Nikolaus Rajewsky

1

Metazoan microRNAs

• Nikolaus Rajewsky
• Biology Mathematics
• Center for Comparative Functional Genomics
• Department of Biology, New York University
• Max Delbruck Centrum for Molecular Medicine,
  Berlin

2
Acknowledgements
Collaborators Markus Stoffel (Rockefeller) Kris
Gunsalus Fabio Piano (NYU) Pino Macino
(Rome) Lior Pachter (Berkeley) T. Tuschl
(Rockefeller) Nick Socci (Sloan-Kettering)
Lab Azra Krek (Dominic Grün) Sabbi Lall Kevin
Chen Marc Friedlander Pranidhi Sood Thadeous
Kacmarczyk Teresa Colombo Yi-Lu Wang D.
Langenberger, R.Wolf, L. Rosenberg, I. da
Piedade
Administration Alexandra Tschernycheff Funding
NIH.
Thanks to The Genomics Initiative and all the
wonderful people at NYU.
3
We have heard a lot about a certain class of
trans acting regulatory factors Transcription
Factors

• DNA (m)RNA

4

• DNA RNA Protein

But they do not tell us anything about what
happens after mRNA synthesis.
5

• DNA RNA Protein

Trans-acting regulatory genes which act on mRNA
stability and mRNA translation RNA binding
proteins and microRNAs and
6
Metazoan genomes encode a large class of
non-coding, small RNAs that bind to target mRNAs
to repress mRNA levels/protein production
microRNAs.mature microRNA UCGAGGACUUACGUAAGUC
7

hairy 3 UTR
mir-7
xxxxx
Rajewsky and Socci 2004
8
But what is their biological function?

• A happy case where experiments and
• computation are arguably equally
• important..

9
My two lectures

• background in mRNAs and history of small RNAs
• microRNA biology and function
• microRNA gene finding
• microRNA target finding and experimental
  validation
• microRNA knockdown techniques
• insights into microRNA function from
  characterization of targets
• population genetics applied to human gentotyped
  SNPs to get
• insights into function and evolution of
  cis-regulatory sites
• microRNA Literature 2002 5 papers in
  pubmed
• 2005 100
  papers
• 2006 more
  than I can read
• Reviews Bartel Cell 2004, Ambros Nature 2004,
  Rajewsky Nature Genetics 2006

10
Abundance of RNAs in cells
Hershey et al, 40 years ago
11
Ancient RNA

• RNA may be older than DNA (RNA world) dark
  matter (?)

12
Pervasiveness of RNA genesis

• 50 of the human genome is transcribed
• (Affymetrix Science 2004)
• 90 of the yeast genome is transcribed (W. Huber,
  personal communication)

13
Some basic facts about mRNAs
14
CAP structure at 5 end of mRNAs m7G
7-methylguanylate, p phosphate group, N any
base Poly A Tail at 3 end of mRNAs
homopolymeric stretch of 25-200 adenine
nucleotides Secondary structures (hairpins) can
block translation. Internal Ribosomal Entry Sites
can induce CAP independent transl initiation.
upstream Open Reading Frames normally function as
negative regulators. In green binding sites for
proteins/RNAs There are thousands of human genes
with RNA binding domains!
15
Evolution and Complexity of UTR lengths

• Compare C. elegans, fruitfly, humans 5 UTR
  length roughly constant
• But their average 3 UTR length is (in nt)
• 250, 500, 1000
• TFs have often very long and complicated 3
  UTRs.
• Housekeeping genes 3 UTRs typically short.
• Note Human 50-70 of all genes have at
  least 2 different 3 UTR
• isoforms, typically not annotated in current
  databases

16
3UTR length of highly specifically expressed
mRNAs is tissue specific
Average 3 UTR length
17

• Eukaryotic mRNAs are thought to be
• circularized upon translation PABP binds
• directly poly A, eIF4E binds directly cap,
• hold together by eIF4G

18
Circularization of mRNA by translation initiation
factors
From Wells et al, Mol. Cell (1998)
19
Translational control by 3 UTRs ?
From Mazumdar et al, Trends in Biochemical
Sciences (2003)
20
Some basic facts about RNAsecondary structures

