Selection preserves Ubiquitin Specific Protease 4 alternative exon skip** in therian mammals

Vlasschaert, Caitlyn; **a, Xuhua; Gray, Douglas A.

doi:10.1038/srep20039

Selection preserves Ubiquitin Specific Protease 4 alternative exon skip** in therian mammals

Article
Open access
Published: 02 February 2016

Volume 6, article number 20039, (2016)
Cite this article

Download PDF

You have full access to this open access article

Scientific Reports

Selection preserves Ubiquitin Specific Protease 4 alternative exon skip** in therian mammals

Download PDF

Caitlyn Vlasschaert^1,3,4,
Xuhua **a^2,3 &
Douglas A. Gray^1,4

1832 Accesses
17 Citations
2 Altmetric
1 Mention
Explore all metrics

Abstract

Ubiquitin specific protease 4 (USP4) is a highly networked deubiquitinating enzyme with reported roles in cancer, innate immunity and RNA splicing. In mammals it has two dominant isoforms arising from inclusion or skip** of exon 7 (E₇). We evaluated two plausible mechanisms for the generation of these isoforms: (A) E₇ skip** due to a long upstream intron and (B) E₇ skip** due to inefficient 5′ splice sites (5′SS) and/or branchpoint sites (BPS). We then assessed whether E₇ alternative splicing is maintained by selective pressure or arose from genetic drift. Both transcript variants were generated from a USP4-E₇ minigene construct with short flanking introns, an observation consistent with the second mechanism whereby differential splice signal strengths are the basis of E₇ skip**. Optimization of the downstream 5′SS eliminated E₇ skip**. Experimental validation of the correlation between 5′SS identity and exon skip** in vertebrates pinpointed the +6 site as the key splicing determinant. Therian mammals invariably display a 5′SS configuration favouring alternative splicing and the resulting isoforms have distinct subcellular localizations. We conclude that alternative splicing of mammalian USP4 is under selective maintenance and that long and short USP4 isoforms may target substrates in various cellular compartments.

Origin of exon skip**-rich transcriptomes in animals driven by evolution of gene architecture

Article Open access 17 September 2018

Domestication of self-splicing introns during eukaryogenesis: the rise of the complex spliceosomal machinery

Article Open access 01 December 2017

The unusual gene architecture of polyubiquitin is created by dual-specific splice sites

Article Open access 24 January 2024

Introduction

Ubiquitin-specific protease 4 (USP4) is a deubiquitinating enzyme that can edit or remove ubiquitin chains of various topologies. USP4 can remove both degradative K48- and regulatory K63-linked ubiquitin chains from a growing list of protein targets that include key players in a number of signal transduction pathways. Its substrates include the transforming growth factor-β (TGF-β) I receptor^1,2 and the TGF-β-activated kinase 1 (TAK1)³, Wnt/β-catenin pathway transcription factor Nemo-like kinase (NLK)⁴, the p53 antagonist ARF-BP1⁵, anti-viral response mediator RIG-I⁶, TNFα/NF-κB inflammatory pathway mediators TRAF-2 & -6

Results

Bioinformatic analysis

To study exon-intron architecture and splice signal conservation, we downloaded GenBank sequences for well-annotated USP4 sequences for 62 vertebrate species covering major vertebrate taxa (see Supplementary Table S1).

