The RNA-binding protein HuR (Elavl1) is essential for the B cell antibody response

HuR expression is increased upon B cell activation

HuR is the only member of the Elavl family expressed in splenic B cells (Fig. 1a). In order to understand the role of HuR during B cell development and activation we generated B cell-specific conditional KO mice (Elavl1^fl/fl Mb1-Cre, hereafter HuR-cKO mice). Analysis of HuR protein expression by flow cytometry showed that it was deleted from the pro-B cell stage in the bone marrow (BM) and was undetectable in transitional and mature B lymphocytes (Fig. 1b-d). HuR was similarly expressed in all B cell populations present in the spleen. RT-qPCR analysis of GC B cells, formed in the spleen after immunisation, showed a 6-fold increase in HuR mRNA in GC B cells compared to non-GC B cells (Fig. 1e). Intracellular flow cytometry indicated that HuR protein was increased by 2-fold in GC B cells (Fig. 1f). B cell activation with LPS or αCD40 + IL-4 + IL-5 increased HuR protein expression by up to 3-fold (Fig. 1g). Taken together, these data suggest that HuR expression is increased during B cell activation in vitro and in vivo.

B cell development following HuR deletion in pro-B cells

Normal proportions of B cell subpopulations were present in BM, spleen and lymph nodes of HuR-cKO mice (Fig. 2a-c). We found a small decrease in the number of pre-B cells in BM and in the number of transitional and FO B cells in spleens of HuR-cKO mice. HuR-cKO mice had reduced numbers of B1 B cells in the peritoneal cavity (Fig. 2d). Analysis of the incorporation of 2-bromodeoxyuridine (BrdU) into the DNA of rapidly proliferating pre-B cells was not different in HuR-cKO mice (Fig. 2e and Supplementary Fig. 1a and 1b). Furthermore, B cell turnover in the spleen, inferred from BrdU incorporation over a 7-day period, was not different from control mice (Fig. 2f). Thus, with the exception of B1 B cells, there is little requirement for HuR for either B cell development or homeostasis.

B cells require HuR for normal antibody responses

HuR-cKO mice showed a significant reduction in all immunoglobulin (Ig) isotypes present in serum, apart from IgA (Fig. 3a). IgG₁ and IgG_2b titers were reduced by 2.5-fold in HuR-cKO mice, whereas IgG₃ and IgM titers were reduced by 6- and 12-fold, respectively. Decreased IgM titers could result from the reduction of B1 B cells, as these contribute substantially to serum IgM. The biggest difference was found in IgG_2c titers that were reduced by 200-fold in HuR-cKO mice. The decrease in almost all antibody isotypes likely reflects a requirement for HuR in B cells for a functional response following antigen encounter.

To investigate the requirement for HuR during the response of B cells to antigen, we immunised HuR-cKO mice with the thymus (T)-independent antigens, Ficoll or LPS coupled to the hapten 4-hydroxy-3 nitrophenolic acid (NP). NP-reactive IgM titers in serum showed a 5-fold reduction with both antigens (Fig. 3b and 3c). By contrast, we were unable to detect NP-reactive IgG₃ antibodies in serum from HuR-cKO mice. Next, we evaluated the ability of HuR-cKO B cells to participate in a thymus-dependent antibody response by assessing the serum antibody responses following primary and secondary immunisation with NP coupled to keyhole limpet haemocyanin (KLH) (Fig. 3d). Serum NP-reactive IgM titers showed an average 8.5-fold reduction. NP23-reactive IgG₁ titers, which reflect the combination of both low and high affinity antibodies, were 15-times lower in HuR-cKO mice than in control mice. Memory responses evaluated seven days after secondary immunisation were clearly detectable in HuR-cKO mice in the form of IgM antibodies. By contrast, NP2-reactive IgG₁ titers did not increase, suggesting high affinity IgG₁ memory cells did not form, or, if they did, their reactivation failed to generate ASC. As a measure of affinity maturation, we determined the ratio of NP2- and NP23-reactive IgG₁ antibodies, which indicated that NP-reactive antibodies in HuR-cKO mice failed to undergo affinity maturation (Fig 3e).

