Transcript lecture 03 - phylogenetics - Cal State LA

Phylogeny or “tree thinking”
The evolutionary history of extant organisms can be understood
in terms of their shared inheritance
- which extant species evolved from the same ancestor?
- how were ancestral traits modified in different lineages?
A hypothesis of the evolutionary history of a group is called its
phylogeny
- often summarized in a branching diagram called a
phylogenetic tree
Since we can’t travel back in time to identify common ancestors,
relationships of existing species must be estimated or inferred
from data – therefore, a phylogeny is always a hypothesis
Tree thinking and phylogeny
Nodes: branching points
Branches: lines connecting nodes
Phylogenetic tree
Topology: branching pattern
Taxon = any named group of organisms
A+B
Two or more = taxa
C+D
Sister Taxa: two taxa (= named group of organisms) that are
more closely related to each other than either is to a 3rd taxon
recently
diverged
diverged a long time ago
Branches can be rotated at a node, without changing
relationships among the taxa
these 3 trees are all equivalent
Relationships can be resolved or unresolved
A node with more than 2 branches from it is called a polytomy,
and means the relationships are not fully resolved
- may the data aren’t good enough to figure out the true
relationships among descendents of that ancestor
- maybe ancestor had multiple direct descendents -- rapid
speciation can occur during an adaptive radiation
Characters (traits) change across a phylogeny
Plesiomorphy: ancestral (or primitive) character state
Apomorphy: derived character state, modified from the
ancestral (primitive) state
Synapomorphy: derived character shared by more
than one species or group (shared derived character)
Synapomorphy
ancestral
derived
Synapomorphy
(unites all birds)
Feathers
(birds)
Plesiomorphy
Scales
(snakes
)
Scales
(bony fish)
Scales (early
reptile)
Scales (ancestral fish)
Synapomorphy: shared, derived character
- it’s different from the ancestral state
- it’s found in more than one species
Homologous trait: structurally similar, but functionally different
- human arm, dolphin flipper, bat wing are homologous:
all descended from same ancestral tetrapod appendage,
with modification
Homoplasy: structurally different (unrelated),
but functionally similar
- example: streamlined shape of sharks + killer whales
- adaptation to similar environment,
driven by natural selection
flippers = yes
arms = no
similar coding for
these non-relatives
Distinguishing between true homology
and homoplasy is critical if we are to sort out the phylogeny,
or evolutionary relationship, among living things
Homoplasy = problem for phylogenetics
Homoplasy: non-homologous traits look superficially similar
- resemblance is not due to common ancestry, but rather to
convergent evolution in 2 different groups, which tricks us
- you don’t want to code non-relatives as having the same
traits, or they will falsely group together on your tree
wings in birds
wings in bats
Cladistics uses synapomorphies to identify clades, which
(a) are the groups we want to name (true groups)
(b) tell us something about the evolution of a group
For a long time, this was the only way to infer phylogeny,
using morphological character traits
Cladistics uses synapomorphies alone to establish recent
common ancestry
- relies on synapomorphies to define monophyletic groups,
or clades (same thing)
Monophyletic group includes ALL descendants of an
ancestral species
Types of non-monophyletic groups
Paraphyletic: leaves out one or more descendants of an
ancestor
Polyphyletic: derived from 2 or more ancestral taxa (= crap)
Paraphyletic
Polyphyletic
Clade = monophyletic group
Birds = clade
Reptiles: paraphyletic group
(not a clade)
Mammals = clade
Why do we name para- and polyphyletic groups?
Taxonomists often assigned class or family status to groups that
are very morphologically distinctive, such as birds and cetaceans
(whales & dolphins)
- does not reflect their evolutionary status within other groups
what do you think: should scientists call birds and reptiles
both classes of vertebrates, if birds are really a subset of
reptiles?
