Cancer-associated SF3B1 mutations affect alternative splicing by promoting alternative branchpoint usage

Abstract

Hotspot mutations in the spliceosome gene SF3B1 are reported in ∼20% of uveal melanomas. SF3B1 is involved in 3'-splice site (3'ss) recognition during RNA splicing; however, the molecular mechanisms of its mutation have remained unclear. Here we show, using RNA-Seq analyses of uveal melanoma, that the SF3B1^R625/K666 mutation results in deregulated splicing at a subset of junctions, mostly by the use of alternative 3'ss. Modelling the differential junctions in SF3B1^WT and SF3B1^R625/K666 cell lines demonstrates that the deregulated splice pattern strictly depends on SF3B1 status and on the 3'ss-sequence context. SF3B1^WT knockdown or overexpression do not reproduce the SF3B1^R625/K666 splice pattern, qualifying SF3B1^R625/K666 as change-of-function mutants. Mutagenesis of predicted branchpoints reveals that the SF3B1^R625/K666 -promoted splice pattern is a direct result of alternative branchpoint usage. Altogether, this study provides a better understanding of the mechanisms underlying splicing alterations induced by mutant SF3B1 in cancer, and reveals a role for alternative branchpoints in disease.

ARTICLE

Received 15 Sep 2015 | Accepted 5 Jan 2016 | Published 4 Feb 2016

Samar Alsafadi1, Alexandre Houy1, Aude Battistella1, Tatiana Popova1, Michel Wassef2, Emilie Henry3, Franck Tirode1, Angelos Constantinou4, Sophie Piperno-Neumann5, Sergio Roman-Roman3, Martin Dutertre6 & Marc-Henri Stern1

Hotspot mutations in the spliceosome gene SF3B1 are reported in B20% of uveal melanomas. SF3B1 is involved in 30-splice site (30ss) recognition during RNA splicing; however, the molecular mechanisms of its mutation have remained unclear. Here we show, using RNA-Seq analyses of uveal melanoma, that the SF3B1R625/K666 mutation results in

deregulated splicing at a subset of junctions, mostly by the use of alternative 30ss. Modelling the differential junctions in SF3B1WT and SF3B1R625/K666 cell lines demonstrates that the

deregulated splice pattern strictly depends on SF3B1 status and on the 3ss-sequence context. SF3B1WT knockdown or overexpression do not reproduce the SF3B1R625/K666 splice pattern, qualifying SF3B1R625/K666 as change-of-function mutants. Mutagenesis of predicted branchpoints reveals that the SF3B1R625/K666-promoted splice pattern is a direct result of

alternative branchpoint usage. Altogether, this study provides a better understanding of the mechanisms underlying splicing alterations induced by mutant SF3B1 in cancer, and reveals a role for alternative branchpoints in disease.

1 Department of Genetics and Biology of Cancers, INSERM U830, Institut Curie, PSL Research University, Paris 75248, France. 2 Depatment of Developmental Biology and Genetics, CNRS UMR 3215/INSERM U934, Institut Curie, PSL Research University, Paris 75248, France. 3 Translational Research Department, Institut Curie, PSL Research University, Paris 75248, France. 4 Department of Molecular Bases of Human Diseases, CNRS UPR 1142, IGH-Institute of Human Genetics, Montpellier 34090, France. 5 Department of Medical Oncology, Institut Curie, Paris 75248, France. 6 Department of Genotoxic stress and Cancer, CNRS UMR 3348, Institut Curie, PSL Research University, Orsay 91400, France. Correspondence and requests for materials should be addressed to M.H.S. (email: mailto:[email protected]

Web End [email protected] ).

NATURE COMMUNICATIONS | 7:10615 | DOI: 10.1038/ncomms10615 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 1

DOI: 10.1038/ncomms10615 OPEN

Cancer-associated SF3B1 mutations affect alternative splicing by promoting alternative branchpoint usage

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10615

Discovery of recurrent missense mutations in splicing factors in cancers revealed the importance of the spliceosome pathway as a direct actor in carcinogenesis

and questioned functional roles and molecular mechanisms of these mutations. SF3B1 (Splicing Factor 3B Subunit 1A) encodes for a core component of the U2 small nuclear ribonucleoprotein (snRNP) complex of the spliceosome and is involved in early stages of splicing. Alterations in SF3B1 were initially discovered in myelodysplastic syndromes (MDSs) and chronic lymphocytic leukemia (CLL), together with other mutations of splicing factors, such as U2AF1, SRSF2 and ZRSR2 (refs 13). Importantly, these genes encode proteins that are all involved in 30-splice site recognition during RNA splicing4. It has been shown that SF3B1 is mutated in a signicant proportion (B20%) of uveal melanoma (UM), a rare malignant entity deriving from melanocytes from the uveal tract57, and in other solid tumours at lesser frequencies8,9.

RNA splicing is a fundamental process in eukaryotes, which is carried out by the splicing machinery (spliceosome) composed of ve snRNPs and additional proteins10. Introns contain consensus sequences that dene the 50 donor splice site (50ss), branchpoint (BP) and 30 acceptor splice site (30ss), which are initially recognized by the U1 snRNP, SF1 protein and U2AF, respectively. U2AF is a heterodimer composed of U2AF2 (also known as U2AF65) and U2AF1 (also known as U2AF35), which recognize the poly-pyrimidine tract and the well-conserved AG dinucleotide sequence of 30ss, respectively. After binding to the 30ss, U2AF facilitates replacement of SF1 by U2 snRNP at the BP. Interaction between U1 and U2 snRNPs then triggers transesterication joining the 50-end of the intron to the BP, most generally an adenosine located in a loosely dened consensus

B25 nucleotides upstream of the 30ss. The 50ss and 30ss are then ligated together and the branched intron is discarded10.

SF3B1 mediates U2 snRNP recruitment to the BP by interacting with the intronic RNA on both sides of the BP and with U2AF11. Structurally, the SF3B1 protein has an N-terminal hydrophilic region containing U2AF-binding motif and a C-terminal region, which consists of 22 non-identical HEAT (Huntingtin, Elongation factor 3, protein phosphatase 2A, Targets of rapamycin 1) repeats. Cancer-associated mutations in SF3B1 are missense mutations with three major hotspots targeting the fth, sixth and seventh HEAT repeats at codon positions R625, K666 and K700, respectively. Interestingly, K700 mutations are by far the most frequent in haematopoietic malignancies, whereas R625 mutations are prevailing in UM. These alterations affect residues that are predicted to be spatially close to one another and therefore might have a similar functional impact1.

Recently, RNA-sequencing (RNA-Seq) analysis of CLL, breast cancer and UM showed that a global splicing defect in SF3B1-mutated tumours consists in usage of cryptic 30ss (hereafter called

AG0) located 10 to 30 bases upstream of normal 30ss, yet the underlying mechanism has remained poorly understood. It has been proposed that AG0 is located at the end of a sterically protected region in a specic region downstream the BP. Yet, not every potentially located AG0 was used in an SF3B1MUT context12.

In the present study, RNA-Seq analysis of 74 primary UMs, mutated or not for SF3B1, conrmed the SF3B1MUT-promoted pattern identied by DeBoever et al.12, demonstrating the robustness of the deregulated pattern. By constructing in cellulo models, we show that SF3B1MUT is the direct cause of the deregulated splice pattern and could be qualied neither as gain-of-function (that is, hyperactivity) nor loss-of-function, but rather as change-of-function mutants. Our experiments provided evidence that (i) mutant SF3B1 preferentially recognizes alternative BPs upstream of the canonical sites and (ii) the alternative 30ss used in a SF3B1MUT context are less dependent on

U2AF. We propose a model of the SF3B1MUT dysfunctions that sheds new light on the mechanism of splicing dysregulation in cancer. In addition, our data reveal a currently under-appreciated role for recently described alternative branch points13 in alternative splicing and disease.

ResultsSF3B1 mutations promote upstream alternative acceptors. Following initial nding of recurrent mutations of SF3B1 gene in UM5, we set up an independent consecutive series of UM to analyse the effect of SF3B1 hotspot mutations. This series included 74 T2T4 tumours of different histologic types (21 epithelioid cell, 18 spindle cell and 35 mixed cases) treated by primary enucleation. Thirty-eight cases (51%) subsequently developed metastases and 40 patients (54%) died. SF3B1 mutations were found in 16 tumours affecting two hotspotsp.R625 and p.K666 (Supplementary Table 1). No mutation of other genes coding for splicing factors was observed. SNP array analysis did not reveal any chromosome loss or gain in the region containing SF3B1 (2q33.1). The overall mutation rate of 22% (16/74) is comparable to the rate (19%) recently reported for SF3B1 mutations in UM57.

