The RNA-binding protein HuR is required for maintenance of the germinal centre response

The germinal centre (GC) is required for the generation of high affinity antibodies and immunological memory. Here we show that the RNA binding protein HuR has an essential function in GC B cells to sustain the GC response. In its absence, the GC reaction and production of high-affinity antibody is severely impaired. Mechanistically, HuR affects the transcriptome qualitatively and quantitatively. The expression and splicing patterns of hundreds of genes are altered in the absence of HuR. Among these genes, HuR is required for the expression of Myc and a Myc-dependent transcriptional program that controls GC B cell proliferation and Ig somatic hypermutation. Additionally, HuR regulates the splicing and abundance of mRNAs required for entry into and transition through the S phase of the cell cycle, and it modulates a gene signature associated with DNA deamination protecting GC B cells from DNA damage and cell death.

G erminal centres (GC) are the sites for active B cell proliferation and, following somatic hypermutation (SHM), selection of B cell clones with increased affinity for antigen. GC B cells give rise to long-lived memory B cells and antibody-secreting plasma cells that protect against recurrent infections 1 . Based on cell density, GCs are divided in two main areas: the dark zone (DZ) and the light zone (LZ) 2 . DZ B cells are highly proliferative and undergo immunoglobulin (Ig) SHM 3 . They migrate to the LZ to test their newly mutated B cell receptor (BCR) by capturing antigen from follicular dendritic cells and seeking help from T follicular helper (Tfh) cells 4,5 . The scarcity of antigen makes this a highly competitive process where LZ GC B cells expressing BCRs with the highest affinities are subsequently selected for further rounds of proliferation 6 or for terminal differentiation into plasma cells or memory B cells 7 . Establishment of the GC reaction is stringently regulated by transcriptional and post-transcriptional mechanisms.
To achieve their functions, DZ and LZ GC B cells have very distinctive transcriptomes that reflect the differential expression of transcription factors. Bcl6 is required for naive B cell differentiation into GC B cells upon activation by antigen and Tfh cells 8 . E2A, PAX5 and FOXO1 are all required for proliferation and induction of activation-induced cytidine deaminase (AID)mediated SHM in DZ B cells 9 . By contrast, activation of nuclear factor (NF)-kB and expression of Myc are required for GC B cell maintenance [10][11][12] . Myc is mainly expressed by LZ GC B cells expressing high-affinity BCRs that are positively selected by Tfh cells 13 . Myc protein abundance reflects the extent of Tfh cell help received by LZ GC B cells 14 and sets in motion a genetic programme marking further rounds of proliferation and SHM in the DZ 11,12 .
Post-transcriptional regulation of gene expression is also an essential regulatory principle of GC responses, but it is far less well understood than the regulation by transcription factors. MicroRNAs have been found to function in the GC B cell by buffering the toxic effects of Myc and AID-mediated DNA damage 15,16 . RNA-binding protein (RBP) regulate cell fate and function by controlling multiple aspects of gene expression, including transcription, processing, subcellular localisation, stability and translation of mRNA into protein 17 . Recent studies have shown that PTBP1, an RBP that is induced by Myc, has an ancillary function in the Myc-dependent gene expression programme of GC B cells 18 . Also, conditional deletion of the RBP human antigen R (HuR; encoded by Elavl1) from the pro-B cell stage onwards has shown its essential effect in B cell activation and antibody production 19,20 . However, the function that HuR plays in the biology of GC B cells remains unknown.
HuR binds to U-rich elements present mostly in introns and 3' untranslated regions (3'UTR) of its RNA targets 21,22 . HuR shuttles from the nucleus to the cytoplasm where it mediates distinct functions. In the nucleus, HuR controls mRNA splicing, whereas in the cytoplasm, it regulates mRNA stability and translation 23 . HuR has been implicated in the control of the cell cycle 24 . Chemical inhibition of HuR translocation to the cytoplasm or blocking of HuR binding to its mRNA targets reduced T-lymphocyte activation and proliferation of different cell lines 25,26 . HuR modulates mRNA translation of Myc in different cell types, including B cell lymphoma cell lines [27][28][29] . During B cell activation, HuR is required to protect B cells from cell death induced by reactive oxygen species (ROS) 20 . Additionally, HuR may regulate the magnitude of expression of B cell activation markers required for B:Tfh cell interaction during positive selection 19 . Whether these are relevant and contribute to B cell selection in the GC is unknown.
In this work, we show that HuR has a fundamental function in establishing and sustaining the GC B cell responses. HuR is increased in Myc + LZ GC B cells receiving Tfh cell help, and conditional gene deletion leads to a reduction in Myc expression. HuR is required for the expression of hundreds of genes including those of a Myc-dependent transcriptional programme linked to recycling of antigen-selected LZ GC B cells into the DZ. Additionally, HuR further controls the splicing and expression of mRNA targets required for cell cycle entry and transition through the S phase. In the absence of HuR, antigen-specific class-switched GC B cells are severely impaired showing diminished proliferation and increased DNA damage and GC B cell death. Thus, HuR has an important function in GC B responses allowing selection and expansion of antigen-specific GC B cell clones that confer immunity.

Results
GC-dependent humoral responses are impaired in the absence of HuR. To bypass the activation and metabolic defects leading to the death of follicular HuR knockout (KO) B cells upon cell activation 20 and to study the intrinsic role of HuR in GC B cells, we generated Elavl1 fl/fl Aicda-Cre mice (hereafter named as HuR fl/fl AID Cre mice). We used an Aicda-Cre transgenic mouse line expressing a polycistronic mRNA in which the coding sequences of a Cre recombinase and a human CD2 reporter are linked by an internal ribosome entry site (IRES). Analysis of hCD2 and HuR expression in HuR fl/fl AID Cre mice showed progressive and efficient gene deletion in GC B cells after establishment of the GC reaction ( Supplementary Fig. 1a, b) allowing us to study the intrinsic functions of HuR in GC B cells and GCmediated humoral responses.
