRID: AB_528796
Anti-mouse CD38 Pacific Blue	BioLegend	Cat # 102720; RRID: AB_10613468
Anti-mouse CD95 PE-Cy7	BD Biosciences	Cat # 557653; RRID: AB_396768
Anti-mouse CD138 PE	BioLegend	Cat # 142504; RRID: AB_10916119
Anti-mouse CD138 BV711	BD Biosciences	Cat # 563193; RRID: AB_2631190
Anti-mouse CD21/CD35 BV421	BD Biosciences	Cat # 562756; RRID: AB_2737772
Anti-mouse CD21/CD35 PerCP-eFluor 710	Thermo Fisher Scientific	Cat # 46–0212-82; RRID: AB_10855041
Anti-mouse CD80 BV421	BioLegend	Cat # 104726; RRID: AB_2561445
Anti-mouse Ly6c BV510	BioLegend	Cat # 128033; RRID: AB_2562351
Anti-mouse Galectin-9 PE-Cy7	BioLegend	Cat # 137913; RRID: AB_2750157
Anti-mouse PD-L2 PerCP-Cy5.5	BioLegend	Cat # 107218; RRID: AB_2728126
Anti-mouse PD-L2 PE-Dazzle 594	BioLegend	Cat # 107216; RRID: AB_2749894
Anti-mouse CD11c APC-Cy7	BioLegend	Cat # 117324; RRID: AB_830649
Anti-mouse BrdU BV510	BD Biosciences	Cat # 563445; RRID: AB_2738210
TotalSeq-C0301 anti-mouse Hashtag 1 Antibody	BioLegend	Cat # 155861; RRID: AB_2800693
TotalSeq-C0302 anti-mouse Hashtag 2 Antibody	BioLegend	Cat # 155863; RRID: AB_2800694
Fixable Viability Stain 700	BD Biosciences	Cat # 564997; RRID: AB_2869637
Fixable Viability Stain 780	BD Biosciences	Cat # 565388; RRID: AB_2869673
InVivoPlus anti-mouse IFNAR-1	Bio X Cell	Cat # BE0241; RRID: AB_2687723
InVivoPlus anti-mouse IFNγ	Bio X Cell	Cat # BE0055; RRID: AB_1107694
InVivoPlus polyclonal Armenian Hamster IgG	Bio X Cell	Cat # BE0091; RRID: AB_1107773
AffiniPure Goat Anti-Rat IgG (H+L) (polyclonal)	Jackson ImmunoResearch	Cat # 112–005-003; RRID: AB_2338090
TruStain FcX (anti-mouse CD16/32) Antibody	BioLegend	Cat # 101319; RRID: AB_1574973
Bacterial and virus strains
Lymphocytic choriomeningitis virus WE strain	Gift from Marc Pellegrini	N/A
Lymphocytic choriomeningitis virus Docile strain	Gift from Marc Pellegrini	N/A
E. coli (BL21 strain)	Sigma-Aldrich	Cat # CMC0014
Biological samples
Mouse Tissue	N/A	N/A
Chemicals, peptides, and recombinant proteins
5-bromo-2’-deoxyuridine (BrdU)	Sigma-Aldrich	Cat # B5002; CAS: 59–14-3
Favipiravir	Biorbyt	Cat # orb640745; CAS: 259793–96-9
BD Cytofix Fixation Buffer	BD Biosciences	Cat # 554655
Streptavidin-R-Phycoerythrin	Agilent	Cat # PJRS25
DyLight 650 NHS Ester	Thermo Fisher Scientific	Cat # 62265
Kanamycin monosulfate	Gold Biotechnology	Cat # K-120–50
Chloramphenicol	Sigma-Aldrich	Cat # R4408–10ML
