Axios Review Decision Letter
Manuscript #: ax-14-0005
Title: Age estimates for an adaptive lake fish radiation, its mitochondrial introgression,
and an unexpected sister group: Sailfin silversides of the Malili Lakes system in Sulawesi
Authors: Björn Stelbrink, Isabella Stöger, Renny K. Hadiaty, Ulrich K. Schliewen, and
Fabian Herder
Comments to Authors:
This manuscript has now been seen by three expert referees, who all agree that this is an
interesting piece of research that is based on an impressive set of samples and that should
ultimately be published in one of the target journals (see below). This is also my opinion.
One drawback noted by all referees is that the study is based on mtDNA only, and it
might be a good idea to discuss this issue in a bit more detail. The referees provide
additional and very constructive suggestions, which should help the authors to further
improve their manuscript. Overall, I find this a well-made study and I am looking
forward to see it published. Judging from the referees’ reports and my own evaluation, it
seems that BMC Evolutionary Biology is the right address for this paper.
Walter Salzburger
Editor, Axios Review
Decisions for each target journal:
1. BMC Evolutionary Biology
It appears that the authors’ first choice. BMC Evol Biol, is also the first choice of the
referees and of myself. I think that this manuscript would nicely fit into this journal,
given that some of the comments raised by the referees have been addressed, which
should, however, not be a problem.
2. Molecular Phylogenetics and Evolution
Given the strong focus of the present work on phylogenetic analyses, MPE would also be
a suitable address for this manuscript. However, as also noted by some of the referees, the
lack of nuclear data might be problematic.
3. Journal of Evolutionary Biology
Although one of the referees suggests this journal as the probably best fit (after
shortening), I rather agree with referee #1 that it is the least fitting journal for this
manuscript.
4. Journal of Biogeography
In its present form, the focus of the manuscript is more on phylogenetics. While the
Journal of Biogeography would probably work for this manuscript, it would require some
rewriting (and maps, as noted by one of the referees).
Reviewer Comments
Referee 1:
General Comments:
The manuscript of Stelbrink et al. presents a time-calibrated phylogeny that focusses on
Telmatherinidae, their radiation in Sulawesi’s Malili Lakes system, and their outgroups.
The paper aims to determine the position of basal telmatherinids, and the timing of their
divergences, and investigates biogeographic scenarios that could have led to the observed
relationships and distributions.
The phylogeny is based on extensive taxon sampling, and a rather moderate amount of
sequence data, that is nevertheless sufficient to answer the questions at hand. Time
calibration is performed according to four different hypotheses, based on biogeography
and age estimates derived from other studies. In clades without fossil evidence, such as
Telmatherinidae and their close relatives, this is the best possible way to obtain timelines
of diversification. The Discussion is relatively long but valuable as it explores the
implications of each of the four different hypotheses used for time-calibration.
Overall, analyses are solid, results are interpreted appropriately, and the manuscript is
generally well-written, so that this study could in principle be published in any of the four
suggested journals. I do have a long list of minor comments and suggestions (below) that
can easily be addressed by the authors, and found only one potential problem that could
require re-running one of the BEAST analyses, but would lead to very minor differences
in the results, and would most likely not require a new interpretation.
The time calibration is based on four different age constraints: [A] the age of the split
between Isonidae and Melanotaeniidae observed in the phylogeny of Setiamarga et al.
(2009), which was time-calibrated with multiple fossils of clades that are not included in
the submitted manuscript, [B] the age of the split between the northern and southern
subgroups of Melanotaeniidae, again according to another time-calibrated phylogeny
(Unmack et al. 2013), [C] a substitution rate estimate obtained by Unmack et al. (2013),
and [D] a geological calibration on the divergence of northern and southern
melanotaeniid subgroups. The two additional analyses [E] and [F] are based on
combinations of [A] and [B] or [A] and [D], respectively. In the absence of direct fossil
evidence, these schemes are well justified, however it seems to me that the constraint of
Setiamarga et al. (2009) has not been implemented correctly in [A]. I could not find any
mention of the isonid-melanotaeniid divergence date estimate in Setimarga et al. (2009),
except for in their Figure 1, in which it appears to be 94-120 Myr old. In the submitted
manuscript, Stelbrink et al. use normally distributed priors to model the estimates of
Setiamarga et al. (2009), Unmack et al. (2013), and geological constraints, which is
reasonable when no other information on the distribution shape is available. However, the
age estimate of Setiamarga et al. (2009), is reported by Stelbrink et al. as 93-113 Mya and
modeled with a mean of 106 Mya and (apparently) a standard deviation of 5 Mya. Thus
the uncertainty present in the estimate of Setiamarga et al. (2009) has been reduced
without explanation, which is the reason why I suggest that this analysis should be
repeated. Given that 40 million MCMC generations were apparently sufficient for
convergence of BEAST analyses, this should not take longer than a few days.
