Influence of Friedreich Ataxia GAA Noncoding Repeat Expansions on Pre-mRNA Processing

Plasmid Construction

The pEDA minigene was previously described³⁶ and contains the genomic fibronectin EDA exon, its flanking introns, and part of the −1 and +1 exons embedded in the human α-globin gene under the control of the SV40 enhancer. To generate the pEDA Nco, pEDA Nde, and pEDA BclI, we individually modified the corresponding unique restriction sites in the downstream EDA intron to the KpnI site by inserting specific oligonucleotides. We inserted a 40 bp spacer derived from pBS in the HindIII site of the second α-globin exon to obtain pEDA BglII. For generating the pBRCA1 ex18 minigene, a BRCA1 exon 17, exon 18, and exon 19 cassette was amplified by PCR and cloned in the BstEII site in the third exon of the α-globin minigene.³⁷ A frataxin cassette, which included 180 nucleotides of the first exon, 1170 bases of the flanking intron, the last 786 bases of the intron, and 78 bases of the frataxin exon 2, was amplified in two fragments with Fra395Dir and Fra1750RevKpn, Fra10314Dir, and Fra11100RevXba, respectively (oligonucleotide sequences are provided in the Supplemental Data), and cloned in pCDNA3 to obtain the pFrx minigene. One hundred GAA or TTC repeats were obtained by PCR amplification with primers C27 dir and C27Rev on genomic DNA and inserted in the unique KpnI site sites of pEDA and pFrx constructs. The same number of repeats were amplified with specific primers and cloned in the unique BamHI and EcoRI sites of pCF ex9³⁸ and pBRA,³⁷ respectively. The human genomic frataxin sequence containing 217 GAA repeats was amplified from human FRDA transgenic YG22 mouse DNA³⁹ with primers GAA1724Dir and GAA2192Rev, digested with KpnI, and cloned in sense and antisense orientation in the pEDANco minigene to generate pEDANcoGAA217 and pEDANcoTTC217. Oligonucleotides with fifteen GAA or TTC repeats were inserted in pEDA Nde to generate pEDANde15TTC and pEDANde15GAA. All minigenes were verified by sequencing, and the correct size of repeat insertions was checked for every plasmid preparation.

Transfection, RT-PCR Analysis, Northern Blotting, and Nuclear Run On Assay

COS7 cells (2.5 × 10⁵) or HeLa cells p100 (5 × 10⁶) were grown in standard conditions and transfected with Effectene reagent with 1.5 μg or 3 μg of plasmid DNA, respectively.⁴⁰ RNA extraction, RT-PCR, and quantitation of the amplified products were done as previously described.⁴⁰ For the analysis of spliced forms, pEDA minigenes were amplified with primers α2-3 and EDA5, pBRCA1 exon 18 was amplified with BRC90 and glo800, pCF exon 9 was amplified with α2-3 and BRA2³⁸, and pFrx was amplified with Fra395Dir and Sp6. For nascent transcript analysis, total RNA was digested with DNase RNase free (Roche), purified with the RNeasy kit (QIAGEN), and subsequently amplified with specific primers. For the EDA minigene, we used the following: for u1, EDA1207D and EDA1251Rev; for d1, EDA+1F and EDA1940Rev; and for d2, EDA2764Dir and glo800. For the pFrx minigene, we used the following oligonucleotides: for U, Fra395Dir and Fra606Rev; for D, Fra10902Dir and Sp6; and for D1, Fra10754Dir and Fra11037Rev. For turnover of splicing intermediates, we amplified the transfected minigenes with α936 and EDA1251R to detect the upstream intermediates and with EDA1207D and glo395 to detect the downstream intermediates. The relative abundance of spliced and unspliced forms in the Frx minigene was analyzed by RT-PCR with primers specific for the flanking exons (Fra395Dir and Sp6) and/or for the intronic sequences (Fra606Rev and Fra10902Dir). To detect the spliced and U unspliced forms, we amplified the transfected pFrx minigenes with Fra395Dir, Fra606Rev, and Sp6 primers. Amplification with Fra395Dir, Fra10902Dir, and Sp6 primers allowed detection of the spliced and D unspliced forms. The amplified products were resolved on 2% agarose gels, and the relative efficiency of splicing was measured as the ratio between the intensity of the band originating from the mature spliced transcript and that originating from the unspliced pre-mRNAs.