To evaluate the effects of SF3B1 mutations on splicing, we performed transcriptome analysis of the UM cohort using RNA-Seq technique. Differential analysis of splice junctions between the SF3B1MUT (n 16) and SF3B1WT (n 56) tumours

using DESeq2 (ref. 14) revealed an overall high level of differences. The top 1,469 differentially spliced junctions with P-values r10 5 (Benjamini-Hochberg) and absolute Log2(fold change)Z1 were selected for further analyses (Supplementary Fig. 1 and Supplementary Data 1). A hierarchical clustering of the 74 tumours using the 1,469 differential splicing junctions showed coherent changes in SF3B1MUT tumours (Fig. 1a). A single SF3B1WT tumour clustered together with SF3B1MUT cases. Manual reanalysis excluded any variant of its SF3B1-coding sequence or any over- or under-expression of the SF3B1 transcript and exome sequencing of this case failed to identify any mutation of the spliceosome genes as a potential genocopy.

Interestingly, 72% (1,060/1,469) of differential splice variants had no Ensembl Transcript identiers (ENST) and these novel splice variants were found almost exclusively in SF3B1 mutants. To be noticed, only 9% (12,866/142,458) of non-differential splice variants had no ENST. The acceptor splice site (30ss) was altered in 1,124 differential junctions (76.5%), whereas the altered donor splice site (50ss) was observed in 186 differential junctions (12.7%;

Fig. 1b). For 159 junctions (10.8%), the novel junction was either ambiguously attributed to the alternative 50 or 30ss, or attributed to both alternative 50ss and 30ss.

The analysis of distances between the alternative and canonical SF3B1MUT-sensitive 3ss showed repetitive peaks of alternative 3ss every three nucleotides (Fig. 1c). Such spacing of two nucleotides suggests that frameshift variants are targeted by nonsense-mediated mRNA decay.

We observed that the majority (765 out of 1,124) of the SF3B1MUT-promoted alternative 30ssthereafter named

AG0were located within 50 nucleotides (nts) that precede the canonical 30ssthereafter named AGwith a clear clustering in the 12 to 24 nt region upstream of the canonical AG

(Fig. 1c). No ENST Identier exists for 675 out of these 765 AG0 alternative acceptor sites (88%).

These results are concordant with a recent study based on RNA-Seq data from CLL, breast cancer and UM samples12. DeBoever et al. showed that 619 cryptic 30ss clustering 1030 nucleotides upstream of canonical 30ss were used in cancers with SF3B1 mutations. Interestingly, we found 327 out of these 619

2 NATURE COMMUNICATIONS | 7:10615 | DOI: 10.1038/ncomms10615 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10615 ARTICLE

a

Count

0

Colour key

and histogram

4

2 0 2 4

Row Z-score

SF3B1K666M

SF3B1K666T

SF3B1R625C

SF3B1R625G

SF3B1R625H

SF3B1R625L

SF3B1WT

Junctions

Samples

b

1,124

159 186

Donors

(5 ss)

Acceptors

(3 ss)

c

Counts

100

95

90

85

80

75 65

60

55

50 40

35

30 20

15

5

0

54

51

42

39

33

30

27

24

21

18

15

12

6

3

70

45

25

AG

10

45

48

36

9

9 15

18

21

24

27

30

33

36

39

42

45

48

Distance (nts)

Figure 1 | Differential splice junctions in SF3B1MUT tumours. (a) Hierarchical clustering and heat-map analysis of differential splice junctions in tumour samples. The colours of squares below the tree denote the subtype of each sample. Below the array tree and the subtype identication row, the heat map of the 1,469 splice junctions is shown. The complete list of up- and downregulated splice junctions can be found in Supplementary Data 1. (b) Venn diagram of differential splice junctions in SF3B1MUTcompared with SF3B1WT tumours. Numbers show the count of alterations within only 50ss (186 events) or only 30ss (1,124 events). The overlapping area represents junctions that are either ambiguously attributed to an alternative donor or acceptor site, or attributed to both alternative 30 and 50 splice sites (159 events). (c) Distances between the alternative and canonical SF3B1MUT-sensitive 30ss. For alternative 30ss within the 50 nts preceding the canonical 30ss (765), the distance between the alternative (AG) and corresponding canonical (AG) 30ss was plotted as a histogram. Negative distances mean the alternative AG upstream of the canonical AG, whereas positive distances mean the AG downstream. The 0 point demarks the position of the canonical AG.

cryptic 30ss (53%) to be differentially expressed in our data set, demonstrating the robustness of this splicing pattern despite the differences in the series of analysed tumours and bioinformatics pipelines.

Sequence context determines acceptor sensitivity to SF3B1MUT. To validate the splice pattern detected in SF3B1MUT tumours and to determine if it is conferred by sequences in the region of the 30ss, we performed minigene splicing assay. We selected seven

NATURE COMMUNICATIONS | 7:10615 | DOI: 10.1038/ncomms10615 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 3

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10615

Table 1 | Sensitive and insensitive 30ss selected for validation in cell lines.

Gene ID Position Alternative splice ratio ([AG/AG]*100)Wild-type SF3B1 Mutated SF3B1MP41 HEK293T Mel202 SF3B1K666T HEK293T

Sensitive 3ssENOSF1 chr18:683,349-683,440 3 3 60 23 TMEM14C chr6:10,724,123-10,725,452 7 8 69 43 DPH5 chr1:101,458,237-101,458,383 1 1 66 15 ZNF76 chr6:35,257,930-35,258,130 3 2 41 12 DLST chr14:75,356,550-75,356,612 1 1 33 14 CHTF18 chr16:844,008-844,060 0 1 55 24 ARMC9 chr2:232,209,626-232,209,696 0 0 65 26

Insensitive 3ssWRAP73 chr1:3552508-3551858 0 0 0 0 VPS45 chr1:150044293-150048311 0 0 0 0

30ss within the top differential splice junctions associated with SF3B1MUT tumours (named as sensitive 30ss) and two 30ss found unaltered in an SF3B1MUT context (named as insensitive 30ss; Table 1). Regions containing the selected 30ss were cloned into an ExonTrap vector and expressed into human cell lines with different SF3B1 status: 2 UM cell lines, MP41 (SF3B1WT) and Mel202 (SF3B1R625G; Supplementary Figs 2 and 3). In addition, to directly determine whether the splice pattern detected in SF3B1MUT tumours is dependent on SF3B1 status, we used two isogenic HEK293T cell lines, SF3B1WT and SF3B1K666T obtained by the CRISPR/Cas9 technology. RNA-Seq data, as visualized by IGV (Integrative Genomics Viewer), showed different mutation rates of 30% in Mel202 cells and 14% in SF3B1K666T HEK293T cells, because of multiple copies of the wild-type SF3B1 allele in these aneuploid cell lines. As expected, Mel202 (SF3B1MUT) clusters with the SF3B1MUT tumours and MP41 (SF3B1WT) clusters with SF3B1WT tumours. Probably due to its low level of SF3B1K666T expression, SF3B1K666T-HEK293T clusters together with HEK293T (WT) and SF3B1WT tumours (Supplementary Fig. 2).

The splice forms corresponding to canonical AG usage were found expressed after transfection of the insensitive 30ss constructs in all cell lines, regardless of their SF3B1 status (Fig. 2a). Likewise, for the seven sensitive 30ss constructs, the splice forms corresponding to canonical AG usage were found expressed in the SF3B1WT cell lines MP41 and HEK293T. By contrast, the SF3B1MUT cell lines Mel202 and SF3B1K666T HEK293T expressed the alternative splice forms using the alternative AG0 in addition to the canonical AG (Fig. 2a).

The correspondence between band sizes and splice usage was veried by Sanger sequencing. Interestingly, the ratio of the alternative AG0 versus canonical AG usage (AG0/AG) based on capillary electrophoresis proles varied according to the SF3B1MUT/SF3B1WT rate in the cell lines (Fig. 2b,c). Of note, a faint but signicant usage of alternative AG0 in SF3B1WT cell lines was detected on the capillary electrophoresis proles for three sensitive 30ss, ENOSF1, TMEM14C and ZNF76 (AG0/AG in

SF3B1WT cell lines 0.2, 0.1 and 0.07, respectively), implying that

the AG0 usage may be reinforced rather than induced de novo in an SF3B1MUT context (Fig. 2b).