>>The range inferred by Setiamarga’s study is 93-119. We therefore used a mean of
106.0 My for the root height representing the split between Iso and Melanotaenia.
Actually, the SD should be 13.0 to cover the full range of Setiamarga’s credibility
interval. However, values for SD are not allowed to be higher than the root square of the
mean, which is c. 10.30. We indeed used a reduced standard deviation (SD) of 5.0 in the
previous analysis to reduce increasing uncertainty for time estimates for the posterior
distribution. Our previous 95% HPD for the root height was 95.22-114.22; though the
upper HPD is c. 5 My younger than the upper HPD from Setiamarga, the lower HPD is
only 2 My younger.
However, we re-run this analysis (2 runs in parallel; same methodology) using a
modified SD of 10.0. The new 95% HPD now ranges from 82.70-122.15 My and thus
showing a much younger lower HPD (compared to those from Setiamarga; >10 My), but
the upper HPD almost corresponds with Setiamarga’s posterior distribution. We use
these updated ages in the present revision (text, Table 2, Fig. 3), though credibility
intervals generally increased among nodes.
I would further suggest that analyses [E] and [F] can be left out of the manuscript. They
are mentioned only very briefly in the text, and age estimates resulting from these
analyses are predictably intermediate between those of the other analyses using only one
of the two constraints applied in [E] and [F]. Thus, hardly anything would be lost if these
analyses were removed, but the manuscript would be more straightforward to read.
>>We agree. Analyses [E] and [F] have been removed.
All other comments can probably be addressed by minor changes to the text:
You’re referring multiple times to the ‘most plausible’ analysis/estimates, sometimes
(lines 37, 323, 427) without explaining why this analysis should be more plausible than
others. I see the point that the fit between the results of analysis [D], especially of its
younger nodes, with the geological history of the region makes these results the most
appealing ones, but at the same time the age estimates of the older nodes appear
surprisingly young (~30 Myr for the split between Isonidae and all other included
atheriniforms). While the results of Setiamarga et al. (2009) may be overestimates, they
are similar for isonid-melanotaeniid divergence and for the not too distantly related
exocoetid-hemiramphid divergence. As Exocoetidae are present in the deposits of Monte
Bolca, this suggests that the divergence of Isonidae and other atheriniforms could be on
the same order, around 50 Myr. The fact that the age estimates for younger nodes in
analysis [D] appear reasonable, while those of older nodes in the same analysis appear
underestimated could possibly be due to the combination of a relatively young age
constraint (~12 Myr) with a mitochondrial sequence data set, if some amount of
saturation was present in the alignment, which would make older branches appear shorter
than they actually are. I suggest that analysis [D] should not generally be considered the
most plausible analysis without pointing out the rather young root in this analysis, and
that the possibility of saturation, and thus underestimation of the age of basal nodes
should be discussed.
>>We now discuss the problems of exclusively using mitochondrial markers, saturation
effects and age underestimations in the Discussion section of analysis [D].
On line 68, it would help if the three main lakes would be introduced by name.
>>Done.
On line 95, comma after both is not needed.
>>Removed.
Are sequences not going to be submitted to Genbank? Journal requirements should be
checked here.
>>Sequences have been submitted to GenBank.
Line 127 should probably start with ‘The GTR+G substitution model…’
>>Done.
On line 132, you write that DAMBE was used to reduce sequence alignments to unique
haplotypes, but it is not clear what this is used for.
>>The use of non-redundant haplotypes is a common procedure prior to performing
molecular clock analyses as there is no need for time estimates within haplotypes of the
same (sub)population (see e.g., Telmatherina bonti TOW T85,T86; Telmatherina opudi
MAT 33153/33154/33157 in Figure 2). Unique haplotypes shared by different species are
far more interesting (see also Reviewer 3 comment below) as shown for Glossolepis
dorityi and G. pseudoinciscus. We now briefly mention this observation, which is
probably due to gene between the two species both species occurring in the same river
system. However, it does not affect the general results of the present study.