Affinity maturation takes place in GC, where helper T cells select B cells with the highest affinity for antigen. GC formation after seven days of immunisation with NP-KLH was diminished in HuR-cKO mice. The number of GC B cells was reduced by more than 6-fold (Fig. 3f). Similarly, following administration of highly immunogenic sheep red blood cells the GC response was defective over time in HuR-cKO mice (Supplementary Fig. 1c and 1d). Consistent with reduced titers of serum antibodies, the number of NP-reactive IgM ASC in the spleen of HuR-cKO mice immunised with NP-KLH was reduced by 2-fold, whereas the number of NP-reactive IgG₁ ASC was reduced by 50-fold (Fig. 3g). A similar reduction was found in the number of NP-reactive IgM and IgG₁ ASC in the spleen of HuR-cKO mice seven days after secondary immunisation with NP-KLH (Fig. 3h). Further analysis of ASC in the BM showed that HuR-cKO mice contained a similar number of NP-reactive IgM ASC, but NP-reactive IgG₁ ASC were barely detectable (>100-fold reduction) when compared to control mice (Fig. 3i). These data show that HuR is required for proper B cell activation and/or differentiation following antigen encounter.

Regulation of the B cell transcriptome by HuR

To examine whether an early stage of B cell activation was affected by the absence of HuR, we measured changes in intracellular Ca²⁺ following crosslinking of cell surface IgM. The Ca²⁺ flux elicited after activation of Mb1-Cre unfloxed controls (defined as “Ctrl.”) and HuR-cKO LN B cells was similar using either a highly crosslinking anti-IgM F(ab’)₂ or a monoclonal anti-IgM (mAb B7.6) (Supplementary Fig. 2a). Additionally, cell surface expression of CD25, CD40, CD69 and CD86 was increased to a similar extent in response to a variety of stimuli (Supplementary Fig. 2b and 2c), indicating that, by these criteria at least, B cell activation was normal in the absence of HuR.

To discover the molecular mechanisms regulated by HuR in resting and activated B cells, we performed an integrated analysis of three high throughput measurements of the B cell transcriptome. Firstly, we searched for differential mRNA expression in HuR-cKO B cells by mRNAseq. Secondly, we determined the differences in mRNA translation between HuR-cKO and control B cells using ribosome footprinting (Ribo-Seq). Thirdly, we characterised HuR-RNA interactions at the single nucleotide level in LPS-activated B cells using iCLIP. Taken together, these measurements have the power to discriminate between changes in RNA isoform and abundance as well as translational regulation, and can relate these changes to HuR binding. Thus detected changes can be ascribed to be direct or indirect effects of HuR:RNA interactions. None of the changes were due to compensatory regulatory mechanisms affecting the abundance or ribosome loading of mRNAs encoding other members of the Elavl-family or of other AU-rich element binding proteins (Supplementary Fig. 3).

Differential expression analysis of mRNAseq and Ribo-Seq followed by pathway enrichment analysis indicated gene signatures related to different aspects of the cell cycle and energy metabolism were enriched in HuR-cKO B cells (Table 1). HuR deficiency influenced expression of genes related to glycolysis, TCA cycle and oxidative phosphorylation, three pathways increased in GC B cells and after B cell activation with LPS (Supplementary Fig. 4a, 4b and 4c). Blockade of these energy pathways using 2-deoxyglucose (2DG), CPI-613 or Oligomycin A decreased B cell survival, cell growth and cell proliferation after mitogen stimulation (Supplementary Fig. 4d and 4e). Analysis of the global fold-change in mRNA expression and translation of genes related to glycolysis, TCA cycle and oxidative phosphorylation showed no change when comparing ex vivo (unstimulated) control versus HuR-cKO B cells. By contrast, global expression and/or translation of these mRNAs were increased in LPS-activated HuR-cKO B cells, suggesting that HuR is only important for regulating the reprogramming of energy metabolism upon activation (Supplementary Fig. 4f). Data correlation between mRNAseq and Ribo-Seq of only those metabolic genes that are differentially translated in LPS-activated HuR-cKO B cells showed that all of them, with the exception of dihydrolipoamide S-succinyltransferase (Dlst), were modestly increased (Fig. 4a and Supplementary Fig. 4g, 4h and 4j). Dlst mRNA was increased in GC B cells when compared to naive B cells (Supplementary Fig. 4i), but its mRNA expression and translation was significantly reduced in LPS-activated HuR-cKO B cells (Fig. 4b).

