Regulation of Multiple Core Spliceosomal Proteins by Alternative Splicing-Coupled Nonsense-Mediated mRNA Decay

Abstract

Alternative splicing (AS) can regulate gene expression by introducing premature termination codons (PTCs) into spliced mRNA that subsequently elicit transcript degradation by the nonsense-mediated mRNA decay (NMD) pathway. However, the range of cellular functions controlled by this process and the factors required are poorly understood. By quantitative AS microarray profiling, we find that there are significant overlaps among the sets of PTC-introducing AS events affected by individual knockdown of the three core human NMD factors, Up-Frameshift 1 (UPF1), UPF2, and UPF3X/B. However, the levels of some PTC-containing splice variants are less or not detectably affected by the knockdown of UPF2 and/or UPF3X, compared with the knockdown of UPF1. The intron sequences flanking the affected alternative exons are often highly conserved, suggesting important regulatory roles for these AS events. The corresponding genes represent diverse cellular functions, and surprisingly, many encode core spliceosomal proteins and assembly factors. We further show that conserved, PTC-introducing AS events are enriched in genes that encode core spliceosomal proteins. Where tested, altering the expression levels of these core spliceosomal components affects the regulation of PTC-containing splice variants from the corresponding genes. Together, our results show that AS-coupled NMD can have different UPF factor requirements and is likely to regulate many general components of the spliceosome. The results further implicate general spliceosomal components in AS regulation.

The production of multiple mRNA variants through alternative splicing (AS) represents a widespread mechanism for the expansion of proteomic diversity, and regulated AS plays important roles in many physiological processes (for reviews, see references 4, 5, and 34). However, sequence-based predictions have revealed that approximately one-third or more of AS events have the potential to introduce a premature termination codon (PTC) that could target the resulting spliced transcript for nonsense-mediated mRNA decay (NMD) (32). While many predicted PTC-introducing AS events do not appear to be conserved or subject to regulation by NMD, AS microarray profiling experiments have revealed that approximately 10 to 20% of these events display substantial changes when NMD is disrupted (42). AS microarray profiling (40, 42) thus provides a basis for identifying new genes and functional processes regulated by AS-coupled NMD and for elucidating the factor requirements for this mode of gene regulation on a large scale.

In mammals, NMD is generally, though not always (6, 35), dependent on a splicing event sufficiently downstream of a termination codon. Termination codons are usually recognized as premature when they occur more than 50 to 55 nucleotides upstream of a final splice junction (39), which is “marked” by the deposition a postsplicing exon junction complex (EJC) (30). NMD is believed to require three core UPF factors which are conserved from yeast to humans, namely, UPF1 (also known as regulator of nonsense transcripts 1 [RENT1]), UPF2 (RENT2), and UPF3. In mammals, UPF2 and presumably one of the two UPF3-related proteins, UPF3 and UPF3X (also known as UPF3A and UPF3B, respectively), associate with spliced mRNA via the EJC (23, 29). UPF1 is a 5′-to-3′ helicase and RNA-dependent ATPase (3) that interacts with UPF2 and with translation release factors (22). Additional conserved NMD factors include, among others (33), SMG1 (suppressor with morphogenetic effect on genitalia 1), SMG5, SMG6, and SMG7, which are implicated in the regulated phosphorylation of UPF1 (for reviews, see references 1, 9, and 21).

Consistent with initial genetic and biochemical evidence that UPF1, UPF2, and UPF3 act in a single, linear pathway, microarray profiling studies of yeast (18, 31) and Drosophila cells (46) have revealed the differential expression of a common set of transcripts when each of these UPF factors is inactivated or depleted. In contrast, more recent studies with human cells have provided evidence for alternative UPF1-dependent branches of the NMD pathway that function independently of either UPF2 or UPF3X (8, 15). However, the factor requirements for AS-coupled NMD are not well understood.

The range of cellular functions regulated by AS-coupled NMD has also not been fully elucidated. There is evidence of an autoregulatory role for AS-coupled NMD in the expression of AS factors (17, 52, 56), ribosomal proteins (11, 12, 37), and other genes (14, 20, 27). In addition, PTC-introducing AS events in genes that encode AS factors, including members of the hnRNP and SR families of splicing factors, have been associated with highly conserved genomic sequences, implying important regulatory roles for these events (28, 40).

In the present study, we have extended the use of AS microarray profiling to compare the effects of small interfering RNA (siRNA)-mediated knockdown of UPF1, UPF2, and UPF3X. Our results reveal that these core NMD factors have overlapping but distinct effects on the degradation of PTC-containing transcripts arising from AS, indicating that AS-coupled NMD may be regulated by different UPF1-dependent mechanisms. Unexpectedly, among the functionally diverse group of genes containing exons regulated by AS-coupled NMD are multiple genes that encode core spliceosomal components, including snRNP-associated proteins and spliceosome assembly factors. Our results therefore provide evidence that AS-coupled NMD plays a much wider role in the regulation of the splicing machinery than previously appreciated. Consistently, expression of the core snRNP SmB/B′ (SNRPB) or the assembly factor SPF30 (SMNDC1) affects the levels of PTC-containing variants from the corresponding genes. The results thus also provide new evidence suggesting that, in addition to well-established roles in constitutive splicing, core spliceosomal proteins can function in the regulation of AS.

MATERIALS AND METHODS

Cell culture, siRNA, and plasmid transfection.