In conclusion, we demonstrate that the aberrant splice pattern is strictly dependent on the SF3B1MUT status and on sequences in the close vicinity of the sensitive 30ss.

SF3B1 hotspot mutations are change-of-function mutations. The mode of action of SF3B1 mutant was then addressed by analysing endogenous DPH5 and ARMC9 transcripts. The different cell lines were transiently transfected with expression

vectors for SF3B1WT and SF3B1K700E and examined 48 h later for the AG0/AG usage of endogenous 30ss (Fig. 3a). We represent the shift from the canonical AG to the alternative AG0 by the AG0/AG index, which is the ratio of mRNA expression of AG0 form to AG form of a validated gene, DPH5 or ARMC9, determined by quantitative reverse transcription (RT)PCR. The overexpression of SF3B1K700E signicantly increased the AG0/AG index in SF3B1WT cell lines (10- and 32-fold increases for DPH5 in MP41 and HEK293T, respectively), whereas overexpression of SF3B1WT had no effect on the AG0/AG index. The overexpression of SF3B1K700E increased by only three-fold the AG0/AG index in SF3B1K666T HEK293T (transcript mutation rate 14%) and did not modify it in

Mel202 cell line (transcript mutation rate 30%), which may

indicate a saturating effect of SF3B1MUT. Similar results were obtained with the endogenous sensitive 30ss of ARMC9. We conclude that SF3B1 mutation does not lead to a hyper-activity of the protein, as its phenotype is not reproduced by SF3B1 overexpression.

To determine whether SF3B1 mutations are loss-of-function mutations, we then assessed the effect of SF3B1 short interfering RNA (siRNA)-mediated knockdown on alternative splicing in SF3B1WT HEK293T and MP41 and in SF3B1MUT Mel202. Non-target siRNA was used as a negative control and siRNA-mediated knockdown was conrmed by immunoblotting (Fig. 3b). As shown in Fig. 3b, SF3B1 siRNA-mediated knockdown did not have any signicant effect on AG0/AG index despite up to 93% of SF3B1 protein level reduction. These ndings demonstrate that the SF3B1MUT splice pattern is not mimicked by SF3B1 knockdown.

Altogether, our results provide the rst evidence that SF3B1 mutants are neither gain- (hyperactive) nor loss-of-function mutants, and suggest change-of-function consequences.

SF3B1MUT-promoted AG0 are weakly dependent on U2AF. As sensitivity to SF3B1 mutants was conferred by sequences in the close vicinity of 30ss (Fig. 2), we searched for a sequence pattern associated with sensitive 30ss. We compared the sequences of alternative AG0, corresponding canonical AG, and insensitive 30ss (Fig. 4a). Two obvious features were found associated with AG0 consensus sequence: the paucity of G nucleotide at 1 position,

and the frequency of A nucleotides at 11 to 14. Only 20% of

G was observed for the alternative AG0, compared with B50% of G for both canonical AG and insensitive 30ss (Fig. 4b). The high proportion of AG-(C/T) in the alternative AG0 site (65% versus 23% in AG0 and AG, respectively) is best explained by the presence of the AG0 within the polypyrimidine tract of the

4 NATURE COMMUNICATIONS | 7:10615 | DOI: 10.1038/ncomms10615 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10615 ARTICLE

a

Insensitive sequences

WRAP73 VPS45 ENOSF1

Sensitive sequences

TMEM14C

AG

MP41

MEL202

293T WT

293T

MP41

MEL202

293T WT

293T 666

MP41

MEL202

293T WT

293T 666

MP41

MEL202

293 TWT

293T 666

AG 3

5

AG+AG

AG+AG

AG

AG

AG AG

AG

AG

AG

Sensitive sequences

ZNF76

MEL202

DPH5 ARMC9

DLST

MEL202

MP41

293T WT

293T 666

MP41

MEL202

293T WT

293T 666

MP41

293T WT

293T 666

MP41

MEL202

293T WT

293T 666

AG+AG

AG+AG

AG+AG

AG+AG

AG AG

AG

AG

AG

AG

AG AG

b

ENOSF 1

MPP41 (SF3B1WT)

40

60

80

100

120

140

160

180

200

220

240

260

280

300

320

340

360

380

400

420

440

460

480

500

520

40

60

80

100

120

140

160

180

200

220

240

260

280

300

320

340

360

380

400

420

440

460

480

500

520

HEK293T (SF3B1WT)

AG

AG

000 500 000 500 000 500 000 500 000 500 000 500 000 500 000 500 000

10,000

9,500 9,000 8,500 8,000 7,500 7,000 6,500 6,000 5,500 5,000 4,500 4,000 3,500 3,000 2,500 2,000 1,500 1,000

500

0

AG

AG

500 000 500 000 500 000 5000

AG

AG

203

303

319 nts

Mel202 (SF3B1MUT)

SF3B1K666T HEK293T

40

60

80

100

120

140

160

180

200

220

240

260

280

300

320

340

360

380

400

420

440

460

480

500

520

40

60

80

100

120

140

160

180

200

220

240

260

280

300

320

340

360

380

400

420

440

460

480

500

520

RFU

9,500 9,000 8,500 8,000 7,500 7,000 6,500 6,000 5,500 5,000 4,500 4,000 3,500 3,000 2,500 2,000 1,500 1,000

5000

AG

AG

nts

ENOSF1

c

TMEM14C

AG/AG

DPH5

ZNF76

DLST

CHTF18

ARMC9

0 0

0.2

SF3B1WT

SF3B1MUT

Mel202

0 3

HEK293T MP41

SF3B1K666T-HEK293T

0.1 0.1

0 0

0 0

0

0.2

0.1 0.1

0.1

0.9 0.6 0.6

0.5 0.5

0.9

0.9

0.9

3.1 1.1 2.4 2.9

0.4 0.4

Figure 2 | In cellulo validation of differential splice junctions. (a) Minigene splice assay of two SF3B1MUT-insensitive 30ss (WRAP73 and VPS45) and6 SF3B1MUT-sensitive 30ss (TMEM14C, ENOSF1, ZNF76, DPH5, DLST, ARMC9). Gel electrophoresis shows the different splicing processes for minigene

ExonTrap constructions in SF3B1WT cell lines (MP41 and HEK293T) and SF3B1MUT cell lines (Mel202 and K666T-SF3B1 HEK293T). The lower band corresponds to the variant generated by the usage of the canonical 30ss (AG). The intermediate band corresponds to a splice generated by the usage of the alternative 30ss (AG). The upper band is the heteroduplex formation from two bands (AG and AG). (b) Analysis of alternative AG and canonical AG usage of the ExonTrap construct (ENOSF1) in cell lines by capillary electrophoresis of RTPCR products. Representative GeneMarker electrophoregrams for fragment analysis of ENOSF1 minigene cDNA expression are shown. The x axis represents molecular size (in nucleotides (nts)) of PCR products, and the y axis indicates relative uorescent units (RFUs). The peak of 203 nts refers to the internal splicing of the pET01 ExonTrap vector using its 30ss and 50ss. The peak of 303 nts corresponds to the usage of the canonical AG, whereas the peak of 319 nts corresponds to the usage of alternative AG WT. (c) Heat-map analysis using the AG/AG ratio of the top differential splice junctions in cell lines as determined by capillary electrophoresis of RTPCR products.

corresponding canonical AG. AG-G is part of the recognition motif of U2AF1 (ref. 15). The low frequency of G nucleotide immediately following alternative AG0 sites suggests lesser dependence to U2AF1 compared with the canonical AG at the

step of 3ss recognition1618. To test this hypothesis, MP41 and HEK293T cells were transiently transfected with siU2AF1 or siU2AF2. Efcient knockdown was conrmed by immuno-blotting. As shown in Fig. 4c, both U2AF1 and U2AF2

NATURE COMMUNICATIONS | 7:10615 | DOI: 10.1038/ncomms10615 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 5

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10615

knockdowns increased the AG0/AG index of DPH5 and ARMC9 in SF3B1WT cells, and had no signicant effect in SF3B1R625/K666 cells (Fig. 4c). These ndings suggest that AG0 is less dependent on U2AF than the competing canonical AG.

