Javascript required
Skip to content Skip to sidebar Skip to footer

During Reverse Transcription the Template Is Read

  • Loading metrics

Credible Non-Canonical Trans-Splicing Is Generated past Reverse Transcriptase In Vitro

  • David Tollervey

Apparent Not-Canonical Trans-Splicing Is Generated by Reverse Transcriptase In Vitro

  • Jonathan Houseley,
  • David Tollervey

PLOS

x

  • Published: August 18, 2010
  • https://doi.org/10.1371/journal.pone.0012271

Abstract

Groundwork

Trans-splicing, the in vivo joining of two independently transcribed RNA molecules, is well characterized in lower eukaryotes, but was long thought absent from metazoans. However, recent bioinformatic analyses of EST sequences suggested widespread trans-splicing in mammals. These apparently spliced transcripts generally lacked canonical splice sites, leading us to question their actuality. Specially, the native ability of reverse transcriptase enzymes to template switch during transcription could produce manifestly trans-spliced sequences.

Principal Findings

Here we report an in vitro system for the assay of template switching in reverse transcription. Using highly purified RNA substrates, nosotros testify the reproducible occurrence of credible trans-splicing betwixt ii RNA molecules. Other reported non-canonical splicing events such as exon shuffling and sense-antisense fusions were also readily detected. The latter caused the production of apparent antisense non-coding RNAs, which are besides reported to be abundant in humans.

Conclusions

We propose that well-nigh reported examples of non-approved splicing in metazoans arise through template switching by reverse transcriptase during cDNA preparation. We farther show that the products of template switching tin vary between opposite transcriptases, providing a simple diagnostic for identifying many of these experimental artifacts.

Introduction

Reverse transcriptases (RTs) are enzymes that synthesize complementary DNA (cDNA) from an RNA template, and have evolved in retroviruses to catechumen single stranded viral RNA into double stranded Deoxyribonucleic acid for integration into host genomes. They are an invaluable tool for molecular biological science, existence used to copy RNA into Deoxyribonucleic acid for assay by PCR (RT-PCR), microarrays and high throughput sequencing. It is possible to sequence RNA direct, however, nigh of the experimentally determined RNA sequences, and all loftier-throughput data, accept been generated past RT-based protocols.

RTs lack a proof reading activity (reviewed in [1]) and consequently typically show a fidelity of nucleotide incorporation that is orders of magnitude lower than that of Dna polymerases. This generally causes few problems, every bit comparison of RNA and genomic DNA sequences allows like shooting fish in a barrel identification of base of operations substitutions. Withal, less hands detectable sequence errors tin be introduced by another intrinsic property of RTs. Retroviral replication is known to require two template switches, where RT 'jumps' to some other template location without terminating Deoxyribonucleic acid synthesis [two], and this ability is also implicated in high retroviral mutability [iii]. Template switching has been repeatedly implicated in the ascertainment of apparent intramolecular splicing events [four], [v], [6], [7], [eight], and prove for its involvement in apparent intermolecular trans-splicing has too been reported [nine], although this is disputed [ten]. A key observation regarding these credible splicing events is that they occur betwixt not-approved splice sites that oftentimes share short homologous sequences. This presumably reflects a requirement for homology betwixt the nascent transcript and the acceptor site to allow RT to prime number continued cDNA synthesis after template switching [4].

Trans-splicing of common mRNA leader sequences has long been known to occur in trypanosomes, nematode worms and bounding main squirts (reviewed in [11]), but appeared to be very rare in mammalian cells (reviewed in [12]). Unexpectedly, still, bio-informatic analyses of mammalian transcripts reported large numbers of ostensible trans-splicing events [thirteen], [14], [15], [16]. The observation that trans-spliced products could be detected from virtually 50% of human genes [14] provided the central evidence underlying the recent suggestion that trans-splicing is a frequently used method of increasing transcriptome complexity in higher eukaryotes [17]. If existent, these trans-splicing events must utilise an as all the same undiscovered splicing mechanism as the exons involved mostly lacked canonical splice sites. Notably, however, they often showed short homologous sequences at the donor and acceptor sites [14].

Here we report the development of an in vitro system to written report the occurrence of template switching events during reverse transcription. Our information greatly extend the range of substrates that can be considered likely to be formed past opposite transcriptase antiquity when encountered as part of high throughput sequencing data sets.

