IgM, IgG, and IgA influenza-specific plasma cells express divergent transcriptomes

Abstract

Antibody secreting cells (ASC) or plasma cells are essential components of the humoral immune system. While antibodies of different isotypes have distinct functions, it is not known if the ASC that secrete each isotype are also distinct. ASC downregulate their surface BCR upon differentiation, hindering analyses that couple BCR information to other molecular characteristics. Here, we developed a methodology using fixation, permeabilization, and intracellular staining coupled with cell sorting and reversal of the crosslinks to allow RNA-seq of isolated cell subsets. Using hemagglutinin (HA) and nucleoprotein (NP) antigen-specific B-cell tetramers and intracellular staining for IgM, IgG, and IgA isotypes, we were able to derive and compare the gene expression programs of ASC subsets that were responding to the same antigens following influenza infection in mice. Intriguingly, while a shared ASC signature was identified, each ASC isotype-specific population expressed distinct transcriptional programs controlling cellular homing, metabolism, and potential effector functions. Additionally, we extracted and compared BCR clonotypes and found that each ASC isotype contained a unique, clonally related CDR3 repertoire. In summary, these data reveal specific complexities in the transcriptional programming of antigen-specific ASC populations.

INTRODUCTION

The immune system is composed of a diverse and complex mixture of specialized cells that protect the host from invading pathogens and facilitate organism homeostasis. Using tools that integrate expression of various surface markers, immune cells can be defined and subdivided based on cell type and/or specialized functions. It is increasingly apparent that phenotyping based on intracellular markers can further specify functionally significant cell populations. For example, CD4 T helper cell subsets are identified by lineage-defining transcription factor expression (1–3); innate lymphoid cells are defined by transcription factors and/or cytokine production patterns (4, 5); and antibody secreting cells (ASC) or plasma cells are classified based on BCR isotypes (6). Additionally, fluorescent dyes can be used to track DNA content, organelle function and number, as well as metabolic status. The identification of intracellular targets for FACS is routinely accomplished by fixation using a formaldehyde solution followed by permeabilization (7, 8). An advantage to fixation-based phenotyping is the inactivation of pathogens that might be present in blood or tissue samples, as well as the stabilization of cellular states over time. When fixation is not possible, surrogate surface markers for intracellular factors can be used. For example, CXCR3 surface expression can be used to identify cells that express the transcription factor Tbx21 (T-BET) (9–11). However, surrogate markers are imperfect and given the widespread use of intracellular assays, there is a need to improve and validate the limited techniques that facilitate the reversal of fixation for downstream readouts such as global gene expression (12, 13).

Upon differentiation of B cells to ASC, the BCR transitions from membrane bound to a secreted form (14–16). For this reason, ASC are particularly challenging to phenotype and are largely identified based on surrogate markers such as CD138 (Syndecan-1) for both mouse and human ASC (17, 18), as well as TACI or SCA-1 (19, 20). The genetic knock in of the GFP fluorescent reporter into the Prdm1 locus has also allowed for the identification of Blimp-1 expressing cells (i.e., ASC) without the need for intracellular staining (21). Using combinations of these markers, RNA-seq based studies have identified ASC-specific gene expression signatures (22). However, none of these markers provide information about the BCR antigenic target or whether the transcriptional signatures are shared among ASC of distinct BCR isotypes. We therefore developed a fixation and staining protocol to identify influenza-responding and specific ASC of distinct BCR subtypes, followed by fixation reversal and isolation of RNA for deep sequencing. Isotype-specific ASC demonstrated unique expression patterns of key signature genes that regulate BCR class-switching and homing to distinct tissues. Furthermore, we were able to extract BCR clonal frequencies and identify VDJ combinations and CDR3 sequences that were specific to each ASC isotype. These data define an approach for transcriptome profiling based on intracellular phenotypes and identify unique features of ASC subsets that correlate with functional differences.

MATERIALS AND METHODS

Cell culture

Raji human Burkitt’s lymphoma cell line was purchased from the American Type Tissue Collection (ATCC, CCL-86) and cultured in RPMI containing 10% FBS and 100 U/ml penicillin and streptomycin.

Mice and influenza infection

C57/BL6J mice were 8–12 weeks of age and infected with 15000 vfu influenza A/PR8/34 (23). Spleen and dLN (mediastinal-lung draining lymph node) were analyzed 14 d after infection. All animal protocols were approved by the Emory Institutional Animal Care and Use Committee.