Isopropyl β-D-1-thiogalactopyranoside (IPTG)	Gold Biotechnology	Cat # 12481C100
BirA enzyme	Made in-house	N/A
Histopaque-1077	Sigma-Aldrich	Cat # 10771
Red Blood Cell Lysing Buffer Hybri-Max	Sigma-Aldrich	Cat # R7757
Deoxyribonuclease I (DNAse I)	Invitrogen	Cat # D5025; CAS: 9003–98-9
Digitonin (5%)	Thermo Fisher Scientific	Cat # BN2006
Proteinase K	Thermo Fisher Scientific	Cat # AM2542
AMPure XP Beads	Beckman Coulter	Cat # A63881
CD40L	R&D Systems	Cat # 8230-CL-050
Goat anti-mouse F(ab’)2 Ig	Southern Biotech	Cat # 1012–01
IL-21	STEMCELL Technologies	Cat # 78116.1
Critical commercial assays
Illumina Tagment DNA Enzyme and Buffer Small Kit	Illumina	Cat # 20034197
KAPA HiFi HotStart ReadyMix	Roche	Cat # 7958927001
KAPA Library Quantification Complete kit	Roche	Cat # 07960255001
EasySep Mouse PE Positive Selection Kit II	STEMCELL Technologies	Cat # 17666
AEC Peroxidase (HRP) Substrate Kit	Vector Laboratories	Cat # SK-4200; RRID: AB_2336076
His•Bind Purification Kit	Millipore	Cat # 70239
Deposited data
Raw data files for scRNA-seq	This paper	GEO GSE225529
Raw data files for scATAC-seq	This paper	GEO GSE225529
Raw data files for bulk ATAC-seq	This paper	GEO GSE226485
Experimental models: Organisms/strains
C57BL6	In house, Monash Animal Research Platform	N/A
Ifnar−/− bone marrow	Gift from Paul Herzog	N/A
μMT bone marrow	Gift from Daniel Utzschneider	N/A
S1pr2-ERT2cre-tdTomato bone marrow	Gift from Tomohiro Kurosaki	N/A
Recombinant DNA
pET30 Plasmid	Genscript Biotech Corp	N/A
Software and algorithms
GraphPad Prism 8	GraphPad	https://www.graphpad.com/scientific-software/prism/
FlowJo	FlowJo, LLC	https://www.flowjo.com/
ImageJ	National Institutes of Health (NIH)	https://imagej.nih.gov/ij/index.html
Gene Set Enrichment Analysis	Subramanian, Tamayo, et al. (2005, PNAS) and Mootha, Lindgren, et al. (2003, Nature Genetics).	https://www.gsea-msigdb.org/gsea/index.jsp
R	R Project	https://www.R-project.org/
Ingenuity Pathway Analysis	QIAGEN	N/A
Other
HisTrap Fast Flow	Cytiva	Cat # GE17–5255-01
Superdex 200 10/300 GL	Cytiva	Cat # GE17–5175-01
Sephacryl S-300 HR	Cytiva	Cat # GE17–0599-01
Dulbecco’s Phosphate-Buffered Saline (DPBS)	Gibco	Cat # 14190144
RPMI 1640 Medium	Gibco	Cat # 11875093
Dulbecco's Modified Eagle Medium (DMEM)	Gibco	Cat # 11965092
Trypsin-EDTA (0.25%), phenol red	Gibco	Cat # 25200056
GlutaMAX™ Supplement	Gibco	Cat # 35050061
Buffer EB	QIAGEN	Cat # 19086

