Evolution of alternative splicing

Mikhail GelfandInstitute for Information Transmission Problems,

Russian Academy of Sciences

Workshop “Gene Annotation Analysis and Alternative Splicing”Berlin, December 2004

Overview

• Exon-intron structure of orthologous genes– human–mouse – Drosophila–Anopheles

• Sequence divergence in alternative and constitutive regions

• Evolution of splicing and regulatory sites • Alternative splicing and protein structure

Alternative splicing of human(and mouse) genes

5% Sharp, 1994 (Nobel lecture)35% Mironov-Fickett-Gelfand, 199938% Brett-…-Bork, 2000 (ESTs/mRNA)22% Croft et al., 2000 (ISIS database)55% Kan et al., 2001 (11% AS patterns conserved in mouse ESTs)

42% Modrek et al., 2001 (HASDB)~33% CELERA, 2001

59% Human Genome Consortium, 200128% Clark and Thanaraj, 2002all? Kan et al., 2002 (17-28% with total minor isoform frequency > 5%)

41% (mouse) FANTOM & RIKEN, 200260% (mouse) Zavolan et al., 2003

• Exon-intron structure of orthologous genes– human – mouse – Drosophila–Anopheles

• Sequence divergence in alternative and constitutive regions

• Alternative splicing and protein structure

Data

• known alternative splicing– HASDB (human, ESTs+mRNAs)– ASMamDB (mouse, mRNAs+genes)

• additional variants– UniGene (human and mouse EST clusters)

• complete genes and genomic DNA– GenBank (full-length mouse genes)– human genome

Methods

• TBLASTN (initial identification of orthologs: mRNAs against genomic DNA)

• BLASTN (human mRNAs against genome)• Pro-EST (spliced alignment, ESTs and mRNA

against genomic DNA)• Pro-Frame (spliced alignment, proteins against

genomic DNA)– confirmation of orthology

• same exon-intron structure• >70% identity over the entire protein length

– analysis of conservation of alternative splicing• conservation of exons or parts of exons• conservation of sites

166 gene pairs

42 84 40

human mouse

Known alternative splicing:

126 124

Elementary alternatives

Cassette exon

Alternative donor site

Alternative acceptor site

Retained intron

Human genes

mRNA EST

cons. non-cons. cons. non-cons.

Cassette exons 56 25 74 26Alt. donors 18 7 16 10Alt. acceptors 13 5 19 15Retained introns 4 3 5 0Total 96 30 114 51Total genes 45 28 41 44

Conserved elementary alternatives: 69% (EST) - 76% (mRNA)

Genes with all isoforms conserved: 57 (45%)

Mouse genes

mRNA EST

cons. non-cons. cons. non-cons.

Cassette exons 70 5 39 9Alt. donors 24 6 17 6Alt. acceptors 15 6 16 9Retained introns 8 7 10 4Total 117 24 82 28Total genes 68 22 30 26

Conserved elementary alternatives: 75% (EST) - 83% (mRNA)

Genes with all isoforms conserved: 79 (64%)

Real or aberrant non-conserved AS?• 24-31% human vs. 17-25% mouse elementary

alternatives are not conserved• 55% human vs 36% mouse genes have at least

one non-conserved variant• denser coverage of human genes by ESTs:

– pick up rare (tissue- and stage-specific) => younger variants

– pick up aberrant (non-functional) variants• 17-24% mRNA-derived elementary alternatives

are non-conserved (compared to 25-32% EST-derived ones)

smoothelin

human

common

mouse

human-specific donor-site

mouse-specific cassette exon

autoimmune regulator

human

common

mouse

retained intron; downstream exons read in two frames

Na/K-ATPase gamma subunit (Fxyd2)

human

mouse

(deleted) intron

com

mon

alternative acceptor site within (inserted) intron

Comparison to other studies.Modrek and Lee, 2003: skipped exons

• 98% constitutive exons are conserved• 98% major form exons are conserved• 28% minor form exons are conserved

• inclusion level is a good predictor of conservation

• inclusion level of conserved exons in human and mouse is highly correlated

Minor non-conserved form exons are errors? No:

• minor form exons are supported by multiple ESTs

• 28% of minor form exons are upregulated in one specific tissue

• 70% of tissue-specific exons are not conserved

• splicing signals of conserved and non-conserved exons are similar

Thanaraj et al., 2003:extrapolation from EST comparisons

• 61% (47-86%) alternative splice junctions are conserved

• 74% (71-78%) constitutive splice junctions are conserved

• the former number is consistent with other studies, whereas the latter seems to be an underestimate

Regulation of alternative splicing: introns

• Brudno et al., 2001: UGCAUG is over-represented downstream of tissue-specific exons (brain, muscle).

