The alternative splicing regulator Tra2b is required for somitogenesis and regulates splicing of an inhibitory Wnt11b isoform

SUMMARY

Alternative splicing is pervasive in vertebrates, yet little is known about most isoforms or their regulation. transformer-2b (tra2b) encodes a splicing regulator whose endogenous function is poorly understood. Tra2b knockdown in Xenopus results in embryos with multiple defects, including defective somitogenesis. Using RNA-seq, we identify 142 splice changes, mostly intron retention and exon skipping, of which 89% are not in current annotations. A previously not described isoform of wnt11b retains the last intron, resulting in a truncated ligand (Wnt11b-short). We show that this isoform acts as a dominant-negative in cardiac gene induction and pronephric tubule formation. To determine the contribution of Wnt11b-short to the tra2b phenotype, we induce retention of intron4 in wnt11b, which recapitulates the failure to form somites but not other tra2b morphant defects. This alternative splicing of a Wnt ligand adds intricacy to a complex signaling pathway and highlights intron retention as a regulatory mechanism.

INTRODUCTION

Genome sequencing projects have shown that complex animals have a similar number of genes to simpler fruit flies and nematodes, raising the question of how increased complexity in anatomy and behavior is encoded. Alternative splicing, which creates multiple transcript isoforms from a single gene, is unusual in simple animals but increases with organismal complexity, raising the possibility of alternative splicing as a mechanism for expanding protein diversity (Nilsen and Graveley, 2010). Indeed, more than 90% of all human multi-exon genes are alternatively spliced (Wang et al., 2008), and many diseases are caused by mutations that perturb either constitutive or alternative splicing (Cooper et al., 2009).

Constitutive pre-mRNA splicing is catalyzed by the spliceosome; a large molecular machine containing U1-U6 RNAs and several hundred proteins (Cooper et al., 2009; Nilsen and Graveley, 2010; Zhou et al., 2002). Alternative uses of splice sites are regulated by auxiliary RNA-binding proteins that bind to the pre-mRNA and either facilitate or repress the use of nearby splice sites (Matlin et al., 2005; Wang et al., 2008).

Vertebrate Transformer-2b (tra2b) is an SR-like protein that contains an RNA Recognition Motif (RRM) flanked by two serine- and arginine-rich (SR) domains (Cooper et al., 2009; Segade et al., 1996). The Drosophila homologue, transformer, promotes splicing and regulates sex-determination through a cascade of alternative splicing (Black, 2003; Will and Luhrmann, 2011). Less is known about the biological function of vertebrate Tra2b, although it has been implicated in several human diseases, including cancer (Matlin et al., 2005; Watermann, 2006). Homozygous tra2b mutant mice die during embryogenesis but the cause is unknown (Mende et al., 2010; Segade et al., 1996). Interestingly, heterozygous mutant mice are morphologically normal, but obese due to dysfunctional lipid metabolism, indicating that the amount of Tra2b protein must be correctly calibrated (Black, 2003; Pihlajamäki et al., 2011). Selective knockout of tra2b in the nervous system results in increased apoptosis and disorganized brain structure (Roberts et al., 2013). However, in none of these cases is it known which splicing changes underlie the biological defects.

We isolated tra2b in a screen for potent mRNA-encoded bioactivities that affect development (Dichmann et al., 2008). Here we show that tra2b is essential for multiple aspects of normal development in Xenopus, including extension of the anterior-posterior axis, somitogenesis, and pronephros formation, all consistent with altered Wnt signaling. Wnt signaling is critical for embryogenesis and tissue homeostasis (MacDonald et al., 2009; Yang, 2012). In Xenopus two wnt11 genes, wnt11 (also called wnt11-r) and wnt11b (Garriock et al., 2005; Mende et al., 2010) encode functionally identical proteins whose expression patterns differ; wnt11b is expressed maternally, and zygotically in the developing mesoderm and somites (Ku and Melton, 1993; Pihlajamäki et al., 2011), whereas both genes are expressed in the neural crest and other tissues at later stages (Garriock et al., 2005; Ku and Melton, 1993; Li et al., 2008; Matthews et al., 2008).