For Northern analysis, total RNA (2 μg) was resolved on 1.2% agarose-formaldehyde gel, and blotted membranes were hybridized with α³²PdCTP-labeled α-globin cDNA or with the 270 bp EDA fibronectin alternative spliced exon.

For NRO analysis, nuclei from HeLa-transfected cells were isolated in hypertonic lysis buffer (10 mM Tris HCl [pH 7.5], 10 mM NaCl, 2.5 mM MgCl₂, 0.5% NP40, and 10% sucrose) and resuspended in transcription buffer (40 mM Tris HCl [pH 7.9], 500 mM KCl, 10 mM MgCl2, 40% glycerol, and 2 mM DTT) in the presence of ³²P pUTP and ³²P pGTP for 15 min at 30°C. After RNA extraction, the labeled nascent RNAs were isolated and hybridized to antisense single-strand M13 probes slot blotted on nylon membrane. Radioactive signals were quantitated on a Cyclon Imager. We obtained the A-to-F M13 probes spanning the pEDA minigene by cloning the different PCR-amplified sequences in M13 plasmid. The VA M13 probe for normalization of transfection efficiencies was a generous gift from N. Proudfoot. Each transfection was VA normalized, and data were expressed for each probe as a percentage relative to the values obtained with the pEDA construct.

Affinity Purification of RNA-Binding Proteins: Pull-Down Assay

Two synthetic (GAA)₁₀ and (UUC)₁₀ RNA oligonucleotides (Integrated DNA Technologies) were used as targets for pull-down assays. An amount of 10 μg of either target RNA oligo was placed in 400 μl of reaction mixture (100 mM NaAcetate [pH 5.0] and Sigma 5 mM sodium m-periodate), incubated for 1 hr in the dark at room temperature, precipitated with ethanol, and resuspended in 100 μl of 100 mM sodium acetate (pH 5.0). Approximately 400 μl of adipic acid dehydrazide agarose beads 50% slurry (Sigma) previously equilibrated with 100 mM sodium acetate (pH 5.0) was added to each periodate-treated RNA, and the mix was incubated for 12 hr at 4°C on a rotator.

The beads with the bound RNA were then washed three times with 1 ml of 2 M NaCl, equilibrated in 1× solution D (20 mM HEPES [pH 7.5], 100 mM KCl, 0.2 mM EDTA, 0.5 mM DTT, and 6% glycerol) and incubated, in a final volume of 500 μl, with 0.5 mg of HeLa cell nuclear extract (C4, Biotech), 10× solution D (200 mM HEPES [pH 7.5], 2 mM EDTA, 5 mM DTT, and 60% glycerol), 160 mM KCl, and Heparin (final concentration 2.5 μg/μl), for 30 min on a rotator at room temperature. The beads were then washed six times with 1 ml of 1× solution D before the addition of SDS sample buffer and loading on SDS-12% polyacrylamide gels. Proteins were visualized by Coomassie brilliant blue staining. Protein sequence analysis of the bands excised from the gel was performed with an electrospray ionization mass spectrometer (LCQ DECA XP-ThermoFinnigam). Protein bands were digested with trypsin, and the resulting peptides were extracted with water and 60% acetonitrile/1% trifluoroacetic acid. Fragments were then analyzed by mass spectrometry, and proteins were identified by analysis of the peptide MS/MS data with Turbo SEQUEST (ThermoFinnigam) and MASCOT (Matrix Science).

