Exon painting working group
Exon painting signal in CAGE
Purpose of this page is to accumulate evidence and analyses on the exon-painting phenomenon. This was identified 7 years ago and is still not fully addressed.
History
Identified in FANTOM3.
Exerts from F3 mailing list.
- Original observation File:FANTOM3.pdf and powerpoint file from July 2004 File:Painting.ppt
- Followup comments File:FANTOM3 a.pdf
- Oligo-cap RACE confirmation of some painting signal File:FANTOM3 b.pdf
Possible explanations
1. capture of true capped 5' ends from initiation events
2. capture of true capped 5' ends from recapping
3. capture of non-capped 5' ends
4. capture of partial cDNA (ie. did not reach the 5' end).
Evidence supporting each explanation
1 or 2. Majority of signal is capped. Oligo-cap RACE confirms some of these events. Promoter-reporter constructs derived from 3'exonic/UTR sequences drive reporter expression (FANTOM3 papers). Recapping work by Gingeras lab (anti-cap antibody validations).
3. rRNA is still found in our libraries. TEX, and more specifically T4PNK phosphorylation then TEX treatment of total RNA reduces exon painting signal
4. ???
FANTOM5 helicos CAGE control experiments
ActinomycinD TEX (5' phosphate dependent exonuclease) T4PNK (polynucleotide kinase to add a phosphate to the 5') File:Tex experiment.xls
Below is a link to three loci for demonstration.
Observations:
1. Actinomycin D removes the majority of the exonic antisense signal (we have shown this in the hCAGE paper). This is consistent with a template switching artifact. Previous CAGEs do not just sequence first strand cDNA and do not sequence as deeply. As F3 and F4 versions of CAGE undergo a second strand synthesis step, tag cleavage step and linker ligation step these template switch artifacts are depleted in previous versions of CAGE.
2. Antisense intronic signal is also depleted (see the region covering the last 3 exons of Pu.1). The introns are painted at a lower level than the exons, but they are still painted. Actinomycin D treatment depletes the antisense intronic signal but has no effect on the sense intronic signal. This would argue to me that there was intronic RNA in our libraries that made it through to a CAGE tag, removal of template switching removes the antisense signal but does not affect the sense signal. This again suggests that the intronic sense signal is measuring RNA (not necessarily capped 5' end here).
3. The RNA-seq shows unspliced tags within the intronic regions where we see hCAGE intron painting. The signal is very strand specific.
4. Carrying out PNK followed by TEX treatment reduces all forms of painting but does not completely remove it. Including ActD removes the artifactual antisense and PNK and TEX reduce the remaining sense signal. (the fraction that is not removed could represent recapping).
I think basically we are sequencing very deeply. CAGE picks up some artifactual signals and the deeper you go the more you will see:
1. Not all signal is capped 5' end - but the VAST majority is capped signal
2. Template switching explains the majority of antisense painting signal in the F5 dataset
3. Given that some uncapped 5' ends make it through and we are sequencing deeply, we expect to see degraded/truncated signal along exons.
4. Given that a much smaller fraction of the RNA has not been fully processed, we expect also to see degraded/truncated signal along introns (but at a much lower rate than exons).
5. Artefactual mapping/mappability of intronic regions can not explain the observation that actinomycinD removes antisense intronic signal as well as exonic. This tells us that the signal is there (ie a cDNA end was sequenced, and that it was mapped correctly).
Note: none of the treatments completely removed exonic/intronic painting signals. This could be because there is recapping (as the Gingeras group have shown), or it could also be that the treatments did not go to completion. I think this a further area of investigation.
I also think it could be useful to mine the painting signal for generating tentative gene models for new regions identified by hCAGE. Although we will have some CAGE-scan and RNA-seq, the hCAGE painting signal often defines where the exons are... is anyone interested in chasing this?
Painting vs promoter signal
Piotr checked it quickly some time ago by simply using RefSeq starts and their exons. Painting correlates well with promoter signal. Also, the antisense tag counts correlate with promoter usage (although this signal is ~50x weaker than the sense one). (this was done using human tissues + primary cells in U11)
Exon Painting in other datasets
For a comparison please see the following configs: FANTOM3: https://fantom5-collaboration.gsc.riken.jp/zenbu/gLyphs/#config=XFdlSK6DHlhnwlovH8EK3;loc=mm9::chr5:143663892..143669304 (Note: mouse beta actin shown, exon and intron painting visible, ANTISENSE present, Note2: F5 and F3 shown. 10X deeper for F5)
FANTOM4: https://fantom5-collaboration.gsc.riken.jp/zenbu/gLyphs/#config=fa1qiko8ER4qn7l-uw0eLD;loc=hg18::chr7:5532442..5537622 (Note: human beta actin shown exon and intron painting visible, little to no antisense)
nanoCAGE with atypical template-switching oligonucleotides
nanoCAGE libraries made with alternative template-switching (to be uploaded to Zenbu at least). See the nanoCAGE page for a discussion about type G and type C libraries.
FASTQ files where are available for download at: https://fantom5-collaboration.gsc.riken.jp/webdav/home/plessy/nanoCAGE/extracted/
Library Type Reads Cell ─────── ──── ───── ──── I56-GA C 1,194,725 HepG2 I56-GA G 975,706 HepG2 I63-GA C 2,292,793 HepG2 I63-GA G 2,441,271 HepG2 I79-GA C 973,013 HepG2 I79-GA G 1,057,835 HepG2 I80-GA C 580,413 HepG2 I80-GA G 774,428 HepG2 I81-GA C 104,029 HepG2 I81-GA G 302,963 HepG2 I83-GA C 1,482,477 THP-1 I83-GA G 823,832 THP-1
Zenbu config: https://fantom5-collaboration.gsc.riken.jp/zenbu/gLyphs/#config=roWMSw6pCZpjMX-24UYAUD;loc=hg19::chr11:75109018..75118275 (RPS3)
Additional experiments that could be done
Uncapped spike RNAs -> THP-1 total RNA on hCAGE
Capped spike RNAs -> THP-1 total RNA on hCAGE
3' blocking of reactive group THP-1 total RNA on hCAGE
Analyses that NEED to be done
Painting vs promoter signal (with mappability of promoter and exons considered)
Comparison of Gene ontology terms for highly painted loci and weakly/no painted loci with matched promoter expression levels (ie. if the loci are highly expressed, highly expressed loci tend to fall into several well known GO terms).
Comparison of gene length painted vs not painted
members
Al Forrest
Piero Carninci
David Hume
FANTOM5 email reference
Exon painting is discussed in some threads on the FANTOM5 mailing list. Each email below identify one discussion thread.
- fantom5:00488
- fantom5:00673
- fantom5:00736
Bibliography
Post-transcriptional processing generates a diversity of 5′-modified long and short RNAs Affymetrix/Cold Spring Harbor Laboratory ENCODE Transcriptome Project* Katalin Fejes-Toth,1,2* Vihra Sotirova,1,2 Ravi Sachidanandam,1† Gordon Assaf,1,2 Gregory J. Hannon,1,2 Philipp Kapranov,3* Sylvain Foissac,3 Aarron T. Willingham,3 Radha Duttagupta,3 Erica Dumais,3 and Thomas R. Gingeras1,3 http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2719882/
Regulated post-transcriptional RNA cleavage diversifies the eukaryotic transcriptome Tim R. Mercer1,4, Marcel E. Dinger1,4, Cameron P. Bracken2,3, Gabriel Kolle1, Jan M. Szubert2, Darren J. Korbie1, Marjan E. Askarian-Amiri1, Brooke B. Gardiner1, Gregory J. Goodall2,3, Sean M. Grimmond1 and John S. Mattick1,5 http://genome.cshlp.org/content/20/12/1639.full
