Transcript General influence of mRNA secondary structure on splicing

Pre-mRNA secondary structures
influence exon recognition
Michael Hiller
Bioinformatics Group
University of Freiburg, Germany
Current model of splicing
enhancer
silencer
Michael Hiller
Secondary structure of (pre-)mRNA
(pre-)mRNA is not a linear sequence:
• structural elements: IRE, IRES, SECIS, A to I editing
• secondary structure and splicing:
– stem structure containing the exon 10 donor leads to exon
skipping of MAPT
– regulation of mutually exclusive exons in FGFR2 and
Drosophila DSCAM
• SR proteins / hnRNPs have “single-stranded RNA
binding domains”
• bind hairpin loops (Nova1, hnRNP A1, SRp55)
Michael Hiller
Fibronectin EDA exon
wt
mutant
General trend for splicing motifs to be single-stranded?
Buratti et al. Mol and Cell Bio. 24(3) 2004
Michael Hiller
1. Data set
Experimentally verified splicing motifs
• AEDB motif database:
– motifs with their natural pre-mRNA sequence context
– only motifs shorter than 10 nt
final set of 77 motifs
intronic/exonic enhancers/silencers from >6 species
including CFTR, FN1, CD44, FGFR1/2, SMN1, tra2beta
Michael Hiller
2. How to measure single-strandedness?
Probability that an mRNA part is completely Unpaired
Eall  Eunpaired
PU = e
•
•
•
•
RT
the higher PU, the higher the single-strandedness
use all (sub)optimal structures
consider the free energies of structures
allow comparison for motifs of the same length
Michael Hiller
3. In which region is pre-mRNA free to fold?
• long range base pairs are less likely
– protein binding
– co-transcriptional structure formation
– need more time
• experimental evidence that folding is limited to ≈ 50 nt
Michael Hiller
3. In which region is pre-mRNA free to fold?
consider short range base pairs
• symmetrical context lengths 11 – 30 nt
• compute average PU
Michael Hiller
Results and Statistics
77 experimentally verified motifs: average PU = 0.25
Results
real data:
control 1:
control 2:
control 3:
control 4:
control 5:
PU = 0.25
PU = 0.15
PU = 0.18
PU = 0.15
PU = 0.12
PU = 0.15
P<0.01
P<0.01
P<0.01
P=0.046
P<0.001
Michael Hiller
Results and Statistics
• negative correlation between
PU value and GC content of the
flanks (r = -0.64)
 all null models have the same
GC content
Consistent results for:
• different measurements for single-strandedness
• different context lengths (11-20 nt and 11-50 nt)
verified motifs are significantly more single-stranded
attributed to the flanks
Michael Hiller
Experimental testing
inserts with known splicing motifs (TAGGGT, hnRNP A1)
Michael Hiller
Experimental testing
secondary structure of ESE / ESS affects splicing
Michael Hiller
Can we detect structural selection on predicted motifs ?
• divide all 4096 hexamers into [Stadler et al. PLoS Genet. 2006]
– enhancers
– splicing neutral
– silencers
• for each hexamer get „overall PU value“ in
– real exons
– pseudo exons
– intronic regions between a real and a decoy donor/acceptor site
Michael Hiller
Selection on structural context of predicted motifs
Compare motifs with equal number of GC´s (e.g. GAAGAA with AACCTA)
higher single-strandedness
- for enhancers in exons
- for silencers in pseudo exons and decoy regions
Michael Hiller
Selection on structural context of predicted motifs
How often is selection strong enough to overcome
the correlation between PU and GC?
structural context has a widespread and general importance
secondary structures are subject to selection
Michael Hiller
Implications – splicing effect of mutations
• SNP can change secondary structures
[Shen et al. PNAS, 1999]
• secondary structure might be important for
– design and interpretion of mutagenesis experiments
– basis of mutations that affect splicing
Michael Hiller
Implications – splicing effect of mutations
human CFTR exon 12:
25GA mutation
reduces exon inclusion
from 80 to 25%
Michael Hiller
Implications – splicing effect of mutations
rat beta-tropomyosin exon 8:
- mutations in the first enhancer
- mutations in the second enhancer
- mutations in the third enhancer
 no effect on splicing
 strong effect
 weak effect
Michael Hiller
Conclusion
• verified splicing motifs are more single-stranded
• structural context of predicted ESEs/ESSs under natural selection
• selection pressure on a coding exon:
– coding sequence
– splicing signals
– structural context for splicing motifs
• another piece for the ‘mRNA splicing code’
Michael Hiller
Thank you
University Freiburg
– Rolf Backofen
University of Erlangen-Nürnberg
– Stefan Stamm
– Zhaiyi Zhang
Michael Hiller