Germinal center output is sustained by HELLS-dependent DNA-methylation-maintenance in B cells

Abstract

HELLS/LSH (Helicase, Lymphoid Specific) is a SNF2-like chromatin remodelling protein involved in DNA methylation. Its loss-of-function in humans causes humoral immunodeficiency, called ICF4 syndrome (Immunodeficiency, Centromeric Instability, Facial anomalies). Here we show by our newly generated B-cell-specific Hells conditional knockout mouse model that HELLS plays a pivotal role in T-dependent B-cell responses. HELLS deficiency induces accelerated decay of germinal center (GC) B cells and impairs the generation of high affinity memory B cells and circulating antibodies. Mutant GC B cells undergo dramatic DNA hypomethylation and massive de-repression of evolutionary recent retrotransposons, which surprisingly does not directly affect their survival. Instead, they prematurely upregulate either memory B cell markers or the transcription factor ATF4, which is driving an mTORC1-dependent metabolic program typical of plasma cells. Treatment of wild type mice with a DNMT1-specific inhibitor phenocopies the accelerated kinetics, thus pointing towards DNA-methylation maintenance by HELLS being a crucial mechanism to fine-tune the GC transcriptional program and enable long-lasting humoral immunity.

Fig. 1. Hells is highly upregulated in human and mouse during the germinal center reaction.

a Quantification by RT-qPCR of DNMT3B, ZBTB24, CDCA7, and HELLS transcripts of cells sorted from pediatric spleens (n = 2). Naive: CD19⁺CD38⁺CD24⁺IgD⁺IgM⁺CD27^neg; transitional: CD19⁺CD38^hiCD24^hi; Marginal Zone B cells (MZB): CD19⁺CD38⁺CD24⁺IgD⁺IgM⁺CD27⁺; Germinal Center B cells (GCB): CD19⁺CD38⁺CD24^neg; Memory B cells (MBC): CD19⁺CD38⁺CD24⁺IgD^negCD27⁺IgM^neg; Plasma cells (PC): CD19⁺CD38^hiCD24^neg. b Quantification by RT-qPCR of transcripts of splenic naive and GCB cells from WT mice (n = 2), and in vitro induced GCB-like (iGB-4) and ASC (iPC-21) (n = 3), according to Nojima et al.. c Heatmap of the expression of the four ICF genes and of Dnmt1 at all stages of B-cell development and activation in the mouse. RNAseq data were collected from Immgen database GSE109125. Bar charts show relative fold-change of each transcript compared to its level in naïve B cells, after normalization to β2-microglobulin. Each dot represents one sample.

Fig. 2. Hells B-cell conditional knockout mice display normal B-cell development but reduced basal antibody titers.

a Percentage of B lymphocytes (B220⁺) among living splenocytes, measured by flow cytometry in control (n = 10) and Mb1Hells^KO mice (n = 7). b Representative flow cytometry plots showing type 1 and 2 transitional B cells (T1: B220⁺IgM⁺CD93⁺CD23^neg; T2: B220⁺IgM⁺CD93⁺CD23⁺), follicular B cells (FoB): B220⁺CD23⁺CD21⁺; and marginal zone B cells (MZB): B220⁺CD21^hiCD23^lo/neg in the spleen of non-immunized control and Mb1Hells^KO mice. c Quantification of the splenic subsets depicted in (b) in control (n = 10, of which 7 were Mb1^Cre/WTHells^F/WT) and Mb1Hells^KO mice (n = 7). d Quantification by ELISA of immunoglobulin titers of non-immunized 12-weeks old mice. Number (n) of control (of which 4 were Mb1^Cre/WTHells^F/WT) and Mb1Hells^KO mice used is indicated in the figure. Data were pooled from two independent experiments. Bar charts and error bars represent the mean ± SD. Unpaired two-tailed t-test (a) and (c) and two-tailed Welch’s t-test (d) were performed, ns: non-significant, exact p values indicated in figures. Source data are provided in Source Data File.