Immunisation of HuR fl/fl (control) and HuR fl/fl AID Cre mice with the thymus-dependent antigen 4-hydroxy-3-nitrophenylacetyl hapten conjugated to Keyhole Limpet Hemocyanin (NP-KLH) in alum followed by analysis of NP-antigen-specific antibodies in the serum of these mice showed that HuR was required for highaffinity, class-switched antibody production. Quantitation of NP23-binding IgM showed no differences between control and HuR fl/fl AID Cre mice, while NP23-binding IgG1 was reduced~5fold in HuR fl/fl AID Cre mice after both primary and secondary immunisation (Fig. 1a). High-affinity NP2-binding IgG1 antibodies increased over time during the immune response in control mice. However, the presence of NP2-binding IgG1 in the serum of HuR fl/fl AID Cre mice was negligible and impeded calculation of antibody titres at all the time points analysed (Fig. 1b). Quantitation of the number of antibody-secreting cells (ASCs) producing NP-specific antibodies in the spleen and bone marrow (BM) of these mice showed a 5-10-fold reduction in the number of NP23binding IgG1 ASCs present in HuR fl/fl AID Cre mice. NP2-binding IgG1 ASCs were not detected in HuR fl/fl AID Cre mice (Fig. 1c). Thus, HuR is required for the generation of ASCs producing highaffinity and class-switched antibodies.
HuR is required for the expansion of GCs. Analysis of GC responses in the spleen and lymph node (LN) of mice immunised with NP-KLH in alum showed that the percentage and number of GC B cells at day 7 post-immunisation was reduced 2-fold in HuR fl/fl AID Cre mice compared to control HuR fl/fl mice or AID Cre mice ( Fig. 2a and Supplementary Fig. 1c-e).
Progression of the GC reaction was severely impaired in HuR fl/fl AID Cre mice, with a 7-fold and a 10-fold reduction in the percentage and number of GC B cells compared to control mice at day 10 and day 14 following immunisation (Fig. 2a). By contrast, early establishment of GC responses was not affected in HuR fl/fl AID Cre mice as the percentage and number of GC B cells found in the draining LNs of these mice at day 3.5, 4.5 and 5.5 after immunisation were similar to those found in control mice ( Supplementary Fig. 1d, e). This was likely a reflection of the time required for transcriptional activation of the AID-Cre transgenic construct and progressive depletion of HuR in GC B cells from HuR fl/fl AID Cre mice after assembly of GCs ( Supplementary  Fig. 1a, b). Thus, to assess the requirement of HuR for in vivo B cell activation and development of GCs, we analysed early GC responses in HuR fl/fl Mb1 Cre mice in which deletion of HuR takes place at the pro B cell stage of development. Compared to control mice, the percentage and number of GC B cells found in the draining LNs of HuR fl/fl Mb1 Cre mice were reduced by 2.5-fold and 9-fold at day 4.5 and 5.5 post-immunisation with NP-KLH in alum, respectively ( Supplementary Fig. 1d, e). GC responses remained severely impaired in HuR fl/fl Mb1 Cre mice at day 7.5 post-immunisation as previously reported 20 . Taken together, these data suggest that HuR has an important function in the establishment, expansion and/or the maintenance of GCs.
Further analysis of GC responses in HuR fl/fl AID Cre mice revealed an~2.5-fold reduction in the percentage and number of total IgG1 + GC B cells compared to control mice at day 7 after immunisation ( Fig. 2b). However the percentage of IgG1 + , IgE + and CD138 + B cells generated after co-culturing naive B cells with 40LB stroma cells 30 was similar in the presence or absence of HuR ( Supplementary Fig. 2a). This indicates that HuR is likely dispensable for Ig class-switch recombination (CSR) and terminal differentiation into plasma cells.
Further characterisation of GC B cells based on cell surface expression of markers of DZ and LZ GC cells revealed a reduction in the percentage of DZ GC B cells and a corresponding increase of LZ GC B cells in HuR fl/fl AID Cre mice compared to control mice at day 7 and day 14 following immunisation ( Fig. 2c and Supplementary Fig. 2b). Visualisation by confocal microscopy of GCs in the spleen of immunised mice at day 7 confirmed a reduction in GC size and the altered GC cell distribution between the DZ and the LZ in the absence of HuR (Fig. 2d).
The proportion of apoptotic GC B cells labelled ex vivo with a viability dye for flow cytometry, and/or with VAD-FMK that recognise active caspases, was increased between 1.5-and 3-fold in HuR fl/fl AID Cre mice compared to control mice ( Fig. 2e and Supplementary Fig. 2c). This was not a consequence of altered mitochondrial metabolism as the analysis of mitochondrial mass, mitochondrial potential or ROS levels showed no changes in GC B cells from HuR fl/fl AID Cre mice compared to control mice immunised with NP-KLH in alum (data shown as mean ± SD, n = 7 mice per group, two-way ANOVA and Bonferroni's post-test, **p < 0.01, ***p < 0.001). b Serum dilution curves for quantitation of high-affinity NP2-IgG1 in blood serum from mice described in a (data shown as mean ± SD, NI = not immunised). c Numbers of NP23-IgM, NP23-IgG1 and high-affinity NP2-IgG1-specific antibody-secreting cells (ASCs) in the spleens of HuR fl/fl and HuR fl/fl AID Cre mice at day 7 and day 14 after immunisation (representative data from 1 of 3 independent experiments, n = 4-9 mice per group per experiment, two-sided Mann-Whitney test). Source data are provided as a Source data file.  Fig. 2c), but it was increased 1.4-fold at day 10 ( Fig. 2e) and 5-fold at day 14 in HuR fl/fl AID Cre mice ( Supplementary Fig. 2c). Taken together, HuR is required for HuR is needed for NP-specific GC B cell selection and expansion. The increased death of LZ GC B cells in HuR fl/fl AID Cre mice suggested a role for HuR during selection and expansion of NP-specific B cell clones. Consistent with this, analysis of the percentage of NP + GC B cells showed a 1.6-fold reduction in HuR fl/fl AID Cre mice compared to control mice at day 10 after immunisation (Fig. 3a). The percentage of IgG1class-switched NP + GC B cells was reduced even further with a 7-fold reduction in HuR fl/fl AID Cre mice compared to control mice. By contrast, the percentage of IgG1-class-switched NP − GC B cells was unaltered, whereas the percentage of non-classswitched NP − GC B cells was significantly increased in HuR fl/fl AID Cre mice ( Fig. 3a and Supplementary Fig. 3a). Importantly, analysis of the appearance of NP + IgG1 + B cells in early GCs formed at days 3.5 and 5.5 did not detect any significant differences in HuR fl/fl AID Cre mice compared to control mice (Supplementary Fig. 3b, c) suggesting that HuR is required primordially for the expansion of antigen-specific class-switched B cells in GCs.