Dlst is one of the three subunits of the αKGDH enzymatic complex, which is essential for maintaining tricarboxylic acid (TCA) cycle flux and cell energy supply. In order to understand the role of HuR in Dlst mRNA regulation, we examined mRNAseq data and plotted the reads mapped across the Dlst locus as Sashimi plots (Fig. 4c). These mRNA splicing profiles showed that a single mRNA transcript was generated after RNA splicing in ex vivo and LPS-activated control B cells. In the absence of HuR, Dlst mRNA showed two alternative splicing events: intron 10 retention and alternative inclusion of a cryptic exon between exon 10 and 11. iCLIP data showed that HuR binds to several locations along Dlst RNA (Fig. 4c and Supplementary Fig. 5a-c). Peak calling analysis showed that HuR binds preferentially to introns, including the poly-pyrimidine tract found downstream the 3′ splice site of the cryptic exon present within intron 10 (Supplementary Fig. 5d). Taken together, these data demonstrate that HuR binding to Dlst pre-mRNA might promote Dlst mRNA expression and translation in HuR-cKO B cells. The modest change in translation of other components of cell energy pathways may reflect a compensatory mechanism.

HuR binding to introns modulates alternative intron usage

To gain a mechanistic insight into the role of HuR in mRNA splicing in B cells we further examined the HuR iCLIP data obtained from LPS-activated B cells. Analysis of unique read counts in all three iCLIP experiments showed that 75% of HuR-RNA crosslink sites were mapped to introns (Fig. 5a and Supplementary Fig. 5e and 5f). Visualisation of HuR crosslink sites close to the exon-intron boundaries indicated that HuR preferentially binds to introns, and showed a significant binding enrichment between the branch point and the 3′ splice site (Fig. 5b). These data suggested that HuR might be a splicing regulator in B cells, thus we studied whether HuR modulates pre-mRNA splicing by further analysis of mRNAseq data from LPS-activated B cells. Differential exon analysis using DEXSeq did not reveal significant changes in exon usage of protein coding transcripts in the absence of HuR, and failed to identify the alternative splicing events associated with Dlst mRNA (Supplementary Tables 1-5). Thus, we performed an intron-centric analysis of the mRNAseq data (Supplementary Fig. 6a), which showed that 530 introns belonging to 375 genes were differentially used in LPS-activated HuR-cKO B cells compared to control B cells (padj<0.1, Supplementary Fig. 6b). HuR was bound to 85% of these 375 genes in, at least, two of the three independent HuR iCLIP experiments (Fig. 5c). Dlst was found amongst these genes. Taken together data correlation from the intron-centric analysis and HuR iCLIP experiments identifies alternative intron usage in the absence of HuR.

HuR modulates mRNA expression and translation via splicing

Expression and translation analysis of all 375 genes with differential intron usage in HuR-cKO B cells (group 1) showed no differences globally (Supplementary Fig. 6c). Individually, 64 genes (group 2) out of these 375 were differentially expressed in LPS-activated HuR-cKO B cells and bound to HuR (Fig. 5d). A similar data correlation showed that 71 out of the 375 genes (group 3) were both differentially translated and bound to HuR (Fig. 5e). Only 25 of these genes (group 4) were both differentially expressed and translated in HuR-cKO B cells (Fig. 5f). When expression of genes in groups 1, 2 and 3 was analysed globally, no changes in mRNA abundance was observed when comparing HuR-cKO versus control B cells (Fig. 5g). By contrast, global translation of these mRNAs was significantly reduced in HuR-cKO B cells, suggesting that, even though HuR-dependent regulation of alternative splicing might not necessarily affect overall mRNA levels, HuR is required for mRNA translation (Fig. 5h). Global mRNA expression and translation of the genes in group 4 were both reduced by up to 50%. Closer examination at individual genes indicated that both mRNA expression and translation of 76% of genes in group 4 (19 out of 25) were reduced in the absence of HuR, including Dlst (Fig. 5i and 5j). In summary, differential intron usage analysis and its correlation with differential expression in mRNAseq and Ribo-Seq allowed us to discover the alternative intron events associated with HuR binding. Visualization of the read coverage revealed high complexity in mRNA splicing associated with these genes, detecting up to 8 different alternative splicing events associated with the solute carrier family 25 (mitochondrial thiamine pyrophosphate carrier), member 19 (Slc25a19) in the absence of HuR (Supplementary Fig. 6d and 6e). Genetic deletion of either Dlst or Slc25a19 has been previously associated with reduced αKGDH enzymatic activity and alterations in the TCA cycle^{23, 24}.

HuR regulates Dlst mRNA splicing

To understand the role of HuR in mRNA splicing, we selected Dlst mRNA for detailed analysis and evaluated usage of intron 10 with two RT-PCR assays. HuR-cKO B cells showed increased intron retention compared to control B cells (Fig. 6a). Read coverage analysis of the Dlst gene locus using mRNAseq data from HuR-cKO B cells identified a novel alternative exon splicing event located after exon 10. This alternative exon 10b was previously annotated in Ensembl as the start exon of predicted transcript ENSMUST00000165575, a transcript for which evidence of its existence is lacking. Detection of reads mapped across the exon 10/alternative exon 10b junction suggested that this was an alternative exon inclusion rather than an alternative first exon. RT-PCR validated this observation and indicated that over 50% of Dlst transcripts within HuR-cKO B cells included the alternatively spliced exon 10b (Fig. 6b). mRNAseq data showed little difference in Dlst mRNA expression between ex vivo or LPS-activated HuR-cKO and control B cells (Fig. 6c). However, the number of ribosome footprints mapped to Dlst mRNA was reduced by 2-fold in HuR-cKO B cells, indicating reduced mRNA translation (Fig. 6d). Visualization of ribosome footprints as Sashimi plots failed to identify ribosome-protected reads mapped to the alternative exon 10b, suggesting that transcripts containing this exon were not translated (Supplementary Fig. 7a). Consistent with this, inclusion of alternative exon 10b introduces several stop codons into the coding sequence (Supplementary Fig. 7b). Immunoblot analysis of B cell protein extracts from ex vivo and LPS activated B cells confirmed a 50% reduction in Dlst protein abundance in HuR-deficient B cells (Fig. 6e and 6f). Antibodies reactive with the N-terminus of Dlst could not detect any alternative protein isoforms produced as a consequence of exon 10b insertion in the absence of HuR (Supplementary Fig. 7c). Decreased Dlst protein expression was associated with a reduction in αKGDH enzymatic activity in HuR-cKO B cells (Fig. 6g). Taken together, HuR binding to Dlst intron 10 prevented alternative exon inclusion, thus guaranteeing Dlst mRNA translation into enzymatically active protein.

αKGDH is required for B cell proliferation and Ig CSR

Dlst, as part of the αKGDH complex, is highly regulated in order to maintain the flux through the TCA cycle. Inhibition of the αKGDH enzymatic activity using αKG analogs (succinyl phosphonate (SP) and phosphonoethyl ester of succinyl phosphonate (PESP)) significantly reduced B cell viability (Fig. 7a). Analysis of B cell proliferation following LPS + IL-4 stimulation showed that αKGDH inhibition reduced proliferation in a dose-dependent manner (Fig. 7b and 7c). Further analysis of CSR showed that the number of IgG₁⁺ cells was significantly reduced both after chemical inhibition of αKGDH enzymatic activity or in B cells with a single copy of the Dlst gene (Dlst^+/− B cells) (Fig. 7d and 7e). Thus, αKGDH enzymatic activity is required for in vitro B cell survival, proliferation and CSR.

ROS scavengers rescue B cell proliferation and Ig CSR

Analysis of HuR-cKO B cell cultures showed that cell viability was reduced in the absence of HuR (Fig. 8a and 8b). Viability of HuR-cKO B cells could not be rescued by blocking caspase-dependent apoptosis or programed necrosis using Q-VD-OPh or necrostatin-1 respectively, but was reversed by the H₂O₂ scavenger catalase (Fig. 8b and c). Consistent with this, the fluorescence of the ROS reporter dye H₂DCFDA showed that viable HuR-cKO B cells produced larger amounts of ROS compared to Mb1-Cre⁺ controls. Addition of catalase reduced H₂DCFDA fluorescence, indicating diminished ROS concentration (Fig. 8b).

HuR-cKO B cells failed to proliferate when cultured with mitogens in vitro. Media supplementation with catalase as well as with ROS scavengers like sodium pyruvate, N-acetyl cysteine and EUK134 enhanced the proliferation of HuR-cKO B cells in response to a variety of stimuli (Fig. 8d and supplementary Fig. 8). Quantitation of the number of B cells recovered showed that HuR-cKO B cells undergo a similar number of cell divisions; however the number of cells in each division was reduced, indicating that ROS scavengers did not completely restore proliferation (Fig. 8e). ROS scavengers led to a recovery in the percentage and number of IgG1⁺ and CD138⁺ (ASC) HuR-cKO B cells (Fig. 8f and 8g). However NAC and EUK134 marginally inhibited the yield of IgG1⁺ and CD138⁺ cells in control cultures. Taken together, our data indicate that HuR-cKO B cells contained elevated amounts of ROS and these compromised B cell survival. Mitigating ROS allowed the HuR-cKO B cells to proliferate, undergo CSR and form ASC in vitro.