Fig. 3. Hells expression in B cells is mandatory for efficient TD humoral immune responses and establishment of high-affinity MBCs.

a, b Mice were immunized i.p. with alum-adsorbed NP-CGG and boosted 6 weeks later. Serum was collected at day 7, 14, and 41 after the primary immunization, and 5 days after the boost. NP-specific IgM (a), and IgG1 (b) from Mb1Hells^KO and control animals (n for numbers of mice used is indicated in figures) were quantified by ELISA (arb.units, arbitrary units). c Representative Flow Cytometry (FC) plots of splenic IgG1-secreting cells 5 days after antigenic boost, and quantification of these cells (n = 8 for Mb1Hells^KO, n = 9 for controls). d Enumeration by ELISPOT assay of splenic IgG1-ASC and NP-specific IgG1-ASC, 5 days after antigenic boost. Each well is representative of a triplicate (n = 8 for Mb1Hells^KO, n = 9 for controls). e NP₄-BSA and NP₂₃-BSA-binding IgG1 were quantified by ELISA in the serum of the animals described in (b) (n = 8 for each genotype) and the NP₄/NP₂₃ ratio was calculated. f Representative FC plots of splenic NP-specific IgG1⁺ MBC 28 days after immunization with NP-CGG, and quantification of NP-specific IgG1⁺ MBCs per million B220⁺ B cells in the spleen, 7, 14, 28 and 43 days after immunization (n = 7 for each genotype). g Proportion of VH186.2 BCR sequences from single-cell sorted NP-specific IgG1 MBCs at day 28 post NP-CGG immunization that bear W33L or K59R substitutions in control (n = 7) and Mb1Hells^KO (n = 8) animals. The numbers of sequences (W33L or K59R/total) are indicated at the center of the pies. Experiments (a) to (f) were done twice, (g) was done three times, and data were pooled. Bar charts and error bars represent the mean ± SD. Unpaired two-tailed Welch’s t-test were performed for (a), (b) and (e), and unpaired two-tailed t-test were performed for (c), (d) and (f). Source data are provided in Source Data File.

Fig. 4. GC B cells devoid of Hells form and proliferate normally but collapse prematurely.

a GC kinetics in the spleen after immunization. Representative Flow cytometry (FC) plots and quantification of GC B cells in Mb1Hells^KO and control animals, 7-, 10- and 14-days post NP-CGG immunization (n for numbers of mice used is indicated in the figure). b Percentage of GC B cells among living Peyer’s patches B cells measured by FC (n = 19 for Mb1Hells^KO and n = 33 for control mice, of which 4 were Mb1^Cre/WTHells^F/WT, pooled from multiple independent experiments). c 2D-cell cycle analysis of day 14 GC B cells after in vivo EdU labeling. Representative FC plots (left) and quantification of cell cycle phases. Data shown for n = 7 animals of each genotype (2 controls were Mb1^Cre/WTHells^F/WT). d Distribution of splenic GC B cells between Light Zone (LZ) and Dark Zone (DZ) 14 days after immunization. Representative FC plots for LZ and DZ B cells (left), and quantification of their percentages within each zone (right, n = 8 for each genotype). e Somatic hypermutation frequency in GC B cells. Bar charts show the frequency of intron JH4 mutations per 100 pb in GC B cells 14 days after immunization with NP-CGG. n = 4 animals of each genotype. f Proportion of VH186.2 BCR sequences from sorted GC B cells at day 14 post NP-CGG immunization that bear W33L or K59R substitutions in control and Mb1Hells^KO animals. The numbers of sequences (W33L or K59R/total) pooled from 4 mice of each genotype are indicated at the center of the pies. Experiments were done twice; data were pooled for (a), (c), (d), (e), and (f), and one representative experiment was shown for (c). Bar charts and error bars represent the mean ± SD. Unpaired two-tailed t-test were performed for (a), (b), (c), (d) and (e); ns: non-significant. Source data are provided in Source Data File.