Immunoblots were performed onto PVDF membrane (Amersham Bioscience), which was blocked by incubation in 5% dried milk in 0.1 M PBS, 0.1% Tween-20, or for Mab 1H4, in 1× Western Blocking Reagent (Roche). Membranes were probed with primary antibodies raised against hnRNPA1 and hnRNPA2 (kindly provided by R. Klima, ICGEB, Italy; 1:1000), PTB (a kind gift from E. Buratti, ICGEB, Italy; 1:1000), ZNF 9 (kindly provided by B. Cardinali; 1:1000), PABPN1 (a kind gift from E. Wahle, Halle, Germany; 1:1000), ASF/SF2 (Zymed Laboratories; 1:1000), SR proteins (Map 1H4, Zymed Laboratories; 1:1000) and Tra2β (Abcam; 1:1000). Immunoreactive bands were detected with enhanced chemiluminiscence reagent (ECL; Amersham Biosciences).

GAA Repeats Did Not Affect Transcription Elongation or Pre-mRNA Transcript Abundance in Transfected Cells

To test the effect of GAA repeats on PolII elongation and pre-mRNA transcript abundance, we used the alternatively spliced fibronectin pEDA minigene. This minigene has been extensively studied^36,41 and provides a good model for studying the process of cotranscriptional splicing and splicing regulatory sequences.^42–44 One hundred GAA or TTC repeats were introduced in the NcoI site of the pEDA construct (Figure 1A). This size expansion, found in affected individuals, has been shown to form abnormal DNA structures in vitro^13,14,45 and to inhibit prokaryotic polymerase transcription.¹⁶ We performed a nuclear run on (NRO) assay to determine whether some differences in the transcription level could be observed between minigenes carrying the GAA or TCC expansion and the empty minigene. Analysis was performed with single-strand, antisense DNA probes that cover the minigene sequences upstream and downstream of the triplet repeats (Figure 1A). Cotransfection with the VAI-expressing plasmid was used as a control for transfection efficiency. NRO signals obtained for the GAA, TTC, or empty pEDA minigenes were comparable relative to the VAI control signal, resulting in no changes in nascent transcription (Figure 1B). Quantitation of three independent VAI normalized transfections confirmed that the insertion of either the GAA or TTC repeat did not affect pre-mRNA transcription levels (Figure 1C), as shown by the presence of similar amounts of transcribed RNA along the two minigenes and the empty pEDA. We also evaluated the relative abundance of nascent transcript before and after the triplet insertion in each of pEDA minigenes by RT-PCR. Pre-mRNA sequences, upstream (u1) and downstream (d1 or d2) of the GAA/TTC insertion, were coamplified (u1 plus d1 and u1 plus d2) from pEDA Nco, pEDA NcoGAA, and pEDA NcoTTC transfections. The ratio of two bands resulting from of u1-d1 and u1-d2 coamplifications showed similar patterns in the pEDA Nco, pEDA NcoGAA, and pEDA NcoTTC, indicating that the type of insertion has no effect on the relative amount of nascent transcripts before and after the insertions (Figure 1D). Similarly, an increased number of triplets (n = 217) inserted in the pEDA minigene did not affect the relative abundance of the nascent transcripts before and after the insertion (Figure S1). In summary, these results clearly demonstrate that the insertion of a disease-causing GAA repeat expansion in a mammalian reporter gene did not affect the polII transcription elongation or pre-mRNA transcript abundance.

GAA Repeats Affect Pre-mRNA Splicing in a Position-Dependent Manner in a Heterologous Hybrid Minigene System

The lack of any effect on transcription in the pEDA NcoGAA minigene was unexpected, and we decided to explore in more detail the effect of the repeats on the mRNA biosynthesis. The GAA or TTC repeats (n = 100) were cloned in different intronic positions along the entire pEDA construct, and the resulting minigenes were transfected in COS cells. The quality of the resulting RNAs was evaluated in parallel by two complementary approaches: by RT-PCR analysis with specific primers (Figure 2C) and by Northern analysis (Figure 2D). RT-PCR analysis can provide an easy way to analyze aberrant splicing, but as a result of the intrinsic characteristics of PCR, large-intron retention or peculiar aberrant splicing products can be missed by this technique. Interestingly, the insertion of the GAA repeats in the NcoI position (pEDA NcoGAA) induced two major events in comparison with the corresponding TTC minigene: complete exclusion of the EDA exon, as can be seen by RT-PCR and Northern analysis (Figures 2C and 2D), and retention of upstream and downstream introns. This latter event can be well appreciated only by Northern blot (aberrant band number 1, Figure 2D). To rule out potential interference with splicing regulatory elements at the cloning Nco site, which is in close proximity of the EDA exon, we inserted the repeats 400 bp downstream at the NdeI site. pEDA Nde GAA again showed an aberrant splicing product. In RT-PCR the normal EDA inclusion product disappeared and was replaced by a higher MW band (band 4 in Figure 2C). Sequencing of band 4 showed that it contained the EDA exon and a cryptic exon, which surprisingly included the entire GAA sequence (Figure 2B). In addition, Northern analysis showed two more bands (numbers 3 and 3b in Figure 2D). Their size and hybridization pattern indicates that band 3 corresponds to complete retention of the downstream intron (Figure 2B), whereas the faint 3b band probably corresponds to intron retention of the segment upstream or downstream of the GAA insertion. Lastly, insertion of the repeats farther downstream had no effect on pre-mRNA processing (pEDA BclI constructs, Figure 2) suggesting that the number of GAA repeats affects splicing in a position-dependent manner.

The Number of Repeats Affects Abnormalities in Pre-mRNA Splicing

Because the number of GAA repeats is associated with disease severity in FRDA patients, we tested the effect of different repeat lengths on pre-mRNA splicing in hybrid minigenes. We created two additional pEDA minigenes by embedding 15 GAA or 15 TTC repeats within the NdeI restriction site and evaluated their splicing patterns. Compared to the construct with 100 TTC repeats, pEDA Nde(TTC)₁₅ had no effect on the normal splicing pattern (Figure 3A). On the contrary, pEDA Nde(GAA)₁₅ shifted the pre-mRNA processing toward normal splicing in comparison with pEDA Nde(GAA)₁₀₀, although this low number of repeats still has a pathological effect in the minigene in that it induces an aberrant splicing form with retention of the downstream intron (Figure 3, band 3).

GAA Repeats Affect Pre-mRNA Processing in Two Additional Heterologous Minigenes

To test whether the effect of the repeats on pre-mRNA processing was a peculiarity of the alternatively spliced fibronectin EDA minigene, we extended the study to two additional minigenes (Figure 4A). The GAA or TTC repeats were inserted downstream of the constitutively included BRCA1 exon 18 and downstream of the aberrant alternatively spliced CFTR exon 9 (Figure 4A). mRNAs were then evaluated by RT-PCR and Northern analysis. In the presence of the GAA repeats but not TTC, both BRCA1 exon 18 and CFTR exon 9 showed aberrant splicing products similar to those found in the EDA minigene. In particular, the pBRCA1 ex18GAA produces an aberrant band (number 1) that contains the GAA repeats followed by the downstream intronic sequences (Figure 4B). The pCF ex 9 GAA revealed a complex aberrant splicing pattern in which exon 9 was skipped, GAA was used as a cryptic exon, and downstream introns were retained (Figure 4). These different aberrant splicing products were also observed in Northern blots (Figure 4D) and are represented schematically in Figure 4E.

Turnover of Pre-mRNA Splicing Intermediates in the pEDA Nco GAA Minigene

To investigate the GAA-dependent aberrant splicing mechanism more in detail, we evaluated the turnover of pre-mRNA splicing intermediates in pEDA NcoGAA and pEDA NcoTTC minigenes. Preferential activation of one pre-mRNA splicing intermediate (i.e., of an upstream or downstream intron) has been recently shown to regulate splice-site selection through differential binding of a neuronal specific splicing factor to intronic or exonic sequences.³⁵ In order to study the splicing intermediates in the pEDA Nco minigenes carrying the TTC or GAA repeats, we used two pairs of primers: the 1207D and glo395R pair and the α936 and EDA1251R pair. The first pair of primers detects intermediates containing the intron upstream of the EDA exon, whereas the second primer pair detects intermediates containing the intron downstream of the EDA exon. Amplification with EDA 1207D and glo395R showed a band corresponding to the removal of the TTC-containing intron of the downstream intermediate only in the pEDA TTC minigene (Figure 5B). The absence of any amplified band in pEDA NcoGAA is consistent with mature mRNA in the Northern analysis, in which this minigene showed complete exon skipping and the retention of the GAA-containing intron. On the contrary, amplification with α936 and EDA1251R showed two bands in both pEDANco GAA and pEDA NcoTTC plasmids (Figure 5C, lanes 1 and 2). The major upper band corresponds to the splicing intermediate in which the GAA- or TCC-containing intron is present and all the upstream introns are correctly removed (iUP) (Figure 5C, lanes 1 and 2, respectively), whereas the faint lower band was due to partial skipping of the −1 hybrid globin-fibronectin exon. In light of the fact that the iUP intermediate represents one of the two alternative pathways leading to exon inclusion (i.e., the intermediate in which the upstream intron is skipped), its presence in the GAA minigene was unexpected because the mature mRNA completely lacks the EDA (Figure 2). To quantify the relative amount of the iUP pre-mRNA, we set up a cotransfection experiment with the pEDABglII minigene. This pEDABglII internal-control minigene without repeats contains a 40-bp-long insertion in the second α-globin exon, and RT-PCR amplification results in a slightly higher intermediate, iUPBg (Figure 5C, lane 3). When equal amounts of the pEDA NcoGAA and pEDA NcoTTC were individually cotransfected with the same amount of pEDABglII (250 ng), the iUPBg intermediate was only evident in the pEDA NcoTTC lane, suggesting that the iUP produced by the pEDA NcoGAA is significantly reduced (Figure 3C, compare lanes 4 and 5). To confirm these data, we cotransfected pEDA NcoGAA with increasing amounts of pEDABglII (1 μg and 2.5 μg) (Figure 5, lanes 6 and 7). In light of the iUP intermediate band, cotransfection experiments showed that approximately ten times more pEDABglII plasmid is necessary to compete with the the pEDA NcoGAA than with the pEDA NcoTTC minigene (Figure 5C, lanes 5–7). Thus, the amount of iUP is significantly higher in the pEDA NcoGAA than in the pEDA NcoTTC. This indicates that the GAA repeats induce a block in the turnover of the upstream intermediate, which accumulates and is not processed into mature EDA+ mRNA, as shown by Northern blot analysis (Figure 2D). One possibility is that the preferential accumulation of pre-mRNA splicing intermediates might occur through binding of trans-acting splicing factors in the intron.³⁵

Binding Properties of GAA-Repeat Transcripts

In order to identify trans-acting factor(s) whose binding to GAA repeats could give us an explanation for the observed changes in the splicing pattern, we performed a pull-down analysis with synthetic RNA oligonucleotides containing either 10 GAA or 10 UUC trinucleotide repeats. We detected at least four protein bands that were enriched in the GAA-repeat fraction (data not shown). These protein bands of approximately 50, 35, and 18 kDa were excised, analyzed by mass spectrometry, and eventually identified as PABPN1 (Uniprot accession number Q86U42), a protein that is involved in nuclear polyadenylation, might contribute to mRNA export⁴⁶, and has an antiapoptotic role⁴⁷; as hnRNPA1 and hnRNPA2 (Uniprot accession numbers P09651 and P22626, respectively), a well-known factor involved in splicing inhibition³¹; and as ZNF9 (Uniprot accession number P62633), whose function is yet to be understood.⁴⁸ The protein doublet associated with UUC oligonucleotide corresponds to the PTB (Uniprot accession number P26599), an important regulatory splicing protein, whose prevalent inhibitory effect on alternative splicing as a result of the recognition and binding to UC-rich sequences has already been well established.⁴⁹ The differential binding properties of these proteins were further confirmed by immunoblot analysis (Figure 6). We have also included in the analysis some members of SR protein family because these splicing factors have been previously reported to specifically interact with GAA-rich exonic-splicing regulatory elements.^41,50,51 Our results indicate that the GAA repeats represent the target for many different factors involved in pre-mRNA processing. More precisely, we observed their binding capacity not only to PABPN1, hnRNPA1, hnRNPA2, and ZNF9 proteins but also to several SR proteins (SF2/ASF, SRp30c [Uniprot accession number Q13242], SRp40 [Uniprot accession number Q13243], SRp55 [Uniprot accession number Q13247], and Tra2β [Uniprot accession number P62995]), yet notably not to SRp20 (Uniprot accession number P84103), which interacted with the UUC oligonucleotides (Figure 6). A similar association between GAA-rich RNA molecules and splicing factors (including several orthologous SR proteins and PAPBN1) has been detected also in Xenopus oocyte extracts.^52,53

GAA Repeats Inhibit Splicing in the Frataxin Minigene

We tried to assess whether the observed effect of GAA repeats could be reproducible when the triplets were inserted in the frataxin minigene. The main characteristic of this minigene is that the GAA repeats are inserted in the first intron, which does not allow the potential skipping of the upstream exon. We made a frataxin minigene construct that carried exon 1 with 1170 bases of downstream intronic sequences and exon 2 with 786 bases of upstream intronic sequences under transcriptional control of the CMV promoter (Figure 7A). Either this empty minigene (Ø) or the (GAA)₁₀₀- and (TTC)₁₀₀-containing minigenes were transfected into COS cells. In order to evaluate the relative abundance of spliced and unspliced forms, we analyzed these constructs by RT-PCR with primers specific for the flanking exons and/or for the intronic sequences. We conducted the analysis with coamplification experiments to detect the relative efficiency of splicing measured as the ratio between the mature spliced transcript and the unspliced pre-mRNAs (see Material and Methods). Figure 7 clearly shows that the relative splicing efficiency of the frataxin minigene was significantly inhibited by the GAA repeats. In fact, the spliced-to-unspliced ratio is markedly reduced in the GAA minigene, and this is evident for both the upstream (S/U) and downstream (S/D) ratio. This effect was GAA-sequence specific given that the construct with the TTC triplets showed a splicing efficiency comparable to that of the empty one. Additional controls showed that there was no interference from contaminating DNA (lanes with minus RT) and that there was no background from endogenous Frataxin mRNA (data not shown).

In parallel, we also evaluated in the frataxin minigene the relative abundance of the nascent transcripts upstream and downstream of the triplet insertion (U and D1 in Figure 7A). As found for the EDA minigene (Figure 1D), the two bands resulting from U and D1 coamplifications showed a similar ratio independently of the type of insertion, indicating that the GAA repeats did not change the mRNA transcript abundance in the frataxin minigene (Figure 7C). On the other hand, analysis by RT-PCR and Northern blot did not reveal any significant effect on the mature mRNA (data not shown), indicating that the GAA repeats in their original context do not affect the quality of the final transcript. Thus, we conclude that the GAA-repeat-rich region can act as an intronic-splicing regulatory element that inhibits the splicing efficiency of the first frataxin intron.