• Basepairing Watson-Crick (AU, GC) but also GU
  wobbles. GC free energy 2.5 AU free energy
• Popular programs to fold or hybridize RNA MFold
  (Zucker), Vienna package, RNAhybrid (Rehmsmeier)

21
RNA secondary structure predictions

• Golden rule Never trust a computationally
  predicted
• secondary structure of a sequence gt300 basepairs.
• Terminology

• Hairpin loop
• Internal loop
• Bulge loop
• Multibranched loop
• Stem
• Pseudoknot

22
Nussinov algorithm

• Simplest approach find the configuration with
  the greatest number of paired bases
• Impossible to score all possible structures
• Dynamic programming (Ruth Nussinov et al 1978)
  look at all ways to extend an optimal
  substructure.

23

• Note
• (a) Hairpin loops are treated as if they could
  have any length. In reality, RNA need to have a
  minimum of 3 nucleotides to form a hairpin
• (b) Any number of bulges/internal loops/stems of
  any size are possible (for example a stem of
  length 1- but nature demands a minimal stem
  length)
• (c ) no parameters for loops, stacking energies,
  no temperature,
• A better algorithm MFOLD (M. Zucker, see his
• review in Nucleic Acids Res. 2003)

24
An excerpt out of the 110 pages of MFOLD
parameters
Y Y
Y Y Y
Y ------------------
------------------ ------------------
------------------ ------------------
------------------ A C G U A
C G U A C G U A
C G U A C G U A
C G U ------------------
------------------ ------------------
------------------ ------------------
------------------ 5' --gt 3'
5' --gt 3' 5' --gt 3'
5' --gt 3' 5' --gt 3'
5' --gt 3' X X
X X
X X
A A A C
A G A U
A G A U U U
U G U C
U A U U
U G YA
YA YA
YA YA
YA 3' lt-- 5' 3' lt-- 5'
3' lt-- 5' 3' lt-- 5'
3' lt-- 5' 3' lt-- 5' 3.90
3.70 3.10 5.50 3.20 3.00 2.40 4.80 3.20
3.00 2.40 4.80 3.90 3.70 3.10 5.50 3.90
3.70 3.10 5.50 3.90 3.70 3.10 5.50 3.80
3.70 5.50 3.70 3.10 3.00 4.80 3.00 3.10
3.00 4.80 3.00 3.80 3.70 5.50 3.70 3.80
3.70 5.50 3.70 3.80 3.70 5.50 3.70 3.20
5.50 2.30 5.50 2.50 4.80 1.60 4.80 2.50
4.80 1.60 4.80 3.20 5.50 2.30 5.50 3.20
5.50 2.30 5.50 3.20 5.50 2.30 5.50 5.50
5.50 5.50 5.50 4.80 4.80 4.80 4.80 4.80
4.80 4.80 4.80 5.50 5.50 5.50 5.50 5.50
5.50 5.50 5.50 5.50 5.50 5.50 5.50
Y Y
Y Y
Y Y ------------------
------------------ ------------------
------------------ ------------------
------------------ A C G U A
C G U A C G U A
C G U A C G U A
C G U ------------------
------------------ ------------------
------------------ ------------------
------------------ 5' --gt 3'
5' --gt 3' 5' --gt 3'
5' --gt 3' 5' --gt 3'
5' --gt 3' X X
X X
X X
A A A C
A G A U
A G A U U U
U G U C
U A U U
U G YC
YC YC
YC YC
YC 3' lt-- 5' 3' lt-- 5'
3' lt-- 5' 3' lt-- 5'
3' lt-- 5' 3' lt-- 5' 3.60
3.20 3.10 5.50 2.90 2.50 2.40 4.80 2.90
2.50 2.40 4.80 3.60 3.20 3.10 5.50 3.60
3.20 3.10 5.50 3.60 3.20 3.10 5.50 3.70
4.00 5.50 3.70 3.00 3.30 4.80 3.00 3.00
3.30 4.80 3.00 3.70 4.00 5.50 3.70 3.70
4.00 5.50 3.70 3.70 4.00 5.50 3.70 5.50
5.50 5.50 5.50 4.80 4.80 4.80 4.80 4.80
4.80 4.80 4.80 5.50 5.50 5.50 5.50 5.50
5.50 5.50 5.50 5.50 5.50 5.50 5.50 3.70
5.50 2.80 4.80 3.00 4.80 2.10 4.80 3.00
4.80 2.10 5.50 3.70 5.50 2.80 5.50 3.70
5.50 2.80 5.50 3.70 5.50 2.80
25