HuR was efficiently deleted in over 90% of GC B cells from HuR fl/fl AID Cre mice at day 7 after immunisation ( Fig. 3b). However, only~70% GC B cells from HuR fl/fl AID Cre mice lacked the expression of HuR at day 14, suggesting a depletion of HuRinsufficient B cells as the GC reaction progresses. To interrogate whether HuR deletion confers a competitive disadvantage to GC B cells, we reconstituted the immune system of Rag2 KO mice by transferring BM from wild type (CD45.1 + ) and HuR fl/fl (HuR + CD45.2 + ) or HuR fl/fl AID Cre (HuR − CD45.2 + ) mice in a ratio of 1:3 ( Supplementary Fig. 3d). Analysis of GC responses in these chimeric mice at day 7 and day 10 after immunisation with NP-KLH showed a progressive reduction in the percentage of HuR − CD45.2 + GC B cells and a concomitant increase of HuR + CD45.1 + GC B cells over time (Fig. 3c). Analysis of the distribution of GC B cells in these chimeric mice showed a 2-fold increase in the percentage of HuR − CD45.2 + GC B cells into the LZ compared to controls (Fig. 3d) reinforcing the evidence that HuR expression was required for B cell distribution within GCs. Additionally, we observed that antigen-specific classswitched HuR − GC B cells were highly reduced with NP + IgG1 + HuR − CD45.2 + B cells representing only a 10% of the total number of NP + IgG1 + B cells present in GCs at day 10 ( Fig. 3e). This was a 7-fold reduction in the percentage of NP + IgG1 + CD45.2 + GC B cells found in mice reconstituted with control cells and prompted us to investigate whether selection of antigenspecific GC B cells was affected in the absence of HuR. To this end, we crossed HuR fl/fl AID Cre mice with a Myc GFP/GFP (GFP-Myc) reporter mouse 31 . Quantitation of GFP-Myc + GC B cells revealed a 1.5-fold reduction in percentage and a 2-fold reduction in cell number in HuR fl/fl Myc GFP/GFP AID Cre compared to control mice (Fig. 3f). Taken together, our data indicate that HuR expression is required for the selection, expansion and/or maintenance of antigen-specific GC B cells and modulates their distribution within the DZ and LZ compartments.

HuR expression is increased in positively selected GC B cells.
To define more precisely how HuR is involved in the selection of GC B cells, we sought to understand further the regulation of HuR in GC B cells. To this end, we reanalysed previously published RNAseq data sets that identified three different subsets of LZ GC B cells with different BCR affinity and capacity to establish stable contacts with Tfh cells 7 . Elavl1 mRNA (which encodes for HuR) was significantly increased 4-fold in Bcl6 hi CD69 hi LZ GC B cells with lower-affinity BCRs and expressing high levels of Myc and Tfap4 mRNA ( Fig. 4a and Supplementary Fig. 4a). Similarly, analysis of the transcriptome of sorted Myc + AP4 + LZ GC B cells 32 showed a 2-fold increase in the expression of Elavl1 mRNA when compared to Myc − AP4 − LZ GC B cells ( Supplementary  Fig. 4b). These data suggest that HuR mRNA increases in LZ GC B cells following T cell help.
Consistent with this, we found Elavl1 mRNA to be increased in B cells co-cultured with Tfh cells compared to naive or GC B cells (Fig. 4b). Myc and Tfap4 mRNAs were also increased in these B cells indicating that they were efficiently receiving help from Tfh cells. Importantly, blocking help provided by Tfh cells by adding Tfr cells into the co-cultures prevented the induction of Elavl1, Myc and Tfap4 mRNA. Activation through the CD40 receptor and the BCR rapidly increased the expression of HuR in B cells in vitro ( Supplementary Fig. 4c). Thus, to confirm that HuR protein expression was augmented in LZ GC B cells that received a positive selection signal, we immunised Myc GFP/GFP reporter mice with NP-KLH in alum. HuR protein abundance was similar between DZ and LZ GC B cells ( Fig. 4c) but increased 2-fold in GFP-Myc + GC B cells compared to total GC B cells (Fig. 4c) indicating that HuR is induced in GC B cells in response to T cell help. 14 . In cultured cell lines, HuR has been implicated in the posttranscriptional regulation of Myc [27][28][29]33 prompting the hypothesis that HuR controls Myc expression in GC B cells. In support of this, we sought evidence of a direct interaction between HuR and Myc mRNA by performing individual-nucleotide crosslinking immunoprecipitation (iCLIP) in primary B cells. Analysis of HuR:RNA interactions showed that HuR was bound to U-rich elements present in the 3'UTR of Myc in both resting and mitogen-activated B cells (Fig. 4d). This suggested that HuR might be required for the induction of Myc upon GC B cell positive selection by Tfh cells. Indeed, quantification of Myc protein in positively selected GC B cells from HuR fl/fl Myc GFP/GFP AID Cre and HuR fl/fl Myc GFP/GFP mice revealed a significant reduction in the expression of Myc in the absence of HuR (Fig. 4e). Additionally, HuR expression was required for the complete induction of Myc upon in vitro activation of B cells with mitogens ( Fig. 4f). This indicates that HuR plays a direct role in the induction of Myc expression in B cells even in the absence of Tfh cell help.

