Report

RNA-binding proteins ZFP36L1 and ZFP36L2 promote cell quiescence

See allHide authors and affiliations

Science  22 Apr 2016:
Vol. 352, Issue 6284, pp. 453-459
DOI: 10.1126/science.aad5978

Reducing the risk of rearrangement

As lymphocytes develop, they rearrange their antigen receptor genes and proliferate extensively, potentially putting their genomes at risk. Galloway et al. found that two RNA-binding proteins, ZFP36L1 and ZFP36L2, ensure careful entry and exit into the cell cycle. This helps developing B lymphocytes maintain their genomic integrity. Mice deficient in ZFP36L1 and ZFP36L2 exhibited a profound block in B cell development. ZFP36L1 and ZFP36L2 suppress mRNAs that help B cells progress through the cell cycle, ensuring that cells can enter quiescence and keep their genomes safe when they undergo the risky process of rearranging their antigen receptors.

Science, this issue p. 453

Abstract

Progression through the stages of lymphocyte development requires coordination of the cell cycle. Such coordination ensures genomic integrity while cells somatically rearrange their antigen receptor genes [in a process called variable-diversity-joining (VDJ) recombination] and, upon successful rearrangement, expands the pools of progenitor lymphocytes. Here we show that in developing B lymphocytes, the RNA-binding proteins (RBPs) ZFP36L1 and ZFP36L2 are critical for maintaining quiescence before precursor B cell receptor (pre-BCR) expression and for reestablishing quiescence after pre-BCR–induced expansion. These RBPs suppress an evolutionarily conserved posttranscriptional regulon consisting of messenger RNAs whose protein products cooperatively promote transition into the S phase of the cell cycle. This mechanism promotes VDJ recombination and effective selection of cells expressing immunoglobulin-μ at the pre-BCR checkpoint.

Lymphocyte development is characterized by dynamic shifts between quiescence and proliferation. Quiescence promotes variable-diversity-joining (VDJ) recombination, the process that generates immunoglobulin and T cell receptor genes, because RAG2 protein expression is restricted to the G0-G1 phase of the cell cycle (13). In B cells, VDJ recombination leads to expression of an immunoglobulin-μ (Igμ) heavy chain that, together with the surrogate light chains, forms a precursor B cell receptor (pre-BCR). Signals from the pre-BCR terminate the recombination process and trigger rapid proliferation associated with passage through the pre-BCR checkpoint (4). Later signals from the pre-BCR reestablish quiescence, allowing immunoglobulin light-chain recombination (fig. S1A) (5, 6).

The ZFP36 family of RNA-binding proteins (RBPs) regulate gene expression posttranscriptionally by promoting mRNA decay (7). This requires their direct binding to AU-rich elements (AREs) located in the 3′ untranslated regions (3'UTRs) of mRNAs. ZFP36 destabilizes cytokine mRNAs and exerts an anti-inflammatory function (8, 9). In addition, ZFP36 antagonizes Myc-induced lymphomagenesis (10), and its paralogs ZFP36L1 and ZFP36L2 have redundant roles in preventing T cell leukemia in mice (11). The pathways controlled by these RBPs, however, have remained poorly understood.

In our study, all three ZFP36-family mRNAs were expressed throughout B cell development (fig. S1B). Conditional genetic deletion demonstrated redundant roles for ZFP36L1 and ZFP36L2 in early B cell development that could not be compensated for by endogenous ZFP36 and that were independent of NOTCH1, a known target of these RBPs (fig. S2) (11). We therefore generated mice in which Zfp36l1 and Zfp36l2 were deleted in pro-B cells (fig. S3). For simplicity, Zfp36l1fl/fl Zfp36l2fl/fl Mb1cre/+ mice will be referred to as DCKO (double conditional knockout) mice, and their Zfp36l1fl/fl Zfp36l2fl/fl Mb1+/+ littermates will be referred to as control mice. DCKO mice displayed reduced cellularity from the pre-B stage onward, culminating in a 98% reduction in the number of mature B cells (Fig. 1, A and B). The proportion of CD43+ cells expressing Igμ was greatly diminished (Fig. 1C), and a variable proportion of DCKO cells transited the pre-BCR checkpoint without Igμ expression (Fig. 1D). Within the compartment enriched for pro-B cells, DCKO mice had reduced proportions of cells containing one or two V-to-DJ-recombined IgH alleles (Fig. 1E and fig. S4). At later developmental stages, the DCKO mice failed to increase the proportion of recombined IgH alleles to control levels. The decrease in the proportions of V-to-DJ-recombined IgH alleles and Igμ+ cells within the pro-B and early pre-B cell compartment of DCKO mice was similar, suggesting that the failure to express Igμ is not due to an increase in nonproductive joints. Similarly, V-to-J recombination of the Igκ light-chain locus was reduced in DCKO late pre-B cells (Fig. 1F). Thus, DCKO mice display reduced B cell numbers, delayed VDJ recombination, and failure of the pre-BCR checkpoint.