• mfold (m multiple) server www.bioinfo.rpi.edu
  /applications/mfold
• Original version (Zucker Science 1989)
• Computes minimum free energy and optimal fold,
  suboptimal folds, melting temperatures
• Handles temperature, and has a set of ionic
  conditions
• Can also hybridize 2 strands of nucleic acids
• Pseudoknots are not allowed
• Uses dynamic programming. Still written in
  fortran (!)
• Although Mfold uses a plethora of experimental
  binding parameters, many parameters are guessed
  by fitting secondary structures deduced from
  phylogenetic evidence
• Varying the mfold parameters in their error range
  significantly changes mfold output(R. Bundschuh,
  unpublished)

26

• Note the availability of 17 vertebrate genomes,
  12 fly genomes, etc gives very interesting
  opportunities to detect conserved RNA secondary
  structure (check Sean Eddys lab)
• Tools are being developed to compare RNA
  secondary structure (check Giegerich lab), for
  example to identify structural motifs

27
History of regulatory RNAs

• 1960s Jacob Monod
• Idea intermol. RNA-RNA interactions involved in
  gene regulation Britten Davidson 1969
• Small translation control RNAs bind to inactive
  mRNA Heywood and Kennedy 1976
• lin-4 supresses lin-14 translation phenotypes,
  expression, first target sites in 3 UTRs
  Ambros, Lee, Feinbaum, Ruvkun, etc early 90s
• First big surprise microRNAs are abundant
  Tuschl, Bartel, Ambros labs early 2000
• Second big surprise 2005 microRNA are likely to
  regulate at least 30 of all human genes (three
  groups Bartel, Lander, Rajewsky)

28
The classic phenotype heterochronic microRNAs in
C. elegans
In-vivo knockdown of microRNAs by
anti-oligonucleotides
Let-7 knock down phenotype (injection)
Let-7 knock down phenotype (soaking)
29
microRNAs

• Large class of regulatory
• genes (gt200 in humans)
• Initial phenotypic,
• expression, and
• evolutionary analysis
• suggest important function
• Number of microRNAs
• with known function small

Fig. from Bartel CELL 2004
30
microRNAs interesting for computational people
because

• likely to regulate the posttranscriptional
  expression of at least 30 of all (human) genes.
• microRNA gene finding and microRNA target finding
  are at least in part computable.
• A rare case where we can computationally look at
  evolution of cis-regulatory sites and
  trans-acting factors.

Note Many other RNAs siRNAs, rasis, sncRNAs,
tcRNAs, .
31
microRNA genomic organization

• Typically transcribed by pol II
• Can reside in introns, can come in operons
• Many seem to have regular promoters
• Transcription start easily hundreds of bases
  upstream of the hairpin
• Many come in several slightly different copies

32
Processing transcripts the making of mature
microRNAs
Note mature microRNAs presumably freely diffuse
between cytoplasm/nucleus.
33

hairy 3 UTR
mir-7
xxxxx
Rajewsky and Socci Dev. Biol. 2004
34
Mechanistic function of mature microRNAs
From Bartel, CELL 2004
(already outdated).
35
(No Transcript)
36
microRNA expression detection

• Cloning and sequencing 454 sequencing
• Northerns
• In-situs (Exiqon)
• Microarrays (bead based oligo based)
• PCR (Applied Biosystems)
• Reporter gene assays
• By bioinformatic analysis of microarray data for
  regular genes
• (later)

37
microRNA microarrays
Baskerville and Bartel, RNA 2005
38

• Most miRNA genes within 50 kb of each other have
  highly correlated expression patterns, consistent
  with the idea that they are processed from
  polycistronic primary transcripts.

39
Classifying cancer (Golub lab, Nature 2005)
40
microRNA in situs (Exiqon)
Plasterk group, Science 2005
41
Reporter gene activity driven by microRNA
upstream sequences( see also in situs by Eric
Lai PNAS 2005 )
miR-7 reporter
miR-1 reporter
42

• A puzzle microRNAs seemingly not needed and not
  expressed during vertebrate body patterning, but
  microRNAs frequently and specifically expressed
  during fly body patterning.

Giraldez et al Science 2005 Eric Lai lab PNAS
2005
43
Known microRNA functions
44
microRNA gene finding

• Experimental cloning, sequencing (Tuschl/Bartel
  labs), bioinformatics, Northerns.
• no false positives, many false negatives.
  Also implicit expression analysis.
• Computational
• miRscan (Burge lab), miRseeker (Rubin lab,
  fly) score conservation structure of hairpins.
• 50 false positives, less false negatives,
  methods rely on training data.
• New methods (machine learning) Mihaela Zavolan
  (Biozentrum Basel,
• Nature methods 2005) Rosetta Informatics
  (Nature Genetics 2005),
• do not rely on cross-species conservation.

45
Vertebrate microRNA precursor evolution
46
Vertebrate mature microRNA evolution
47
miRscan
From Lim et al Genes Development 2003
48
Zavolan lab (from Pfeffer et al , Nature Methods
2005 viral microRNAs) No cross-species
comparisons. 71 true positives, 3 false
positives.
49

• Rosetta Informatics (Nature genetics 2005)
  roughly 800 human microRNAs, many of them primate
  specific (?)

50

Metazoan microRNA targets

• Nikolaus Rajewsky
• Biology Mathematics
• Center for Comparative Functional Genomics
• Department of Biology, New York University
• Max Delbruck Centrum for Molecular Medicine,
  Berlin

Review/perspective Rajewsky, Nature Genetics 2006
51
microRNAs

• Large class of regulatory
• genes (gt200 in humans)
• Initial phenotypic,
• expression, and
• evolutionary analysis
• suggest important function
• Number of microRNAs
• with known function small

Fig. from Bartel CELL 2004
52

• animal microRNAs bind target mRNAs and
• can degrade mRNAs/inhibit translation.
• to understand function of microRNAs
• find and characterize their targets.
• Note also a beautiful system to computationally
• study function and evolution of trans-acting
  factors
• AND cis-regulatory sites.

53
Known microRNA functions
54
mature microRNA

• UGCAUCCGACCUUUACCAGGA

55

hairy 3 UTR
mir-7
From Rajewsky and Socci, 2004
xxxxx
56
A Meeting at the New York Academy of Sciences
2005
The Rules
57
The participants
Algorithm (alphabetical order)
Representative Speaker
Hatzigeorgiou Sander Rajewsky Rigoutsos
Rehmsmeier Bartel

• DIANA_MicroT.1
• miRanda-2.0
• PicTar1.0
• rna22
• RNAhybrid_helix_2_7TE
• targetscanS

58
(We removed all identifiers not starting with NM)
59
The first round of target predictions (2003/2004)

• Brennecke et al (Cohen lab)
• Enright et al (Sander/Marks labs)
• Kirikadiou et al (Hatzigeorgiou lab)
• Lewis et al (Bartel/Burge labs)
• Rajewsky and Socci

Important Eric Lai Nature Genetics 2002
60
miRanda

• Enright et al, Genome Biology 2003
• Phase 1 Smith-Waterman type.
• Complementarity parameters 5 for GC, 5 for
  AU, 2 for GU and -3 for others. Affine gap model
  in addition scaling factor of 2.0 for first 11
  positions. Finally no mismatches at positions
  2-4 fewer than 5 mismatches at 3-12 at least 1
  mismatch between 9-(L-5) fewer than 2 mismatches
  in the last five positions of the alignment.
• Phase 2 free energy (Vienna) minimium -14
  kcal/mol.
• Phase 3 conservation 80 identity
• No notion of significance. No experimental
  validation.