HeLa cell culture and transient siRNA transfection were performed as previously described (42). Control, UPF1, UPF2, and UPF3X siRNAs were also as previously described (24). For transient protein overexpression, cDNAs that encode SNRPB, SMNDC1, and DDX42 from the human ORFeome collection (Open Biosystems) were cloned into pMT3989 (Marcia Roy, M. Tyers laboratory) with Gateway LR Clonase II (Invitrogen). The resulting N-terminally Flag₃-tagged constructs were transfected into HeLa cells with Lipofectamine 2000 (Invitrogen) or Fugene6 (Roche Applied Science) according to the manufacturer's instructions.

RT-PCR assays and Western blotting.

Reverse transcription (RT)-PCR assays with primers specific for constitutive exons flanking the alternative exon (see Fig. 2) were performed as previously described (42), and Western blotting for UPF factors was also performed as previously described (24). RT-PCR assays (see Fig. 5) with one primer specific for the alternative exon and a second primer specific for either the 5′ untranslated region or another coding exon were performed as described above, with the following changes. Ten nanograms of input total cell RNA (isolated with the RNeasy Mini kit; Qiagen) was used, and ethidium bromide-stained bands were quantified with QuantityOne (Bio-Rad). Total cell protein lysates in radioimmunoprecipitation assay buffer were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis. Western blotting (see Fig. 5) was performed with anti-FlagM2 (F3165; Sigma) or anti-α-tubulin (T6074; Sigma) as a loading control.

FIG. 2.

Representative RT-PCR assays showing effects of UPF protein knockdown on levels of PTC-introducing alternative exons. RT-PCR assays were performed with primers specific for the constitutive exons flanking the alternative exons. The RNA samples analyzed are indicated above the lanes and correspond to cells transfected with either a control siRNA (−) or an siRNA specific for the indicated UPF factor (+). Arrows to the right of each gel indicate the expected sizes of included and skipped products. The color bar below each gel panel shows the quantification of the knockdown-dependent changes in percent skipping predicted by the microarray (upper color) and measured by RT-PCR (lower color). The change in percent skipping is shown with the same color scale as in Fig. 1B. For details on these AS events, see Table S2 in the supplemental material (PTC-upon-inclusion AS events 3, 146, 957, 557, 2315, and 2372; PTC-upon-skipping AS events 750, 1268, 76, 188, 1909, and 2914).

FIG. 5.

SNRPB (also known as SmB/B′) or SMNDC1 (also known as SPF30) overexpression leads to increased levels of the respective PTC-containing (PTC+) alternative transcript. Either Flag-tagged SNRPB or SMNDC1 was transiently overexpressed in HeLa cells, and the endogenous PTC-containing transcript of SNRPB (A, left top panel) or SMNDC1 (A, right top panel) was amplified by RT-PCR with a forward primer specific for the 5′ untranslated region or an upstream exon and a reverse primer specific for the PTC-introducing alternative exon. RT-PCR amplification of conserved, PTC-containing transcripts of SF1, TRA2A, and SR140 were not affected to the same degree by the overexpression of SNRPB (A, left lower panels) or SMNDC1 (A, right lower panels). The overexpression of another Flag-tagged spliceosome-associated protein (DDX42) in parallel had little effect on the level of the SNRPB and SMNDC1 transcripts. (B) Western blot assay showing the expression of each Flag-tagged protein (left, approximate molecular sizes of markers in kilodaltons; right, arrowheads indicate expected protein sizes). Data are representative of three independent transfections, and RT-PCR assays were performed in triplicate.

Microarray design and hybridization.

Cassette AS events were identified in cDNA and expressed sequence tag (EST) sequences and filtered as previously described (41, 43). Probes representing 3,055 human AS events (of which 2,923 were unique) were designed as previously described (19, 43) and printed on a 22K microarray (Agilent Technologies). Cy3- and Cy5-labeled cDNAs were prepared from poly(A)⁺ RNA in duplicate by fluor reversal and hybridized to Agilent microarrays as previously described (19, 43). For the UPF1 knockdown, the same samples and microarray data described in reference 42 were reanalyzed.

Microarray data analysis.

Microarrays were scanned, and signal intensities were detrended and normalized as previously described (58). Relative inclusion levels (percent skipping) and confidence ranks of alternative exons monitored on the microarray were determined with the “Generative model for the Alternative Splicing Array Platform” (GenASAP) algorithm (48). For this study, analysis was limited to events with confidence ranks in the top half of the data in at least two of the six samples (three control siRNA and three knockdown siRNA), representing 1,704 of the 2,923 unique AS events on the microarray (see Table S1 in the supplemental material).

Annotation of PTC-introducing AS events.

Open reading frame (ORF) information for genes represented on the microarray was obtained from the header line of the corresponding cDNA sequences in UniGene (ftp://ftp.ncbi.nih.gov/repository/UniGene/). AS events were then separated into categories as follows (counts are given for the entire array, followed by the count for “detectable” events in parentheses): of 2,171 (1,360) unique events mapping to a cDNA with a known ORF, 434 (249) were excluded from further PTC analysis because the alternative exon was upstream of, or overlapping with, the start codon or because the termination codon was not located in the last exon. Termination codons were mapped to the ORF sequences in relation to the splice junctions for the remaining 1,737 (1,111) events. After the application of the “50-nucleotide rule” (39), 735 (495) of the AS events had potential PTCs. Of these, 282 (164) were PTC-upon-inclusion events and 453 (331) were PTC-upon-skipping events (see also reference 42). Gene identifiers, aliases, and gene ontology annotation (see Tables 1 and 2; see Table S2 in the supplemental material) were obtained manually or with SOURCE (http://source.stanford.edu) (13) or Gene/Clone ID Converter (38).

