Probing early eukaryotic evolution using phylognetic methods

1
Probing early eukaryotic evolution using
phylognetic methods
2
The Universal SSU rRNA TreeWheelis et al. 1992
PNAS 89 2930
3
The SSU Ribosomal RNA Tree for Eukaryotes
Mitochondria?
Prokaryotic outgroup
Archezoa
4
The Archezoa HypothesisT. Cavalier-Smith (1983)

• Archezoa are eukaryotes which primitively lack
  mitochondria
• The nucleus was invented before the mitochondrion
  was acquired
• The first eukaryotes were anaerobes
• Archezoans might provide insight into the nature
  of ancestral eukaryotic genomes and biology

5
The Archezoa HypothesisT. Cavalier-Smith (1983)

• The Archezoa hypothesis would fall if
• Find mitochondrial genes on archezoan genomes
• Find mitochondrion-derived organelles in
  archezoans
• Find that archezoans branch among aerobic species
  with mitochondria

6
Is Trichomonas an Archezoan?
7
Trichomonas chaperonin 60 shares common ancestry
with mitochondrial chaperonins
Mitochondria
alpha - proteobacteria
8
Mitochondrial genes in Archezoa
Proteins of mitochondrial origin
Archezoa
Giardia / Spironucleus Trichomonas Microsporidia
Heat shock 70, Chaperonin 60 Heat shock 70,
Chaperonin 60 Heat shock 70
defined as forming a monophyletic group with
mitochondrial homologues in a non-controversial
species phylogeny
9
Chaperonin 60 Protein Maximum Likelihood Tree
(PROTML, Roger et al. 1998, PNAS 95 229)
Note 100 Bootstrap support
A case of Eukaryote Eukaryote HGT?
10
Long branches may cause problems for phylogenetic
analysis

• Felsenstein (1978) made a simple model phylogeny
  including four taxa and a mixture of short and
  long branches

TRUE TREE
WRONG TREE
p gt q

• Methods which assume all sites change at the same
  rate (e.g. PROTML) may be particularly sensitive
  to this problem

11
Chaperonin 60 Protein Maximum Likelihood Tree
(PROTML, Roger et al. 1998, PNAS 95 229)
Longest branches
12
A simple experiment

• Does the Cpn60 tree topology change
• If we remove long-branch outgroups
• If we remove sites where every species has the
  same amino acid

13
Cpn-60 Protein ML tree (PROTML) from variable
sites with outgroups removed
14
Competing Hypotheses for Microsporidia
Microsporidia Fungi Tubulin, mitHSP70
Microsporidia Early SSU rRNA, EF-1 alpha EF-2

• HGT from Fungi to Microsporidia? (Sogin, 1998)
• Another artefact of the method of analysis?

15
Microsporidia have a number of unusual features

• Absence of mitochondria and peroxisomes
• 70s ribosomes - most eukaryotes have 80S
• 5.8S and 23S rRNA genes are fused - like in some
  prokaryotes
• Lack 9 2 microtubule structures

16
Alternative explanations of Microsporidia unusual
features

• Retention of ancestral features of the eukaryote
  cell at an early stage of evolution?
• Or are they
• Adaptations to an obligate intracellular
  lifestyle?

17
Elongation Factor 2 protein ML tree (PROTML)
(Hashimoto et al. 1997 Arch. Protist. 148287)
Entamoeba
Dictyostelium
88
Eukaryote root
75
Microsporidia
Archaebacteria outgroups
Also note that in PROTML the amino acid
substitution process is assumed to be homogeneous
across the tree
18
Shared nucleotide or amino acid composition
biases can also cause problems for phylogenetic
analysis
Aquifex
Thermus
Aquifex (73)
Bacillus (50)
True tree
Wrong tree
16S rRNA
Bacillus
Thermus (72)
Deinococcus
Deinococcus (52 GC)
Aquifex
The correct tree can be obtained if a model is
used which allows base/aa composition to vary
between sequences -LogDet/Paralinear
Distances Heterogeneous Maximum Likelihood
Bacillus
Thermus
Deinococcus
19
LogDet/Paralinear distances for EF-2 DNA variable
sites codon positions 12
Animals
Chlorella
70
Note that root has changed
Trypanosoma
Trichomonas
60
Giardia
Dictyostelium
25
Entamoeba
Sacharomyces
76
45 GC
Microsporidia
Cryptosporidium
Sulfolobus
Archaebacteria outgroups
Methanococcus 44
Halobacterium 58
20
A combination of factors (outgroup GC content and
site rate heterogeneity) influence the EF-2 DNA
tree
Methanococcus outgroup (low GC)
Halobacterium outgroup Higher GC
100
100
80
80
LogDet Bootstrap values
ML estimate
60
60
40
40
20
20
0
0
0
20
40
60
80
100
0
20
40
60
80
100
Fraction of constant sites removed
(Microsporidia, outgroup)
21
A combination of factors (outgroup GC content
site rate heterogeneity) influence the EF-2 DNA
tree
Methanococcus outgroup (low GC)
Halobacterium outgroup Higher GC
100
100
80
80
Bootstrap values
60
60
40
40
20
20
0
0
0
20
40
60
80
100
0
20
40
60
80
100
Fraction of constant sites removed
(Microsporidia, outgroup)
(Microsporidia, Fungi)
22
A combination of factors (outgroup GC content
site rate heterogeneity) influence the EF-2 DNA
tree
Methanococcus outgroup (low GC)
Halobacterium outgroup Higher GC
100
100
80
80
Bootstrap values
60
60
40
40
20
20
0
0
0
20
40
60
80
100
0
20
40
60
80
100
Fraction of constant sites removed
(Giardia, Trichomonas, outgroup)
23
Competing hypotheses for Microsporidia
Microsporidia Fungi Tubulin, RNA
polymerase, LSU rRNA, HSP70, TATA binding
protein, EF-2, EF-1 alpha
Microsporidia Early SSU rRNA
The best supported hypothesis for Microsporidia
is a relationship to fungi - why does SSU rRNA
place them deep?
24
Summary I

