Transcript The long-range interaction landscape of gene promoters

Paper review: The long-range
interaction landscape of gene
promoters
——Li Yanjian
2012/9/19
Outline
• Why we study DNA-DNA interaction
• 3C and 5C technology
• Experiment results—Interaction landscape
1.
2.
3.
4.
Experiment design
Data validation
Analysis by cell lines and states
Important features
• Conclusion
• Q&A
Why we study DNA-DNA interaction
• How target genes interact with distal
regulatory elements is still unknown.
• Promoters and distal elements can form
looping interactions which have been
implicated in gene regulation.
• Chromosome is not simply linear and has its
special spatial structure. To learn DNA-DNA
interactions is the first step to know
chromosome’s 3D structure in vivo.
3C and 5C technology
• 3C (Chromosome
Conformation Capture) is
the first technology to
detect DNA-DNA
interaction invented by
Job Dekker
3C and 5C technology
• 3C can only detect one pair of
interaction at a time by PCR, so
they improved it and invent 5C
(Chromosome Conformation
Capture Carbon Copy)
• The experiment detail is quite
complicated, so you can simply
focus on the aim of 5C: to detect
lots of interactions at a time
Interaction landscape——
Experiment design
• Using 5C to detect 44 ENCODE region’s (0.5~1.9Mb,
30Mb in total) DNA-DNA interaction in 3 cell lines
(GM12878, K562, HeLa-S3)
• Analysing interactions between 628 TSS regions and
4535 distal regions
Interaction landscape——
Data validation
• Interaction strength:
1. Within region > Between region
2. Within ENCODE region > Merely neighbour in
genome
3. Different regions from same chromosome >
Different regions from different chromosome
• Consistent with previous 4C and Hi-C data
Interaction landscape——
Analysis by cell lines and states
• Authors defined 7 distinct chromatin states based
on histone modifications, the presence of DHSs
and the localization of proteins such as RNA
polymerase II and CTCF
1. enhancer (E)
2. weak enhancer(WE)
3. TSS
4. predicted promoter flanking regions (PF)
5. insulator element (CTCF)
6. predicted repressed region (R)
7. predicted transcribed region (T).
Interaction landscape——
Analysis by cell lines and states
• ACSL6 region in K562 cell
Interaction landscape——
Analysis by cell lines and states
• γ-δ globin region in K562 cell
Interaction landscape——
Analysis by cell lines and states
• α-globin region in K562 cell
• Important regulatory interaction can be found
Interaction landscape——
Analysis by cell lines and states
• α-globin region in
GM12878 and
HeLa-S3 cells
• Same interactions
were not detected
because these 2
cells express little
or no globin
Interaction landscape——
Analysis by cell lines and states
• Conclusion: The 5C data shown in this paper
consists well with previous study, so it’s
convincing.
• Interactions found by 5C are very likely to be
functional
• Good Pearson correlation coefficient between
replicates (>90%)
Interaction landscape——
Analysis by cell lines and states
• ~60% of the interactions only occurred in one
cell line
Interaction landscape——
Analysis by cell lines and states
• Authors defined 7 distinct chromatin states based
on histone modifications, the presence of DHSs
and the localization of proteins such as RNA
polymerase II and CTCF
1. enhancer (E)
2. weak enhancer(WE)
3. TSS
4. predicted promoter flanking regions (PF)
5. insulator element (CTCF)
6. predicted repressed region (R)
7. predicted transcribed region (T).
Interaction landscape——
Analysis by cell lines and states
• Then they categorized interactions into 4
broader functional groups:
1. Putative enhancer (‘E’ (E or WE))
2. Putative promoter (‘P’ (TSS or PF))
3. CTCF-bound element (CTCF)
4. Not contain any elements belongs to the
above 3 groups (‘U’, unclassified)
• This is non-exclusive classification
Interaction landscape——
Analysis by cell lines and states
• Regions which have interactions usually enrich
active functional markers
Interaction landscape——
Analysis by cell lines and states
• Many U group regions have
active marker——
conservative segmentation
approach
Interaction landscape——
Analysis by cell lines and states
• Conclusion: Unclassified group is relatively
large and still enriched in active marker such
as H3K4me1
• The restriction used by the author is very strict,
so only very significant interactions can be
taken into consideration (high false negative
rate)
Interaction landscape——
Analysis by cell lines and states
• We found that TSS–E and TSS–P interactions are
more cell-type specific than TSS–CTCF interactions
Only one: more
than one
TSS-E/TSS-P
TSS-CTCF
~4:1
~1:1
Interaction landscape——
Analysis by cell lines and states
• Conclusion: TSS-CTCF interactions are more
conservative among different cell types
Interaction landscape——
Analysis by cell lines and states
• Looping interactions with E elements were
significantly enriched for those that involved
expressed TSSs
Interaction landscape——
Analysis by cell lines and states
• Conclusion: TSSs interacted with E elements
are more likely to be expressed
Interaction landscape——
upstream or downstream
• Long-range interaction is asymmetric
• A peak at 120kb upstream of TSSs
Interaction landscape——
upstream or downstream
• Conclusions: Interactions between TSS and
distal fragments are asymmetric
Interaction landscape——
Affect of elements order
• Only,7% of the looping
interactions are between an
element and the nearest TSS
(for active TSS, it goes up to
22%)
• 27% of the distal elements
have an interaction with the
nearest TSS, and 47% of
elements have interactions
with the nearest expressed
TSS.
Interaction landscape——
Affect of elements order
• Conclusion: Interactions don’t always occur
between nearest TSS and distal fragment
Interaction landscape——
CTCF’s insulation function
• We found that 79% of longrange interactions are
unaffected by the presence of one or more CTCFbound sites
• 58% of looping interactions skip sites co-bound by
CTCF and cohesin
Interaction landscape——
CTCF’s insulation function
• Conclusions: CTCF or CTCF&cohesin binding
seems to have little affect on interactions’
forming
• Other factors are needed to complete
insulation function
Interaction landscape——
Multiple interactions
• 50% of TSSs display one or more long-range
interaction, with some interacting with as many as 20
distal fragments
• 10% of distal fragments interacted with one or more
TSS
Interaction landscape——
Multiple interactions
• an example of the complex long-range interaction
networks in the ENr132 region in K562 cell
Conclusion
1. Generate a rich data set reflecting specific geneelement interactions
2. Interactions between TSS and distal elements are
correlated with expression
3. Interactions between TSS and distal elements prefer to
occur in the upstream (~120kb)
4. Interactions are often not blocked by CTCF and cohesin
5. Very few interactions occur between genes and its
nearest elements
6. Promoters and distal elements are engaged in multiple
interaction networks
Q&A