Staining, Fixation, and Sorting

Splenocytes were washed and resuspended at 25 × 10⁷ cells/ml in PBS with 1% BSA and 2 mM EDTA (FACS). Cells were stained with anti-CD138-APC (Biolegend, clone 281–2) at 4° for 30 mins. Cells were then washed with 10 ml MACS buffer (1× PBS, 0.5% BSA, 2 mM EDTA) and incubated with 35 μl anti-APC microbeads (Miltenyi 130–090-855) at 4° for 15 mins. Cells were washed with 10 ml PBS + 0.5% BSA and 2mM EDTA and run over a magnetic column, as per manufacturer’s instructions. Enriched cells were resuspended in 100 μl FACS + surface antibody cocktail for 30 mins on ice. Surface stains were as follows: Biolegend - Zombie Yellow Viability dye (77168), CD11b APC-Cy7 (clone M1/70), Thy½ APC-Cy7 (clone 30-H12), F4/80 APC-Cy7 (clone BM8), CD98 PE-Cy7 (clone RL388), CD36 PE-Cy7 (clone HM36); Tonbo - B220 PE-Cy7 (clone RA3–6B2); Invitrogen LAG-3 PE-Cy7 (clone eBioC9B7W). Cells were washed in 1ml FACS before fixation and permeabilization with BD Cytofix/Cytoperm (554714) as per manufacturer’s instructions. Intracellular staining mix included 20 minutes of staining first with a decoy reagent (Streptavidin- PE-AF647, Invitrogen S20992) to eliminate intracellular streptavidin:biotin interactions with the tetramers. Then, cells were washed and incubated with HA-PE and NP-PE B cell tetramers (24); Invitrogen - IgM v450 (clone eB121–15F9), IgA FITC (clone mA-6E1); and Biolegend - IgG PerCp-Cy5.5 (clone Poly4053), BCL-2 PE-Cy7 (clone Bcl/10C4). Cells were sorted using a FACSAria II directly into 300 μl RLT buffer (Qiagen) with 1% 2-mercaptoethanol, snap frozen, and stored at −80°C. FCS files were analyzed using FlowJo (v10).

Decrosslinking and RNA-seq

To reverse the formaldehyde crosslinking, the samples in RLT buffer (Qiagen, Inc.) were heated to 65°C for the indicated amount of time, total RNA purified using the Quick-RNA MicroPrep kit (Zymo Research), and eluted in 10 μl. RNA-seq libraries were generated using the SMARTer Stranded Total RNA-Seq kit v2 – Pico input Mammalian (Takara) according to the manufacturer’s instructions and sequencing on a NextSeq500 using paired-end 75 base chemistry at the New York University Genome Technology Center and Emory Integrated Genomics Core.

RNA-seq analysis

Reads were mapped to the mm10 version of the mouse genome using STAR (25) and processed as previously described (26). Genes were filtered for detection based on being expressed at >= 10 reads per million in all samples of one group (i.e., IgM, IgG, or IgA) resulting in 8,766 genes. Differential expression between ASC subsets was determined using edgeR v3.18.1 (27) and an FDR < 0.05 and log₂FC > 1. Principal component analysis was performed using the vegan package v2.4–3 (http://vegan.r-forge.r-project.org/). To determine Igh expression the immunoglobin gene segments were downloaded from the IMGT (28) and expression annotated using the GenomicRanges v1.30.3 (29). All RNA-data is available from the NCBI Gene Expression Omnibus (GEO) (https://www.ncbi.nlm.nih.gov/geo/) under accession GSE124578.

Repertoire analysis

Clonotype information was extracted for each sample with MiXCR v2.1.12 (30) using the ‘align -p rna-seq -s mmu -OallowPartialAlignments=true’ options. Aligned clones were exported using the ‘exportClones -c IGH -o -t’ options. Clone information was analyzed using VDJtools (31) and custom R/Bioconductor scripts.

