Splenic fibroblasts control marginal zone B cell movement and function via two distinct Notch2-dependent regulatory programs

Abstract

Innate-like splenic marginal zone (MZ) B (MZB) cells play unique roles in immunity due to their rapid responsiveness to blood-borne microbes. How MZB cells integrate cell-extrinsic and -intrinsic processes to achieve accelerated responsiveness is unclear. We found that Delta-like1 (Dll1) Notch ligands in splenic fibroblasts regulated MZB cell pool size, migration, and function. Dll1 could not be replaced by the alternative Notch ligand Dll4. Dll1-Notch2 signaling regulated a Myc-dependent gene expression program fostering cell growth and a Myc-independent program controlling cell-movement regulators such as sphingosine-1 phosphate receptor 1 (S1PR1). S1pr1-deficient B cells experienced Notch signaling within B cell follicles without entering the MZ and were retained in the spleen upon Notch deprivation. Key elements of the mouse B cell Notch regulome were preserved in subsets of human memory B cells and B cell lymphomas. Thus, specialized niches program the poised state and patrolling behavior of MZB cells via conserved Myc-dependent and Myc-independent Notch2-regulated mechanisms.

Keywords: B cells; Notch; chemotaxis; fibroblastic reticular cells; marginal zone; spleen.

Figure 1.. Dll1 but not Dll4 ligands control Notch-regulated transcription in MZB cells and mediate their retention in the spleen

C57BL/6 mice received isotype control, anti-Dll1, anti-Dll4 or anti-Notch2 antibodies 12-48 hours before analysis of splenic B cells. (A) Contour plots gated on CD19⁺CD93⁻ mature B cells showing CD1d^lowCD23^hi FoB and CD1d^hiCD23^low MZB cells 48 hours after antibody administration. (B) %CD1d^hiCD23^low MZB cells after treatment with isotype control, anti-Dll1 or anti-Dll4 antibodies 12-48 hours earlier. (C) %Dll4/Notch1-dependent CD25^hiCD44^low double negative 3 (DN3) cells thymocytes at 48 hours. (D) Histograms and contour plots for surface CD21 in CD19⁺CD93⁻ B cells. (E) CD21 mean fluorescence intensity (MFI). n=2-4/group. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001, ns=non-significant (two-way ANOVA, Bonferroni correction). (F) RNA-seq of CD1d^hiCD23^low MZB cells. Best-fit Euler plots for genes with decreased mRNA upon Dll1 and Dll4 blockade (12-48 hours) vs Notch2 blockade (24-48 hours). (G) Differential expression induced by Dll1 and Dll4 blockade for genes with decreased mRNA upon Notch2 blockade. Genes with a log₂ fold change (FC) ≥1 are in purple (left) and Benjamini-Hochberg-adjusted p-value<0.05 in orange (right). (H) Transcripts per Million (TPM) for selected Notch transcriptional target genes in MZB and FoB cells after Dll1 and Dll4 blockade, with mean (bar), 75% (box) and 95% confidence interval (whisker) plus individual data points (jitter).

Figure 2.. f Dll1 and Dll4 expression in splenic Ccl19-Cre⁺ fibroblastic reticular cells

Flow cytometry and imaging identify Ccl19-Cre⁺ FRCs expressing mCherry in lymph nodes and spleen of Ccl19-Cre⁺ROSA^eYFP mice carrying Dll1-mCherry or Dll4-mCherry reporters. (A-C) Collagenase-digested lymph node (A) and spleen (B) showing eYFP⁺ non-hematopoietic/non-endothelial (CD45⁻Ter119⁻CD31⁻) cells, followed by detection of Dll1-mCherry or Dll4-mCherry vs gp38/Podoplanin. Contour plots (A-B), quantification (C). Data representative of 4 experiments. *p<0.05, ****p<.0001, ns=non-significant (two-way ANOVA, Bonferroni correction). (D) eYFP and Dll1-mCherry among spleen non-hematopoietic/non-endothelial cells and type 3 innate lymphoid cells (ILC3s) in Ccl19-Cre⁺ROSA^eYFP+Dll1-mCherry⁺ mice. Gating on CD45⁻NK1.1⁻GR-1⁻Ter119⁻CD3⁻TCRβ⁻TCRγδ⁻B220⁻CD19⁻ live singlets (non-hematopoietic cells); and CD45⁺NK1.1⁻CD5⁻Gr-1⁻Ter119⁻CD3⁻TCRβ⁻TCRγδ⁻B220⁻CD19⁻CD127⁺ live singlets (ILC-enriched). **p<0.01 (unpaired t-test, Welch’s correction). (E) Imaging of spleen cryosections from Ccl19-Cre⁺ROSA^eYFP+Dll1-mCherry⁺ (top) and Ccl19-Cre⁺ROSA^eYFP+Dll1-mCherry⁺ mice (bottom). CD169 identifies macrophages (dotted line, marginal sinus). eYFP⁺mCherry⁺ FRCs were defined using Imaris (white). Scale bars: 100μm.