• Sorek and Ast, 2003: Enhanced conservation (between human and mouse) in intronic sequences flanking alternatively spliced exons. UGCAUG is over-represented in conserved regions.

• Exon-intron structure of orthologous genes– human – mouse – Drosophila–Anopheles

• Sequence divergence in alternative and constitutive regions

• Alternative splicing and protein structure

Fruit fly and mosquito

• Technically more difficult than human-mouse:– incomplete genomes– difficulties in alignment, especially at gene

termini– changes in exon-intron structure irrespective of

alternative splicing (~4.7 introns per gene in Drosophila vs. ~3.5 introns per gene in Anopheles)

Filtering of the

dataset

FlyBase

alternatively spliced fruit fly gene and all its protein isoforms

Non-canonical sites:exclude isoform

Pro-Frame alignment of all isoforms with the fruit fly genome. Frameshift or in-frame stop for at least one isoform:

exclude gene

No constitutive segments inside gene:exclude gene

List of orthologous

pairs

List of filtered fruit fly genesENSEMBL

Pro-Frame alignment of all fruit fly isoforms with the mosquito genome mosquito

genesSimilarity for all isoforms <30%: exclude

orthologous pair

Poly-N within aligned region in the mosquito genome for at least one isoform:

exclude orthologous pair

Set of filtered orthologous pairs

Classification of exons and coding segments • for each pair of isoforms define: mutually Exclusive

exon, Cassette exon, retained Intron, alternative Acceptor site, alternative Donor site; then merge these definitions over all pairs for a gene

--I--

----D ---AD -C-------- EC------A-E----

Left marginal coding segments

Internalcoding segments

isoform 1

isoform 2

isoform 3

- exon - alternative coding segment - constitutive coding segment

----D

E----

E----

E----

E----

----D

----D

-----

-----

---A-

---A-

EC---

--I--

---AD

---AD

-C--- -C---

-C---

constitutive exon

---AD

Right marginal coding segments

Left marginal exons

Internal exons

Right marginal exons

How to define conservation of fruit fly alternative exons

• Alignment of an exon may depend on the isoform. In the cases listed below, shorter exons are assumed to be conserved, whereas longer ones are considered missing

isoform 1

isoform 2

- similarity in alignments of all isoforms including this segment was less than 35%

- similarity in alignment of at least one isoform including this segment was greater than 35%

**missing exon **missing exon *missing exon ***missing exon

Conservation of fruit coding segments in the

mosquito genome. Small (curated) sample

Type of segment

Missing Conserved Total

left marginal (alternative)

46 (77%) 14 (23%) 60 (12%)

internal alternative

22 (55%) 18 (45%) 40 (8%)

internal constitutive

83 (24%) 264 (76%) 347 (69%)

right marginal (alternative)

31 (56%) 24 (44%) 55 (11%)

Total 182 (36%) 320 (64%) 502 (100%)

Conservation of fruit coding segments in the

mosquito genome. Large (non-curated) sample

Type of segment

Missing Conserved Total

left marginal (alternative)

858 (57%) 639 (43%) 1497 (23%)

internal alternative

215 (55%) 178 (45%) 393 (6%)

internal constitutive

903 (23%) 2999 (77%) 3902 (59%)

right marginal (alternative)

414 (53%) 369 (47%) 783 (12%)

Total 2390 (36%) 4185 (64%) 6575 (100%)

Classification of slice events for fruit fly exons

• divided exon• joined exon• exactly conserved exon• mixed;

d eDr

An

- slice

j jj d m jd j m m j

- exon

Different types of events for the same exon dependent on an isoform

dDr (isoform 1)