Experimental model and subject details

Mice

C57Bl/6 mice were bred and housed under specific-pathogen-free conditions at the Monash Animal Research Platform. Bone marrow from S1pr2-ERT2cre-tdTomato mice was kindly provided by T. Kurosaki (Osaka University), bone marrow from μMT mice kindly provided by D. Utzschneider (Peter Doherty Institute), and bone marrow from Ifnar^−/− mice kindly provided by P. Hertzog (Hudson Institute). Both males and females, between 6 and 24 weeks old were used in this study. The Monash Animal Ethics Committee approved all experimental procedures. Mice were culled at indicated time intervals as detailed in the figure legends.

Viruses

LCMV-WE and LCMV-Docile viral strains were originally sourced from M. Pellegrini (WEHI). They were propagated and prepared as previously described^4,104. Briefly, 2×10⁵ cells/ml baby hamster kidney cells were seeded in 20ml DMEM media (10% FCS, 100 U penicillin, 0.1mg/ml streptomycin, 2mM Glutamine). When the cells were approximately 70% confluent, culture media was removed, and cells washed with 10ml of PBS that was pre-warmed at 37°C. Cells were then infected with LCMV-WE at multiplicity of infection (MOI) 0.01 or LCMV-Docile at MOI 0.05 in 5ml DMEM media, incubated at 37°C and another 5ml DMEM media was added to the flask. After at least 48 hours, the supernatant was harvested, filtered and quantified.

Method Details

Viral Infections

C57Bl/6 mice that were approximately 6–12 weeks of age were infected intravenously (i.v.) with either 3 × 10³ PFU of LCMV-WE to induce an acute LCMV infection, or 2 × 10⁶ PFU of LCMV-Docile to induce a chronic LCMV infection.

Mixed Bone Marrow Chimeras:

Bone marrow from μMT and Ifnar^−/− mice, or μMT and wild-type mice, were mixed in an 80:20 ratio and injected i.v. into irradiated (2 × 450Gy) recipient mice. For S1pr2 experiments, bone marrow was sourced from S1pr2-ERT2cre-tdTomato mice. To aid effective reconstitution, S1pr2-ERT2cre-tdTomato bone marrow was mixed with μMT bone marrow and injected i.v. into irradiated (2 × 450Gy) recipient mice. Mice were rested for at least a 6-week reconstitution period before viral infection.

BrdU administration

5-bromo-2’-deoxyuridine (BrdU; Sigma-Aldrich) was diluted to 5mg/ml in sterile PBS. 200uL (1mg) was injected intraperitoneally into each recipient mouse at indicated labelling time points in order to label proliferating cells. PBS alone was used for unlabelled control mice.

Favipiravir treatment

Favipiravir (Biorbyt) was diluted in sterile PBS 10% DMSO to a concentration of 10mg/ml. Mice were administered 100mg/kg/day of the Favipiravir solution by i.p. injection, daily for up to 8 days.

IFN blocking antibody treatment

Mice were treated initially with 1mg and then 500μg every 3 days of BioXcell InVivoMAb anti-mouse IFNγ (cat: BE0055, clone: XMG1.2), InVivoMAb anti-mouse IFNAR-1 (cat: BE0241, clone: MAR1–5A3) or InVivoMAb polyclonal Armenian Hamster IgG isotype control (cat: BE0091) for up to 2 weeks during indicated treatment windows.

Tetramer generation and conjugation

A decoy tetramer was used to gate out cells binding to irrelevant tetramer components. The decoy tetramer consisted of DyLight 650 (DL650) conjugated to SA-PE loaded with an irrelevant protein, specifically biotinylated MHC-peptide (HLA-A2-M1).

Biotinylated protein production

Recombinant protein was expressed and produced by Escherichia coli (E.coli; BL21 strain). Transfected E. coli bacteria were selected in kanamycin (pET30 plasmid) or chloramphenicol containing medium overnight at 37°C, and then passaged to cultures at 1:50 and incubated at 37°C to reach an OD600nm of ~0.6. Isopropyl β-D-1-thiogalactopyranoside (IPTG; Sigma-Aldrich) was added to cultures prior to further incubation at 37°C. Cells were centrifuged, lysed with lysozyme (10 mg per litre of expression) in the presence of DNAse. Supernatant was separated via centrifugation. NP peptidase was purified either by passing supernatant over a HisTrap Fast Flow column (Cytiva) or using a His-Bind purificatioin kit (EMD Chemicals). Purified protein was separated from E. coli proteins using a size exclusion column (Cytiva or GE Healthcare). For biotinylation of the protein, the recombinant protein was incubated with BirA enzyme and reaction buffer (50 mM Bicine pH 8.3, 10mM ATP, 10 mM magnesium acetate (MgOAc), and 50μM d-Biotin) at 4°C overnight. 1mg of protein per 10μg of enzyme was used and biotinylation efficiency was quantified using gel electrophoresis.