HuR binds to Myc mRNA and regulates Myc protein expression. Myc protein abundance in LZ GC B cells is proportional to the strength of help received from Tfh cells and it controls growth and division destiny of LZ GC B cells after entry into the DZ
To establish the importance of reduced Myc expression to the phenotype of HuR fl/fl AID Cre mice, we introduced the MYC-IRES-CD2 transgene that is expressed constitutively from the Rosa26 locus 11   LZ GC cell type, but not on genotype ( Supplementary Fig. 5a). We confirmed this was not due to inefficient Cre-mediated deletion of HuR in HuR fl/fl AID Cre GC B cells as we did not detect reads mapped to the floxed exon 2 of the Elavl1 gene in libraries from HuR KO cells (Supplementary Fig. 5b). Thus, we conclude that HuR deficiency does not globally alter the transcriptome of GC B cells but might affect mRNA abundance in a gene-specific manner.
Differential gene expression analysis showed that the expression of >2000 genes changed between DZ and LZ GC B cells (false discovery rate (FDR) < 0.05) (Fig. 5a and Supplementary Data 1).  (Fig. 5c)   as targets of HuR in B cells stimulated through the BCR or the CD40 receptor ( Fig. 5c and Supplementary Fig. 5c), even with a variable depth in the different iCLIP data sets ( Supplementary  Fig. 5d). Further analyses showed that the proportion of HuR mRNA targets was not increased among those genes that were DE in HuR KO GC B cells compared to all genes. Thus, we concluded that HuR selectively controls the abundance of hundreds of mRNAs in DZ and LZ GC B cells.
HuR recognition of RNA regulatory elements in 3'UTRs has been associated with mRNA stabilisation and altered translation into protein. To assess whether the directionality of changes in gene expression in HuR KO GC B cells was associated with the binding of HuR, we grouped genes based on whether their mRNA transcripts were found associated with HuR in the 3'UTR in our HuR iCLIP data. Empirical cumulative distribution function analysis showed that HuR binding to 3'UTRs correlated with an increase in mRNA abundance of its targets in DZ and LZ GC B cells, particularly of those that were more abundantly expressed ( Fig. 5d and Supplementary Fig. 5e, f). Similarly, 21 out of 39 DE genes of which mRNA transcripts were bound by HuR were less abundant in HuR KO DZ GC B cells compared to control DZ GC cells. By contrast, most DE genes that were not targeted by HuR (36 out of 51) were increased in HuR KO DZ GC B cells (Fig. 5e). We did not observe any differences in the direction of changes in gene expression when LZ GC B cells were analysed (Fig. 5e). This was likely due to the expression of most genes being diminished in HuR KO LZ GC B cells compared to control LZ GC B cells, and/or the fact that our iCLIP data from mitogen-activated B cells did not fully recapitulate the transcriptome of LZ GC B cells. In summary, we observed that HuR is required to maintain mRNA abundance in DZ and LZ GC B cells.
RNA splicing controls the production of alternative mRNA transcripts, and if dysfunctional, it can promote RNA degradation by non-sense-mediated mRNA decay. Thus, we assessed the impact of the loss of HuR on alternative mRNA splicing (AS) in GC B cells. Comparison of the transcriptome of DZ and LZ WT GC B cells showed 333 significant AS events affecting to 310 genes (rMATS, FDR < 0.05 and absolute changes in inclusion levels >10%) ( Fig. 6a and Supplementary Data 2). In the absence of HuR, the number of AS events rose by 1.8-fold in DZ GC B cells and by 3-fold in LZ GC B cells ( Fig. 6a and Supplementary Data 2). A total of 595 splicing events affecting 489 genes were differentially regulated in DZ KO GC B cells compared to control DZ GC B cells. By contrast, 991 AS events affecting 806 genes were identified in LZ KO GC B cells. Of these AS events, 208 events associated with 154 genes were commonly found in DZ and LZ GC B cells (Fig. 6b). This was just 13% of all 1586 AS events that were detected in HuR KO DZ and LZ GC B cells, which suggested that HuR was involved in mRNA editing programmes that were cell-type-dependent. Integration of our RNA splicing analysis with HuR iCLIP data revealed that >75% of AS genes in DZ and LZ GC B cells were bound by HuR, a fold enrichment compared to the total number of genes targeted by HuR at introns (Fig. 6c and Supplementary Fig. 5g). However, AS was not associated with changes in mRNA abundance in the absence of HuR as only 7 and 14 genes were both alternatively spliced and DE in HuR KO DZ and LZ GC B cells. Taken together, our data show that HuR is an important splicing modulator in DZ and LZ GC B cells.
HuR controls pathways linked to cell cycle and DNA damage in GC B cells. Next, we assessed the biological consequences of the identified changes in splicing and transcript abundance in HuR KO GC B cells. Gene set enrichment analysis (GSEA; Fig. 6d and Supplementary Data 3) showed that changes in splicing in HuR KO DZ B cells associated with RNA and DNA metabolic processes linked to chromatin organisation (FDR = 1.16 × 10 −4 ), mRNA processing (FDR = 3.71 × 10 −4 ), cell cycle (FDR = 7.91 × 10 −3 ) and DNA recombination (FDR = 2.19 × 10 −2 ). In HuR KO LZ GC B cells, changes in splicing were linked to the same biological functions. We also identified gene signatures associated with cellular responses to DNA damage (FDR = 9.04 × 10 −13 ), signal transduction by p53 (FDR = 4.70 × 10 −8 ), cell cycle (FDR = 4.70 × 10 −8 ) and DNA repair (FDR = 6.42 × 10 −8 ). Thus, HuR modulates splicing programmes in GC B cells associated with GC B cell proliferation and the response to DNA damage.