Results

During the RT-PCR analysis of a yeast non-coding RNA, IGS1 R [eighteen], nosotros detected an apparent splicing issue removing a 117 nt intron from about 30% of transcripts (Fig. 1A). Surprisingly, the putative intron lacked conserved sequences usually present at the intron co-operative signal, 5′ and 3′ splice sites, which are highly conserved between yeast pre-mRNAs. It was, however, flanked by 2 short homologous sequences predicted to lie at the base of a hairpin in the unspliced RNA (Fig. 1B). Previous analyses had suggested that apparent splicing might arise from template switching by RT and we therefore tested whether changing the contrary transcription conditions would alter the consequence. Increasing the reaction temperature has been reported to suppress template switching [4], [vii] simply had no upshot on the apparent abundance of spliced IGS1 R (Fig. 1C). Yet, the putative spliced product was observed following RT-PCR using Superscript II (a Moloney Murine Leukemia Virus derived RT) only not with AMV (an Avian Myeloblastosis Virus derived RT) (Fig. 1D). This demonstrated that the apparent splicing of IGS1 R arises from an RT artifact, which is dependent on the specific RT used.

thumbnail

Effigy 1. An apparent not-canonical intron in the IGS1 R non-coding RNA.

A: 35 cycle RT-PCR beyond the credible intron on cDNA synthesized with Superscript Ii and genomic DNA. cDNA was produced from a trf4Δ strain where this non-coding RNA is stabilized. B: Hairpin construction of IGS1 R, short homologous repeats are underlined in grey. C: 35 cycle RT-PCR across the apparent intron on trf4Δ cDNA synthesized using Superscript II at 42°C or Superscript III at 55°C. D: 35 cycle RT-PCR across the credible intron on trf4Δ cDNA synthesized using Superscipt 2 or AMV. Control shows 30 bicycle RT-PCR reaction across the ASC1 mRNA intron on the aforementioned cDNA samples.

https://doi.org/x.1371/periodical.pone.0012271.g001

The ascertainment that non-canonical splicing events tin can reproducibly occur betwixt short repeated sequences atomic number 82 us to question whether many recently reported trans-splicing events are in fact due to template switching artifacts. Proving that whatsoever particular splicing event does non occur at low levels is problematic, so we instead attempted to reproduce apparent trans-splicing using RT in vitro. From the five budding yeast trans-splicing events reported by Li et al. (2009), we arbitrarily selected GenBank sequence M14410, a fusion between KRE29 and HXK1, every bit a substrate for in vitro analysis (Fig. 2A). Regions spanning a few hundred base of operations pairs either side of the credible trans-splicing sites in both genes were amplified from genomic DNA and cloned, providing sequence-verified Deoxyribonucleic acid templates. RNAs were transcribed with T7 RNA polymerase and purified past gel extraction (Fig. 2B). The ii individual RNA molecules were mixed, diluted one∶thou with HeLa full RNA, then reverse transcribed from random hexamers using Superscript Two. The substrate RNAs were diluted in HeLa full RNA to mimic the high complexity of the RNA population in existent RT reactions, and to ensure that template switching was non being driven by the presence of only the donor and recipient.

thumbnail

Figure 2. An in vitro system for the assay of apparent trans-splicing.

A: HXK1 and KRE29 substrate RNAs showing primer locations. Template switching events produced by Superscript 2 and AMV are indicated. B: Purified substrate RNAs. C: RT-PCR using primers complementary to each RNA on three independent RT reactions (lanes ane–3), and a no RT control (Lanes 4–5). The template for the Deoxyribonucleic acid control (lanes half-dozen–7) was HeLa cDNA with restriction fragments encompassing the entire sequence of the substrate RNAs. Upper panel 35 cycles, other panels 25 cycles. D: PCR reactions performed as in C on cDNA produced with AMV reverse transcriptase. Upper panel 35 cycles; lower panels 25 cycles.

https://doi.org/ten.1371/journal.pone.0012271.g002

PCR was performed with primers designed to notice trans-splicing events and this produced the same product in three independent RT reactions performed on three unlike occasions (Fig. 2C lanes 1–3). Sequencing of this product revealed an apparent trans-splicing event from near the end of the HXK1 RNA to the middle of the KRE29 RNA. Formation of this product required RT (Fig. 2C lanes 4–5), and was not a PCR artifact, as it was not amplified from HXK1 and KRE29 Deoxyribonucleic acid mixed with HeLa cDNA (Fig. 2C lanes 6–seven). Using AMV RT, multiple RT-PCR reactions yielded different products to those observed with Superscript II (Fig. 2D), which were shown to represent at least three unlike apparent trans-splicing events by sequencing (Fig. 2A). Nosotros conclude that both Superscript and AMV reproducibly generate credible trans-spliced products on the HXK1 and KRE29 template pair, just with singled-out preferred fusion sites.