Fig. 1 Conditional knockout of Zfp36l1 and Zfp36l2 in pro-B cells abrogates pre-B cell development.

(A) Representative scatter plots from flow cytometric analysis of B cell development in control and DCKO bone marrow. Numbers on plots indicate percentages of plotted cells in the gate (axes are in arbitrary units of fluorescence intensity). Purple arrows show the cell gating strategy. (B) Quantification of B cell developmental subsets in control (n = 5) and DCKO (n = 5) bone marrow from flow cytometry data shown in (A). Control and DCKO populations were compared by an analysis of variance (ANOVA) with Sidak’s post-test. Data are representative of two independent experiments. (C and D) Flow cytometry measuring intracellular Igμ in control (n = 6) and DCKO (n = 5) pro- and early pre-B cells (C) and control (n = 7) and DCKO (n = 7) late pre-B cells (D). Representative histograms of flow cytometry data and summary data are shown. Control and DCKO populations were compared using a Student’s t test. Data are representative of four (C) or two (D) independent experiments. (E) Quantification by DNA FISH (fluorescence in situ hybridization) of cells with zero, one, or two V-to-DJ-recombined IgH alleles within the CD24low pro-B and early pre-B (enriched for pro-B cells; n = 3 biological replicates), CD24high pro-B and early pre-B (enriched for early pre-B cells; n = 1 biological replicate), and late pre-B (n = 3 biological replicates) populations of control and DCKO mice. Data are from a single experiment; mean values and standard deviations (error bars) are shown. (F) Abundance of recombined Igκ alleles in the late pre-B cells of control (n = 11) and DCKO (n = 11) mice, measured by qPCR (quantitative polymerase chain reaction). Control and DCKO samples were compared using a Student’s t test. Data are included from three independent experiments. (G and H) Proportion of control (n = 5) and DCKO (n = 5) pro- and pre-B cells (G) or Igμ cells appearing in the late pre-B cell gate (H) that were in early apoptosis, measured by staining for activated caspases with FITC-VAD-FMK (fluorescein isothiocyanate-valyl-alanyl-aspartyl-[O-methyl]-fluoromethylketone). Control and DCKO samples were compared by an ANOVA with Sidak’s post-test. Data are representative of two independent experiments. In (B) to (D) and (F) to (H), circles indicate biological replicates, bars represent means [geometric means in (B)], and P values are shown at the top of each panel.

Expression of a productively rearranged Igμ transgene failed to restore late pre-B, immature B, or mature B cell numbers in DCKO mice, indicating that reduced VDJ recombination of IgH is not the sole defect (fig. S5). Consistent with this, we observed an increase in apoptosis of DCKO Igμ+ late pre-B cells (Fig. 1G). The increased apoptosis was not susceptible to inhibition by BCL2, and we did not detect a DNA damage response (fig. S6). In both DCKO and control mice, a substantial number of Igμ cells in the late pre-B cell gate were apoptotic, indicating that although Igμ DCKO cells were aberrantly selected, they did not bypass the requirement of pre-BCR expression for continued survival (Fig. 1H).