GSEA showed that genes with altered expression in the absence of HuR were associated with similar gene sets to the ones linked with changes in splicing ( Fig. 6e and Supplementary Data 3). Genes that were DE between DZ and LZ GC B cells only in control mice were associated with cell activation (FDR = 6.84 × 10 −4 ), DNA-binding transcription factor activity (FDR = 6.84 × 10 −4 ) and cell cycle (FDR = 4.2 × 10 −3 ). By contrast, gene sets associated with mitochondria metabolism (FDR = 9.48 × 10 −3 ) were enriched in HuR KO GC B cells but we did not observe changes in mitochondrial mass, mitochondrial potential or ROS levels compared to control GC B cells ( Supplementary Fig. 2d). Thus, our data suggest that HuR might be required for activation, proliferation and to control DNA damage in GC B cells.
HuR controls entry and progression through the cell cycle. Myc expression rescues GC B cell clones from apoptosis and sets a transcriptional programme that allows GC B cell transition from the LZ to the DZ as well as entry into the cell cycle 12,14 . In line with the reduced expression of Myc in positively selected HuR KO GC B cells, the global analysis of gene expression changes between LZ and DZ GC B cells showed in the absence of HuR a significant reduction in the expression of a set of genes directly regulated by Myc (Fig. 7a). Further clustering of Myc-dependent genes based on HuR binding to their mRNA transcripts revealed a significant reduction in the expression of these genes in both HuR KO LZ and DZ GC B cells compared to those genes that were not bound by HuR (Fig. 7b). This is consistent with the hypothesis that HuR controls a gene programme dependent on Myc for GC expansion.
GSEA also showed a reduced expression of genes associated with the cell cycle in HuR KO DZ B cells ( Fig. 7c and Supplementary Fig. 6a, b). Detailed analysis of gene expression changes associated with each of the phases of the cell cycle showed that HuR was required for the expression of genes associated with the G1 phase and entry into S phase ( Supplementary Fig. 6c). By contrast, HuR was dispensable for the expression of genes in the S, G2 and M phases as changes in the expression of HuR target genes was similar to those genes that were not bound by HuR (Supplementary Fig. 6c). Rb1, Prim2, Ccnd2 and Ccnd3 were among those regulators of cell cycle progression that were reduced in HuR KO GC B cells, whereas the expression of the cell cycle inhibitors at G1/S phase Cdkn1a and Cdkn1b was increased in these cells ( Fig. 7d and Supplementary Fig. 7a, b). Taken together, our data suggest that HuR is required for GC B cell proliferation by directly controlling the expression of Myc and cell cycle-associated genes required for cell cycle entry and progression.
To establish the function of HuR as a regulator of GC B cell proliferation, we studied by flow cytometry the incorporation of bromodeoxyuridine (BrdU) into proliferating GC B cells from control and HuR fl/fl AID Cre mice immunised with NP-KLH for 7 days (Fig. 7e). The overall percentage of BrdU + GC B cells in HuR fl/fl AID Cre mice was reduced 1.25-fold compared to control mice. This was due to a significant reduction in the percentage of proliferating DZ GC B cells but not LZ GC B cells. Sequential in vivo labelling of GC B cells with 5-ethynyl-2′-deoxyuridine (EdU) and BrdU to follow cell cycle progression showed a decrease in the percentage of HuR KO GC B cells entering S phase as well as a reduction in the transition from early to late S phase (Fig. 7f, g). Consistent with this, HuR also promoted B cell cycle progression in vitro. Proliferation of HuR KO B cells cocultured with 40LB stroma cells was reduced 3-fold compared to control cells (Supplementary Fig. 6d). Analysis of BrdU incorporation and DNA content showed an increased percentage of HuR KO GC B cells remaining in the G0-G1 phase with a lower proportion of cells progressing through S and G2-M phases compared to control cells (Supplementary Fig. 6e). The defect in  Supplementary Fig. 6e), of the Ig isotype expressed by the cells, and of cellular differentiation into plasma cells ( Supplementary  Fig. 6f). In summary, HuR has a fundamental function for the expression of cell cycle genes that timely control the entry and progression of GC B cells through the cell cycle.

HuR limits deamination in GC B cells. GC B cell proliferation is required for Ig SHM and high-affinity antibody diversification.
Our transcriptomics analysis highlighted that HuR might regulate the splicing of genes involved in DNA recombination and DNA damage repair (DDR) in GC B cells (Supplementary Data 4) without affecting mRNA abundance between HuR KO GC B cells and control cells (Fig. 8a). However, HuR KO DZ B cells failed to increase the expression of Chek2 and other genes of the DDR associated specifically with the G1 phase of the cell cycle ( Fig. 8a and Supplementary Fig. 7b). This defect could not be ascribed to a global regulation of genes selectively bound by HuR. Additionally, a gene signature associated with deamination activity was increased both in DZ GC B cells and LZ GC B cells from HuR fl/fl AID Cre mice compared to controls (Fig. 8b, c). Consistent with these observations, the quantitation of somatic mutations in the JH4 intron revealed an increased in the incidence and number of mutations in GC B cells from HuR fl/fl AID Cre mice compared to controls (Fig. 8d, e). Aicda mRNA expression was not increased in HuR KO GC B cells compared to control cells, but other family members such as Apobec1 were significantly increased in the absence of HuR (Supplementary Fig. 7b).
Next, we assessed further DNA damage in HuR KO GC B cells. The expression of the DNA damage sensor Trp53 was not detectably increased in HuR KO GC B cells compared to control GC B cells. However, the expression of Cdkn1a and Cdkn1b, two genes induced upon p53 activation in response to DNA damage, was elevated in HuR KO DZ GC B cells (Supplementary Fig. 7b). Quantitation of the DNA damage marker phospho-histone H2A.X (pS319) by flow cytometry also showed a significant increase in HuR KO GC B cells compared to control cells (Fig. 8f). This increase was significant in LZ GC B cells and classswitched antigen-specific GC B cells (Fig. 8g and Supplementary  Fig. 7c). Thus, we conclude that HuR is required for optimal expression of genes involved in SHM and DDR, and in its absence, GC B cells fail to restrict DNA damage leading them into apoptosis.

Discussion
RNA interactome capture 34 has described >1500 RBPs encoded in the mammalian cell genome 35 . However, the importance of regulation of gene expression by RBPs in the humoral response is poorly understood. This study shows the importance of the RBP HuR in initiating and maintaining GC responses and the production of high-affinity antibodies. HuR is dispensable for CSR and the terminal differentiation of B cells, but selection, maintenance and expansion of antigen-specific GC B cells is dependent upon HuR actions in GC B cells.
Our data provide evidence that HuR is required for the full expression of Myc and Myc-associated transcriptional programmes that allow migration of LZ GC B cells to the DZ and prepares LZ GC B cells to enter into the cell cycle. The GC reaction is not sustained in the absence of HuR in a manner that resembles the collapse of GCs in the absence of Myc 11,12 . Myc protein abundance reflects the magnitude of Tfh cell help received by LZ GC B cells during positive selection by Tfh cells 14 . Activation of phosphoinositide-3 kinase (PI3K)-AKT and NF-κB signalling pathways are both required for productive upregulation of Myc in GC B cells 13 . These pathways, or BCR cell surface expression, were not found deregulated in our transcriptomics analyses or by fluorescence-activated cell sorting (FACS). PI3K or NF-κB activation has the potential to induce HuR in tumour cells 36 , and they might be also responsible for the increased HuR expression in positively selected LZ GC B cells. It is likely that different transcriptional and post-transcriptional mechanisms cooperate to induce the expression of Myc and Myc-dependent genes in LZ GC B cells. We have previously shown that PTBP1 is required for mRNA biogenesis and GC B cell proliferation induced by Myc 18 . HuR associates with RNA transcripts from 134 Myc-regulated genes in B cells, in which overall mRNA abundance was increased in DZ B cells compared to LZ B cells and significantly reduced in the absence of HuR. By contrast, mRNAs not bound by HuR did not change significantly between HuR KO DZ and LZ B cells. This highlights that HuR, PTBP1 and, maybe, others 37 could be part of a more extended post-transcriptional programme that acts downstream of gene transcription for global remodelling of the transcriptome and sets the fate of GC B cells upon positive selection.
HuR might control Myc mRNA translation without affecting Myc mRNA abundance in GC B cells, although this needs further exploration. The molecular mechanism by which HuR controls Myc mRNA translation is cell-and context-dependent. In cell lines, HuR cooperates with microRNAs let-7 or miR-17-19b to block Myc mRNA translation 27,28 . However, HuR promotes Myc mRNA translation under hypoxia 33 or in the presence of polyamines 29 . HuR is extensively regulated at the posttranslational level by multiple mechanisms. Caspase-dependent cleavage 33 and phosphorylation by SphK1 and Chk2 29,38 have been suggested as potential mechanisms controlling HuRdependent Myc mRNA translation and cell proliferation. Thus, it is possible that Tfh:B cell encounter not only induces a coordinated expression of Myc and HuR but also promotes HuRdependent Myc mRNA translation.
Overexpression of Myc failed to rescue the GC reaction in the absence of HuR, highlighting the multiple functions of HuR in GC B cells. Our transcriptomic analysis revealed that HuR controls splicing of hundreds of mRNAs that can modulate GC responses in a Myc-independent manner. This is in line with our    Numerous studies have reported the importance of HuR in cell proliferation and apoptosis 41,42 . HuR is increased in late S and G2-M compared to non-cycling GC B cells remaining at G0-G1. Moreover, HuR-dependent RNA binding, stabilisation and translation are linked to HuR subcellular location and phosphorylation. HuR shuttles from the nucleus to the cytoplasm during the cell cycle 43 . This process is controlled by the cell cycleassociated kinase Cdk1 44 and the DDR kinase sensor Chk2 29 . Induced deletion of HuR with tamoxifen increases sensitivity to gamma irradiation and cell death 42 . HuR regulates the G1/S cell cycle-regulator Cdkn1a 45 and the DNA damage checkpoint master regulator p53 46,47 . These checkpoints might be inappropriately active in GC B cells in the absence of HuR, as suggested by the increased expression of Cdkn1a and Cdkn1b, triggering GC B cell death.
AID activity and cell cycle progression are closely linked. Nuclear AID is highly stable at the G1 phase when DNA damage is mostly repaired by error-prone mechanisms 48 . Increased DNA deamination and altered DDR in the absence of HuR can have a negative effect over cell cycle progression and cell viability of GC B cells. DNA damage of Ig genes in DZ GC B cells promotes cell cycle arrest at G1-S and, if unresolved, can trigger cell apoptosis 49 . Our iCLIP data show that HuR binds to Aicda and Apobec1 mRNAs and has the potential to regulate the expression of the latter. Aicda mRNA abundance remained constant in HuR KO GC B cells and the low degree detection of Aicda exon 2 skipping was most likely due to the AID Cre transgene used for HuR deletion. Overactivation of AID is commonly found in GCderived B cell lymphomas and it is associated with genome instability and increased mutagenesis 50 . The increased mutagenesis in the JH4 intron reflects hyperactivation of deaminases in HuR KO GC B cells, although the rapid loss of antigen-specific GC B cells in HuR fl/fl AID Cre mice prevented us from assessing whether increased mutagenesis contributed to affinity maturation. Similarly, the increased phosphorylation of H2A.X and apoptosis and reduced expression of Chk2 in HuR KO GC B cells suggests that HuR is a key modulator in synchronising mutagenesis, DDR and cell cycle progression. These results are in line with previous observations in which HuR is at the centre of an ATM kinase-dependent regulatory mechanism that allows DDR in B cells 51 .