Reagents, antibodies and oligonucleotides

Information is detailed in Supplementary Table 6, 7 and 8.

Mouse strains and animal procedures

Mice on the C57BL/6 background used in this study are: CD79a^{tm1(cre)Reth 51} and Elavl1^{tm1Dkon 16}. Dlst^+/− mice were generated by Lexicon Pharmaceuticals, Inc. (The Woodlands, Texas)²⁴. Immunisations were performed by intraperitoneal administration of Sheep Red Blood Cells (SRBCs, 2 × 10⁸ cells/mouse), NP-Ficoll (25 μg), NP-LPS (25 μg) or alum-precipitated NP-KLH (100 μg). For in vivo experiments, the number of animals was decided based on preliminary studies using small cohorts to assess the magnitude of the changes (2-3 mice per genotype). 6 to 10 mice (12-16 weeks old) were considered a sufficient number of mice in final studies to reach sufficient statistical power analysis. No animals were excluded due to a lack of responsiveness to immunisation. Randomization, but not experimental blinding, was performed in these studies. All animal procedures at the Babraham Institute were approved by the Babraham Institute AWEEC, the UK Home Office. Local Ethical Committee for Animal Experiments of KU Leuven, Belgium approved experimentation with Dlst^+/− mice under license number 000/2014.

ELISA and ELISPOT assays

Serum immunoglobulins and NP-specific antibodies were detected by ELISA⁵². NP-specific antibody end point titers were used as a measure of relative concentration. NP-specific antibody secreting B cells (ASC) were detected by ELISPOT assays performed as previously described⁹.

Flow cytometry

Analysis of B cell populations was performed using specific antibodies for cell surface markers. B cell population analysed in bone marrow were: Pre-Pro-B cells - B220^lo IgM⁻ CD25⁻ c-kit⁺ CD19⁻; Pro-B cells - B220^lo IgM⁻ CD25⁻ c-kit⁺; Pre-B cells - B220^lo IgM⁻ CD25⁺ CD19^lo; immature-B cells - B220⁺ IgM⁺ IgD⁻; and mature-B cells - B220⁺ IgM⁺ IgD⁺. B cell subsets identified in spleen were: T1 - B220⁺ CD93⁻ CD23⁻ IgM^hi; T2 - B220⁺ CD93⁻ CD23⁺ IgM^hi; T3 - B220⁺ CD93⁻ CD23⁺ IgM⁺; MZ - B220⁺ CD93⁺ CD21^hi CD23⁻, FO - B220⁺ CD93⁺ CD21⁺ CD23⁺ and GC B cells - B220⁺CD95⁺PNA⁺.

Fixable Viability Dye eFluor^® 780 or DAPI was used to test cell viability. For intracellular staining of HuR, the BD Cytofix/Cytoperm^™ Fixation/Permeabilization Solution Kit was used. In vitro B cell proliferation was tested by flow cytometry after the labelling of ex vivo B cells with the CellTrace^™ Violet dye. ROS levels in B cells were measured by using OxyBURST^® Green H2DCFDA (1 μM) added into the B cell culture 30 min before cell analysis by flow cytometry.

Calcium flux analysis

Purified LN B cells were loaded with Fluo-4 AM (3 μM) in PBS + 0.5% BSA (5 × 10⁶ cells/ml) for 30 min at RT in the dark. After washing, cells were incubated in 0.5% BSA/PBS containing 1 mM CaCl₂ for further 30 min to allow de-esterification of intracellular Fluo-4 AM ester. Changes in fluorescence emission (at 494 nm) after surface Ig crosslinking were measured using a PerkinElmer LS 55 Fluorescence spectrometer.

B cell isolation and cell culture

B cells from spleen or peripheral lymph nodes were isolated by using the B Cell Isolation Kit from Miltenyi Biotec. B cells were cultured in RPMI 1640 Medium (Dutch Modification) plus 5% FCS, antibiotics, 2 mM L-glutamine and β-mercaptoethanol (5 μM). ROS scavengers, Q-VD-OPh, Necrostatin-1 (Nec-1), 2-deoxy-glucose (2-DG), CPI-613, Oligomycin A, Succinyl phosphonate (SP) and Phosphonoethyl ester of succinyl phosphonate (PESP) were added to the cell culture media when indicated in the figure legend. Mitogen activation of B cells was carried out using either LPS (E.coli 0127:B8), αCD40 (3/23 clone) or αIgM (B7.6 clone) in the presence or not of IL4 and IL5.

Protein separation and Western Blot analysis

Total cell extracts were prepared by incubating cells in RIPA buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% NP-40, 0.1% SDS and 0.5% Sodium deoxycholate) supplemented with protease inhibitors. 10% polyacrylamide-SDS gels were loaded with 20 μg of total protein extracts per lane. Gel electrophoresis and protein transfer to nitrocellulose membranes was performed prior detection of HuR, Dlst and β-actin by Western blot using specific antibodies and enhanced chemiluminescence or IRDye detection.

RNA extraction and RT-PCR assays

Total RNA extraction from purified B cells was performed using TRIzol (LifeTech). 1 μg of RNA was treated with DNase I prior reverse transcription into cDNA. Analysis of Dlst alternative splicing events were performed by RT-PCR using specific primers (table S3). Intron retention was quantified by qPCR whereas alternative exon inclusion was quantified after gel densitometry of the PCR products. qPCR assays were performed using Platinum^® SYBR^® Green qPCR SuperMix (Life Technologies). mRNA expression of Dlst and Elavl genes was quantified by the ΔΔCT method (comparative threshold cycle) and normalised to the expression of 18S rRNA.

Immunofluorescence microscopy

Ex vivo or mitogen- activated splenic B cells were cultured for 20 minutes on cover slips coated with 0.01% Poly-L-Lysine prior detection of HuR by immunofluorescence microscopy as previously described⁵³. Images from three independent experiments were captured with an Olympus FV1000 System and processed using Image-J.

Individual-nucleotide resolution Cross-Linking and ImmunoPrecipitation (iCLIP)

iCLIP experiments were performed as described ²¹, using protein lysates from splenic B cells activated with LPS for 48 hours (n=3). Briefly, splenic B cells were irradiated with UV- light to crosslink protein-RNA interactions (150 mJ/cm², Stratalinker 2400). After washing cells with ice-cold PBS, total cell extracts were obtained using RIPA buffer and sonication (x3). Extract supernatants was treated with RNase I (0.167 U/ml) for 3 minutes at 37°C. Immunoprecipitation of HuR-RNA complexes was performed using 2 μg. of αHuR mAb coupled to protein G dynabeads. B cell extracts from HuR-cKO B cells were used as a negative control. After washing twice with high-salt buffer (50 mM Tris-HCl pH 7.4, 1 M NaCl, 1 mM EDTA, 1% NP-40, 0.1% SDS and 0.5% sodium deoxycholate) and once with PNK washing buffer (20 mM Tris-HCl pH 7.4, 10 mM MgCl₂, 0.2% Tween-20), one tenth of the sample were labelled with ³²P-ATP using PNK, whereas an RNA Linker was ligated to the rest of the sample after RNA dephosphorilation. RNA-protein complexes were separated by electrophoresis (SDS-Page) and transfer to nitrocellulose membranes. HuR-RNA complexes higher than 55 KDa were marked after visualization by autoradiography. RNA extraction was performed by incubating the nitrocellulose fragment at 37°C for 10 minutes with proteinase K in PK buffer (100 mM This-Cl pH 7.5, 50 mM NaCl and 10 mM EDTA). 200 μl of PK buffer containing urea (7 M) was added to each sample and further incubated at 37°C for 20 minutes. RNA was isolated by phenol/clorophorm extraction and ethanol precipitation. RNA was retro transcribed into cDNA using RCLIP primers and SuperScript III reverse transcriptase. After cDNA purification using 6% TBE-urea gels, cDNA was circularised and amplified by PCR using Solexa P5/P7 primers. cDNA libraries from three independent experiments were prepared and analysed. RCLIP primers included a 7 bases long barcode at the 5′end (three known bases plus four unknown nucleotides). Additional two bases (AT) were added to the 3′end. Barcodes allowed us to multiplex different cDNA libraries for Illumina sequencing and identify PCR duplicate reads.