61
TargetScan, Embl algorithm (2003)Lewis et al
Cell 2003. Brennecke et al 2003.

• Perfect W/C complementarity in the 5 end of the
  microRNA (exception Brennecke et al), free energy
  cutoff of the entire site
• Both have experimental validation (Brennecke in
  vivo, Lewis et al in cell lines)
• Lewis et al introduce the concept of shuffled
  microRNAs as a control

62
Rajewsky and Socci

• Training set 25 known target sites in C. elegans
  
• Screened a number of alignment algorithms to best
  recover the sites versus random sites
• Discovery of a nucleus of consecutive W/C
  basepairings, length 6-8
• Remaining microRNA bases partially basepaired in
  predicted in silico secondary structure of
  miRNAmRNA duplex

63
Nucleus based approach specifically recovers
known target sites
counts
nucleus score
64
Model for target site recognition

• kinetic step the nucleus (consecutive W/C
  basepairings) allows rapid zip up of
  microRNA/target to overcome thermal diffusion
• thermodynamics further basepairings stabilize
  the duplex
• also free energy is a bad predictor for finding
  targets
• nucleus often in 5 end of microRNA
• ran algorithm on fly genes and predicted
  conserved microRNA targets.

Rajewsky and Socci 2004
65
Next generation of target finding algorithms

• TargetscanS (Lewis et al 2005)
• PicTar (Krek et al 2005)
• http//pictar.bio.nyu.edu
• Essentially much more power due to massive
• cross-species comparisons.
• Both algorithms similar. PicTar is a HMM and
• can naturally score combinations of sites also
• detects imperfect nuclei sites.

66
Single target site recognition 2005

• A stretch of perfect consecutive W-C basepairings
  between miRNAmRNA, termed nucleus is crucial to
  target recognition
• Nuclei are 7 bps long and start at the 1st or 2nd
  position from the 5 end of the microRNA
• Imperfect nuclei are allowed if compensated
  (therefore, at least 2
• classes of target sites class 1 seems much
  larger)
• However, this model still cannot explain
  specificity of target
• recognition too many false positives. Need
  cross-species
• comparisons.

67
Stark et al, PLoS Biology 2005
68
PicTar algorithm

• PicTar
• Probabilistic
• Identification of
• Combinatorial
• Target sites
• http//pictar.bio.nyu.edu

69
Using genome wide alignments of 8 vertebrate
genomes (UCSC database)
Multiple alignments of 3 UTRs (Number of
nucleotides in megabases)

• 90 of all 20.255 human RefSeq transcripts
  are aligned between
• human/chimp/mouse/rat/dog. Low error rate
  (human/mouse 5).
• 20 of all 20.255 human RefSeq transcripts
  aligned between all 8 species

70
Identification of anchor sites
12345678------------
21 microRNA 5 TGATACAggcgtaggttaaac 3
microRNA 5 tGATACAGgcgtaggttaaac 3

• hs AAAAAATTATT----GTATCATTTTAA
• pt AATAAATTATT----GTATCATTATGA
• mm ------TTACT----GTATCTGAAAAA
• rn ------TTATT----GTATCA---AAA
• cf -----ATAATT----GTATCA---AAA
• gg -----ATAATTTTGCCCCTCAGTTAAA

71

• Controls randomize bases 1-8 of each microRNA
• Pick 8mers such that abundance of 7mer 1-7, and
  2-8 is comparable in abundance to real microRNA
  (lt15)
• Experimented with many other ways of randomizing,
  with comparable results

72
Predicting single microRNA target sites
(anchors)
Total number of conserved target sites for real
and randomized microRNAs
1.8
2.3
3.3
4.0
73
Experimental validation of 7/13 predictions (not
picked by score!)
74

• Thus, thousands of genes are predicted to be
  targeted by microRNAs.
• Some microRNAs target hundreds of genes.