RESULTS

Optimization of RNA isolation and sequencing from formaldehyde fixed cells

To develop methods that allow the transcriptome to be interrogated from cells that have been fixed and permeabilized, we first determined the optimal conditions for reversing the crosslinking. Nucleic acid fixation with formaldehyde is reversible by incubation at 65°C (32). We first tested the size distribution and quantity of RNA obtained following incubation of 50,000 Raji B cells that were fixed and permeabilized using a formaldehyde-based protocol (See Materials and Methods). Incubation at 65°C for 0, 5, 15, and 30 min followed by RNA purification resulted in increasing yield of total RNA (Fig. 1A). Closer assessment of the RNA size distribution revealed an enrichment of small RNA fragments that likely represent fragmentation of RNA from the fixation and heating process (Fig. 1B). Low RNA yields were obtained from samples that were not subjected to incubation at 65°C, indicating the heating process was necessary to release the RNA from fixation. Thus, we demonstrated that RNA can be successfully isolated from cells that had undergone fixation and permeabilization.

Figure 1. Optimized RNA-seq libraries following fixation and intracellular staining.

(A) Concentration of purified RNA from formaldehyde fixed Raji B cells following incubation at 65°C for the indicated number of minutes. (B) Size distribution of purified RNA from A. The location of the standard in each sample and size in bp is indicated. FU, fluorescent units. (C) Scatterplot of normalized expression levels for 13,545 detected genes in each sample. The Pearson’s correlation r value is indicated for each comparison. FPKM, fragments per kilobase per million.

To further determine the efficiency of decrosslinking, RNA-seq libraries were generated from Raji cells from each of the three conditions tested above, as well as unfixed cells as a control. Consistent with unfixed cells, we detected the expression of 13,545 genes (RPM > 3) in each of the conditions. Importantly, an almost perfect correlation was observed in the expression levels of genes between each of the fixed samples (Fig. 1C). Comparison of the 30 min fixed to unfixed cells revealed a high correlation of gene expression that was the same for all the other fixed samples (data not shown). These data demonstrate the feasibility of performing global transcriptome profiling by RNA-seq from cells that have undergone fixation and permeabilization protocols that are based on formaldehyde.

BCR isotype subsetting of ASC by intracellular staining following influenza infection

Upon differentiation, ASC use alternative mRNA processing to convert the sameBCR to a secreted isoform (14–16). Therefore, ASC have low to undetectable surface staining with anti-isotype antibodies and B cell tetramers (33–35). To directly phenotype isotype-specific ASC we used a model of influenza in which mice were infected intranasally with influenza strain A/PR8/34 (PR8). At day 14 post-infection, all CD138⁺ cells from the spleen and lung draining lymph node were magnetically enriched. (Fig. 2A). Following enrichment, cells were first stained with extracellular surface markers for phenotyping followed by fixation and intracellular staining using B cell tetramers and antibodies against distinct BCR isotypes. After intracellular staining, cells were washed and phenotyped by flow cytometry. Analysis of ASC by isotype revealed distinct populations representing IgM, IgG, and IgA. The intracellular staining provided a clear separation between isotypes (Fig. 2B–E). Although a B cell tetramer was used in the initial gating strategy, non-specific intracellular interactions, which were not present in non-B cell lineage cells (Fig. 2C), precluded the analysis of antigen specificity without decoy-tetramer reagents (see below). Therefore, antigen specificity was excluded from the analysis of these cell populations. The relative frequency of total cells expressing IgM, IgG, and IgA varied somewhat between mice (Fig. 2F). The early emergence of IgM ASC has been reported for model antigens and IgG ASC tend to accumulate later in infection (36). However, the kinetics of IgA ASC are not as clearly defined. These results demonstrate that single isotype ASC can be readily identified.

Figure 2. Identification and phenotyping of isotype-specific ASC.

(A) Schematic detailing the experimental procedure used to identify, fix, sort, reverse crosslinks and sequence the cells of interest. (B) Flow cytometry gating strategy for identification of ASCs from the spleen and dLN at d14 post infection. Cells were enriched on CD138-APC, stained with surface markers, and fixed/permeabilized for intracellular staining. (C) Flow cytometry gating of PE-HA and PE-NP tetramers in non-B cell lineage cells from B. (D) Sorted ASC populations from B (colored gates) were overlaid to show staining relative to each isotype. (E) Overlaid histograms of the gate drawn for each subset relative to other isotypes from B. (F) Summary of the relative frequency of each isotype from B for three independent mice.