In summary, here we present evidence for the essential function of post-transcriptional regulation by HuR in GC maintenance and high-affinity antibody production. The interaction of GC and Tfh cells during positive selection is likely to trigger important transcriptional and post-translational changes inducing HuR expression and possibly affecting its subcellular location, binding and function. In a possible model, HuR is likely part of an extended network of RBPs that controls mRNA biogenesis, stability and translation in GC B cells. Conditional deletion of HuR in GC B cells leads to ample changes in mRNA splicing and abundance of hundreds of genes. Among these genes, we identified that HuR is required for the expression of Myc and a Mycdependent programme that enables antigen-based GC B cell selection by Tfh cells. HuR also modulates the expression of cell cycle genes required by GC B cells to progress from the G1 to the S phase of the cell cycle. It is then possible that the extended permanence of cells into the G1 phase leads to an increased residence of AID in the nucleus and mutagenesis 52 . This, in combination with high expression of Apobec1 in HuR KO GC B cells, will result in increased DNA damage and cell death and the loss of antigen-specific GC cells that differentiate into plasma cells producing high-affinity antibodies. Deeper study of how HuR controls the expression of important transcriptional and posttranscriptional programmes in the GC reaction is certainly needed and it will open new opportunities to understand further the complex molecular biology of humoral immune responses.

Methods
Mice and animal procedures. Elavl1 tm1dkon (HuR fl/fl ) mice 53 were crossed with Tg(Aicda-cre)9 Mbu (AID Cre ) mice 54 or with CD79a tm1(cre)Reth (Mb1 Cre ) mice 55 . HuR fl/fl AID Cre mice were crossed with Myc tm1Slek (Myc GFP/GFP ) 31 or with Gt(ROSA)26Sor tm13(CAG-MYC,-CD2*)Rsky 11 . B6.Rag2 tm1.1Cgn (Rag2-KO) and C57BL/6-Ly5.1 (CD45.1 + ) mice were kindly provided by Dr. A. Saoudi and Dr. N. Blanchard (both from INFINITy, Toulouse, France). All mice were maintained on a C57BL/6 background strain. The number of animals was decided on the basis of preliminary data and statistical power calculations. Data fulfil a log distribution, the null hypothesis is that the medians of the two samples are identical and assuming a 2-fold difference between means, a standard deviation of 1/3 of the mean value, a ratio between groups of 1 and a type I error rate of 5% we need a minimum of 6-7 mice per group and experiment for a 90% statistical power. Mice of 8-16 weeks of age were used. Littermate control mice (Cre negative) were used in most experiments. No animals were excluded due to a lack of responsiveness to immunisation. Randomisation, but not experimental 'blinding', was set in these studies. All animal procedures at the Babraham Institute were approved by the local animal welfare and experimentation committee and by the UK Home Office. Experiments at INFINITy were approved by the local ethical committee and by the French Ministry of Education, Research and Innovation (project number 16731-2018090610469877v4). No primary pathogens or additional agents listed in the FELASA recommendations have been confirmed during health monitoring surveys of the mouse stock holding rooms maintained in the Babraham Institute Biological Support Unit and at the CREFRE in Toulouse since the opening of these barrier facilities. Ambient temperature was~19-21°C and relative humidity 52%. Lighting was provided on a 12-h light: 12-h dark cycle, including 15 min 'dawn' and 'dusk' periods of subdued lighting. After weaning, mice were transferred to individually ventilated cages with 1-5 mice per cage. Mice were fed CRM (P) VP diet (Special Diet Services) ad libitum and received seeds (e.g. sunflower, millet) at the time of cage cleaning as part of their environmental enrichment.
Immune responses were analysed upon intraperitoneal (ip) or subcutaneous (sc, in the hook) injection of 100 or 20 μg of NP-KLH (Biosearch Technologies) precipitated in aluminium hydroxide gel (Serva). No adjuvant was used to analyse NP-KLH recall responses. Alternatively, immune responses were elicited using sheep red blood cells (SRBCs; 2 × 10 8 SRBCs in Alsevers, ip, TCS Bioscience). For in vivo analysis of cell cycle, animals were injected (ip) with 0.5 mg of EdU (Sigma Aldrich) and 1.5 h later with BrdU (2 mg, Sigma Aldrich). Animals were analysed 2.5 h after first injection. Alternatively, mice were injected with BrdU only and culled 2.5 h later.
In vitro derived GC B cells were generated as previously described 30 . Briefly, 17.5 × 10 3 cells were co-cultured over a monolayer of 40LB stroma cells irradiated (120 Gy) and seeded in a 24-well plate 24 h before. IL-4 (2 ng/ml) was added to the RPMI medium that was replaced every 2 days. At day 4 of cell culture, iGC B cells were recovered and seeded on freshly irradiated 40LB stroma cells. At day 4, IL-4 (2 ng/ml) or IL-21 (10 ng/ml, Peprotech) was added to the media. Analysis of cell cycle was performed after adding 10 μM BrdU 30 min before harvest.