One hypothesis to explain why U2AF knockdown partially mimicked the effect of SF3B1MUT was that the SF3B1U2AF2 interaction11 might be decreased in the case of mutant SF3B1. However, U2AF2 and U2AF1 antibodies immunoprecipitated

pCMV-3tag-1A

SF3B1 SF3B1 SF3B1 SF3B1 WT MUT

250 kDa

130 kDa

a

SF3B1K666T HEK293T

MP41

Mel202

HEK293T

Ctl WT MUT

Ctl WT MUT

Ctl WT MUT

Ctl

pCMV-3tag-1A Mock pCMV-3tag-1A Mock pCMV-3tag-1A

Mock pCMV-3tag-1A

Mock pCMV-3tag-1A Mock pCMV-3tag-1A Mock pCMV-3tag-1A

Mock pCMV-3tag-1A

Mel202

5 siCtl 4 5 siCtl 4 5

71 38

80

100

7

70

100

15

Flag

-Actin

55 kDa

12

*** *

NS NS ***

NS

50

4

1.6

NS

NS

10

40

3

1.2

8

30

DPH5

AG/AG index

6

2

0.8

20

4

1

2

10

0.4

0

0

0

0

Empty

WT-SF3B1

MUT-SF3B1

Empty

WT-SF3B1

MUT-SF3B1

Empty

WT-SF3B1

MUT-SF3B1

Empty

WT-SF3B1

MUT-SF3B1

Mock

MP41

HEK293T

Mel202

K666T- SF3B1 HEK293T

*** *

NS NS

NS NS

50

160

1.6

NS

NS

1.6

AG/AG index

40

120

1.2

1.2

30

ARMC9

80

0.8

0.8

20

10

40

0.4

0.4

0

0

0

0

Empty

WT-SF3B1

MUT-SF3B1

Empty

WT-SF3B1

MUT-SF3B1

Empty

WT-SF3B1

MUT-SF3B1

Empty

WT-SF3B1

MUT-SF3B1

K666T- SF3B1 HEK293T

SF3B1WT cell lines SF3B1MUT cell lines

MP41

HEK293T

Mel202

b

HEK293T MP41

siSF3B1

siSF3B1

siCtl

siSF3B1

4

SF3B1

-Actin

250 kDa 130 kDa

55 kDa

(%)

100

2

2

2

AG/AG index

1.5

1.5

1.5

DPH5

1

1

1

0.5

0.5

0.5

0

0

0 siControl siSF3B1_4 siSF3B1_5 siControl siSF3B1_4 siSF3B1_5 siControl siSF3B1_4 siSF3B1_5

HEK293T

MP41

Mel202

6 NATURE COMMUNICATIONS | 7:10615 | DOI: 10.1038/ncomms10615 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10615 ARTICLE

The observed consequences of TMEM14C mutants were the following: (i) the -17A4G_-16A4G-double mutation completely abolished the usage of the alternative AG0 in both recipient cell lines; (ii) the -13A4G mutation had no consequence on 30ss usage; (iii) the -6A4G mutation completely abolished the usage of the canonical AG; (iv) the -2A4G mutation completely abolished the usage of the alternative AG0 as it destroyed the AG0 site.

We interpret these data as indicating that the A at 30 nts upstream of AG is the BP0 for AG0, as its mutation allowed only the use of the canonical AG regardless of SF3B1 status.Alternatively, -6A4G mutation switched the usage of the AG to the usage of AG0, arguing that this site may serve as the BP for the canonical AG. Our data therefore support the existence of two branchpoints, BP and BP0, differentially used depending on the

SF3B1 status.

Similar observations were obtained with the ENOSF1 construct conrming the existence of two different BPs. Specically, we could determine the BP0 of AG0 at 29 nts preceding the AG (-29A4G). The -18A4G mutation disturbed the usage of both AG0 and AG, whereas the -17A4G mutation disturbed AG usage and inhibited the usage of AG0. In fact, the later mutation created another alternative acceptor site replacing the AG0 (AAG4AGG), and competing with the canonical AG. The potential BP was loosely dened for ENOSF1, because of the multiple adenosines in the vicinity of AG0 participating to the usage of the canonical AG. To be noticed, the -15A4G mutation of the nucleotide immediately following the AG0 strengthened the alternative site, presumably by reinforcing the binding of U2AF1 to the AG0-G site. Thus, the analyses of both the TMEM14C and ENOSF1 genes indicate that SF3B1MUT affects the 30ss choice by promoting the use of alternative BPs.

SF3B1 plays a major role in U2 snRNP recruitment to the BP.

To determine whether the potential of BP sequences to form base-pairing interaction with U2 snRNA (small nuclear RNAs) can modulate the sensitivity to SF3B1 mutations, BP and BP0 mutants of TMEM14C were generated (Supplementary Fig. 5).The strength of the resulting BPs was estimated by their SVM score20 (Fig. 5c). The TMEM14Cmut1 allows a perfect base-pairing of BP with U2 snRNA21; TMEM14Cmut2 contains a suboptimal BP; TMEM14Cmut3 contains a defective alternative BP0; TMEM14Cmut2 3 includes both TMEM14Cmut2 and

TMEM14Cmut3 mutated regions and TMEM14Cswap contains swapped endogenous BP and BP0 sequences.

The consequences of these mutants were then assessed by the AG0/AG index (RTPCR). Enhancing the base-pairing of the

BP region (TMEM14Cmut1), disrupting BP0 (TMEM14Cmut3) or combining a disrupted BP0 with a suboptimal BP (TMEM14Cmut2 3) led to a total inhibition of AG0 usage, regardless of the SF3B1 status. Decreasing the strength of

BP (TMEM14Cmut2) led to a reinforcement of AG0 usage (Fig. 5c). Interestingly, swapping the BP and BP0 sequences

Figure 3 | Overexpression and underexpression of wild-type SF3B1 do not reproduce the splice pattern of SF3B1 hotspot mutations. (a) Effect of overexpression of wild-type and mutated SF3B1 on the AG/AG ratio of DPH5 and ARMC9 in SF3B1WTand SF3B1MUTcell lines. MP41, Mel202, HEK293Tand SF3B1K666T-HEK293Tcell lines were transiently transfected with expression vectors for SF3B1WTand SF3B1K700E. The protein overexpression was conrmed by immunoblotting with anti-Flag using b-actin as a loading control (upper panel). Ratios of expression levels of alternative AG and canonical AG forms (AG/AG) of DPH5 and ARMC9 were determined by quantitative RTPCR (lower panel). The results are average of three replicates and are represented as means.d. Paired t-test was used to generate the P-values comparing each condition to the empty vector transfection: NS, non-signicant; *P40.05;

***Po0.001. (b) Effect of siRNA-mediated knockdown of SF3B1 on the AG/AG ratio in cell lines. HEK293T, MP41 and Mel202 cells were transiently transfected with non-target control siRNA or two different siSF3B1: 4and 5. Proteins and RNA were extracted at 48 h after transfection. siRNA-mediated knockdown was conrmed by immunoblotting with anti-SF3B1, using anti-b-actin as a loading control. Numbers represent the protein band intensity normalized to b-actin and expressed as a percentage of control samples (upper panel). Ratios of expression levels of alternative AG form to the expression level of canonical AG form (AG/AG) of DPH5 were determined by quantitative RTPCR (lower panel). The results are average of three replicates and are represented as means.d.

equally SF3B1WT and SF3B1K700M, implying no detectable alteration in the SF3B1MUTU2AF interaction (Supplementary Fig. 4).

Considering that U2AF1 mutations are reported in B10% of patients with MDS and associated with partial functional impairment in regulated splicing2,19, we tested two MDS samples each harbouring one of the two U2AF1 hotspot mutations, S34F and Q157P. Neither of these MDS samples presented any increase of the AG0/AG index of DPH5, demonstrating that U2AF1 and SF3B1 hotspot mutations do not lead to the same aberrant splicing phenotype (Fig. 4d).

Overall, our results exclude a defective SF3B1MUTU2AF interaction and show that U2AF1MUT and SF3B1MUT can induce different splicing patterns. Importantly, the rarity of G at the position 1 after AG0 as well as the increase of AG0/AG

transcript ratio when U2AF is depleted suggest that SF3B1MUT-promoted 30ss (AG0) is less dependent on U2AF as compared with downstream canonical 30ss (AG).

Alternative BP usage in an SF3B1MUT context. The second feature characterizing the AG0 consensus sequence is the presence of frequent adenosines at a distance of 1114 nts preceding the AG0, which could represent alternative BPs (Fig. 4a). Exploring this hypothesis, we investigated whether SF3B1MUT alters BP choice.