Antigen-specific ASC show isotype-specific gene expression profiles

Using the staining and gating strategy detailed above, we FACS isolated ASC that intracellularly expressed either IgM, IgG, or IgA. Following sorting, formaldehyde crosslinking was reversed by treating for 15 min at 65°C. The 15 min time point was chosen as a balance of maintaining a short RNA extraction workflow with enough time to adequately reverse the crosslinking (Fig. 1). The resulting RNA was purified and RNA-seq performed to analyze the transcriptomes of the ASC. Overall, we detected the expression of 8,766 genes in at least one ASC isotype. Principal component analysis (PCA) of all detected genes demonstrated clear separation of ASC by isotype, suggesting that unique transcriptional differences were present (Fig. 3A). Pairwise differential analysis between isotypes identified 1,438 significantly differentially expressed genes (DEG) (FDR < 0.05 and log₂FC > 1) (Fig. 3B, Supplemental Table 1). Consistent with the PCA clustering, IgG and IgM ASC were the most distinct. Interestingly, IgA ASC expressed a set of genes that were shared with IgM or IgG. Therefore, ASC of distinct isotypes have both shared and unique gene expression programs.

Figure 3. Distinct ASC isotypes contain unique transcriptomes.

(A) Principal component analysis (PCA) of 8766 detected genes in each population. Percent of variation covered indicated on axes with circles denoting 99% confidence intervals. (B) Heatmap of 1438 differentially expressed genes (DEGs). (C) Volcano plots for each pair-wise comparison. Number of DEG in each comparison is indicated. (D) Bar plots showing expression of the indicated gene in each population. (E) Overlap of bone marrow ASC signature genes (22) and ASC isotype dataset described here. (F) Heatmap of ASC overlap genes from E that are differentially expressed between any two isotype subsets. (G) Example bar plots of DEG from the genes displayed in F. Samples represent ASC subsets from three independent mice.

Further analysis of the genes that were expressed in an isotype-specific manner revealed important functional differences between the subsets. IgA ASC expressed high levels of Ccr9, Ccr10, and Itgb7 (Fig. 3D). Each of these three genes are important for homing to mucosal sites. CCR9 binds to CCL25, which is constitutively expressed by the cells of the intestinal epithelium (37, 38). ITGB7 can heterodimerize with integrin alpha-4 to make α₄β₇, which together with CCR9/CCL25 interactions strengthen gut homing signals (39, 40). CCR10 can bind to CCL27 and CCL28, which canonically directs skin homing, but has also been shown to position IgA ASC in the gut lamina propria (41). IgM and IgG ASC expressed significantly higher levels of Cxcr4 (the chemokine receptor responsive to CXCL12) than IgA ASC. CXCL12 is prevalent in the bone marrow and is pivotal in establishing the bone marrow microniche for long-lived ASC (42, 43). IgG ASC expressed Cxcr3, which serves to direct antigen-specific cells to the lung during influenza infection (44).

Additionally, we compared our ASC expression profiles to a previously defined bone-marrow ASC gene expression signature (22) to determine the overlap of ASC from the periphery with those of the bone marrow. Of the reported ASC signature genes, 96% (289/301) were also expressed in IgM, IgA, and IgG ASC (Fig. 3E). These data show an overlapping gene expression signature for ASC in the periphery and in the bone marrow. We also determined whether ASC signature genes were similarly expressed between isotypes and found 15% (43/289) of the overlapping genes were DEG between at least two of the comparisons (Fig. 3F). Many of the gene expression differences were consistent with homing properties of cells of each isotype. As mentioned above, IgA ASC upregulated Ccr9, Ccr10, and Itgb7, which facilitate homing to mucosal sites. Additionally, IgA ASC expressed high levels of Tgfbr1, which can induce switching to IgA (45). Furthermore, Tnfrsf17, which encodes BCMA, was more highly expressed in IgM and IgA ASC; and Ada, which encodes adenosine deaminase, an enzyme involved in purine metabolism, was more highly expressed in IgG ASC (Fig. 3G). Thus, each ASC isotype has a distinct transcriptional signature that reflects their developmental history, homing, and specialization potential and reinforces the importance of identifying the transcriptomes of more highly subdivided ASC.