TABLE 1.

Selected PTC-introducing AS events in genes with functions related to RNA processing^a

NMD feature and RNA processing function	Gene name/alias(es)	AS event no.	Conserved AS in mouse?	No. of nucleotides overlapping PCb (upstream/downstream)
PTC upon inclusion
Spliceosome assembly	SMNDC1/SMNrp/SPF30	2372		150/69
Sm core	SNRPB/SmB/B′	957	Yes	80/83
U1-snRNP component	SNRP70/U1-70K	951	Yes	120/75
U2-snRNP component	SF3B1/SF3B155/SAP155	3218	Yes	150/150
Second-step catalysis	PRPF18/hPrp18	557		150/125
Other RNA metabolism	EIF4A2	909	Yes	142/115
Other RNA-binding	RBM18	2315		120/134
	IVNS1ABP/NS1BP	3	Yes	150/96
SR-related family	TRA2A	146	Yes	150/150
SR domain containing	BCLAF1/BTF	1624		150/150
	LUC7L	879	Yes	131/150
PTC upon skipping
Branch point recognition	SF1/ZNF162	188	Yes	150/124
U2-snRNP component	DDX42/SF3B125	2914	Yes	115/138

^aThese alternative exons show changes in inclusion level upon UPF knockdown(s) and are flanked by highly conserved intron sequences.

^bPC, phastCons conserved elements (49).

TABLE 2.

Conserved, PTC-introducing AS events identified in transcripts from spliceosome-associated proteins^a

Functional group and name/alias	Description	PTC upon inclusion	PTC upon skipping
U1 + U2 snRNP
SF3B1	Splicing factor 3b, subunit 1, 155 kDa	X
SF3B3	Splicing factor 3b, subunit 3, 130 kDa	X
SNRP70	Small nuclear ribonucleoprotein 70-kDa polypeptide (RNP antigen)	X
SR140	U2-associated SR140 protein	X	X
U11 + U12 snRNP, C16ORF33	U11/U12 snRNP 25K protein	X
Sm, SNRPB	Small nuclear ribonucleoprotein polypeptides B and B1	X
U2AF
U2AF2/U2AF65	U2 small nuclear RNA auxiliary factor 2	X
U2AF1L3	U2 small nuclear RNA auxiliary factor 1-like 4	X
DEAD
DDX17/p72	DEAD (Asp-Glu-Ala-Asp) box polypeptide 17	X
DDX26B	DEAD/H (Asp-Glu-Ala-Asp/His) box polypeptide 26B	X
DDX39/Urh49	DEAD (Asp-Glu-Ala-Asp) box polypeptide 39	X	X
DDX46/Prpf5	DEAD (Asp-Glu-Ala-Asp) box polypeptide 46	X
DHX35	DEAH (Asp-Glu-Ala-His) box polypeptide 35		X
DHX8/Prp22	DEAH (Asp-Glu-Ala-His) box polypeptide 8		X
CELF/CUG-BP
BRUNOL4	Bruno-like 4, RNA-binding protein (Drosophila)	X	X
CUGBP1	CUG triplet repeat, RNA-binding protein 1		X
CUGBP2	CUG triplet repeat, RNA-binding protein 2		X
NOVA, NOVA1	Neuro-oncological ventral antigen 1	X
CLK
CLK1	CDC-like kinase 1		X
CLK3	CDC-like kinase 3		X
CLK4	CDC-like kinase 4	X	X
SR
SFRS2/SRp30b	Splicing factor, arginine/serine-rich 2	X
SFRS3/SRp20	Splicing factor, arginine/serine-rich 3	X
SFRS4/SRp75	Splicing factor, arginine/serine-rich 4	X
SFRS6/SRp55	Splicing factor, arginine/serine-rich 6	X
SFRS7/9G8	Splicing factor, arginine/serine-rich 7, 35 kDa	X
SFRS10/Tra2b	Splicing factor, arginine/serine-rich 10 (Tra2 homolog, Drosophila)		X
SFRS11/p54	Splicing factor, arginine/serine-rich 11	X
SFRS12/SRrp86	Splicing factor, arginine/serine-rich 12	X
hnRNP
CIRBP	Cold-inducible RNA-binding protein		X
HNRPD/AUF1	Heterogeneous nuclear ribonucleoprotein D/AUF1	X
HNRPDL	Heterogeneous nuclear ribonucleoprotein D-like	X
HNRPH1	Heterogeneous nuclear ribonucleoprotein H1 (H)		X
HNRPLL	Heterogeneous nuclear ribonucleoprotein L-like	X
HNRPM	Heterogeneous nuclear ribonucleoprotein M		X
HNRPR	Heterogeneous nuclear ribonucleoprotein R	X	X
HNRPU/SAFA	Heterogeneous nuclear ribonucleoprotein U		X
MSI2	Musashi homolog 2 (Drosophila)	X
MYEF2	Myelin expression factor 2		X
PCBP3	Poly(rC)-binding protein 3		X
PTBP1	Polypyrimidine tract-binding protein 1		X
PTBP2/nPTB	Polypyrimidine tract-binding protein 2		X
RBM3	RNA-binding motif (RNP1, RRM) protein 3	X	X
ROD1	ROD1 regulator of differentiation 1 (Schizosaccharomyces pombe)		X
Other
ACIN1	Apoptotic chromatin condensation inducer 1	X	X
ARS2	Arsenate resistance protein 2		X
C21ORF66	GC-rich sequence DNA-binding factor candidate		X
CDC40/PRP17	Cell division cycle 40 homolog (Saccharomyces cerevisiae)	X
CROP/LUC7A	Cisplatin resistance-associated overexpressed protein	X	X
CTNNBL1/NAP	Catenin, beta-like 1	X
DNAJC8/SPF31	DnaJ (Hsp40) homolog, subfamily C, member 8		X
ET/MFSD11	Major facilitator superfamily domain-containing 11	X
FUBP3	Far upstream element (FUSE)-binding protein 3		X
FUS	Fusion [involved in t(12;16) in malignant liposarcoma]		X
ILF3/DRBF	Interleukin enhancer-binding factor 3, 90 kDa		X
KIAA1429	KIAA1429		X
PPIL2	Peptidylprolyl isomerase (cyclophilin)-like 2		X
PRPF18	PRP18 pre-mRNA processing factor 18 homolog (S. cerevisiae)	X
RBM15	RNA-binding motif protein 15	X
RBM5	RNA-binding motif protein 5	X	X
RDBP	RD RNA-binding protein	X	X
SF1	Splicing factor 1	X	X
SIAHBP1/PUF60	Poly(U)-binding splicing factor, 60 kDa		X
SSB/La	Sjogren syndrome antigen B (autoantigen La)		X
TAF15	TATA box-binding protein (TBP)-associated factor, 68 kDa	X	X
TCERG1/CA150	Transcription elongation regulator 1	X
TIA1	TIA1 cytotoxic granule-associated RNA-binding protein	X	X