Somite segregation employs multiple signaling pathways, in particular those of FGF, Notch, Retinoic Acid, and Wnt. Several Wnt ligands have been shown to function during somite formation, including Wnt3a and other canonical ligands (Dequéant and Pourquié, 2008). Wnt11 has been shown to function after initial somite formation to direct differentiation of the dermatome and myotome organization (Geetha-Loganathan, 2006; Gros et al., 2008; Morosan-Puopolo et al., 2014). In Xenopus, somitogenesis is initially highly skewed towards muscle differentiation (Gaspera et al., 2012), and Wnt signaling has been implicated in both muscle formation and axial extension (Heisenberg et al., 2000; Hoppler et al., 1996; Tada and Smith, 2000).

Here we analyze the splicing changes after tra2b knockdown, including changes in wnt11b that induce expression of a dominant negative ligand. This identifies a previously unknown layer of regulation of the already complex Wnt signaling pathway and highlights the capacity of intron retention to expand the cell’s proteomic repertoire.

RESULTS

tra2b morphants have developmental defects in all germ layers

To determine the function of Tra2b, we designed two morpholino-oligonucleotides (MOs) to knockdown Tra2b in either X. laevis (tra2b-MO1) or X. tropicalis (tra2b-MO2) (Figure 1A). X. laevis embryos injected with tra2b-MO1 showed delayed gastrulation and a broadened neural plate at stage 14 (Figure 1B,C). At stage 18, tra2b morphants had completed gastrulation, but failed to close the neural tube (Figure 1D,E), and as development proceeded morphants failed to extend the anterior-posterior axis, resulting in shortened embryos (Figure 1F,G). In addition, the endoderm of tra2b morphants dissociated and leaked out of the blastopore prior to hatching (Figure 1G,H). This severe and pleiotropic phenotype points to an essential role for Tra2b in multiple processes during embryogenesis, consistent with the broad tra2b expression during development (Figure S1). Injection of tra2b-MO2 in X. tropicalis resulted in an identical phenotype (data not shown). In addition, injection of tra2b-MO1 into X. tropicalis, or tra2b-MO2 into X. laevis as mismatch controls (each contains five mismatches to the target sequence in the different species), yielded no phenotype supporting the specificity of Tra2b knockdown (data not shown).

Figure 1. Tra2b is required for somite formation and normal embryogenesis.

(A) A translation blocking MO was used in X. laevis, and a splice blocking MO was used in X. tropicalis. (B,C) Delayed gastrulation in tra2b morphants at stage 14; red arrow indicates protruding mesendoderm in morphants (103/110 embryos). (D,E) Neural tube closure defects at stage 18. White arrows indicate fused neural folds in control embryos; red arrows point to neural folds in the open neural plate of morphants (98/101 embryos). (F-H) Axis elongation defects and endoderm detachment at stage 23. Red arrows in (G) and (H) indicate endoderm detaching from the embryo through the blastopore (87/99 embryos). (B-E) Dorsal view with anterior up. (F,G) Lateral view with anterior to the left. (H) Posterior view with dorsal up. (I-R) ISH on control and tra2b morphants. (I-L) Neural plate morphology (sox2) and mesoderm specification (t/bra) at stage 15. (M,N) Paraxial mesoderm forms in tra2b morphants but does not segregate into segmented muscle blocks. White arrow in (M) indicates segregated muscle block in control embryo. (O,P) Pre-somitic mesoderm (psm) is present in tra2b morphants (white brackets) but pre-somitic stripe formation is compromised. White arrow in (O) indicate normal stripe pattern, red arrow in (P) points to smaller and fewer stripes in tra2b morphants. (Q,R) Mature somites marked by hey1 are almost completely absent in tra2b morphants. (S,T) Quantification of hey1-positive somites and pcdh8-positive stripes in control and tra2b morphants. Bars show mean + SD. (***) Indicate that difference compared to control is significant at p < 2.2*10^−16 (t-test). Number of embryos used for quantification: 63 (ctrl, hey1), 58 (tra2bMO, hey1), 68 (ctrl, pcdh8), and 59 (tra2bMO, pcdh8). Embryos shown are X. laevis. See also Figure S1.