Influenza-specific ASC display similar gene expression profiles to bulk ASC

The development of B cell tetramers facilitates the tracking of antigen-specific B cells during immune responses (24, 36). We sought to compare the gene expression changes observed in isotype-specific ASC to isotype and antigen-specific ASC. Therefore, mice were infected with PR8, enriched, and stained as detailed above (Fig. 2A). In addition to the staining panel above, we intracellularly stained with a decoy streptavidin phycoerythrin (PE)-Alexa647 reagent, as well as phycoerythrin (PE) labeled hemagglutinin (HA) and nucleoprotein (NP) B-cell tetramers that are specific for PR8 (H1N1) (24). This allowed the exclusion of cells from the analysis that bound nonspecifically to the biotin, streptavidin, or the PE fluorophore present in the tetramers (Fig. 4A). There were fewer overall ASCs identified in naïve mice than d14 PR8-infected mice; and importantly, no tetramer-positive cells were present in naïve mice. As anticipated, naïve mice have a large proportion of IgM ASC, and this frequency is decreased in PR8-infected mice, signifying the expansion of class-switched cells following infection. At d14 post infection, ~15% of cells remained IgM+, up to 60% of ASC expressed IgG, and ~5% were IgA+ from the spleen and dLN (Fig. 4B).

Figure 4. Influenza-specific ASC show complementary transcriptomes to bulk ASC.

(A) Flow cytometry gating strategy for identification of influenza-specific ASCs from the spleen and dLN at d14 post infection. Cells were enriched on CD138-APC, stained with surface markers, and fixed/permeabilized for intracellular staining. Naïve mice are used as a staining control and to draw gates for tetramer binding cells. (B) Summary of ASC from naïve (n=5) and influenza infected (n=10) mice, displayed as the frequency of parent gate. ASC are from the total CD11b, F4/80, Thy1.2- (Fig. 2B, panel 1); Tet (Tetramer+) are of the total ASC; IgM, IgG, IgA are of total Tet+ ASC. (C) PCA of 13665 genes identified by RNA-seq in cells colored in A from two independent mice. (D) Mean fluorescence intensity (MFI) of CD36, BCL-2, LAG-3, and CD98 on total CD138+ cells and Tetramer+ isotype-specific cells as indicated. (E) Bar plot summarizing MFI data from D. Data is representative of 10 mice from two independent groups. (F) Bar plots displaying mRNA expression in FPKM for each gene. Influenza responding ASC (Fig. 3) are present in closed circles. Tetramer+Decoy- ASC are included in open circles.

We performed RNA-sequencing on IgA, IgG, and IgM populations of influenza-specific ASC. Similar to bulk ASC, PCA of all detected genes showed influenza-specific IgM and IgA to have transcriptomes more closely related to each other than to IgG (Fig. 4C). The expression of specific DEGs identified between the three subsets were validated by flow cytometry (Fig. 4D, 4E) and compared to expression levels derived by RNA-seq (Fig. 4F). CD98, which is a protein composed of two subunits Slc3a2 and Slc7a5, was highly expressed on IgM and IgA ASC but was reduced on IgG expressing ASC. CD98 is a neutral amino acid transporter which has been shown to correlate with Ki67 expression in multiple myeloma plasma cells (46). Cd36 was upregulated in IgM ASC and has been reported previously to be highly induced upon stimulation, but downregulated in class-switched ASC in vivo (47). Anti-apoptotic protein BCL-2 was more highly expressed on IgM and IgA ASC. BCL-2 had previously been reported to be expressed on normal and malignant plasma cells with varying intensity (48). Its variable expression between the three isotypes suggests that the variations in intensity seen previously may be reflective a mixture of various isotype classes. Lag3 expression was identified specifically on IgM ASC in the periphery at d14; however, Il10 expression, which has been shown to correlate with LAG-3 protein expression was not detected at this time point (49).