Also, it should be acknowledged that the tree prior models (Yule, Birth-Death) in BEAST
are speciation priors that assume each tip to represent a species, and that the mix of
population level and species level samples in your phylogeny violates these models. This
could lead to bias in age estimates, but it’s not clear to what extend and in which
direction. Actually, your data set would be well suited for analyses with *BEAST even
though that type of analysis would probably not be needed to answer your questions.
>>Thanks for pointing that out. We acknowledge that different tree priors might bias
time estimates, however, we think that using the Yule prior for the present dataset is
suitable as likewise stated on the BEAST FAQ site: “This tree prior [Yule] is most
suitable for trees describing the relationships between individuals from different
species”. We agree that the present dataset might be already suited for a species tree
reconstruction; however, we think that a more comprehensive dataset also including
nuclear markers would be more appropriate.
Some more information should be given for phylogenetic analyses. What’s the
percentage of missing data in the alignment? Did you use a single concatenated partition
for RAxML, MrBayes, and BEAST? Did you use a strict molecular clock or the
uncorrelated relaxed lognormal clock? Did you constrain the ingroup as monophyletic in
BEAST? How large were ESS values? Some more wording would help on line 135.
>>We are sorry for not mentioning several priors for different analyses. We have
extended this section by more details such as analysis settings.
When you refer to the split between Pseudomugilidae and Telmatherinidae, and to
diversification events within these families on lines 180-183, you do not account for the
observed paraphyly of Pseudomugilidae. Technically, the first diversification event
within Pseudomugilidae is the same as the split between the two families, node 11. Also,
as only relatively few representatives of Pseudomugilidae are sampled, the first
diversification of that family could have been earlier.
>>We agree and rephrased this section: “Phylogenetic reconstructions (ML and BI)
show a well to highly supported clade comprising members of the genus Pseudomugil (P.
signifer, P. reticulatus, P. furcatus, and P. pellucidus; note that the Pseudomugilidae are
paraphyletic; see also below), and the Telmatherindae.
[...]
Estimated divergence times range from 18.1-59.7 My for the split of Telmatherinidae and
the most recent Pseudomugilidae clade (node 11), and are quite similar for the first
diversification events within each family (12.3-39.4 My for node 12 for the most recent
Pseudomugil clade, and 12.9-44.3 My for node 13, respectively).”.
Lines 186-187 are a bit confusing, as it’s unclear which MRCA is referred to, the one of
Paratherina, or that of Paratherina and Telmatherina. Also, the abbreviation TMRCA
should first be introduced. The same applies to line 188.
>>The abbreviations ‘TMRCA’ and ‘speciation’ are explained. We also clarify that the
given TMRCA refers to the most recent common ancestor of Paratherina+Telmatherina
and Tominanga+Telmatherina, respectively.
On multiple occasions, e.g. lines 190, 192, 193, 199, 209, and in the legend of Figure 3,
you refer to nodes as if these would represent clades, but they rather represent the MRCA
of these clades. Please rephrase.
>>Thanks. Clarified in text and figure caption.
The double parentheses on line 200 should be rephrased.
>>Done.
The ‘preliminary age estimations’ on line 220 probably refer to the origin of the lake, but
this is not clear from the sentence.
>>Clarified: “preliminary lake age estimations”.
Isn’t there any more citable information for the age of the Malili lake system then the
personal communication with Robert Hall (line 225)? In any case, it would be good to
know why this is considered ‘plausible’ from a geological point of view.
>> Unfortunately, there is no published age estimates for these lakes yet. The proposed
lake ages are estimations based on seismological data and geological settings in that
area. However, precise data will be available in the course of the Towuti Drilling Project
(http://www.icdp-online.org/projects/world/asia/lake-towuti).
On line 228, ‘more closely’ sounds better than ‘closer’.
>>Done
Not sure whether one can call two monophyletic clades ‘respectively monophyletic’ as on
line 242.
>>We removed “respectively”.
Priapiumfishes should be lower case on line 248.
>>Done.
Into what should Phallostethidae be included (line 249)?
>>Clarified: “within the atherinomorphs as distant…”
What do you mean with ‘all remaining Melanotaenidae’ on line 267.
>>We removed “remaining.”