- slice

j

- exon

An

d

Dr (isoform 2)j

An

j

j

e

e

Types of elementary alternatives and conservation of fruit fly exons in the mosquito genome. Large (non-curated) sample, internal exons

missing mixed joined divided exact

constitutive 728 (23%) 212 (7%) 754 (23%) 407 (13%) 1356 (42%)

Donor site 229 (50%) 21 (5%) 52 (11%) 47 (10%) 130 (28%)

Acceptor site 390 (43%) 45 (5%) 133 (15%) 124 (14%) 250 (28%)

retained Intron 37 (70%) 3 (6%) 2 (4%) 8 (15%) 6 (11%)

Cassette exon 90 (59%) 4 (3%) 9 (6%) 6 (4%) 50 (33%)

Exclusive exon 10 (15%) 1 (1%) 1 (1%) 1 (1%) 55 (82%)

Types of elementary alternatives and conservation of fruit fly exons in the mosquito genome. Large (non-curated) sample, internal exons

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

CONSTANTexon

Donor site Acceptorsite

retainedIntron

Cassetteexon

Exclusiveexon

EXACT

divided

joined

mixed

MISSING

Fruit fly and mosquito

• The general results are the same as for the human-mouse comparison: more conservation of constitutive segments than alternative ones:– 75% const. and 45% alt. segments are

conserved– constitutive exons: >50% conserved exactly,

~25% intron in drosophila, ~8% intron in anopheles

– conservation of alternatives: 36% cassette exons, 51% donor sites, 63% acceptor sites, 83% mutually exclusive exons

• Exon-intron structure of orthologous genes– human – mouse – Drosophila – Anopheles

• Sequence divergence in alternative and constitutive regions

• Alternative splicing and protein structure

Concatenates of constitutive and alternative regions in all genes: different evolutionary rates

Columns (left-to-right) – (1) constitutive regions; (2–4) alternative regions: N-end, internal, C-end

0,1760,199

0,187

0,301

0,00

0,10

0,20

0,30

Constitutive N-endalternative

Internalalternative

C-endalternative

d N/dS

0,886 0,874 0,878

0,807

0,7

0,8

0,9

Constitutive N-endalternative

Internalalternative

C-endalternative

Am

ino-

acid

iden

tity

• Relatively more non-synonimous substitutions in alternative regions (higher dN/dS ratio)

• Less amino acid identity in alternative regions

Genes with length of both const. and alt. reg. > 80 nt

• Horizontal axis: difference in dN/dS in const. and alt. regions• Vertical axis: number of genes• Violet : dN/dS in const. regions > dN/dS in alt. regions • Yellow: dN/dS in const. regions < dN/dS in alt. regions

658

207

79

27 19 27

773

333

140

7144 58

0

100

200

300

400

500

600

700

800

900

0.0-0.1 0.1-0.2 0.2-0.3 0.3-0.4 0.4-0.5 >0.5

279 proteins from SwissProt+TREMBL with “varsplic” features

constitutive alternative % alt. to all

length 199270 66054 25%all SNPs 1126 368 25%synonymous 576 (51%) 167 (45%) 22%benign 401 (36%) 141 (38%) 26%damaging 149 (13%) 60 (16%) 29%

again, there is some evidence of positive selection towards diversity. This is not due to aberrant ESTs

(only protein data are considered).

• Exon-intron structure of orthologous genes– human – mouse – Drosophila – Anopheles

• Sequence divergence in alternative and constitutive regions

• Alternative splicing and protein structure

Data• Alternatively spliced genes (proteins) from

SwissProt– human– mouse

• Protein structures from PDB• Domains from InterPro

– SMART– Pfam– Prosite– etc.

a)

6%10%

15%37%

40%

34%

21%

19%

6%13%

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Expected Observed

Non-domain functional units partially

Domains partially

No annotated unit affected

Non-domain functional units completely

Domains completely

Alternative splicing avoids disrupting domains (and non-domain units)

Control:

fix the domain structure; randomly place alternative regions

… and this is not simply a consequence of the (disputed) exon-domain correlation

0

1

Rat

io(o

bser

vere

d/ex

pect

ed)