Fig. 5. Hells^KO GC B cells undergo deep hypomethylation and derepression of repeated non-coding sequences and germline genes.

a DNA methylation levels of naïve (day 0), day 7 and 14 GC B cells sorted from Mb1Hells^KO and control animals. The distribution of mCpG/CpG ratio is shown by violin plots for 200 CpGs genomic tiles (left) and different classes of repetitive elements (right). The average of two biological replicates is shown. b Correlation plots of CpG methylation levels at CpG-dense promoters (% CpG > (0.25)²*100) between Mb1Hells^KO and control animals. Promoters with WT methylation >75% at all time points were clustered based on the kinetics of methylation changes in Mb1Hells^KO using k-means clustering. c MA plots displaying the differential expression of RNA Transposable Elements (TEs) between Mb1Hells^KO and control samples at the three time points. X and Y axis show the average of normalized count values and the shrunken log2 fold changes, respectively. Colored dots have adjusted p-value < 0.05, calculated by a two-tailed Wald test. d Correlation between CpG methylation loss in Mb1Hells^KO samples and reactivation kinetics (DESeq2 shrunken log2 fold change) for full-length RNA TEs (LINE1 > 5 kb, IAP > 6 kb and MMERVK10C > 4.5 kb). e Relative amounts of IAPEz, minor satellite and LINE-1 transcripts in naïve and day 10 GC B cells from Mb1Hells^KO mice compared to controls (n = 2 for each genotype). Transcripts were quantified by RT-qPCR and normalized to GAPDH. f Differential gene expression between Mb1Hells^KO and control day 14 GC B cells measured by bulk RNAseq (n = 4 for each genotype). Upregulated (fold change>1.5, p < 0.05) and downregulated genes (fold change < 0.65, p < 0.05) are represented in blue and black, respectively. Orange dots represent derepressed genes, whose transcripts are detected only in Mb1Hells^KO cells. g Classification of the top-50 derepressed genes, according to the presence of an IAPEz copy, or to expression restricted to immunoprivileged tissues. Source data are provided in Source Data File.

Fig. 6. Accelerated acquisition of pre-MBC markers in the absence of Hells.

a Heatmap representation of scaled counts of selected genes in control and Mb1Hells^KO mice. b GSEA showing enrichment of a pre-MBC signature in Mb1Hells^KO vs. control day 14 GC B cells. The pre-MBC signature consists in the top 200 upregulated genes in pre-MBC vs. GC B cells in GSE89897. c Representative flow cytometry (FC) plots showing the CD38⁺CCR6⁺ pre-MBC population within GC B cells (B220⁺CD95⁺GL7⁺), and quantification of the pre-MBC subset in Mb1Hells^KO (n = 6) and control (n = 6) mice. d Experimental setup for in vivo EdU pulse labeling of GC B cells and follow-up by flow cytometry. e Representative FC plots showing the phenotype of EdU positive live B cells 24 h after the in vivo pulse in one Mb1Hells^KO and one control mice. Frequencies of GC (CD95⁺GL7⁺), pre-MBC (CD95⁺GL7⁺CD38⁺CCR6⁺) and MBC (GL7^-CCR6⁺CD38⁺) were quantified in n = 8 Mb1Hells^KO and n = 9 control mice. Experiments c and e were done twice, and data were pooled. Bar charts and error bars represent the mean ± SD. Unpaired two-tailed t-test were performed. Source data are provided in Source Data File.

Fig. 7. Mb1Hells^KO GC B cells overexpress an mTORC1/ATF4-dependent metabolic signature.

a GSEA showing enrichment of signatures related to mTORC1 activation, UPR, ATF4 targets and cholesterol biosynthesis in Mb1Hells^KO day 14 GC B cells. NES normalized enrichment score, FDR false discovery rate. b Venn Diagram of upregulated genes in day-14 Mb1Hells^KO GC B cells compared to mTORC1 targets upregulated in activated B cells (GSE141423) and ATF4-dependent mTORC1 targets (GSE158605). Genes commonly upregulated in day-14 Mb1Hells^KO GC B cells and in GSE141423 and/or GSE158605 are represented in blue on the volcano plot (gray and blue = p < 0.05, calculated by a two-tailed test). Source data are provided in Source Data File.