‘Details on terrane movements and distribution of land and sea’ sounds a bit too general
for the information given in references 46 and 47 (line 273-274). Something like ‘in the
Indo-Australian archipelago/Wallacea’ should be added.
>>Done.
A closing bracket is missing on line 282.
>>Done.
‘Terrane raft’, mentioned on line 286, has four hits on Google. Maybe the term should be
explained in more detail?
>>We re-arranged and re-phrased this section. We also discuss the p
A comma is not needed on line 354.
>>Done.
On line 381, you write ‘Melanotaenia lacustris is comparatively terminal…’, however all
extant taxa are equally terminal. You could say that M. lacustris is deeply nested within
Melanotaenidae.
>>Done.
Regarding lines 382-384, it’s not clear to me why sparse taxon sampling and resulting
long branches in itself should lead to age overestimation. I agree that the estimates of
Setiamarga et al. (2009) are likely too old, but this probably results more from the
combination of mitochondrial sequence data (and therefore saturation over long
branches) and relatively old calibration points.
>>We think that a sparse taxon sampling might affect divergence time estimates, too, but
we also agree that saturation due to the use of mtDNA and old calibration points are
more important issues in Setiamarga’s study. We therefore re-phrased this section.
Lines 421-422 could be rephrased to make clear that the prior age of [D] is younger than
the posterior of [A], as no prior is specified on node 9 in [A]. Also, it would be good to
add that the two values given (36.2 and 10.9 My) represent the respective distribution
means.
>>Clarified.
Line 424 is also a bit unclear the way it is written. Of course there’s a discrepancy
between different nodes, but the older ages in [E] and [F] compared to [D] are due to the
relatively old constraint for node 1 in [A], [E], and [F], which is missing in [C].
>>We agree and removed analyses [E] and [F].
The sentence on lines 425-426 could also be made clearer. When stating ‘results in
generally older ages’ in the previous sentence, you referred to [E] and [F] compared to
[D]. In this sentence, you seem to refer to [E] and [F] compared to [A]. Some more
wording would help. Also, two commas are not needed here.
>>Analyses [E] and [F] removed.
Reference 39 could be cited again at the end of line 444.
>>Done.
On line 458, two commas are not needed.
>>Done.
Line 486 is grammatically not correct, it should rather be ‘as the age of the introgressed
haplotypes clearly postdates the initial…’
>>Done.
The caption of Figure 1 should include the explanation that grey text background
indicates representatives of Telmatherinidae.
>>Done.
The caption of Figure 2 should state from which of the four BEAST analyses the MCC
tree is derived. Also, the abbreviation ‘MCC’ should be explained.
>>Done.
Table S1 still includes ‘Awaiting information from collaborator’. I hope the collaborator
has provided that information by now. Also, accession numbers will need to be updated.
>>Now updated, Genbank numbers included.
Book titles in the reference should better not be abbreviated. Thus ‘Evolution since
Darwin: The First 150 Years’ instead of ‘Evol since Darwin first 150 years’ and
‘Evolution in Action - Adaptive Radiations and the Origins of Biodiversity’ instead of
‘Evol Action’.
Reference 7 (Friedman et al. 2013) is no longer ‘in press’.
Reference 27 has a typo: ‘phaloostethid’
For the Tracer reference (38), a web address should be added.
Reference 42 also has a typo: ‘fisheswithin’
>>Done.
Other than that, I’d like to congratulate the authors to an interesting study that can
hopefully soon be published.
Michael Matschiner
Comments on the suitability of this paper for:
1. BMC Evolutionary Biology
BMC Evol Biol regularly publishes studies similar to the submitted manuscript in terms
of analysis and data set, and would therefore be well suited.
2. Molecular Phylogenetics and Evolution
As the focus of the submitted manuscript is very phylogenetic, Mol Phyl Evo would also
be very well suited.
3. Journal of Evolutionary Biology
Of the four proposed journals, J Evol Biol might be the least fitting, as its focus is less
phylogenetic than the other three journals.
4. Journal of Biogeography
The submitted manuscript combines biogeography and phylogenetics and could therefore
also be published in J Biogeo.
Referee 2:
General comments:
a few sentences on the aims and results of the paper
This paper uses mtDNA sequence data to resolve relationships among the Malili Lakes
Telmatherinids, other extant Sulawesi and extra-Sulawesi Telmatherinids, and
Rainbowfishes; and provide both age estimates for the Malili Lakes telmatherinid
radiation and for introgression events of sharpfin/riverine Telmatherinid groups. I
enjoyed reading this paper, as I think it asks a number of important questions that have
lingered since the appearance of first molecular-based Malili Lakes fish papers of the last
10 years. Having said that, I was disappointed by the mtDNA only approach the authors
took, as hybridization has played a rather large role in shaping the Malili Lakes fish
system—and since this review is to strengthen the paper and see it published, I will try to
make my comments more constructive than simply critical.
The wide range divergence estimates provided here in conjunction with the messy,
composite geology of Sulawesi and lack of resolution/discrepancies in ages of the lakes,
lake bedrock and lake systems themselves leaves the authors with a broad range of
‘acceptable’ or possible (perhaps even equally plausible) colonization/diversification
scenarios. That being said, given the broad ‘window’ of split times (some ranges here
~50 MY) and the sole reliance on mtDNA and tree-based molecular clocks is limited.
an overview of the paper’s strengths and weaknesses
the authors provide an impressive mtDNA dataset in terms of both sequence length
(2000 bp), and taxonomic and geographic coverage. The multiple comparative methods
(both geologic calibration and indirect calibrations) for estimating node ages, despite
their variability is a particular strong point in their analysis. However, and this should
come as no surprise to the authors, my primary criticism involves the sole reliance on
maternally-inherited markers to address two of the major points of the paper – 1)
identification of the sister taxon to Telmatherina; 2) to date ages of introgression events
among sharpfin Telmatherinids. Even though this (the mtDNA) is explicitly expressed in
the title of the paper, I can’t help but wonder why nuclear markers weren’t also used
particularly given the complex evolutionary relationships known among the groups
within the system. Surely 1-2 nuclear genes could be employed at the very least to assist
in resolving age estimates. Unfortunately, this problem (the mtDNA-sided story) has
persisted across a number of taxa within this system, despite substantial evidence of
remarkable divergence, endemism, and above all – hybridization.
>>We completely agree with the referee, that phylogenetic reconstructions targeting the
evolutionary history of species within the Sulawesi species flocks clearly require the
incorporation of nuclear markers. However, the present manuscript addresses research
questions which do not necessarily require incorporation of nuclear markers. We do not
aim at testing for the presence of introgressive hybridization within the Malili Lakes
radiation – we employ molecular clock approaches for estimating its haplotype lineage
age, based on results of earlier phylogenetic studies which already had included nuclear
markers. However, we address this issue now explicitly in the Discussion, and think there
remains no confusion or uncertainty.
an in-depth discussion of whatever major problems you have identified
I don’t want to beat this point to death, but the lack of nuclear sequence data from at least
one gene is problematic for a few key points. First, identification of the sister taxon to
Telmatherina – at the very least, the authors could (perhaps they have!) sequence
Marosatherina, Telmatherina, and Kalyptatherina at EPIC markers, introns, or other
nuclear loci to resolve this for comparative purposes. The timing of such events across
2000 bp of mtDNA from multiple regions with varying substitution rates raising
questions for the application of a molecular clock – and given the ranges indicated here
this is evident. Also, and this may be of particular importance given the authors have
indicated that full sequences (12S-16S) were not amplified for all Kalyptatherina
specimens (how many? Only 4 listed in Table S1, is it 3 of 4?) – this is big deal given
that one of primary findings of this paper is the unexpected sister relationship of
Kalypatherina and Malili telmatherids.
>>Unfortunately, the 12S-16S fragment could not be amplified for all Kalyptatherina
helodes specimens as mentioned in the Materials and Methods section. We now run an
independent analysis [D] with ND2 only. The general topology is the same:
Kalyptatherina is sister to the Malili clades, while Marosatherina is sister to this group
(see new MCC tree on next page). However, some intra-Malili relationships differ, which
is probably related to the more conserved and shorter ND2 fragment: * T. bonti
Tominanga 33070 is not sister to roundfins and sharpfins A, but is rather sister to T.
bonti Nuha 33128 (sharpfins clade); ** T. bonti Nuha 33129 is not sister to introgressed
T. celebensis, but sister to T. bonti TOW T85/T86. Divergence time estimates differ, but
not substantially except for basal nodes (see Table below with age estimates for selected
nodes of the original analysis [D] and ‘[D] ND2 only’).
Analysis
Node
1
9
11
13
14
15
17
18
19
22
26
[D]
[D] ND2 only
31.31 (16.40, 48.90)
10.87 (6.70, 14.98)
18.11 (9.05, 27.98)
12.93 (6.18, 20.36)
8.45 (3.79, 13.45)
5.18 (2.15, 8.53)
1.86 (0.63, 3.27)
1.01 (0.31, 1.92)
0.93 (0.27, 1.80)
0.37 (0.07, 0.83)
0.70 (0.18, 1.37)
26.06 (12.72, 41.00)
10.85 (6.75, 15.08)
15.16 (6.85, 24.79)
11.44 (5.03, 18.87)
8.03 (3.37, 13.43)
5.53 (2.23, 9.47)
2.98 (0.80, 5.67)*
1.67 (0.30, 3.51)
1.20 (0.15, 2.88)
0.73 (0.05, 1.94)
1.34 (0.13, 3.07)**
BEAST MCC tree, analysis [D], ND2 only
The other major non-nuclear data concern I have is regarding the 3rd major aim of the
paper - dating the age of introgression. The data presented in the paper provides examples
of taxonomic-mtDNA discordance and refers to the senior authors’ previously published
work as evidence for introgression (which their clearly is!—it seems to me that the most
appropriate avenue for corroborating the dating of introgressive hybrid events is with the
inclusion of a suite of nuclear markers. We know from the literature that AFLPs were
used to uncover the hybridization among telmatherinids in earlier papers – has there been
no follow up with nuclear sequencing at a few loci?
>>See above.
Ok, so let’s suppose the authors do not have nuclear data or it’s not an option to acquire
it. Is there any other analysis that I think may seriously strengthen their paper particularly
with respect to their 3rd aim which is to date the Sharpfin hybridization events – the
answer is yes!
An IM analysis would be an appropriate and interesting technique to explore this data,
primarily given the broad-range of age estimates from the molecular clock estimates. The
beauty of using an IM analysis here is that divergence times for three introgressed groups
can each be estimated independently and together in IMa2. The BEAST/Phylogenybased age estimates would provide good comparative estimates to test against using
IMa2, with sequence data this shouldn’t take too long with sequence data (however, this
is relative; IMa2 with microsat data can take months!!!) – but the authors could probably
get good coalescent time estimates with IMa and relate them to the node estimates from
their molecular clock analyses.
It seems to me that the authors’ mtDNA would be very well suited for IM analysis. There
is introgression – so there is evidence of gene flow. Secondly, the authors wish to
characterize primary and secondary hybridization events – their tree shows these
branching patterns very nicely. This can certainly be used to their advantage in IMa2, in
particular given the variation among haplotypes and well-resolved clades (I use the term
loosely here), there is likely a lot of information to work from.
Furthermore, the use of ND5 and the 16S-12S regions can be used both jointly and
separately, with various mutation rates applied to To scalars to get a better sense of the
information contained in the data (lines 403-405 – the authors highlight varying
substitution rates). Furthermore, the Ne estimates provided by IM may also yield some
insights drift effects and possibly shed insights on demographic expansion/contraction in
relation to other hypotheses (e.g., Marosatherina marine dispersal) the authors have
speculated on.
>>We agree that IMa2 is potentially interesting for a more comprehensive sharpfin
dataset. However, the IMa2 analysis would require – in our view – a substantially better
prior information on certain parameters for the included species. The Lake Matano
species assemblage is highly diverse, and it is characterized by ancient and still ongoing
gene-flow between (comparatively ill-defined) species, i.e. by changes of not only
demographic population parameters, but also by mosaic-like and changing composition
of population entities, that led to the present-day species assemblage. It is highly likely
that this situation has led in the past to quite drastic changes in demographic population
parameters, which cannot be estimated using IMa2. Therefore, and also because of the
limitation to comparatively sparse mtDNA data, we refrain from using IMa2So far, the
most interesting result inferred from our analyses is that the introgressions observed in
several clades are substantially more recent than the original/ancient Lake Matano
radiation.
The Kalypatherina-Telmatherina split at 8.5-28.8 MYA and the terrane-rafting
hypothesis. The range in the split date is factor of nearly 4! Is there a way the authors can
reconcile this other than simply the overestimate biases they allude to in their discussion.
The geological history of Sulawesi is indeed complex—however, is the terrane-rafting
hypothesis consistent across this rather wide-range of estimates—a bit less hand-waving
is needed here for clarification.
>>We agree and re-worded this paragraph: “It appears most plausible that the
population ancestral to the Malili sailfin silversides originates from the Sahul Shelf area,
and was dispersed on such a ‘terrane raft’ when the Sula Spur was extended and moved
westwards until this fragment (‘raft’) collided with West Sulawesi. However, given the
temporal uncertainties in both the separation of the Sula Spur from the Sahul Shelf (c. 15
Mya; see also [48]) and divergence time estimates among the four analyses, it remains
difficult to test this hypothesis. Mean ages and credibility intervals suggest that this
scenario might be plausible for analysis [A] and [B] (15.0-43.5 My and 9.6-30.4 My; see
Table 3), while the credibility intervals obtained from analysis [C] and [D] (5.5-13.1 and
3.8-13.5) would slightly post-date the estimated age of the Sula Spur separation.
However, the lack of fossil remains requires denser sampling and the incorporation of
multiple markers to explain the relationships of present-day geographically distant
groups, which probably have been in vicinity in the past.”
Furthermore, the marine dispersal hypothesis for Marosatherina colonization of the
Maros karst vs. the relict population retention is somewhat unresolved. This again may
represent an interesting use of IM or even a MIGRATE analysis, particularly if enough
wild-caught individuals have been sampled. Estimates of Ne may provide a relative
estimate of standing genetic variation – and comparison of these estimates under
scenarios of relic population or a founding colonization followed by expansion may be
interesting.
>>We agree, but there are not enough wild-caught individuals available.
The authors need a figure with map, perhaps more than one – if only to highlight the
current geographic relationships among taxa. Given the dynamics of this region, a map
will go far in assisting readers in following the story, it will also aid in convincing the
plausibility of the hypotheses explored here. The long reaching ‘arms’ of Sulawesi and
their proximity to other island, etc, would be very helpful especially alongside the
‘Biogeographic Implications’ section of the Discussion. The authors should also consider
figures that display the arrangement of Sulawesi at important time frames drawn from
their analysis.
>>The current distribution of Telmatherina, Marosatherina and Kalyptatherina is now
shown in Figure 2. We agree that the visualization of SE Asia’s geological reconstruction
is quite helpful in understanding the geological changes and therefore refer to important
publications. In addition, we now include Figure 4 showing palaeogeographic maps from
the mid Oligocene (30 Mya) until the early Pliocene (5 My) in 5 My-steps.
a list of any other minor issues
I think the paper could use some further insights (e.g., mechanisms, or scenarios) with
respect to marine dispersal pathways (Marosatherina and Kalypatherina origins). For
example, papers focusing on colonization patterns of other insular freshwater faunas
(e.g., Guppies and guapin killifishes in Trinidad) have hypothesized that seasonal river
flooding and the decreased coastal salinities facilitated colonization spatially variable and
repeated colonizations. Any similar patterns for Australia-to-Sulawesi routes?
>>We think that this would open a speculative discussion. We now include Fig. 4
showing palaeogeographic maps, which might be helpful for the reader to understand
geological settings in the past in that region.
Justification for the use of GTR+G model (jModelTest? MrModelTest?), perhaps the
authors should consider the use of other model selection criteria with respect to relaxed
molecular clock analysis – see Baele et al. 2012 Mol Biol Evol 30: 239-243.
>>Now clarified: “Substitution models were selected using jModeltest ([34]; AIC and
BIC both resulted in GTR+G for both partitions).”
Comments on the suitability of this paper for:
1. BMC Evolutionary Biology
It appears that the authors have formatted with this journal in mind. I think that
conceptually the paper is ideal for BMC Evol Biol as the journal’s scope is rather broad
in terms of interest. However, many recent and very n ice BMC Evol Biol papers have
used a combination of mtDNA and many nuclear loci to study age estimations,
hybridization events – Sequeira et al. (2011) BMC Evol Biol 11:264 comes to mind, I
understand the questions are quite different – however, it appears to be that a suite of
markers is more appropriate for this journal. Do the IMa and I think you’ve got a good
shot at BMC Evol Biol.
2. Molecular Phylogenetics and Evolution
The lack of nuclear markers will likely be a problem for publication in MP&E, reviewers
will definitely highlight the one-sidedness of the dataset.