We used the online tools SVM (Support Vector Machine algorithm)-BPnder and the Human Splicing Finder to predict the BP of the 744 sensitive 30ss. The predicted BP clustered together at B22 nts upstream the insensitive 30ss. However, the predicted BP for the alternative and canonical 30ss showed a bimodal distribution centred at 5 and 15 nts upstream the AG0, and at 20 and 35 nts upstream AG (Fig. 5a). Using the experimentally determined BP data set recently reported by Mercer et al.13, we found 286 out of the 744 sensitive 3ss with a determined BP in an SF3B1WT context (Fig. 5a). Remarkably, we found that 37% (105/ 286) of the A of the BP coincided with the A of AG0. Most of the other BP were closely distributed upstream of AG0 at an average of 5 nts. This superimposition of BP and AG0 made unlikely the usage of the same BP for both the canonical AG and alternative AG0. We thus suspected the usage of alternative branchpoint (BP0), possibly corresponding to the second peak of predicted BP around 35 nts 50 to the AG and B15 nts 50 to the AG0 (Fig. 5a).

To explore such hypothesis, we mutated all adenosines within a region of 30 nts preceding the canonical AG in two sensitive sequences, TMEM14C and ENOSF1. These two minigenes were selected for their low frequency of adenosines upstream the 30ss in order to limit the number of required site-directed mutations. We then expressed these variant acceptor sequences in MP41 (SF3B1WT) and Mel202 (SF3B1MUT) cells followed by RTquantitative PCR and fragment analysis by capillary electrophoresis as described above (Fig. 5b).

NATURE COMMUNICATIONS | 7:10615 | DOI: 10.1038/ncomms10615 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 7

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10615

a b

Alternative AG

1.00.5

0.0 50 45 40 35 30 25 20 15 10 5

Acceptor sequences (748)

Acceptor sequences (748)

Probability

Sensitive

sites

100%

0 5 10 15 20 25 30 35 40 45 50

80%

Canonical AG

Acceptor sequences (765)

60%

Probability

1.00.5

0.0 50 45 40 35 30 25 20 15 10 5 0 5 10 15 20 25 30 35 40 45 50

40%

20%

0%

Insensitive

sites

1.00.5

0.0 50 45 40 35 30 25 20 15 10 5 0 5 10 15 20 25 30 35 40 45 50

Alternative Canonical Control

3 ss

A C T G

Probability

c

SF3B1K666T

SF3B1K666T

MP41

Mel202 Mel202

MP41

HEK293T

HEK293T HEK293T HEK293T

7 siCtl 6 7 siCtl 6 7 siCtl 6 7

siU2AF1

siU2AF1 siU2AF1

siU2AF2

2 9 siCtl 2 9 2 9

siU2AF2

siU2AF2 siU2AF2

siCtl

siU2AF1

6

siCtl

siCtl 2 9

siCtl

U2AF35 U2AF65

25 kDa

70 kDa

55 kDa

55 kDa

-Actin -Actin

(%) (%)

100 100

25

22

20

20

100 64 37 100 22 15 100 39 38 100 53 38

100

20

11

100

11

7

8

7

4

5

7

6

6

4

5

DPH5

AG/AG index

3

5

4

3

4

2

3

3

2

2

2

1

1

1

1

0

0 siCtl 6 7 2 9

siCtl 6 7 2 9

0 siCtl 6 7 2 9 siCtl 6 7 2 9 siCtl 6 7 2 9

siU2AF1 siU2AF2

siU2AF1 siU2AF2

siU2AF1 siU2AF2

HEK293T

siU2AF1 siU2AF2

Mel202

0

siU2AF1 siU2AF2

K666T-SF3B1-HEK293T

MP41

4

5

2

3

1.8

3

4

1.6

ARMC9

AG/AG index

1.4

3

1.2

2

2

1

2

0.8

1

0.6

1

1

0.4

0.2

0

0

0 siCtl 6 7 2 9 siCtl 6 7 2 9

siU2AF1 siU2AF2 siU2AF1 siU2AF2

0 siCtl 6 7 2 9

siU2AF1 siU2AF2

MP41

HEK293T

Mel202

K666T-SF3B1-HEK293T

d

1,600

DPH5

AG/AG index

1,200

800

400

0

WT-SF3B1 MUT-SF3B1 MUT-U2AF35S34F

MUT-U2AF35 Q157P

MDS

UM

Tumour samples

Figure 4 | Characterization of alternative 30ss (AG) sequences. (a) Comparison of sequence logos of 30ss sensitive to SF3B1 status with canonical (AG) and alternative (AG) sequences and 30ss insensitive to SF3B1 status. One-hundred-nucleotide-long sequences surrounding the 30ss were used to generate sequence logos with WebLogo. The height of each letter indicates the preference strength for that nucleotide at each position. (b) The proportion of the nucleotides immediately following the alternative (AG), the canonical (AG) and insensitive 30ss. (c) Effect of U2AF35 and U2AF65 siRNA-mediated knockdown on the AG/AG ratio in cell lines. MP41 and HEK293T (SF3B1WT) and Mel202 and SF3B1K666T-HEK293T (SF3B1MUT) cells were transiently transfected with non-target control siRNA, siU2AF35 or siU2AF65. Proteins and RNA were extracted at 48 h after transfection. siRNA-mediated knockdown was conrmed by immunoblotting with anti-U2AF35 and anti-U2AF65, using anti-b-actin as a loading control. Numbers represent the protein band intensity normalized to b-actin and expressed as percentage of control samples (upper panel). Ratio of expression levels of alternative AG form to the expression level of canonical AG form (AG/AG) of DPH5 and ARMC9 was determined by quantitative RTPCR (lower panel). The results are average of three replicates and are represented as means.d. (d) Effect of U2AF35 hotspot mutations on the AG/AG ratio of DPH5 in MDS tumours. Ratio of expression levels of alternative AG form to the expression level of canonical AG form (AG0/AG) of DPH5 was determined by quantitative RTPCR in two

MDS samples, each harbouring one of the two U2AF35 hotspot mutations, S34F and Q157P and compared with mutated and wild-type SF3B1 uveal melanoma (UM) samples.

8 NATURE COMMUNICATIONS | 7:10615 | DOI: 10.1038/ncomms10615 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10615 ARTICLE

a

Alternative 3 ss (AG)

Canonical 3 ss (AG)

Alternative BP (202)

Canonical BP (162)

Insensitive BP (359)

**

Branchpoint

Validated Predicted

Probability

1.00.50.0

AG

AG

Sensitive

3 ss

100

75

50

25

0

100

75

50

25

0

100

75

50

25

0

50

Counts

Probability

1.00.50.0

5

Insensitive

3 ss

Probability

1.00.50.0

0 5

0

0 5

*** ***

45

40

35

30

25

20

15

10

5

0

Distance (nts)

b

TMEM14C Exontrap cassette

ENOSF1 Exontrap cassette

17_16 13 29 15

17

18

6

2

TTTAACTACCTCTGATCCAGCTTGTTTTCTGCAGG

BP BP AG AG

CTGACCTCCCTGTGAAGAGTCTCTTTTTGCAGA

BP AG AG

MP41

13A>G

Mel202 Mel202

MP41

No mutation

17A>G_-16A>G

17A>G_-16A>G

No mutation

29A>G

18A>G

17A>G

15A>G

2A>G

No mutation

No mutation

6A>G

13A>G

6A>G

29A>G

18A>G

17A>G

15A>G

2A>G

AG

AG

AG

AG

AG

AG

AG

AG

AG%

9 0 4 91 0 46 0 35 100 0

AG% 11

0 56 0 47 83 0 8 0 100

c

Mel202

TMEM14C Exontrap cassette

BP BP AG

AG

4

1

P

M

2

0

2

l

e

M

5

+2

Strength of BP

(SVM score)

3

No mutation

mut1

mut2

mut3

Mut2+3

Swap

No mutation

mut1

mut2

mut3

Mut2+3

Swap

No mutation TMEM14C mut1

TMEM14C mut2

TMEM14C mut3

TMEM14C mut2+3

TMEM14C swap

AG

AG AG

AG

1 100

0

MP41

AG%

Figure 5 | Identication of alternative branchpoint usage in an SF3B1MUTcontext. (a) Analysis of distances between the branchpoints and the associated alternative (AG), canonical (AG) or insensitive 30ss. Left panel: Zero represents the position of the AG, AG or insensitive 30ss. Red bars represent the clusters of branchpoints predicted by the tool of SVM, blue bars represent the experimentally determined BP data set reported by Mercer et al.13. Right

panel: the extracted branchpoint sequence logos generated by WebLogo. **Signicantly different in the three branchpoint patterns (w2 test, Po0.01), ***Signicantly different in the three branchpoint patterns (w2 test, Po0.001). (b) Point mutations of branchpoints in TMEM14C and ENOSF1 ExonTrap constructs. All adenosines within a region of 30 nts preceding the canonical AG were mutated into guanines. Mutant constructs were expressed in MP41 (SF3B1WT) and Mel202 (SF3B1MUT) cells followed by RTqPCR. The lower band corresponds to the variant generated by the usage of the canonical 30ss (AG). The intermediate band corresponds to the variant generated by the usage of the alternative 30ss (AG). The upper band is the heteroduplex formation from two bands (AG and AG0). The numbers represent the ratio of AG usage as determined by capillary electrophoresis. (c) Base-pairing potential mutants of TMEM14C. Mutant constructs (sequences shown in Supplementary Fig. 5) were expressed in MP41 (SF3B1WT) and Mel202 (SF3B1MUT) cells followed by

RTqPCR. The lower band corresponds to the variant generated by the usage of the canonical 30ss (AG). The upper band corresponds to the variant generated by the usage of the alternative 30ss (AG). A schematic presentation of the strength of the resulting branchpoints as estimated by their SVM score is shown on the right panel. The ratio of AG usage as determined by capillary electrophoresis in MP41 and Mel202 cells is shown as a heat map.

