Sequential and coordinated control of human plasma cell differentiation by IRF4 and BLIMP1 utilizing a discriminating ISRE/EICE motif lexicon

Louis C Lau
Sneha Lal
Jingyu Fan
Nicholas A Pease author has email address
Harinder Singh author has email address

Center for Systems Immunology and Department of Immunology, University of Pittsburgh, Pittsburgh, United States
Program in Microbiology and Immunology, University of Pittsburgh, Pittsburgh, United States
Department of Computational and Systems Biology, University of Pittsburgh, Pittsburgh, United States

https://doi.org/10.7554/eLife.112254.1

Open access
Copyright information

Figures and data

A stepwise human B cell in vitro differentiation system generates plasma cell precursors that mature into PCs.
(A) Overview of the in vitro PC differentiation culture system. Naive human B cells isolated from healthy donor PBMCs were cultured through sequential phases of activation/expansion (D0-D4), PB/prePC generation (D4-D7), PC differentiation (D7-D14), and PC maturation/maintenance (D14+). (B, C) PB and PC differentiation were monitored by surface immunophenotyping. (B) Representative flow cytometry plots show gating strategies used to quantify actBs (CD20⁺CD38⁺CD27^-), PBs (CD20^-CD38⁺CD27⁺ICAM2^+/-CD138^-), and PCs (CD20^-CD38⁺CD27⁺ICAM2⁺CD138⁺) at D7 and D21. (C) Bar graphs show PB and PC frequencies among live cells at the indicated timepoints (Donor 1). Dots indicate independent experiments, and error bars indicate mean +/- SD (n=3-4; *p<0.05, **p<0.01; two-way ANOVA with Tukey’s post-hoc test). (D, E) Intracellular flow cytometry was used to measure key transcription factors (BLIMP1, IRF4, IRF8) and proliferative status (Ki-67) along the PB/PC trajectory. To relate intracellular measurements to the surface-defined states in Fig. 1B, cells were gated as CD20⁺ (blue) or CD20^- (orange) and displayed for the indicated proteins and timepoints. (D) Biplots showing BLIMP1 and IRF4 protein levels at D7 (left) and D21 (right). (E) Mode-normalized histograms of IRF8 and Ki-67 at D7 (dashed) and D21 (solid). (F) Schematic illustrates experimental design used to determine which D7 compartment contains functionally competent prePCs. The actB and PB populations were sorted at D7, placed back in culture, and analyzed by flow cytometry at D21. Histograms normalized to modal counts show CD138 expression of sorted D7 actB (blue) and PB (green) populations before and after their re-culture through D21. (G) At D21, the PCs generated from the two sorted populations were assayed by IgG ELISpot analyses (n=4 technical replicates; actB-generated PCs input=375 cells; PB-generated PCs input=236 cells).

Single-cell transcriptional profiling delineates prePCs and demonstrates correspondence of their PC progeny with BMPCs.
(A) In vitro-derived cells were profiled by scRNA-seq at D7 and D19. After Leiden clustering (Fig. S2A), cells were consolidated into three major states based on marker-gene expression: actB, PB/prePC, and PC. UMAP displays aggregated cells from both timepoints and donors (Donors 1 and 2), annotated by cell state (left). Heatmap shows the top 100 marker genes per cell state, representing actB, PB/prePC, and PC gene modules (right), with representative genes from each module indicated. (B) UMAP projections show module scores for the indicated cell states (left) and timepoint annotation (right). (C) CCA-like integrated UMAPs show in vitro-derived cells jointly embedded with human BMPCs based on scRNA-seq profiles. (D) PC module scores, derived from the PC gene module in Fig. 2A, are shown on the integrated UMAP (left) and as violin plots (right) for in vitro-derived cells and BMPCs grouped according to the source annotation (Pop A, Pop B, Pop D).

Stage-specific perturbations reveal distinct functions of IRF4 and BLIMP1 in prePCs.
(A) Stage-specific IRF4 and PRDM1 perturbations were analyzed by immunophenotyping during PC differentiation. At D7, cells were nucleofected with CRISPR/Cas9 RNP complexes containing non-targeting control gRNAs (Control) or gRNAs targeting IRF4 (IRF4 KO) or PRDM1 (PRDM1 KO). Representative flow cytometry plots show PB (CD38⁺CD27⁺ICAM2^+/-CD138^-) and PC (CD38⁺CD27⁺ICAM2⁺CD138⁺) gating at D14 (left). Bar plots show mean PB and PC frequencies among live cells at D14 and D21 for Donor 1 (n=2-3 independent experiments; **p<0.01, ***p<0.001; two-way ANOVA with Dunnett’s post-hoc test for KO versus Control). (B) Effects of IRF4 and PRDM1 perturbation on IRF4 and BLIMP1 protein abundance were analyzed by intracellular flow cytometry. Representative plots show IRF4 and BLIMP1 levels at D9, 2 days post-KO (left). Bar plot shows BLIMP1^hiIRF4^hi frequencies at D9 for Donor 1 (right). (C) scRNA-seq analysis of Control, IRF4 KO, and PRDM1 KO cells at D9. UMAP projections show aggregated cells annotated by Leiden cluster or sample identity. Bar plot shows the distribution of cluster cell numbers across samples at D9 for Donor 1 (right). (D, E) Venn diagrams show intersections of genes repressed (D) or activated (E) by IRF4 and BLIMP1, inferred from DEGs. Bar plots show selected gene ontology pathway enrichments for IRF4-only (green) and BLIMP1-only (orange) DEGs, with representative genes annotated (see Methods for DEG analysis details).