^aEach AS event either has at least 35 out of 150 nucleotides of flanking intron overlapping conserved nucleotides (phastCons) (49) or is conserved between human and mouse based on cDNA/EST analysis.

Analysis of conservation of flanking intron sequence and conserved AS.

Genomic coordinates of microarray-profiled alternative exons, and the 150 nucleotides of intronic sequences flanking these exons, were determined by aligning the exon sequences to the human genome (hg18, March 2006 assembly, University of California, Santa Cruz, genome browser at http://genome.ucsc.edu/) with BLASTN. Conserved elements identified by the phastCons algorithm (49) were retrieved from the University of California, Santa Cruz, human genome browser “phastConsElements17way” table of the “17-Way Most Conserved” track (25). The overlap of the flanking intron sequences with these elements was determined with the Galaxy web application (16). The conservation of the AS events in mouse ESTs was determined as previously described (43).

Identification of AS events in spliceosomal and control gene sets.

Unigene identifiers for a curated list of spliceosome-associated proteins (2) were retrieved with MatchMiner (7) or manually. To generate the “control” gene set, for each spliceosome-associated factor with n sequences in its Unigene cluster, all other Unigene clusters with n sequences were grouped. One cluster, other than the one associated with the spliceosomal factor, was then randomly selected from this group to be in the control set (or if no other such clusters were found, from the next group with n + 1 sequences). Cassette AS events were mined as previously described (41, 43), and PTCs were annotated as described above and in reference 42.

Statistical analysis.

Microarray-profiled AS events were classified by overlap with phastCons elements and by UPF dependence (see Fig. 3). Fisher's exact test was used to determine whether or not these two groupings were independent. This test was also used to compare sets of AS events showing pronounced dependence on one or another UPF factor (see Fig. S1 in the supplemental material). AS events from the spliceosomal and control groups of genes were classified by conservation and by PTC status (see Fig. 4), and the chi-square test was used to test whether or not these grouping were independent. The number (n) of AS events considered in each case is given in Results and/or the appropriate figure legend.

FIG. 3.

PTC-upon-inclusion alternative exons that show UPF1- or UPF2-dependent changes (at least a 5% difference) in inclusion level are significantly often flanked by highly conserved intronic sequences (*1, P = 2·10⁻⁴; *2, P = 3·10⁻², Fisher's exact test; compare the proportions marked by asterisks to the proportion for the total group [All]). The stacked bar graph shows the proportion of PTC-upon-inclusion AS events that overlap (black) or do not overlap (white) phylogenetically conserved sequences, as identified by the phastCons algorithm (49). Overlap requires that at least 35 of the first 50 nucleotides of both the upstream and downstream intron sequences flanking the alternative exon overlap phastCons elements. “All” represents all detectable PTC-upon-inclusion events (n = 164), and “KD” represents PTC-upon-inclusion events with at least a 5% difference in the indicated direction (more inclusion or more skipping) and knockdown (UPF1KD more inclusion, n = 46; UPF2KD more inclusion, n = 16).

FIG. 4.

Conserved AS events in genes for spliceosomal proteins are more often PTC introducing than are those in genes from the control set (compare proportions marked by asterisks; P = 3·10⁻³, chi-square test). Conserved AS events either have flanking introns overlapping (≥35 nucleotides) phastCons conserved elements or are conserved in mouse based on an analysis of cDNAs/ESTs. PTC-introducing AS events in the spliceosomal protein set are also more often conserved than nonconserved (compare proportions marked by closed circles). Genes with more than one AS event of the same type (PTC introducing or non-PTC introducing) were only counted once.

RESULTS

Individual knockdowns of NMD factors UPF1, UPF2, and UPF3X result in overlapping but distinct effects on PTC-introducing AS events.