Within mRNAs that were overrepresented in DCKO late pre-B cell transcriptomes, there was a strong enrichment for pathways promoting cell cycle progression (Fig. 2, A and B, and tables S1 to S4). Because VDJ recombination is inhibited by cell cycle progression, and the pre-BCR checkpoint is mediated by the selective proliferation of Igμ+ cells, we hypothesized that these aspects of the phenotype might be explained by uncontrolled cell cycle progression in DCKO mice. The apoptosis of DCKO late pre-B cells could be due to delayed light-chain recombination or overexpression of cell cycle regulators; these cells are destined to be quiescent and therefore may not tolerate activation of the E2F pathway, which regulates transition through the cell cycle and DNA synthesis. Loss of E2f1 has been demonstrated to increase late pre-B cell numbers (12).

Fig. 2 Zfp36l1 and Zfp36l2 control the cell cycle during B cell development.

(A) Gene set enrichment analysis of transcripts that were significantly increased in DCKO late pre-B cells. The 20 pathways with the lowest false discovery rates (FDR), calculated using the Benjamini-Hochberg correction, are shown (FDR < 10−6 for all gene sets shown). Numbers beside bars represent the number of overlapping genes between the gene set and the transcripts that were increased in DCKO late pre-B cells. Data set abbreviations are as follows: REAC, Reactome; BIOC, Biocarta; Wiki, Wikipathways; and KEGG, Kegg pathways. (B) Scatter plot showing the average reads per kilobase per million (RPKM) in DCKO and control late pre-B cells for all genes (gray) and genes in the Reactome Cell Cycle pathway gene set that were significantly increased (red), unchanged (black), or significantly decreased in DCKO mice (blue). (C) Representative scatter plots from intracellular flow cytometry measuring BrdU (bromodeoxyuridine) incorporation after 2.5 hours of labeling in vivo in control and DCKO pro- and pre-B cells (DAPI, 4′,6-diamidino-2-phenylindole). (D) Proportion of control (n = 5) and DCKO (n = 5) pro- and pre-B cells in the S phase (incorporating BrdU), measured by flow cytometry as shown in (C). Data are representative of two independent experiments. (E) Representative scatter plots from intracellular flow cytometry measuring p27 in control and DCKO pro- and pre-B cells. (F) Proportion of control (n = 6) and DCKO (n = 5) pro- and pre-B cells in G0 (expressing p27), measured by flow cytometry as shown in (E). Data are representative of two independent experiments. Numbers on flow cytometry plots indicate the percentages of plotted cells in the gate. Flow cytometry data for control and DCKO samples were compared by an ANOVA with Sidak’s post-test. In (D) and (F), symbols indicate biological replicates, bars represent means, and P values are shown at the top of each panel.

Cell cycle analysis showed an increase in the proportion of DCKO pro-B cells in the S phase (Fig. 2, C and D). The cyclin-dependent kinase inhibitor p27 (also known as CDKN1B) is known to regulate the cell cycle during pre-B cell development (5), and high expression of p27 (p27high) is a marker of cellular quiescence (13). The proportion of p27high G0 cells was markedly reduced in DCKO pro-B cells and was moderately reduced in DCKO early and late pre-B cells (Fig. 2, E and F). Thus, ZFP36L1 and ZFP36L2 impose quiescence on developing B cells and inhibit entry into the S phase before expression of the pre-BCR.

We demonstrated that the transcription factors induced by the pre-BCR that promote quiescence in late pre-B cells were expressed in DCKO mice and that p27 mRNA was induced; additionally, factors mediating VDJ recombination were expressed, and the Igκ locus was transcriptionally active (fig. S7). Therefore, DCKO late pre-B cells are transcriptionally poised to enter quiescence and undergo VDJ recombination, but posttranscriptional regulation mediated by ZFP36L1 and ZFP36L2 is required for the full activation of these processes.

Consistent with their role in promoting mRNA decay, ZFP36L1-bound transcripts identified by cross-linking immunoprecipitation (iCLIP) (14) were more abundant in DCKO late pre-B cells (Fig. 3A). Increased mRNA abundance in DCKO late pre-B cells was also associated with AREs in the 3'UTRs of mRNAs (fig. S8A) and with ZFP36 binding sites in the human homologs of mouse transcripts (fig. S8, B and C) (15). Thus, the specificity of ZFP36-family proteins is generally conserved across family members, species, and cell types.

Fig. 3 Cell cycle mRNAs are direct targets of ZFP36L1 and ZFP36L2.