Tetramer conjugation

The Decoy tetramer was prepared by conjugating the core fluorochrome SA-PE to DL650 (Sigma-Aldrich) according to the manufacturer’s protocol for 60 minutes at room temperature. The free DL650 was removed by centrifugation in a 100-kD molecular weight cut off Amicon Ultra filter (Millipore). The SA-PE*DL650 complex concentration was calculated by measuring the absorbance of PE at 566nm using a NanoDrop ND-1000 spectrophotometer (Thermo Fisher Scientific). The SA-PE*DL650 complex was then incubated with 10-fold molar excess of biotinylated HLA-A2-M1 for 30 minutes at room temperature. The solution was diluted to 1μM based on the absorbance of PE at 566 nm (divided by the extinction coefficient = 1.96 cm−1μM−1). The LCMV-specific antigen tetramer was prepared by conjugating the biotinylated LCMV-NP to SA-PE (Prozyme PJRS25) at a concentration ratio of 4:1 respectively. Following addition of SAPE to the biotinylated LCMV-NP, the mixture was first incubated in the dark at room temperature for 3 hours, then at 4°C overnight. The tetramer fraction was centrifuged in a 100-kD molecular weight cutoff Amicon Ultra filter. The concentration of tetramer was calculated by measuring the absorbance of PE at 566 nM (divided by the extinction coefficient = 1.96 cm−1μM−1).

Flow cytometric analysis and Fluorescence-activated cell sorting (FACS)

Single cell suspensions were isolated from each spleen. To collect peripheral blood mononuclear cells (PBMCs), needles were coated with heparin (StemCell) prior to cardiac puncture. Blood was layered onto an equal volume of Histopaque 1077 (Sigma-Aldrich) and centrifuged at 400g for 30 minutes at room temperature without brakes.

Antigen-specific B cell enrichment:

The EasySep Mouse PE positive selection kit II (Stemcell) was used to enrich antigen-specific B cells using the manufacturer’s protocol. Enriched cells were filtered and resuspended in 1ml PBS 2% FCS. 2–5 × 10⁶ cells were added to FACS tubes and centrifuged (1500 RPM, 5 mins, 4°C). Cells were first stained with viability dye, Fixable Viability Stain 700 (FVS700 - BD Biosciences 564997) or FVS780 (BD Biosciences 565388). Cells were washed and resuspended in the relevant surface antibody staining cocktail. Following antibody staining, LCMV-infected samples were fixed using BD Cytofix according to the manufacturer’s protocol. Samples were analysed using a BD-LSRFortessa X-20 flow cytometer with FACSDiva software (BD Biosciences). FCS files were analysed using FlowJo software (FlowJo, LLC). Cell cycle analysis: 5-Bromo-2’-deoxyuridine (BrdU, Sigma Aldrich) staining was performed as described previously¹⁰⁵.

In vitro cell culture

Splenic Ly6c⁺B220⁺IgD^loCD38⁺CD138⁻ and Ly6c⁻B220⁺IgD^loCD38⁺CD138⁻ cells were sort-purified from mice infected with LCMV-WE and LCMV-Docile 4 weeks prior. 5000 cells per well were stimulated with 50 ng/mL CD40L (R&D Systems), 1.25 μg/mL goat anti-mouse F(ab’)2 (Southern Biotech) and 50 ng/mL IL-21 (STEMCELL Technologies) for 4 days. 10⁴ Calibrite beads (BD) were added to each well immediately before harvesting to enable enumeration of cells. Cells were harvested and stained for flow cytometry. Plasmablasts were defined as CD138⁺ B220^lo/int and the total number per well was calculated based on the known number of beads added per well.