Fig. 8. Hells^KO GC B cells overexpress an mTORC1/ATF4-dependent metabolic pathway upregulated during ASC differentiation.

a Single-cell transcriptional analysis of day 7 and 14 GC B cells. UMAP displaying 429 day-14 GC B, 466 day-7 GC B Mb1Hells^KO, 414 day-14 GC B, and 448 day-7 GC B control cells, colored by shared nearest neighbor clusters collected. Bar charts show the frequencies of GC B cells within the 4 clusters. n = 2 mice for each genotype and timepoint; one of the two control mice for day-7 GC B cells was Mb1^Cre/WTHells^F/WT. b Selected gene expression for the 4 different clusters shown in (a), presented in dot plot scaled on normalized UMI counts. c Representative CD98 staining of CD95⁺GL7⁺CCR6^negCD38^neg GC B cells and quantification of CD98 geometric MFI (Mean Fluorescent Intensity). d Representative intracellular ATF4 staining within CD95⁺GL7⁺CCR6^neg GC B cells and quantification of ATF4 geometric MFI (Mean Fluorescent Intensity). e Representative CD98 staining of B220⁺ day-8 iGB cell cultures and quantification of CD98 geometric MFI (Mean Fluorescent Intensity). f Expression of Asns and Chac1 in sorted B220⁺CD138^neg day-8 iGB cells measured by RT-qPCR after normalization to Ubc. g Representative CD98 staining of GFP⁺ and GFP^neg day-8 iGB cell cultures after transduction of empty or ATF4 coding retroviral vectors, and quantification of CD98 geometric MFI (Mean Fluorescent Intensity). h GSEA showing enrichment of ASC signature in Mb1Hells^KO vs. control day 14 GC B cells. The ASC signature consists in the top 200 genes upregulated in GSE60927. NES normalized enrichment score, FDR false discovery rate. n = 4 for each genotype in (c), (d), (e), (f) and (h); experiments were done twice and one representative experiment is shown. n = 5 for (g); experiment was done twice and data were pooled. Bar charts and error bars represent the mean ± SD. Unpaired two-tailed t-test were performed for (c) to (g). Source data are provided in Source Data File.

Fig. 9. Pharmacological inhibition of DNMT1 phenocopies the impact of HELLS loss-of-function on GC kinetics.

a Experimental setup for DNMT1 inhibition in vivo. b Representative flow cytometry (FC) plots showing splenic GC B cell population after treatment with vehicle or DNMT1 inhibitor GSK3685032, and quantification of GC B percentage. c Percentage of unmethylated CpGs inferred from bisulfite sequencing of 16–20 LINE1 copies amplified from sorted GC B cells after 5 days of treatment with DNMT1i or vehicle. d Representative FC plots showing the pre-MBC population within GC B cells after treatment with vehicle or GSK3685032, and quantification of pre-MBC percentage. e Representative CD98 staining within CD95⁺GL7⁺CCR6^negCD38^neg GC B cells after treatment with vehicle or GSK3685032, and quantification of CD98 geometric MFI (Mean Fluorescent Intensity). f Representative Cell Trace Violet staining after 3 days of in vitro culture with LPS + IL-2 + IL-5, in the presence of 300 nM GSK3685032 or DMSO. Distribution of living cells according to the number of cell divisions, for increasing concentrations of GSK3685032 or DMSO. g Representative FC plots showing ASC (CD138⁺B220⁺) differentiation after 3 days of in vitro culture as in (e). Frequencies of ASCs as a function of cell division number, for increasing concentrations of GSK3685032 or DMSO (n = 2 independent cultures). h Representative CD98 staining of cells that underwent 4 divisions after 3 days of in vitro cultures as in (e) and quantification of CD98 geometric MFI (Mean Fluorescent Intensity) as a function of cell division number and of GSK3685032 or DMSO concentrations (n = 2 independent cultures). All experiments were done twice; data were pooled for (b) to (e); one representative experiment is shown for (f), (g), (h). n = 12 mice for each treatment in (a), (b), (d). n = 4 mice for each treatment were used in (c). Bar charts and error bars represent the mean ± SD. Unpaired two-tailed t-test were performed for (b), (c), (d), (e). Source data are provided in Source Data File.