Direct competition between hnRNP C and U2AF65 protects the transcriptome from the exonization of Alu elements

Abstract

There are ~650,000 Alu elements in transcribed regions of the human genome. These elements contain cryptic splice sites, so they are in constant danger of aberrant incorporation into mature transcripts. Despite posing a major threat to transcriptome integrity, little is known about the molecular mechanisms preventing their inclusion. Here, we present a mechanism for protecting the human transcriptome from the aberrant exonization of transposable elements. Quantitative iCLIP data show that the RNA-binding protein hnRNP C competes with the splicing factor U2AF65 at many genuine and cryptic splice sites. Loss of hnRNP C leads to formation of previously suppressed Alu exons, which severely disrupt transcript function. Minigene experiments explain disease-associated mutations in Alu elements that hamper hnRNP C binding. Thus, by preventing U2AF65 binding to Alu elements, hnRNP C plays a critical role as a genome-wide sentinel protecting the transcriptome. The findings have important implications for human evolution and disease.

Graphical abstract

Figure 1

Examples of U2AF65 and hnRNP C Binding at the 3′ Splice Sites of Constitutive or hnRNP-C-Repressed Exons within the CD55 Gene (A) Genome browser view of the CD55 gene displaying the iCLIP data (crosslink events per nucleotide) of hnRNP C (blue) and U2AF65 (purple) as well as the RNA-seq data (overlapping reads per nucleotide; green) from control and HNRNPC knockdown HeLa cells. The red arrowhead marks the hnRNP-C-repressed alternative Alu exon. RefSeq transcript annotations (gray) and Alu elements in antisense orientation to the shown strand (orange) are depicted below. (B) Enlargement of the genomic region containing the 5′ UTR and the first four exons. Red arrowheads mark U2AF65 peaks at 3′ splice sites. (C) Enlargement of the region around the 3′ splice site of the hnRNP-C-repressed Alu exon (marked in A) including the underlying genomic sequence. The red arrowhead marks the site of increased U2AF65 occupancy in the HNRNPC knockdown. See also Figures S1 and S2.

Figure 2

hnRNP C and U2AF65 Compete for Binding on U-Tracts and at Regulated Exons (A) The average ratio of U2AF65 occupancies from knockdown (KD) over control is shown for U2AF65-binding sites that do not overlap with hnRNP C (no binding) or that overlap with hnRNP-C-binding sites within different ranks of hnRNP C occupancy (i.e., rank 10 contains the 10% strongest hnRNP-C-binding sites). Error bars indicate the 95% confidence interval of the mean. (B) Plot showing the ratio of U2AF65 occupancies against the total number of U2AF65 crosslink events for individual binding sites. Binding sites that show an at least 4-fold change in occupancy and overlap with hnRNP C binding are depicted in red. (C) Plots depicting the frequency of overlapping binding sites of ten other RNA-binding proteins around the summit of U2AF65-binding sites (position 0) that overlap with hnRNP C. Proteins that show increased crosslinking are colored. (D) Plot as in (A) showing U2AF65-binding sites that overlap with hnRNP C compared to sites that overlap with both hnRNP C and TIA1, TIAL, TDP-43, or PTB (indicated below) and sites that overlap with only hnRNP C and none of the other proteins. The percentage of shared hnRNP-C-U2AF65-binding sites within the different categories is indicated above. (E) Weighted Venn diagrams depicting the overlap of U2AF65-binding sites that are bound by hnRNP C (blue) and/or either TIA1/TIAL, TDP-43, or PTB (green). Absolute numbers are given within each segment. (F) Weblogos showing the relative nucleotide frequency around the summits (position 0) of hnRNP-C- and U2AF65-binding sites. (G) Plots comparing the pentamer fold-enrichment around crosslink sites from replicate experiments with hnRNP C and U2AF65 from control and HNRNPC knockdown HeLa cells. The three panels compare iCLIP data from (i) experiments with both proteins from untreated HeLa cells (Ctrl; left), (ii) replicate experiments with U2AF65 from Ctrl cells (middle; see also Figure S3D), and (iii) experiments with U2AF65 from HNRNPC knockdown (KD1) and Ctrl cells (right). (H) Autoradiograph from an in vitro UV crosslinking assay using recombinant hnRNP C1 (33 kDa) and U2AF65^RRM12 proteins (21 kDa). A stable amount of U2AF65^RRM12 plus increasing concentrations of hnRNP C1 (indicated above in μM) were UV crosslinked to radioactively labeled wild-type (U₁₀, lanes 1–7) and mutant (U₂CU₄CU₂, lanes 8–14) RNA oligonucleotides (100 nM) and analyzed by denaturing gel electrophoresis. Radioactive signals of RNA crosslinked to hnRNP C1 or U2AF65^RRM12 are marked on the left. Asterisks indicate C-terminal hnRNP C1 truncations (^∗) and GST (^∗∗). Coomassie staining of the same gel (bottom) serves as loading control. Note that there is an additional hnRNP C1 signal likely representing an hnRNP C1 dimer, which is only shown in Figure S3H. (I) RNA maps showing the total number of crosslink events of U2AF65 in control HeLa (light purple) and HNRNPC knockdown cells (dark purple) relative to the 3′ splice sites of all exons that (i) are repressed and bound by hnRNP C (left), (ii) are repressed but not bound by hnRNP C (middle), and (iii) are not subject to any regulation in the HNRNPC knockdown (fold change < 1.1; right). The number of exons in each category is indicated above. See also Figure S3.