Figure 3.. Dll4 cannot substitute for Dll1 to support MZB cells, even when expressed in Ccl19-Cre⁺ FRC niches

Ccl19-Cre⁺ splenic FRCs were engineered to express Dll1 or Dll4 ligands, with/without endogenous Dll1/Dll4 gene inactivation. (A) Experimental scheme depicting the impact of Ccl19-Cre on floxed Dll1 and Dll4 and Cre-inducible Hprt-Dll1 or Hprt-Dll4 transgenes in spleen FRCs. (B) Contour plots showing CD21^lowCD23^hi FoB and CD21^hiCD23^low MZB cells (top), and imaging of spleen cryosections stained for CD1d, IgD, CD169 and CD3 (bottom). Full/empty arrows point to the presence/absence of CD1d^hi B cells in the MZ. Scale bars: 50μm. (C) Data from 6 experiments (n=9-14/group). (D) Experimental scheme depicting the impact of Ccl19-Cre on Cre-inducible Hprt-Dll1 or Hprt-Dll4 transgenes in the presence of wild-type Dll1 and Dll4 alleles. (E)Contour plots showing CD21^lowCD23^hi FoB and CD21^hiCD23^low MZB cells (top), and imaging of spleen cryosections stained for CD1d, IgD, CD169 and CD3 (bottom). Full/empty arrows point to the presence/absence of CD1d^hi B cells in the MZ. Scale bars: 50μm. (F) Data quantification from 4 experiments (n=11-14/group). *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001, ns=non-significant (one-way ANOVA, Brown-Forsynthe and Welch correction). See Figure S1.

Figure 4.. Splenic MZB cells receive Notch2 signals in B cell follicles, and not only in the MZ

(A) Hes1-GFP among spleen CD1d^hiCD23^low MZB and CD1d^lowCD23^hi FoB cells from Hes1-GFP reporter mice. Left: contour plots. Right: quantification (n=4 mice/group), *p<0.05 (Mann-Whitney test). (B) Imaging showing Hes1-GFP (pink) among CD1d^hi cells internal (in the follicle, Fo) and external to the marginal sinus (in the MZ). The marginal sinus was outlined via laminin staining. Arrows point to GFP within CD1d^hi cells in the MZ (top) and follicle (bottom). (C-E) C57BL/6 mice received isotype control, anti-Dll1 or anti-Notch2 antibodies 6 hours before analysis. Data representative of two experiments. (C) Imaging of cryosections with staining for B220, IgD and N2-ICD. Dotted lines outline the marginal sinus based on B220/IgD staining (showing MZB cells as B220⁺IgD^low). Full arrowheads show N2-ICD in B220⁺ B cells of isotype control-treated mice (consistent with cleaved nuclear +/− cytoplasmic N2-ICD). Empty arrowheads point to cell surface N2-ICD in B cells of anti-Dll1 and anti-Notch2-treated mice, both in the MZ and follicle. Images representative of 2 experiments (n=3 mice/group). (D) Imaging flow cytometry (ImageStream) showing nuclear vs cytoplasmic patterns of N2-ICD in MZB cells from mice treated with isotype control vs anti-Dll1 antibodies 1 or 4 hours before analysis. After 5’ intravenous labeling with anti-CD19 antibodies, data were gated on i.v.CD19⁺ vs i.v.CD19⁻ MZB cells with/without direct access to blood circulation. Pool of 2 experiments, n=2-4/group. **p<0.01, ***p<0.001, ****p<0.0001 (one-way ANOVA). See Figures S2–S3.

Figure 5.. Identification of a Notch-regulated Myc-independent transcriptional program controlling B cell retention in the MZ

(A) Experimental strategy to dissociate Myc and Notch effects in B cells. hCD20-TamCre⁺Myc^f/f (Flox) or hCD20-TamCre⁺Myc^+/+ (Cre) mice received tamoxifen for 4 weeks, then isotype control vs anti-Notch2 antibodies 24 hours before analysis. (B) Myc inactivation in MZB cells of tamoxifen-treated hCD20-TamCre⁺Myc^f/f mice. RNA-seq reads are displayed as BigWig tracks showing loss of loxP-flanked Myc exons 2-3 (n=4 mice/group). (C) Contour plots showing CD21^lowCD23^hi FoB and CD21^hiCD23^low MZB cells. *p<0.05, ns=non-significant (two-way ANOVA, Bonferroni correction). (D) Imaging of spleen cryosections stained for CD1d (MZB cells), CD3 (T cells), CD169 (macrophages) and Ter119 (erythroid cells). Arrows point to the presence (full arrows) or absence (empty arrows) of CD1d^hi B cells in the MZ. Scale bars: 100μm. (E) %B cells with a CD1d^hiCD23^low MZB cell phenotype in blood for each genotype with/without Notch2 blockade. ***p<0.001 (two-way ANOVA, Bonferroni correction). (F) hCD20-TamCre⁺Myc^+/+ (Cre) or hCD20-TamCre⁺Myc^f/f (Flox) mice received tamoxifen for 4 weeks, then isotype control vs anti-Notch2 antibodies 24 hours before analysis. FoB and MZB cells were sort-purified for RNA-seq. Hierarchical clustering of 440 genes differentially expressed in MZB cells from mice treated with anti-Notch2 vs isotype control antibodies (z-score with clustering by Pearson correlation). Differential expression was defined by log2 FC≥1 and adjusted p-value<0.05. (G) Transcripts per Million (TPM) for selected Myc-independent Notch-regulated genes (clusters 4-5 in F). Mean (bar), 75% confidence interval (box) and 95% confidence interval (whisker) plus individual data points (jitter). MZ: MZB cells. FO: FoB cells. See Figures S4–S5, Supplemental Table 1.