For assessment of cell viability, we incubated LN cells with CaspACE™ FITC-VAD-FMK in Situ Marker (1 μM per 10 7 cells) for 30 min at 37°C in complete RPMI medium. Zombie NIR Fixable Viability was used to stain dead cells and in combination with an anti-mouse Fc Receptor Blocking antibody (clone 2.4G2) for 15 min at 4°C prior to staining of cell surface markers. For intracellular staining, cells were fixed and permeabilised with the BD Cytofix/Cytoperm™ Fixation and Permeabilization Solution from BD Biosciences. EdU staining was carried out after DNA digestion with TurboDNase (ThermoFisher Scientific) using the Click-iT™ EdU Pacific Blue™ Flow Cytometry Assay Kit (Cat. No. C10418, ThermoFisher Scientific) followed by incubation with Anti-BrdU antibody [clone MoBu-1]. BrdU and DNA staining were carried out using the FITC BrdU Flow Kit from BD Biosciences (Cat. No. 559619). Data were collected using a BD Fortessa and analysed using the software FlowJo vX.
Imaging. Spleens from immunised mice was embedded in Tissue-Tek O.C.T. (Sakura Finetek) before snap-freezing and cryopreservation. Ten-micrometre sections were obtained in a Cryostat, fixed with acetone for 15 min at −20°C, dried at room temperature, blocked with phosphate-buffered saline (PBS) with 0.3% Triton X-100 and 5% bovine serum albumin for 1 h. at room temperature. Staining with specific antibodies against IgD (coupled to AF488, 1 in 400), Ki67 (coupled to AF488, 1 in 100) and CD21/35 (coupled to biotin, 1 in 400) was performed overnight at 4°C in a wet chamber. After washing, tissue samples were incubated with streptavidin-PE (1 in 800) for 45 min at room temperature. Sections were then washed (3×) with PBS with 0.3% Triton X100, dried and mounted on slides using ProLong Gold Antifade Mountant (Thermo Scientific). Images were collected in a Leica SP8 confocal microscope using the Leica Application Suite X (LAS X) software and analysed with the FIJI Image J software.
Analysis of SHM. Mutations in the JH4 intronic region were analysed as previously described 18 . Briefly, DNA was extracted from FACS-sorted GC B cells at day 7 after mouse immunisation with NP-KLH alum. GC B cells from individual mice (n = 2 per group) were analysed. JH4 intronic regions were amplified by PCR using PfuUltra II Fusion HS DNA Polymerase (Agilent Technologies), primers JH4 (fwd), TCCTAGGAACCAACTTAAGAGT and JH4 (rev) TGGAGTTTTCT-GAGCATTGCAG (Supplementary Table 2) and 35 cycles of denaturation 95°C (30 s), annealing 57°C (30 s) and extension 72°C (30 s). cDNA was cloned using the Zero Blunt® TOPO® PCR Cloning Kit (Cat. No. 450245, ThermoFisher Scientific). Mutation frequencies were calculated by dividing the total number of mutations identified in each replicate by the total length of amplified DNA (amplicon of 565 bp).
RNA sequencing. Transcriptomics analysis of LZ and DZ GC B cells from HuR fl/fl and HuR fl/fl AID Cre mice immunised with NP-KLH alum for 7 days was performed by FACS cells as previously described 18 and summarised in Supplementary Fig. 8. Briefly, GC B cells were pre-enriched by depleting IgD + , CD3e + , Gr1 + and Ter119 + cells with biotinylated antibodies, anti-biotin magnetic beads and MACS LS columns. GC B cells were labelled with specific antibodies (LZ GC B cells = CD19 + CD38 − CD95 + GL7 + CXCR4 − CD86 + ; DZ GC B cells = CD19 + CD38 − CD95 + GL7 + CXCR4 + CD86 − ). Four biological samples (two from males and two from females) for each genotype were sorted and RNA was isolated using the RNeasy Micro Kit from Qiagen (Cat. No. 74004). RNA quality was analysed on a 2100 Bioanalyzer (Agilent) and libraries were prepared with 50 ng of RNA using the SureSelect Strand-Specific RNA Library Preparation Kit for Illumina (Cat. No. G9691A, Agilent). Libraries were multiplexed and sequenced across three lanes on an Illumina HiSeq 2500 platform (100 bp paired-end mode). Other RNA sequencing data sets used in this study are previously described (GSE109732, GSE82003, GSE80669) 7,32,58 .
In vivo interactions of HuR and RNA in splenic B cells treated with LPS for 48 h has been previously described (GSE62148) 20 . In parallel, naive B cells were treated or not with αCD40+IL4 or αIgM+IL4 to detect HuR physical interactions with Myc 3'UTR. Briefly HuR:RNA interactions were detected using iCLIP and sequencing. After annotation of reads to the mouse genome, nucleotide −1 was annotated as the crosslink binding site. Peak call analysis was performed as described 60 . The number cDNA counts associated with a single crosslink site was quantified based on read length and the sequence of the 5′ barcode. Annotation of crosslink sites into different genomic features (introns, exon, 5'UTR, 3'UTR or CDS) was performed with R (v3.4.1), bioMart (v2.40.4), GenomicRanges (v1.36.0) and Mus_musculus.GRCm38.90.gtf. Only those crosslinks with a FDR < 0.05 and >5 cDNA counts annotated to the same nucleotide were considered as significant (if not stated otherwise).
Gene ontology enrichment analysis was performed using either ToppGene 61 or GSEA, 62 using default settings. Selected gene sets were from Hallmark, Reactome or AmiGO2. Data from our sequencing data sets and gene sets were extracted and plotted in R (v3.4.1) using ggplot2 (v3.2.1).
Statistics and reproducibility. Statistics were performed in R or using Prism. Statistical tests performed are indicated in each figure legend. Briefly, Mann-Whitney or Wilcoxon tests were used for pairwise comparisons; analysis of variance was used when having more than two groups to compare. Two-sided analyses were performed in all cases. Kolmogorov-Smirnov test was used to assess equality of one-dimensional probability cumulative distributions. Benjamini and Hochberg test was used for multiple testing and FDR calculation (if not stated otherwise). All experiments were reproduced at least twice successfully and whenever possible.