Figure 3

The HNRNPC Knockdown Leads to Widespread Exonization of Antisense Alu Elements (A) Box plots summarizing the change in normalized expression of Alu exons compared to downstream non-Alu exons as well as all exons. (B) Pie chart summarizing the regulation of all 1,903 Alu exons detect from our RNA-seq data. Upregulated and downregulated exons are further subdivided into those called by DEXSeq or displaying a more than 2-fold change in the HNRNPC knockdown. (C) Plot depicting the mean inclusion levels in control HeLa cells (open diamonds) and both HNRNPC knockdowns (KD1, filled circles; KD2, filled diamonds) of 55 Alu exons that were measured by RT-PCR (Data S1A and S1B). (D) Bar chart showing the percentage of binding sites of hnRNP C and ten other RNA-binding proteins (indicated below) that overlap with antisense Alu elements. (E) Semiquantitative RT-PCR analyses of four cryptic and four Ensembl-annotated Alu exons (indicated on the left) upon knockdown of HNRNPC (two independent knockdowns, KD1 and KD2), TIA1/TIAL, PTB/nPTB, HNRNPA1, and TDP-43 (labeled as KD with the respective gene[s] indicated above) as well as in control HeLa cells (Ctrl). Known target exons of these proteins can be found in Data S1C. (Right) Gel views of capillary electrophoresis of the PCR products with the fragments including (in) or excluding (ex) the Alu exon marked on the right. (Left) Bar diagrams depicting the mean inclusion (gray) and exclusion (white) level in each sample. Asterisks indicate the significance level (Student’s t test) relative to control: n.s., nonsignificant; ^∗p value < 0.05; ^∗∗p value < 0.001; ^∗∗∗p value < 0.0001. Error bar represents SDM; n = 3. (F) Schematic representations of hnRNP C crosslink events per nucleotide (top) and of Alu exon locations (bottom) along the Alu consensus sequence. Exons that extend beyond the Alu element end with a blue dash. (G) Plots depicting the ratio of the cumulative frequencies of U-tracts of a given length (e.g., at least five uridines) in exonized Alu elements (red line) compared to nonexonized Alu elements within the same genes. Analyses are separately shown for Alu exons from the first or second arm of the Alu element (gray rectangle above) as well as for the upstream and linker U-tracts (magnifier icon). Nonexonized Alu elements of all other genes (gray line) serve as control. Black dots, p value < 0.05 (Pearson’s chi-square test). See also Figures S4 and S5, Data S1, and Table S3.

Figure 4

The Competition of hnRNP C with U2AF65 at 3′ Splice Sites Represses Alu Exon Inclusion (A) Heatmaps comparing the amount of crosslinking of hnRNP C (left; different shades of blue) and U2AF65 (different shades of purple) in control (middle) and HNRNPC knockdown cells (right) relative to the 3′ splice sites of Alu exons (indicated by a dashed line). U2AF65 iCLIP data were corrected for differences in library sizes. Each row corresponds to one of 681 Alu exons that contain at least five crosslink events within the analyzed region (from −50 nt to +10 nt relative to the first nucleotide of the exon). The bar diagram on the right shows the fold change in Alu exon inclusion (KD1 over wild-type; differentially regulated exons according to conditional thresholding are shown in red). (B) Heatmaps as in (A) for a subset of 200 control non-Alu exons that lie downstream within the same genes (full set in Figure S5D). See also Figure S5.

Figure 5

Point Mutations that Impair hnRNP C Binding Promote Inclusion of the Alu Exon in the CD55 Minigene (A) Schematic overview of the minigene including the Alu exon (gray square), intronic regions (black lines), and two flanking exons (white squares) from the CD55 gene. The original sequence (WT) as well as the mutated sequence surrounding the 3′ and 5′ splice sites (3^mut and 5^mut, respectively; splice sites marked by arrowheads) are depicted below. Introduced point mutations are highlighted in black. (B) RT-PCR monitoring inclusion or suppression of the Alu exon in the minigenes with wild-type (WT) or mutated sequences (3^mut, 5^mut) in HNRNPC knockdown (KD1 and KD2) and control HeLa cells (Ctrl). The corresponding capillary electrophoresis data is given in a gel-like representation with Alu exon inclusion and suppression indicated schematically on the right. (C) Average Alu exon inclusion in percent from three replicate RT-PCR experiments. Lines indicate relevant comparisons with asterisks representing different levels of significance (^∗p value < 0.05; ^∗∗p < 10⁻³; Student’s t test). Error bars represent SDM. See also Figure S6 and Table S5.

Figure 6

hnRNP C Repression of Alu Exonization in the PTS Gene Is Relevant for Disease (A) Genome browser view including the disease-relevant Alu element (orange) within the PTS gene. iCLIP data for hnRNP C (blue) and U2AF65 (purple) from HNRNPC knockdown (KD1 and KD2) and control HeLa cells as well as RNA-seq data (green) are shown above. The corresponding isoforms are schematically indicated: Alu suppression in isoform S, usage of the downstream 3′ splice site (open arrowhead) in isoform 1 (light gray; this isoform is produced as a result of the disease-associated deletion which removes the upstream 3′ splice site together with the U-tract) and usage of the upstream 3′ splice site (filled arrowhead) in isoform 2. Wild-type sequence (WT) and disease-associated deletion are shown below. (B) Gel-like view of capillary electrophoresis of RT-PCR analyses of minigenes containing the Alu element described in (A) with the two flanking exons. The different isoforms are schematically indicated on the right. (C) Average Alu exon inclusion in percent from three replicate RT-PCR experiments. Lines indicate relevant comparisons with asterisks representing different levels of significance (^∗∗∗p < 10⁻⁴; n.s., not significant; Student’s t test). Error bars represent SDM. (D) Bar diagram depicting the fraction of downregulated genes within sets of genes carrying Alu exons with different levels of upregulation in KD1 (intervals of the fold change in inclusion in KD1 are given below). See also Table S5.

Figure 7

hnRNP C Safeguards the Transcriptome from the Exonization of Alu Elements In normal cells, hnRNP C prevents recognition of the Alu elements through U2AF65, thereby ensuring accurate splicing. In the HNRNPC knockdown, U2AF65 can bind to the U-tracts and promote Alu exonization. Similarly, disease-associated mutations in the U-tracts can favor Alu exonization in the presence of hnRNP C by impairing hnRNP-C-binding. The resulting nonfunction transcripts are likely either targeted by nonsense-mediated decay (NMD) or give rise to nonfunctional proteins. Once an exon acquires beneficial changes during evolution, similar mutations accumulate to relieve hnRNP C repression, opening the possibility for new protein functionalities.

Figure S1

iCLIP Data Show Binding of hnRNP C and U2AF65 at 3′ Splice Sites of Target Exons, Related to Figure 1 (A) The majority of hnRNP C and U2AF65 crosslink events is located within introns. Pie chart summarizing the fraction of crosslink events within different genomic regions. 3‘ splice site (3′ ss) and 5‘ splice site (5′ ss) indicate the regions 200 nt upstream and downstream of Ensembl-annotated exons, respectively. (B) Binding sites on longer U-tracts show higher hnRNP C occupancy. Top panel: absolute numbers of hnRNP C-bound U-tracts of different length in the transcriptome (open circles) and the corresponding fraction of bound tracts (filled circles). Lower panel: heatmap showing the fraction of U-tracts of a given length that is allocated to the ten different ranks of hnRNP C occupancy. 3-nt U-tracts show the highest enrichment at the 10% weakest hnRNP-C-binding sites (rank 1), whereas U-tracts of nine or more nucleotides show the highest enrichment at the 10% strongest binding sites (rank 10). (C) hnRNP C binding is enriched immediately upstream of the 3′ splice sites of hnRNP-C-repressed exons. RNA map showing the percentage of exons that have a crosslink nucleotide at a certain position relative to the 3′ and 5′ splice site of all exons that are repressed by hnRNP C (taken from our RNA-seq data, Table S2). (D) U2AF65 binding is strongly enriched immediately upstream of the 3′ splice sites of all annotated exons. RNA map as in (C) visualizing U2AF65 and hnRNP C iCLIP data relative to all 3′ and 5′ splice sites annotated in the Ensembl database.

Figure S2

The Protein Abundance and RNA-Binding Ability of U2AF65 Are Not Affected by the HNRNPC Knockdown, Related to Figure 1 (A) The total amount of crosslinking of U2AF65 to RNA is not altered in the absence of hnRNP C. Analysis of crosslinked U2AF65-RNA complexes using denaturing gel electrophoresis and membrane transfer. Protein extracts were prepared from UV-crosslinked (UV+) control HeLa cells (Ctrl) and HNRNPC knockdown cells (KD1 and KD2), and RNA was partially digested using low (+) or high (++) concentrations of RNase. U2AF65-RNA complexes were immuno-purified with a mouse antibody against U2AF65 (α U2AF65). To allow visualization of the protein-RNA complexes, the 5′ ends of the RNAs were radioactively labeled. The complexes were size-separated using denaturing gel electrophoresis and transferred to a nitrocellulose membrane. The upper panel shows the autoradiograph of this membrane. U2AF65-RNA complexes are shifted upward from the size of the protein (53 kilo Dalton, kDa; lane 6: the red box indicates the region that was extracted for subsequent analyses). This shift is focused when high RNase concentrations are used (lane 5). A similar pattern with comparable intensity is observed for the HNRNPC knockdown cells (lanes 7-10), indicating that crosslinking is not generally affected. As a control, no signal is observed in experiments where the antibody was omitted during immunoprecipitation (lanes 3 and 4). Importantly, when omitting UV irradiation, no shifted U2AF65-RNA complexes can be observed. The remaining radioactive signal at the size of U2AF65 (marked by ^∗) in these samples indicates that part of the protein is labeled under the used conditions. The lower panel shows the Western blot analysis of the same immunoprecipitations with a rabbit antibody against U2AF65 (Rb α U2AF65). (B) Analysis of PCR amplified iCLIP cDNA libraries using gel electrophoresis. RNA recovered by proteinase K digestion from the nitrocellulose membrane as indicated in (A) was reverse transcribed and size-selected using denaturing gel electrophoresis (not shown). Three of the cDNA size fractions (H: 100-170 nt; M: 85-100 nt; L: 75-85 nt) were further processed and PCR amplified (17 cycles of amplification) to obtain the iCLIP libraries. In the gel image, PCR products of different sizes can be observed, according to the size of the input fractions from control HeLa cells (Ctrl, lanes 13-18) and HNRNPC knockdown samples (KD1 and KD2, lanes 4-12). When no antibody was used in the immunoprecipitation, no signal is observed (lanes 1-3). Positions of a size standard are given on the left, and the position of the PCR primers is indicated by an asterisk. (C) The protein abundance of U2AF65 is not altered in the HNRNPC knockdown. Western blot analyses with antibodies against U2AF65, hnRNP C and GAPDH (indicated on the right) comparing the protein abundances in control HeLa cells (Ctrl, lanes 1-3) and HNRNPC knockdown cells (KD1 and KD2, lanes 4-9) in triplicates. The HNRNPC knockdown efficiency was estimated to about 20% based on comparison with Ctrl lanes containing 50%, 20% and 10% of input material (lanes 10-12). A protein size marker in kDa is indicated on the left.