Assessment of viremia

The right lobe of the liver was collected from experimental mice. Livers were weighed and cDMEM was added to obtain 500 mg/ml of liver suspension. One stainless steel bead (Qiagen) was added per sample and liver tissues were homogenized using the TissueLyser LT (Qiagen) for 5 minutes at 50 oscillations/second. The beads were removed, and samples were centrifuged. The supernatant was collected for each sample and 10-fold serial dilutions were added to a 96-well culture plate (Falcon). A plaque forming assay was carried out as described previously⁴.

scRNA-seq

Multiplexing samples with hashtag antibodies:

100,000 cells were sort-purified per mouse. Cells were resuspended in washing buffer (PBS 0.04% BSA) and incubated with mouse Fc blockers (FcX, Biolegend). Each sample was then incubated with either TotalSeq anti-mouse Hashtag Antibody 1 (Biolegend #155861) or 2 (Biolegend #155863) for 20 minutes at 4°C. Cells were washed 3 times with 1ml of washing buffer to remove unbound antibodies. Following the final wash, one LCMV-WE and one LCMV-Docile sample were pooled together at equal ratios in 1ml washing buffer. Pooled samples were then resuspended in PBS 0.04% BSA 0.2 U/ml RNAse inhibitor.

Droplet-based scRNA-seq:

Reverse transcription, cDNA amplification and library preparation for single cell gene expression and V(D)J clonotypes were performed based on the manufacturer’s protocol using the Chromium Single Cell 5′ Library & Gel Bead Kit v3 (10X Genomics, Pleasanton, CA). Libraries were pooled at equimolar ratios and gene expression libraries were sequenced on the Illumina NovaSeq 6000 System, while HTO (hashtag oligo) and V(D)J (BCR) libraries were sequenced on Illumina NextSeq 500.

Raw data processing:

Cellranger (v3.1.0) was used to align the raw scRNAseq data to the mouse reference genome (mm10–3.0.0). Cellranger count was run under default parameters. Cellranger VDJ (v3.1.0) was run under default parameters on the VDJ libraries to assemble and annotate VDJ sequences using the VDJ compatible reference (cellranger-vdj- GRCm38-alts-ensembl-3.1.0). To process the CITE-seq data, CITE-seq count (v1.4.3) was run with the following parameters (‘--sliding window -cbf 1 -cbl 16 -umif 17 -umil 26 -cells $(number_submitted_cells)’). Each sample was processed individually with each pipeline.

Analysis of scRNA-seq data:

The Seurat package (v3.1.4) was used with R (v3.6)¹⁰⁶. Count matrices from the CITE-seq pipeline and cellranger count were loaded into R and Seurat was used to demultiplex the data with the HTODemux function^107,108. Cells that lacked a corresponding hashtag were discarded. The two samples per condition were merged after demultiplexing. 7199 cells were then stringently filtered for a minimum number of 1000 UMI, at least 500 genes detected and less than 5% mitochondrial reads for a cell to be kept. All doublets and empty cells were filtered out of the data, and a total of 2795 cells (acute condition) and 2591 cells (chronic condition) were retained. The data was scaled, log normalised and the number of principle components (PC) to be used for dimensionality reduction was determined using an elbow-plot. The first 10 PCs were selected, though the inclusion of more PCs was examined and found to have no change on the UMAP generated. The FindClusters function was used to cluster the data and was run with several different resolutions before settling on a resolution of 0.4, all other parameters were set to default. For all differential tests, the Mann-Whitney U test was used, and only positive markers were returned.