To assess the relative requirements of the three core UPF proteins, UPF1, UPF2, and UPF3X, in AS-coupled NMD, we individually knocked down these factors in HeLa cells with transfected siRNAs that have been previously characterized and reported to specifically and efficiently target these factors (24). Western blot assays of lysates from HeLa cells transfected with UPF factor-specific siRNAs or with a control siRNA were probed with antibodies specific for UPF1, UPF2, or UPF3X (Fig. 1A; see also Fig. 2A in reference 42). Based on comparisons of UPF protein levels in the knockdowns (right panels) and in serial dilutions of control siRNA-treated cells (left panels), the knockdown levels of UPF1, UPF2, and UPF3X were estimated to be between ∼5% and 30% of the levels of the proteins detected in the corresponding control lysates.

FIG. 1.

Overlapping but distinct effects of UPF protein knockdowns on PTC-introducing AS events. (A) Western blot assays of HeLa cell lysates following siRNA-mediated UPF factor knockdowns. Western blots were probed with UPF2- or UPF3-specific antibodies (upper panels) and with calnexin- or β-actin-specific antibodies as loading controls (lower panels). Serial threefold dilutions of control lysates are shown in the left panels. An estimation of the knockdown efficiency following targeting siRNA treatment relative to the control siRNA treatment is shown below the gels (right panels). For a Western blot assay showing knockdown of UPF1, see reference 42. (B) The change in alternative exon inclusion level (percent skipping) is shown for AS events that introduce a PTC upon inclusion (left panel) or upon skipping (right panel). The change in percent skipping is represented by the cyan-black-yellow color scale shown and is calculated for each AS event as the percent skipping in the UPF1, UPF2, or UPF3X knockdown minus the percent skipping for the same AS event in the corresponding control siRNA treatment. Events for which we detected at least a 5% change in percent skipping upon UPF1 knockdown are shown, and rows are ordered according to the median of the percent skipping change across the three knockdowns.

We then used quantitative AS microarray profiling (42, 43, 48) to examine changes in AS patterns in cells transfected with UPF factor-specific siRNAs, compared to cells transfected with control siRNAs. Fluor-labeled cDNAs were hybridized to custom microarrays containing sets of exon body and splice junction probes for monitoring the inclusion levels of approximately 3,000 human cassette alternative exons mined from EST and cDNA sequence alignments. These AS events were mapped to ORFs to establish whether or not they have the potential to introduce a PTC in transcripts as a result of either their inclusion (referred to below as PTC upon inclusion) or skipping (referred to below as PTC upon skipping). PTC-upon-skipping events arise when the excluded exon results in a frameshift and the introduction of a downstream PTC, whereas PTC-upon-inclusion events arise when a termination codon is located within the inserted exon or introduced downstream as a result of a frameshift. The PTC-upon-inclusion and PTC-upon-skipping groups represent 15% and 30%, respectively, of the profiled AS events for which sufficient ORF information was available to map PTCs and which also met our detection criteria (total number of events = 1,111 [see Materials and Methods]).

We previously reported that PTC-introducing alternative exons are overrepresented among the alternative exons showing inclusion level changes upon UPF1 knockdown (42). These changes in exon inclusion level resulted in an increase in the abundance of the PTC-containing splice variant in approximately 90% of such cases, consistent with a loss of NMD activity. In the present study, we initially examined the effect of knocking down UPF2 or UPF3X on the inclusion levels of PTC-introducing alternative exons that displayed at least a 5% change in inclusion level following UPF1 knockdown (Fig. 1B). For each event shown, the difference in percent exon skipping level in each UPF factor knockdown compared to the level in the corresponding control sample is indicated by a color scale, where brighter yellow indicates an increase in the relative abundance of the exon-included splice variant and brighter cyan indicates an increase in the relative abundance of the skipped splice variant.

In comparing the effects of knockdowns of UPF1, UPF2, and UPF3X (Fig. 1B), we find that AS events affected by any one of the knockdowns are significantly more likely than expected by chance (P < 2·10⁻¹², Fisher's exact test; see Fig. S1 in the supplemental material) to be similarly affected (more skipping or more inclusion) by knockdown of either of the other two factors. Similar results were obtained when considering only the PTC-introducing AS events (see Fig. S2 in the supplemental material). Among the subset of AS events similarly affected when comparing any pair of knockdowns, on average, approximately half (a minimum of 26% and a maximum of 86%) are also similarly affected by the third knockdown (see Fig. S1 in the supplemental material; data not shown). However, while we observe significant overlaps between AS events affected by pairs of knockdowns, variable proportions of which are affected in all three knockdowns, some events appear to be significantly affected by only two knockdowns (e.g., UPF1 and UPF3X knockdowns or UPF1 and UPF2 knockdowns), and other events show a pronounced effect only upon UPF1 knockdown (see Fig. S1 and S2 in the supplemental material).

The microarray predictions for both distinct and overlapping effects of the UPF protein knockdowns were validated by RT-PCR assays with primers specific for sequences in the constitutive exons flanking the PTC-introducing exons. Representative RT-PCR data for 12 AS events (out of 128) assayed in the three UPF knockdowns and corresponding controls are shown in Fig. 2. The percent exon skipping measured by RT-PCR assays agrees very well with the microarray data (R² = 0.8 to 0.9; see Fig. S3 in the supplemental material). For knockdown-dependent AS level changes predicted by the microarray data to be at least 5% different in exon inclusion level, RT-PCR assays showed a change in the same direction in 83% of the cases tested (see Fig. S3 in the supplemental material). For the quantifications of a larger subset of RT-PCR assays corresponding to the microarray data in Fig. 1B, see Fig. S4 and Table S2 in the supplemental material. Although pronounced changes in percent skipping are more often observed upon knockdown of UPF1 than upon knockdown of UPF2 or UPF3X for the PTC-upon-inclusion or PTC-upon-skipping events, it is important to note that the frequencies of changes in non-PTC-introducing AS events are similar for the three knockdowns (see Fig. S5 in the supplemental material) and that more pronounced changes upon UPF2 or UPF3X than upon UPF1 knockdown are observed for a subset of AS events (see Fig. S1 and S2 in the supplemental material). In addition, the overall frequencies of pronounced AS level changes in all microarray-profiled events are similar in the UPF1 and UPF3X knockdowns and only slightly lower in the UPF2 knockdown (see Fig. S1 and S5 in the supplemental material). These results provide evidence that AS-coupled NMD has differential dependencies on core UPF proteins and that these differential dependencies are unlikely due to possible differences in the efficiencies of the siRNA knockdowns of the three factors.

PTC-introducing AS events affected by UPF knockdowns are often flanked by highly conserved sequences.

Conservation of sequence elements among species is generally associated with selection pressure to preserve function, and regulated alternative exons are often associated with conserved flanking intron sequences (50, 51, 57). Accordingly, to identify PTC-introducing AS events that are most likely associated with conserved gene regulatory functions, we measured the association between introns flanking PTC-introducing exons that display UPF factor-dependent changes in inclusion levels and conserved sequence elements. Alternative exons represented on our array were aligned to human genome sequences, and the coordinates of the upstream and downstream flanking intron sequences (50 or 150 nucleotides) were identified. We then calculated the proportion of nucleotides that overlap with conserved elements detected by the phastCons algorithm (49). In this analysis, we scored overlaps with the most-conserved sequence elements identified by the phastCons algorithm, which are derived from alignments of genomic sequences from 17 species. We observed a statistically significant enrichment for overlap with these phastCons elements in introns flanking PTC-upon-inclusion alternative exons that display a UPF1 knockdown-dependent increase in relative abundance (35% with phastCons overlap, n = 46), compared with all PTC-upon-inclusion exons (16% with phastCons overlap, n = 164; P = 2·10⁻⁴, Fisher's exact test [Fig. 3]). Similarly, a significant enrichment for overlap with phastCons elements in introns flanking the UPF2-dependent group of PTC-upon-inclusion exons (38% with phastCons overlap, n = 16) was observed (P = 0.03 [Fig. 3]). However, no such enrichment was observed for the UPF3X-dependent events. Differences between the conservation levels of intron sequences flanking PTC-upon-inclusion alternative exons that do or do not show UPF protein knockdown-dependent changes in percent skipping are also shown in empirical cumulative distribution function plots (see Fig. S6 in the supplemental material).

Some PTC-upon-skipping alternative exons displaying UPF knockdown-dependent changes in AS levels also overlapped phastCons conserved elements (Table 1; see Table S2 and Fig. S4 in the supplemental material); however, a statistically significant enrichment was not observed (data not shown). Some enrichment of phastCons elements was also observed among AS events which display an increase in skipping upon knockdown of UPF2 or UPF3X but which do not introduce a PTC (see Fig. S7 in the supplemental material). Further experimentation is necessary to determine whether or not these effects are associated with NMD. In summary, the analysis described above reveals that a significant number of PTC-introducing AS events regulated by one or more UPF proteins, particularly those in the PTC-upon-inclusion category, have flanking introns that overlap highly conserved sequences. This observation suggests that these AS events and their associated flanking intronic sequences likely have important regulatory roles.

New regulatory targets for AS-coupled NMD.

We examined the list of functions represented by genes with PTC-introducing AS events that show UPF factor knockdown-dependent changes in percent skipping (Fig. 1B). Among the subset flanked by conserved sequences (Fig. 3; see Fig. S4 in the supplemental material) are 13 AS events located in transcripts from genes with RNA processing-related functions (Table 1). Other such AS events are located in genes that represent a range of other cellular functions, including nucleotide metabolism (e.g., the NT5C3 gene), signaling (e.g., the RAB5A gene), and sumoylation (e.g., the SENP1 gene) (for additional examples, see Table S3 in the supplemental material). Surprisingly, in addition to AS regulators such as SR and hnRNP proteins which have recently been associated with conserved PTC-introducing events (28, 40), regulated PTC-introducing AS events were found in core spliceosome components. On this list are genes that encode the U1 snRNP-specific 70-kDa protein (U170K) and the common snRNP Sm protein SmB/B′ (SNRPB) and other genes involved in spliceosome formation (SF1, PRPF18, and SMNDC1 gene [Table 1]). Consistent with conserved regulatory roles of the PTC-introducing AS events in these genes implied by the detection of overlaps with phastCons elements, most of these AS events were also detected in alignments of mouse ESTs (Table 1). Moreover, the microarray-predicted change in relative inclusion levels of these exons was validated by RT-PCR assays in at least one of the UPF factor knockdowns (Fig. 2; see Fig. S4 and Table S2 in the supplemental material).

The identification of conserved, UPF-regulated, PTC-introducing AS events in genes of core spliceosomal components led us to investigate whether such events are a more general feature of genes that encode spliceosomal factors, including spliceosomal factor genes that are not represented on our AS microarray. To this end, we computationally mined PTC-introducing and non-PTC-introducing cassette-type AS events from EST/cDNA data representing a curated (2) list of 253 genes that encode spliceosome-associated proteins and other splicing factors. For comparison, a set of non-spliceosome-related Unigene clusters with similar EST coverage was also analyzed in parallel. We identified 443 AS events in 149 Unigene clusters in the spliceosome-associated set (see Table S4 in the supplemental material) and 547 events in 161 Unigene clusters in the control set (see Table S5 in the supplemental material). Both sets had similar distributions of AS events per gene (data not shown), with a median of two AS events per gene. Available sequence information for these transcripts allowed the annotation of whether the AS events were PTC introducing or non-PTC introducing for 73% and 74% of the events in the spliceosomal and control gene groups, respectively.