The abundance of mRNAs in DCKO and control late pre-B cells was measured by RNA sequencing and was analyzed in DESeq. ZFP36L1 target mRNAs were identified by iCLIP in mitogen-stimulated lymph node B cells. (A) The moderated log2 fold change in the abundance of RNA in DCKO relative to control late pre-B cells, grouped according to the number of ZFP36L1 iCLIP reads within significant peaks (FDR < 5%) in the 3'UTR of each gene. The number n of genes in each group is indicated. Boxes show medians and interquartile ranges; whiskers indicate the 5th and 95th percentiles. Groups of mRNAs were compared by an ANOVA with Tukey’s post-test. (B) Venn diagram of cell cycle mRNAs, showing the overlap between mRNAs identified as ZFP36L1 targets by iCLIP, mRNAs with WWAUUUAWW motifs in their 3'UTRs, and mRNAs that were significantly increased in DCKO relative to control late pre-B cells, as determined by a negative binomial test with a Benjamini-Hochberg correction for multiple testing in DESeq. The moderated log2 fold changes in mRNA abundance in DCKO late pre-B cells for selected groups of mRNAs are shown. (C) Sequence conservation among Homo sapiens, Mus musculus, Loxodonta africana, Pteropus vampyrus, Canis lupus familaris, and Bos taurus CDS (coding sequences), 3'UTRs, and WWAUUUAWW motifs for selected mRNAs that encode cell cycle regulators. (D) Proportion of pro- and pre-B cells from ROSA26L1/L1 (n = 7) and ROSA26L1/L1 CD2cre (n = 10) mice in the S phase (incorporating BrdU), determined by flow cytometry after 2.5 hours of labeling in vivo. Flow cytometry scatter plots are shown in fig. S11A. Data are combined from two independent experiments. (E) Proportion of ROSA26L1/L1 control (n = 7) and ROSA26L1/L1 CD2cre (n = 6) pro- and pre-B cells in G0 (expressing p27). Flow cytometry scatter plots are shown in fig. S11B. Data are from a single experiment in G0 (expressing p27). Flow cytometry data for control and DCKO samples were compared by an ANOVA with Sidak’s post-test. In (D) and (E), circles indicate biological replicates, bars represent means, and P values are shown at the top of each panel.

ZFP36L1 binding sites were typically associated with AREs (table S5). In the mRNAs identified by iCLIP, cell cycle pathways were strongly enriched (tables S6 and S7), evidence that further connects ZFP36L1 to cell cycle regulation. From among mRNAs implicated in cell cycle control, we identified several candidate targets that have ZFP36L1 binding sites or AREs and that exhibited significant increases in mRNA abundance in DCKO late pre-B cells (Fig. 3B and table S8). We validated the activity of the ZFP36L1 binding site in Ccne2 (fig. S9). Among the putative targets are the mRNAs encoding PIM-family kinases and components of the CDK2–cyclin E complex that phosphorylates p27, promoting its destruction (16, 17); this mechanism is consistent with the reduced p27 protein but equivalent p27 mRNA in DCKO late pre-B cells, relative to those of control mice. Furthermore, the AREs in putative target mRNAs that have roles in cell cycle progression are very highly conserved in mammals (Fig. 3C). These data strongly suggest that ZFP36L1 and ZFP36L2 directly regulate an evolutionarily conserved posttranscriptional regulon that controls cell cycle progression (18).

A posttranscriptional mechanism for enforcing quiescence is well suited to the events surrounding the pre-BCR checkpoint, because it can be reversed more rapidly than changes that are mediated at the level of transcription. ZFP36L1 and ZFP36L2 are phosphorylated by MAPKAP2 downstream of p38 MAPK (mitogen-activated protein kinase), and this inhibits their mRNA-destabilizing effects (19). The activity of p38 is induced downstream of the pre-BCR, providing a mechanism to relieve the repression by ZFP36L1 and ZFP36L2 of mRNAs encoding cell cycle regulators (20). To examine the effects of ZFP36L1 overexpression at the pre-BCR checkpoint, we generated mice that conditionally expressed ZFP36L1 fused at its N terminus to green fluorescent protein (GFP); we refer to the allele as ROSA26L1 (fig. S10). There was a significantly reduced proportion of S-phase cells and a significantly increased proportion of G0 cells in ROSA26L1/L1 CD2cre pre-B cells, compared with controls (Fig. 3, D and E, and fig. S11, A and B). Thus, enforced expression of ZFP36L1 suppresses proliferation at the pre-BCR checkpoint.

