Geological Dates and Molecular Rates: Fish DNA Sheds Light on Time
Christopher P. Burridge,* Dave Craw,? David Fletcher,? and Jonathan M. Waters*
*Department of Zoology, University of Otago, Dunedin, New Zealand; ?Department of Geology, University of Otago, Dunedin,
New Zealand; and ?Department of Mathematics and Statistics, University of Otago, Dunedin, New Zealand
Knowledge of DNA evolution is central to our understanding of biological history, but how fast does DNA change?
Previously, pedigree and ancient DNA studies—focusing on evolution in the short term—have yielded molecular rate
estimates substantially faster than those based on deeper phylogenies. It has recently been suggested that short-term,
elevated molecular rates decay exponentially over 1–2 Myr to long-term, phylogenetic rates, termed ‘‘time dependency of
molecular rates.’’ This transition has potential to confound molecular inferences of demographic parameters and dating of
many important evolutionary events. Here, we employ a novel approach—geologically dated changes in river drainages
and isolation of fish populations—to document rates of mitochondrial DNA change over a range of temporal scales. This
method utilizes precise spatiotemporal disruptions of linear freshwater systems and hence avoids many of the limitations
associated with typical DNA calibration methods involving fossil data or island formation. Studies of freshwater-limited
fishes across the South Island of New Zealand have revealed that genetic relationships reflect past, rather than present,
drainage connections. Here, we use this link between drainage geology and genetics to calibrate rates of molecular
evolution across nine events ranging in age from 0.007 Myr (Holocene) to 5.0 Myr (Pliocene). Molecular rates of change
in galaxiid fishes from calibration points younger than 200 kyr were faster than those based on older calibration points.
This study provides conclusive evidence of time dependency in molecular rates as it is based on a robust calibration
system that was applied to closely related taxa, and analyzed using a consistent and rigorous methodology. The time
dependency observed here appears short-lived relative to previous suggestions (1–2 Myr), which has bearing on the
accuracy of molecular inferences drawn from processes operating within the Quaternary and mechanisms invoked to
explain the decay of rates with time.
The ‘‘molecular clock’’ concept has underpinned re-
cent advances in evolutionary biology, facilitating hypoth-
esis testing and elucidating timescales of biodiversification
(Arbogast et al. 2002; Bromham 2003; Bromham and
Penny 2003). In parallel, use of coalescent theory has en-
abled molecular biologists to estimate demographic param-
eters such as effective population sizes and migration rates
(Kuhner et al. 1995; Beerli and Felsenstein 1999; Hey
2005). But the accuracy of such studies depends on knowl-
edge of underlying mutation rates. In the majority of cases,
rates of DNA change have been determined by calibrating
differences in DNA sequences against an estimate of line-
age divergence time, based on independent fossil or pale-
obiogeographic data, typically of Tertiary age (1.5–65 Myr;
e.g., Brown et al. 1979, 1982; supplementary table S1, Sup-
plementary Material online). More recently, rates of DNA
change have been estimated using precisely dated subfossil
material, upto60kyrold(e.g.,Lambert etal.2002;Shapiro
et al. 2004; Edwards et al. 2007). At the shallowest level,
rates have been calibrated directly from the number of
changes that have accumulated along pedigrees or mutation
accumulation lines (e.g., Denver et al. 2000; Howell et al.
2003; Santos et al. 2005).
Although it has long been recognized that different
taxa and genes experience different rates of DNA change
(Arbogast et al. 2002; Bromham and Penny 2003; Gillooly
et al. 2005), recent studies have also suggested that rates
decline with increasing evolutionary timescale: the so-
called ‘‘time dependency of molecular rates’’ (Ho et al.
2005; Ho and Larson 2006) or ‘‘lazy J’’ curve (Penny
2005). With respect to the mitochondrial DNA (mtDNA)
control region in humans, for example, pedigree rates are
fastest (e.g., 0.51 mutations/site/Myr; Santos et al. 2005),
followed by ancient DNA (aDNA) rates based on 10.3
kyr remains (0.34–0.44 mutations/site/Myr; Kemp et al.
2007), and then ‘‘phylogenetic’’ estimates derived from
Neogene primate divergences (0.05–0.24 mutations/site/
Myr; Santos et al. 2005; Emerson 2007). Although some
of this discrepancy in rates may be explained by differences
in methodologies and their errors (Bromham 2003;
Bromham and Penny 2003; Howell et al. 2003; Santos
et al. 2005), it appears that appropriate rates need to be ap-
plied not only with respect to taxon and gene, but also the
timescale of the question at hand (Ho et al. 2005; Ho and
Larson 2006). However, the time of transition between fast
‘‘pedigree-like’’ rates and asymptotic ‘‘deep-phylogeny’’
rates is under question (Ho et al. 2005; Emerson 2007;
Ho, Shapiro, et al. 2007). Satisfactory resolution of this is-
odology across a time frame ranging from thousands to
millions of years, in a group of closely related species.
In the current study, therefore, we assess molecular evolu-
tionary rates across a broad temporal scale (0.007–5.0
Myr), based on 9 geological events effecting isolation of
of galaxiid fishes (Waters and Wallis 2001; Waters and
New Zealand’s South Island (fig. 1) is a geologically
dynamic and mountainous region that has experienced ex-
tensive glaciation and tectonic uplift during the last 5 Myr
Key words: molecular clock, mutation rate, calibration, time
dependency, purifying selection.
Mol. Biol. Evol. 25(4):624–633. 2008
Advance Access publication February 14, 2008
? The Author 2008. Published by Oxford University Press on behalf of
the Society for Molecular Biology and Evolution. All rights reserved.
For permissions, please e-mail: firstname.lastname@example.org
(Sutherland 1996). This isolated, active landscape presents
a superb system for studying the effects of geological ac-
tivity on biological evolution. Recent geological studies
have revealed dynamic histories for many South Island
river systems (fig. 1; e.g., McAlpin 1992; Mortimer and
Wopereis 1997; Turnbull 2000; Craw and Norris 2003).
These histories include a number of river reversal and riv-
er capture events (fig. 2)—driven by tectonics and
glaciation—that involved wholesale transfer of streams be-
tween adjacentrivercatchments.Suchriver evolutioneffects
vicariant isolation of freshwater-limited fish populations
that were previously connected (fig. 2) and can lead to clad-
ogenesis and the formation of new species (e.g., Waters and
Wallis 2001; Burridge et al. 2006; Waters et al. 2007). In
addition, the extent of genetic divergence among popula-
tions is strongly correlated with their age of physical
isolation, as determined using geological techniques—
correlation of river terrace profiles with Oxygen Isotope
glacial stages (Craw and Waters 2007). Our study uses this
causal link between drainage geology and genetics to quan-
tify rates of DNA change. This novel calibration method is
based on precise spatiotemporal disruptions of linear fresh-
water systems and hence avoids many of the limitations as-
sociated with typical DNA calibration methods, where the
link between a geological event (e.g., fossil deposition, is-
land formation) and population isolation is not necessarily
Here, we use 7 geologically dated river evolution
events—ranging from 7,000 to 500,000 years old—to
measure rates of mtDNA change in New Zealand
FIG. 1.—The study area. Numbers represent dated freshwater isolation events employed during calibration of molecular rates in New Zealand
fishes (see table 1 for details). Blue arrows represent historical river directions, whereas red arrows represent contemporary directions. Scale bars are
50 km. (A) South Island, New Zealand, showing the Alpine Fault and eastern uplift of the Southern Alps. The Chatham Islands were formed approximately
800 km east of New Zealand. (B) Detailed view of the northeast South Island, showing river capture, reversal, and valley drowning events (numbers 6,
5, and 9, respectively). (C) Detailed view of the southwest South Island, showing river capture (numbers 4, 7, and 8) and reversal (number 3) events.
(D) Examples of fish phylogeographic patterns reflecting historical rather than contemporary drainage connections. Nevis (3) and Von (8) rivers contain
Galaxias gollumoides (yellow circles) rather than Galaxias sp. (green circles), reflecting the former southern connections of these drainages.
Time Dependency of Fish Molecular Rates625
freshwater-limited fishes. In addition, we employ 2 older
events: the Pliocene uplift of the Southern Alps and the
mid-Pleistocene emergence of the Chatham Islands (fig.
tectonic events lack robust upper and lower bounds of time
since population isolation but in combination provide
a deeper level contrast of mtDNA rates. In total, 13 inde-
pendent estimates of molecular rates are calculated for gal-
axiid fishes (Galaxiidae) and 3 for an eleotrid fish species
(Eleotridae), spanning 0.007–5.0 Myr (table 1). In each
case, mitochondrial cytochrome b and control region se-
quences have been obtained from populations with known
geological isolation histories, and a consistent coalescent
methodology is employed to estimate associated rates of
Materials and Methods
Fish sampling (galaxiids, eleotrids) associated with
each calibration point is summarized in table 1 and was ap-
proved by the University of Otago Animal Ethics Commit-
tee. At least 50 individuals were analyzed for 13 of the 16
independent calibrations performed. This does not include
region to confirm that genetic relationships were consistent
with thehypothetical paleodrainage connections rather than
some unexpected geological history. Sampling effort was
geographically spread throughout the catchments that
had undergone modification to their paleodrainage bound-
aries, such that the majority of genetic diversity would be
encompassed. Intracatchment phylogeographic structuring
is typically absent in these taxa (e.g., Burridge et al. 2006;
Waters et al. 2007) and appears only in conjunction with
large in-stream features such as gorges (Burridge et al.
2007), which are lacking within the systems we have em-
ployed for calibrations. Consequently, the geological age of
isolation events should be applicable to the patterns of mo-
lecular variation observed. Laboratory protocols for the se-
quencing of mitochondrial cytochrome b gene and control
region were as previously described (Waters et al. 2007).
For galaxiids, cytochrome b sequences were at least 762
bp and control region 664 bp. Corresponding Gobiomor-
phus breviceps (Eleotridae) sequences were 900 and
385 bp. Potential heteroplasmies were scored as ambiguous
nucleotides, which were subsequently ignored during coa-
lescent estimation of rates. Sequence data are available
from GenBank under accession numbers reported previ-
et al. 2007) plus EU309048–469. Phylogeographic rela-
tionships were reconstructed via maximum parsimony,
maximum likelihood, and Bayesian analysis and were con-
cordant with the geological histories of river isolation
predicted a priori (Waters and Wallis 2001; Burridge
et al. 2006, 2007; Waters et al. 2007). An overview of
phylogenetic relationships among galaxiids, and inferred
geological isolation events, is given in supplementary
figure S1, Supplementary Material online. For Gobiomor-
phus breviceps, Smith et al. (2005) detail the divergence
populations on either side of the Southern Alps, and Waters
et al. (2007) describe the relationships among the Pelorus,
Kaituna, and Wairau rivers.
FIG. 2.—River capture and river reversal. Freshwater-limited populations located in stream sections ‘‘A’’ and ‘‘B’’ become genetically isolated
following changes to river drainage geometry. ‘‘River capture’’ represents the displacement of stream sections between adjacent catchments, vicariantly
isolating populations that previously inhabited the same system. Similarly, ‘‘river reversal’’ within part of a drainage will result in vicariant isolation of
populations on either side of the new drainage divide, as will marine inundation of a stream junction (data not shown).
626 Burridge et al.
The majority of calibration points were provided by
changes in river drainage geometry (table 1), and the ages
of such events were inferred by correlation of river terraces
with previous glacial stages. River terraces are remnants of
old river gravel deposits preserved on valley walls. These
gravel deposits accumulated at times of high sediment sup-
ply inthe catchments, typicallyduring glaciations, and their
gradients reflect paleodrainage direction. River terraces of-
ten occur as a series of ‘‘steps’’ up a valley wall, reflecting
the past series of glaciations, and their relative ages can be
geomorphic features throughout river catchments. Absolute
of global glaciation that have been determined with oxygen
isotope ratios in the marine sedimentary record. A well-
calibrated oxygen isotope timescale has been defined
internationally (Chappell et al. 1996). Optically stimulated
luminescence and radiocarbon dating of material from river
terraces were employed to verify correlations with the in-
ternational timescale. Three calibration dates were esti-
mated from uplift rates and age of island emergence
(table 1). Many of the paleodrainage events have ages con-
strained by both older (higher) and younger (lower) river
terraces, thereby providing both objective and temporally
accurate upper and lower bounds of population isolation,
as opposed to events such as island colonization, where up-
per bounds may be erroneous (Emerson 2007) and lower
bounds are subjective.
Coalescent simulations under the 4-parameter isola-
tion model (Wakeley and Hey 1997) were employed to es-
timate ratesofDNAchangeusing theIMsoftware.Because
the mitochondrial cytochrome b and control regions are
linked during inheritance and their proportions of variable
sites were similar (cytochrome b relative to control region:
0.73–2.24 in galaxiid data sets, 0.62–1.00 in G. breviceps
data sets), they were combined for the estimation of a single
mutation rate, to maximize the information content of
the data. Nucleotide change was assumed to follow the
Hasegawa-Kishino-Yano (HKY) model, which is the only
model implemented by IM that can accommodate multiple
substitutions at sites. This model appears adequate for the
majorityof the data sets, as HKY distances increase in a lin-
ear manner with those obtained under models selected from
56 candidates using a hierarchical likelihood ratio test
(Posada and Crandall 1998)(supplementary fig. S2,Supple-
mentary Material online). The nonlinear relationships ob-
served from 3 of the older calibration data sets represent
underestimation of molecular change at deeper divergence
levels when using the HKY model, which could introduce
slight error during the estimation of mutation rates.
Using the IM software, genealogical topologies were
simulated and updated along a Markov chain Monte Carlo
during which the 4 model parameters—divergence time (t),
2 contemporary and 1 ancestral population sizes (h)—were
were set such that posterior distributions were fully
Summary of Freshwater Isolation Events Employed for the Calibration of Mutation Rates
Isolation Event(number in fig. 1)
Age (Myr), Mechanism
1 versus 2
p distance HKY distance Bayesian
1. Southern Alps
2.0–5.0, mountain uplift
Galaxias divergens, G. paucispondylus
Wellman (1979); Sutherland (1996)
2. Chatham Is.
,1.0, island formation
Neochanna burrowsius, N. rekohua
Campbell et al. (2006)
0.300–0.500, river reversal Galaxias gollumoides
Waters and Wallis (2001); Youngson et al. (2002)
0.145–0.240, river capture
Turnbull (2000); Craw, Burridge, et al. (2007)
0.070–0.130, river reversal G. divergens
Lauder (1970); Mortimer and Wopereis (1997);
Craw, Anderson, et al. (2007); Craw and Waters (2007)
6. Clarence–Wairau 0.010–0.020, river capture
McAlpin (1992); Burridge et al. (2006);
Rattenbury et al. (2006)
,0.020, river capture
0.012, river capture
Turnbull (1980, 2000); Craw and Norris (2003)
0.007, valley drowning
Gibb (1986); Singh (1994)
aTest not possible due to polyphyletic relationships.
bMaximum intrapopulation divergence exceeds maximum interpopulation divergence.
*P , 0.05.
Time Dependency of Fish Molecular Rates627
contained within them, with peaks representing maximum
likelihood estimates. A minority of h posteriors contained
infinite nonzero tails, but repeating analyses with different
Carlo searches employed 4 chains, 3 of which were incre-
mentally heated to promote broader searching of parameter
space. After a burn-in period of 105generations, parameter
trendline plots were examined for the attainment of statio-
ery 10 generations. Independence of samples collected
along a chain was assessed via their autocorrelation statis-
tics, and runs were continued until the effective sample size
exceeded 100 for all parameters. Independent runs using
different random number seeds were conducted to assess
convergence upon the true stationary distribution. Indepen-
dence among parameter estimates was checked. Rates of
DNA change (u) were calculated based on geologically de-
rived estimates of maximum generations elapsed since pop-
ulation isolation (t), using the relationship t 5 tu. We
employed female generation times of 1 and 2 years for gal-
axiids and eleotrids, respectively (Hopkins 1971; Staples
1975; McDowall 2000; Hamilton and Poulin 2001).
Simple estimates of rates were also derived from net
sequence distance, Dnet50:5dxy? 0:5maxfdx;dyg; where
d is the sequence divergence either between (dxy) or within
(dx, dy) populations y and x. We employed both uncorrected
(p) distances and HKY distances. The latter provided com-
parisons for calibrations derived from the 4-parameter iso-
lation model when the genealogies underlying the data sets
were likely to be reciprocally monophyletic, as the time pa-
rameter (t) may be strongly and negatively correlated with
ancestral h under such conditions (Wakeley and Hey 1997).
Tests of purifying selection were employed using the
likelihood scores approach of Hasegawa et al. (1998). Like-
lihood scores were calculated for cytochrome b data sets
gle ratio of nonsynonymous to synonymous nucleotide
changes (dN:dS) or 2 such ratios, with one constrained
to population clades (e.g., catchments) and the other con-
strained to the branch linking population clades. Tree topol-
ogies andinitialbranch lengthswerederivedfromBayesian
analysis of cytochrome b and control region data, but
branch lengths were subsequently reestimated based on
the cytochrome b data set and the M0 codon model (Yang
et al. 2000). If purifying selection contributed to any time
model will explain the data significantly better than the
1-ratio model as the interpopulation branch should have
a lower dN:dS than the intrapopulation clades; this was as-
sessed by likelihood ratio tests.
Results and Discussion
Time Dependency in Galaxiid and Eleotrid Fishes
The analysis of freshwater fish vicariance suggests
a decline in mtDNA evolutionary rates with increasing
age of calibration. Specifically, from coalescent analysis
we observe rates of 0.031–0.125 changes/site/Myr from
river isolation events younger than 200 kyr in galaxiid
fishes (table 1 and fig. 3). In contrast, galaxiid rates derived
from older events are slower and less variable, in the order
of 0.011–0.026 changes/site/Myr (table 1 and fig. 3), and
compatible with most fish mtDNA rates based on older iso-
lation events (supplementary table S1, Supplementary
Material online). Confidence intervals of model parameters
derived from weighted nonlinear least squares regression
rejected a constant rate of molecular change with time (sup-
plementary text S1, Supplementary Material online). The
transition in coalescent-based rate estimates is also ob-
served for rates derived from net sequence divergences
derived from p or HKY distances are broadly consistent
with the coalescent rate estimates, but the larger discrep-
ancies that exist may reflect inappropriate correction for
sequence divergence present at the time of population iso-
lation when using maximum contemporary intrapopulation
divergence as a proxy. Strong negative correlations be-
tween the time parameter (t) and ancestral h are expected
during coalescent analysis of data sets likely to represent
reciprocally monophyletic genealogies (Wakeley and Hey
1997), which could bias calibrations based on older isolation
events. However, such correlations were positive or only
slightly negative (e.g., ?0.12) for the majority of the recip-
rocally monophyletic fish data sets, and rate estimates were
similartosimpleestimates derived underthesame mutation
model (HKY; table 1 and fig. 3).
The time-dependent transition observed here cannot
be explained by interspecific rate variation (Gillooly
et al. 2005); the calibrations are based on a closely related
Material online; Galaxiidae: Galaxias and Neochanna spp.;
of 8 transitions from a younger calibration event to the next
oldest calibration within the same species of galaxiid were
accompanied by a decline in molecular rate (fig. 3); the ex-
ception was the transition from 7 to 20 kyr calibrations for
Galaxias divergens. ‘‘Time dependency’’ of a similar mag-
nitude was also observed for a completely unrelated taxon,
an eleotrid fish (G. breviceps), confirming that this pattern
is not peculiar to the Galaxiidae (fig. 3).
Sequencing errors in the order of 1 in 1,000 bases
(Wesche et al. 2004) could explain some of the elevated
molecular rates we observed from calibrations of 20 kyr
or younger, assuming that the true molecular rate in these
fishes approximates the asymptotic estimate obtained here
for galaxiids (0.01876 changes/site/Myr; supplementary
text S1, Supplementary Material online). However, it
should be reiterated that the rates presented here were de-
rived from maximum ages of population isolation and
hence are likely to have been underestimated, perhaps sub-
stantially. In addition, much higher sequencing error rates
are required to explain elevated molecular rates obtained
from older calibrations (see also Ho et al. 2005). Sequenc-
ing errors are also likely to be substantially less influential
than aDNA damage, which has recently been refuted as
a major contributor to elevated molecular rates observed
from calibrations across short timescales (Ho, Heupink,
et al. 2007).
628Burridge et al.
Finally, errors in estimated ages of population isola-
molecular rates with time, given the temporal constraints
provided by geological information. We therefore conclude
that time dependency is the best explanation for the calibra-
tion data presented here.
The Duration of Time Dependency
Theoretical (Penny 2005) and some empirical (Howell
et al. 2003; Santos et al. 2005) evidence for time depen-
dency is strong, but determining the point at which
short-term (e.g., pedigree, aDNA) molecular rates decay
to long-term (i.e., deep-phylogenetic) levels is crucial for
genetic estimates of evolutionary timescales and demo-
graphic parameters (Ho et al. 2005; Ho and Larson
2006; Ho, Shapiro, et al. 2007). In galaxiid fishes, we ob-
serve ‘‘deep-phylogenetic’’ mtDNA rates (0.011–0.026
changes/site/Myr) beyond 200 kyr (fig. 3). Phylogenetic
rates may also be attained earlier than 200 kyr, given the
larger uncertainty surrounding rates from progressively
younger calibration points. Therefore, our study suggests
that elevated mtDNA rates might not persist much beyond
200 kyr in galaxiid fishes.
FIG. 3.—Molecular clock calibrations (nucleotide changes/site/Myr) derived from dated isolation events of New Zealand freshwater fish
populations. Blue symbols represent galaxiid divergences (Galaxiidae), and brown symbols represent Gobiomorphus breviceps (Eleotridae). Letters
indicate rates derived from multiple events for the same species (g, Galaxias gollumoides; d, Galaxias divergens; s, Galaxias ‘southern’; p, Galaxias
paucispondylus; b, Gobiomorphus breviceps). Where ages are represented by both minimum and maximum estimates (table 1), we employed the latter,
yielding minimum estimates of rates. The top graph represents coalescent-based estimates (error bars are the 90% highest posterior density). The red
lines are vertically translated exponential decay curves for galaxiid data representing best-fit estimates (solid line, y 5 0.01876 þ 0.03911 ? e-5.25878x)
and upper and lower 95% confidence intervals of model parameters (supplementary text S1, Supplementary Material online). The bottom graph
represents ‘‘simple’’ rates based on net sequence divergences under either uncorrected (p distance, triangles) or HKY (diamonds) models of nucleotide
substitution. Two young calibration events yielded negative simple rates (maximum intracatchment divergence exceeded intercatchment divergence,
table 1) and are not shown on the lower graph.
Time Dependency of Fish Molecular Rates629
mtDNA rates at ?200 kyr in galaxiid fishes (see above)
would contrast with time dependent curves proposed for
that the duration of time dependency in mtDNA extended
up to 2 Myr for primate protein-coding genes and up to 1
Myr for primate control region and avian protein-coding
genes. While the duration of time dependency may vary
among taxa, reflecting differences in generation time, effec-
tive population size, instantaneous mutation rate, and selec-
tion, a reanalysis of the Ho et al. (2005) data by Emerson
(2007) provided no evidence of elevated rates for events
older than 100 kyr, more consistent with our suggestions
for galaxiid fishes of ?200 kyr. Although a recent analysis
of cichlid fish mtDNA control region sequences (Genner
et al. 2007) suggested a pattern similar to that proposed
by Ho et al. (2005), with attainment of asymptotic molec-
ular rates after ?1 Myr, there were no calibrations older than
50 kyr until 0.7 Myr, and the age of this and the next oldest
calibration point were inferred from a molecular clock
rather than directly from geological data. Therefore, an ear-
More recently, Ho, Shapiro, et al. (2007) argued for
elevated mtDNA evolutionary rates in birds at scales
greater than 100 kyr. However, their rock partridge rate
of 0.125 substitutions/site/Myr—derived from a 238-kyr
calibration point—is actually a lineage specific rate of
0.062 substitutions/site/Myr based on a 2-Myr isolation
event, whose age was itself inferred from a molecular clock
(Randi et al. 2003). Ho, Shapiro, et al. (2007) additional
calibrations beyond 100 kyr yielding elevated rates are
based on island emergence events and limited sampling
(maximum 3 individuals per island), which could lead to
overestimates of rates for several reasons, such as lineages
predating island emergence or extinction of the closest
mainland relative (Emerson 2007). Some elevated rates
from events within the last 100 kyr may also require rein-
terpretation. For example, the Bison aDNA (0–60 kyr) rates
reported by Ho, Shapiro, et al. (2007) are consistent with a
phylogenetic rate derived from Bison–Bos divergence 1 Myr
before present (Troy et al. 2001; see also Shapiro et al.
2004; Edwards et al. 2007). Hence, the duration of time de-
pendency in birds, mammals, and cichlid fishes may match
our suggestion for galaxiid fishes (?200 kyr), or in all
cases, couldbemuch morerecent.Furtherworkonthissub-
ject is required, employing robust calibration points.
of asymptoticor deep-phylogenetic
Mechanisms of Time Dependency
The suggestion of a relatively rapid (,200 kyr) at-
tainment of asymptotic (deep-phylogenetic) rates in fishes
from this study, which may also apply to birds, mammals,
and cichlid fishes (see above), has implications for the
mechanistic cause of time dependency. Recent studies have
suggested that slightly deleterious mtDNA mutations con-
tribute to rapid divergences across shallow timescales prior
to their removal by purifying selection (Howell et al. 2003;
Ho et al. 2005; Penny 2005; Santos et al. 2005), which is
consistent with observations of higher ratios of nonsynon-
ymous to synonymous nucleotide changes (dN:dS) at shal-
low versus deep phylogenetic levels (Hasegawa et al. 1998;
Ho et al. 2005; Burridge et al. 2006; Kivisild et al. 2006;
Rocha et al. 2006). Although recent simulations suggested
that purifying selection alone could only explain time
dependency if there were many significantly deleterious
mutations and effective population sizes were implausibly
large (Woodhams 2006), these results hinged on the decay
curves reported by Ho et al. (2005). But if elevated rates do
not persist much beyond 100 kyr, as we contend, time
dependency might be readily explained by purification se-
lection under plausible population sizes.
Analyses of dN:dS revealed significant signatures of
purifying selection for 3 of the galaxiid data sets, represent-
ing5Myr,0.5Myr,and 0.24-Myrisolationevents (table1).
Even in the nonsignificant data sets, dN:dS within catch-
ments (xW) was greater than that along the branch between
catchments (xB, table 1). Consequently, the suggestion of
purifying selection is quite persuasive, given the limited
amount of nonsynonymous variation present in our data
sets,reducing statistical power. At this stage,however, con-
tributions by some alternative mechanisms cannot be dis-
counted; these involve drift and random fluctuations in
population size (Zhivotovsky et al. 2006) or selective
sweeps via linkage to sites experiencing positive selection,
as recently proposed for mtDNA (Bazin et al. 2006). None
of the above mechanisms are mutually exclusive.
Mutational hot spots—sites that undergo molecular
change at very short time intervals but quickly become sat-
urated such that the apparent accumulation of change
decreases with time—have previously been invoked for ob-
servations of elevated rates across short timescales (Pa ¨a ¨bo
1996; Jazin et al. 1998). However, it does not appear that
relates predominantly to mutational hot spots. We have as-
sessed the location of variable sites and the numbers of nu-
of galaxiids. However, there is no similarity or clustering in
the locations of intraclade polymorphisms in either cyto-
chrome b or control region, and very few polymorphic sites
exhibit more than 2 nt states, as would be anticipated if
rapid molecular divergence across short timescales was
the result of mutational hot spots (Burridge et al. 2006).
However, mutational hot spots appear to make contribu-
tions to elevated rates in other data sets, where regions
of repeated nucleotide sequences exist (Denver et al.
2000), although in that instance, as we suggest in ours,
the presence of mutational hot spots does not appear to
be a prerequisite for the observation of elevated rates across
Implications of Time Dependency in Fishes
A major implication of time dependency is that esti-
mates of population divergence times and demographic pa-
rameters may require reestimation (Ho et al. 2005; Ho and
Larson 2006). For example, female effective population
size (Nef) in shortnose sturgeon (Acipenser brevirostrum),
derived from aphylogeneticmtDNAcontrolregioncalibra-
tion of 0.018 changes/site/Myr, exceeded a census pop-
ulation estimate by over an order of magnitude (Quattro
et al. 2002), whereas the reverse relationship is expected
(Turner et al. 2006). In contrast, a calibration of
630 Burridge et al.
0.109–0.131 changes/site/Myr—derived from a postglacial
transmontanus)—yielded Neffor the endangered Chinese
sturgeon (Acipenser sinensis) consistent with observational
estimates of effective size (Zhang et al. 2003). The white
sturgeon calibration also produced molecular estimates
of east Atlantic colonization by Acipenser oxyrinchus that
closely matched a Holocene first appearance in the archae-
ological record (Ludwig et al. 2002), whereas a phyloge-
netic rate of 0.018 would have exceeded this by 10 kyr.
Similarly, many studies have invoked previously unex-
pected glacial refugia to accommodate molecular estimates
of population divergence time that predate the last glacial
2006). However, the time-dependent rates observed here
will convert such molecular divergences into inferences
of post-Pleistocene population isolation. This problem
may be particularly important for studies using genetic di-
vergences toformulatepredictionsof biological responseto
future climate change.
of time dependency in rates of molecular change for
mtDNA regions in freshwater-limited fishes. The duration
of time dependency in these fishes appears shorter than
some suggestions for other taxa (Ho et al. 2005; Genner
et al. 2007; Ho, Shapiro, et al. 2007), but we contend that
shorter durations may apply to these groups as well. Under
such a scenario, time dependency may be more readily ex-
plained by purifying selection than previously inferred
(Woodhams 2006), although other mechanisms are also
likely to contribute (drift and random fluctuations in pop-
ulation size, positive selection, and mutational hot spots
associated with repeated nucleotide sequences). However,
a shorter duration of time dependency does not remove the
need for a careful reassessment of published estimates of
population divergence times and demographic parameters
derived from deep-phylogenetic molecular rate calibra-
tions. However, the news is not all bad; using established
molecular techniques, but with an appreciation of time de-
pendency, we can now address important evolutionary
questions that were previously considered too recent for
are available at Molecular Biology and Evolution online
We thank the following people for assistance with the
collection of specimens: R. Allibone, S. Charteris,
K. Garrett, D. Jack, R. McDowall, P. Ravenscroft,
D. Rowe, M. Rutledge, and G. Wallis. D. Gleeson provided
DNA from her published study of Neochanna burrowsius.
L. Anderson, T. King, and D. Rowe assisted with
DNA sequencing of specimens. Geographx (http://www.
geographx.co.nz/) provided the images for figure 1, and
K. Miller assisted with the production of illustrations.
The research was funded by Marsden contract UOO0404
(Royal Society of New Zealand) and a University of Otago
Research Grant. H. Edmonds, S. Charteris, and E. Edwards
(N.Z. Department of Conservation) facilitated the provision
of collection permits, and the researchwas performed under
Otago University Animal Ethics permit 16/03. Jody Hey
and 2 anonymous reviewers made comments that improved
Arbogast BS, Edwards SV, Wakeley J, Beerli P, Slowinski JB.
2002. Estimating divergence times from molecular data on
phylogenetic and population genetic timescales. Ann Rev
Ecol Syst. 33:707–740.
Bazin E, Glemin S, Galtier N. 2006. Population size does not
influence mitochondrial genetic diversity in animals. Science.
Beerli P, Felsenstein J. 1999. Maximum-likelihood estimation of
migration rates and effective population numbers in two
Bromham L. 2003. Molecular clocks and explosive radiations. J
Mol Evol. 57:S13–S20.
Bromham L, Penny D. 2003. The modern molecular clock. Nat
Rev Genet. 4:216–224.
Brown WM, George M, Wilson AC. 1979. Rapid evolution of
animal mitochondrial-DNA. Proc Natl Acad Sci USA.
Brown WM, Prager EM, Wang A, Wilson AC. 1982.
Mitochondrial-DNA sequences of primates—tempo and mode
of evolution. J Mol Evol. 18:225–239.
Burridge CP, Craw D, Waters JM. 2006. River capture, range
expansion, and cladogenesis: the genetic signature of
freshwater vicariance. Evolution. 60:1038–1049.
Burridge CP, Craw D, Waters JM. 2007. An empirical test of
Campbell HJ, Begg JG, Beu AG, Carter RM, Davies G, Holt K,
Landis CA, Wallace C. 2006. On the turn of a scallop. Geol
Soc N Z Misc Publ. 121:9.
Chappell J, Omura A, Esat T, McCulloch M, Pandolfi J, Ota Y,
Pillans B. 1996. Reconciliation of late Quaternary sea levels
derived from coral terraces at Huon Peninsula with deep sea
oxygen isotope records. Earth Planet Sci Lett. 141:
Craw D, Anderson L, Rieser U, Waters JM. 2007. Drainage
reorientation in Marlborough Sounds, New Zealand, during
the last interglacial. N Z J Geo Geophys. 50:13–20.
Craw D, Burridge C, Anderson L, Waters JM. 2007. Late
Quaternary river drainage and fish evolution, Southland, New
Zealand. Geomorphology. 84:98–110.
Craw D, Norris RJ. 2003. Landforms. In: Darby J, Fordyce RE,
Mark A, Probert K, Townsend C, editors. The natural history
of southern New Zealand. Dunedin (New Zealand): Univer-
sity of Otago Press. p. 17–34.
Craw D, Waters J. 2007. Geological and biological evidence
for regional drainagereversal
transport, Marlborough, New Zealand. J Geol Soc. 164:
river capture.Mol Ecol.
during lateral tectonic
Time Dependency of Fish Molecular Rates631
Culling MA, Janko K, Boron A, Vasil’Ev VP, Cote IM,
Hewitt GM. 2006. European colonization by the spined loach
(Cobitis taenia) from Ponto-Caspian refugia based on
mitochondrial DNA variation. Mol Ecol. 15:173–190.
Davey ML, O’Brien L, Ling N, Gleeson DM. 2003. Population
genetic structure of the Canterbury mudfish (Neochanna
burrowsius): biogeography and conservation implications.
N Z J Mar Freshw Res. 37:13–21.
Denver DR, Morris K, Lynch M, Vassilieva LL, Thomas WK.
2000. High direct estimate of the mutation rate in the
mitochondrial genome of Caenorhabditis elegans. Science.
Edwards CJ, Bollongino R, Scheu A, et al. (40 co-authors). 2007.
Mitochondrial DNA analysis shows a Near Eastern Neolithic
origin for domestic cattle and no indication of domestication
of European aurochs. Proc Biol Sci. 274:1377–1385.
Emerson BC. 2007. Alarm bells for the molecular clock? No
support for Ho et al.’s model of time-dependent molecular
rate estimates. Syst Biol. 56:337–345.
Genner MJ, Seehausen O, Lunt DH, Joyce DA, Shaw PJ,
Carvalho GR, Turner GF. 2007. Age of cichlids: new
dates for ancient fish radiations. Mol Biol Evol. 24:
Gibb JG. 1986. A New Zealand regional Holocene eustatic sea-
level curve and its application to determination of vertical
tectonic movements. R Soc N Z Bull. 24:377–395.
Gillooly JF, Allen AP, West GB, Brown JH. 2005. The rate of
DNA evolution: effects of body size and temperature on the
molecular clock. Proc Natl Acad Sci USA. 102:140–145.
Gum B, Gross R, Kuehn R. 2005. Mitochondrial and nuclear
DNA phylogeography of European grayling (Thymallus
thymallus): evidence for secondary contact zones in central
Europe. Mol Ecol. 14:1707–1725.
Hamilton WJ, Poulin R. 2001. Parasitism, water temperature and
life history characteristics of the freshwater fish Gobiomor-
phus breviceps Stokell (Eleotridae). Ecol Freshw Fish.
Hasegawa M, Cao Y, Yang ZH. 1998. Preponderance of
slightly deleterious polymorphism in mitochondrial DNA:
nonsynonymous/synonymous rate ratio is much higher
within species than between species. Mol Biol Evol. 15:
Hey J. 2005. On the number of New World founders: a population
genetic portrait of the peopling of the Americas. PLoS Biol.
Ho SYW, Heupink TH, Rambaut A, Shapiro B. 2007. Bayesian
estimation of sequence damage in ancient DNA. Mol Biol
Ho SYW, Larson G. 2006. Molecular clocks: when times are
a-changin’. Trends Genet. 22:79–83.
Ho SYW, Phillips MJ, Cooper A, Drummond AJ. 2005. Time
dependency of molecular rate estimates and systematic
overestimation of recent divergence times. Mol Biol Evol. 22:
Ho SYW, Shapiro B, Phillips MJ, Cooper A, Drummond AJ.
2007. Evidence for time dependency of molecular rate
estimates. Syst Biol. 56:515–522.
Hopkins CL. 1971. Life history of Galaxias divergens (Salmo-
noidea: Galaxiidae). N Z J Mar Freshw Res. 5:41–57.
Howell N, SmejkalCB, Mackey
Turnbull DM, Herrnstadt C. 2003. The pedigree rate of
sequence divergence in the human mitochondrial genome:
there is a difference between phylogenetic and pedigree rates.
Am J Hum Genet. 72:659–670.
Jazin E, Soodyall H, Jalonen P, Lindholm E, Stoneking M,
Gyllensten U. 1998. Mitochondrial mutation rate revisited:
hot spots and polymorphism. Nat Genet. 18:109–110.
Kemp BM, Malhi RS, McDonough J, et al. (14 co-authors).
2007. Genetic analysis of early Holocene skeletal remains
from Alaska and its implications for the settlement of the
Americas. Am J Phys Anthropol. 132:605–621.
Kivisild T, Shen PD, Wall DP, et al. (17 co-authors). 2006. The
role of selection in the evolution of human mitochondrial
genomes. Genetics. 172:373–387.
Kuhner MK, Yamato J, Felsenstein J. 1995. Estimating
effective population size and mutation rate from sequence
data using Metropolis-Hastings sampling. Genetics. 140:
Lambert DM, Ritchie PA, Millar CD, Holland B, Drummond AJ,
Baroni C. 2002. Rates of evolution in ancient DNA from
Ade ´lie penguins. Science. 295:2270–2273.
Lauder WR. 1970. Ancient drainage of Marlborough-Sounds.
N Z J Geo Geophys. 13:747–749.
Ludwig A, Debus L, Lieckfeldt D, Wirgin I, Benecke N,
Jenneckens I, Williot P, Waldman JR, Pitra C. 2002. When
the American sea sturgeon swam east—a colder Baltic Sea
greeted this fish from across the Atlantic Ocean in the Middle
Ages. Nature. 419:447–448.
McAlpin JP. 1992. Glacial geology of the upper Wairau Valley,
Marlborough, New Zealand. N Z J Geo Geophys. 35:
McDowall RM. 2000. The Reed field guide to New Zealand
freshwater fishes. Auckland (NZ): Reed.
Mortimer N, Wopereis P. 1997. Change in direction of the
Pelorus river, Marlborough, New Zealand: evidence from
composition of Quaternary gravels. N Z J Geo Geophys.
Pa ¨a ¨bo S. 1996. Mutational hot spots in the mitochondrial
microcosm. Am J Hum Genet. 59:493–496.
Penny D. 2005. Evolutionary biology—relativity for molecular
clocks. Nature. 436:183–184.
Posada D, Crandall KA. 1998. MODELTEST: testing the model
of DNA substitution. Bioinformatics. 14:817–818.
QuattroJM, GreigTW, Coykendall
Baldwin JD. 2002. Genetic issues in aquatic species man-
agement: the shortnose sturgeon (Acipenser brevirostrum) in
the southeastern United States. Conserv Genet. 3:155–166.
Randi E, Tabarroni C, Rimondi S, Lucchini V, Sfougaris A.
2003. Phylogeography of the rock partridge (Alectoris
graeca). Mol Ecol. 12:2201–2214.
Rattenbury MS, Townsend DB, Johnston MR. 2006. Geology of
the Kaikoura area. 1:250000 Geological map 13. Lower Hutt
(New Zealand): New Zealand Institute of Geological and
Nuclear Sciences Limited.
Rocha EPC, Smith JM, Hurst LD, Holden MTG, Cooper JE,
Smith NH, Feil EJ. 2006. Comparisons of dN/dS are time
dependent for closely related bacterial genomes. J Theor Biol.
Santos C, Montiel R, Sierra B, Bettencourt C, Fernandez E,
Alvarez L, Lima M, Abade A, Aluja MP. 2005. Understanding
mtDNA mutation rate: a model using families from the Azores
Islands (Portugal). Mol Biol Evol. 22:1490–1505.
Shapiro B, Drummond AJ, Rambaut A, et al. (27 co-authors).
2004. Rise and fall of the Beringian steppe bison. Science.
Singh LJ. 1994. Late Quaternary history of Pelorus Sound from
shallow seismic reflection and core analysis. Geol Soc N Z
Misc Publ. 80A:165.
Smith PJ, McVeagh SM, Allibone R. 2005. Extensive genetic
Zealand. J Fish Biol. 67:627–639.
Staples DJ.1975. Production
Philypnodon breviceps Stokell in a small New Zealand lake.
632Burridge et al.
1. Life-history, food, feeding and activity rhythms. J Fish Download full-text
Sutherland R. 1996. Transpressional development of the
Australia-Pacific boundary through southern South Island,
New Zealand: constraints from Miocene-Pliocene sediments,
Waiho-1 borehole, South Westland. N Z J Geo Geophys.
Troy CS, MacHugh DE, Bailey JF, Magee DA, Loftus RT,
Cunningham P, Chamberlain AT, Sykes BC, Bradley DG.
2001. Genetic evidence for Near-Eastern origins of European
cattle. Nature. 410:1088–1091.
Turnbull IM. 1980. Sheet E42AC—Walter Peak (West). Geo-
logical map of New Zealand 1:50 000. Wellington (New
Zealand): Department of Scientific and Industrial Research.
Turnbull IM. 2000. Geology of the Wakatipu Area. Lower Hutt
(New Zealand): New Zealand Institute of Geological and
Nuclear Sciences Limited.
Turner TF, Osborne MJ, Moyer GR, Benavides MA, Alo D.
2006. Life history and environmental variation interact to
determine effective population to census size ratio. Proc Biol
Wakeley J, Hey J. 1997. Estimating ancestral population
parameters. Genetics. 145:847–855.
Waters JM, McDowall RA. 2005. Phylogenetics of the Austral-
asian mudfishes: evolution of an eel-like body plan. Mol
Phylogenet Evol. 37:417–425.
Waters JM, Rowe DL, Apte S, King TM, Wallis GP,
Anderson L, Norris RJ, Craw D, Burridge CP. 2007.
Geological dates and molecular rates: rapid divergence of
rivers and their biotas. Syst Biol. 56:271–282.
Waters JM, Wallis GP. 2001. Cladogenesis and loss of the
marine life-history phase in freshwater galaxiid fishes
(Osmeriformes: Galaxiidae). Evolution. 55:587–597.
Weiss S, Antunes A, Schlotterer C, Alexandrino P. 2000.
Mitochondrial haplotype diversity among Portuguese brown
trout Salmo trutta L. populations: relevance to the post-
Pleistocene recolonization of northern Europe. Mol Ecol.
Wellman HW. 1979. An uplift map for the South Island of New
Zealand, and a model for the uplift of the Southern Alps. In:
Walcott RI, Cresswell MM, editors. The origin of the
Southern Alps. Wellington (NZ): Royal Society of New
Zealand. p. 13–20.
Wesche PL, Gaffney DJ, Keightley PD. 2004. DNA sequence
error rates in Genbank records estimated using the mouse
genome as a reference. DNA Seq. 15:362–364.
Woodhams M. 2006. Can deleterious mutations explain the time
dependency of molecular rate estimates? Mol Biol Evol.
Yang ZH. 1997. PAML: a program package for phylogenetic
analysis by maximum likelihood. Comput Appl Biosci.
Yang ZH, Nielsen R, Goldman N, Pedersen AMK. 2000. Codon-
substitution models for heterogeneous selection pressure at
amino acid sites. Genetics. 155:431–449.
Youngson JH, Wopereis P, Kerr LC, Craw D. 2002. Au-Ag-Hg
and Au-Ag alloys in Nokomai and Nevis valley placers,
northern Southland and Central Otago, New Zealand, and
their implications for placer-source relationships. N Z J Geo
Zhang SM, Wang DQ, Zhang YP. 2003. Mitochondrial DNA
variation, effective female population size and population
history of the endangered Chinese sturgeon, Acipenser
sinensis. Conserv Genet. 4:673–683.
Zhivotovsky LA, Underhill PA, Feldman MW. 2006. Difference
between evolutionarily effective and germ line mutation rate
due to stochastically varying haplogroup size. Mol Biol Evol.
Jody Hey, Associate Editor
Accepted November 20, 2008
Time Dependency of Fish Molecular Rates633