(TMEM14Cswap) decreased AG0 usage, which could be interpreted as a higher strength of BP0 as compared with BP to form base-pairing interactions with U2 snRNA.

We extended this nding by an in silico comparison of the sequence patterns of alternative, canonical and insensitive BPs. We show that the canonical, alternative and insensitive BP

NATURE COMMUNICATIONS | 7:10615 | DOI: 10.1038/ncomms10615 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 9

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10615

presented distinct patterns (Fig. 5a), with signicant sequence differences at positions 2, 4 and 6 of the motif ( 5 being

the A of the BP).

These data suggest that SF3B1MUT favours the use of BP0 with stronger base-pairing potential with U2 snRNA compared with the downstream BP.

Collectively, in an SF3B1MUT context, the stronger afnity of BP0 for U2 snRNA when compared with BP may allow the use

BP0 with suboptimal AG0 (not followed by G, Fig. 4b) and may explain the lower dependence of AG0 on U2AF (Fig. 4c).

DiscussionHere, we addressed the consequences of SF3B1 hotspot mutations on splicing in UM and its underlying mechanisms. First, we observed that SF3B1 hotspot mutations in UM are associated with deregulation of a restricted subset (B0.5%) of splice junctions, mostly caused by the usage of alternative 30ss (AG0) upstream of the canonical 30ss (AG). This nding is concordant with a recent publication12, implying the robustness of the deregulated splice pattern in SF3B1MUT tumours. Furthermore, this pattern is shared by tumours having for origin different cell lineages12,22. Second, we show here that SF3B1MUT pattern was reproduced neither by knockdown nor by overexpressing wild-type SF3B1, indicating that SF3B1 mutants could be qualied as change-of-function mutants. Third, and important, our data provide signicant progress in understanding the molecular mechanisms underlying alternative 30ss regulation by SF3B1MUT.

We show that this mechanism involves a misregulation of BP0 usage, which have been largely overlooked in previous studies of alternative splicing and have been identied only recently on a large scale13.

Based on in silico data, DeBoever et al. proposed that SF3B1MUT-induced alternative 30ss (AG0) is located at the end of a sterically protected region in a specic region downstream the canonical BP. Yet, not every potentially well-located AG0 was used in an SF3B1MUT context, suggesting additional or different requirements for SF3B1MUT selectivity12. They hypothesized no alternative BP usage as a mechanism of AG0 selection, because of the observed limited distances between AG and AG0. Strikingly, however, we showed here that mutagenesis of the predicted BP0 and the predicted canonical BP abrogated usage of AG0 and AG, respectively, conrming the existence of two BPs differentially used according to SF3B1 status. Interestingly, since submission of our work, Darman et al. reported ndings fully conrming our results. They also showed the consequences of SF3B1 mutations on transcription through the generation of nonsense-mediated mRNA decay-sensitive aberrant spliced transcripts23.

Until recently, only few examples of alternative branchpoints were reported, such as in human XPC and rat bronectin genes24,25. However in 2015, the genome-wide identication of BPs revealed that one-third (32%) of introns have at least two BPs13, but little is known about their regulation. Our ndings provide the rst evidence that misregulation of alternative BPs is involved in physiology or pathology.

Our ndings indicate that SF3B1MUT-induced alternative 30ss usage relies on three properties: an AG0 with lower afnity to U2AF than the canonical 30ss, the presence of an BP0 with a higher afnity to U2 snRNA than canonical BP and the location of BP0 at a distance of 1114 nts preceding the AG0. Based on these ndings and on current understanding of SF3B1 function, we propose the following model for how SF3B1MUT exerts its effects (Fig. 6). Because BP0 potential to form base-pairing interactions with U2 snRNA is generally superior to that of canonical BP, we suggest that U2 snRNP containing SF3B1MUT has more stringent requirement for BP sequences than U2 snRNP-containing SF3B1WT. Consistently, the hotspot mutations

of SF3B1 target the HEAT repeats of SF3B1, which form helical structures that occlude the binding surface for RNA recognition motif of p14, a component of U2 snRNP that binds the BP1,26. The hotspot mutations of SF3B1 in the HEAT repeats occur on the inner surface of the structure and may induce a conformational change in the U2 snRNP complex altering its selectivity for BPs. It is likely that stronger BP0 (in terms of

U2 snRNA complementarity) can compensate for lower AG afnity to U2AF, leading to the recognition of BP0 in a U2AF-independent manner (or less dependent than in the case of canonical BP). This model is supported by our U2AF depletion experiments and is consistent with previous ndings that BP recognition may depend or not on AG binding to U2AF35, according to BP and 30ss sequence and organization11,1618. In contrast, SF3B1WT may allow a more promiscuous binding of U2 snRNA to both canonical and alternative BPs, and in this case, the nal choice of BP may be determined by context, especially 30ss afnity for U2AF.

Further work is required to evaluate the molecular mechanism by which the mutations of the SF3B1 HEAT domains may inuence the base-pairing potential of U2 snRNA. The functional impact of SF3B1MUT-deregulated splicing pattern on oncogenesis also remains to be understood. Meanwhile, our study opens new possibilities for applying the deregulated splicing pattern as a screening tool as well as for targeting the splicing deregulation as a therapeutic strategy in UM and other SF3B1MUT-associated diseases27.

BP

BP

AG(Y)

5 ss

3 ss

3ss

Pre-mRNA

AG(G)

U2

U2AF

AG(G)

U2 snRNA

SF3B1WTT

65 35

p14

5 ss

5 ss

3 ss

BP Y(X=15)

BP Y

Y((XX=115)XX

BP

BP

G(G

AG(G

U2 snRNA BP U2AF AG Strong

Weak

U2

U2AF

65

SF3B1MUT

BP

35

U2 snRNA

p14

3 ss

Y(X=8)

AG(Y)

U2 snRNA BP U2AF AG Weak

Strong

Figure 6 | A model for alternative splicing dysregulation induced by SF3B1 hotspot mutations. The 30ss contains a segment, which is rich in pyrimidines (Y), a well-conserved AG dinucleotide and a branchpoint (BP)

sequence recognized by the U2 snRNP. The U2 snRNP complex binds to the intron through base-pairing interactions between the BP sequence and the U2 snRNA, and through interactions between intron sequences, SF3B1 and p14. The HEAT repeats of SF3B1 form helical structures that occlude the surface of RNA recognition motif of p14. U2 snRNP containing SF3B1WT

recognizes the canonical U2AF-dependant BP. The hotspot mutations of SF3B1 targeting the HEATrepeats occur on the inner surface of the structure and might induce a conformational change in the U2 snRNP complex altering its selectivity for BPs. U2 snRNP containing SF3B1MUT has more

stringent requirement for BP sequences and less for U2AF-dependent sequences, leading to the binding of alternative branchpoints (BP) with high potential of base-pairing with U2 snRNP. AG, canonical 30ss; AG, alternative 30ss; x, average number of pyrimidines; Y, pyrimidine.

10 NATURE COMMUNICATIONS | 7:10615 | DOI: 10.1038/ncomms10615 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10615 ARTICLE

Methods

Patient cohort. A series of 109 consecutive patients diagnosed for UM without metastasis at diagnosis and treated by primary enucleation at the Institut Curie between January 2006 and December 2008 was assembled. RNA extracted from the tumour specimens was qualied for 74 (74/109) cases, which dened the patient cohort for this study.