High throughput sequencing and library preparation

mRNAseq libraries were obtained using TruSeq Stranded mRNA Sample Prep Kit (Illumina Inc). Splenic B cells from individual HuR^f/f control or HuR-cKO mice were independently processed for RNA extraction (ex vivo samples, n=4 per genotype) or were stimulated with LPS for 48 hours in RPMI media plus 1 mM NaPyr (n=3-4 per genotype).

Previously published mRNAseq libraries were used to analyse the expression of genes involved in glycolysis, TCA cycle and electron transport in naive and GC B cells (GEO deposition number - GSE47705)⁵⁴.

Ribosome footprinting profiling (Ribo-Seq) assays were performed using ARTseq^™ Ribosome Profiling Kit (Epicentre, Illumina). Ex vivo or LPS-activated B cells (n=4-5 per genotype and condition) were treated with cyclohexamide (CHX, 100 μg/ml) three minutes prior cell extract preparation. cDNA libraries were sequenced using either GAIIX (iCLIP) or HTSeq2000 (mRNAseq and Ribo-Seq) Illumina technology. The type of sequencing performed was: iCLIP, 40 bp SE; mRNAseq, 100 bp SE; and Ribo-seq, 50 bp SE.

Bioinformatics and statistics

Mapping and peak call analysis of HuR iCLIP data was performed as described previously ²¹. Briefly, RCLIP primers were used to generate the cDNA libraries from HuR iCLIP experiments. Data demultiplexing involved identification of the three known bases of the 5′ barcode. Quality analysis of sequencing data was done using FastQC. Data analysis was done in the iCount pipeline. First, the four unknown bases of the 5′ barcode were used to remove PCR duplicate reads. 3′ barcode (AT) was used to define the length of the reads. Both parameters were important to identify unique reads. Then, reads were trimmed to remove 5′ and 3′ barcodes along with any adaptor sequence before mapping to the genome using Bowtie. Data was originally annotated to mm9 genome, but was lifted over to mm10 genome using Galaxy. Once genome annotation was completed, nucleotide −1 was located and used for data visualization in UCSC genome viewer. This base is the HuR- crosslink binding site. The number cDNA counts associated to a single HuR- crosslink site was quantified base on read length and unique 5′ barcode sequence. Peak call analysis was performed to identify unique HuR-crosslink binding sites with a FDR<0.05. Peak call analysis was performed in iCount by genome wide analysis of 30 base windows defined after annotation at single nucleotide resolution of HuR-RNA interactions (15 nucleotides flanking a cross link site). Peak enrichment analysis and FDR calculation was performed allowing 100 permutations and multiple hypothesis testing to correct for multiple comparisons.

Raw sequencing data from mRNAseq and Ribo-Seq was demultiplexed, FastQ analysed and mapped to mm10 genome annotation using Tophat. Differential gene expression analysis of mRNAseq and Ribo-Seq datasets was performed using DESeq and DESeq2 packages⁵⁵. Briefly, data was adjusted to a negative binomial distribution after correction by library size factor. Differential expression analysis, p value determination and Benjamini-Hochberg multiple test correction of the p values were performed to obtain final padj values. DEXSeq package was used for exon and intron usage analysis⁵⁶. First, we define exons and introns by using the Ensembl exon annotation for protein coding transcripts (GRCm38.72) and we independently counted reads within the exon and intron bins by using HTSeq. Read counting along exon bins was required in order to estimate the library size factors. We then used these size factors to normalise the intron bins and assessed differential intron usage with DEXSeq. A hypergeometric statistical test followed by Benjamini-Hochberg correction of the p- values was performed for pathway enrichment analysis^{57, 58}. UCSC and IGV were used to visualise read coverage (see Supplemental Information). Sequencing data was visualised in UCSC genome viewer. Read coverage of mRNAseq and Ribo-Seq data was visualised in the Integrated Genome Browser (IGV).

Unpaired t tests or Mann-Whitney tests were performed for statistical analysis of non-sequencing data. More details about sample size and the statistical test performed in each experiment are indicated in each figure legends.

Data deposition

High throughput sequencing data was deposited in Gene Expression Omnibus (GEO). Identification numbers are: GSE62129 for mRNAseq, GSE62134 for Ribo-seq and GSE62148 for iCLIP.