Another reason is, mistakes can be made in defining
relationships depending on the character that is used
Cladistics and parsimony
Techniques that identify monophyletic groups based on shared
derived characters are cladistic methods
- based on the idea that the most related groups will have
the most traits in common
Parsimony is a principle that simpler explanations are more
likely to be correct than complicated explanations
- when choosing among hypotheses, we should accept the
one that explains the data most simply and efficiently
- assumes convergence is less likely than shared descent
Cladistics and parsimony
When a new trait appears, all descendants will share it as a
synapomorphy (barring secondary loss of the trait)
feathers
Bird evolution
Cladistics and parsimony
Maximum parsimony = method of phylogeny reconstruction,
process by which we infer the evolutionary history of a group
based on the traits we see today
- the best phylogenetic tree is the one which requires the
fewest changes in traits (characters) to account for modern
character states in surviving lineages
- i.e., assumes that the minimum number of changes
is what really happened over the course of evolution
Using parsimony to infer evolutionary relationships
How are the major types of algae-eating sea slugs related?
some have shells;
seem “primitive”
because they retain
a plesiomorphy, the
ancestral snail shell
some groups have
frilly flaps on their
backs...
are they the most derived =
different from the ancestor?
“code” the different states for each morphological character,
for each existing taxon
“has a shell” = 0, “no shell” = 1
# of characters you can use is unlimited
the characters
you are using
a coded
representation of
one genus
feed all this into a
computer; ask it to
give you the tree that
requires the fewest
number of overall
changes, to produce
the distribution of traits
seen in modern taxa
Parsimony analysis of morphological traits gave this answer
for how all genera in this group are related
Homoplasy = problem for phylogenetics
Problem: Morphological traits may be unreliable for inferring
evolutionary relationships if they are frequently subject
to convergent evolution
- homoplasy may fool us if we rely too heavily on morphology
can we find traits to use in phylogeny-building that are
largely invisible to selection?
DNA sequence data & phylogenetics
Most modern phylogenetic trees are based partly or entirely on
molecular data, usually DNA sequence information
- you can combine molecular and morphological data in a tree
- you can combine data from different genes
Some changes in DNA are largely “invisible” to selection;
mutations steadily accrue over time in a clock-like manner
- more distant relatives have more changes in their DNA than
close relatives
- true for non-coding junk DNA, introns, & silent substitutions
ATTCGTATTC
ATTCGTTTTC
ATTCGTTTTC
ATTCGTTTTC
ATTCGTATTC
Changes provide insight into patterns of relatedness (phylogeny)
DNA sequence data & phylogenetics
Drawback: any given site in a DNA sequence can only occupy
one of 4 possible “character states” -- A, T, G, or C
- “character” is the nucleotide at a given position in the DNA
- “state” is what the character looks like in a particular species
(a “G” at position 137 of the actin gene)
Mutations can change the nucleotide sequence, but then later
mutations can change the sequence back to the original one
such reversions can fool us (= homoplasy)
ATTCGTATTC
ATTCGTTTTC
ATTCGTTTTC
ATTCGTATTC
ATTCGTATTC
ATTCGTTTTC
ATTCGTATTC
DNA sequence data & phylogenetics
Mutations can change the nucleotide sequence, but then later
mutations can change the sequence back to the original one
This leads to a form of homoplasy when comparing DNA
sequences, since there are only 4 possible character states
- unlike with morphology, at least this kind of homoplasy
cannot result from convergent evolution; it’s “accidental”
Amino acids have 20 possible states, which is a bit better
- less chance of homoplasy from reverse-mutations to the
ancestral amino acid
- however, they can also be under selection  possibility
of convergent evolution exists
DNA sequence data & phylogenetics
Mutations can change the nucleotide sequence, but then later
mutations can change the sequence back to the original one
This leads to a form of homoplasy when comparing DNA
sequences, since there are only 4 possible character states
- unlike with morphology, at least this kind of homoplasy
cannot result from convergent evolution; it’s “accidental”
Take-home:
DNA sequence data & phylogenetics
Advantages of DNA-based methods:
- DNA sequences do not undergo convergent evolution, so
there’s no homoplasy like in morphological characters
- you can cheaply acquires 1000’s of nucleotides; each
position is potentially information
- hopefully, the huge amount of good data will drown out the
mis-information present due to “accidental” homoplasy

DNA sequence data & phylogenetics
Disadvantages of DNA-based methods:
- homoplasy arises from reverse-mutations, which makes
some nucleotide positions misleading
- data can be generated very quickly, but analysis is slow
and can require months of computing time (more to follow)
Evolution of whales: who’s the nearest relative?