75
T is a tiling of sequence S with binding sites w
Probability for this tiling is
Likelihood to observe S given pwi
Z is optimized using standard minimization
procedures. PicTar score -log(Z/Zbackground).
76
Combinatorial scoring across species

• hs -AAAAAATTATT----GTATCATTTTAACCCAAAAATTA
  TCAAACGATTTTA
• pt -AATAAATTATT----GTATCATTATAACCCATAAATTA
  TCAAACGATTATG
• mm ------ATTACT----GTATCTTATAACA-------TTA
  TTAACCCAGAAAA
• rn -------TTATT----GTATCA--TAACA-------TTA
  TTAACCCA---AA
• cf ------ATAATT----GTATCA--TAACA------ATAA
  TTAACCCA---AA
• gg ------ATAATTTTGCCCCAATTTTAAAC------ATAA
  TTAACCCACAAAA
• Score phylo average (scores) primates
  rodents dog

77
An early application

• -- the novel microRNA miR-375 is specifically
  
• expressed in pancreatic islet cells
• -- miR-375 down-regulates insulin secretion.
• Targets??
• 
  (Poy et al Nature 2004)

78
Testing target predictions with Westerns in cell
lines
79

• Further experimental results in pancreatic beta
  cell lines
• Knock-down of miR-375 increases insulin secretion
  
• Overexpression of miR-375 reduces insulin
  secretion
• Knock-down of myotrophin reduces insulin
  secretion
• Point mutation of the putative target site in
  myotrophin abolishes knock-down effect of miR-375

80
Krek et al. 2005 Grun et al. 2005 Lall et al. 2006
81
Combinatorial control by microRNAs?

• Synergism between target sites for single
  microRNAs has been experimentally shown
• Some experimental evidence for coordinate control
  by different miRNAs targeting the same mRNA (Krek
  et al, 2005)
• and

82
Figure 1. GY-boxes (green circles), Brd-boxes
(blue squares), and K-boxes (red triangles) in
the 3'-UTRs of Notch target genes of the
Brd-Complex and the E(spl)-Complex
Eric C. Lai et al. Genes Dev. 2005 19 1067-1080
83
Promoter/enhancer
Gene
DNA
Transcription
cis-elements
3 UTR
a
mRNA
Nnnnnn
Translation
miRNAs, RBPs
Transcription factors
Protein
Cell-type specific TF and miRNAs modulates
cellular gene expression
Cell type 2
Cell type 1
TF code
TF code
Gene Battery 2
Gene Battery 1
miRNA code
miRNA code
(from Hobert 2004, modified)
84
PicTar predictions and experimental validation in
nematodes
Lall et al Curr Biology 2006
C. elegans C.
briggsae C. remanei

85

• results
• improved algorithm
• at least 10 of all elegans genes are
• regulated by microRNAs
• new experimental in vivo platform for target
• testing 2 out of 3 new let-7 predictions
• validated by site mutagenesis
• system for embedding microRNA targets
• into regulatory networks

86
N-Browse Integration of target predictions with
C. elegans functional genomics data
87
Experimental system for testing 3 UTR mediated
gene regulation in vivo
88
Time
Time
let-7
let-7
89
Advantages of this in vivo system

• assays 3 UTR mediated, endogenous target
• expression within endogenous miRNA
• expression
• promoterome already available, 3 UTRome
• possible
• cloning and transgenics scalable

90
Body Wall Muscle
Somatic Gonad
Hypodermis
Vulval Muscle
Ant. Neurons
Seam Cells
Germ Line
Intestine
VNC
T14B1.1 3UTR
91
Body Wall Muscle
Somatic Gonad
Hypodermis
Vulval Muscle
Ant. Neurons
Seam Cells
Germ Line
Intestine
VNC
T14B1.1 3UTR
unc-54 3UTR
92
Expression of predicted let-7 regulated 3UTRs
relative to unc-54 3UTR
93
(incomplete) summary of experiments

• most of the 12 tested let-7 targets show 3
  UTR dependent
• expression
• 4 of these show temporal expression
  consistent with
• textbook let-7 expression
• 2/3 targets validated by site mutagenesis

94
Broad rewiring of posttranscriptional control by
microRNAs during metazoan evolution
Grun et al PLoS Comp Bio 2005 Chen Rajewsky,
submitted
95
All 16,000
heptamers microRNA
families
Significance of
conservation across flies
Significance of conservation
across mammals
96

• many microRNAs and recognition motifs deeply
  conserved
• but regulatory relationships only weakly
  conserved (for example 100 between flies and
  vertebrates).
• lots of microRNA target rewiring during
  organismal evolution
• microRNAs maybe part of genes that drive the
  diversity of life (?)