Cyclin D3, cyclin E2, and their partner kinases were identified among candidate ZFP36L1 and -2 targets in DCKO late pre-B cells. Cyclin D3 has an essential role in pre-BCR–mediated proliferation (21). Elevated protein expression of cyclin D3 and cyclin E2 was confirmed in DCKO pro-B and early pre-B cells (Fig. 4, A and B, and fig. S12, A and B), indicating that ZFP36L1 and ZFP36L2 limit the induction of cyclin D3 and cyclin E2 to cells that are transiting pre-BCR selection. In contrast, ROSA26L1 CD2cre mice failed to properly induce cyclin D3 or cyclin E2 at the proliferative early pre-B cell stage, a result that further reinforces the role of ZFP36L1 in controlling the cell cycle at the point of pre-BCR selection (Fig. 4, C and D, and fig. S12, C to E).

Fig. 4 Cyclin expression is repressed by ZFP36L1 and ZFP36L2 in B cell development, and this mechanism is required for efficient light-chain recombination.

(A) MFI (median fluorescence intensity) of flow cytometry stains for cyclin D3 in control (n = 4) and DCKO (n = 5) pro- and pre-B cells. Scatter plots of flow cytometry are shown in fig. S12A. Data are representative of two independent experiments. (B) MFI of flow cytometry stains for cyclin E2 in control (n = 5) and DCKO (n = 4) pro- and pre-B cells. Scatter plots of flow cytometry are shown in fig. S12B. Data are representative of two independent experiments. (C) MFI of flow cytometry stains for cyclin D3 in ROSA26L1/L1 control (n = 5) and CD2cre (n = 5) pro- and pre-B cells. Scatter plots of flow cytometry are shown in fig. S12C. Data are from a single experiment. (D) MFI of flow cytometry stains for cyclin E2 in ROSA26L1/L1 control (n = 5) and CD2cre (n = 5) pro- and pre-B cells. Scatter plots of flow cytometry are shown in fig. S12D. Data are from a single experiment. Flow cytometry data for (A) to (D) were compared using an ANOVA with Tukey’s post-test. (E) Quantification by DNA FISH of pro- and early pre-B cells with zero, one, or two V-to-DJ-recombined IgH alleles in control and DCKO mice after treatment with vehicle control or 150 mg/kg of palbociclib daily for 2 days. Data are from a single experiment (n = 3 for each group); mean values and standard deviations (error bars) are shown. (F) Abundance of recombined Igκ alleles, measured by qPCR, in the late pre-B cells of control and DCKO mice treated with vehicle control or 150 mg/kg palbociclib daily for 2 days. Control and DCKO samples were compared using an ANOVA with Tukey’s post-test. Data are from a single experiment. (G) The fold change (from published data sets) in expression of Zfp36 family members, comparing resting (quiescent) and stimulated dividing (nonquiescent) follicular B cells (n = 1), light-zone (quiescent) and dark-zone (nonquiescent) germinal center B cells (n = 2), naïve (quiescent) and effector (nonquiescent) CD8+ T cells (n = 1), adult (quiescent) and fetal (nonquiescent) hematopoietic stem cells (HSCs) (n = 2), p27high (quiescent) and p27low (nonquiescent) fibroblasts (n = 4), and resting (quiescent) and activated (nonquiescent) muscle satellite cells (n = 3). An asterisk indicates a P value less than 0.05. Solid lines separate independent data sets; dashed lines separate biological replicates. (H) Expression of ZFP36L1, cyclin D3, cyclin D1, and HSP90 proteins in parental and ZFP36L1−/− HCT116 colon carcinoma cells after serum starvation and restimulation (C, continuous culture; SF, serum free for 24 hours; Hrs, hours; FBS, fetal bovine serum). Data are representative of three experiments. In (A) to (D) and (F), symbols represent biological replicates, bars indicate means, and P values are shown at the top of each panel.

Overexpression of cyclin D3 can inhibit VDJ recombination in pre-B cells through a mechanism involving loss of quiescence (22). Therefore, we treated DCKO and control mice with the CDK4 and -6 inhibitor palbociclib, which inhibits activation of the E2F pathway. Palbociclib treatment increased V-to-DJ recombination at the IgH locus of cells in the pro-B and early pre-B compartment (Fig. 4E and fig. S13, A and B). Consistent with increased recombination and reduced cell division, the proportion of pro- and pre-B cells containing excised signal circles was increased after palbociclib treatment (fig. S13, C and D). Conversely, the frequency of recombination at the IgH locus of late pre-B cells was not restored by palbociclib treatment, reflecting the inhibition of the cell cycle, which prevents the proliferative selection of cells into the late pre-B cell pool (fig. S13E). Igκ recombination was also restored in DCKO late pre-B cells after palbociclib treatment (Fig. 4F and fig. S13F). This indicates that the delays in VDJ recombination are caused by loss of quiescence in DCKO pro- and pre-B cells.

We found that increased Zfp36–family member mRNA expression was typically associated with quiescent cell phenotypes (Fig. 4G). Therefore, we generated a ZFP36L1−/− HCT116 human colorectal carcinoma cell line and measured the expression of cyclin D3 and of cyclin D1, a putative ZFP36L1 target that is not expressed in B cells (15, 23). Expression of both D-type cyclins was increased in ZFP36L1−/− HCT116 cells compared with the parental line (Fig. 4H). Additionally, genetic experiments have shown that the loss of Zfp36l2 leads to depletion of hematopoietic stem cells (24), and the loss of Zfp36 is associated with increased muscle satellite activation (25). Thus, these RBPs probably form part of a general mechanism for the posttranscriptional regulation of quiescence.

As many as 10% of human mRNAs contain AREs (26); this may enable interdependent cellular processes to be coordinated by the ZFP36 family. The dynamics of the transition between the G0-G1 and S phases are characterized by switching behavior that is mediated by positive feed-forward regulation in the E2F pathway (27); thus, this pathway may be particularly sensitive to moderate changes in the abundance of its components, as measured in DCKO pre-B cells (Fig. 3A). We propose that ZFP36L1 and ZFP36L2 suppress the expression of limiting factors for E2F pathway activation (16, 2831) and DNA replication licensing, thus providing a robust mechanism for reversibly stabilizing the G0-G1 state. This mechanism would contribute to the ability of the progenitor cell populations to respond appropriately and dynamically to both mitogenic and antiproliferative signals.

Supplementary Materials

www.sciencemag.org/content/352/6284/453/suppl/DC1

Materials and Methods

Figs. S1 to S13

Tables S1 to S16

References (3250)

References and Notes

  1. Acknowledgments: We thank K. Vogel, R. Newman, C. Tiedje, and S. Bell for advice and comments on the manuscript; L. Matheson, A. Stark, and R. Venigalla for technical advice; S. Bell, K. Bates, D. Sanger, N. Evans, and The Babraham Institute’s Biological Support Unit, Flow Cytometry Core Facility, and Next Generation Sequencing Facility for expert technical assistance; J. Ule and T. Curk for help with iCLIP; L. Dolken for the gateway-compatible psi-check vector; and R. Brink, D. Kioussis, and M. Reth for mice. This work was funded by the Biotechnology and Biological Sciences Research Council, a Medical Research Council (MRC) Collaborative Award in Science and Engineering (CASE) studentship with GlaxoSmithKline, a MRC centenary award (A.G.), and project grants from Bloodwise. D.J.H. was supported by a MRC Clinician Scientist Fellowship. The data from this study are tabulated in the main paper and in the supplementary materials. Sequencing data from the RNA sequencing and iCLIP experiments have been deposited in the National Center for Biotechnology Information’s Gene Expression Omnibus (GEO) and are accessible under GEO Series accession number GSE78249 (www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE78249). Zfp36l1fl, Zfp36l2fl, and ROSA26L1 mice are available from The Babraham Institute under a material transfer agreement.
View Abstract

Navigate This Article