RNA samples were obtained from surgical residual tumour tissues. In accordance to the national law on the protection of individuals taking part in biomedical research, patients were informed by their referring oncologist that their biological samples could be used for research purposes and they gave their verbal informed consent. All analyses done in this work were approved by the Institutional Review Board and Ethics Committee of the Institut Curie Hospital Group.

DNA and RNA sequencing. Tumour DNA and RNA were provided by the Biological Resource Center of the Institut Curie. The DNA was extracted from frozen tumour or formalin-xed parafn-embedded samples using a standard phenol/chloroform procedure. SF3B1 was sequenced by Sanger methods as previously described5. Primers used for Sanger sequencing are: (forward) 50-CCA

ACTCATGACTGTCCTTTCT-30 and 50-TGGAAGGCCGAGAGATCATT-30.

The total RNA was isolated from frozen tumour samples using a NucleoSpin Kit (Macherey-Nagel). cDNA synthesis was conducted with MuLV Reverse Transcriptase in accordance with the manufacturers instructions (Invitrogen), with quality assessments conducted on an Agilent 2100 Bioanalyzer. Libraries were constructed using the TruSeq Stranded mRNA Sample Preparation Kit (Illumina) and sequenced on an Illumina HiSeq 2500 platform using a 100-bp paired-end sequencing strategy. An average depth of global sequence coverage of 114 million and a median coverage of 112 million was attained.

RNA-Seq analysis. TopHat (v2.0.6)28 was used to align the reads against the human reference genome Hg19 RefSeq (RNA sequences, GRCh37) downloaded from the UCSC Genome Browser (http://genome.ucsc.edu

Web End =http://genome.ucsc.edu). Read counts for splicing junctions from junctions.bed TopHat output were considered. Differential analysis was performed on junction read counts using DESeq2 (ref. 14). Only alternative acceptor splice sites (two or more 30ss with junctions to the same 50ss)

and alternative donor splice sites (two or more 50ss with junctions to the same 30ss) were considered for this analysis.

Fifty-nucleotide-long sequences surrounding the splice acceptor sites were extracted to generate sequence logos using WebLog 3 (http://weblogo.threeplusone.com/

Web End =http://weblogo.threeplusone. http://weblogo.threeplusone.com/

Web End =com/ )29 with the default parameters, the classic colour scheme and the unit frequency being plotted as probability.

The data set supporting the results of this article is available on ArrayExpress repository under the accession E-MTAB-4097.

BP sequence analysis. The online tools SVM-BPnder20 (http://regulatorygenomics.upf.edu/Software/SVM_BP/

Web End =http://regulatorygenomics. http://regulatorygenomics.upf.edu/Software/SVM_BP/

Web End =upf.edu/Software/SVM_BP/ ) and the Human Splicing Finder30 (http://www.umd.be/HSF/

Web End =http://www.umd.be/ http://www.umd.be/HSF/

Web End =HSF/ ) were used to predict the BPs. The SVM_BP code was altered to allow for BP six base pairs from the 30ss by setting minidist3ss 6 in svm_getfeat.py.

Minigene constructs. For each selected candidate gene alternative AG0-centred sequence of B200 nucleotides was PCR amplied from the genomic DNA of HEK293T cells using Phusion Hot Start II High Fidelity DNA Polymerase (Thermo Fisher Scientic). The primer sequence information is provided in Supplementary Table 2. We introduced 15 bases of homology with the ends of the linearized vector at the 50-end of the forward and reverse primers. Using In-fusion HD cloning kit (Clontech), we cloned the amplicon into the BamH1 site of pET01 ExonTrap vector (Mobitec) containing a functional splice donor site (Supplementary Fig. 3).

Wild-type and mutated SF3B1 constructs. A pCMV-3tag-1A vector containing wild-type SF3B1 was synthesized by Genscript Corporation. Because mammalian SF3B1 cDNA sequence was found unclonable in bacteria, a synthetic sequence was generated after codon-optimization for expression in bacteria. The full sequence of codon-optimized SF3B1 is available upon request. K700E mutation in SF3B1 was introduced using QuikChange II Site Directed Mutagenesis Kit (Stratagene). All constructs were veried by DNA sequencing. Primers used for generating the mutated SF3B1 are: (forward) 50-CTGGTGGATGAGCAGCAGGAGGTCAGAA

CCATCTCTGC-30 and (reverse) 50-GCAGAGATGGTTCTGACCTCCTGCTG CTCATCCACCAG-30.

BP mutant constructs. Mutations of potential BP in TMEM14C and ENOSF1 ExonTrap constructs were introduced using QuikChange II Site Directed Mutagenesis Kit (Stratagene) and veried by DNA sequencing. The primer sequences used to generate the mutations are provided in Supplementary Table 3.

Cell culture and transfection. Mel202 cell line was purchased from the European Searchable Tumour Line Database (Tubingen University, Germany) and MP41 (derived at Institut Curie and described in ref. 31) UM cell lines were cultured in

RPMI-1640 supplemented with 10% fetal bovine serum. A point mutation in SF3B1 resulting in K666T amino-acid substitution was introduced using CRISPR/ CAS9-stimulated homology-mediated repair to generate isogenic HEK293T cell lines and was veried by Sanger sequencing. A donor template encoding a puromycin selection cassette was transfected at a 1:1:1 ratio with Cas9 (Addgene 41815) and a SF3B1-specic gRNA (built from gRNA cloning vector, Addgene 41824). The selection cassette was removed by ippase-mediated excision. All cell lines were tested and proved to be Mycoplasma free. Authentication of the cell lines was veried by Sanger sequencing for their mutational status and by RNA-Seq.

Plasmid transfections were carried out in cell lines using 500 ng of plasmid construct and LipofectAMINE 2000 reagent (Invitrogen) according to the manufacturers instructions. After 24 h, total RNA was extracted with NucleoSpin RNA kit (Macherey-Nagel). The quantity and quality of RNA was determined by spectrophotometry (NanoDrop Technologies). Five hundred nanograms of RNA was used as a template for cDNA synthesis with the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Twenty-ve nanograms of the synthesized cDNA was used as a template for RTPCR amplication with specic primers.

HEK293T and MP41 cells were transfected with the following siRNA obtained from Qiagen: SF3B1 (Cat.No. SI00715932 and Cat. No. SI04154647), U2AF1 (Cat.No. SI04158049 and Cat. No. SI04159547); U2AF2 (Cat.No. SI00754026 and Cat. No. SI04194498) or control siRNA (Cat.No. S103650318). The cells were transfected with 50 nM of siRNA using lipofectamine RNAiMAX (Invitrogen). After 48 h, total RNA was extracted and was used as a template for cDNA synthesis. Twenty-ve nanograms of the synthesized cDNA was used for RTPCR amplication with specic primers. PCR products were separated on a 23% agarose gel (Supplementary Figs 6 and 7).

Immunoblot analysis. Cells were lysed in radioimmunoprecipitation assay buffer, and proteins were quantied using a BCA Protein Assay (Pierce). Equal amounts were separated on SDSpolyacrylamide gel electrophoresis gels. Proteins were transferred to nitrocellulose membranes followed by immunoblotting with specic primary antibodies for SF3B1 (1:1,000; #170854; Abcam), Flag (1:1,000, #3165; Sigma), U2AF1 (1:500; #19961; Santa Cruz Biotechnology), U2AF2 (1:500; #53942; Santa Cruz Biotechnology) and b-actin (1:2,000; #5313; Sigma). The membrane was then incubated at room temperature for 1 h with either goat anti-rabbit or goat anti-mouse Odyssey secondary antibodies coupled to a 700 or 800 nm. Immunolabelled proteins were detected using the Odyssey Infrared Imaging System (Li-cor). Quantications were performed using the ImageJ Software. b-Actin immunoblotting was used to quantify and normalize results.

Co-immunoprecipitation. HEK293T cells were co-transfected with pcDNA3.11.-Myc-U2AF2, kindly given by Edwin Chan32, and either pCMV-3tag-1A-SF3B1WT or pCMV-3tag-1A-SF3B1K700M. After 48 h, proteins immunoprecipitated by anti-Flag gel afnity (A2220, Sigma) were separated by SDSpolyacrylamidegel electrophoresis and probed in a western blot assay with anti-U2AF2 and anti-U2AF1 antibody (Supplementary Fig. 8).