Figure S3

hnRNP C Competes with U2AF65 Binding at Long U-Tracts and Cryptic Exons, Related to Figure 2 (A) Scatter plots comparing the U2AF65 iCLIP crosslink events within binding sites (purple) or Ensembl transcripts (black) for all replicate experiments from control HeLa cells (Ctrl) and HNRNPC knockdown cells (KD1 and KD2). Sample types and replicate numbers (in brackets) are given along the diagonal. The Spearman’s rank correlation (r) for each pair is given in the upper left corner of the respective panel. (B) The changes in U2AF65 occupancies on individual binding sites are highly correlated between both knockdowns. Scatter plot comparing the fold changes in U2AF65 occupancy in the two different HNRNPC knockdowns (KD1 and KD2). The occupancy was calculated by normalizing the number of crosslink events within each binding site by the total number of crosslink events within the corresponding gene. The Pearson’s product-moment correlation (r) is given in the upper left corner. (C) U2AF65 shows significantly changed occupancy on a large number of binding sites in the HNRNPC knockdown. Plot showing the ratio of U2AF65 occupancy from the combined HNRNPC knockdowns (KD) over Ctrl against the sum of U2AF65 crosslink events in all samples. The 5,500 binding sites that show an at least 4-fold change in U2AF65 occupancy are shown in yellow. The dashed lines mark a 4-fold change in either direction. (D) hnRNP C preferentially crosslinks to UUUUU pentamers. Scatter plot comparing the pentamer fold-enrichment at crosslink sites from two hnRNP C iCLIP replicate experiments. hnRNP C prefers the pentamer UUUUU (red), but shows no particular enrichment for further pyrimidine combinations (purple) or other pentamers (black). (E) U2AF65-binding sites that overlap with a long U-tract show increased U2AF65 occupancies in the HNRNPC knockdown. The average ratio of U2AF65 occupancy from HNRNPC knockdown (KD) over control HeLa cells (Ctrl) is shown for U2AF65-binding sites that overlap with U-tracts of varying lengths. Error bars indicated the 95% confidence interval of the mean. (F) Sites of hnRNP C-U2AF65 competition are strongly enriched at deep-intronic locations (p value < 0.001 compared with all other U2AF65-binding sites, hypergeometric test), in particular at cryptic exons and within Alu elements. Pie charts showing the fraction of binding sites located at annotated Ensembl exons as well as at deep-intronic positions (subdivided into positions at cryptic exons, within Alu elements and other). Graphs are shown from left to right for: all hnRNP C, all U2AF65 binding sites, U2AF65 binding sites that overlap with hnRNP C binding, and U2AF65 binding sites that show an at least 4-fold increase in occupancy in the HNRNPC knockdown and overlap with hnRNP C (competition sites). (G) Recombinant U2AF65^RRM12 shows comparable crosslinking to the wild-type (U₁₀) and mutant (U₂CU₄CU₂) RNA oligonucleotides that resemble the upstream U-tract of the Alu exon in the NUP133 minigene (Figure S6). Increasing concentrations of U2AF65^RRM12 (21 kDa, concentration indicated above in μM) were incubated with radioactively labeled wild-type or mutant RNA oligonucleotide (100 nM), UV crosslinked and analyzed by denaturing gel electrophoresis. The radioactive signal from the RNA crosslinked to the protein can be observed in the autoradiograph (U2AF65^RRM12-RNA, top panel). Coomassie staining of the same gel serves as loading control (lower panel). (H) Recombinant hnRNP C1 crosslinking is drastically reduced to the mutant RNA oligonucleotide. Experimental setup as in (G) but using increasing concentrations of hnRNP C1 (33 kDa, concentration indicated above). Note that in addition to the signal derived from hnRNP C1 crosslinked to RNA (hnRNP C1-RNA), an additional crosslinking signal is visible at about 65 kDa. This signal is likely derived from two hnRNP C1 proteins crosslinked to one RNA molecule and therefore labeled as dimer-RNA. Impurities are indicated by asterisks on the left (^∗ C-terminal truncations of hnRNP C1; ^∗∗ GST).

Figure S4

RNA-Seq Identifies a Large Number of hnRNP-C-Repressed Alu Exons, Related to Figure 3 (A) The detected splicing changes show a strong correlation between both HNRNPC knockdowns. Scatter plot comparing the fold changes (log₂) in normalized exon expression (HNRNPC knockdown over control HeLa cells) from both HNRNPC knockdowns (KD1 and KD2). The Pearson’s product-moment correlation (r) is given in the upper left corner. (B) The majority of changed exons show higher inclusion in the HNRNPC knockdown. Plot indicating the fold change (log₂) in normalized exon expression in KD1 against the sum of RNA-seq reads detecting the exon in KD1 and control samples. Exons that are called by DEXSeq are indicated in red (p values used for conditional thresholding p_a = 0.01 and p_b = 0.05). (C) The Alu exons show widespread upregulation in the HNRNPC knockdown. Plot as in (B). DEXSeq-called Alu exons are highlighted in red, and all other Alu exons are depicted in yellow. (D) The Alu exons arise from canonical splice sites. Weblogos indicating the consensus sequence at the 3′ and 5′ splice sites of the Alu exons. (E) Bar diagram summarizing the usage of different 3′ (black) and 5′ (gray) splice sites in the Alu consensus sequence (color coding as for enlarged sequence below). The region around the most commonly used 3′ splice sites is enlarged below (black arrowheads; the number of exons using each splice site is given below the arrowhead). (F) Scatter plot comparing changes in Alu exon inclusion (x axis) with changes in the corresponding transcript levels. The linear regression line (red) indicates the negative correlation between both data sets (Pearson’s product-moment correlation r = −0.231).

Figure S5

Polypyrimidine Tracts of Established Bona Fide Alu Exons Exhibit Shorter Contiguous U-Tracts, Related to Figures 3 and 4 (A) Established Alu exons accumulate point mutations that disrupt the long U-tracts. Heatmaps visualizing the polypyrimidine tract of Alu exons originating from the first (top) and second (bottom) arm of Alu elements (separated by a horizontal black line). Each line represents the polypyrimidine tract of an individual Alu exon, with the different colors indicating the four bases (T: red; C: blue; G: yellow; A: green). In the left panel, the 81 pyrimidine tracts of established bona fide Alu exons are shown that are highly included and show a less than 1.5-fold change in either direction, while the right panel shows the same regions of a randomly selected subset of hnRNP-C-repressed Alu exons. The exons were sorted according to the longest contiguous U-tract, as indicated in the bar chart on the right. (B) Established Alu exons harbor weaker hnRNP-C-binding sites. Box plot comparing the distribution of ranks of hnRNP-C-binding sites in front of 81 established or 81 randomly selected hnRNP-C-repressed Alu exons (p value < 0.1, Student’s t test). (C) U2AF65-binding sites at established Alu exons show hardly any change in occupancy upon knockdown of HNRNPC. Box plot depicting the ratio of U2AF65 occupancies from HNRNPC knockdown (KD) and control HeLa cells (Ctrl) of U2AF65-binding sites in front of 81 established or 81 randomly selected Alu exons (p value < 10⁻⁹, Student’s t test). (D) Control exons show little hnRNP C crosslinking and no increase in U2AF65 crosslinking in the HNRNPC knockdown. Heatmaps comparing the crosslinking of hnRNP C and U2AF65 relative to the 3′ splice site of the non-Alu exons downstream of the Alu exons shown in Figure 4. Each row corresponds to one of 652 exons. The left panel shows the crosslink events of hnRNP C at each nucleotide (nt) depicted in different shades of blue as indicated aside. The middle and right panels visualize U2AF65 crosslinking (different shades of purple) in control HeLa (Ctrl) and HNRNPC knockdown cells (KD). The bar diagram on the right (RNA-seq) shows the corresponding fold change in exon inclusion (KD1 / WT). Fold change values found to be significant according to the conditional thresholding approach are shown in red.

Figure S6

hnRNP C Suppresses the Exonization of an Alu Element in the NUP133 Gene through Competition with U2AF65, Related to Figure 5 (A) Genome browser view including an Alu element (orange) plus the downstream exon within the NUP133 gene on chromosome 1 (229,601,700-229,601,100). hnRNP C iCLIP data (blue) show crosslinking at the upstream and the linker U-tract of the Alu element (corresponding thymidines are highlighted in red at the bottom). Little U2AF65 crosslinking (purple) is detected at the U-tracts in control HeLa cells, while strong crosslinking is observed in the HNRNPC knockdowns (KD1 and KD2). RNA-seq data (green) show exonization of the Alu element in both knockdowns, giving rise to three different Alu exon variants. The corresponding isoforms are schematically indicated: Alu suppression in isoform S, Alu inclusion as a cassette exon in isoform 1 and 2 (differing in the usage of an upstream or downstream 5′ splice site, respectively) and Alu inclusion using an alternative 3‘ splice site for the downstream exon (isoform 3). The wild-type (WT) sequence surrounding the 3′ and 5′ splice sites (arrowheads) including the introduced point mutations (3^mut and 5^mut) are shown below. The positions of the point mutations are highlighted in the sequences by black squares. (B) U-to-C transitions promote Alu exonization in the presence of hnRNP C. Shown is the gel-like view of capillary electrophoresis of RT-PCR analysis of minigenes containing the Alu element in the NUP133 gene described in (A) with two flanking exons. The wild-type (WT) minigene shows no Alu exonization in control HeLa cells (Ctrl) and a significant increase in isoforms 1-3 in the HNRNPC knockdown (KD1 and KD2). A mutant minigene containing three T-to-C transitions in the upstream U-tract (3^mut) shows significant inclusion of isoforms 1-3 in control HeLa cells. Additional T-to-C transitions in the linker U-tract (5^mut) further elevate the inclusion of isoforms 1-3 in control HeLa cells and prevent any further regulation by hnRNP C (KD1 and KD2). The sizes corresponding to the different analyzed isoforms are schematically indicated on the right. (C) Average Alu exon inclusion in percent for three replicate RT-PCR experiments as described in (B). The different isoforms are indicated by shades of gray (light gray: isoform 1; medium gray: isoform 2; dark gray: isoform 3). Lines indicate relevant comparisons with asterisks indicating different levels of significance for changes in the summed inclusion isoforms (^∗: p value < 0.05; ^∗∗: < 10⁻³; n.s.: not significant; Student’s t test). Error bars represent SD of the mean; n=3.