Figure 6.. S1pr1-deficient MZB cells experience Notch signaling within B cell follicles and are retained in the spleen upon Notch2 inhibition

(A) Contour plots gated on CD19⁺CD93⁻splenic B cells showing CD1d^lowCD23^hi FoB and CD1d^hiCD23^low MZB cells in Mb1^+/Cre;S1pr1^+/+ vs. Mb1^+/Cre;S1pr1^f/f mice. (B) Frequency of MZB and FoB cells for each genotype (n=4-5/group, representative of three experiments). **p<0.01, ***p<0.001 (unpaired 2-tailed Student’s t-test). (C) Imaging of spleen cryosections with staining for CD1d, CD169 and CD3. Scale bars: 50μm. (D) BM chimeras were generated by transplantation of CD45.2⁺ BM cells (3x10⁶) from Mb1^+/Cre;S1pr1^+/+ vs. Mb1^+/Cre;S1pr1^f/f mice into irradiated B6-CD45.1 recipients (2x5.5 Gy). Ten weeks later, recipients received isotype control vs anti-Notch2 antibodies 48-96 hours before analysis. (E) Histograms and (F) MFI for CD21 in CD1d^hi MZB cells after isotype control vs anti-Notch2 antibodies for 48-96 hours. **p<0.01, ***p<0.001, ****p<0.0001 (two-way ANOVA, Bonferroni correction). Data representative of three experiments. (G) Volcano plots showing the impact of Notch2 blockade in S1pr1-deficient MZ-like B cells defined in (A). Highlighted in black/red are genes up/downregulated by Notch2 blockade in MZB cells. Genes in red showed at least −1 log2 FC and Benjamini-Hochberg-adjusted p-values<0.05. (H, I) Flow cytometric analysis of donor-derived CD1d^lowCD23^hi FoB and CD1d^hiCD23^low MZB cells in Mb1^+/Cre;S1pr1^+/+ (H) vs Mb1^+/Cre;S1pr1^f/f mice (I), and spleen cryosections stained for CD1d, CD169, CD3 and Ter119. Scale bars: 100μm. (J) Ratio of the absolute number of MZB cells 48 or 96 hours after anti-Notch2 blockade over numbers in isotype control mice for each genotype. “One” indicates no change with Notch2 blockade (dashed line). Data pooled from 2 experiments. **p<0.01, ****p<0.0001 (unpaired 2-tailed Student’s t-test). See Figure S6–S7.

Figure 7.. Conservation of Notch-regulated transcriptional programs from mouse MZB cells to subsets of human memory B cells and non-Hodgkin lymphomas

(A) Genome tracks showing chromatin binding of N1-ICD (active Notch paralog in Rec1) and RBPJκ, active enhancer histone mark H3K27Ac load, transcription, and SMC1-mediated looping at MYC (left), S1PR1 (middle) and CD21-encoding gene CR2 (right) in the Notch-driven human mantle cell lymphoma (MCL) cell line Rec1 with (red) or without (black) γ-secretase inhibitor. Bottom track: Ensembl gene positions. (B) Notch2-driven genes in mouse MZB cells compared to human B cells. ChIP-verified Notch2-driven genes in Rec1 cells (Ryan et al., group1-up) were filtered for those with mouse orthologs (n=210). Euler plots show genes from the MCL Notch signature whose expression was up in human vaccinia-specific CD21^hi memory B vs naïve B cells (Chappert et al., n=51) and down upon anti-Notch2 antibody treatment in S1pr1-sufficient or deficient MZB cells (see Figure 6 and S7, n=59). See also Supplemental Table 6. (C) Dynamic volcano plots depicting differential expression changes of ChIP-verified Notch2-driven genes (defined in Rec1 cells) between MZB and FoB cells after 48 hour in vivo anti-Notch2 antibody treatment for control or S1pr1-deficient animals (Figure 6 and S7). Arrows depict changes in log2 FC and log₁₀ adjusted p-value from isotype control (origin) to 48h anti-Notch2 blockade (arrowhead). Colors indicate decreased (blue) and increased (red) log₂ FC vs FoB cells. (D) Myc-independent, Myc-dependent and total Notch2-driven MZB cell gene sets from Figure 5F were converted to human orthologs, before gene set enrichment analysis in human vaccinia-specific CD21^hi memory vs naïve B cells. NES, normalized enrichment score. FDR-q: q-value with 1000 gene-set permutations. See Supplemental Table 6.