Mouse Human Mouse Human Mouse Human

nonAS_Exons AS_Exons AS

AS&Exon boundaries and SMART domains

inside domainsoutside domains

Positive selection towards domain shuffling (not simply avoidance of disrupting domains)

a)

6%10%

15%37%

40%

34%

21%

19%

6%13%

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Expected Observed

Non-domain functional units partially

Domains partially

No annotated unit affected

Non-domain functional units completely

Domains completely

b)

Domains completely

Non-domain units

completely

No annotated

units affected

Expected Observed

Short (<50 aa) alternative splicing events within domains target protein functional sites

a)

6%10%

15%37%

40%

34%

21%

19%

6%13%

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Expected Observed

Non-domain functional units partially

Domains partially

No annotated unit affected

Non-domain functional units completely

Domains completely

c)

Prosite patterns

unaffected

Prosite patterns

affected

FT positions

unaffected

FT positions affected

Expected Observed

An attempt of integration

• AS is often young (as opposed to degenerating)

• young AS isoforms are often minor and tissue-specific

• … but still functional– although unique isoforms may be result of aberrant

splicing• AS regions show evidence for positive

selection – excess damaging SNPs– excess non-synonymous codon substitutions

What to do

• Each isoform (alternative region) can be characterized:– by conservation (between genomes)– if conserved, by selection (positive vs negative)

• human-mouse, also add rat; compare species of Drosophila and Caenorhabditis

– pattern of SNPs (synonymous, benign, damaging)– tissue-specificity

• in particular, whether it is cancer-specific– degree of inclusion (major/minor)– functionality (for isoforms)

• whether it generates a frameshift• how bad it is (the distance between the stop-codon and

the last exon-exon junction)

What to expect

• Cancer-specific isoforms will be less functional and more often non-conserved

• Set of non-conserved isoforms will contain a larger fraction of non-functional isoforms; and this may influence evolutionary conclusions on the sequence level

• Still, after removal of non-functional isoforms, one would see positive selection in alternative regions (more non-synonymous substitutions compared to constant regions etc.), especially in tissue-specific ones

ReferencesNurtdinov RN, Artamonova II, Mironov AA, Gelfand MS (2003)

Low conservation of alternative splicing patterns in the human and mouse genomes. Human Molecular Genetics 12: 1313-1320.

Kriventseva EV, Koch I, Apweiler R, Vingron M, Bork P, Gelfand MS, Sunyaev S. (2003) Increase of functional diversity by alternative splicing. Trends in Genetics 19: 124-128.

Brudno M, Gelfand MS, Spengler S, Zorn M, Dubchak I, Conboy JG (2001) Computational analysis of candidate intron regulatory elements for tissue-specific alternative pre-mRNA splicing. Nucleic Acids Research 29: 2338-2348.

Mironov AA, Fickett JW, Gelfand MS (1999). Frequent alternative splicing of human genes. Genome Research 9: 1288-1293.

Acknowledgements

• Discussions– Vsevolod Makeev (GosNIIGenetika)– Eugene Koonin (NCBI)– Igor Rogozin (NCBI)– Dmitry Petrov (Stanford)

• Support– Ludwig Institute of Cancer Research– Howard Hughes Medical Institute– Russian Fund of Basic Research– Russian Academy of Sciences

Authors• Andrei Mironov (Moscow State University) – spliced alignment• Ramil Nurtdinov (Moscow State University) – human/mouse

comparison• Irena Artamonova (Institute of Bioorganic Chemistry, now

Institute of Bioinformaics, GSF) – human/mouse comparison, MAGEA family

• Dmitry Malko (GosNIIGenetika) – Drosophila/Anopheles comparison

• Inna Dubchak (Lawrence Berkeley Lab) – sites• Michael Brudno (UC Berkeley, now Stanford) – sites• Ekaterina Ermakova (Moscow State University) – evolution of

alternative/constitutive regions• Vasily Ramensky (Institute of Molecular Biology) – SNPs• Eugenia Kriventseva (EBI, now BASF) – protein structure• Shamil Sunyaev (EMBL, now Harvard University Medical

School) – protein structure