Transcript ppt

Genome evolution:
a sequence-centric approach
Lecture 13: epistasis: RNA,
enhancers, networks
(Probability, Calculus/Matrix theory, some graph theory, some statistics)
Simple Tree Models
HMMs and variants
PhyloHMM,DBN
Context-aware MM
Factor Graphs
DP
Sampling
Variational apx.
LBP
EM
Generalized EM
(optimize free energy)
Probabilistic models
Genome structure
Inference
Mutations
Parameter estimation
Population
Tree of life
Genome Size
Elements of genome
structure
Elements of genomic
information
Models for populations
Drift
Selection and fixation
Draft
Protein coding genes
Inferring Selection
Today refs: Papers cited
TFBSs
Epistasis
Assume we have two loci, each bearing two alleles (Aa and Bb)
Assume that the basal state of the population is homogenous with alleles ab
f(A) - The relative fitness of A is defined using the growth rate of the genome Ab
f(B) - The relative fitness of B is defined using the growth rate of the genome aB
What is the fitness of AB?
If the two loci are unrelated, we can expect it to be: f(Ab)*f(aB)
When f(A)=1+s, f(B)=1+s’, and s,s’ are small, f(A)*f(B)~(1+s+s’)
Epistasis is defined as the deviation from such linearity/independence:
f(AB) > f(Ab)*f(aB): synergistic loci
f(AB) < f(Ab)*f(aB): antagonistic loci
A
+
AB
B
A
-
B
AB
How widespread is epistasis? Is it positive or negative in general?
and how it affect evolution in general?
Testing epistasis in viruses: directed mutagenesis
47 genotypes of vesicular stomatitis virus carrying pairs of nucleotide substitution mutations (filled)
15 genotypes carrying pairs of beneficial mutations (empty circles)
Sanjuan, PNAS 2004
Testing epistasis in viruses: HIV-1 isolated drug resistant strains
Comparing growth in drug-free media (extracting viral sequence and reintegrating it in a virus model)
Sequencing strains, comparing to some standard
Plotting fitness relative to the number of mutations:
For each pair of loci, compute average fitness for aa,aB,Aa and BB, then estimate epistasis. To assess
significance, recompute the same after shuffling the sequences
Mean is significantly higher than randomized means
Effect is stronger when analysis is restricted to
59 loci with significant effect on fitness
Results suggesting that epistasis tends to be positive (at least in these viruses and in this condition)
Bonhoeffer et al, science 2004
Functional sources for epistasis:
•
Protein structure (interacting residues)
•
Different positions in the same TFBS
•
Two interacting TFBSs
•
TF DNA binding domain and its target site
•
Two competing enzymes
•
Two competing TFBS
•
RNA paired bases
•
Groups of TFBSs at co-regulated promoters
RNA folds and the function of RNA moelcules
•RNA molecular perform a wide
variety of functions in the cell
•They differ in length and class, from
very short miRNA to much longer
rRNA or other structural RNAs.
•They are all affected strongly by
base-pairing – which make their
structural mostly planar (with many
exceptions!!) and relatively easy to
model
Simple RNA folding energy:
number of matching basepairs or sum over basepairing weights
More complex energy (following Zucker):
each feature have an empirically determined parameters
stem stacking energy (adding a pair to a stem)
bulge loop length
interior loop length
hairpin loop length
dangling nucleotides and so on.
Pseudoknots (breaking of the basepairing hierarchy) are typically forbidden:
Predicting fold structure
Due to the hierarchical nature of the structure (assuming no pseudoknots), the situation can be analyzed
efficiently using dynamic programming.
We usually cannot be certain that there is a single, optimal fold, especially if we are not at all sure we are
looking at a functional RNA.
It would be better to have posterior probabilities for basepairing given the data and an energy model…
This can be achieved using a generalization of HMM called Stochastic Context Free Grammar (SCFG)
EvoFold: considering base-pairing as part of the evolutionary model
Once base-pairing is predicted, the evolutionary model works with pairs instead of single
nucleotides.
By neglecting genomic context effects, this give rise to a simple-tree model and is easy to
solve.
If we want to simultaneously consider many possible base pairings, things are becoming
more complicated.
An exact algorithm that find the best alignment given the fold structure is very expensive
(n^5) even when using base pairing scores and two sequences.
Pedersen PloS CB 2006
EvoFold: considering base-pairing as part of the evolutionary model
Whenever we discover compensatory mutations, the prediction of a functional RNA
becomes much stronger.
Evolution of a regulatory module:
eve stripe 2 in D. melanogaster and D. pseudoobscura
mel
While the two enhancers drive a conserved
expression patter, we cannot mix and match them
between species!
Evolution therefore continuously compensate for
changes in one part with changes in the other.
pseudo
Ludwig, Kreitmen 2000
Evolution of a regulatory module
Eve staining in 4 species
Orthologous stripe 2 enhancer
reporters in a melanogaster embryo
D. Melanogaster
D. Yakuba
D. Erecta
The D.
Erecta S2E
is forming
much weaker
stripe in D.
Mel.
D. Pseudoobscura
Ludwig,..,Kreitmen 2005
Sequence conservation and divergence in eve stripe 2 and around it
D. Melanogaster
Enhancer functional in mel.
D. Yakuba
Enhancer not functional in mel.
D. Erecta
Enhancer functional in mel.