• Making trees is not easy
• Among-site rate heterogeneity, fast clock
  species, shared nucleotide or amino acid
  composition biases
• Different data sets may be affected by individual
  phenomena to different degrees
• Biases need not be large if phylogenetic signal
  is weak

25
Summary II

• Are Archezoa ancient offshoots?
• Microsporidia are related to fungi
• Evidence for Giardia and Trichomonas branching
  deeper than other eukaryotes is based on trees
  made using unrealistic assumptions
• PLUS
• For the same reasons we dont know where the root
  lies on the eukaryote tree
• So arguments about early or late branching are
  probably premature anyway

26
Can we make a robust unrooted tree for eukaryotes?

• Combining different genes in a single analysis
  may provide a more robust eukaryotic tree
• One argument is that phylogenetic signal should
  be additive whereas gene-specific noise will
  pull in different directions

27
DNA ML tree found using a model which allows both
base composition and site rates to vary across
the tree
Animals fungi slime moulds
Ciliates plus apicomplexa
Giardia and Trichomonas
Red and green algae/plants
ActintubulinEF-2
28
Hydrogenosomes

• Strange anaerobic eukaryotic organelles which
  make hydrogen

29
(No Transcript)
30
(No Transcript)
31
Anaerobic eukaryotes from different phylogenetic
groups make hydrogen - how?
32
Origin(s) of Hydrogenosomes

• Is a 2 part problem
• The organelle (the bag)
• The biochemistry to produce hydrogen particularly
  hydrogenase

33
CO2
hsp70
Protein import
ME
Malate
Pyruvate
cpn60
Transit peptides
AAC
NAD(P)
NAD(P)H
ATP
N
ADP
NAD(P)-FO
FeHyd
H2
2Fd
2Fd-
2H
ASCT
Acetyl-CoA
Acetate
PFO
Succinyl-CoA
Succinate
CO2
Double membrane
CoASH
STK
Fungi and Trichomonas
ADP Pi
ATP
Enzyme found also in mitochondria
Alpha-proteobacterial ancestry
Schematic Map of Hydrogenosomes (after Muller
1993)
Unknown ancestry
34
Spironucleus
35
(No Transcript)
36
DNA ML tree for Fe hydrogenases
37
A likelihood ratio test of monophyly
(Huelsenbeck, Hillis Nielson 1996)

• The Test Statistic (d) lnL1 - lnL0
• Where lnL1 is the likelihood of the best tree and
  lnL0 is the likelihood of the best monophyly tree
• The null (eukaryote monophyly) distribution of d
  is generated by simulation under an appropriate
  model (parametric bootstrapping)

38
Parametric Bootstrapping to estimate a test
distribution
What might the test statistic distribution look
like if the Fe hydrogenases were monophyletic?
Estimate ML model parameters using original data
Simulate 1000 new sequence data sets using this
model over the best monophyly tree
Calculate d for original data and compare to
distribution - if it falls outside of the 95
interval it is bigger than expected by chance and
monophyly can be rejected
For each new data set estimate L0 and L1 using
ML, with model re-optimised each time
Plot d for each of the 1000 data sets to give the
test distribution and the 95confidence interval
39
The likelihood ratio test rejects the hypothesis
that eukaryotic hydrogenases are monophyletic
d lnL1 - lnL0
d for original data
d Original data
95
9.64
95
d (lnL1 - lnL0) distribution from 1000
simulations of the Fe hydrogenase data on the
best monophyly tree
40
(No Transcript)
41
Iron hydrogenase ML tree
Eukaryotic compartment
Trichomonas Hydrogenosome
Cytosolic?
Plastid
Ciliate Hydrogenosome
42
Conclusions I

• Hydrogenosomes share common ancestry with
  mitochondria
• Hydrogenase has been acquired at least twice and
  can be targeted to different cell compartments in
  different eukaryotes
• Humans, plants and fungi also contain remnants of
  iron hydrogenases
• There is no evidence from phylogenetic analysis
  that the bag and hydrogenase share a common
  origin from the mitochondrion endosymbiont

43
Conclusions II

• Phylogeny is hard, there are lots of potential
  problems with data, so we need to be careful in
  our interpretations of what trees mean - includes
  inferences of HGT
• Better methods hold promise of more reliable
  trees (allowing re-analysis of SSUrRNA data)
• Archezoa contain genes which originated with
  mitochondrion endosymbiont and the jury is still
  out on whether former archezoa have lost the
  mitochondrial bag
• We dont know which eukaryotes are early
  branching - for this we need a rooted tree

44
The mitosome, a novel organelle related to
mitochondria in Entamoeba histolytica
Tovar et al., 1999.
Slide shows epitope tagged recombinant cpn60
localised to mitosome
45
Are there still organelles of common ancestry
with mitochondria in Giardia and Microsporidia?

• Giardia
• What are the ovoid pellicular bodies (in
  Giardia)? The study made suggests that they might
  be nothing but changed mitochondria with a few
  crysts or tubules
• The ultrastructure of mitochondria may be
  related with the oxygen deficiency in Lamblia
  environment
• 
  (Cheissen, 1965)
• Microsporidia
• There are reports of mitochondria-like
  structures in several microsporidia
• 
  (Vavra, 1976)