IPA:

Genes from scRNAseq datasets with a log fold change (FC) ≥2 with a false discovery rate (FDR, adjusted P-value) of <0.05 were defined as differentially expressed genes (DEGs) and assessed using IPA software (Ingenuity Systems, QIAGEN). DEG data were uploaded to IPA for core analysis and analysed with the global molecular network in the Ingenuity Pathways Knowledge Base (IPKB). IPA identified canonical pathways, diseases and functions and molecular networks enriched by differentially expressed genes (DEGs)¹⁰⁹.

Gene set enrichment analysis (GSEA):

Gene expression profiles were generated using Loupe Browser (v5.0 – 10X Genomics) from LCMV-WE and LCMV-Docile samples. Genes were ordered in a ranked list based on differential expression between conditions. Gene set enrichment analysis (GSEA) (v4.1.0 - Broad Institute) was conducted with Signal2Noise values to determine where a priori sets of genes were distributed throughout the ranked list¹¹⁰. Genes related to a previously published phenotypic distinction will distribute at the top or bottom of the ranked list. Gene matrix files were obtained from MSigDB (Molecular Signature Database)¹¹¹. The number of permutations was 1000, and the permutation type was set to gene set with FDR q-value less than 0.05.

Processing of single cell (V(D)J) datasets:

FASTQ files for Chromium Single Cell V(D)J libraries were processed with 10x Genomics Cell Ranger 6.0.2 using the mouse VDJ reference (refdata-cellranger-vdj-GRCm38-alts-ensembl-5.0.0). Contigs generated by Cell Ranger (filtered_contigs.fasta) were post-processed with stand-alone IgBLAST (version 1.14.0) [doi: https://doi.org/10.1093/nar/gkt382] against the mouse V, D and J germline datasets obtained from the IMGT Reference Directories [website: https://www.imgt.org/vquest/refseqh.html] (accessed 16-Jan-2020) to produce tab-delimited output files (--outfmt 19).

Cell Ranger and IgBLAST output were merged with a perl script to provide a single line summary of heavy and light chains for each cell barcode. Where multiple chains were present at a single locus (i.e. IGH or IGK/L) the chain with the highest UMI count was retained and the presence of an additional chain was noted. VDJ data was merged by cell barcodes with single cell gene expression (GE) analysis in R version 4.2.1¹¹² using the tidyverse package¹¹³. R was used within RStudio IDE¹¹⁴ and additional packages for data manipulation, analysis and visualisation included ggsci (colour palettes)¹¹⁵ and ggpubr (figure panel layouts)¹¹⁶.

Analysis for immunoglobulin repertoire features was restricted to cells with paired heavy and light chains. V gene somatic hypermutation (SHM) was derived from the IgBLAST output using the v_identity field (100 – v_identity). Isotype usage was determined from the Cell Ranger c_gene field for IGH chains with all IgG subclasses (IGHG1, IGHG3, IGHG2B/C) combined to a single IgG group and IgM sequences were split into ‘mutated’ (> 0% SHM) and ‘unmutated’ (0% SHM).

scATAC-seq

Droplet-based scATAC-seq:

100,000 cells were sort-purified per mouse. Nuclei isolation, barcoding and library preparation were performed according to the manufacturer’s protocol using the Chromium Next GEM Single Cell ATAC Reagent Kit v1.1 (10X Genomics). Libraries were prepared using the MGIEasy v3 chemistry kit and sequenced using the MGITech MGISEQ2000RS sequencer (MGI Tech).

Raw data processing:

The raw sequence data was processed using cellranger-atac-2.0.0 for each sample against mm10, then aggregated using cellranger with normalize=none.

