The evolution of expression patterns in the Arabidopsis genome

1
The evolution of expression patterns in the
Arabidopsis genome

• Todd Vision
• Department of Biology
• University of North Carolina at Chapel Hill

2
Driving forces in genome evolution

• Proximate vs. ultimate explanations
• Deleterious mutations are frequent and selection
  cannot effectively act on all of them
• Substitutions
• Insertions and deletions
• Duplications
• Transpositions
• Cellular processes will be affected by this rain
  of mutations
• At the molecular level, we must entertain
  ultimate explanations that do not invoke adaption

3
An example Codon bias

• Genes differ in the frequency that they use the
  preferred codon for a given amino acid, thereby
  affecting
• Translational efficiency
• Translational accuracy
• The strongest codon bias is typically seen in
  short, highly expressed genes under strong
  purifying selection
• Realized codon bias is a balance between
  selection for preferred codons and a continual
  rain of mutations toward unpreferred codons

4
What are the consequences of mutational rain on
the regulatory networks that modulate gene
expression?
5
Outline

• Arabidopsis gene expression (MPSS)
• Two evolutionary issues in the evolution of
  expression profiles
• Physical clustering of co-expressed genes
• Divergence of duplicated genes

6
Digital expression profiling

• Bar-code counting raises fewer concerns about
  cross-hybridization, probe selection, background
  hybridization, etc.
• Serial Analysis of Gene Expression (SAGE)
• Count occurrence of 10-12 bp mRNA signatures
• Long SAGE 21-22 bp signatures
• Uses conventional sequencing technology
• Massively Parallel Signature Sequencing (MPSS)
• Count occurrence of 17-20 bp mRNA signatures
• Cloning and sequencing is done on microbeads
• Commercialized by Lynx Therapeutics

7
MPSS library construction
Brenner et al., PNAS 971665-70.
GATC
8
MPSS library construction
AAAAAAA
AAAAAAA
AAAAAAA
AAAAAAA
AAAAAAA
AAAAAAA
AAAAAAA
AAAAAAA
AAAAAAA
AAAAAAA
AAAAAAA
AAAAAAA
AAAAAAA
AAAAAAA
AAAAAAA
AAAAAAA
AAAAAAA
AAAAAAA
Brenner et al., PNAS 971665-70.
AAAAAAA
AAAAAAA
AAAAAAA
AAAAAAA
AAAAAAA
AAAAAAA
AAAAAAA
AAAAAAA
AAAAAAA
Sort by FACS to remove empty beads
The result of the library construction is a set
of microbeads. Each bead contains many DNA
molecules, all derived from the 3 end of a
single transcript. Beads are loaded in a
monolayer on a microscope slide for the
sequencing of 17 20 bp from the 5 end.
9
MPSS Sequencing
Brenner et al., Nat. Biotech. 18630-4.
10
MPSS Sequencing
Each bead provides a signature of 17-20 bp
Signature Sequence
of Beads (Frequency)
Tag
GATCAATCGGACTTGTC GATCGTGCATCAGCAGT GATCCGATACAGCT
TTG GATCTATGGGTATAGTC GATCCATCGTTTGGTGC GATCCCAGCA
AGATAAC GATCCTCCGTCTTCACA GATCACTTCTCTCATTA GATCTA
CCAGAACTCGG . . GATCGGACCGATCGACT
2 53 212 349 417 561 672 702 814 . . 2,935
1 2 3 4 5 6 7 8 9 . . 30,285
Total of tags gt1,000,000
Two sets of signatures are generated from each
sample in different reading frames staggered by
two bases
11
A catalog of signatures in the Arabidopsis genome
Hits At genome of total Random
1 748204 87.407 845057 2 88392 10.326 6134 3 11
019 1.287 21 4 3512 0.410 0 5 1452 0.170 0 6 87
4 0.102 0 7 470 0.055 0 8 326 0.038 0 9 237 0.0
28 0 10 192 0.022 0 11 158 0.018 0 12-20 707 0.
083 0 21-30 247 0.029 0 31-50 124 0.014 0 gt
50 86 0.010 0 Total 851,212 851,212
All potential signatures (GATC 13 bp) are
identified on both strands of the genomic
sequence. There is one potential signature
appx. every 293 bp on each strand of genome A
signature is classified according to its position
relative to the 29,084 genes pseudogenes in the
TIGR annotation Signatures may not be unique.
The number of hits in the genome is recorded
12
Classifying signatures
Typical signatures
13
Arabidopsis signatures
Based on TIGR annotation (release 3.0, July 2002)
Class in genome of total 1 sense
exonic 203,174 24.0 2 3UTR, lt500 bp
44,202 5.2 3 anti-sense exonic
197,065 23.3 4 inter-genic 288,109 34.0 5
intronic 60,817 7.2 6 anti-sense intronic
57,845 6.8 TOTAL 851,212 100.5
355 genes lack potential Class 1 or 2 signatures
(undetectable) On average, there are 8.5 class 1
2 signatures per gene 8422 genomic signatures
have secondary classes due to overlap or near
overlap of two genes in the TIGR annotation.
14
Core Arabidopsis MPSS librariessequenced by Lynx
for Blake Meyers, U. of Delaware
Signatures Distinct Library sequenced signatur
es Root 3,645,414 48,102 Shoot 2,885,229 53,396
Flower 1,791,460 37,754 Callus 1,963,474 40,903
Silique 2,018,785 38,503 TOTAL 12,304,362 133,37
7
15
Catalog of expressed signatures
Class Position Count 1 or 2 Exon or
3UTR 25,568 3 through 6 Elsewhere in
genome 14,424 0 No match in genome! 10,871
Counting only signatures with abundance 4 PPM
in at least one library. Total is for for 7
libraries (core 1 new root flower library)
16
Genome-wide expression profiling Arabidopsis
Of the 29,084 gene models, 14,674 match unique,
expressed signatures
17
http//www.dbi.udel.edu/mpss