To understand the tra2b morphant phenotype in detail we analyzed the expression of developmentally regulated transcripts by in situ hybridization. Expression of the pan-neural marker sox2 confirmed that neural induction and neural plate morphology was largely normal (Figure 1I,J), and expression of the mesodermal marker t (bra) surrounding the blastopore and notochord during neurulation was normal (Figure 1K-L). In contrast, although myod was expressed correctly in the paraxial mesoderm, it failed to show proper segmentation (Figure 1M,N). The failure in axial segmentation was also revealed by pcdh8, which is expressed broadly in the pre-somitic mesoderm (PSM), as well as in stripes where the somites will form (Kim et al., 1998), and hey1, which marks somites after their formation (Pichon et al., 2002). The PSM, marked by pcdh8, was present, but did not show the normal pre-somitic stripes in tra2b morphants (Figure 1O,P). Furthermore, tra2b morphants failed to form mature somites as judged by hey1 expression (Figure 1Q,R,S,T). These results argue that defective somitogenesis constitutes an important part of the tra2b morphants phenotype.

RNA-seq identifies intron retention and exon skipping as the primary splice changes in tra2b morphants

We used RNA-seq to identify splice changes in tra2b morphants. Since X. laevis is pseudotetraploid, we used the closely related diploid species, X. tropicalis, to simplify our RNA-seq experiments (Hellsten et al., 2010). We employed Cufflinks (Trapnell et al., 2010) to assemble transcripts, using the current JGI transcript annotation (v7.2) as guide. Importantly, we assembled transcripts separately for the control and MO conditions to increase sensitivity in detecting novel isoforms and only used paired-end sequenced fragments where both reads aligned uniquely (Figure 2A, Table S1 and S2). We then merged those condition-specific transcript sets to a final transcriptome assembly containing annotated as well as novel transcripts. Finally we used DEXSeq (Anders et al., 2012) to detect differential exon expression based on the final annotation. This strategy is highly sensitive for detecting novel isoforms that are predominantly expressed in one set of samples and thereby significantly augments the existing annotation. We also found the count-based test for differential exon expression used by DEXSeq more reliable at predicting alternative splice changes than the isoform estimation performed by Cuffdiff2 (a part of the trinity package containing Tophat and Cufflinks). As a final validation we inspected the read distribution for each predicted splice change on a genome browser and discarded artifactual events (usually a result of incomplete gene models).

Figure 2. Analysis of alternative splicing in tra2b morphants shows intron retention as the most common splice change.

(A) Outline of RNA-seq analysis pipeline. Condition-specific transcript assemblies are merged with JGI annotation resulting in an augmented transcriptome assembly that forms the basis of DEXSeq testing of differential exon expression. (B) Table showing number of novel transcripts found in this study and number of significant splice changes in tra2b morphants. (C) RNA-seq reads from X. tropicalis control and morphants in tra2b locus show MO-induced intron retention (blue box) and an increase in variable exon 2 (black arrow). (D) DEXSeq output showing fitted splicing (a proxy for number of reads aligned) across all exonic regions. Control (black) and morphant (red) exon expression is similar except for the variable exon 2 and MO induced retained intron, where the graphs diverge. Both events are significant (purple exons indicate adjusted p < 0.05). (E) Alternative splicing changes in morphants grouped by category show intron retention (RI, red) as the most common event, followed by skipped or included exons (ESI, yellow). (F) Plot of individual alternative splicing events shows that retained introns are always included in morphants, whereas ESI in all but two instances are included in normal embryos. (G) Most alternative splicing events detected are novel and are not described in the annotation. See also Figure S2.

Using the Cufflinks/Cuffmerge protocol, we found 14,416 unannotated isoforms in 8,083 already annotated genes, which added increased depth to the genome annotation. The DEXSeq test for differential exon use and subsequent curation identified 142 events that were changed at least 1.5 fold in number of reads aligned to a region and were significant at adjusted p < 0.05 (Figure 2B). The splice changes occurred in 133 different genes. Importantly, the analysis verified the MO induced intron retention in tra2b transcripts, and discovered increased inclusion of a previously unannotated alternative exon 2 in morphants (Figure 2C,D; Figure S2A). This demonstrates our ability to confirm known splice changes in tra2b morphants as well as detect new ones.

A prominent pattern of changes was also shown by the analysis (Figure 2E): the vast majority was either whole intron retention (RI; 61 individual events or 43% of total) or whole exon skipping or inclusion (ESI; 48 or 34%) with the rest divided among alternative 5’ or 3’ splice site usage (A5SS; 6% and A3SS; 2%), combinations of alternative last exons or 3’UTRs (ALE/UTR; 6%) or complex/other (OTHER; 5%). Furthermore, we identified five events where an exon was skipped and the surrounding introns were retained (RI+SE; 4%).

