• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptNIH Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
RNA Biol. Author manuscript; available in PMC Feb 17, 2010.
Published in final edited form as:
Published online Nov 22, 2009.
PMCID: PMC2823379

The epithelial splicing factors ESRP1 and ESRP2 positively and negatively regulate diverse types of alternative splicing events


Cell-type and tissue-specific alternative splicing events are regulated by combinatorial control involving both abundant RNA binding proteins as well as those with more discrete expression and specialized functions. Epithelial Splicing Regulatory Proteins 1 and 2 (ESRP1 and ESRP2) are recently discovered epithelial-specific RNA binding proteins that promote splicing of the epithelial variant of the FGFR2, ENAH, CD44 and CTNND1 transcripts. To catalogue a larger set of splicing events under the regulation of the ESRPs we profiled splicing changes induced by RNA interference-mediated knockdown of ESRP1 and ESRP2 expression in a human epithelial cell line using the splicing sensitive Affymetrix Exon ST1.0 Arrays. Analysis of the microarray data resulted in the identification of over a hundred candidate ESRP regulated splicing events. We were able to independently validate 38 of these targets by RT-PCR. The ESRP regulated events encompass all known types of alternative splicing events, most prominent being alternative cassette exons and splicing events leading to alternative 3' terminal exons. Importantly, a number of these regulated splicing events occur in gene transcripts that encode proteins with well-described roles in the regulation of actin cytoskeleton organization, cell-cell adhesion, cell polarity and cell migration. In sum, this work reveals a novel list of transcripts differentially spliced in epithelial and mesenchymal cells, implying that coordinated alternative splicing plays a critical role in determination of cell type identity. These results further establish ESRP1 and ESRP2 as global regulators of an epithelial splicing regulatory network.

Keywords: alternative splicing, exons, ESRP, splicing regulatory network, epithelial to mesenchymal transition, exon arrays


Alternative splicing is a highly utilized process in higher eukaryotic cells whereby multiple functionally distinct mRNAs and proteins can be generated from a single gene, greatly enhancing the coding potential of a genome.1 Recent deep sequencing of mRNAs across several human tissues has revealed that up to 94% of human transcripts undergo alternative splicing, indicating that splicing regulation is a vital and widespread means of regulating gene expression.2,3 Coordinated patterns of alternative splicing involving numerous gene transcripts generate a complex protein expression pattern that in turn contributes to cellular identity.4 Alternative splicing is under the control of RNA binding proteins that bind cis-regulatory elements surrounding the splice sites resulting in enhanced or suppressed utilization of the regulated splice site.5,6 These regulators include the largely ubiquitous serine-argenine (SR) rich family and hnRNP proteins as well as proteins with more limited expression such as the Fox, Nova and CELF families of splicing factors.7,8

We recently identified epithelial cell type-specific splicing regulators ESRP1 and ESRP2 using a high-throughput cDNA expression screen.9 These two proteins are highly conserved paralogs containing three RNA Recognition Motif (RRM) domains. Strict, cell-type specific expression was demonstrated by in situ hybridization in mouse tissue sections. Induction of an epithelial-to-mesenchymal transition (EMT) in a mammary epithelial cell line resulted in downregulation of both ESRP1 and ESRP2. As a consequence, a switch occurred from the epithelial to the mesenchymal splicing pattern in the FGFR2, CD44, CTNND1 (p120-Catenin) and ENAH transcripts. Notably, the same switch in splicing of these four transcripts was observed during siRNA mediated downregulation of the ESRPs in epithelial cells.9 In addition, ectopic expression of a cDNA encoding mouse Esrp1 (mEsrp1) in a mesenchymal cell line induced the opposite changes in splicing in all four of these transcripts. Because the proteins encoded by these gene transcripts have well documented rolls in the EMT, we sought to further expand the list of alternative splicing events regulated by the ESRPs. One guiding hypothesis is that additional gene transcripts that are regulated by these proteins may similarly play important roles in epithelial to mesenchymal transitions during development as well as pathophysiologic conditions such as cancer metastasis and tissue fibrosis.

A number of technological advances have led to the development of tools that can be used to facilitate large scale profiling of alternative splicing, including several splicing sensitive microarray platforms.10 One such example of this technology is the Human Exon ST1.0 Array from Affymetrix, a high density expression array designed to tile probes over all known and predicted exons in the human genome.11,12 This platform has previously been used to assay alternative splicing between different tissues and in normal vs. cancer cells.1315 This array has also been used in more directed approaches to catalogue alternative splicing events under the control of specific splicing factors including PTB, hnRNPL and hnRNPL-like.1619 While the design of this platform facilitates an unbiased, near genome-wide search for regulated splicing events, the large number of probes necessary to achieve this goal introduces a significant source noise in the form of background and cross-hybridization. Since there are no more than four probes per exon in the current design, this noise and the differential hybridization efficiencies can make the estimated exon expression levels unreliable. A recently described tool, Microarray Analysis of Differential Splicing (MADS), was developed to overcome the challenges in analyzing the data extracted from the Exon array. The MADS tool was designed to perform background correction and to detect and remove probes with sequence-specific cross-hybridization to off-target transcripts in order to predict differentially spliced exons with a low false positive rate.18

We used siRNAs to deplete ESRP1 and ESRP2 in a human epithelial cell line and utilized the Human Exon Array 1.0 and the MADS tool to identify a novel set of alternative splicing events regulated by these splicing factors. The efficacy of this experimental system to yield significant splicing changes was previously shown, including a near complete switch from use of the epithelial to the mesenchymal cassette exon in FGFR2.9 This analysis yielded a broad list of novel candidate ESRP targets encompassing all known types of alternative splicing events. Many of the alternative splicing events identified in this study are in genes that encode proteins with well-described roles in processes such as actin cytoskeleton organization, cell adhesion and cell motility. The majority of the candidates that were subjected to independent validation were demonstrated to be regulated by the ESRPs, suggesting that most of these candidates contain splice variants that are indeed components of an epithelial splicing signature. In sum, this study greatly expands the list of alternative splicing events regulated by ESRP1 and ESRP2 and suggests that the ESRPs orchestrate an epithelial splicing regulatory program.


An approach using siRNA-mediated depletion and exon microarrays to identify splicing events regulated by ESRP1 and ESRP2

We previously demonstrated that depletion of ESRP1 and ESRP2 from the human prostatic epithelial cell line PNT2 is sufficient to robustly alter the epithelial splicing pattern of four gene transcripts towards the mesenchymal splicing pattern. We therefore made use of this experimental system to carry out a near genome-wide based approach to identify a broader set of splicing events regulated by the ESRPs. Combined depletion of ESRP1 and ESRP2 in PNT2 cells was carried out with siRNAs against each transcript or with siRNAs against green fluorescent protein (GFP) as a control. Each condition was performed in four biological replicates and total RNA, each processed and prepared separately, was converted into labeled cDNA for hybridization to the Affymetrix Human Exon Array 1.0 ST Microarray (Fig. 1A). The efficacy of siRNA-mediated knockdown of ESRP1 and ESRP2 was demonstrated by real-time RT-PCR (Fig. 1B). Because there are presently no suitable antibodies to specifically detect endogenous ESRP protein, we were unable to verify protein depletion by immunoblotting. However, we were able to demonstrate that ESRP1 and ESRP2 were functionally depleted by showing a nearly complete splicing switch from use of the mutually exclusive epithelial IIIb exon, to the mesenchymal IIIc exon in the FGFR2 transcript in all four replicates of ESRP knockdown compared to the control knockdowns (Fig. 1C). Furthermore, we previously validated the specificity of this approach as being due to protein depletion by showing that the epithelial splicing pattern could be rescued by expression of a mouse Esrp1 protein from an RNAi resistant cDNA.9

Figure 1
Outline and validation of an siRNA and exon based microarray approach to identify alternative splicing events regulated by ESRP1 and ESRP2. (A) A flowchart illustrating the experimental design. (B) Quantitative RT-PCR showing a greater than 80% decrease ...

Exon array analysis identifies many novel splicing targets of ESRP1 and ESRP2

In order to identify novel alternative splicing events regulated by the ESRPs, we used the MADS tool to identify probesets that were differentially expressed between the two conditions, subtracting out probeset differences that were the result of differences in overall transcript abundance. The resulting probesets were ranked according to their MADS p-value and the top 500 differentially expressed probesets were selected for further analysis and validations. We first manually examined all probesets using the UCSC genome browser and discarded any probesets that did not correspond to annotated alternative splicing events as supported by mRNA or EST evidence. We also discarded examples of alternative promoters, including cases in which the exon array data could not distinguish between alternative promoters and true alternative splicing events. For the remaining probesets we carefully analyzed graphical output from the MADS algorithm as well as the Gene View of all filtered probesets in Partek Genomics Suite. Probesets for which neither analysis of the graphical data output supported an interpretation of a true splicing change were also discarded. Most of the discarded probesets were present in intronic regions, the majority of which lacked evidence supporting the existence of an exon. By discarding a number of probesets without well-annotated alternative splicing it is thus clearly possible that some of the discarded probesets may nonetheless represent novel alternatively spliced exons that are missed by using this filter. Our evaluation also included a collective interpretation of differential probeset signal, determination as to whether probeset differences were more consistent with overall transcript differences, and a general analysis of the overall pattern of probeset signals. Such an assessment thus also included, in some cases, an assessment as to whether reciprocal changes in several transcripts’ probesets also supported differential splicing. For example, several cases of mutually exclusive exons or alternative 3' ends in which opposite changes in probeset signal were supportive of alternative splicing events were included in the final set of high confidence alternative splicing events. While some of this assessment necessarily was partly subjective, we sought to conservatively include only examples in which we had high confidence that the alternative splicing events were indeed regulated by the ESRPs in this experiment. Although some false positives may be present in this final list, the validation rate from the limited number of validated events strongly suggests that most are true positives.

After manual annotation of the top 500 MADS probesets and deletion of poorly supported alternative events, we narrowed the list to 148 alternative splicing events from 171 probesets, in a total of 134 different genes (Suppl. Material, Table S1). Satisfyingly, this list included detection of ESRP-regulated alternative splicing events that we previously demonstrated in the CD44, CTNND1 and ENAH transcripts.9 In the case of CD44, probesets corresponding to tandem variable exons 2, 4, 9, 5 and 7 ranked 21st, 26th, 100th, 101st and 122nd, respectively. For CTNND1 (p120-catenin), a probeset for alternative exon 2 ranked 139th. The ENAH alternative cassette exon 11a ranked 206th. The probesets for these transcripts are shown schematically using software from the Partek Genomics Suite (Fig. 1D). Probesets corresponding to exons IIIb and IIIc of FGFR2 were filtered from the analysis because the overall transcript level of FGFR2 did not pass the expression threshold, and hence, did not appear in the list of top 500 probesets. Nonetheless, the p-values associated with both of these mutually exclusive exons (IIIb: 9.4 × 10−5 and 1.6 × 10−3; IIIc: 2.9 × 10−4 and 7.4 × 10−2) likewise strongly supported differential splicing of these exons as shown by RT-PCR.

We categorized the different types of alternative splicing events and found that there is at least one example of every known type of alternative splicing event in the resulting list of high-confidence ESRP-regulated splicing events (Fig. 2). Notably, the computational Alternative Conserved Exon (ACEScan) tool predicts 25 of the 143 events were ACEScan positive (including 21 out of the 68 cassette exons), indicating that they were highly likely (or known) to be evolutionarily conserved.20 The ACEScan data set was derived based on intronic sequence conservation between human and mouse genes that is indicative of conserved splicing events in both species and likely also reflects the presence of intronic binding sites for splicing regulatory factors. The enrichment of such exons indicates that alternative splicing of these exons is likely to be physiologically relevant. The largest group of splicing events consists of alternative cassette exons. However, we also noted a large class of alternative splicing or polyadenylation events that result in the use of alternative 3' ends. These events can be sub-divided into alternative polyadenylation (polyA) sites or alternative 3' terminal exons. We further categorized the alternative 3' terminal exons into two types. We defined Type I events as those where an alternative 5' splice site is in competition with a polyA signal in the immediate downstream intron. Thus, in these events, when the 5' splice site is not used, the associated exon becomes the 3' terminal exon, whereas its use leads to use of an exon further downstream as the 3' terminal exon, often also including numerous additional downstream cassette exons. Type II events are defined as those in which competing 3' splice sites in different exons lead either to an upstream 3' terminal exons, or selection of a downstream exon as either the 3' terminal exon or continued use of additional 3' cassette exons (see schematic in Fig. 2).

Figure 2
Numerous novel alternative splicing targets representing all known types of alternative splicing are identified among a high confidence set of genes identified using MADS analysis of the Exon Array data. On the left are schematics depicting each type ...

The ESRPs regulate inclusion and skipping of internal cassette exons

The ESRPs were predicted by the Exon Array data to both enhance and repress splicing of alternative cassette exons and an example of each is illustrated in Figure 3A and 3B. A single probeset in the SLK transcript corresponding to exon 13 showed decreased expression in the ESRP knockdown (Fig. 3A). This suggests that inclusion of SLK exon 13 is enhanced by the ESRPs. RT-PCR confirmed this prediction, showing that exon 13 inclusion decreases from 30% in the control to 2% when ESRP1 and 2 are downregulated (Fig. 3A). We further validated examples of ESRP enhanced cassette exons in the LOXL2 and TRIP10 transcripts. Inclusion of the LOXL2 and TRIP10 regulated exons decreased from 4.9% to 1.3% and 2.6% to 0.2%, respectively, upon ESRP knockdown (Fig. S1). An example of a cassette exon predicted to be repressed by the ESRPs is exon 16 in the SCRIB transcript. The single probeset associated with this exon showed increased levels of inclusion in the ESRP knockdown compared to the control (Fig. 3B). RT-PCR shows the inclusion level of the exon increasing from 11% to 77% upon ESRP knockdown, validating this exon as being a target of ESRP repression (Fig. 3B).

Figure 3
Validation of alternative cassette exons and alternative splice site selection events. Shown at the top of each panel are core and extended probesets from the exon arrays corresponding to the ESRP and control knockdown samples as shown in the Gene View ...

The ESRPs regulate the utilization of alternative 3' and 5' splice sites within an exon

In addition to exon inclusion or skipping, our analysis also revealed that the ESRPs can regulate alternative 5' or alternative 3' splice site selection. Figure 3C and 3D illustrates a specific example for each case. CHRNA5 exon five has three alternative 5' splice sites and the use of these splice sites results in exons of 49 nt, 350 nt and 836 nt. Four probesets corresponding to the portion of the exon created by utilization of the most downstream 5' splice site, three of which were in the top 500 MADS probesets, show increased expression upon ESRP knockdown (Fig. 3C). This data indicates that the ESRPs repress splicing at the most downstream 5' splice site in CHRNA5 exon five. RT-PCR confirmed this prediction as the splice variant of CHRNA5 with the longest form of exon five increases from 5% to 18% upon ESRP knockdown. Utilization of both upstream 5' splice sites decreased when the ESRPs were downregulated. The amount of the shortest form and intermediate form of exon five decreased from 19% to 10% and 76% to 72%, respectively (Fig. 3C).

As an example of an alternative 3' splice site choice regulated by the ESRPs, we validated the predicted change in 3' splice site utilization in FAM62A exon 14. A single probeset that corresponds to the portion of exon 14 formed by splicing at the upstream 3' splice site showed decreased expression upon ESRP knockdown relative to the control suggesting that ESRP promotes splicing of this exon at the upstream 3' splice site. RT-PCR analysis clearly confirmed this prediction, with the level of transcripts including the long form of exon 14 decreasing from 23% to 5% (Fig. 3D). We also observed that in four of the five examples of alternative 5' splice site selection, the ESRPs are predicted to favor splicing at the upstream 5' splice site. Similarly, in four of the five examples of alternative 3' splice site selection, the ESRPs are predicted to favor splicing at the upstream 3' splice site. Whether these examples indicate that the ESRPs more generally follow this pattern in the regulation of alternative 3' and 5' splice sites will require further study. Should this prove to be the case, however, such a preferential function may provide clues as to how the ESRPs regulate splicing at the molecular level.

The ESRPs regulate splicing of mutually exclusive cassette exons

Within the high-confidence list of ESRP regulated splicing events was one example of a mutually exclusive splicing event. A single probeset from the MADS data predicted increased expression of an exon directly upstream of exon 4 in the OGDH transcript. OGDH mRNA and EST alignments strongly suggest that the MADS exon, which we termed exon 4a, and exon 4 are spliced in a mutually exclusive manner. Despite the probeset for exon 4 being absent from the MADS top 500 list we performed an RT-PCR assay to validate the predicted increase in exon 4a inclusion and the reciprocal decrease in exon 4 inclusion. To distinguish between the two amplicons of similar size, we digested the PCR products with Pvu II and BstE II restriction enzymes that uniquely cut exon 4 and exon 4a, respectively. The RT-PCR along with a bar graph illustrating the splicing quantification is shown in Figure 4A. Splicing of exon 4a is increased in the ESRP knockdown, from 19% to 48% inclusion, as predicted by MADS. Conversely, exon 4 is included in 81% of the OGDH transcripts in the control but decreases to 52% inclusion when the ESRPs are downregulated. Furthermore, there was no indication of double-inclusion or double-skipping of the two exons. This result validates the MADS prediction as well as our interpretation that the alternative splicing event detected by the MADS is of the mutually exclusive type. Interestingly, this is not analogous to the role the ESRPs play in the regulation of FGFR2 mutually exclusive splicing where the ESRPs favor splicing of the upstream mutually exclusive exon and silencing of the downstream exon. In the case of OGDH, the ESRPs appear to promote splicing of the downstream exon over that of the upstream exon (but see also below).

Figure 4
Validation of mutually exclusive exons and pairs of tandem cassette exons. (A) Mutually exclusive splicing of exon 4a and exon 4 of the OGDH gene. MADS analysis predicted increased splicing of exon 4a (probeset 2999969) after ESRP depletion. The probeset ...

The ESRPs co-regulate splicing of tandem sets of exons

We previously demonstrated that the ESRPs regulated multiple, tandem exons within the CD44 and CTNND1 gene transcripts. The MADS analysis revealed additional tandem alternative exons regulated by the ESRPs and two examples are illustrated here. Two probesets in MYO1B corresponding to exons 24 and 25, showed increased inclusion in the ESRP knockdown, indicating that these exons are normally silenced by the ESRPs (Fig. 4B). RT-PCR confirmed that both of these exons are targets of ESRP repression. In the control, products containing exon 24 and/or exon 25 make up 14% of the MYO1B transcripts compared to 39% in the ESRP knockdown sample (Fig. 4B). We verified that the single inclusion product is a mix of products including either the upstream or the downstream exon alone (data not shown).

The second example of a pair of tandem exons regulated by the ESRPs is in the ADAM15 transcript where the probesets for exons 20 and 21 were shown to decrease in the knockdown sample compared to the control as detected in our MADS analysis (Fig. 4C). Unlike the case of the MYO1B exons, the two exons in ADAM15 were predicted to be enhanced by the ESRPs. RT-PCR analysis revealed three major splice isoforms corresponding to skipping of both exons, single inclusion of one or the other exon, and inclusion of both exons (Fig. 4C). As predicted by the array data, the splice variants containing one or both of these exons decrease in the ESRP knockdown while those that skip both exons increase in abundance. This result confirms that splicing of exons 20 and 21 in ADAM15 are enhanced by the ESRPs.

The ESRPs regulate splicing of alternative terminal exons

An unexpectedly prevalent regulatory function of the ESRPs unveiled by the MADS analysis was alternative splicing of 3' terminal ends or alternative polyadenylation (see description above and Fig. 2). A recent study showed that these events are widespread, occurring in up to 20% of human genes.21 Many of these types of events dramatically change the size and sequence of the resulting proteins more than is typically achieved through alternative splicing of cassette exons. Thus, it is perhaps more likely that some of these events may more significantly affect the physiological functions of the alternative isoforms. A total of 36 of these events were in our list of high-confidence alternative splicing targets. Identification of these events was made possible not only by the exon probesets identified by MADS, but by using the complete set of probesets for each transcript to detect reciprocal changes in expression between the upstream and downstream terminal exons as analyzed in the Geneview feature of the Partek software program. We selected several alternative 3' terminal exon events, including both Type I and Type II events, for independent RT-PCR validation. We used a competitive PCR approach that employed a common forward primer and reverse primers specific to each alternative 3' end. The disadvantage of this method is that it cannot be used as accurately to determine the true ratio of the alternative splice variants as in the case of cassette exons using common PCR primers. However, this approach was sufficient to detect robust changes in splicing mediated by the ESRPs in the validation assays.

Exon two of the SF1 gene transcript has a 5' splice site in competition with a polyA signal. When the 5' splice site in exon two is used, it is spliced to exon three and the remaining downstream exons. Conversely, when this 5' splice site is not used, the SF1 transcript is polyadenylated and terminates in the intron immediately downstream of exon two, thereby encoding a significantly truncated isoform of SF1 (Fig. 5A). Our analysis of the MADS data predicted the Type I alternative 3' end splicing of the SF1 gene to be a target of ESRP regulation. Three of the five probesets corresponding to the portion of exon two created by skipping of its 5' splice site were present in the MADS top 500 and showed increased expression in the knockdown while the probesets corresponding to the downstream exons display decreased expression (Fig. 5A, gene view). Together, these results suggest that the ESRPs enhance splicing of exon two to the downstream exons and when they are depleted there is increased polyadenylation in the intron downstream of exon two. This prediction was confirmed by RT-PCR. In the control, the predominant splicing product is the full length SF1 transcript. When the ESRPs are downregulated, we observe a decrease in the full-length transcript and an increase in the truncated splice variant (Fig. 5A). We validated an additional Type I alternative 3' end splicing target of the ESRPs in the SF3B1 gene transcript (Suppl. Material, Fig. S1).

Figure 5
Regulation of alternative 3' terminal exons by the ESRPs. (A) Validation of ESRP regulated alternative 3' end—Type I splicing in the SF1 gene. Five probesets, three of which are present in the MADS top 500 (3377077, 3377078 and 3377080), target ...

The EPB41L5 gene transcript undergoes Type II alternative 3' end splicing. In the EPB41L5 transcript, exon 16 can be spliced to the upstream terminal exon 17, with polyadenylation and termination occurring in the intron downstream of exon 17. Alternatively, exon 17 can be skipped and exon 16 is spliced to exon 18 and the numerous remaining exons in the transcript. Similar to SF1, splicing to exon 17 in the EPB41L5 transcript would result in a significantly truncated protein. Four probesets corresponding to the downstream exons in EPB41L5 were in the MADS top 500 list (Fig. 5B). These four probesets as well as the rest of the probesets corresponding to the EPB41L5 exons downstream of exon 17 showed increased expression. The two probesets corresponding to the alternative terminal exon 17 showed a converse decrease in expression. This data indicate that the ESRPs enhance splicing of exon 17, therefore promoting the shorter EPB41L5 isoform. RT-PCR confirmed this to be true (Fig. 5B). In the control, almost all EPB41L5 terminates at exon 17. However, when the ESRPs are downregulated, we observe a mix of both the short and long splice variants of EPB41L5. We also validated two additional Type II alternative 3' end splicing targets of the ESRPs (CUL4A, and GIT2) and established that the ESRPs promote splicing of the upstream terminal exon in these two cases as well (Fig. S1).

Additional validation of predicted ESRP regulated splicing events

In total, we tested 18 alternative splicing events from the list of high-confidence ESRP targets by RT-PCR. Of these 18, we were able to validate 14 novel alternative splicing events regulated by the ESRPs (Table 1). We note, however, that one of the events that was not validated was a predicted retained intron in RBM39. While this retained intron was not validated, an increase in the inclusion of a known alternative cassette exon in RBM39 flanked by this intron was observed in response to ESRP knockdown. To further validate changes in splicing associated with downregulation of the ESRPs, we submitted RNA samples for high-throughput (HT) RT-PCR using high-resolution capillary electrophoresis.22 This analysis was used to investigate an additional 35 targets, performing two different PCR reactions per target. To facilitate direct analysis of the products and simplify the design of PCR primers for this approach, we focused use of this method towards analysis of simple, single alternative cassette exons. Due to space limitations, we present only a few representative examples here. A complete summary of the (HT) RT-PCR data can be found in the Supplementary Material (Table S2). All the data from these experiments, including PCR design, primer sequences, the sizes and molarities of the products, and the electrophoregrams are accessible online (see Materials and Methods). Electrophoregrams of three pairs of RT-PCR reactions are shown in Figure 6. These examples demonstrate clear changes in splicing of the ESRP enhanced exon in the FNIP1 gene transcript and the ESRP silenced exons in OSBPL3, and GOLGA4 gene transcripts. In total, 24 of the 35 events tested were confirmed to be valid changes in splicing as predicted by the array data. Four of the ten targets that we were unable to validate (NASP, TGFBR2, PRMT1, GPR126 [probesets 2334420, 2615384, 3838818 and 2928487 respectively]) suffered from errors or ambiguities in primer design and were thus removed from the overall analysis as no conclusions could be drawn. Exon two of the NT5C3 gene transcript (probeset 3045024) was predicted to be enhanced by the ESRPs, but RT-PCR results showed that exon inclusion increased upon ESRP depletion. This contradiction might be explained by the presence of alternative first exons, only one of which was targeted by the PCR primers. Four events (OSBPL3, DZIP1, ASPH and TIA1 [probesets 3041901, 3521435, 3137642 and 2558539, respectively]) demonstrated small changes in splicing that did not pass our threshold of an 1.0% change in splicing or a change was detected in only one of two PCR reactions. Several of these exons may in fact be regulated by the ESRPs, but without additional replicates, we cannot confidently conclude that these are validated splicing targets. Combining this data with the validations summarized in Table 1, we validated a change in splicing in 38 out of 49 tested splicing events that were identified in the top 500 of the MADS analysis. Thus, at a minimum, 78% of the splicing events that we subjected to validation could be confidently concluded to represent true ESRP regulated splicing events.

Figure 6
HT RT-PCR validation of additional targets of ESRP splicing regulation. HT RT-PCR reactions characterizing the alternative splicing patterns of three genes are shown as electrophoregrams. Sizes and molarities of the RT-PCR products are shown above the ...
Table 1
Summary of ESRP target events tested by RT-PCR

Depletion of ESRP1, but not ESRP2, is sufficient to induce switching towards mesenchymal splicing patterns

To further investigate functional redundancy between ESRP1 and ESRP2 as well as the relative requirements for each factor in maintaining epithelial splicing pathways, mRNAs encoding each protein were individually depleted as well in combination, as was previously done for the exon array. We then assayed splicing of two gene transcripts with simple cassette exons that displayed robust changes in splicing in response to ESRP1 and ESRP2 depletion (Fig. S2A). Depletion of ESRP1 alone was sufficient to induce a partial increase in splicing of the SCRIB exon and combined depletion of both ESRP1 and ESRP2 caused a further increase in exon inclusion. In contrast, depletion of ESRP2 alone was not sufficient to induce a change in splicing of this exon. In the case of SLK, depletion of ESRP1 alone partially induced exon skipping which was similarly augmented by combined depletion of ESRP1 and ESRP2 (Fig. S2A). However, depletion of ESRP2 alone was unable to induce exon skipping, and if anything, a slight increase was observed. The specificity and efficacy of knocking down each factor was validated by quantitative RT-PCR (Fig. S2B). These results thus suggest that while ESRP1 and ESRP2 generally have functionally redundant activities, expression of ESRP1 would appear to have more essential functions in splicing regulation (see also below).

Ectopic expression of mEsrp1 in a mesenchymal cell line induces reciprocal changes in splicing of ESRP target gene transcripts

To further investigate the role of the ESRP1 and ESRP2 in regulation of these novel splicing targets, we tested whether ectopic expression of the ESRPs was also sufficient to promote the opposite changes in splicing in a mesenchymal cell line as those seen in epithelial cells upon ESRP depletion. Our previous work demonstrated that ectopic expression of an mEsrp1 cDNA in the mesenchymal human breast cancer cell line MDA-MB-231 was sufficient to induce the opposite changes in splicing of FGFR2, CD44, ENAH and CTNND1 from those seen with ESRP depletion in the PNT2 cell line.9 Using this experimental system, we tested the effect of mEsrp1 expression on splicing of the 18 splicing events that we could test ourselves using the primers sets we had for RT-PCR (Table 1). A subset of these data is presented in Figure 7. The ESRP regulated exons in the FLNB and SLK transcripts are included at low levels in the control 231 cells (EV). Expression of mEsrp1 greatly enhances splicing of these exons (Fig. 7, top). Two examples of exons that are suppressed by the ESRPs are also shown (Fig. 7, middle). The SCRIB alternative exon is included in 58% of transcripts in the control cells where cells expressing mEsrp1 only splice this exon in 9% of transcripts. In the case of MYO1B, inclusion of the tandem exons is strongly silenced as the percentage of transcripts containing one or both exons decreases from 31% to 8% upon mEsrp1 expression. Finally, an example of mEsrp1 being sufficient to alter alternative 3' end splicing is shown for the EPB41L5 transcript (Fig. 7, bottom). In the control cells, there is a nearly equal mix of the short and long splice variant of EPB41L5. However, when mEsrp1 is expressed in these cells, the long splice variant is almost completely downregulated in favor of the truncated splice variant. In total, out of the 15 targets that were confirmed in the PNT2 ESRP knockdown experiment, 8 showed reciprocal changes in splicing upon ectopic expression of mEsrp1 in the MDA-MB-231 cells (Fig. 7 and Fig S3). Interestingly, while not a true “validated” splicing change, the cassette exon in RBM39 also showed a reciprocal change from that observed upon ESRP knockdown (Fig. S3). Unexpectedly, the direction of change in splicing of the mutually exclusive exons in the OGDH transcript was the same in the mEsrp1 expression system as it was in the ESRP knockdown experiment (Fig. S3). One possible explanation for this unexpected finding could be that the ESRPs have differential effects on splicing of one or both exons in different cellular milieus, or that they may indirectly regulate these exons. Given that most exons are under regulation by complex combinatorial control by multiple regulatory factors, it is possible that antagonistic functions of the ESRPS may be due to interactions with different regulatory proteins in different cell types. However, at present we cannot provide a simple explanation to account for this seemingly contradictory observation.

Figure 7
Changes in alternative splicing in MDA-MB-231 cells ectopically expressing Esrp1. Total RNA was isolated and prepared after transduction of the mesenchymal MDA-MB-231 cells with retrovirus encoding FLAG-tagged mouse Esrp1 or a control empty vector (EV). ...

Six of the validated splicing changes observed in response to ESRP knockdown showed no change in response to mEsrp1 expression in mesenchymal cells. One possible explanation was that some of these events might instead be specifically regulated by ESRP2. To address this possibility, we analyzed the splicing of these targets in MDA-MB-231 cells when mEsrp1, mEsrp2, or both proteins together were ectopically expressed. In five of these six cases (FAM62A, CUL4A, TRIP10, SF3B1 and SF1) neither mEsrp1, mEsrp2, nor a combination of both induced splicing changes. In the case of the CHRNA5 alternative 5' splice site there did appear to be evidence that mEsrp2 could induce a change in splicing that was not achieved by mEsrp1, but the effect was small (Fig. S4A). As controls, we also tested the effects of ectopic expression of both paralogs on splicing of SLK and SCRIB. In both cases we noted that mEsrp1 induced more profound changes in splicing than mEsrp2 and coexpression of both proteins showed no evidence of additive functions (Fig. S4B). Thus, together with the results from independent ESRP1 and ESRP2 knockdown, these results suggest that these proteins have similar, redundant functions, but that ESRP2 may be a less robust splicing factor than ESRP1. It is possible, particularly for the alternative 3' terminal exon splicing events that are often difficult to validate, that these assays may have been unable to detect some mEsrp1 induced splicing changes. However, it may be more likely that for some of these types of splicing events, the ESRPs are necessary for expression of epithelial isoforms, but ectopic expression of mEsrp1 alone may not be sufficient to promote a switch in splicing from the mesenchymal to the epithelial pathway. Here too it is possible that some of the discrepancies might be due to indirect effects on splicing in one or both experimental systems. Nonetheless, taken together the results from the ectopic expression experiments support the conclusion that the ESRPs are in fact bona fide regulators of most of the splicing changes identified in the MADS analysis of the microarray data.

ESRP target genes share common functions and biological processes, including numerous examples of genes that function in cell-cell adhesion and cell motility

We were interested in investigating whether the set of ESRP splicing targets we identified could be shown to function in any common pathways and have related biological roles. As a first step in such analysis, the 134 genes from the list of high-confidence splicing events within the MADS top 500 probesets were analyzed for enriched Gene Ontology (GO) terms using the DAVID Bioinformatics Resource.23 DAVID was unable to map the following four genes (MNAT, AGFG1, PMS2L14, ARFGAP2) and thus, these are absent in the analysis. Compared to the background set of all genes represented by probesets on the Exon Arrays, we noted several GO terms that were enriched among the genes in the high confidence list, although the list of genes is still not large enough to have strong statistical power (Fig. 8A). This analysis suggested that a number of the ESRP target genes encode proteins involved in cytoskeleton structure, cell adhesion, RNA splicing, and the other categories presented.

Figure 8
A subset of ESRP target genes regulate cytoskeleton structure and cell adhesion. (A) Gene Ontology (GO) analysis of ESRP targets. The table lists the number of ESRP targets in each GO category, the enrichment for each category, and the corresponding p-value. ...

Though not all of the predicted splicing events in the 134 genes analyzed have been validated, the validations performed suggest that most of the genes in this list are true targets. Furthermore, any false positives would not be predicted to cluster in a way that would further support the observed enrichment.

We wished to further investigate whether relevant biological functions or themes could be identified among the list of ESRP target genes and thus we also carried out an extensive literature review of these gene products. A subset of these genes could be shown to encode proteins with important roles in the regulation of the actin cytoskeleton, cell-cell adhesion, and cell motility and migration (Suppl. Material, Table S3). Among these gene products were a number that have well documented and essential roles in epithelial to mesenchymal transitions and maintenance of cell polarity in addition to the previously reported FGFR2, p120-catenin, CD44 and ENAH. Such genes include, for example, EPB41L5, SCRIB, MACF1 (ACF7), GIT2, LOXL2, FAT and PTPRM. We also noted a number of cases in which these gene products have been shown to be involved in common regulatory pathways by physically associating with one another (See discussion below). These observations thus suggest that many of these targets function in common pathways and imply that the different splice variants may well prove to have different, or even opposing functions in epithelial and mesenchymal cells as has already been demonstrated for FGFR2 and p120-catenin.


We used the Affymetrix Exon Arrays to identify novel alternative splicing events that are controlled by the epithelial splicing regulators ESRP1 and ESRP2. The results of our analysis significantly expanded the number of known and candidate alternative splicing events that are under the control of these cell-type-specific splicing regulators. Based on our previous work demonstrating its efficacy, we employed the MADS tool to detect differentially expressed probesets.18 Genome wide analysis of alternative splicing in the human prostatic epithelial cell line PNT2 following ESRP depletion yielded thousands of exon probesets with differential expression as detected by this approach. For the purposes of this study we limited our analysis to the 500 exon probesets that showed the most significant differential expression as detected using MADS. Further analysis was limited to a subset of 148 high confidence ESRP-regulated splicing events in which well annotated alternative splicing has been documented. We selected 52 of these events for independent validation. Based on our conservative estimate, most of these events showed true changes in splicing upon ESRP depletion. Furthermore, in the 15 cases tested, eight alternative splicing events showed reciprocal changes in splicing upon ectopic expression of mEsrp1 in a mesenchymal cell line, indicating that the ESRPs are not only necessary, but often sufficient for the regulation of many of these events.

One of our more notable observations was that the ESRPs regulate every major category of the known types of alternative splicing. While some of these types of events can be difficult to validate, we were able to validate at least one of each of these types of alternative splicing. While most previous studies on splicing factor targets have predominantly focused on cassette exons, we were particularly intrigued by the number of cases in which the events yielded alternative 3' ends. A number of the latter types of events are predicted to significantly change the size and amino acid sequence of the encoded proteins, indicating a high likelihood that the isoforms have distinct functions. It also merits consideration that the expression of mRNA transcripts with different 3' UTRs may further subject them to differential regulation by other RBPs and microRNAs.24 Several of the validated events demonstrate very robust changes in splicing including cases in which the ESRPs enhanced or silenced exon inclusion. For example, the inclusion of cassette exons in FLNB, SLK, ARFGAP2 and FNIP1 are strongly enhanced while cassette exons in SCRIB, MYO1B, OSBPL3 and GOLGA4 are strongly repressed by the ESRPs. The robust changes in inclusion levels of these evolutionarily conserved alternatively spliced exons thus give us reason to suspect that the resulting protein isoforms are likely to have different functions that are critical to perform appropriate cell-type-specific roles.

While the primary goal of this study was to identify a set of novel ESRP splicing targets and not to validate the bioinformatic approach or microarray platform, we point out several limitations of this study. First, we restricted the potential number of target events we could identify by profiling the transcriptome of only one cell-type. It is likely there are many genes with ESRP-regulated splicing events that are not expressed in the PNT2 epithelial cell line, as well as in genes that are preferentially expressed in mesenchymal cells. Second, a number of relevant alternative splicing events lead to premature termination codons that will cause mRNA degradation through nonsense-mediated decay (NMD).25 By eliminating transcripts with more than a 2.0 fold change in expression, we are biased against the detection of these types of events. Third, we limited our analysis of the data to just the top 500 ranked probesets that showed differential expression by MADS. This was partly due to one drawback in analysis of the exon array data: the probesets are not specifically designed to target defined splicing events and many of the differential probesets did not target events that appeared to represent alternative splicing. Thus, the selection of a list of “high confidence” events required time consuming case by case analysis and alignment of each probeset. Even among the 500 probesets we selected for analysis, many mapped to introns or intergenic regions and thus no interpretation of a splicing event could be made, further limiting our analysis. However, while many probesets in the top 500 list were not validated to be ESRP regulated events, there are undoubtedly additional regulated events among the list of genes further down the list of those selected for differential exon inclusion. Therefore, given the large number of novel splicing events identified in our rather limited analysis, we suspect that there remain many more targets of ESRP splicing to be identified.

The advantage of this exon-tiling array is the high density of probes targeting all known and predicted exons in the human genome, offering a global, non-biased means to identify regulated splicing events. An alternative approach to the exon arrays are splice junction microarrays, which incorporate probesets that target the exon junctions formed by alternative splicing. The inverse correlation between probes for exon skipping and inclusion may facilitate easier and more accurate data analysis using some types of junction arrays. The relative advantages and disadvantages of these different microarray designs have been reviewed elsewhere.10 However, we point out that most existing exon junction arrays are less likely to identify novel alternative splicing events and are less capable of detecting some of the alternative 3' end types of events. Our detection of numerous examples of alternative 3' ends thus illustrates how an exon tiling approach may have some advantages compared to other array formats. In addition, detection of these events is greatly facilitated when used in conjunction with a software application such as the Geneview feature in the Partek software that allows all probesets signals for a given transcript to be viewed simultaneously. Recently, use of deep sequencing of transcriptomes has been utilized to probe changes in alternative splicing and as this technology matures may supplant some uses of splicing sensitive microarrays to globally profile splicing.2,26 However, there remain some limitations to the application of this technology for comprehensive identification of alternative splicing.27

Several previous studies have been undertaken to comprehensively profile alternative splicing events regulated by both ubiquitous splicing factors as well as those regulated by cell-type or developmental-stage specific splicing factors.10,28,29 The splicing regulatory networks characterized in these studies found that many of the co-regulated splicing events occur within transcripts that encode proteins that function in biologically coherent pathways and/or physically interact with one another in manners that are relevant to the tissue or cell type in which the splicing factor is expressed. Such coordinated modules of post-transcriptional regulation by RNA binding proteins and the transcripts they regulate have been described as “post-transcriptional operons” or “regulons”.30 In one example, identification of neuronal splicing events in control mice versus mice lacking the brain specific splicing factor Nova demonstrated that the Nova regulated target mRNAs encode proteins involved in neurogenesis and synaptic transmission.31 A splicing regulatory network has also been established in the developing vertebrate heart where developmental differences in expression of the CELF and MBNL splicing factors were shown to regulate changes in splicing in genes that function in cell structure and muscle contraction.32 Splicing regulatory networks have also been established for the Fox family of splicing factors. One study combined bioinformatics tools and sequence conservation with splicing microarray data to identify thousands of targets of Fox1/2 splicing regulation.33 A second study focused on the splicing regulatory functions of Fox2 in human embryonic stem cells through a combined cross-linking immunoprecipitation and sequencing approach.34 Both studies found that the targets of Fox regulation shared functions involved in regulating the neuromuscular junction as well as mRNA splicing.

By analogy with the previous studies that profiled coordinated splicing events, our study similarly provides strong evidence that the ESRPs are global regulators of an epithelial-specific splicing regulatory network. A number of the genes in our list share common functions and biological processes such as cytoskeleton binding and cell adhesion, both of which are important features in defining cellular differences between epithelial and mesenchymal cells. A hypothesis for which this study also provides further evidence is that the different splice variants the ESRPs regulate will be shown to have distinct properties that contribute to the distinct functions and characteristics of epithelial vs. mesenchymal cell types. Such isoform differences in function are furthermore likely to play important roles in epithelial to mesenchymal transitions during development as well as in pathophysiologic conditions. We previously discussed work demonstrating that the mesenchymal isoforms of p120-catenin promote cell migration and invasion, in contrast to the non-mesenchymal forms that promote E-cadherin-based cell-cell adhesion.9,35 We anticipate that isoform-specific differences in the functions of several other targets identified here will also be shown to resolve similar paradoxes in which seemingly antagonistic roles of the same gene to promote cell-cell-adhesion vs. migration have been observed. One potential example is in the EPB41L5 gene products that encode distinct long and short isoforms of the protein due to splicing changes that lead to different 3' ends (Fig. 8B). EPB41L5 has been shown to promote the EMT through downregulation of E-cadherin mediated cell-cell adhesion through a mechanism that appears to involve binding to p120-catenin.36 Mice with EPB41L5 loss of function mutations have defects in the EMT during gastrulation.36,37 However, the same gene is also expressed endogenously in epithelial cells and thus it has been unclear why it would not promote the EMT in this cell context. Furthermore, EBP41L5 and its orthologs in zebrafish and Drosophila control epithelial polarity.3840 The presence of a paxillin binding domain within the alternative C-terminus of the mesenchymal long isoform might contribute to isoform-specific differences that could resolve this paradox. It is of interest that another ESRP target gene, GIT2, also lacks a C-terminal paxillin binding site (Fig. 8C).41 The GIT proteins, GIT1 and GIT2, have been shown to regulate cytoskeletal dynamics during cell migration and mesenchymal long isoforms of GIT2 as well as EPB41L5 that can bind to paxillin might differentially regulate focal adhesion complex formation and turnover.41,42

Scribble (Scrib) is another ESRP regulated gene with evolutionarily conserved roles in the regulation of cell polarity and maintenance of cell-cell adhesion.43 Depletion of Scrib was shown to disrupt E-cadherin mediated cell-cell adhesion and to promote migratory properties and a mesenchymal appearance in Madin Darby Canine Kidney cells.44 However, Scrib has also been suggested to be required for cell migration suggesting that this gene also has context-dependent functions.45 Here again, it may be that the expression of epithelial vs. mesenchymal splice variants could resolve such a contradiction. While the alternative cassette exon in Scrib that is alternatively spliced does not affect any obvious protein motifs, this represents an evolutionarily conserved alternative splicing event and the changes in splicing upon ESRP depletion or ectopic expression are quite robust. Filamin-B also demonstrates very robust changes in splicing of a cassette exon whose inclusion is strongly ESRP dependent. This exon encodes a well-described hinge region (H1) located between stretches of filamin repeats in the filamin B protein and increased skipping of this exon was shown during myogenesis (Fig. 8D).46 The filamin proteins are actin binding proteins that play important roles in cytoskeletal organization and the filamin B isoform lacking the H1 region was further shown to be differentially localized at focal complexes and associated with B1 integrins at tips of actin stress fibers.46 Thus, expression of these different filamin B isoforms was proposed to differentially regulate cell morphology and differentiation. Additional examples of ESRP targets with roles in the EMT or regulation of cell-cell adhesion and/or migration are detailed in Table S3.

The identification of a number of physiologically relevant ESRP target genes that function in common cellular processes thus provides evidence that they also are part of an “RNA operon” that adds another functionally coregulated layer of gene expression control. As a further example of this, we also identified a number of cases in which the published literature has previously shown physicial interactions among some of the ESRP targets we identified. Perhaps the best examples of such protein-protein interactions involve p120-catenin which has been shown to have interactions with EPB41L5 as well as CASK, PTPRM, CDH3, GSK3B and ACTG1.36,4751 We anticipate that identification of an even larger subset of ESRP targets may yield a larger protein interactome that defines more genes involved in pathways that regulate cell morphology and motility.

Ongoing studies will be needed to begin to unravel the mechanism by which the ESRPs regulate splicing, including a determination as to the means by which they can both promote and repress the inclusion of different exons. Investigations of other splicing factors that both enhance and silence exons have shown that the location of splicing factor binding with respect to the regulated exon often correlates with the effect it will have on splicing of the exon.3234,52 Such “RNA maps” have been shown to be capable of identifying novel splicing targets of some of these regulators as well as to predict whether they will promote exon inclusion or skipping of cassette exons. In the case of the Nova and Fox family of splicing factors, binding upstream of a regulated exon usually leads to exon silencing, whereas they usually enhance splicing from downstream binding sites.33,53 Genome-wide studies of these splicing factors have benefited from prior knowledge of their respective binding sites or motifs in order to validate direct targets of splicing. We previously identified a GU-rich binding site for the ESRPs within the FGFR2 pre-mRNA. 9 Relative to this binding site, the ESRPs enhance splicing of the upstream IIIb exon and silence the downstream IIIc exon suggesting that similar rules for position dependent effects on splicing may be a general rule for the ESRPs as well. Consistent with this possibility, the introns downstream of the epithelial exons v8 and v9 in the CD44 pre-mRNA also contain “GU” rich splicing enhancers.54 However, experimental identification of a more defined ESRP binding sequence or motif is needed to guide future studies to investigate whether an RNA map of ESRP binding sites can explain their global effects on splicing. Given our findings that numerous non-cassette exons are regulated by the ESRPs we hope to derive predictive rules that can account for their functions in other types of alternative splicing events as well. However, it is also likely that the ESRP-mediated regulation of the events defined here and others still to be described will also involve cooperative interactions with other splicing factors that regulate the same alternative exons. Furthermore, we identified several splicing factors among the ESRP targets, as has also been shown for other splicing regulators. Thus it is possible that some of the splicing changes we identified may not represent direct targets of the ESRPs. Therefore, several layers at which different splicing regulators converge to regulate subsets of alternatively splicing exons may preclude the determination of a simple RNA map to account for the mechanisms by which the ESRPs regulate numerous types of splicing events. Future identification of direct binding sites for the ESRPs using approaches such as high-throughput sequencing of RNA isolated by cross-linking and immunoprecipitation (HITS-CLIP; also known as CLIP-Seq) should help to clarify a consensus ESRP binding site and lead to a high resolution RNA map to account for ESRP functions.34,52,55 Such studies will begin to uncover the layers at which different splicing regulators converge to regulate subsets of alternatively spliced exons and to provide a “splicing code” that can describe and predict patterns of alternative splicing in diverse cells and tissues. Identification of the set of ESRP regulated exons defined in this study should facilitate future studies to investigate the complex and combinatorial mechanisms by which they greatly expand ribonomic diversity.

Materials and Methods

Cell culture

PNT2 and MDA-MB-231 cells were maintained as described.9

RNA interference

Transfections of siRNAs in PNT2 cells were carried out as described.9 The sequences of the sense strands of the siRNAs targeting ESRP1 and ESRP2 are 5'-CAC AAU GAC AGA GUA UUU AAA-3' and 5'-AGC CCG AGG UGA UAA AGC A-3', respectively.

Exon microarray target preparation, array hybridization and data analysis

Total RNA was isolated using Trizol (Invitrogen). Ribosomal RNA was removed using the RiboMinus kit (Invitrogen). Biotinylated sense-strand DNA targets were prepared using the Affymetrix GeneChip Whole Transcript (WT) Sense Target Labeling Assay according to the manufacturer’s directions. Hybridization of end-labelled ssDNA to the Affymetrix Human Exon Array ST 1.0 was performed according to the manufactur’s directions.13 Arrays were scanned using the Affymetrix GCS 3000 7G and GeneChip Operating Software (GCOS) to produce .CEL intensity files. The exon array data can be accessed at NCBI Gene Expression Omnibus repository (accession number GSE17468)

Exon 1.0 array data were processed by MADS (Microarray Analysis of Differential Splicing) to identify exons differentially spliced between control and ESRP knockdown groups. MADS is a tool we developed for exon array analysis of alternative splicing. In our previous study of the splicing regulator PTB (polypyrimidine tract binding protein), we used MADS to identify novel exon targets of PTB at an experimental validation rate of 90%.18 The MADS calculation involves the following steps: (1) correction of background noise;56 (2) expression index calculation;57 (3) detection and removal of probes with potential cross-hybridization to off-target transcripts;58 (4) calculation of probeset p-values for differential alternative splicing. The key feature of MADS is to model and remove microarray noise in order to achieve a high accuracy in exon-level analysis. As in our MADS analysis of PTB,18 we removed genes with low expression levels in all samples (i.e., expression index of smaller than 500), as well as genes showing more than two-fold change in expression levels between control and ESRP knockdown groups. For remaining probesets, we ranked all probesets according to their MADS p-values. The top 500 probesets were manually inspected in the UCSC Genome Browser for their genomic locations and gene structures.

Reverse transcription and PCR

Reverse transcription, PCR and splicing analysis were performed essentially as described.9 Quantification of most splicing events was calculated as the long isoform divided by the sum of the long isoform and short isoform. In the examples of tandem cassette exons, quantification was calculated as the sum of single exon and double exon inclusion products divided by the sum of the inclusion products and skipped products. In the example of CHRNA5 where exon five can be spliced at one of three alternative 5' splice sites, the longest form of the exon was quantified as the longest isoform divided by the sum of the longest exon isoform, intermediate exon isoform, and shorted exon isoform. For OGDH we quantified exon 4a inclusion as Pvu II resistant PCR products divided by the sum of Pvu II resistant products and BstE II resistant products. Corrections of molar equivalents for differences in size were made prior to quantification calculations. Amplifications of splicing targets with alternative 3' ends was carried out using a universal forward primer and equimolar quantities of two reverse primers unique to each of the splice variants. PCR primer sequences are available in the Supplementary Material (Table S4). Real time RT-PCR assays for ESRP1 and ESRP2 mRNA quantification was performed as described.9


HT RT-PCR, capillary electrophoresis, and data analysis were performed by the RNomics Platform of Genome Quebec at the University of Sherbrooke as previously described.22 Quantification of splicing changes was calculated the sum of “expected” inclusion product molarities divided by the sum of all “expected” product molarities. PCR primer locations, sequences and electropherograms of all reactions can be accessed on the web at http://palace.lgfus.ca/login.

Esrp1 and Esrp2 ectopic expression

Flag-tagged mouse Esrp1 or Esrp2 was expressed using a retroviral system as previously described.9

Supplementary Material

Supplementary Figures and Table S3

Table S1

Table S2


We thank Don Baldwin and the staff at the Penn Microarray Facility for assistance with sample preparation and microarray hybridizations, John Tobias for technical support on exon array data analysis, Roscoe Klinck and Philippe Thibault for expert assistance with the HT RT-PCR, David Eichmann, Ben Rogers and the University of Iowa Institute for Clinical and Translational Science for computer support, and Behnam Nabet, Wing Hung Wong and Kimberly Dittmar for technical support and helpful discussion.

Financial disclosure

This work was supported by the National Institutes of Health (R01 CA093769 to R.P.C., R01HG004634 to Y.X.); the United States Department of Defense (PC 991539 to R.P.C.); a Penn Genome Frontiers Institute Seed Grant to R.P.C.; a research startup fund from the University of Iowa to Y.X.; and a Genetics Training Grant from the National Institutes of Health (T32-GM08216 to C.C.W.).


1. Hartmann B, Valcarcel J. Decrypting the genome’s alternative messages. Curr Opin Cell Biol. 2009;21:377–386. [PubMed]
2. Wang ET, Sandberg R, Luo S, Khrebtukova I, Zhang L, Mayr C, et al. Alternative isoform regulation in human tissue transcriptomes. Nature. 2008;456:470–476. [PMC free article] [PubMed]
3. Pan Q, Shai O, Lee LJ, Frey BJ, Blencowe BJ. Deep surveying of alternative splicing complexity in the human transcriptome by high-throughput sequencing. Nat Genet. 2008;40:1413–1415. [PubMed]
4. Stamm S, Ben-Ari S, Rafalska I, Tang Y, Zhang Z, Toiber D, et al. Function of alternative splicing. Gene. 2005;344:1–20. [PubMed]
5. Matlin AJ, Clark F, Smith CW. Understanding alternative splicing: towards a cellular code. Nat Rev Mol Cell Biol. 2005;6:386–398. [PubMed]
6. Black DL. Mechanisms of alternative pre-messenger RNA splicing. Annu Rev Biochem. 2003;72:291–336. [PubMed]
7. Gabut M, Chaudhry S, Blencowe BJ. SnapShot: The splicing regulatory machinery. Cell. 2008;133:191–192. [PubMed]
8. Hanamura A, Caceres JF, Mayeda A, Franza BR, Jr, Krainer AR. Regulated tissue-specific expression of antagonistic pre-mRNA splicing factors. RNA. 1998;4:430–444. [PMC free article] [PubMed]
9. Warzecha CC, Sato T, Nabet B, Hogenesch JB, Carstens RP. ESRP1 and ESRP2 are epithelial cell-type-specific regulators of FGFR2 splicing. Mol Cell. 2009;33:591–601. [PMC free article] [PubMed]
10. Moore MJ, Silver PA. Global analysis of mRNA splicing. RNA. 2008;14:197–203. [PMC free article] [PubMed]
11. Affymetrix. 2005b
12. Affymetrix. 2005a
13. Gardina PJ, Clark TA, Shimada B, Staples MK, Yang Q, Veitch J, et al. Alternative splicing and differential gene expression in colon cancer detected by a whole genome exon array. BMC Genomics. 2006;7:325. [PMC free article] [PubMed]
14. Yeo GW, Xu X, Liang TY, Muotri AR, Carson CT, Coufal NG, Gage FH. Alternative splicing events identified in human embryonic stem cells and neural progenitors. PLoS Comput Biol. 2007;3:1951–1967. [PMC free article] [PubMed]
15. Clark TA, Schweitzer AC, Chen TX, Staples MK, Lu G, Wang H, et al. Discovery of tissue-specific exons using comprehensive human exon microarrays. Genome Biol. 2007;8:R64. [PMC free article] [PubMed]
16. Hung LH, Heiner M, Hui J, Schreiner S, Benes V, Bindereif A. Diverse roles of hnRNP L in mammalian mRNA processing: a combined microarray and RNAi analysis. RNA. 2008;14:284–296. [PMC free article] [PubMed]
17. Oberdoerffer S, Moita LF, Neems D, Freitas RP, Hacohen N, Rao A. Regulation of CD45 alternative splicing by heterogeneous ribonucleoprotein, hnRN-PLL. Science. 2008;321:686–691. [PMC free article] [PubMed]
18. Xing Y, Stoilov P, Kapur K, Han A, Jiang H, Shen S, et al. MADS: a new and improved method for analysis of differential alternative splicing by exon-tiling microarrays. RNA. 2008;14:1470–1479. [PMC free article] [PubMed]
19. Wu Z, Jia X, de la Cruz L, Su XC, Marzolf B, Troisch P, et al. Memory T cell RNA rearrangement programmed by heterogeneous nuclear ribonucleoprotein hnRN-PLL. Immunity. 2008;29:863–875. [PMC free article] [PubMed]
20. Yeo GW, Van Nostrand E, Holste D, Poggio T, Burge CB. Identification and analysis of alternative splicing events conserved in human and mouse. Proc Natl Acad Sci USA. 2005;102:2850–2855. [PMC free article] [PubMed]
21. Tian B, Pan Z, Lee JY. Widespread mRNA polyadenylation events in introns indicate dynamic interplay between polyadenylation and splicing. Genome Res. 2007;17:156–165. [PMC free article] [PubMed]
22. Klinck R, Bramard A, Inkel L, Dufresne-Martin G, Gervais-Bird J, Madden R, et al. Multiple alternative splicing markers for ovarian cancer. Cancer Res. 2008;68:657–663. [PubMed]
23. Dennis G, Jr, Sherman BT, Hosack DA, Yang J, Gao W, Lane HC, Lempicki RA. DAVID: Database for Annotation, Visualization, and Integrated Discovery. Genome Biol. 2003;4:3. [PMC free article] [PubMed]
24. Sandberg R, Neilson JR, Sarma A, Sharp PA, Burge CB. Proliferating cells express mRNAs with shortened 3’ untranslated regions and fewer microRNA target sites. Science. 2008;320:1643–1647. [PMC free article] [PubMed]
25. McGlincy NJ, Smith CW. Alternative splicing resulting in nonsense-mediated mRNA decay: what is the meaning of nonsense? Trends Biochem Sci. 2008;33:385–393. [PubMed]
26. Castle JC, Zhang C, Shah JK, Kulkarni AV, Kalsotra A, Cooper TA, Johnson JM. Expression of 24,426 human alternative splicing events and predicted cis regulation in 48 tissues and cell lines. Nat Genet. 2008;40:1416–1425. [PMC free article] [PubMed]
27. Li H, Lovci MT, Kwon YS, Rosenfeld MG, Fu XD, Yeo GW. Expression of 24,426 human alternative splicing events and predicted cis regulation in 48 tissues and cell lines. Proc Natl Acad Sci USA. 2008;105:20179–20184. [PMC free article] [PubMed]
28. Blencowe BJ. Alternative splicing: new insights from global analyses. Cell. 2006;126:37–47. [PubMed]
29. Ben-Dov C, Hartmann B, Lundgren J, Valcarcel J. Genome-wide analysis of alternative pre-mRNA splicing. J Biol Chem. 2008;283:1229–1233. [PubMed]
30. Keene JD. RNA regulons: coordination of post-transcriptional events. Nat Rev Genet. 2007;8:533–543. [PubMed]
31. Ule J, Ule A, Spencer J, Williams A, Hu JS, Cline M, et al. Nova regulates brain-specific splicing to shape the synapse. Nat Genet. 2005;37:844–852. [PubMed]
32. Kalsotra A, Xiao X, Ward AJ, Castle JC, Johnson JM, Burge CB, Cooper TA. A postnatal switch of CELF and MBNL proteins reprograms alternative splicing in the developing heart. Proc Natl Acad Sci USA. 2008;105:20333–20338. [PMC free article] [PubMed]
33. Zhang C, Zhang Z, Castle J, Sun S, Johnson J, Krainer AR, Zhang MQ. Defining the regulatory network of the tissue-specific splicing factors Fox-1 and Fox-2. Genes Dev. 2008;22:2550–2563. [PMC free article] [PubMed]
34. Yeo GW, Coufal NG, Liang TY, Peng GE, Fu XD, Gage FH. An RNA code for the FOX2 splicing regulator revealed by mapping RNA-protein interactions in stem cells. Nat Struct Mol Biol. 2009;16:130–137. [PMC free article] [PubMed]
35. Yanagisawa M, Huveldt D, Kreinest P, Lohse CM, Cheville JC, Parker AS, et al. A p120 catenin isoform switch affects Rho activity, induces tumor cell invasion, and predicts metastatic disease. J Biol Chem. 2008;283:18344–18354. [PMC free article] [PubMed]
36. Hirano M, Hashimoto S, Yonemura S, Sabe H, Aizawa S. EPB41L5 functions to post-transcriptionally regulate cadherin and integrin during epithelial-mesenchymal transition. J Cell Biol. 2008;182:1217–1230. [PMC free article] [PubMed]
37. Lee JD, Silva-Gagliardi NF, Tepass U, McGlade CJ, Anderson KV. The FERM protein Epb4.1l5 is required for organization of the neural plate and for the epithelial-mesenchymal transition at the primitive streak of the mouse embryo. Development. 2007;134:2007–2016. [PubMed]
38. Gosens I, Sessa A, den Hollander AI, Letteboer SJ, Belloni V, Arends ML, et al. FERM protein EPB41L5 is a novel member of the mammalian CRB-MPP5 polarity complex. Exp Cell Res. 2007;313:3959–3970. [PubMed]
39. Laprise P, Beronja S, Silva-Gagliardi NF, Pellikka M, Jensen AM, McGlade CJ, Tepass U. The FERM protein Yurt is a negative regulatory component of the Crumbs complex that controls epithelial polarity and apical membrane size. Dev Cell. 2006;11:363–374. [PMC free article] [PubMed]
40. Christensen AK, Jensen AM. Tissue-specific requirements for specific domains in the FERM protein Moe/Epb4.1l5 during early zebrafish development. BMC Dev Biol. 2008;8:3. [PMC free article] [PubMed]
41. Schmalzigaug R, Garron ML, Roseman JT, Xing Y, Davidson CE, Arold ST, Premont RT. GIT1 utilizes a focal adhesion targeting-homology domain to bind paxillin. Cell Signal. 2007;19:1733–1744. [PMC free article] [PubMed]
42. Mazaki Y, Hashimoto S, Okawa K, Tsubouchi A, Nakamura K, Yagi R, et al. An ADP-ribosylation factor GTPase-activating protein Git2-short/KIAA0148 is involved in subcellular localization of paxillin and actin cytoskeletal organization. Mol Biol Cell. 2001;12:645–662. [PMC free article] [PubMed]
43. Yamanaka T, Ohno S. Role of Lgl/Dlg/Scribble in the regulation of epithelial junction, polarity and growth. Front Biosci. 2008;13:6693–6707. [PubMed]
44. Qin Y, Capaldo C, Gumbiner BM, Macara IG. The mammalian Scribble polarity protein regulates epithelial cell adhesion and migration through E-cadherin. J Cell Biol. 2005;17:1061–1071. [PMC free article] [PubMed]
45. Dow LE, Kauffman JS, Caddy J, Zarbalis K, Peterson AS, Jane SM, et al. The tumour-suppressor Scribble dictates cell polarity during directed epithelial migration: regulation of Rho GTPase recruitment to the leading edge. Oncogene. 2007;26:2272–2282. [PubMed]
46. van der Flier A, Kuikman I, Kramer D, Geerts D, Kreft M, Takafuta T, et al. Different splice variants of filamin-B affect myogenesis, subcellular distribution, and determine binding to integrin [beta] subunits. J Cell Biol. 2002;156:361–376. [PMC free article] [PubMed]
47. Xia X, Mariner DJ, Reynolds AB. Adhesion-associated and PKC-modulated changes in serine/threonine phosphorylation of p120-catenin. Biochemistry. 2003;42:9195–9204. [PubMed]
48. Lehtonen S, Lehtonen E, Kudlicka K, Holthofer H, Farquhar MG. Nephrin forms a complex with adherens junction proteins and CASK in podocytes and in Madin-Darby canine kidney cells expressing nephrin. Am J Pathol. 2004;165:923–936. [PMC free article] [PubMed]
49. Reynolds AB, Daniel JM, Mo YY, Wu J, Zhang Z. The novel catenin p120cas binds classical cadherins and induces an unusual morphological phenotype in NIH3T3 fibroblasts. Exp Cell Res. 1996;225:328–337. [PubMed]
50. Zondag GC, Reynolds AB, Moolenaar WH. Receptor protein-tyrosine phosphatase RPTPmu binds to and dephosphorylates the catenin p120(ctn) J Biol Chem. 2000;275:11264–11269. [PubMed]
51. Martinez MC, Ochiishi T, Majewski M, Kosik KS. Dual regulation of neuronal morphogenesis by a delta-catenin-cortactin complex and Rho. J Cell Biol. 2003;162:99–111. [PMC free article] [PubMed]
52. Licatalosi DD, Mele A, Fak JJ, Ule J, Kayikci M, Chi SW, et al. HITS-CLIP yields genome-wide insights into brain alternative RNA processing. Nature. 2008;456:464–469. [PMC free article] [PubMed]
53. Ule J, Stefani G, Mele A, Ruggiu M, Wang X, Taneri B, et al. An RNA map predicting Nova-dependent splicing regulation. Nature. 2006;444:580–586. [PubMed]
54. Galiana-Arnoux D, Del Gatto-Konczak F, Gesnel MC, Breathnach R. Intronic UGG repeats coordinate splicing of CD44 alternative exons v8 and v9. Biochem Biophys Res Commun. 2005;336:667–673. [PubMed]
55. Sanford JR, Wang X, Mort M, Vanduyn N, Cooper DN, Mooney SD, et al. Splicing factor SFRS1 recognizes a functionally diverse landscape of RNA transcripts. Genome Res. 2009;19:381–394. [PMC free article] [PubMed]
56. Kapur K, Xing Y, Ouyang Z, Wong WH. Exon arrays provide accurate assessments of gene expression. Genome Biol. 2007;8:82. [PMC free article] [PubMed]
57. Xing Y, Kapur K, Wong WH. Probe selection and expression index computation of Affymetrix Exon Arrays. PLoS ONE. 2006;1:88. [PMC free article] [PubMed]
58. Kapur K, Jiang H, Xing Y, Wong WH. Cross-hybridization modeling on Affymetrix exon arrays. Bioinformatics. 2008;24:2887–2893. [PMC free article] [PubMed]
PubReader format: click here to try


Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...


Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...