Ostensible, not-approved trans-splicing events evidence a significant bias towards splicing between transcripts from the same locus. This has been taken to support their authenticity, since these sequences would be in shut proximity in vivo but not in the RT reaction [14]. These events are classified equally either exon shuffles (where exon order in the transcript differs from that in the genomic Deoxyribonucleic acid), or fusions between sense mRNA and antisense not-coding RNA. Notwithstanding, the ability of contrary transcriptase to jump frontwards on a template (yielding apparent non-canonical cis-splicing), suggested that backwards jumps could generate exon shuffles. Moreover, trans-splicing between sense and antisense transcripts could be formed by a template switch from the RNA to the cDNA being produced by another RT on the same RNA (Fig. 3A).

thumbnail

Figure iii. In vitro formation of sense-antisense fusions.

A: Proposed machinery of sense-antisense fusion formation. B: Schematic of SPT7 RNA showing primer bounden sites and observed sense-antisense fusions. C: Purified SPT7 substrate. D: RT-PCR experiments performed on SPT7 substrate performed as in Fig. 2C. Upper panel shows a 32 cycle PCR reaction, other panels show 25 cycles. E: Schematic of SPT7 RNA showing primer binding sites and observed exon shuffling events. F: RT-PCR experiments performed as in d. Sequenced bands are indicated past *.

https://doi.org/10.1371/journal.pone.0012271.g003

To test these possibilities, we arbitrarily selected another yeast clone, GenBank sequence T37598, representing a sense-antisense fusion produced from the SPT7 locus (Fig. 3B). As before, the region surrounding the apparent splice site was amplified from genomic DNA, cloned, transcribed, purified (Fig. 3C), and diluted with HeLa RNA prior to reverse transcription. To observe sense-antisense fusions, PCR reactions were performed using two primers complementary to the same Deoxyribonucleic acid strand. This consistently generated the same prepare of product bands (Fig. 3D lanes ane–3). Sequencing of prominent bands from ii independent RT-PCR reactions revealed multiple sense-antisense fusion events (depicted in Fig. 3B). Formation of these products required RT enzyme and did not occur during PCR on a DNA template (Fig. 3D lanes 4–seven). Therefore, sense-antisense fusion events readily and reproducibly occur during opposite transcription in vitro.

A different primer pair was designed to discover exon shuffling (Fig. 3E). This consistently detected multiple species from the same opposite transcription reactions (Fig. 3F). Sequencing of the two clearly defined bands (marked * in Fig. 3F) confirmed the occurrence of apparent exon shuffling events involving SPT7 RNA (Fig. 3E). We conclude that both types of trans-splicing seen at a single locus can exist readily reproduced on a purified template in vitro using reverse transcriptase.

As for the IGS1 intron, reverse transcription temperature did not alter the observance of sense-antisense fusions (Fig. 4A). In contrast, the prominent bands representing both types of trans-splicing at a single locus were not observed when AMV was substituted for Superscript Two, although PCR products were still obtained, suggesting that some template switching occurs (Fig. 4B). Even so, the affluence of these products was too depression for us to sequence, and so we cannot dominion out their arising from PCR mis-priming. The fact that prominent template-switching events were not obtained with AMV excludes the possibility that sense-antisense RNAs are produced by T7 RNA polymerase during transcription and survive the gel extraction stride. Were this the example they should be amplified with like efficiency past either RT.

thumbnail

Effigy 4. Generality of trans-splicing artifacts.

A: PCR reactions for detecting sense-antisense fusion were performed on SPT7 substrate RNA reverse transcribed with Superscript Ii or III at the given temperatures. Upper panel 30 bicycle PCR reactions; lower panel 25 cycles. B: PCR reactions for detecting sense-antisense fusion (upper panel) and exon shuffling (middle console) were performed on SPT7 substrate RNA reverse transcribed with Superscript II or AMV. Upper panels 32 cycle PCR reactions; lower panel 25 cycles. C: RT-PCR using primers to detect sense-antisense fusions on poly(A) tailed RNA. Upper to lower panels bear witness 32, 25 and 30 cycle PCRs. D: PCR reactions performed equally in B using Superscript Two, products from two RT reactions with and two RT reactions without vi µg ml−one Actinomycin D are shown.

https://doi.org/ten.1371/journal.pone.0012271.g004

Our proposed machinery for the germination of sense-antisense fusions requires two RT enzymes to be active on the same RNA molecule. This volition occur frequently if opposite transcription is primed from random hexamers, only is expected to exist less common when oligo(dT) is used to prime synthesis from the poly(A) tail. To exam the effect of this modify, SPT7 RNA was incubated with ATP in the presence or absenteeism of Eastward. coli poly(A) polymerase to add together a poly(A) tail. These substrates were then used in vitro for contrary transcription every bit above but primed from oligo(dT). This produced the same pattern of products seen in previous experiments, which now depended on the presence of poly(A) polymerase (Fig. 4C). The pp1a control (a homo mRNA present in the HeLa RNA) is presented to testify that RT efficiency was like in the presence and absence of poly(A) polymerase. Note that some reverse transcription of the SPT7 RNA however occurs in the absence of poly(A) polymerase due to priming of the oligo(dT) on short, encoded oligo(A) stretches in the substrate. These data show that template switching events detected in random hexamer primed RT reactions also occur during oligo(dT) primed cDNA synthesis.

Actinomycin D was at once routinely added to RT reactions to suppress the formation of sense-antisense fusions acquired past RT changing strand on hairpin structures at the iii′ end of the cDNA. This was, yet, thought unnecessary after the introduction of RT lacking RNase H activity [xix]. Although this ability is weak in RNase H deficient enzymes, it has recently been shown that actinomycin D can suppress the formation of some artifactual antisense RNAs [20]. Addition of actinomycin D to SPT7 reactions reduced the aberrant products in some experiments, but conspicuously did not eliminate template switching (Fig. 4D).

Whereas the patterns of bands observed on gels following in vitro reactions were highly reproducible, sequencing of multiple products rarely revealed identical splice sites. Rather, the fusion sites varied by minor numbers of nucleotides (Tabular array S1). Similarly, the precise splice sites observed in the private GenBank clones selected were non observed, but fusions were observed in close vicinity.

Discussion

Reverse transcriptases have been invaluable tools in RNA analyses. It is, however, articulate that these enzymes are error prone and the frequent introduction of point mutations past RT has been widely recognized. In contrast, their ability to generate artifacts that resemble splicing products remains largely unappreciated, despite existence first reported many years ago [21]. One consequence of template switching is the germination of sense-antisense fusion transcripts. This would lead to the detection of apparent antisense ncRNAs in loftier throughput experiments. Reported antisense ncRNAs that share the splicing pattern of the cognate sense mRNA are particularly likely to be artifacts [22]. Most template switching events are rare merely the huge volume of transcriptome data currently beingness produced ensures that their contamination of cDNA databases will increment. Moreover, on particularly skillful substrates, such as the yeast IGS1 R ncRNA or the FOXL2 mRNA [4], template switching occurs in a large fraction of cDNAs produced.

In our easily, the different template switching propensities of Superscript and AMV provided a useful diagnostic tool for identifying artifactual splicing events. More often than not, withal, our data show that all putative not-approved splicing events and antisense ncRNAs require verification by non-RT based methods, e.g. northern blot or RNase protection, prior to their inclusion in farther analyses. Other known methods to suppress template switching, notably elevated reverse transcription temperature and actinomycin treatment, failed to suppress SPT7 sense-antisense fusion.

It is worth noting that some cases of trans-splicing observed in mammalian cells have been verified by non-reverse transcriptase based methods [23], [24]. However, these events occurred at canonical splice sites in contrast to the vast majority of reported trans-splicing events. Nosotros did not find splicing at canonical splice sites in our in vitro system, and almost events occurred between brusk directly repeats. However, direct repeats were not an absolute requirement, peculiarly for AMV, equally we detected a number of trans-splicing events with little or no visible homology between donor and acceptor sequences.

Here nosotros have confirmed a previous, controversial, report that reverse transcriptase can generate apparent trans-splicing [9]. We extended this analysis to prove that two other oftentimes encountered non-colinear splicing events, exon shuffling and sense-antisense fusion, tin can also be generated every bit reverse transcriptase artifacts. Furthermore we present a simple test for identifying many template switching events based on comparison of MMLV and AMV reverse transcription products.

Materials and Methods

Substrates for in vitro assays were amplified from genomic DNA with Phusion (NEB) and cloned into pGEM-T (Promega). Oligonucleotides used were HXK1 F1/R1 for HXK1, KRE29 F1/R1 for KRE29 and SPT7 F1/R1 for SPT7; sequences of oligonucleotides are given in Table 1. Plasmids were linearized with XhoI and 1 µg transcribed using T7 RNA polymerase (NEB) for two h at 37°C. Gels were run in 1x TBE, acrylamide gels contained 8 Thou urea. Gels were stained with SYBR Condom and imaged using a Fuji FLA5100 scanner. RNA was eluted from acrylamide gel slices by crushing and soaking for 4 h in 0.five Thousand NaOAc/one mM EDTA/0.one% SDS, followed by phenol/chloroform extraction and ethanol precipitation with 1 µg glycogen. PCR reactions on reverse transcribed fabric were performed with Phire (Neb), details of cycle number are given in individual figure legends. Annealing temperature was 50°C for IGS1 and HXK1/KRE29 PCR and 53°C for SPT7 PCR. For poly(A) tailing, 50 ng RNA was incubated with 5U poly(A) polymerase (Neb) and 1 mM ATP, so cleaned on QIAQuick columns (QIAgen).

Superscript II RT: 0.5 ng substrate RNA, 500 ng HeLa RNA (Invitrogen), 125 ng random hexamers and 0.5 µl 10 mM dNTPs in six.5 µl full volume were denatured at 65°C for 5 min before 2 min on ice. 2 µl 5x starting time strand buffer and 1 µl of 0.one M DTT were added followed by 0.five µl (100 U) Superscript Two (Invitrogen). Reactions were incubated 10 min at room temperature, 42°C for l min and seventy°C for fifteen min. For oligo(dT) priming, 250 ng oligo(dT)xviii was added in identify of hexamers, and reactions were heated to 42°C prior to enzyme addition. Superscript III reactions were performed as per manufacturer's instructions at the indicated temperatures. Dna template controls were cDNA from 500 ng HeLa RNA produced as to a higher place, with 0.5 ng gel purified XhoI-PvuI fragments of the template plasmids. Where indicated, Actinomycin D (Calbiochem) was added to Superscript 2 RT reactions afterward the 65°C step at 6 µg ml−i from a 1 mg ml−1 stock solution.

AMV RT: 0.5 ng RNA, 500 ng HeLa RNA (Invitrogen) and 125 ng random hexamers in full volume 8.25 µl were heated 5 min at 70° and 5 min on water ice. i.25 µl 10 mM dNTPs and ii.5 µl 5x buffer were added followed by 0.5 µl (5 U) AMV (Promega). Reactions were incubated for 1 h at 37°C.

Supporting Data

Table S1.

Sequencing information. All sequences obtained in this projection are shown. Regions of the sequence take been colour coded red, green and blue to indicate that they emanate from different molecules or unlike regions of the same molecule. Overlapping regions are shown in purple. Sequences were obtained either by direct sequencing of band-purified PCR products, or for complex PCR products by sequencing multiple clones ligated in pGEM-T.

https://doi.org/10.1371/periodical.pone.0012271.s001

(0.04 MB DOC)

Acknowledgments

We would similar to thank Nicola Stead for critical reading of the manuscript and the Schirmer group for reagents.

Author Contributions

Conceived and designed the experiments: JH DT. Performed the experiments: JH. Analyzed the information: JH DT. Wrote the newspaper: JH DT.

References

  1. 1. Svarovskaia ES, Cheslock SR, Zhang West-H, Hu Westward-Southward, Pathak VK (2003) Retroviral mutation rates and reverse transcriptase fidelity. Frontiers in bioscience 8: d117–134.
  2. 2. Gilboa E, Mitra SW, Goff Due south, Baltimore D (1979) A detailed model of reverse transcription and tests of crucial aspects. Cell eighteen: 93–100.
  3. 3. Temin HM (1993) Retrovirus variation and reverse transcription: aberrant strand transfers result in retrovirus genetic variation. Proc Natl Acad Sci U S A ninety: 6900–6903.
  4. 4. Cocquet J, Chong A, Zhang G, Veitia RA (2006) Reverse transcriptase template switching and false alternative transcripts. Genomics 88: 127–131.
  5. 5. Geiszt M, Lekstrom G, Leto TL (2004) Analysis of mRNA transcripts from the NAD(P)H oxidase 1 (Nox1) gene. Prove against production of the NADPH oxidase homolog-i short (NOH-1S) transcript variant. J Biol Chem 279: 51661–51668.
  6. 6. Mader RM, Schmidt WM, Sedivy R, Rizovski B, Braun J, et al. (2001) Contrary transcriptase template switching during opposite transcriptase-polymerase concatenation reaction: bogus generation of deletions in ribonucleotide reductase mRNA. J Lab Clin Med 137: 422–428.
  7. 7. Ouhammouch M, Brody EN (1992) Temperature-dependent template switching during in vitro cDNA synthesis by the AMV-reverse transcriptase. Nucleic Acids Res 20: 5443–5450.
  8. viii. Roy SW, Irimia M (2008) When good transcripts get bad: artifactual RT-PCR 'splicing' and genome analysis. Bioessays 30: 601–605.
  9. ix. Zeng XC, Wang SX (2002) Prove that BmTXK beta-BmKCT cDNA from Chinese scorpion Buthus martensii Karsch is an antiquity generated in the reverse transcription procedure. FEBS Lett- 520: 183–184; author reply 185.
  10. 10. Zhu S, Li W, Cao Z (2002) Does MMLV-RT lacking RNase H activeness take the capability of switching templates during contrary transcription? FEBS Letters 520: 185–185.
  11. 11. Hastings KEM (2005) SL trans-splicing: like shooting fish in a barrel come or piece of cake go? Trends in genetics 21: 240–247.
  12. 12. Horiuchi T, Aigaki T (2006) Alternative trans-splicing: a novel style of pre-mRNA processing. Biology of the prison cell 98: 135–140.
  13. 13. Herai RH, Yamagishi ME (2010) Detection of human interchromosomal trans-splicing in sequence databanks. Brief Bioinform 11: 198–209.
  14. 14. Li X, Zhao L, Jiang H, Wang Due west (2009) Curt homologous sequences are strongly associated with the generation of chimeric RNAs in eukaryotes. J Mol Evol 68: 56–65.
  15. 15. Niu X, Luo D, Gao Due south, Ren G, Chang 50, et al. (2010) A conserved unusual posttranscriptional processing mediated past short, direct repeated (SDR) sequences in plants. J Genet Genomics 37: 85–99.
  16. 16. Unneberg P, Claverie JM (2007) Tentative mapping of transcription-induced interchromosomal interaction using chimeric EST and mRNA data. PLoS 1 ii: e254.
  17. 17. Gingeras TR (2009) Implications of chimaeric non-co-linear transcripts. Nature 461: 206–211.
  18. 18. Houseley J, Kotovic K, El Hage A, Tollervey D (2007) Trf4 targets ncRNAs from telomeric and rDNA spacer regions and functions in rDNA copy number control. Embo J 26: 4996–5006.
  19. 19. Sambrook J, Russell DW (2001) Molecular cloning: a laboratory manual. Cold Spring Harbor, N.Y.: Cold Jump Harbor Laboratory Press.
  20. 20. Perocchi F, Xu Z, Clauder-Munster Due south, Steinmetz LM (2007) Antisense artifacts in transcriptome microarray experiments are resolved by actinomycin D. Nucleic Acids Res 35: e128.
  21. 21. Luo GX, Taylor J (1990) Template switching by opposite transcriptase during DNA synthesis. Journal of virology 64: 4321–4328.
  22. 22. Kapranov P, Drenkow J, Cheng J, Long J, Helt Grand, et al. (2005) Examples of the complex architecture of the human transcriptome revealed by RACE and loftier-density tiling arrays. Genome Res fifteen: 987–997.
  23. 23. Li H, Wang J, Mor G, Sklar J (2008) A neoplastic gene fusion mimics trans-splicing of RNAs in normal human cells. Science 321: 1357–1361.
  24. 24. Rickman DS, Pflueger D, Moss B, VanDoren VE, Chen CX, et al. (2009) SLC45A3-ELK4 is a novel and frequent erythroblast transformation-specific fusion transcript in prostate cancer. Cancer Res 69: 2734–2738.

abbottcomitaxby.blogspot.com

Source: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0012271