Strikingly, all 61 Tra2b regulated intron retention events occurred in morphants and none were detected in controls (Figure 2F). Conversely, in the category of skipped or included exons, all but two events consisted of exons being skipped in tra2b morphants (one of the included exons was the variable exon 2 in tra2b transcript itself, as described above). Other categories showed more variation or their smaller sample sizes made it difficult to determine if they were preferentially used in morphants relative to controls (Figure 2F). Together, these results demonstrate that Tra2b is principally needed to remove a subset of whole introns and to retain a subset of exons.

Among the alternative splicing changes caused by Tra2b knockdown 126 of the 142 (89%) isoforms changed in tra2b morphants have not been previously described/annotated. Among these, 58 out of 61 retained intron and 39 out of 48 exon skipping/inclusion events were previously unannotated (Figure 2G).

Muscle related gene expression is reduced in tra2b morphants

In addition to changes in isoform use, we also tested for differential gene expression in tra2b morphants to determine if changes in splicing lead to secondary changes in gene expression. Two software packages, Cuffdiff2 (Trapnell et al., 2012) and DESeq2 (Love et al., 2014), together found 155 differentially expressed genes (p < 0.05) with at least 1.5 fold change in expression (Figure 3A and Table S3). In agreement with the observed defects in mesoderm differentiation we found several myogenic and muscle-related genes down-regulated in tra2b morphants (Figure 3B). This was confirmed by Gene Ontology (GO) term analysis of differentially expressed genes, where 77 out of 155 frog genes were successfully mapped with human GO terms. GO terms related to muscle function were significantly enriched in all categories (Figure 3C). Using RT-qPCR we confirmed significant inhibition of 10/11 muscle-related transcripts (Figure S3). In summary, our analysis of gene expression changes and observed somitogenesis defects support a function for Tra2b in mesoderm development.

Figure 3. Differential gene expression in tra2b morphants confirms reduction of muscle transcripts.

(A) Comparison of Cuffdiff2 and DESeq2 programs in calling significant changes in gene expression (X. tropicalis). (B) Heatmap showing muscle related genes that are repressed in tra2b morphants. (C) Table showing top five enriched GO terms in differentially expressed genes with muscle-related GO terms indicated in red. See also Figure S3.

Intron retention in wnt11b results in expression of a dominant-negative ligand

Given the prominence of intron retention in tra2b morphants, we determined whether the failure to form somites could be caused by intron retention in specific transcripts. Indeed, we identified a previously uncharacterized isoform of wnt11b that showed dramatic retention of the last intron (Figure 4A; here named wnt11b-in4ret , encoding Wnt11b-short protein). This splice change introduces a premature stop codon immediately within the retained intron, resulting in a truncated protein. Interestingly, this truncated isoform is similar to an engineered dominant-negative Wnt11b (Wnt11b-dn; Figure 4B) (Tada and Smith, 2000), raising the possibility that some of the defects observed in tra2b morphants are a result of inhibited Wnt signaling. In frogs and zebrafish, wnt11 orthologues are known to function during early development and gastrulation (Heisenberg et al., 2000; Kofron et al., 2007; Tada and Smith, 2000; Walentek et al., 2013). However, wnt11b is also expressed in the pre-somitic mesoderm and somites, consistent with a function during somitogenesis (Figure 4C). In contrast to wnt11b, we found no changes in wnt11 (wnt11-r) transcript level or splicing (data not shown).

Figure 4. tra2b knockdown reveals a novel inhibitory wnt11b isoform.

(A) RNA-seq read profile and Cufflinks assembled transcripts on wnt11b locus show retention of intron 4. Top panel shows JGI gene model for wnt11b; middle and bottom panels show read profiles and Cufflinks assembled transcripts from control and morphants. (B) Intron 4 retention results in a truncated protein (red, Wnt11b-short), resembling a dominant-negative (yellow) that lacks 57 C-terminal residues compared to normal (black). (C) wnt11b is expressed in the pre-somitic mesoderm/circumblastoporal region and somites. Embryos are shown in dorsal view with anterior to the left. (D) wnt11b-short mimics wnt11b-dn in pronephric tubule inhibition assay. Embryos were injected unilaterally in the prospective lateral mesoderm and examined by ISH for atp1a1, which marks the developing pronephros. Arrows point to the proximal pronephric tubules on the injected side. (E) Summary of pronephros tubule inhibition assay. Number of embryos scored: 144 (control), 72 (tra2bMO), 87 (wnt11b-dn), and 69 (wnt11b-short). (F) RT-qPCR for induced cardiac gene expression (gata4) or axial mesoderm (t/bra) on animal caps injected with combinations of activin, wnt11b, wnt11b-dn and wnt11b-short, showing that wnt11b-short acts similar to wnt11b-dn and counters the effect of wnt11b. Bar plots show mean of three independent experiments + SEM of normalized fold-induction compared to activin injected embryos. (A) Shows data from X. tropicalis, (C-F) show data from X. laevis. See also Figure S4.

To test if Wnt11b-short can act as a dominant-negative ligand we tested its ability to inhibit pronephric tubule formation like Wnt11b-dn (Tételin and Jones, 2010). First, we injected embryos with a low dose of tra2b MO targeted to the lateral/intermediate mesoderm, which allowed development to stage 40 when nephrogenesis can be investigated. These morphants had severe defects in forming the proximal pronephric tubules, consistent with tra2b MO affecting nephrogenesis by inducing expression of an inhibitory Wnt11b (Figure 4D). Crucially, injection of a synthetic mRNA (wnt11b-short) that encodes Wnt11b-short, in the lateral/intermediate mesoderm also repressed proximal tubule formation, at least to the same degree as wnt11b-dn (Figure 4E). This in vivo assay suggests that wnt11b-in4ret encodes a ligand functionally equivalent to Wnt11b-dn.

We also tested the ability of wnt11b-short to block cardiac induction in an animal cap assay. Animal caps injected with activin mRNA become mesoderm as marked by expression of t/bra, but when stimulated with both activin and wnt11b they express the cardiac mesoderm gene gata4 or gata6 and reduce t/bra expression (Afouda et al., 2008; Pandur et al., 2002)(Figure 4F, Figure S4A). Neither wnt11b-dn nor wnt11b-short coinjected with activin induced gata4 or reduced t/bra gene expression, but importantly, co-injection of wnt11b-short prevented induction of gata4 or gata6 by activin and wnt11b. The endodermal marker sox17 was, as expected, induced in all activin injected samples but not in those injected with wnt11b, wnt11b-dn, or wnt11b-short alone (Figure S4A). Together these animal cap and pronephric tubule experiments strongly argue that wnt11b-in4ret encodes a dominant-negative Wnt11b ligand.

While analyzing wnt11b, we discovered that the X. tropicalis genome contains an unannotated wnt11b gene duplication (Figure S4B). This gene (xetro.H00536) is adjacent to the annotated wnt11b, but is transcribed in the opposite direction. Interestingly, xetro.H00536 also displays retention of the last intron in tra2b morphants, suggesting that the target sequences are conserved in the duplicate.

Retention of wnt11b intron4 recapitulates the failure to form somites in tra2b morphants

To test if Wnt11b-short contributes directly to the phenotype of tra2b morphants we specifically induced splicing of wnt11b-in4ret using a MO targeting the last exon-intron junction in wnt11b in X. laevis (Figure 5A). Indeed, RT-PCR on single embryos injected with increasing doses of the wnt11b-in4MO showed that the MO was at least as effective at inducing wnt11b intron4 retention as tra2MO1 (Figure 5B). Sequencing the amplicons verified that they originated from retention of intron4.

Figure 5. Intron retention in wnt11b is responsible for somite defects in tra2b morphants.

(A) Diagram showing X. laevis wnt11b gene structure and wnt11b-in4 MO (red). (B) RT-PCR on stage 19 single embryos injected with wnt11b-in4 MO or tra2b MO shows efficient retention of intron 4. Top part shows RT-PCR with primers in wnt11b exon 4 and intron 4, bottom panel shows internal control odc. (C) in situ hybridization for mesodermal gene expression in wnt11b-in4 MO and tra2b MO injected embryos. Black arrows in pcdh8 stained embryos indicate presence of somitic stripes in control and wnt11b-in4 MO injected embryos. Red arrows in hey1 stained embryos indicate mature somites in control embryos, which are absent in wnt11b-in4 and tra2b morphants. Asterisk (*) in hey1 samples indicates non-somitic midline hey1 expression, which is exposed because of neural tube closure defect. All pictures show dorsal views with anterior up. (D-E) Quantification of ISH results showing mean and SD of number of hey1+ somites (D) and pcdh8+ stripes (E). *** Indicate difference is statistically significant from control at p < 2.2 * 10^−15 (t-test). Number of embryos used for quantification: 41 (ctrl, hey1), 30 (tra2bMO, hey1), 43 (wnt11bMO, hey1), 42 (ctrl, pcdh8), 32 (tra2bMO, pcdh8), and 43 (wnt11bMO, pcdh8). Data shown are from X. laevis.

Next, we analyzed mesoderm and somite development in wnt11b-in4 morphants and compared them to tra2b morphants (Figure 5C). The resulting embryos were short, but did not show some of the other defects of the tra2b knockdown, such as endodermal loss. At the molecular level, both control embryos and wnt11b-in4 morphants had similar axial mesoderm formation at stage 20 as judged by myod expression, whereas in tra2b morphants myod was severely reduced, as observed earlier. In addition to the near-normal myod expression, wnt11b-in4 morphants showed segmentation stripes of pcdh8, though these extended less than in control embryos. In contrast and as observed earlier, tra2b morphants had reduced pcdh8 expression and little sign of segmentation. Despite the milder effects on early mesoderm development in wnt11b-in4 morphants, these embryos failed to form mature somites, as judged by hey1 expression. Quantification confirmed that wnt11b-in4 morphants were similar to control embryos in number of pcdh8 expressing stripes whereas tra2b morphants were reduced to 26% of normal (Figure 5D). Likewise, quantification of hey1+ somites in the three classes similarly confirmed that they were reduced to 5% of normal in tra2b morphants and 27% in wnt11b-in4 morphants (Figure 4E). These results demonstrate that wnt11b-in4 morphants have a more normal mesoderm development prior to somitogenesis than tra2b morphants, yet fail to form hey1-positive somites. In frogs, as in other vertebrates, there are two transformer-2 genes (tra2a and tra2b) raising the possibility of Tra2b regulating wnt11b splicing through changes in tra2a. However, we did not detect any significant changes in splicing or expression levels of tra2a in our RNA-seq data, (data not shown), indicating that Tra2b regulates splicing of wnt11b intron 4 even when tra2a expression is normal.

In aggregate, these experiments demonstrate that retention of intron4 in wnt11b is responsible for aspects of defective segmentation in tra2b morphants. However, wnt11b-in4 morphants did not display many of the other defects of tra2b morphants – notably the endoderm dissociation defects prior to hatching. This is unlikely to be a result of low efficacy of the wnt11b-in4 MO since intron4 inclusion was more efficient in wnt11b-in4 injected embryos than in tra2b morphants (Figure 5B). This suggests that other splice changes underlie endoderm dissociation and other defects in tra2b morphants.

DISCUSSION

This is the first systematic analysis of splice changes regulated by Tra2b in a vertebrate. While mouse tra2b mutants have been reported and shown to die during early gestation, the causes and the underlying splice changes are unknown (Mende et al., 2010). However, the early embryonic death of tra2b mouse mutants is consistent with the severe developmental defects that we observe in Xenopus tra2b morphants, so Tra2b may fulfill a conserved regulatory role.

Intron retention has often been dismissed as a result of erroneous splicing, but recent discoveries have shown that intron retention has important regulatory functions during granulocyte and nervous system development (Colak et al., 2013; Wong et al., 2013). In those cases, intron retention was tied to transcript destruction through nonsense mediated decay. However, retention of the final intron, as here with wnt11b, would not activate such decay, and in such cases truncated protein isoforms may act as inhibitors. Thus, the proper removal of the final wnt11b intron mediated by Tra2b provides another example where splicing activities are required to regulate differentiation. Tra2b levels appear to be tightly regulated as evidenced by the metabolic defects in mouse tra2b heterozygotes, and different splicing events may differ widely in their requirement for Tra2b.

We have employed a modified version of the standard RNA-seq analysis and by keeping control and tra2b morphant datasets separate we increase the sensitivity for isoforms that would otherwise fall below the level of detection. In fact, we significantly augment the existing annotation with more than 14,000 novel isoforms in more than 8,000 genes, even though Tra2b regulates only a small subset of these (Figure 2B). This, combined with our use of DEXSeq to test for differential expression of exons, rather than whole isoforms, has enabled us to discover novel splice variants and test them for differential expression with high sensitivity. One surprising observation in this study is that a large majority of the isoforms that we found differentially expressed in tra2b morphants had not previously been described. These observations emphasize the strict requirement for Tra2b in normal development and for splicing of a highly restricted set of isoforms.

The structure of wnt11b-in4ret makes it indistinguishable from a partially spliced transcript but its presence in our RNA-seq data (which is poly-A selected) and inhibition of somitogenesis in wnt11b-in4 morphants strongly argue that it encodes a functional protein. Detection of endogenous Wnt proteins is notoriously difficult, and we have not successfully detected either Wnt11b isoform. Nonetheless, the inferred presence of an inhibitory Wnt11b ligand adds an additional layer of regulation to the complex Wnt pathway. In addition to having a function during the normal development or later tissue-homeostasis, Wnt11b-short dysregulation may also have clinical impact since regulation of alternative splicing is often changed in cancer (David and Manley, 2010).

Wnt signaling is known to be necessary for somitogenesis. In other vertebrates, Wnt11 been shown to influence later aspects of somite differentiation and be involved in epithelial transition of the dermomyotome (Geetha-Loganathan, 2006; Morosan-Puopolo et al., 2014). Our experiments with induced expression of Wnt11b-short in the PSM and somites, using wnt11b-in4 MO, strongly suggest a function for Wnt11b in early phases of somite formation.

Alternative splicing expands proteomic complexity by allowing multiple proteins to be generated from a single locus. More than 90% of all human multiexon genes give rise to more than one isoform (Wang et al., 2008). However, most isoforms have no assigned - or even hypothesized - function, partly because of the lack of a suitable system in which to study alternative splicing in a native organismal context. Here we exploited the pairing of embryology with global transcriptome analysis for studying alternative splicing in Xenopus. We specifically connected the splice change in wnt11b to defects in somitogenesis, whereas other defects in tra2b morphants are unrelated to this splice change. Likely, some of the 141 other splice changes we detected underlie these aspects of the tra2b phenotype.

EXPERIMENTAL PROCEDURES

Microinjection of Xenopus embryos

Morpholino-oligonucleotides (GeneTools) used were as follows: tra2b-MO1 (X. laevis; translation blocking) 5’-CTCCGCTATCACTCATCTTGTCGTC-3’; tra2b-MO2 (X. tropicalis; exon3-intron3 junction) 5’-AAGTTGCATACCCTGTTTCCAACAT-3’; wnt11b-in4-MO (X. laevis; exon4-intron4 junction) 5’-GACACAGGACAGGTAAGCTTATCCT-3’, and standard fluorescein-labelled control MO as tracer. Capped mRNA for microinjection was prepared using mMessage mMachine Kit (Ambion).

For pronephric tubule formation assays, whole embryos were injected laterally in the two right blastomeres at the four-cell stage (to unilaterally target intermediate mesoderm). The total dose was 3 ng of wnt11b-dn or wnt11b-short, or 41 ng tra2b-MO1. Fluorescent Rhodamine-B Dextran lineage tracer (0.5–1.0 mg/ml; Molecular Probes) was used to confirm targeting prior to fixation. Analysis of pronephric tubule formation was performed by comparing injected versus uninjected sides of control and manipulated embryos as previously described (Walentek et al., 2012).

The plasmid for synthesizing wnt11b-short was generated by PCR cloning from Wnt11b-pCS2+ plasmid (Tada and Smith, 2000) as template, using the following PCR forward and reverse primers: 5'-AAAAAAATCGATATGGCTCCGACCCGTCAC-3', 5'-AAAAAAGAATTCTTACCTGTCCTGTGTCCCATATG-3'. The PCR product was cloned into pCS108 using ClaI and EcoRI. A pCS2+ plasmid encoding Wnt11b-dn (Tada and Smith, 2000)), was used for wnt11b-dn synthesis.

RNA isolation and RT-PCR/qPCR

RNA was isolated from single embryos using Trizol (Invitrogen) followed by isopropanol precipitation, one or two phenol-chloroform extractions, and a final ethanol precipitation. Standard RT-PCR: random or oligo-dT primed cDNA was synthesized using SuperScriptII (Invitrogen), and PCR used PlatinumTaq (Invitrogen). RT-qPCR on animal caps: embryos were injected in all blastomeres at the four-cell stage near the animal pole, and animal caps were prepared from 15 (+/−3) embryos per sample (Sive et al., 2000). The following total doses (4 × 10nl) of mRNA were used: 1 ng of Wnt11b, Wnt11b-dn (both (Tada and Smith, 2000)), or Wnt11b-short (this study); 0.4 pg of activin (Thomsen et al., 1990). Following RNA isolation, cDNA was synthesized using iScript (Bio-Rad), and qPCR used SsoAdvanced SYBR Green reagents (Bio-Rad) in technical triplicate on a Bio-Rad CFX96 RT-System C1000 Touch Thermo-cycler. Expression levels were normalized to housekeeping genes ef1-α and odc. Expression was calculated relative to uninjected controls and normalized to the level of activin induction in the specific experiment. Triplicate biological replicates were performed. Oligos used are in listed in Table ST4. RT-PCR to confirm alternative splice changes in X. tropicalis were performed on oligo-dT primed cDNA (for ESI events) or as described above for RT-qPCR (for RI events).

RNA-seq library construction and sequencing

Multiplexed Illumina libraries were synthesized using TruSeq RNA Sample Preparation Kit v2 (Illumina) from X. tropicalis stage 14 single embryo RNA isolated using Trizol (Invitrogen). Libraries were prepared from control or tra2b-MO2 injected in triplicate. Samples were sequenced at the Vincent J. Coates Genomics Sequencing Laboratory at the University of California at Berkeley on Illumina HiSeq2000 machines. Each library was paired-end sequenced on two independent flow-cells resulting in 2×75 and 2×60 bp reads after quality trimming of 3’ends. Summary of sequence yield and alignment is listed in Table S2.

Xenopus embryos and microinjection

X. laevis embryos were obtained, cultured, and microinjected by standard methods (Sive et al., 2000). X. tropicalis embryos were collected from natural matings and injected in both blastomeres at the two-cell stage (Khokha et al., 2002). Embryos were staged according to the standard table (Nieuwkoop and Faber, 1994). All experiments were performed at least in triplicate. Statistical testing of reduction in somites in morphants was performed in R (R Core Team, 2014) using the t.test function.

Bioinformatics

Gene names for the JGI v7.2 annotation were obtained from the names of the previous v7.1 annotation (Dichmann and Harland, 2012). Paired-end RNA-seq reads from either control or tra2b morphants that passed basic quality filters were aligned using Tophat2.0.9 (Trapnell et al., 2009) to X. tropicalis genome v7.1 with JGI v7.2 annotation as guide allowing novel splice junctions to be discovered. Only sequenced fragments where both read-mates aligned uniquely were used to assemble transcripts using Cufflinks2.1.1 (Trapnell et al., 2012) for either control or morphants. The resulting two condition-specific transcriptome assemblies were merged with the JGI annotation to produce a merged annotation using Cuffmerge2.1.1. The merged annotation was filtered so that only stranded and named transcripts with class codes “j” or “=“ were retained resulting in the final annotation. This final annotation was used as basis for query for differential exon expression using DEXSeq v1.8 (Anders et al., 2012). DEXSeq output was filtered for changes of less than 50% and erroneous gene models based on inspection on genome browsers GBrowse2 (www.gmod.org) and IGV (Robinson et al., 2011). Visualization of read profiles and transcripts for figures were captured from GBrowse2. Number of novel isoforms was determined from the final annotation based on transcripts flagged by Cufflinks with class_code “j” (indicating a novel, spliced transcript sharing at least one junction with an annotated transcript). Human GO terms were mapped to X. tropicalis genes and hypergeometric test for enriched terms performed using GOstats bioconductor package (Falcon and Gentleman, 2007). The aligned reads have been submitted to the NCBI Short Read Archive under BioProjectID PRJNA266550. The final annotation of all transcript isoforms can be accessed from Xenbase.org (James-Zorn et al., 2013) under user-submitted data.

Whole Mount RNA in situ hybridization probes

The following plasmids for synthesizing anti-sense probes for in situ hybridization (Sive et al., 2000) have been described previously: atp1a1 (Tran et al., 2007), bra(t) (Smith et al., 1991), myod (N D Hopwood, 1989), pcdh8 (Kim et al., 1998), sox2 (Grammer et al., 2000), and wnt11b (Tada and Smith, 2000). A bluescript plasmid containing a 900 bp fragment of the CDS was used to synthesize hey1 probe (Pichon et al., 2002).