• Query by
• Sequence
• Arabidopsis gene identifier
• chromosomal position
• BAC clone ID
• MPSS signature
• Library comparison
• Site includes
• Library and tissue information
• FAQs and help pages

18
Outline

• Arabidopsis gene expression (MPSS)
• Two evolutionary issues in the evolution of
  expression profiles
• Physical clustering of co-expressed genes
• Divergence of duplicated genes

19
Physical clustering of co-expression

• Caenorhabditis elegans Roy et al., (2002) Nature
  418, 975
• Lercher et al (2003) Genome Research 13, 238
• Drosophila melanogaster Boutanaev et al (2002)
  Nature 420, 666
• Spellman and Rubin (2002) J Biology 1, 5
• Homo sapiens Caron et al (2001) Science 291,
  1289
• Lercher et al (2002) Nature Genetics 31, 180
• Saccharomyces cerevisiae Cohen et al (2000)
  Nature Genetics 26, 183
• Hurst et al (2002) Trends in Genetics 18, 604
• Mannila et al (2002) Bioinformatics 18, 482
• What are the proximate explanations?
• shared cis-regulatory elements
• chromatin packaging, etc.
• What are the ultimate explanations?
• Adaptive greater transcriptional
  efficiency/accuracy?
• Maladaptive mutational rain chipping away at
  insulators and other mechanisms that over-ride
  regional controllers of gene expression?

20
Measuring expression distance
21
Clustering of tissue-specific expression
Chromosome 1
Flower (red)Silique (violet)Leaf (green)Root
(blue)Callus (white)
22
Statistical tests of coexpression clustering

• Measured median pairwise expression distance
  (MPED) in non-overlapping windows of 20 genes
• Summed unique class 1 and 2 signatures for each
  gene
• Only one gene within each tandemly arrayed family
  was counted
• Out of 100 shuffles of gene order
• Zero shuffles had as many windows with small MPED
  (less than 1.5) as the unshuffled data
• Zero shuffles had as large a variance in MPED
  among windows as the unshuffled data

23
Coexpression in Arabidopsis
24
Coexpression in Arabidopsis
25
Coexpression in Arabidopsis
26
Selection and recombination

• In regions of low recombination
• deleterious mutations can hitch-hike to high
  frequency along with favorable ones
• favorable mutations are kept at low frequency by
  linkage to deleterious ones
• Therefore, the effectiveness of natural selection
  is causally related to recombination rate
• Are clusters more concentrated in regions of
• high recombination (i.e. are they adaptive)
• low (i.e. are they maladaptive)?

27
Measuring recombination rate
Chromosome 1
28
Co-expression is greater in low recombination
regions
29
Co-expression clusters

• MPSS data provides evidence for clusters of
  co-expression among non-related genes in
  Arabidopsis
• Co-expression is greater in regions of low
  recombination
• Thus, co-expression clusters may be maladapative,
  at least on average

30
Outline

• Arabidopsis gene expression (MPSS)
• Two evolutionary issues in the evolution of
  expression profiles
• Physical clustering of co-expressed genes
• Divergence of duplicated genes

31
Divergence of duplicated genes
Expression distance
Age of duplication
32
Duplicated genes in Arabidopsis
33
Modes of gene duplication

• Tandem (unequal crossing-over)
• Dispersed (transposition)
• Segmental (polyploidy)

34
Divergence of duplicated genes

• All gene families of size 2 in Arabidopsis were
  classified as dispersed, segmental or
  tandem
• Expression distance was calculated for each
• The number of silent (i.e. synonymous)
  substitutions per site was calculated for each
  (as a proxy for age since duplication)

35
Divergence and mode of duplication
36
Divergence of duplicated genes

• Almost all expression divergence occurs during
  (or immediately following) duplication
• Initial expression divergence is more extreme for
  tandem than dispersed duplicates
• Tandem and dispersed duplicates with the most
  divergent expression profiles are quickly lost
• Segmental duplicates plateau at a lower level of
  expression divergence than dispersed duplicates
• The average divergence in relative expression
  level in each tissue is about 8-fold.

37
Lessons learned

• Clusters of co-expression in Arabidopsis may be
  largely the result of a rain of weakly
  deleterious mutations that homogenize the
  expression profiles of neighboring genes
• Divergence in expression profile between
  duplicated genes is dependent on the nature of
  the mutation that gave rise to the duplication

38
Thanks!

• UNC Chapel Hill
• Jianhua Hu
• University of Delaware
• Blake Meyers
• NSF Plant Genome Research Program
• DBI-01103267 (TJV)
• DBI-0110528 (BCM)

39
(No Transcript)