D. Pseudoobscura
Coregulation: epistasis of transcriptional modules
•
•
•
Transcriptional modules are crucial for the organization and function of
biological system
Gene co-regulation give rise to major epistatic relations among
regulatory loci
epistasis reduces evolvability
Co-regulation
Is advantageous
Disruption of regulation
Is deleterious
Rugged evolutionary
landscape
Regulation
Scheme 1
Regulation
Scheme 2
S phase
S. cerevisiae
S. cerevisiae
Ribosomal Proteins
Ribosome biogenesis
45
genes
P<10-56
S. Pombe
7 genes
P<10-9
S. pombe
S. cerevisiae
Amino acid met.
114
genes
P<10-151
S. Pombe
32
genes
P<10-29
S. Pombe
S. cerevisiae
Cis-elements underlying conserved TMs
Putative
Orthologous
Module (POM)
S. cerevisiae
S. bayanus
S. castellii
C. glabrata
S. kluyverii
K. waltii
K. lactis
A. gossypii
D. hansenii
C.albicans
Y. lypolitica
N. crassa
A. nidulans
S. pombe
Phylogenetic cis-profiling with 17 yeast species
S phase
Respiration Amino acid
metabolism
Conserved
cis-elements
MCB
S. cerevisiae
S. paradoxus
S. mikatae
•Conserved FM are sometime
regulated by remarkably
conserved cis elements
S. kudriavzevii
S. bayanus
S. castellii
C. galbrata
S. kluyveri
K. waltii
•Conserved cis elements are
bounded by conserved TFs
K. lactis
A. gossypii
D. hansenii
C. albicans
Y. lipolytica
N. crassa
A. nidulans
S. pombe
Tanay et al. PNAS, 2005
HAP2345
GCN4
Ribosomal Protein Module:
Evolutionary change via
redundancy
Redundant
mechanism
Rap1
emergence
Homol-D
loss
S. cerevisiae (133)
112
38
S. parad. (75)
46
31
S. mikatae (88)
57
46
S. kudriavz .(94)
48
40
S. bayanus (118)
54
40
S. castellii (89)
53
45
40
C. glabrata (69)
29
21
45
S. kluyveri (61)
30
29
32
K. waltii (54)
34
31
30
K. lactis (75)
35
A. gossypii (73)
64
D. hansenii (73)
17
52
41
RAP1
Homol-D
IFHL
Y. lipolytica (70)
46
30
C. albicans (41)
Homol-D
based
32
51
53
N. crassa (67)
46
A. nidulans (72)
49
S. pombe (74)
73
44
Rap1 evolution in trans
S. cerevisiae
S. castelii
New TA domain
Co-emerged with
Rap1 role in RP
regulation
K. waltii
A. gossypii
C. albicans
N. crassa
A. nidulans
S. pombe
H. sapiens
BCRT
Myb
Silencing
TA
Redundant cis-elements are spatially clustered: RP genes in A. gossypii
5’
3’
6bp
Homol-D
RAP1
Evolution of the IFHL
element
Drift…
Reverse
complement
duplication
sacc. et al.
hansenii
albicans
lypolityca
crassa
Conservation
Tandem duplication
nidulans
pombe
S. cerevisiae (225)
S. parad. (215)
S. mikatae (187)
Evolution of the
Ribosomal biogenesis
module
S. kudri. (196)
S. bayanus (195)
S. castellii (204)
C. glabrata (214)
157
187
175
159
136
151
152
163
151
159
152
167
180
166
137
157
163
181
59
122
200
145
122
171
163
126
152
110
S. kluyveri (178)
K. Waltii (230)
K. lactis (225)
A. gossypii (226)
D. hansenii (219)
C. albicans (214)
Y. lipolytica (208)
RRPE
PAC
TC
N. crassa (193)
51
154
132
?
A. Nidulans (187)
S. pombe (196)
159
99
83
79
a, S. cerevisiae and C. albicans transcribe their
genes according to one of three programs, which
produce the a-, - and a/ -cells.
The particular cell type produced is determined by
the MAT locus, which encodes sequence-specific
DNA-binding proteins.
In S. cerevisiae, a-type mating is repressed in a-cells
by a2.
In C. albicans, a-type mating is activated in a-cells by
a2.
In both species, a-cells mate with a-cells to form a/a
-cells, which cannot mate.
a2 is an activator of a-type mating over a broad
phylogenetic range of yeasts.
In S. cerevisiae and close relatives, a2 is missing and
a2 has taken over regulation of the type.
Mating genes
a2
Albicans
a2
Cerevisiae
Tsong et al. 2006
A transition of motifs is
observed between
Cerevisiae and albicans
Innovation in a2 is observed
along with the emergence of
possible mcm2 interaction
A redundant intermediate may
have enable the switch
Phenotypic innovation through regulatory adaptation
After S. Carroll
After S. Carroll
Rockman, Plos Biol, 2005
Ihmels Science, 2005
481 segment longer than 200bp that are absolutely conserved between human, mouse and rat (Bejerano et al 2005)
What are these elements doing? Why they are completely conserved? 4 Knockouts are not revealing significant
phenotypes..
Ahituv et al. PloS Biolg 2007
Population genetics do suggest ultraconserved elements are under selection
Separating mutational effects from
selective effect is still a challenge…
Katzman et al., Science 2007