97
Knockdown of mouse miR-122 in vivo

• experiment
• standard affymetrix analysis before and after
• knockdown with antagomirs (3 animals each)
• result
• mRNAs for roughly 300 genes go up, 300 down.
• Krutzfeldt et al, Nature 2005

98
miR-122 nucleus abundance


99
Significance of enrichment of miR-122 nuclei
compare no change UTRs to upregulated UTRs.

actcca
cactcc
- log ( p value )

100
The number of miR-122 nuclei correlates with
log(foldchange) cooperativity (?)
genes sorted
by foldchange
101
Significance of depletion of miR-122 nuclei
compare downregulated genes versus no change
genes
-log(pvalue)

- log ( p value )

102
miR-122 function?

• Analyzed all up regulated genes (target set)
  No significant
• over-representation of functional categories.
• Analyzed all down regulated (anti-targets)
  genes
• Top scoring GO category (plt10-11 ) cholesterol
  biosynthesis.
• Experiments cholesterol levels are indeed
  down by 40.

103
1. miR-122 regulates cholesterol levels2.
genes activated by miR-122 execute regulation
of cholesterol levels 3. these genes are
depleted in miR-122 recognition motifs4.
many likely targets (although coserved!)
currently not predicted because sites present
but not aligned in orthologous 3 UTRs.

104

• Cell type specific signatures of microRNAs on
  steady state mRNA levels

Sood et al PNAS 2006
105
(No Transcript)
106
(No Transcript)
107

• tissue specific microRNAs leave broad signatures
  on steady state mRNA levels of hundreds of genes
• can predict where some specific microRNAs are
  expressed
• microRNAs likely to be involved in tissue identity

108
Detecting functional 3 UTR motifs by
correlatingmotifs with mRNA changes in
microarray experiments

• Fit all 3 UTR motifs in all 3 UTRs
  simultaneously to all mRNA
• expression data (linear regression model,
  REDUCE Bussemaker et
• al 2001)
• Both absence and presence of each motif modeled
• No cross-species comparisons necessary
• The model explains 50-70 of of the expression
  data and predicts
• direct targets as well as functional 3 UTR
  motifs (and thus active
• microRNAs)

Sood et al, 2006
109
(No Transcript)
110
Acknowledgements
Collaborators Markus Stoffel (Rockefeller) Kris
Gunsalus Fabio Piano (NYU) Pino Macino
(Rome) Lior Pachter (Berkeley) T. Tuschl
(Rockefeller) Nick Socci (Sloan-Kettering)
Lab Azra Krek (Dominic Grün) Sabbi Lall Kevin
Chen Marc Friedlander Pranidhi Sood Thadeous
Kacmarczyk Teresa Colombo Yi-Lu Wang D.
Langenberger, R.Wolf, L. Rosenberg, I. da
Piedade
Administration Alexandra Tschernycheff Funding
NIH.
Thanks to The Genomics Initiative and all the
wonderful people at NYU.
111
Lineage specific target predictions
7,200 conserved sites in all 9 flies
2,392 sites only in Sophophora
3,446 sites only in Drosophila
Phylo-Pictar, unpublished
112
micro-evolution primate evolution (unpublished)
25 MYA
6 MYA
0.2 MYA
113

• This part of the talk will be re-posted as soon
  as the paper is (hopefully) out

114
Acknowledgements 2
Kevin Chen
Thanks to M. Zavolan (Biozentrum Basel), R.
Borowsky (NYU), E. Halperin (Berkeley), P.
Andolfatto (UCSD), M. Siegal (NYU), R. Nielsen
(Copenhagen)