Fragment analysis by capillary electrophoresis. Minigene fragments were amplied by RTPCR using a 5 FAM-forward primer and reverse-specic primers (Supplementary Table 2). One microlitre of RTPCR product was added to 18.5 ml of deionized formamide and 0.5 ml HD400 marker (Applied Biosystems). The mixture was then denatured 3 min at 95 C, immediately put on ice, and separated using an ABI 3130xl Genetic Analyzer. The data were analysed using GeneMarker software (SoftGenetics).

References

1. Quesada, V. et al. Exome sequencing identies recurrent mutations of the splicing factor SF3B1 gene in chronic lymphocytic leukemia. Nature Genet. 44, 4752 (2012).

2. Yoshida, K. et al. Frequent pathway mutations of splicing machinery in myelodysplasia. Nature 478, 6469 (2011).

3. Wang, L. et al. SF3B1 and other novel cancer genes in chronic lymphocytic leukemia. N. Engl. J. Med. 365, 24972506 (2011).

4. Zhang, J. & Manley, J. L. Misregulation of pre-mRNA alternative splicing in cancer. Cancer Discov. 3, 12281237 (2013).

5. Furney, S. J. et al. SF3B1 mutations are associated with alternative splicing in uveal melanoma. Cancer Discov. 3, 11221129 (2013).

6. Harbour, J. W. et al. Recurrent mutations at codon 625 of the splicing factor SF3B1 in uveal melanoma. Nature Genet. 45, 133135 (2013).

7. Martin, M. et al. Exome sequencing identies recurrent somatic mutations in EIF1AX and SF3B1 in uveal melanoma with disomy 3. Nature Genet. 45, 933936 (2013).

8. Maguire, S. L. et al. SF3B1 mutations constitute a novel therapeutic target in breast cancer. J. Pathol. 235, 571580 (2015).

9. Kong, Y., Krauthammer, M. & Halaban, R. Rare SF3B1 R625 mutations in cutaneous melanoma. Melanoma Res. 24, 332334 (2014).

10. Wahl, M. C., Will, C. L. & Luhrmann, R. The spliceosome: design principles of a dynamic RNP machine. Cell 136, 701718 (2009).

NATURE COMMUNICATIONS | 7:10615 | DOI: 10.1038/ncomms10615 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 11

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10615

11. Gozani, O., Potashkin, J. & Reed, R. A potential role for U2AF-SAP 155 interactions in recruiting U2 snRNP to the branch site. Mol. Cell Biol. 18, 47524760 (1998).

12. 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).

13. Mercer, T. R. et al. Genome-wide discovery of human splicing branchpoints. Genome Res. 25, 290303 (2015).

14. Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 15, 550 (2014).

15. Wu, S., Romfo, C. M., Nilsen, T. W. & Green, M. R. Functional recognition of the 3 splice site AG by the splicing factor U2AF35. Nature 402, 832835 (1999).

16. Guth, S., Martinez, C., Gaur, R. K. & Valcarcel, J. Evidence for substrate-specic requirement of the splicing factor U2AF(35) and for its function after polypyrimidine tract recognition by U2AF(65). Mol. Cell Biol. 19, 82638271 (1999).

17. Guth, S., Tange, T. O., Kellenberger, E. & Valcarcel, J. Dual function for U2AF(35) in AG-dependent pre-mRNA splicing. Mol. Cell Biol. 21, 76737681 (2001).

18. Pacheco, T. R., Coelho, M. B., Desterro, J. M., Mollet, I. & Carmo-Fonseca, M. In vivo requirement of the small subunit of U2AF for recognition of a weak 30 splice site. Mol. Cell Biol. 26, 81838190 (2006).

19. Shao, C. et al. Mechanisms for U2AF to dene 3 splice sites and regulate alternative splicing in the human genome. Nat. Struct. Mol. Biol. 21, 9971005 (2014).

20. Corvelo, A., Hallegger, M., Smith, C. W. & Eyras, E. Genome-wide association between branch point properties and alternative splicing. PLoS Comput. Biol. 6, e1001016 (2010).

21. Corrionero, A., Minana, B. & Valcarcel, J. Reduced delity of branch point recognition and alternative splicing induced by the anti-tumor drug spliceostatin A. Genes Dev. 25, 445459 (2011).

22. Gentien, D. et al. A common alternative splicing signature is associated with SF3B1 mutations in malignancies from different cell lineages. Leukemia 28, 13551357 (2014).

23. 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, 10331045 (2015).

24. Norton, P. A. Polypyrimidine tract sequences direct selection of alternative branch sites and inuence protein binding. Nucleic Acids Res. 22, 38543860 (1994).

25. Khan, S. G. et al. Two essential splice lariat branchpoint sequences in one intron in a xeroderma pigmentosum DNA repair gene: mutations result in reduced XPC mRNA levels that correlate with cancer risk. Hum. Mol. Genet. 13, 343352 (2004).

26. Schellenberg, M. J. et al. Crystal structure of a core spliceosomal protein interface. Proc. Natl Acad. Sci. USA 103, 12661271 (2006).

27. Bonnal, S., Vigevani, L. & Valcarcel, J. The spliceosome as a target of novel antitumour drugs. Nat. Rev. Drug Discov. 11, 847859 (2012).

28. Kim, D. et al. TopHat2: accurate alignment of transcriptomes in the presence of insertions, deletions and gene fusions. Genome Biol. 14, R36 (2013).

29. Crooks, G. E., Hon, G., Chandonia, J. M. & Brenner, S. E. WebLogo: a sequence logo generator. Genome Res. 14, 11881190 (2004).

30. Desmet, F. O. et al. Human Splicing Finder: an online bioinformatics tool to predict splicing signals. Nucleic Acids Res. 37, e67 (2009).

31. Amirouchene-Angelozzi, N. et al. Establishment of novel cell lines recapitulating the genetic landscape of uveal melanoma and preclinical validation of mTOR as a therapeutic target. Mol. Oncol. 8, 15081520 (2014).

32. Tsoi, H., Lau, C. K., Lau, K. F. & Chan, H. Y. Perturbation of U2AF65/NXF1-mediated RNA nuclear export enhances RNA toxicity in polyQ diseases. Hum. Mol. Genet. 20, 37873797 (2011).

Acknowledgements

We thank Drs Edwin Chan, Olivier Bernard, Vronique de Dalle, Raphael Margueron,

David Gentien, Lisa Golmard, Olivier Kosmider and Professor Michaela Fontenay for

providing key reagents and Drs Didier Auboeuf, Stephan Vagner and Sophie Bonnal and

Juan Valcarcel for helpful discussion and reviewing the manuscript. This work was

supported by INSERM, the French National Cancer Institute (INCa) and the Programme

Incitative et Coopratif Mlanome uveal from Institut Curie. S.A. is supported by the

INCa grant MeluGene.

Author contributions

S.A. designed and performed majority of experiments, interpreted the data and wrote the

manuscript. A.H. performed bioinformatics and statistical analyses. A.B. prepared

primary patient specimens. T.P. provided support for data analysis. M.W. and E.H.

generated the CRISPR/Cas9 cell model, F.T. provided support for data analysis,

A.C. provided critical advice, S.P.-N. and S.R-R. provided patient specimens and critical

advice. M.D. provided critical advice and wrote the manuscript, M.-H.S. conceived and

guided the study, interpreted the data and wrote the manuscript. All authors reviewed

and approved the nal manuscript.

Additional information

Accession codes: The RNA-seq data have been deposited in the ArrayExpress repository

under the accession code E-MTAB-4097.

Supplementary Information accompanies this paper at http://www.nature.com/naturecommunications

Web End =http://www.nature.com/

http://www.nature.com/naturecommunications

Web End =naturecommunications Competing nancial interests: The authors declare no competing nancial interests. Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/

Web End =http://npg.nature.com/

http://npg.nature.com/reprintsandpermissions/

Web End =reprintsandpermissions/

How to cite this article: Alsafadi, S. et al. Cancer-associated SF3B1 mutations affect

alternative splicing by promoting alternative branchpoint usage. Nat. Commun. 7:10615

doi: 10.1038/ncomms10615 (2016).

This work is licensed under a Creative Commons Attribution 4.0

International License. The images or other third party material in this

article are included in the articles Creative Commons license, unless indicated otherwise

in the credit line; if the material is not included under the Creative Commons license,

users will need to obtain permission from the license holder to reproduce the material.

To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

Web End =http://creativecommons.org/licenses/by/4.0/

12 NATURE COMMUNICATIONS | 7:10615 | DOI: 10.1038/ncomms10615 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications