Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Altered splicing and cytoplasmic levels of tRNA synthetases in SF3B1-mutant myelodysplastic syndromes as a therapeutic vulnerability

Abstract

Myelodysplastic syndromes (MDS) are haematopoietic malignancies that are characterised by a heterogeneous clinical course. In recent years, sequencing efforts have uncovered recurrent somatic mutations within RNA splicing factors, including SF3B1, SRSF2, U2AF1 and ZRSR2. The most frequently mutated gene is SF3B1, mutated in 17% of MDS patients. While SF3B1 mutations and their effects on splicing have been well characterised, much remains to be explored about their more far-reaching effects on cellular homeostasis. Given that mRNA splicing and nuclear export are coordinated processes, we hypothesised that SF3B1 mutation might also affect export of certain mRNAs and that this may represent a targetable pathway for the treatment of SF3B1-mutant MDS. We used CRISPR/Cas9-genome editing to create isogenic cellular models. Comprehensive transcriptome and proteome profiling of these cells identified alterations in the splicing and export of components of the translational machinery, primarily tRNA synthetases, in response to the SF3B1 K700E mutation. While steady-state protein synthesis was unaffected, SF3B1 mutant cells were more sensitive to the clinically-relevant purine analogue, 8-azaguanine. In this study, we also demonstrated that 8-azaguanine affects splicing. Our results suggest that the simultaneous targeting of RNA metabolism and splicing by 8-azaguanine represents a therapeutic opportunity for SF3B1-mutant myelodysplastic syndromes.

Introduction

Myelodysplastic syndromes (MDS) are myeloid haematopoietic malignancies that are characterised by an extremely heterogeneous clinical course, ranging from indolent disease without progression, to acute onset requiring immediate treatment. The variable course of MDS is mainly driven by molecular heterogeneity. There are many genetic defects associated with different prognoses, including numerous deletions and chromosomal losses1, implying that the dysplasia of MDS is generally due to loss or alteration of gene function.

In recent years, genome and exome-wide sequencing studies in MDS have uncovered a number of recurrent somatic mutations within mRNA splicing factors, including SF3B1, SRSF2, U2AF1 and ZRSR2. The most frequent of these were in SF3B1, representing 17% of all mutations in MDS2, which in isolation generally infer a better prognosis. This is likely due to their high frequency in the refractory anaemia with ringed sideroblasts (RARS) subtype of MDS, which has a clinically more benign phenotype, with a lower likelihood of transformation. Indeed, mutations of SF3B1 occur in up to 90% of patients with RARS and in 70% of those with refractory cytopenia with multilineage dysplasia and ring sideroblasts (RCMD-RS). The presence of ringed sideroblasts, which arise from abnormal iron deposits, was recently demonstrated to be directly related to the presence of SF3B1 mutations3. At the molecular level, mutant SF3B1 results in abnormal splicing of several genes, primarily due to misrecognition of 3′ splice sites4. Many of the resulting aberrant mRNAs undergo nonsense-mediated mRNA decay (NMD), leading to reduced gene expression. This was shown to affect many genes important for iron metabolism in haematopoietic cells, which likely explains the iron transport defects observed in these cells5,6.

While the connection between SF3B1 mutations and its effects on splicing at the molecular level has been well characterised7, much remains to be explored about its more far-reaching effects on cell homeostasis. It has been known for many years that mRNA splicing and nuclear export are coordinated processes, that are tightly-linked8,9,10. More recent research has begun to demonstrate a direct connection between alternative splicing and cytoplasmic abundance of transcripts as a mechanism of control11,12. Therefore, we hypothesised that SF3B1, being a critical part of the spliceosome, might also affect cytoplasmic levels of mRNA species. We sought to investigate whether this role of SF3B1 represented a strategy for targeting mutant cells for clinical benefit.

Our data propose that SF3B1 mutations lead to defects in the splicing and export of mRNAs encoding components of the translational machinery. While steady-state protein synthesis appears unaffected, SF3B1 mutant cells were more sensitive to the clinically-relevant purine analogue, 8-azaguanine. This sensitivity suggests that simultaneous targeting of both RNA metabolism and splicing by this single compound represents a therapeutic opportunity for patients suffering from SF3B1 mutant myelodysplastic syndromes.

Results

CRISPR/Cas9-edited cells express K700E mutant SF3B1 at equivalent mRNA and protein ratios

Whilst a few cell lines harbouring SF3B1 mutations do exist, none is derived from haematopoietic tissues. Therefore, to study the effects of the SF3B1 K700E mutation in isolation, we set out to create isogenic models of this mutation in haematopoietic cell lines. K-562 cells were edited using CRISPR/Cas9 and single-stranded oligodeoxynucleotides (ssODN) to introduce an A > G substitution in codon 700 of the SF3B1 gene, the mutation observed in the majority of MDS patients. A synonymous, tracking mutation was also introduced at codon 701, creating a new MspI restriction site (Fig. S1A).

Successful editing of the locus was identified through restriction fragment length polymorphism (RFLP), as digestion by MspI would create two fragments instead of one (Fig. S1B). Sanger sequencing of successfully edited cells showed a double peak at both the K700E A > G and V701V T > C nucleotides (Fig. 1A). Pyrosequencing of DNA and RNA showed that approximately 30% of both DNA and RNA reads contained the mutant A > G allele (Fig. 1B). These mutated cells are henceforth designated SF3B1K700E.

Figure 1
figure 1

(A) Sanger sequencing of the targeted genomic region from both wildtype and K700E mutated K-562 cells. Double chromatogram peaks representing different nucleotides are labelled in red. (B) (DNA) Pyrosequencing of the targeted genomic region from both wildtype and K700E mutated K-562 cells. The calculated allelic ratio is displayed for both the A > G (K700E) and T > C (V701V) nucleotides. The other ratios in light grey represent control reactions that ideally yield zero. (RNA) Pyrosequencing of cDNA via RT-PCR representing the ratio of RNA species for the same nucleotides. (C) Fluorescent in situ hybridization (FISH) of metaphase spreads from normal lymphocytes (NBM), H-2595 (K700E), Panc0504 (K700E) and K-562 cells. Blue – DAPI; Green – Whole chromosome 2 paint; Red - Fosmid G248P85642F7 [SF3B1]. All A minimum of 100 cells scored per cell line. All samples showed ≥3 signals for SF3B1 in ≥85% of cells.

Our initial attempts to target a number of other MDS and AML cell lines, including OCI/AML-3 and MDS-L, failed. Given that the mutant allele burden in our modified cells deviated significantly from 50%, and an anecdotal finding by Zhou et al.13, we hypothesised that it may have been due to these cell lines only harbouring two copies of SF3B1. Therefore, we used FISH to assess SF3B1 copy number in these cell lines. This revealed that K-562 cells carry at least 3 copies of chromosome 2 and, accordingly, 3 copies of the SF3B1 locus (Fig. 1C). Similarly, testing of two other human cancer-derived cell lines that carry SF3B1 K700E mutations (H-2595 and Panc0504) identified three or more copies of SF3B1. The finding that all reported SF3B1-mutant cell lines (including ESS-1) carry three copies of chromosome 2 may reflect a requirement of immortal cell lines for a lower dosage of the mutant protein.

Further validation of the mutant clones by T/A cloning of the locus agreed with the FISH and pyrosequencing data, showing that mutant cells possessed the A > G transition on only one of three alleles (data not shown). One clone, negative by RFLP and designated SF3B1+/Δ/Δ, carried two different out-of-frame deletions; retaining only a single functional copy of SF3B1 (Fig. S1C).

In order to verify that the mutated allele was translated into protein, SF3B1 was immunoprecipitated (using an N-terminal antibody) from wildtype and SF3B1K700E cells (Fig. S1D) and analysed by mass spectrometry. As the K700E mutation destroys a tryptic cleavage site in SF3B1, we were able to confirm the presence of the glutamic acid-containing peptide in the successfully mutated clone. In accordance with mRNA expression levels, label-free quantitation of peptide fragments estimated that between 32% and 46% of the immunoprecipitated SF3B1 protein contained the K700E substitution. This implies that the K700E change does not drastically affect the translation or stability of the protein.

SF3B1K700E has similar effects on splicing in primary cells and cell lines

We next set out to determine if SF3B1 mutation led to altered mRNA-splicing in this model system and whether this altered splicing reflected the patterns that are seen in MDS patients and other models of the same mutation. Therefore, publically available RNA-Seq data from CD34+ bone marrow cells from MDS patients14 and NALM-6 cell line models4 were re-analysed and directly compared.

Initial analysis identified strong inclusion of a cryptic exon within the SNURF/SNRPN gene in SF3B1 mutant cells in both public data sets. We determined whether our SF3B1K700E modified cells showed similar inclusion of this cryptic exon ‘2b’ by qPCR. We found significantly higher expression levels of the cryptic exon ‘2b’ in the SF3B1K700E clone compared to wildtype or +/Δ/Δ cells (Fig. 2A). To enable more comprehensive global comparisons, we performed total RNA-Seq on our isogenic cell line model. Encouragingly, increased alternate 3′ splice site usage within the ABCB7 gene (chrX:75,071,693-75,071,704 - hg38) was observed in the SF3B1K700E cells (Fig. 2B,C), in agreement with what has been identified by a number of groups4,6.

Figure 2
figure 2

(A) Relative quantities of SNURF exon 2b measured by qPCR across ex2-2b and normalized to ex2-3 levels. Error bars represent SEM. (B) RNA-Seq, total read count-normalized, coverage plots at the ABCB7 exon previously published as having an alternative 3′ splice site. Note: gene is on reverse strand. (C) RT-PCR of cDNA from wildtype, SF3B1K700E & SF3B1+/Δ/Δ whole cell (T), nuclear (N) and cytoplasmic fractions (C). Amplicon spans ABCB7 exons 8–9 (RefSeq NM_004299). Percentages represent RNA-seq reads that span the alternative 3′ splice site as a fraction of the total number of reads spanning exons 8 and 9. (D) Venn diagram representing overlap in genes identified as having altered splicing by dSpliceType (all types) in SF3B1K700E vs wildtype samples. Comparing RNA-Seq data from Darman et al. (NALM-6 cells, n = 3), Dolatshad et al. (primary MDS, n = 4) and this study (n = 2). (E) Venn diagram representing overlap in genes identified as downregulated at the transcript-level by DESeq. 2 analysis in SF3B1K700E vs wildtype samples. Comparing RNA-Seq data from Darman et al. (NALM-6 cells), Dolatshad et al. (primary MDS) and this study.

Investigation of differentially spliced transcripts identified significant overlaps between the model generated in this study and those previously published. In particular, 44.6% of genes that were identified as differentially-spliced in our isogenic model were also altered in primary MDS samples with SF3B1K700E mutation (Fig. 2D). This suggests a degree of consistency in the effects of SF3B1K700E on the splicing of particular genes, despite different cellular contexts. Thus demonstrating the relevance of cell lines as models to study this mutation and validating our isogenic model of the SF3B1K700E mutation to physiologically represent changes in RNA splicing and expression as seen in primary patient material.

Interestingly, the majority of differential events in all comparisons were skipped exons, representing more than 45% of total events in this analysis of the RNA-Seq data (Table 1). This suggests an overall disruption of exon recognition, despite the more commonly described effect of SF3B1K700E on 3′ splice site recognition15. In contrast to the splicing analysis, only 6 down-regulated transcripts overlapped between this study and primary MDS samples (Fig. 2E). This illustrates the limitations of whole-transcript level quantitative RNA-Seq in comparing the effects of spliceosome mutations on the transcriptome.

Table 1 Differentially spliced or expressed genes for data sets analysed.

On a functional level, enrichment analysis (using Enrichr) of the 548 common differentially-spliced genes in all three data sets (Fig. 2D) identified the ribosome, RNA transport and spliceosome pathways (Table 2), as significantly altered in SF3B1K700E cells.

Table 2 Enrichr – KEGG 2016 pathway enrichment from 548 genes identified as differentially-spliced in all three studies; primary MDS, NALM-6 cell line & this study.

SF3B1K700E leads to decreased levels of cytoplasmic mRNA

Numerous studies have demonstrated a tight link between RNA splicing and export, demonstrating that efficient splicing is required before RNA export can occur16,17,18. To examine whether SF3B1K700E expression had a global effect on mRNA export (Fig. 3A), we used polyA-targeting FISH to localise total mRNA species in both our isogenic K-562 cells and two similar mesothelioma cell lines (Fig. 3B). Quantification revealed that those cells harbouring SF3B1K700E displayed a significant increase (~20%) in nuclear mRNA staining intensity (Fig. 3C) compared to their SF3B1WT counterparts, suggesting a defect in mRNA export.

Figure 3
figure 3

(A) Schematic illustrating the effects of SF3B1 mutation on splicing and predicted roles in mRNA export. (B) FISH of cells fixed and stained with DAPI and Oligo d(T)50-Cy5. H-2595 (SF3B1K700E) & H-2591 (SF3B1WT) cells were grown directly on slides. K-562 cells were spun onto slides. Images for the same fluorophore were captured using identical exposure times and gain. Raw monochrome images colorized using ImageJ and overlaid in the final panels. (C) A plot representing the nuclear to cytoplasmic ratio of the Oligo d(T)50-Cy5 signal. Multiple FISH images were captured from replicate stainings and cytoplasmic Cy5 signal was quantified as a ratio of nuclear signal using an image reduction algorithm written in ImageJ. WT values set as 1. Error bars represent SEM.

Based on these observations, we set out to examine if SF3B1K700E expression was associated with reduced cytoplasmic levels of specific mRNAs. We isolated cytoplasmic and nuclear RNA from WT and SF3B1K700E expressing cells. To assess the purity of the samples, we quantified the levels of the nuclear lncRNAs MALAT1 and NEAT119, which were ≥30-fold enriched in nuclear vs. cytoplasmic fractions (Fig. S2A). Subsequent RNA-seq and data analysis confirmed a much greater coverage of MALAT1 and NEAT1 in nuclear RNA fractions compared to cytoplasmic fractions (Fig. S2B). The small nucleolar RNA (snoRNA) SNORD47 was also highly enriched in the nuclear fraction (Fig. S2B).

RNA-Seq analysis revealed that a higher number of genes were decreased in the cytoplasmic RNA fraction of SF3B1 mutant cells than increased (358 vs 168 respectively) (FDR < 0.05 & FC ± 2 | Fig. 4A). An interaction network of the under-represented genes (FDR < 0.05 & FC < −1.5 | Fig. 4B) illustrates that not all of the export deficient transcripts are actually mis-spliced. This suggests that the SF3B1-related export defect is not always related to altered splicing, but may reflect a temporal constraint on co-ordinated splicing and export. The specific clusters of genes identified as being down-regulated include components of Aminoacyl-tRNA biosynthesis (blue cluster), Biosynthesis of amino acids (mauve cluster) and the TCA cycle (orange cluster). Among the mis-spliced/lesser-exported genes in the network (49 genes) the pathways Aminoacyl-tRNA biosynthesis (KEGG hsa00970) and Cytosolic tRNA aminoacylation (Reactome HSA-379716) were highly enriched (red labels). Sequence analysis of tRNA synthetase splicing junctions that were altered in SF3B1K700E cells shows that they frequently possess an AG-GT donor-acceptor consensus site at position −28 (Fig. S2C). Strikingly, we found that 21 of the 44 human tRNA synthetases (48%) were significantly depleted in the cytoplasmic RNA fraction of SF3B1 mutant cells (Fig. S2D | Fig. 4C). Validation by qPCR of tryptophanyl-tRNA synthetase (WARS) showed accumulation of nuclear transcripts, but an almost 20% reduction at the cytoplasmic level (Fig. S2E). Taken together, this implies a deficiency in tRNA synthesis, caused by the defective splicing and/or export of crucial transcripts in SF3B1 mutant cells.

Figure 4
figure 4

(A) Scatter plot of mean cytoplasmic RNA-Seq read counts per gene, as normalised by DESeq2, comparing K700E cells to wildtype cells (n = 2). Axes plotted as biexponential (red = FDR < 0.05 FC > 2, blue = FDR < 0.05 & FC < −2). (B) Network of directly interacting proteins based on the set of genes identified as depleted in the cytoplasmic fraction of mutant cells (FDR0.05 FC < −1.5). Transcripts identified as having altered splicing in mutant cells are indicated as diamonds. Genes belonging to tRNA synthesis pathway (KEGG hsa00970) are labelled in red. Nodes coloured according to cluster assigned by Reactome FI. Protein-protein interactions identified by STRING-DB. Table lists top three enriched pathways for highlighted gene clusters and their FDR-adjusted p-values. (C) Plot of –log10(FDR-adjusted p-value) against log2(fold-change) for cytoplasmic RNAs WT vs SF3B1K700E –only log2(fold-change) ±1.5 shown (n = 2). Genes from tRNA synthetase pathway (KEGG hsa00970) plotted as green triangles. Significantly differentially-expressed tRNA synthetases are labelled.

SF3B1K700E leads to reduced protein levels of components of the translational machinery and tRNA metabolism, but to an increase in spliceosome components

Given the changes in mRNA splicing and export observed in the SF3B1K700E cells, we set out to determine the impact of these changes on protein expression. To investigate the effects of the SF3B1K700E mutation on the proteome, we used Stable Isotopic Labelling by Amino acids in Culture (SILAC). This quantitative mass spectrometry approach identified a number of proteins whose steady-state levels were altered significantly in SF3B1K700E cells compared to wildtype cells. While there was no correlation between transcript and protein levels at the whole cell level (Fig. S3A), changes between cytoplasmic transcript and whole cell proteome showed a positive correlation (Pearson = +0.294 | Fig. 5A). This is also reflected in the divergence of some transcripts from a general correlation between the three subcellular fractions (Fig. S3B).

Figure 5
figure 5

(A) Scatter plot of log2-transformed protein fold change (SILAC, n = 1, 2702 unique protein hits passing QC) vs cytoplasmic transcript log2-transformed fold change (RNA-Seq - DESeq. 2, n = 2). Pearson correlation = +0.294. Red = FC > 2 | Blue = FC < −2 in both comparisons. (B) Venn diagram representing overlap in genes identified as being depleted in the cytoplasm (FDR < 0.05) and down at the protein level by SILAC, comparing SF3B1K700E vs wildtype samples. (C) Network of directly interacting proteins based on those identified as being decreased in the SF3B1K700E cells by SILAC. Nodes coloured according to cluster assigned by Reactome FI. PPIs identified by STRING-DB (direct, 0.5 confidence). Table identifies top 4 significantly enriched gene ontology terms from Reactome, all related to translation (green cluster). (D) Network of directly interacting proteins based on those identified as being increased in the K700E cells by SILAC. Nodes coloured according to cluster assigned by Reactome FI. PPIs identified by STRING-DB. Table identifies top 4 significantly enriched gene ontology terms from Reactome, including oxygen transport (red) and splicing (purple). (E) Flow cytometry histograms of cellular fluorescence representing OP-Puromycin (OP-Puro) uptake in wildtype, +/Δ/Δ and +/+/K700E cells.

Intersection of RNA-seq and SILAC-MS datasets showed that 355 genes had both decreased cytoplasmic levels and reduced protein expression upon SF3B1 mutation (Fig. 5B). Indeed, 10 of the 21 tRNA synthetases that showed significantly reduced cytoplasmic mRNA levels, were also decreased at the protein level (Table S9). Moreover, an interaction network of the proteins down-regulated upon SF3B1 mutation shows enrichment for a cluster of genes related to translation (Fig. 5C), the same pathway enriched in all cytoplasm-depleted genes (FDR < 0.05). The translation machinery was found among the most strongly down-regulated proteins (>3-fold) with some ribosomal proteins (RPS/L) and elongation initiation factors (EIF) decreased at the protein level by more than 10-fold (Table S5).

In contrast, the most significantly enriched process among an interaction network of proteins upregulated in the SF3B1K700E cells (Fig. 5D) was oxygen transport (GO:0015671). This represents the haemoglobin chains (red cluster) — of which 4 subunits were increased >5-fold. In keeping with this, cell pellets from SF3B1 mutant cells were consistently redder in colour, reflecting increased haemoglobin levels (Fig. S4A). Components of the RNA splicing machinery were also enriched among up-regulated proteins (purple cluster), suggesting a possible feedback loop as the cell attempts to compensate for global mis-splicing.

Despite a significant reduction in a number of key ribosomal components at protein, mRNA and tRNA levels, SF3B1K700E cells did not show obvious defects in steady-state protein synthesis, as measured via similar rates of OP-Puro incorporation into the ribosome (Fig. 5E). Further, the clone possessing only a single functional copy of SF3B1 (+/Δ/Δ) also displayed similar rates of protein synthesis, suggesting that SF3B1 copy number alone does not influence global protein synthesis rates. Concomitantly, these data imply that SF3B1 mutation leads to depletion of tRNA pools through reduced tRNA synthetase levels, combined with a deficiency in translational machinery components, without drastically affecting steady-state protein synthesis in non-stressed cells.

SF3B1 mutation increases sensitivity to RNA analogs

Our data show that SF3B1K700E cells have defects in the splicing and cytoplasmic export of tRNA synthetases and RNA metabolism-related factors. Given this, together with the apparent decrease in ribosomal proteins, we hypothesised that SF3B1K700E cells might be hypersensitive to drugs targeting a compromised translational machinery. The nucleoside analogue 8-azaguanine is known to incorporate into rRNA and tRNA, and can block regular guanine incorporation20,21,22. This artificial base leads to the production of non-functional RNA species, thereby interfering with ribosome function. Thus we hypothesised that 8-azaguanine may selectively target SF3B1K700E cells over ribosome/translation replete wildtype cells. Indeed, SF3B1K700E cells were more sensitive to treatment with varying doses of 8-azaguanine (Fig. 6A).

Figure 6
figure 6

(A) Dose-response curve of 8-azaguanine treatment for 5 days; plotted as % viability as ATP levels normalised to vehicle-treated controls. The table represents calculated 50% inhibitory concentrations (IC50) and 95% confidence values (n = 3, technical duplicates). (B) Percentage of cells showing wildtype level of splicing, as measured by ratio of red to green fluorescence, after treatment with the indicated concentration of 8-azaguanine. At least 100 cells scored by high-content analysis.

As 8-azaguanine is also incorporated into mRNA species and as guanine is the most frequent base that is cleaved before transesterification in the predominant GT/AG cleavage sites, we suspected that its selective effect on SF3B1 mutant cells may also be due to an effect on mRNA splicing. To quantify 8-azaguanine-induced changes in correct intron removal, we used an exon-intron-exon two-colour fluorescent splicing reporter. Treatment of cell lines U2OS and 293-T harbouring this reporter with 8-azaguanine led to a decrease in relative fluorescence levels and therefore correct splicing, suggesting that 8-azaguanine does inhibit normal mRNA-splicing functions (Fig. 6B).

Discussion

As a health burden, myelodysplastic syndromes represent a considerable proportion of haematological malignancies, constituting 5.7% of total incidence23. Considering that the median age at diagnosis is 75.8 years, the use of high-intensity chemotherapy regimens is often not possible due to the lack of fitness of these patients to tolerate such therapies, so novel therapeutic options are desperately needed. The discovery of spliceosome mutations in myelodysplastic syndromes has led to an explosion of research in this field. Indeed, screening for these mutations has identified their presence in many other cancers, including chronic lymphocytic leukaemia (CLL)24, uveal melanoma25 and breast cancer26. While a number of novel compound classes have been discovered that target the spliceosome machinery directly, such as Spliceostatin A, Pladienolides and more recently H3B-8800, these drugs are far from clinical application and some have even proven difficult to synthesise27,28,29. Therefore, alternative means of selectively targeting the SF3B1 mutated cells in patients would be of significant clinical benefit.

There already exist a multitude of recent publications, which investigate the effects of SF3B1 mutations on splice-site selection4,7,15,30. Many of these studies have demonstrated that mutant SF3B1 leads to differential 3′ splice site preference during splicing. Another recent paper also investigated altered splicing as a signature of SF3B1 mutations, showing that it led to a pattern distinct from other spliceosome mutations31. In this study, we aimed to identify the downstream effects of SF3B1K700E mutation on cellular homeostasis through modelling of isogenic cell lines.