Sequence and length conservation of USP4 exons

Both long and short isoforms of USP4 comprise a bi-partite catalytic domain and a regulatory DUSP-UBL1 domain, where the seventh exon forms part of the unstructured flexible linker between these two (Fig. 1B). Sequence identity (Fig. 1C) and length (Fig. 1D) of USP4 exons are highly conserved for exons that encode structured functional domains and less well for exons that correspond to unstructured regions. E₇ exhibits greater length variation than its neighbors. Relative to mammals, E₇ of birds and fish generally encodes one and two additional amino acids respectively, while the length of E₇ within each of these clades is variable; this suggests multiple indel events in USP4 during the evolutionary diversification of vertebrates. Sequence conservation among aligned USP4 exons, quantified using Shannon entropy in Fig. 1C, reveals that E₇ is located within a highly variable region. The entropic nature of E₇ conforms to the general pattern that alternatively spliced exons exhibit less sequence conservation³⁰. In Fig. 2A, we have derived a more sensitive D_to3′.opt of 20–40 nt based on the locations of consensus BPS sequences (YURAY) within all 15770 introns of human chromosome 22 protein-coding genes. This slightly larger window for D_to3′.opt will reduce the false negative rate in detecting true BPSs in the neighbouring introns of E₇. If a strong BPS is absent from the upstream intron (I₆), a strong downstream BPS within I₇ could lead to exon skip** (Fig. 1A). We therefore evaluated the relative strengths of USP4 BPS₆ and BPS₇ for 14 well-studied species representing the major vertebrate taxa.

The location of YURAY motifs in I₇ (Fig. 2B) suggests a strong BPS₇ among all studied mammalian species. A strong BPS₇ is also present in I₇ of zebrafish (Danio rerio), but is missing in the frog (Xenopus tropicalis), the chicken (Gallus gallus) and the Chinese turtle (Pelodiscus sinensis). The latter two species retain candidate YURAY motifs which can be eliminated from consideration due to their greater distance from the 3′SS (73nt for the chicken, 85 nt for the Chinese turtle) and to the presence of a downstream non-3′SS AG (non-3′SS AG dinucleotides between a BPS and a 3′SS AG are avoided as the first AG following the branch-point is generally used as the 3′SS AG).

YURAY motifs are absent from the 40 last nucleotides of USP4-I₆ in all mammalian species. Taken together, mammalian species in general have a weak BPS₆ and a relatively strong BPS₇. This lends plausibility to the first hypothesis from Fig. 1A, wherein E₇ splicing results from alternative pairing of 5′SS₆ and BPS₇ (E₇ skip**) or of 5′SS₇ and BPS₇ (E₇ inclusion). In contrast, non-mammalian species tend to have a strong BPS6 (except for the Chinese turtle). YURAY motifs are located near D_to3′.opt in I₆ of the chicken (D_to3′ = 47 nt) and zebrafish (D_to3′ = 16 nt) and frog (three YURAY motifs with D_to3′ = 36, 42 and 53 nt, respectively). This suggests that E₇ skip** should be less likely in non-mammalian than in mammalian species, though the Chinese turtle may be an exception.

Relative strengths of proximal and distal splice signals

In addition to a weak BPS₆ relative to BPS₇, a stronger 5′SS₆ than 5′SS₇ and/or stronger 3′SS₇ than 3′SS₆ would favor E₇ skip** (Fig. 1A(i)). Position weight matrices (PWM) measures site-specific nucleotide usage bias in a motif alignment, where the consensus motif typically has the highest PWM score (PWMS). PWMSs are routinely used to characterize the signal strength of splice sites^31,32.

A PWM for the 5′SS sequences of human chromosome 22 introns (Table 1) shows a consensus 5′SS consistent with what has been documented in the literature, i.e., a core motif of AG|GUAAGU, where “|” indicates the exon-intron junction. PWMs derived from zebrafish and chicken chromosomes are almost identical to the human 5′SS matrix in Table 1(upper panel). Thus, this PWM can be used to generate PWMSs as comparable measures of signal strength at 5′SS₆ and 5′SS₇ from the 14 representative vertebrate USP4 sequences, where significantly larger scores indicate stronger splice signal strength. Given Table 1(upper panel), the maximal PWMS is 16.9. PWMSs are consistently larger for 5′SS₆ than 5′SS₇ for the ten mammalian USP4 sequences (Table 1(lower panel)), with the mean PWMS being 9.2314 for 5′SS₆ and 5.455 for 5′SS₇ (t = 19.759, df = 11, p < 0.0001, paired-sample t-test). This lends support for the scenario depicted in Fig. 1A(i), where a stronger 5′SS₆ favors E₇ skip** in mammals. In constrast, for chicken and zebrafish, PWMS₆ is greater than PWMS₇ while Pelodiscus sinensis again conforms to the mammalian pattern (PWMS₆ > PWMS₇).

Table 1

Full size table

There were no significant differences in flanking 3′SS PWMSs (as might be expected since 5′SS and BPS are most important for determining exon-intron boundaries³³).

Length of the introns flanking exon 7

Because exons flanked by long introns tend to be skipped during the splicing process^22,25, which would suggest that the the potential contribution of the long I₆ to E₇ skip** in mammals should not be ignored.

Experimental tests of alternative hypotheses

Determination of splicing mechanism

Although our results are consistent with the hypothesis that E₇ skip** results from BPS₇ strengthening relative to BPS₆ and 5′SS₆ stronger than 5′SS₇, they do not exclude the possibility that the longer intron I₆ or other potentially complicating factors in USP4 pre-mRNA may contribute to E₇ skip**. To test the hypothesis that E₇ skip** is due to differential splice signals at 5′SS and BPS, we created a minigene construct by inserting the human USP4 genomic sequence encompassing E₇ (together with 75 nt at the 3′ tail of I₆ and 51 nt at the 5′ end of I₇) into the well characterized splicing reporter pXJ41 (the generous gift of Dr. Sushma Grellsheid, Durham University). In the resulting minigene the human E₇ genomic fragment resides in the second intron of the rabbit beta hemoglobin gene (Fig. 4A). If E₇ skip** is due to the long upstream I₆, then we should observe no E₇ skip** in this construct with short upstream intron. The minigene mimics the scenario in Fig. 1A(i) with the differential signal strength of splice sites: 5′SS_a (Fig. 4A) is stronger (PWMS = 7.5358) than 5′SS₇ (PWMS = 5.0898) and BPS_b (Fig. 4A) is stronger than BPS6, with the former having a CUAAC (YURAY) sequence located 35 nt from the 3′ end of the intron and the latter having no YURAY at D_to3′.opt. If E₇ skip** in its natural USP4 mRNA is due to such differential strength of splice signals, then we should observe E₇ skip** in the minigene mRNA.

When the minigene was expressed in human U2OS osteosarcoma cells, RT-PCR analysis using primers upstream and downstream of the rabbit exons revealed two isoforms of the size predicted for retention and exclusion of E₇ (Fig. 4B). Similar results were obtained in the unrelated HeLa cell line (not shown). This finding excludes exon skip** as a consequence of intron length but is consistent with E₇ skip** due to the differential strength of splice signals. We therefore explored the contributions of the BPS₆ and 5′SS₇ elements by site-directed mutagenesis of the E₇ minigene construct, inserting a consensus YURAY sequence at D_to3′.opt in I_a (upstream of E₇) and/or engineering an optimized 5′SS₇ element in I_b (depicted as BP and SS respectively in Fig. 4C). The mutated versions of the minigene were transfected into U2OS cells and RT-PCR analysis was performed as before. Whereas the introduction of the consensus YURAY sequence had no effect on the ratio of exon retained and exon excluded products, the latter was undetectable in RNA isolated from cells transfected with minigenes in which the 5′SS₇ element had been optimized (Fig. 4D).

Phylogenetic distribution of USP4 alternative splicing

In contrast to mammalian species, 5′SS₆s of chicken and zebrafish USP4 are weaker than 5′SS₇s, as indicated by their PWMS values (Table 1). To verify whether (as would be predicted) E₇ skip** does not occur in such species, RT-PCR analysis was performed on RNA isolated from primary chick fibroblast cultures (the generous gift of Dr. J. S Diallo, Ottawa Hospital Research Institute). Primers corresponding to sequences in exons 6 and 8 were used to detect the presence or absence of the seventh exon as depicted in Fig. 5A. We detected only the exon-retained version of the transcript (Fig. 5B). However, when the human minigene was introduced into chick embryo fibroblasts by transfection both isoforms were detected (Fig. 5C). The absence of exon skip** in the chicken cells could thus be directly attributed to the primary sequence of the chicken USP4 pre-mRNA. Our data exclude the possibility that E₇ retention occurs in the chicken as a consequence of an altered repertoire of splicing factors in avian versus mammalian cells (see Discussion). By similar logic we predict that exon skip** would not occur in the zebrafish gene; RT-PCR analysis of RNA from larval stage zebrafish (the generous gift of Dr. Marc Ekker, University of Ottawa) confirmed the presence of a single exon-retained isoform (Fig. 5D). In support of this, performing a BLASTn of USP4 exons 5–13 against the 600,432 chicken EST sequences recovered four sequences with E₇ but no sequence without E₇. Among the 1,488,339 zebrafish ESTs, seven have E₇ but none are without E₇. In contrast, searching the 8,704,868 human ESTs recovered five sequences with and seven without E₇. The corresponding numbers from the 4,853,570 mouse EST sequences are 20 and 8, respectively. Our conceptual framework based on the relative strengths of 5′ splice signals thus correctly predicted splicing propensity in these model organisms, confirmed by both database and experimental analyses.

As is shown in Fig. 4D, the optimization of three nucleotides in the 5′ splice site downstream of USP4-E₇ according to the consensus sequence, namely −3G → C, −2G → A and +6A → T, proved sufficient to eliminate exon skip** in the human USP4 minigene. Among species observed in Table 1, the nine therian mammal 5′SS₆s feature optimal nucleotides −2A and +6T while suboptimal −3G, −2G and +6A penalize the 5′SS₇ PWMSs of all members of this lineage. These residues are identical in the Chinese turtle and are likely thus responsible for the observed alternative E₇ skip** in this distant relative. In contrast, both flanking splice sites of E₇ in zebrafish feature optimal nucleotides (5′SS6: −3A, −2A, +6T; 5′SS7: −3C, −2A, +6C), which preclude E₇ exclusion. Curiously, in chicken, these determinant nucleotides are identical to those of mammals which produce E₇ skip** with the exception of the 5′SS₆ +6N site, which is weak (+6A). The upstream and downstream 5′SSs in chicken, though weak, are equivalent and prevent exon skip** as in zebrafish. The +6N site may thus be the discriminant factor in E₇ skip** propensity. To verify this, we expanded the scope of our analysis to include all sequenced genomes bearing USP4 to see whether the 5′SS mismatching (in particular +6 site mismatching) predicts splicing proclivity. While direct expansion of our analytical framework is limited by insufficient EST data and biological sample unavailability, we can infer splicing patterns from RNA-seq datasets. Similar to the methodology used for EST mining, we performed a BLASTn of available RNA-seq data from the Sequence Read Archive (SRA) using the USP4 coding sequence with E₇ removed as a query. In the absence of hits crossing the exon 6–8 boundary for multiple, sufficiently large expression datasets, species were deemed to forgo short isoform production.

Figure 6A summarizes USP4 splicing patterns in a phylogenetic context with corresponding flanking 5′SS sequence logos indicated. According to the PWM in Table 1A, +6A and +6G weaken the 5′SS while +6T is optimal and +6C is neutral (weighted consensus illustrated in Fig. 6C). For all tetrapods, when the downstream +6 site is stronger than the upstream +6 site, there is alternative splicing of E₇. This correlation is particularly apparent in the avian phylum: chicken and turkey have no E₇ skip** (+6A; +6A), all other birds either exhibit skip** (+6C/T; +6A) or loss of E₇. What is more, some members of sister taxa have lost the ability to produce the long isoform: E₇ is deleted in Corvus brachyrhynchos but present in Corvus cornix cornix; absent from Adelie penguin but present in Emperor penguin, for example. In contrast to this substantial variability, all mammals retain an optimal E₇ skip** configuration, +6T; +6A (with the exception of the clade root: platypus USP4 has +6G; +6A and, consistent with our model, does not undergo skip**). In theory, many nucleotides substitutions could disrupt the 5′SS if alternative splicing were the result of drift. Since the same splice site configuration is maintained throughout 220 million years of mammalian evolution there may be selection for this particular +6 configuration. In Fig. 6C and Supplemental Figure 1, we show the effects of downstream +6 site point mutation from native +6A to +6T, +6C and +6G in human and mouse cell lines. In each case, the splicing propensity changed in direct relation with the estimated fitness in our PWM: alternative splicing was nearly eliminated in +6T, slightly reduced in +6C and increased in +6G. Therefore, we propose that the highly conserved +6A site within 5′SS₇ is under natural selection to maintain both long and short USP4 isoforms in therian mammals.

Differential localizations and roles of spliced isoforms

The evidence supporting alternative splicing selection in mammalian USP4 is strong; we would consequently expect the two isoforms to have distinct cellular roles. Indeed, we observed distinct subcellular localizations of long and short USP4 isoforms in both single- and double-transfections of HeLa, U2OS, 293T and 3T3 cells (see Fig. 6D). While the short isoform was distributed throughout the cell, the localization of the long isoform was largely cytoplasmic in most if not all cells in the four cell lines examined. Potential implications of this observation are discussed below. Altogether, our results suggest that the two major USP4 isoforms generated by alternative skip** of its seventh exon may not be functionally redundant as previously suggested.

Discussion

It has been proposed that mutations that weaken the 5′ splice site are responsible for the evolutionary shift from constitutive to alternative splicing in many vertebrate genes, as reviewed in Keren et al.²¹ and compelling evidence has been presented in support of this hypothesis³⁴. While most minor splice variants are attributable to noisy splicing^35,36, USP4 constitutes a rare case wherein selective pressure acts to conserve differential 5′SS strengths leading to exon skip** in therian mammals. The approach we presented here focuses on these cis-acting splice sites, which offers a more basic but more direct framework towards understanding the splicing code. 5′ splice sites can recruit trans-acting alternative splicing factors for intrinsic splice regulation. For example, deleterious exon skip** in survival of motor neuron (SMN) pre-mRNA can be attributed to recruitment of splice repressor U2AF65 by a weak downstream 5′SS³⁷. While trans-acting factors interacting with the 5′SSs of USP4 may similarly regulate E₇ skip**, our model explains USP4-E₇ splicing propensity independent of other cis-regulatory sequences such as exonic splice enhancers (ESEs), which may or may not be selectively co-optimized in USP4 alternative splicing. Our study also highlights the importance of experimental verification of alternative hypotheses. Although the bioinformatics framework alone cannot distinguish between the two mechanisms proposed in Fig. 1(i,ii), the experimental results demonstrate that relative 5′SS strengths are far better predictors of alternative splicing than BPSs or upstream intron lengths. Further, our combinatorial in silico and experimental approach identified the +6 site within the 5′SS as the splicing discriminant. Intronic +6 site mutations have been reported as splicing instigators in other genes such as SMN1^38,39. E₇ skip** in SMN1 leads to spinal muscular atrophy (SMA) and SNPs that cause this aberrant skip** have been identified in patients at the downstream 5′SS₇ at the +6 site (+6T → G). SMN1 has a very close paralog, SMN2, that is incapable of rescuing SMN1 deficiency in SMA because its E₇ is also skipped due to a WT nucleotide variant, 5′SS₇ +6G. Thus, +6T at the downstream 5′SS₇ of SMN1/2 promotes upstream exon inclusion while +6G promotes near-total upstream exon skip**. In mammalian WT USP4, 5′SS₆ has +6T while 5′SS₇ has +6A. As reflected in the PWM in Table 1 and observed in Fig. 6C, the strengths of +6 site nucleotides are predicted to be as follows: T > C ≥ A > G, where a stronger 5′SS₆ +6 relative to 5′SS₇ +6 correlates with splicing proclivity. It is curious that +6A ≠ +6G and that the former was selected as the weak downstream nucleotide of USP4. A plausible mechanism for +6A-dependent alternative skip** may involve U1C, a component of the U1 snRNP that preferentially recognizes a 5′SS motif with +6A, GTATAA⁴⁰ and can interact with splicing regulator TIA-1^41,42 to promote exon retention, for example in SMN2^43,44. Several genes undergo U1C-dependent alternative splicing^45,46. Based on linear changes in the relative abundances of the short and long isoforms observed during differentiation of P19 embryonic carcinoma cells (Gray, unpublished) we postulate that E₇ of USP4 may be subject to regulated alternative splicing in therian mammals.

Retention or exclusion of the amino acids encoded by exon 7 does not affect the protease activity of the USP4 enzyme (using a synthetic substrate¹⁴) and the ubiquitin-exchange regulatory mechanism proceeds equally in both USP4 isoforms¹⁶. We have nonetheless shown that there is selection for alternative splicing maintenance in mammalian USP4. Establishing the molecular selection driver should be highly informative. We show that the two isoforms display distinct subcellular localizations, which suggests that (1) propensity and/or (2) capacity for substrate interactions may differ. First, vital cytoplasmic (e.g. TGF-β pathway¹) and nuclear (e.g. spliceosomal^11,12) substrates have been reported for USP4 (isoforms specificities not declared). Long and short USP4 isoform production may be advantageous for simultaneous, collective targeting of key substrates in various cellular compartments. On the other hand, the two isoforms almost certainly have some distinct interactors. Cytoplasmic retention of USP4-long is mediated by phosphorylation of a ubiquitously conserved serine (S445); the apparent absence of this regulation in USP4-short may reflect a lack of phosphorylation by Akt². Exon 7 of mammalian USP4 is serine-rich (16 out of 47 residues; human sequence: RSSTAPSRNFTTSPKSSASPYSSVSASLIANGDSTSTCGMHSSGVSRG) and also contains six constitutively charged residues (five positively charged and one negatively charged). While the +4 net charge difference between isoforms likely affects substrate interactions, the serine-rich exon also retains multiple phosphorylation sites (underlined) that are conserved among all mammals. USP4-E₇ may be phosphorylated in conjunction with Ser445 for nuclear exclusion or may be required for interaction with Akt and other substrates. For instance, SART3, a spliceosomal factor and deubiquitination target of both USP4¹² and its paralog USP15⁴⁷, has been reported to interact with serine-rich (not to be confused with serine/arginine-rich) domains of proteins⁴⁸. Interestingly, USP15 contains an analogous, serine-enriched alternatively spliced seventh exon (SPGASNFSTLPKISPSSLSNNYNNMNNR; reported phosphorylation sites underlined). Splice boundaries and amino acid sequence differ between E₇ of USP4 and USP15, suggesting that alternative splicing arose independently in these, though they both maintain significant proportions of serines. This may be a case of stabilizing selection acting on clusters of phosphorylation sites⁴⁹. There may be an important feedback loop involving the splicing and subsequent localization of USP4 and USP15, two DUBs that critically interact with the spliceosome. Long and short USP4 production and thus DUB modification of isoform-specific substrates may differ across tissue types. It remains to be seen whether such isoform-specific substrates drove evolutionary conservation of the dual isoforms within placental mammals.

To summarize, we have shown that the long and short isoforms of USP4 have distinct properties and their contributions to cellular networks should be considered separately. Most proteins have more than one reported isoform and though most may be considered non-essential noise, distinct functional variants, such as in USP4, must not be grouped as one protein. The roles of all significantly expressed minor splice variants should be studied more carefully.