Our microarray results above suggested that PTC-containing AS transcripts affected by UPF factor knockdowns in HeLa cells are often flanked by conserved sequences (Fig. 3) and many show conservation with respect to mouse based on ESTs (Table 1). These features were assessed for the AS events mined from the spliceosomal and control Unigene clusters. As described above for the microarray events, we calculated the proportions of the 150 nucleotides of intron sequences flanking the human alternatively spliced exons that overlap phastCons elements. In parallel, we mined mouse ESTs to assess which of the human AS events are “conserved” in mouse. We found that among the AS events showing flanking intron sequence and/or conservation based on an analysis of EST/cDNA sequences (n = 84 and n = 54 for the spliceosomal and control groups, respectively), the distribution of PTC-introducing and non-PTC-introducing AS events was significantly different between the two groups, with a higher proportion of PTC-introducing events in the spliceosomal group than in the control group (71% versus 46%, P = 3·10⁻³, chi-square test [Fig. 4]). For differences in the distribution of conserved sequence elements in introns flanking the alternative exons in the spliceosomal and control groups, also see Fig. S8 in the supplemental material. In addition, within the spliceosomal group, there is a higher proportion of PTC-introducing AS events in the conserved group than in the nonconserved group (71% versus 61% [Fig. 4]). Although a subset of the genes contains more than one type (PTC category) of AS event, similar results were obtained when these genes were removed from the analysis (data not shown). AS factors including SR and hnRNP proteins, which have been previously associated with conserved, PTC-introducing AS events (28, 40), were also identified by our analysis (Table 2) . However, these did not account for the entire enrichment of conserved, PTC-introducing events in the spliceosomal genes, and similar results were also obtained when these genes were removed (data not shown). Thus, among the genes encoding spliceosomal proteins and other splicing factors are examples not known previously to be associated with AS-coupled NMD (Table 2). Moreover, the association of conserved, PTC-introducing AS events with genes that encode these spliceosomal factors suggests that these AS events play important regulatory roles.

Autoregulation of core splicing factors by AS-coupled NMD.

Previous studies have shown that several AS regulatory factors, such as SR and hnRNP proteins, can autoregulate their expression levels via AS-coupled NMD (see the introduction). However, such a function for core spliceosomal proteins has not been previously established. The identification of conserved, PTC-introducing alternative exons in core spliceosomal factors suggested that the levels of these proteins might be subject to autoregulation. Accordingly, we tested the effect of increased expression of SNRPB or SMNDC1 on the inclusion levels of the PTC-introducing exons in their respective transcripts. Vectors that encode Flag epitope-tagged versions of these proteins and, as a control, the parental vector or a vector that encodes a Flag epitope-tagged U2 snRNP-associated protein, DDX42 (54), were transiently transfected into HeLa cells. RT-PCR assays were performed with primer pairs designed to specifically amplify the endogenous PTC-containing splice variants (Fig. 5). Comparable levels of expression of the three Flag epitope-tagged proteins were obtained (Fig. 5B). Increased expression of SNRPB and SMNDC1 led to a reproducible (observed in three independent experiments) approximately twofold increase in the level of the PTC-containing splice variant transcript arising from alternative exon inclusion, specifically, in SNRPB and SMNDC1 transcripts, respectively (Fig. 5A). In contrast, overexpression of DDX42 in parallel had a comparatively little effect on the levels of the exon-including transcript. Overexpression of SNRPB or SMNDC1 also had lesser effects on levels of PTC-containing transcripts from other splicing-related genes (SF1, TRA2A, and SR140 genes) (Fig. 5A). Thus, artificially increasing the amount of SNRPB or SMNDC1 protein leads to an increase in the level of endogenous PTC-containing, alternative exon-including SNRPB or SMNDC1 transcripts, respectively. These results are thus consistent with the autoregulation of expression levels of these factors via AS-coupled NMD. However, further experimentation is necessary to establish whether this autoregulation occurs via direct or indirect mechanisms.

DISCUSSION

By quantitative AS microarray profiling, we compared the effects of individually knocking down each of the three core NMD factors, UPF1, UPF2, and UPF3X, on the inclusion levels of alternative exons that do or do not introduce PTCs. In pairwise comparisons of the effects of the knockdowns of these factors, significant overlaps were observed between the PTC-introducing AS events that display inclusion level changes. In addition, a substantial proportion (on average, ∼50%) of the AS events similarly affected by any two knockdowns are also affected by the third knockdown. This enrichment for AS events that have overlapping UPF protein requirements is consistent with the operation of these factors in a common pathway in many cases. However, at least one-third of PTC-introducing AS events displayed distinct effects in the different knockdowns, with many of the UPF1-dependent events showing little to no detectable dependence on UPF2 and/or UPF3X. The PTC-introducing AS events displaying common and differential UPF protein requirements were validated at a high rate by semiquantitative RT-PCR assays. Our results thus indicate that AS-coupled NMD may have differential requirements for UPF proteins and may occur via more than one UPF1-dependent branch of the NMD pathway. Moreover, most of the genes we have profiled that contain PTC-introducing exons displaying UPF protein-dependent inclusion level changes were not known previously to be regulated by AS-coupled NMD. Of particular interest are the findings that several core snRNPs and other general spliceosomal proteins are on this list and that, upon overexpression, at least those factors tested can affect AS of their own transcripts. This observation supports previous findings (10, 44, 45) (see below) that core components of the spliceosome can participate in the regulation of AS, in addition to their well-established roles in constitutive splicing. In summary, our results provide new information on the factor requirements for AS-coupled NMD and many interesting examples of AS-coupled NMD and also reveal that core spliceosomal proteins can autoregulate their expression levels.

Alternative branches of the mammalian NMD pathway.

Accumulating evidence indicates that there are multiple branches of the NMD pathway with differing requirements for EJC and UPF proteins (8, 15). Experiments involving tethering of different EJC proteins to reporter transcripts, in combination with siRNA knockdowns of UPF1 or UPF2, initially suggested the existence of a UPF2-independent branch of the NMD pathway (15). Results from conventional microarray profiling studies have indicated that some endogenous transcripts displaying altered steady-state levels upon knockdown of UPF1 (36) are not affected by UPF2 (55) or UPF3X (8). However, in most cases, the relationship between the differential effects of knocking down different UPF proteins and the PTC status of the affected transcripts was not clear. Our results extend these previous studies by linking the effects of the knockdown of different NMD proteins to specific PTC-containing transcripts arising from AS and reveal that core UPF proteins have overlapping but distinct effects on the degradation of the PTC-containing splice variants.

Both UPF1 and UPF2 are encoded by unique genes, whereas UPF3X (also UPF3B) and UPF3 (also UPF3A) are paralogs. This raises the question of whether UPF3X-independent NMD could be due to the existence of possible redundant functions with UPF3. While we have not specifically tested this possibility, on the basis of previous studies we believe that functional redundancy between these proteins, at most, plays a limited role in explaining the effects that we have observed. Despite sharing 42% and 60% amino acid sequence identity and similarity, respectively (47) and both associating with spliced mRNA (23), UPF3X and UPF3 have divergent C-terminal regions, and the activation of NMD by tethered UPF3X is more efficient than it is by tethered UPF3 and requires different factors (26). Moreover, the majority of UPF3X-independent targets validated in the aforementioned microarray profiling study were not affected by knockdown of UPF3 alone or in combination with UPF3X (8). Nevertheless, some degree of redundant effects may exist between these factors in AS-coupled NMD. In addition, since truncating mutations associated with X-linked mental retardation were found to eliminate the expression of UPF3X through NMD, UPF3X may have partially redundant functions in NMD (53).

AS-coupled NMD and regulation of core spliceosomal proteins.

An important question concerning gene regulation is how components of the spliceosome are regulated at appropriate levels in different cell types and under different growth conditions. Unexpectedly, our initial results stemming from microarray profiling revealed that AS-coupled NMD is likely involved in the regulation of core spliceosomal components or assembly factors. We initially identified PTC-introducing alternative exons in transcripts from a subset of spliceosomal component or assembly factor genes (Table 1), some of which are conserved in mouse and all of which respond to knockdown of UPF1 and in some cases knockdown of UPF2 and/or UPF3X as well. Moreover, the presence of highly conserved intronic sequences flanking the exons of these UPF protein-responsive splice variants is also indicative of important regulatory roles. In the examples we have analyzed (Fig. 5), the overexpression of each spliceosomal protein affected the level of the PTC-containing splice variant produced from its corresponding gene, suggesting autoregulatory functions for these core spliceosomal proteins.

Since our microarray did not represent a comprehensive set of AS events in splicing factor genes, we performed a computational search for AS events in a list of spliceosomal factor genes curated based on experimental evidence supporting a functional or physical association with spliceosomal complexes (2). Strikingly, we found that conserved, PTC-introducing AS events were significantly enriched in these genes, even when excluding previously identified examples of regulatory splicing factor genes with PTC-introducing AS events. These results therefore significantly extend the recent identification of UPF1-dependent, PTC-introducing AS events in highly conserved regions of genes that encode defined AS factors of the SR and hnRNP families (28, 40). Besides providing an extensive list of spliceosomal proteins that are likely also subject to regulation by AS-coupled NMD, the experiments in the present study provide evidence that the increased expression of some of these proteins can activate autoregulatory mechanisms.

Our results indicating that proteins involved in the formation of the core spliceosome are regulated by AS-coupled NMD and the results implicating these proteins in autoregulatory loops extend previous findings that components of the “basal” splicing machinery can function in the regulation of AS. For example, previous microarray profiling studies have shown that yeast strains harboring mutations or deletions of various core splicing components displayed transcript-specific effects on intron retention (10, 45). An RNA interference screen for AS regulators revealed that knockdown of several core splicing factors in Drosophila cells result in transcript-specific AS effects (44). Among these factors was SPF30, the ortholog of human SMNDC1, which we have shown contains a conserved PTC-introducing alternative exon that is regulated by UPF1 and when overexpressed affects the levels of its PTC-containing transcript (Table 1 and Fig. 5). In this regard, while our results point to autoregulatory functions of core spliceosomal proteins operating via NMD, it is also interesting to consider that these and other proteins can regulate additional splicing events, including, for example, AS-coupled NMD events that are associated with other splicing factors. Such auto- and cross-regulation could provide an important mechanism for ensuring both the appropriate absolute and relative levels of proteins that make up the core splicing machinery.