The most immediate effect of introducing the K700E mutation was a change in cell colour, which was reflected by an increase in foetal haemoglobin proteins in the SF3B1 mutant clone. This mirrors other research revealing alterations in haemoglobin synthesis in MDS patients32. Recent screening in healthy elderly populations has identified clonal haematopoiesis of indeterminate potential (CHIP) with an increase in the frequency of spliceosome mutations in haematopoietic stem cells33. While a relatively new concept, some hypothesise that CHIP may expand a failing stem cell pool, helping maintain sufficient levels of red blood cells34,35,36. Therefore SF3B1K700E, despite its association with malignant diseases, may initially be beneficial in increasing haemoglobin levels during age-related CHIP.

In comparing our genome-edited cells to existing data sets, we identified the cryptic exon ‘2b’ in the SNURF/SNRPN gene as strongly preferred in all SF3B1K700E samples considered in this study. This may represent a more sensitive surrogate biomarker for SF3B1 mutation in either the clinic or the research laboratory, where clonal diversity might mask small minor allele frequencies.

Given the intimate relationship between splicing and nuclear export, we hypothesised that the SF3B1 mutation might lead to a defect in RNA export. While analysis of cytoplasmic RNA-Seq supported this hypothesis, it also showed that the effect was not uniform for all transcripts. Reassuringly, the RNA-Seq data from all three fractions (total, nuclear, cytoplasmic) correlate reasonably well between wildtype and SF3B1-mutant cells. In contrast, the abundances of tRNA synthetase genes were inversely correlated between whole cell extracts and nuclear/cytoplasmic fractions. This suggests that lower levels of certain cytoplasmic RNAs may not simply be due to decreased export, but also decreased transcription in the nucleus or increased exclusion from efficient translation, perhaps by sequestration into processing bodies (P-bodies). In line with this, we demonstrated cytoplasmic transcript levels correlate with actual protein levels better than total transcript levels, as many mRNAs have recently been shown to be only repressed in P-bodies and not decayed37. It has also been shown that other components of the nuclear export machinery (NXF1/XPO1) are also mutated in CLL, mutually-exclusive of SF3B1 mutations, suggesting that altered export may phenocopy SF3B1 mutation38. However, caution should be used when comparing RNA-Seq quantification between different cellular compartments, as it is known that different fractionation methods can yield different estimated transcript levels39. Therefore, we focused our downstream enrichment analysis on the cytoplasmic RNA levels.

The most interesting finding was a defect in the splicing and export of mRNAs encoding tRNA synthetases and ribosomal components. Additionally, we observed decreased expression of alpha-aminoacyl-tRNA binding proteins in the SILAC-MS data, concomitant with tRNA depletion. While direct quantification of tRNAs is difficult in regular RNA-Seq datasets, this warrants further investigation. Despite these changes, steady-state levels of protein synthesis remained unaffected in SF3B1 mutant cells, presumably through up-regulation of compensatory proteins, such as BCLAF1 and SON, which have roles in ribosome biogenesis and mRNA processing40. These proteins may have been upregulated to meet the increased demand of proliferative capacity and thus become required to maintain high proliferation rates in SF3B1 mutant cells making them attractive targets for splicing-deficient neoplasms.

The relationship between tRNA maintenance and MDS is not entirely new. Sideroblastic anaemia has been connected to a mitochondrial tRNA defect in a rare mitochondrial disorder known as mitochondrial myopathy, lactic acidosis and sideroblastic anemia (MLASA) syndrome41. Interestingly, one of the functional clusters in the network of export-deficient transcripts includes MT-ATP6, whose mutation was associated with MLASA42. Furthermore, there is a difference in tRNA levels in MDS compared to healthy haematopoietic tissue43. However, these tRNA defects have not yet been associated with altered SF3B1 status.

We theorized that the deficiency in tRNAs and ribosomal components might represent a chemically-tractable target in SF3B1 mutant cells. The synthetic nucleoside analogue 8-azaguanine has been investigated for over 70 years44 and has been reported to have anti-tumour activity in mouse models45. More recent data has shown that it is able to induce differentiation in leukaemia cell lines, an effect that was likely mediated through its incorporation into tRNA species46. We found that the SF3B1 mutant cells were more sensitive to this compound when compared to cells with wild-type SF3B1. While the high doses of 8-azaguanine that were needed for efficacy against leukaemia as a single agent led to side effects, it might be used at significantly lower dose in MDS, a much more indolent disease, which typically requires lower doses of chemotherapy.

In addition to its well-established effect on tRNA function, our data also show that 8-azaguanine leads to a defect in mRNA splicing. Therefore, SF3B1 mutations leading to simultaneous defects in both splicing and tRNA/ribosome stability might expose a unique vulnerability for dual-targeting using 8-azaguanine. Additionally, 8-azaguanine does not cross the blood-brain barrier nor is it intercalated or incorporated into DNA. This reduces the risk of cerebellar toxicity or therapy-induced leukaemias, which often result from DNA-damaging chemotherapies, such as cytarabine and daunorubicin. The dual-targeting effect of 8-azaguanine may also partially explain the greater effectiveness of azacytidine in SF3B1 mutant MDS patients as it is also incorporated into RNA/tRNA47,48.

In conclusion, using isogenic cell systems, we show that the SF3B1K700E mutation induces consistent alterations in splicing regardless of cellular context. Additionally, it creates a defect in the nuclear export of polyadenylated mRNAs and a dysregulation of tRNA and ribosome homeostasis. Based on its ability to target both tRNA metabolism and splicing, we propose a new therapeutic application for the riboside analogue 8-azaguanine in treating patients harbouring SF3B1 mutations, such as breast cancer, uveal melanoma, myelodysplastic syndromes, acute myeloid leukaemia and beyond.

Materials and Methods

Published data analysis

RNA-Seq data from SF3B1 mutant NALM-6 cells was published by Darman et al.4 (NCBI PRJNA295064; GEO: GSE72790)4. RNA-Seq data from MDS patients and healthy donors was published by Dolatshad et al. (NCBI PRJNA268220; GEO: GSE63569)14. Both datasets were downloaded as raw FASTQ reads and processed using the same pipeline.

Cell lines

K-562 cells, obtained from DSMZ, were maintained in RPMI-1640 (Sigma Aldrich), supplemented with FBS Superior (Biochrom AG) and 1% Penicillin/Streptomycin (Gibco) at 37 °C 5% CO2. In the case of long-term culture for dilution cloning, complete media was supplemented with 2 mM HEPES (Sigma-Aldrich) and 1 mM UltraGlutamine I (Lonza Biochem).

Plasmids

The pSpCas9(BB)-2A-GFP (PX458) plasmid was a gift from Feng Zhang49 (Addgene plasmid # 48138). The pmKaxxte plasmid (now available as Addgene plasmid #113630), used to check gRNA efficiency, was generated using the pmaxGFP vector (Lonza) as a backbone and the mKate2.5-C1 — a gift from Michael Davidson (Addgene plasmid # 54828) — to generate overlapping insert fragments and a multiple cloning site.

The RFP2GFP-ATM construct was generated by inserting RFP sequence from mKate2.5-C1 into the pmaxGFP vector between KpnI and AgeI sites. The sequence encoding the ATM exon20-intron20-exon21 was PCR amplified from 293 T derived genomic DNA using primer sequences in Table S1 and cloned into the vector’s ORF between RFP and GFP tags using HindIII and SalI enzymes.

CRISPR/Cas9

Guide RNAs (gRNA) were chosen using the online tool designed by Zhang et al. ssODNs were designed that contained the desired substitutions together with 77 bp of homology either side of the targeted substitutions. The gRNAs were tested using the mKaxxte plasmid (Addgene #113630) containing a cloned fragment of the intended target genomic region. This assessed the effectiveness of DSB creation by the gRNA by triggering HR-repair of the template to re-establish fluorescence. The final gRNA chosen yielded bright fluorescence in the majority of control cells, compared to those without the gRNA, which showed no background fluorescence (Fig. S4C).

The ssODN and gRNA/SpCas9-expressing plasmid (gRNA cloned into Addgene #48138) were nucleofected into K-562 cells. Cells were then sorted to purity for GFP-positivity and sub-cloned by dilution. The mutant allele burden and mRNA expression level was checked by Pyrosequencing.

FISH & T/A cloning

FISH analysis was performed at the Wellcome Centre for Human Genetics, Oxford. For polyA FISH, cells were stained with fluorescent Oligo(d)T50 probes, as previously described16.

The targeted genomic region was amplified by PCR and then T/A cloned. At least 9 clones from each sample were sequenced.

Proteomics

For SILAC, cells were grown in either “light” or “medium” RPMI and crude lysates analysed by FingerPrints Proteomics, Dundee. Data were analysed using MaxQuant software. For protein synthesis measurements, the EZClick™ Global Protein Synthesis Assay Kit (BioVision Inc., USA) kit was utilised.

Splicing reporter assay

In the assay, a reporter plasmid encodes green and red fluorescent proteins separated by a human ATM intron fragment containing stop codons in all frames (RFP2GFP-ATM); successful splicing leads to an equivalent ratio of green and red fluorescence. U2OS and 293 T cells were seeded in 6 well glass bottom plates. The following day, cells were treated with 0.1, 0.5, 1 or 5 μM 8-Azaguanine. Approximately 5 hours later, cells were transfected with the RFP2GFP-ATM construct and incubated for further 48 hours. Following this, single cell RFP and GFP fluorescence intensity was acquired in 10,000 cells/well using high-content imaging and splicing was assessed as the GFP/RFP fluorescence intensity ratio relative to control/untreated cells.

RNA-Seq

Nuclear and Cytoplasmic RNAs were separated using a hypotonic buffer. RNA-Seq libraries were prepared using the SMARTer® Stranded Total RNA Sample Prep Kit (634873 Clontech). Sequencing reads were aligned using STAR50, transcripts quantified using DESeq251 and differential splicing analysed by dSpliceType52. Random subsampling analysis via RSeQC53 showed that the read coverage approached saturation of identified splice junctions, meaning that, statistically, all known and many of the novel junctions would have likely been identified at the coverage level achieved for all samples (Fig. S4D).

Data Availability

The RNA-Seq dataset generated during the current study is available in the ArrayExpress (EMBL-EBI) repository, under accession E-MTAB-7192. The proteomics dataset generated during the current study will be made available in the ProteomeXchange repository. All other data generated and analysed during this study are included in this published article (and its Supplementary Information files).

References

  1. Rizzieri, D. A. MDS: unraveling the mystery. Blood 120, 4906–4908 (2012).

    CAS  Article  Google Scholar 

  2. Damm, F. et al. Mutations affecting mRNA splicing define distinct clinical phenotypes and correlate with patient outcome in myelodysplastic syndromes. Blood 119, 3211–8 (2012).

    CAS  Article  Google Scholar 

  3. Dolatshad, H. et al. Disruption of SF3B1 results in deregulated expression and splicing of key genes and pathways in myelodysplastic syndrome hematopoietic stem and progenitor cells. Leukemia 29, 1092–103 (2015).

    CAS  Article  Google Scholar 

  4. Darman, R. B. et al. Cancer-Associated SF3B1 Hotspot Mutations Induce Cryptic 3′ Splice Site Selection through Use of a Different Branch Point. Cell Rep. 13, 1033–1045 (2015).

    CAS  Article  Google Scholar 

  5. Cazzola, M. et al. Biologic and clinical significance of somatic mutations of SF3B1 in myeloid and lymphoid neoplasms. Blood 121, 260–9 (2013).

    CAS  Article  Google Scholar 

  6. Dolatshad, H. et al. Cryptic splicing events in the iron transporter ABCB7 and other key target genes in SF3B1-mutant myelodysplastic syndromes. Leukemia 30, 2322–2331 (2016).

    CAS  Article  Google Scholar 

  7. Alsafadi, S. et al. Cancer-associated SF3B1 mutations affect alternative splicing by promoting alternative branchpoint usage. Nat. Commun. 7, 10615 (2016).

    ADS  CAS  Article  Google Scholar 

  8. Luo, M. J. & Reed, R. Splicing is required for rapid and efficient mRNA export in metazoans. Proc. Natl. Acad. Sci. USA 96, 14937–14942 (1999).

    ADS  CAS  Article  Google Scholar 

  9. Herzel, L., Ottoz, D. S. M., Alpert, T. & Neugebauer, K. M. Splicing and transcription touch base: Co-transcriptional spliceosome assembly and function. Nat. Rev. Mol. Cell Biol. 18, 637–650 (2017).

    CAS  Article  Google Scholar 

  10. Zhou, Z. et al. The protein Aly links pre-messenger-RNA splicing to nuclear export in metazoans. Nature 407, 401–405 (2000).

    ADS  CAS  Article  Google Scholar 

  11. Müller-McNicoll, M. et al. SR proteins are NXF1 adaptors that link alternative RNA processing to mRNA export. Genes Dev. 30, 553–66 (2016).

    Article  Google Scholar 

  12. Wegener, M. & Müller-McNicoll, M. Nuclear retention of mRNAs - quality control, gene regulation and human disease. Semin. Cell Dev. Biol. 79, 131–142 (2018).

    CAS  Article  Google Scholar 

  13. Zhou, Q. et al. A Chemical Genetics Approach for the Functional Assessment of Novel Cancer Genes. Cancer Res. 75, 1949–1958 (2015).

    CAS  Article  Google Scholar 

  14. Pellagatti, A. et al. Association Between Gene Expression Profiles and Commonly Mutated Genes In The Hematopoietic Stem Cells Of Patients With Myelodysplastic Syndromes. Blood 122, 2779- (2013).

    Google Scholar 

  15. Kesarwani, A. K. et al. Cancer-associated SF3B1 mutants recognize otherwise inaccessible cryptic 3′ splice sites within RNA secondary structures. Oncogene 36, 1123–1133 (2017).

    CAS  Article  Google Scholar 

  16. Vohhodina, J. et al. The RNA processing factors THRAP3 and BCLAF1 promote the DNA damage response through selective mRNA splicing and nuclear export. Nucleic Acids Res. 45, 12816–12833 (2017).

    CAS  Article  Google Scholar 

  17. Hautbergue, G. M. RNA Nuclear Export: From Neurological Disorders to Cancer. Adv. Exp. Med. Biol. 1007, 89–109 (2017).

    CAS  Article  Google Scholar 

  18. Wickramasinghe, V. et al. Human Inositol Polyphosphate Multikinase Regulates Transcript-Selective Nuclear mRNA Export to Preserve Genome Integrity. Mol. Cell 51, 737–750 (2013).

    CAS  Article  Google Scholar 

  19. Cabili, M. N. et al. Localization and abundance analysis of human lncRNAs at single-cell and single-molecule resolution. Genome Biol. 16, 20 (2015).

    Article  Google Scholar 

  20. Tavitian, A., Uretsky, S. C. & Acs, G. Selective inhibition of ribosomal RNA synthesis in mammalian cells. Biochim. Biophys. Acta - Nucleic Acids Protein Synth. 157, 33–42 (1968).

    CAS  Article  Google Scholar 

  21. Farkas, W. R., Hankins, W. D. & Singh, R. The guanylation of transferRNA: An enzymatic reaction. Biochim. Biophys. Acta - Nucleic Acids Protein Synth. 294, 94–105 (1973).

    CAS  Article  Google Scholar 

  22. Shindo‐Okada, N., Terada, M. & Nishimura, S. Changes in Amount of Hypo‐Modified tRNA Having Guanine in Place of Queuine during Erythroid Differentiation of Murine Erythroleukemia Cells. Eur. J. Biochem. 115, 423–428 (1981).

    Article  Google Scholar 

  23. Smith, A. et al. The Haematological Malignancy Research Network (HMRN): a new information strategy for population based epidemiology and health service research. Br. J. Haematol. 148, 739–53 (2010).

    Article  Google Scholar 

  24. Wang, L. et al. Transcriptomic Characterization of SF3B1 Mutation Reveals Its Pleiotropic Effects in Chronic Lymphocytic Leukemia. Cancer Cell 30, 750–763 (2016).

    CAS  Article  Google Scholar 

  25. Furney, S. J. et al. SF3B1 mutations are associated with alternative splicing in uveal melanoma. Cancer Discov. 3, 1122–9 (2013).

    CAS  Article  Google Scholar 

  26. Maguire, S. L. et al. SF3B1 mutations constitute a novel therapeutic target in breast cancer. J. Pathol. 235, 571–80 (2015).

    CAS  Article  Google Scholar 

  27. Salton, M. & Misteli, T. Small Molecule Modulators of Pre-mRNA Splicing in Cancer Therapy. Trends in Molecular Medicine 22, 28–37 (2016).

    CAS  Article  Google Scholar 

  28. Armstrong, R. N. et al. Splicing factor mutations in the myelodysplastic syndromes: target genes and therapeutic approaches. Adv. Biol. Regul. https://doi.org/10.1016/j.jbior.2017.09.008 (2017).

  29. Seiler, M. et al. H3B-8800, an orally available small-molecule splicing modulator, induces lethality in spliceosome-mutant cancers. Nat. Med. 24, 497–504 (2018).

    CAS  Article  Google Scholar 

  30. DeBoever, C. et al. Transcriptome sequencing reveals potential mechanism of cryptic 3′ splice site selection in SF3B1-mutated cancers. PLoS Comput. Biol. 11, e1004105 (2015).

    Article  Google Scholar 

  31. Qiu, J. et al. Distinct splicing signatures affect converged pathways in myelodysplastic syndrome patients carrying mutations in different splicing regulators. RNA 22, 1535–1549 (2016).

    CAS  Article  Google Scholar 

  32. Conte, S. et al. Aberrant splicing of genes involved in haemoglobin synthesis and impaired terminal erythroid maturation in SF3B1 mutated refractory anaemia with ring sideroblasts. Br. J. Haematol. 171, 478–90 (2015).

    CAS  Article  Google Scholar 

  33. Genovese, G. et al. Clonal hematopoiesis and blood-cancer risk inferred from blood DNA sequence. N. Engl. J. Med. 371, 2477–87 (2014).

    Article  Google Scholar 

  34. Sperling, A. S., Gibson, C. J. & Ebert, B. L. The genetics of myelodysplastic syndrome: from clonal haematopoiesis to secondary leukaemia. Nat. Rev. Cancer 17, 5–19 (2016).

    Article  Google Scholar 

  35. Lichtman, M. A. Clonal hematopoiesis: a “CHIP” off the old block. Blood 126, 1–2 (2015).

    CAS  Article  Google Scholar 

  36. Steensma, D. P. et al. Clonal hematopoiesis of indeterminate potential and its distinction from myelodysplastic syndromes. Blood 126, 9–16 (2015).

    CAS  Article  Google Scholar 

  37. Hubstenberger, A. et al. P-Body Purification Reveals the Condensation of Repressed mRNA Regulons. Mol. Cell 68, 144–157.e5 (2017).

    CAS  Article  Google Scholar 

  38. Ramsay, A. J. et al. Frequent somatic mutations in components of the RNA processing machinery in chronic lymphocytic leukemia. Leukemia 27, 1600–1603 (2013).

    CAS  Article  Google Scholar 

  39. Sultan, M. et al. Influence of RNA extraction methods and library selection schemes on RNA-seq data. BMC Genomics 15, 675 (2014).

    Article  Google Scholar 

  40. Blalock, W. L. et al. Identification of the PKR nuclear interactome reveals roles in ribosome biogenesis, mRNA processing and cell division. J. Cell. Physiol. 229, 1047–1060 (2014).

    CAS  Article  Google Scholar 

  41. Riley, L. G. et al. Mutation of the mitochondrial tyrosyl-tRNA synthetase gene, YARS2, causes myopathy, lactic acidosis, and sideroblastic anemia–MLASA syndrome. Am. J. Hum. Genet. 87, 52–9 (2010).

    CAS  Article  Google Scholar 

  42. Burrage, L. C. et al. Mitochondrial myopathy, lactic acidosis, and sideroblastic anemia (MLASA) plus associated with a novel de novo mutation (m.8969G&gt;A) in the mitochondrial encoded ATP6 gene. Mol. Genet. Metab. 113, 207–212 (2014).

    CAS  Article  Google Scholar 

  43. Guo, Y. et al. Transfer RNA detection by small RNA deep sequencing and disease association with myelodysplastic syndromes. BMC Genomics 16, 727 (2015).

    Article  Google Scholar 

  44. Roblin, R. O., Lampen, J. O., English, J. P., Cole, Q. P. & Vaughan, J. R. Studies in Chemotherapy. VIII. Methionine and Purine Antagonists and their Relation to the Sulfonamides. J. Am. Chem. Soc. 67, 290–294 (1945).

    CAS  Article  Google Scholar 

  45. Maxwell, R. A. & Eckhardt, S. B. Drug Discovery: A Casebook and Analysis. (Humana Press, 2012).

  46. Kretz, K. A., Katze, J. R. & Trewyn, R. W. Guanine analog-induced differentiation of human promyelocytic leukemia cells and changes in queuine modification of tRNA. Mol. Cell. Biol. 7, 3613–3619 (1987).

    CAS  Article  Google Scholar 

  47. Sarkisjan, D., Steenbergen, R. D., Cloos, J. & Peters, G. J. Re-emerging Antimetabolites with Novel Mechanism of Action with Respect to Epigenetic Regulation: Basic Aspects. in Chemotherapy for Leukemia (ed. Ueda, T.) 311–326 https://doi.org/10.1007/978-981-10-3332-2 (Springer Singapore, 2017).

  48. Thépot, S. et al. A randomized phase II trial of azacitidine +/− epoetin-β in lower-risk myelodysplastic syndromes resistant to erythropoietic stimulating agents. Haematologica 101, 918–25 (2016).

    Article  Google Scholar 

  49. Ran, F. A. et al. Genome engineering using the CRISPR-Cas9 system. Nat. Protoc. 8, 2281–308 (2013).

    CAS  Article  Google Scholar 

  50. Dobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29, (15–21 (2013).

    Google Scholar 

  51. Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq. 2. Genome Biol. 15, 550 (2014).

    Article  Google Scholar 

  52. Zhu, D., Deng, N. & Bai, C. A generalized dSpliceType framework to detect differential splicing and differential expression events using RNA-Seq. IEEE Trans. Nanobioscience 14, 192–202 (2015).

    Article  Google Scholar 

  53. Wang, L., Wang, S. & Li, W. RSeQC: quality control of RNA-seq experiments. Bioinformatics 28, 2184–2185 (2012).

    CAS  Article  Google Scholar 

Download references

Acknowledgements

This project was supported by a Bloodwise Project Grant (13051). F.G.L. and K.I.S. were supported by a grant from the Medical Research Council (MR/K018965/1). F.G.L. and K.I.M. were supported by Leukaemia and Lymphoma NI. F.G.L. & F.G. were funded by the European Research Council (ERC-2014-STG 636855). We thank Alex Thompson for his thoughtful comments and assistance with FACS. We thank Daniela Moralli and Catherine Green from the Wellcome Centre for Human Genetics for the FISH analysis. All bioinformatics analyses made use of the Kelvin High Performance Computing Cluster at Queen’s University Belfast.

Author information

Authors and Affiliations

Authors

Contributions

F.G.L., K.I.S. and K.I.M. conceived the study. F.G.L. performed sequencing experiments, bioinformatics analysis and functional assays, generated the figures and tables, and wrote the manuscript. K.L. performed functional assays and generated the figures and tables. E.M.B. and J.V. performed functional assays. F.G., K.L., K.I.S. and K.I.M. edited the manuscript. K.I.S. and K.I.M. led the study.

Corresponding author

Correspondence to Fabio Giuseppe Liberante.

Ethics declarations

Competing Interests

The authors declare no competing interests.

Additional information

Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Liberante, F.G., Lappin, K., Barros, E.M. et al. Altered splicing and cytoplasmic levels of tRNA synthetases in SF3B1-mutant myelodysplastic syndromes as a therapeutic vulnerability. Sci Rep 9, 2678 (2019). https://doi.org/10.1038/s41598-019-39591-7

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1038/s41598-019-39591-7

Further reading

Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Search

Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing