Page 13 - iCC044-201501_March

Based on the timing of Gondwanan frag-
mentation events, crown cichlids should
occur in the Early Cretaceous or Late
Jurassic [3,6,9,10]; however, the Bayesian
random local molecular-clock analyses
place the origin of the modern cichlid
radiation near the Cretaceous-
Palaeogene boundary (figure 1 and
table 1; electronic supplementary mate-
rial, figure S1), with a mean age esti-
mate of 64.9 Ma (95% CI: 57.3-76.0 Ma).
The estimated age of the most recent
common ancestor (MRCA) of cichlids
and their sister lineage, Pholidichthys, is
also younger (mean: 103.7 Ma; 95% CI:
92.0-118.4 Ma) than the initial rifting of
Gondwana at approximately 135 Ma [11].
The mean estimated age of the MRCA of
the African and Neotropical cichlids was
46.4 Ma (95% CI: 40.9-54.9 Ma), post-dat-
ing the final separation of Africa and South
America by more than 40 Myr. The cichlid
time tree confirms ages estimated in previ-
ous studies for the east African [26] (mean:
8.0 Ma; 95%CI: 6.9-9.5 Ma) and Cameroon
crater lake Barombi Mbo [27] radiations
(mean: 1.4 Ma; 95%CI: 0.8-2.3 Ma), verify-
ing relatively young ages for these remark-
able examples of adaptive radiation (figure
1 and table 1; electronic supplementary
material, figure S2). The age estimate in the cichlids that do not assume Gondwanan
10 nuclear gene inferred time tree closest
to the timing of Gondwanan fragmentation
is that of the inclusive (mean: 123.5 Ma;
95% CI: 111.4-136.2 Ma), but unnamed,
percomorph clade that contains more than
one-quarter of all living vertebrate species
(approx. 16 570 species), including cichlids
(see electronic supplementary material, fig-
ure S2).
cichlidpower.org.au
4. Discussion
(a) Congruence between palaeontological and
molecular time scales for cichlid evolution
The application of two contrasting pal-
aeontological approaches in calculating
temporal range extensions yields strikingly
congruent time scales for cichlid evolution.
Both methods provide point estimates for
the origin of the group that range between
57 and 60 Ma (Palaeocene), and strongly
reject the possibility that crown cichlids
are sufficiently old to have been affected
by the initial rifting of Gondwana. Instead,
upper limits for the origin of cichlids lie
consistently within the late Late Cretaceous.
This congruence is particularly compelling
because the methods that yielded these
comparable results share only one similar-
ity in their calculations: both are constrained are early Late Cretaceous [3], pre-dating
by the minimum age for cichlids as imposed
by the oldest fossil example of the group.
Our molecular time tree provides a
mean estimate for the timing of cich-
lid origin in the Palaeocene, but can-
not reject the possibility that the group
arose as early as the Late Cretaceous.
This result is consistent with other recent
molecular-clock estimates for the origin of
vicariance for the group a priori, and which
range in age from Late Cretaceous to
Eocene [28-33]. In terms of point estimates
and surrounding uncertainty, our revised
molecular time scale is entirely consistent
with the ages derived from analyses of the
fossil record alone (figure 1). It is important Jurassic origin of cichlids that is inde-
to note that our relaxed-molecular-clock
analysis shares no palaeontological data
in common with either our analysis of
the distribution of cichlid-bearing fossil
horizons or our database of outgroup-
based age constraints. We interpret the
convergence of these three semi-inde-
pendent approaches, which all deliver age
estimates for cichlids that are within error
of one another, as a consequence of genu-
ine evolutionary signal that strongly contra-
dicts the time scales for cichlids demanded
by hypotheses of Gondwanan vicariance.
(b) The timing of cichlid diversification: congruence
and incongruence
Our estimates for the time of cichlid
origin are congruent not only with one
another, but also with previous molecular
time scales for the evolution of this group
that do not assume a Gondwanan vicari-
ance scenario a priori [4,5,28-33]. The old-
est such estimates from previous work
our proposed time of origin by roughly
35-45 Myr. However, these more ancient
dates derived from analysis of mitochon-
drial sequences, which are characterized
by high rates of nucleotide substitution that
might bias clock analyses towards older
estimated times of divergence [25,34-36].
Generally, the only molecular-clock analy-
ses to deliver time scales consistent with
the predictions of the vicariance hypoth-
esis were themselves calibrated using a
combination of age constraints from the
fossil record and Gondwanan fragmenta-
tion events [3,5,6,37]. There is no pub-
lished relaxed-molecular-clock analysis
that results in an Early Cretaceous or
pendent of the ages implied by the timing
of the fragmentation of western Gondwana.
Our palaeontological time scales
for Cichlidae constrain only the ori-
11