Analysis of scRNA-seq data:

The aggregated data was loaded into R (v4.1.0) and analysed using Seurat (v4.1.0)¹¹⁷, and Signac (v1.5.0)¹¹⁸ QC filtering for peaks and cells was performed using these criteria: peaks present in at least 10 cells, cells that contained at least 200 called peaks and at least 10,000 fragments in peak regions. The standard Signac processing was performed, using dimension reduction with the LSI technique - the first dimension captured sequencing depth variation so was omitted from clustering and UMAP projection. Clustering was computed across several resolutions, and a resolution setting of 0.4 was chosen to give a good, stable clustering of 10 clusters. Clusters identified to contain contaminating cells were filtered out leaving a total of 7 clusters. Differentially accessible regions were found using the FindMarkers function from Seurat using the logistic regression (lr) method, testing regions that had detected fragments in at least 5% of cells in either cluster. The called peaks were then annotated using ChIPseeker¹¹⁹ to label where they occurred relative to the genes in the mm10 reference. This annotation per-peak was then broken down into per-cluster, and per-condition. ChromVAR⁵⁴ was used to calculate significant differential motif activity between clusters. The motifs used were from the JASPAR 2020 database of transcription factor motifs¹²⁰. First motif activity was calculated for each motif in each cell using the runChromVar function, then differential motif activity was determined using the FindMarkers function for a specific cluster compared to all other cells. Peaks found to be differentially accessible using the method described above were then selected to test for motif enrichment. Motif enrichment was tested using the findMotifsGenome method from HOMER⁵³ which tests motifs found in the peaks against a background selected from the genome to balance for GC bias. The scRNA-seq cluster labels were transferred over to the scATAC-seq data using a standard approach. The “Gene Activity” for each gene, and each cell in the ATAC data was calculated based on the called peaks falling within genes. The FindTransferAnchors function was then used to identify mutual nearest neighbours in a projected lower dimensional space between the two datasets, then these anchor cells were used to transfer the scRNA-seq cluster labels to the nearby cells in the scATAC-seq dataset.

Bulk ATAC-seq

Approximately 15,000 cells were pelleted and washed with PBS. One-step permeabilization and tagmentation method was used by resuspending the cell pellets in Digitonin (EZSolution), Tween-20 (Sigma) buffer. High-molecular-weight tagmented DNA was then removed by incubating samples with a 0.7X ratio of Agencourt AMPureXP beads (Invitrogen). The same process was applied using a 1.2X ratio of beads to exclude low-molecular-weight DNA. Purified DNA was then pre-amplified with indexed primers and HiFi HotStart Polymerase Ready Mix (KAPA Biosystems). After estimation of additional PCR cycles required¹²¹, the genomic library was fully amplified using the SYBR qPCR Master Mix with primers included in the Illumina Library Quantification kit for Bio-Rad iCycler (KAPA Biosystems). Libraries were sequenced on the Illumina NextSeq 2000 using 60-nt paired-end Illumina chemistry. Analysis was performed as described previously⁴.

Statistical analysis

Experimental data are presented as mean ± standard error of the mean (SEM) with statistical analysis performed using GraphPad Prism 8 (GraphPad Software). When comparing groups, between-group differences were analysed using the Mann-Whitney non-parametric two-tailed test with a 95% confidence interval.

Supplementary Material

2

Table S1. Genes lists used to compare to transcriptomic differences of antigen-specific B cells responding to acute versus chronic LCMV infection, related to Figures 2, S2 and S3.

3

Table S2. Genes expressed by T-bet-related scRNA-seq cluster, related to Figures 2 and S4F.

4

Table S3. Distinct epigenetic and transcriptomic changes induced by chronic LCMV infection, related to Figure 3.

5

Table S4. chromVar analyses of transcription factor binding motifs, related to Figures 3, 5, 7 and S5A.

6

Table S5. Transcription Factor Motifs in differential chromatin accessibility regions post-IFNAR-1 blockade, related to Figures 5 and 7.

HIGHLIGHTS.

• Single cell ATAC & RNA-seq reveal expansion of an MBC subset in chronic LCMV infection

• Distinct chronic MBC subset is associated with increased IFN-I-associated ISG signature

• Chromatin landscape of MBCs is established during a critical window early in infection

• IFN-I dynamics govern memory B cell epigenome and phenotype in chronic viral infection