Classical taxonomy:
Whales are a
sister group to
the Artiodactyla
common
ancestor of
whales and
other
Artiodactyls
New hypothesis: whales are
sister group
of hippos
4 kinds of DNA-based analysis
There are 4 ways to use DNA info to construct a phylogeny
(1) Genetic distance (fast) - “Neighbor-Joining” or NJ tree
- pairs up the closest sequences (lowest % difference) as sister
taxa, builds a tree from there
- discards info about specific changes, focuses on overall
similarities and differences
- quick and dirty, not very sophisticated
Genetic distance
Fast, especially when you have a
huge amount of sequence data
(otherwise many possible
trees; slow computation)
Branches show how much each of
2 sister taxa has diverged from
hypothetical common ancestor
4 kinds of DNA-based analysis
(2) Maximum parsimony (medium speed)
- uses only informative sites to draw the most parsimonious tree
- finds the minimum set of changes that had to occur to produce
the data you observe in present-day species
- also discards lots of information
Parsimony analysis and DNA sequence data
Can DNA sequence comparisons tell us who’s related to whales?
site 162:
C T
mutation
defines a
clade that
includes
hippos,
whales,
deer + cows
Parsimony analysis and DNA sequence data
Can DNA sequence comparisons tell us who’s related to whales?
site 166:
G C
mutation
defines
hippos +
whales as
a clade
Position 177 shows a reversion: a change from C to T at (1),
and change from T back to C at (2)
T
C
(2)
C
T
(1)
Every time you posit that a reversion occurred, you make
the tree less parsimonious (it costs you)
- doesn’t mean reversions don’t happen; just that they
are less likely than shared inheritance
Tree that requires the smallest number of changes is the
most parsimonious, and therefore the “chosen” phylogeny
Problem: there can be “ties” where multiple trees require the
same number of changes
- no way to chose among equally-parsimonious trees
4 kinds of DNA-based analysis
(3) Maximum likelihood (slow)
- uses a model of DNA sequence evolution to find the most
likely tree, given the data
- slow; searches all possible trees to find the best one
(4) Bayesian inference (medium speed)
- also uses model of DNA sequence evolution
- rather than looking for best tree, looks at many likely trees
(much faster than likelihood)
- tells you how often a particular relationship shows up
 i.e., what % of good trees show the clade [hippo+whale]
What is a model of DNA evolution?
0.01
1.0
0.5
0.1
- model may also include
a rate multiplier that
allows a given site to
10.3 mutate at 10-100 times
slower, or faster, than
the G
T rate
1.0
Model of DNA sequence evolution is a matrix of estimated rates
at which all the different mutations occurred, given your data
 rates are both inferred from your sequence data, and then
used to estimate the most likely tree given your data
4 kinds of DNA-based analysis
Advantage of likelihood/Bayesian methods:
- models of DNA sequence evolution include all sorts of rates
of change, which are estimated from the data and used by
computer programs to infer the likeliest trees
- used to estimate the likelihood of changes at each nucleotide
position, based on the overall sequence data
A
G
type of change
transition
C
G
transversion
likelihood
1.0
some changes are
more likely than others
0.1
C
A
transversion
0.01
DNA-based phylogenetic analysis
A
G
type of change
transition
C
G
transversion
0.1
sometimes happens
A
C
transversion
0.01
rarely ever happens
A
?
C
?
?
A
G
relative likelihood
1.0
often happens
Can we figure out what the
ancestral nucleotide was at a
certain position, if we know
the nucleotide at that position
in four related species today?
DNA-based phylogenetic analysis
A
G
type of change
transition
C
G
transversion
0.1
sometimes happens
A
C
transversion
0.01
rarely ever happens
A A
C
A
A
A
A
G
A
G
relative likelihood
1.0
often happens
C
Parsimony Likelihood
2 changes A
G often
A
C rarely
Only two changes required, but one is a very rare change
(likelihood doesn’t like to let rare things happen)
DNA-based phylogenetic analysis
A
G
type of change
transition
C
G
transversion
0.1
sometimes happens
A
C
transversion
0.01
rarely ever happens
G
A
A G
C
G
G
G
relative likelihood
1.0
often happens
A
G
G
A
C
Parsimony
3 changes,
don’t like
as much
Likelihood
G
A often
G
A often
G
C sometimes
Likelihood may favor more changes if they are likely changes
Tree parameters
In addition to estimaitng the model of DNA sequence evolution,
the computer program will draw the phylogeny that maximizes
the likelihood of seeing your sequence data, by altering 2 things:
1) the topology itself (the tree, or branching relationships)
C
B
C
A
likelihood: 1.2
0.04
B
A
B
C
A
37.1
Tree parameters
In addition to estimaitng the model of DNA sequence evolution,
the computer program will draw the phylogeny that maximizes
the likelihood of seeing your sequence data, by altering 2 things:
1) the topology itself (the tree, or branching relationships)
2) branch lengths (estimate of evolutionary time, or amount
of change, since a split in the tree)
C
B
C
A
C
A
B
likelihood: 10.2
B
B
A
C
A
B
7.1
C
A
C
A
12.5
B
Genes evolve at different rates
Some genes evolve faster than others, so you can pick a gene
appropriate to the problem you are tackling
If comparing closely related species, use fast-evolving
mitochondrial genes to show differences
 mitochondrial lack sophisticated proof-reading enzymes
of the nucleus; accumulate mutations faster
If comparing distantly related taxa, use a slow-evolving
nuclear gene, like 18S ribosomal RNA or histone 3 gene


Numbers are bootstrap support (maximum parsimony or
maximum likelihood) or posterior probabilities (Bayesian)
Statistical measure of how confident we are that a given node
is real (that relationships really are how they appear in the tree)
- significant bootstrap levels are >70%
- significant posterior probabilities are >90%
Numbers are bootstrap support (maximum parsimony or
maximum likelihood) or posterior probabilities (Bayesian)
E. pratensis
E. subornata
this clade is a polytomy:
the relationships of these
4 species are not resolved
(it’s a 4-way tie)
Elysia sp. 2
Elysia tomentosa
Elysia sp. 4, Japan
mitochondrial
COI gene
every slug
has a
slightly
different
sequence
nuclear histone H3 gene
Alderia modesta
all alleles
differ by
silent
substitutions
the species
are 20%
different at
this gene
within a species,
every slug has
the same allele
the two species are
1% different at
this conserved gene
Alderia willowi
Thuridilla
Boselia
basal, or “primitive”
Plakobranchus
Elysia - most
species-rich genus
My phylogeny of sea slug family
Elysiidae, based on 4 genes:
2 fast mitochondrial genes
2 slow-evolving nuclear genes
Sequences of the fast-evolving
mitochondrial COI gene help to
distinguish individuals within a
species from different populations
- Guam versus Australia
- different Caribbean islands
Also resolve the placement of
closely related (recently diverged)
species – i.e., near the tips of tree
Sequences of conserved, slow-evolving
nuclear genes help to resolve ancient
events (relationships among genera)
- things that happened near the root
of the tree, farther back in time
For instance, can help answer
group-level questions:
Who is the sister group of Elysia?
Sequences of conserved, slow-evolving
nuclear genes help to resolve ancient
events (relationships among genera)
- things that happened near the root
of the tree, farther back in time
For instance, can help answer
group-level questions:
Who is the sister group of Elysia?
… a clade of Thuridilla + Plako
Does DNA give you the same
“answer” (= phylogeny) as
morphological traits?
Why or why not, and which is better?
Morphological phylogeny
Morphology says,
Thuridilla is the
sister group of Elysia
4-gene molecular phylogeny
Boselia
Plakobranchus
Thuridilla
Elysiella
Elysia australis
Elysia ornata
Our says, nope –
Thuridilla is sister to
Plakobranchus
DNA sequence data & phylogenetics
Many kinds of molecular data are used in evolutionary biology:
a) DNA sequence comparisons
b) amino acid sequences of protein-coding genes
c) secondary structure of RNA (rRNA, tRNA genes)
d) presence/absence of introns, transposable elements
e) microsatellites
f) gene order on the chromosome
g) gene duplication events (presence/absence of genes)
h) expression patterns of genes (evo-devo)