Differential IRF4 and BLIMP1 genomic occupancy reveals a discriminating ISRE/EICE motif lexicon.
(A) Schematic summarizes the strategy for comparing IRF4- and BLIMP1-bound open chromatin regions (OCRs) linked to KO DEGs (left). Donut chart shows the distribution of IRF4-bound, BLIMP1-bound, and co-bound OCRs (p≈0 for overlapping region frequency, hypergeometric test) (right). (B) Bar plot shows enrichment of ISRE and EICE motifs in co-bound (gray), IRF4-bound (green), and BLIMP1-bound (orange) OCRs (FDR≤0.056). (C) State-specific ChromBPNet contribution-score analysis of ISRE, EICE, and BLIMP1 motif instances across actB, PB/prePC, and PC states. PWMs for the indicated motifs are displayed above histograms showing contribution-score signal at motif instances in the indicated OCR categories. (D, E) Sequence logos show Jensen-Shannon (J-S) divergence between (D) ISRE or (E) EICE motif instances at IRF4-bound OCRs linked to IRF4 KO DEGs and BLIMP1-bound OCRs linked to PRDM1 KO DEGs. Reference motifs above divergence logos indicate degenerate nucleotides of interest (red asterisks), along with nomenclature for variant positions. Discriminating variant positions in J-S divergence logos are denoted by red asterisks. (F) UCSC Genome Browser tracks show an example of an IRF4-bound ISRE linked to the IRF4-repressed gene SEMA4D. (G) UCSC Genome Browser tracks show an example of a BLIMP1-bound EICE linked to the BLIMP1-repressed gene MBD2.

In vitro TF binding assays test base-pair rules and substantiate the ISRE/EICE regulatory logic.
(A, B) Electrophoretic mobility shift assays (EMSAs) were performed to test N₁/N₃ base-pair rules for IRF4 and BLIMP1 binding at EICE and ISRE motif variants. Gels are grouped by G-containing (G/N or N/G), C-containing without G (C/H or H/C), and A- or T-containing (W/W) variants (N₁/N₃). Relevant TF-specific bands are annotated with red stars. Summaries of favored and disfavored N₁/N₃ nucleotides for each TF-probe combination are shown on the right. (A) Cell-free in vitro-transcribed/translated (IVT) lysates containing IRF4 and PU.1 (top), BLIMP1 (middle), or IRF1 (bottom) were assayed for binding to labeled EICE probes. (B) IVT lysates containing IRF4 (top), BLIMP1 (middle), or IRF1 (bottom) were assayed for ISRE variant probe binding. IRF4-ISRE assays were performed with an antibody directed against the epitope-tag and analyzed for super shifted complexes. (C, D) Statistical enrichment analysis of discriminating N₁/N₃ motif variants at IRF4-bound OCRs linked to IRF4 DEGs (green) or BLIMP1-bound OCRs linked to BLIMP1 DEGs (orange). Bar plots show the frequency of (C) IRF4-specific ISRE or EICE motif variants (N₁/N₃=C/C) and (D) BLIMP1-binding competent ISRE or EICE motif variants (N₁/N₃=G/N or N/G) (Fisher’s exact tests; see Table S8 for counts).

Stepwise human B cell differentiation generates PB/prePC and PC populations in vitro.
(A) Upregulation of PB-associated CD38 and CD27 was monitored by flow cytometry at D4 & D7. Representative plots are shown for indicated timepoints. (B, C) PB and PC differentiation were monitored by surface immunophenotyping. (B) Representative flow cytometry plots show gating strategies used to quantify actBs (CD20⁺CD38⁺CD27^-), PBs (CD20^-CD38⁺CD27⁺ICAM2^+/-CD138^-), and PCs (CD20^-CD38⁺CD27⁺ICAM2⁺CD138⁺) at D7 and D21. (C) Bar graphs show PB and PC frequencies among live cells at the indicated timepoints (Donor 2). Dots indicate independent experiments, and error bars indicate mean +/- SD (n=4-6; *p<0.05, **p<0.01; two-way ANOVA with Tukey’s post-hoc test). (D) Histograms comparing CD20 staining and gating frequencies in surface-flow (Fig. 1B and S1B) and intracellular-staining analyses (Fig. 1D, E and S1E, F) at the indicated timepoints. (E, F) Intracellular flow cytometry was used to measure BLIMP1, IRF4, IRF8, and Ki-67. Cells were gated as CD20⁺ (blue) or CD20^- (orange) to relate intracellular measurements of the indicated TFs to the surface marker-defined PB/PC trajectory. (E) Biplots showing BLIMP1 and IRF4 expression at D7 (left) and D21 (right). (F) Histograms normalized to modal counts are shown for IRF8 and Ki-67 at D7 (dashed) and D21 (solid).

Single-cell transcriptional profiling resolves PB/prePC, PC, and isotype-associated states.
(A) UMAP displays aggregated cells across timepoints (D7 and D19) and donors (Donors 1 and 2), annotated by Leiden cluster. (B) Bar plots show the relative frequency of each cell state at the indicated timepoints (actB, p=0.0013; PC, p=0.0011; two-way ANOVA with Sidak’s multiple-comparison test). (C) UMAP projections show normalized expression of selected genes associated with PB/PC differentiation. (D) UMAP projections show cell-cycle module scores and MYC expression (top) and isotype scores derived from immunoglobulin heavy-chain transcript levels (bottom). (E) UMAP projections show normalized expression of selected genes associated with PC homing and survival. (F) UMAP projection shows CCA-like integration of in vitro-derived cells and human BMPCs annotated by Leiden cluster.

Stage-specific perturbations reveal distinct functions of IRF4 and BLIMP1 in prePCs.
(A) Effects of stage-specific IRF4 and PRDM1 perturbations on PC differentiation were analyzed by immunophenotyping. Bar plots show mean PB and PC frequencies among live cells at D14 and D21 for Donor 2 (n=2 independent experiments; **p<0.01, ***p<0.001; two-way ANOVA with Dunnett’s post-hoc test for KO versus Control). (B) Effects of IRF4 and PRDM1 perturbation on IRF4 and BLIMP1 protein abundance were analyzed by intracellular flow cytometry. Bar plots show BLIMP1^hiIRF4^hi frequencies at D9 for Donor 2, related to Fig. 3B. (C) scRNA-seq analysis of Control, IRF4 KO, and PRDM1 KO cells at D9. Bar plots show the distribution of cluster cell numbers across samples at D9 for Donor 2, related to Fig. 3C. (D) UMAP projections show cell-state module scores derived from Fig. 2A, cell-cycle scores, and MYC expression. (E) Correlation between log₂(fold-change) and the change in fraction of cells expressing each DEG after IRF4 or PRDM1 KO relative to the indicated reference groups (IRF4 KO: Stunted PCs versus PCs; PRDM1 KO PCs versus Control PCs, see Methods). Amplitudes are shown as log₂(fold-change) [KO – control], and frequencies are shown as the difference in fraction of expressing cells [KO – control]. (F) Venn diagram shows intersections among IRF4 and BLIMP1 DEGs, with unique and concordant DEG numbers summarized in the table below (left). Scatter plots show log₂(fold-change) relationships for concordant (R²=0.68) and discordant (R²=0.77) KO DEG sets (right). (G, H) Bar plots show ChIP-X and/or gene ontology pathway enrichments for genes co-repressed (G) or co-activated (H) by IRF4 and BLIMP1, with representative genes indicated. (I, J) Bar plots show ChIP-X enrichment of target gene sets for IRF4-only or BLIMP1-only repressed (I) and activated genes (J).

Differential IRF4 and BLIMP1 genomic occupancy reveals a discriminating ISRE/EICE motif lexicon.
(A) Bar plots show motif enrichment analysis for selected IRF-related motifs in co-bound, IRF4-bound, and BLIMP1-bound OCRs. (B, C) Sequence logos show regenerated PWMs for (B) ISRE and (C) EICE motif instances at IRF4-bound OCRs linked to IRF4 KO DEGs or BLIMP1-bound OCRs linked to PRDM1 KO DEGs, related to Fig. 4D and 4E. (D) UCSC Genome Browser tracks show an example of an IRF4-bound ISRE linked to the IRF4-activated gene XBP1. (E) UCSC Genome Browser tracks show an example of a BLIMP1-bound EICE linked to the BLIMP1-activated gene DAD1.

In vitro TF binding assays test base-pair rules and substantiate the ISRE/EICE regulatory logic.
(A) EMSA analysis of IRF4 binding to ISRE probes in the presence of PU.1 instead of Ab directed against IRF4 epitope tag, related to Fig. 5B. (B) Comparison of ISRE and EICE association with IRF4 and BLIMP1 DEG-linked OCRs. Bar plots show the frequency of IRF4 (left) or BLIMP1 (right) DEGs linked to IRF4-bound or BLIMP1-bound OCRs containing ISRE or EICE motifs. The number of DEGs in each category is shown in parentheses above the bar plots.

Sign up for email alerts