Sorted ASC populations are enriched for the target isotype

To analyze the purity of our isolation strategy we used two complementary approaches. First, using the IMGT database of mouse BCR gene segment locations (28), we computed the expression levels of all Igh constant regions and summarized the percentage of total reads that mapped to each region for the samples in each isotype. This approach demonstrated that 80, 95.5, and 90% of the Igh mRNA for IgM, IgG, and IgA ASC mapped to the Ighm Ighg, and Igha segments, respectively (Fig 5A). This demonstrates that each isolated ASC isotype predominantly expressed Igh mRNA specific to the enriched population. As a complementary approach, we used the MiXCR software tool to extract enriched BCR clones from the RNA-seq reads (30). MiXCR successfully extracted the CDR3 sequences containing the VDJ junctions; and for a subset of the clones, identified the associated constant region. The Igh constant region usage for all clones in each ASC isotype was determined and summarized as above for each ASC isotype. Analyzing rearranged clones showed an even higher specificity with 90% of IgM, 99% of IgG, and 96.1% of IgA clones representing the target isotype (Fig 5B). These data confirm the specificity of the target ASC populations.

Figure 5. Flow cytometry accurately captures specific ASC populations.

Pie charts showing the percentage of FPKM normalized RNA-seq mapped reads (A) or MiXCR extracted clones (B) that map to each Igh constant region for the indicated ASC isotype. Data represents the total from each ASC isotype group. For IgM and IgA ASC groups all Ighg isotypes are summarized for simplicity.

ASC isotypes have distinct repertoires

Using the VDJ and CDR3 sequences extracted from each ASC population, we analyzed the repertoire of each ASC isotype to determine clonality within and between isotypes. The preference for specific V-J combinations was determined and revealed that overall, the ASC clones profiled predominantly used the J-2 segment (Fig 6A). However, the second most common J region differed. IgG and IgA ASC were enriched for clones that utilized J-4 segment and unswitched IgM ASC displayed a preference for the J-1 segment. Therefore, usage of specific J regions can be shared or unique to isotype-specific cell populations. Overall, clones for each ASC isotype employed a diverse range of V segments (Fig. 6B). Analysis of the V-J linkages for the ASC isotypes between one example mouse and the influenza-specific ASC revealed that J regions connected with different V region gene segments in each of the three populations of isotype-specific ASCs, contributing to the diversity of the responding repertoire. Focusing on the IgG ASC, the v9–4 segment was enriched as one of the top three V segments in each sample. Additionally, the v9–2 segment was enriched in both influenza-specific IgA and IgG samples. For the influenza-specific IgM, only v1–75 was shared in the top enriched V segments. As anticipated, the influenza-specific ASC were more clonally enriched than either bulk ASC or LPS-generated ASCs (26, 50, 51) that have many clones that each make up only a small proportion of the total (Fig. 6C). We next assessed the relationships between CDR3 amino-acid sequences of the ASC. Using a Jensen-Shannon divergence metric (52) with Euclidean distance and multi-dimensional scaling (MDS), we found a surprising consistency between CDR3 sequences within ASC isotypes that was shared between different mice (Fig. 6D). Clustering the top 50 clones from each sample revealed that despite coming from different mice, clones containing similar CDR3 amino-acid sequences were shared in the bulk ASC repertoire (Fig. 6E).

Figure 6. Repertoire of influenza-specific and bulk ASC of distinct isotypes.

(A) Heatmap of J-segment usage for each ASC cell type. Data are Z-score normalized by column and hierarchically clustered. Bulk ASC are denoted by closed boxes and influenza-specific ASC by open boxes. (B) Circos plots showing the V-J combinations for each ASC isotype from one representative mouse (top) for bulk ASC and each mouse for influenza-specific ASC. The top three-four V segments are indicated. (C) Clonality plot for the top 40% of clones from each ASC population. For each clone, the % contribution to the total clonal population is plotted versus the normalized clone size for each sample. LPS ASC data: Haines et al (50); Guo et al (26); Scharer et al (51). (D) MDS plot of the pairwise Euclidean distance of the Jensen-Shannon divergence. Bulk ASC are denoted by closed circles and influenza-specific ASC by open circles. (E) Heatmap depicting the frequency of CDR3 amino acid sequences for the top 50 clones of each bulk ASC population. Data is Z-score normalized by column and hierarchically clustered. For select clones, the CDR3 amino acid sequence is indicated. n.d., not detected.

DISCUSSION

Here, we took advantage of an optimized intracellular staining panel that utilized B cell tetramers and isotype specific antibodies to identify influenza-specific ASC that express IgM, IgG, or IgA BCR constant regions. This approach was combined with protocols that allowed the extraction of cellular RNA for deep sequencing, facilitating the molecular characterization of ASC subsets. This is a critical advancement in the understanding of B cell biology as we can begin to correlate transcriptional programs of isotype-specific ASC with known developmental and functional differences. The isotypes of ASC responding to NP and HA antigens at d14 post influenza infection was dominated by IgM and IgG subsets in circulation. The early emergence of IgM ASC has been reported for other model antigens (36), and IgG, specifically IgG1, IgG2b, and IgG2c are characteristic of influenza ASC responses in C57Bl/6 mice (53), but the kinetics of IgA ASC are not as clearly defined. These data highlight the diversified ASC response at an early immune time point.

Analysis of the transcriptional programs of isotype-specific ASC revealed that genes expressed in bone marrow plasma cells (22) were also expressed in circulating (spleen and LN) ASC at d14 following influenza infection. However, it was surprising to find 15% (43/289) of genes were differentially expressed between at least two ASC isotypes. This suggests functional diversification of ASC, such as has been indicated for LAG3-expressing regulatory plasma cells (49). In fact, consistent with reported findings (49), LAG3 protein expression was highest in IgM ASCs and correlated with mRNA levels in this study. Further, many of the gene expression differences observed were consistent with homing properties for cells of each isotype. IgA ASC upregulated Ccr9, Ccr10, and Itgb7, which facilitates homing to mucosal sites. Additionally, IgA ASC expressed high levels of Tgfbr1 which induces switching to IgA (45). CD98, a neutral amino acid transporter, was expressed more highly on IgM and IgA ASC than IgG ASC. mRNA expression of the Slc3a2 subunit of CD98 correlates with protein expression; it is highest on IgM ASCs and lowest on IgG ASCs. CD98 expression is associated with dividing cells (46). In this scenario, CD98 expression may be reflective of the proliferative potential of early IgM plasmablasts. Cd36 was also expressed exclusively in IgM ASC, albeit at low levels. CD36 is a scavenger receptor that can bind to a variety of ligands including lipoproteins and collagen (54–56). In concordance with a reported role in marginal zone B cells, CD36 protein expression was found here to be expressed on some IgM ASCs. Finally, we validated BCL-2 expression by flow cytometry as well. BCL-2 was found more highly expressed on IgM and IgA ASC. Given its anti-apoptotic function, divergent expression of BCL-2 may reflect the isotype-switching potential of IgM-expressing ASC and may be expressed on IgA ASC to preserve those cells in harsh or nutrient-deplete environments, such as the gut or the lung, as in the case of influenza infection. Thus, each ASC isotype has a distinct transcriptional signature that reflects their developmental history, homing potential, and potentially specialized functions. These data reinforce the importance of identifying the transcriptomes of more highly subdivided ASC.

We were able to extract information about the BCR V-J segment combinations from the sequencing data. This was used to confirm the purity of our ASC subsets but also to evaluate the clonal similarities and differences in each ASC population. IgG and IgA class switched ASC were enriched in J-4 segment usage while IgM were enriched in J-1 segments. Clustering of CDR3 amino-acid sequences reveals similarity within isotype subsets, even among different mice. This suggests that distinct BCR clones may be selected into different ASC lineages, potentially from a common microbiota. This finding is in line with previous work in which there is evidence for a specific V region preference in the early response to PR8 in Balb/c mice (57, 58). It is important to note that our staining strategy pooled NP and HA tetramers. Therefore, the clonal differences between subsets could be caused by a skewing of one isotype toward HA or NP clones. Nevertheless, these data are generated from a snapshot of the repertoire and these results do suggest that specific VDJ combinations may be preferred for constant-region functional pairing or antigen specificities.

The unique transcriptional features of the ASC with distinct isotypes emphasizes the roles of cytokines and germinal center reactions that shape class switch recombination and indicates that these signals may impart broader molecular differences on ASC populations. Further investigation into the transcriptional networks that shape these phenotypic differences will be critical in vaccine design where one isotypic subset is desired over another.

Key Points.

1. Fix/permeabilization approach allows isolation of antigen- and isotype-specific ASC

2. IgM, IgG, and IgA ASC have shared and distinct transcriptome components

3. Each ASC isotype displays distinct, expanded CDR3 regions

ABBREVIATIONS

ASC

antibody secreting cell

DEG

differentially expressed gene

dLN

draining lymph node

HA

hemagglutinin

NP

nucleoprotein

PCA

principal component analysis