3. Journal of Evolutionary Biology
J Evol Biol publishes a variety of evolutionary papers, but my opinion is this paper will
may be a bit too geographically and system-specific for general readership in its current
form. This could be circumvented with some clever writing, but I really don’t think that
the paper has very many generalizations to draw out – in fact, its strengths lie in the
endemicity and island biogeography, which may make the paper better suited for J
Biogeo.
4. Journal of Biogeography
If the authors do not consider the IMa analysis I suggested above, then it seems to me in
its present form this may be the best place for this paper given the specific hypotheses
addressed and analyses performed.
Referee 3:
General comments:
The manuscript describes a phylogenetic/phylogeographic investigation of the
evolutionary origins of the Telmatherina silverside radiation from Malili lakes system of
Indonesia. Mitochondrial DNA sequences were amplified from representatives of this
radiation and multiple potential outgroup taxa. These data were used for three specific
aims listed at the end of the Introduction, namely 1) the phylogenetic position of
Kalyptatherina, 2) the age of the Malili lakes radiation, and 3) the timing of mtDNA
introgression into “sharpfin” Telmatherina. The authors reconstructed the phylogeny
using standard approaches/software (BEAST, RAxML). Six different analyses were
conducted, each with a different divergence time calibration set. Briefly, the authors
identified Kalyptatherina as the immediate sister lineage to the Telmatherina radiation.
The timing of the divergence events were however rather inconsistent among calibration
sets, and the authors discuss the virtues and implications of each of these in detail.
Overall this is a well-designed piece of research that should be published. Three issues
need to be addressed.
1) The limitations and risks with basing the analyses exclusively on mtDNA need to be
much more clear, including in the abstract
>>See above.
2) The details of the calibration points and why they are potentially valid need to be
provided in the methods, or an appendix. As it stands the justification for their use, and
their limitations, is mostly in the discussion. The discussion should focus on discussing
the results. There is some inconsistency on how many calibration sets were used, was it
four, or six?
>>Details for calibration points are now available in the Methods section. We also
removed analyses [E] and [F].
3) The results and discussion sections are far too wordy and the text drifts from the aims
as stated at the end of the Introduction. Each of these sections can easily be reduced by
50% by eliminating unnecessary detail and focussing on the core three aims. In the
results section we do not need a full description of the characteristics and dates for all
nodes, only those of central interest. The results section should avoid comparisons with
published work - this is discussion material. The discussion should not repeat at length
information in the results , whether it be from text, tables or figures. It needs to be more
focussed.
>>We think that we only present relevant dates and that the current sections are
balanced in terms of word count.
Minor comments.
L25. “fishes have previously”
>>Done.
L26 and elsewhere. Avoid the use of “monophylum”, it reads awkwardly. It is better to
use “monophyletic group”.
>>Done.
L37. Delete “a kind of”
>>Done.
L40 and elsewhere. Use “more recent” rather then “younger”
>>Done.
L45. Use “system from rivers”
>>Done.
L51 “with the shaprfins”
>>Done.
L60 “for investigating speciation”
>>Done.
L64 “such as introgressive”
>>Done.
L70 “fishes that typically possess bright”
>>Done.
L84 “as an outgroup”
>>Done.
L87 delete “of or”
>>Done.
L89 “(i) assess relationship of Kalyptatherina”
>>Done.
L101-107. This section of text on missing taxa would be better in the Discussion.
>>Moved to Discussion section “The closest relative to the sailfin silversides radiation”
L132. Not a significant issue, but I don’t understand why you need to collapse sequences
down to unique haplotypes. It may save some computational time, but you lose
information on the relationships of species in the tree when multiple species share
haplotypes.
>>See above (Reviewer 1).
L137-145. Given the calibrations are so important, these really should be described in full
here, or in an appendix methods text.
>>Calibration points have been more clearly described in the M&M section.
L157 Avoid referring to methods in the results.
>>Removed.
L214-L225. This is all Introduction material, your discussion section should focus on
discussing results.
>>We agree. However, we think that this stylistic device (introductory section) is quite
helpful for the reader. We would like to leave it to the editor’s discretion.
L322, 465 etc. The bias in your writing towards one calibration set reads uncomfortably,
particularly given the strength of evidence.
>>See above (Reviewer 1).
L327-342. Unnecessary generic text.
>>Removed.
L382-384. I am struggling to follow the logic here.
>>See above (Reviewer 1).
L408. This is not the final analysis, it is calibration set 4 of 6 following your methods,
and table 1. However it appears in from this text, and Figure 3, it is calibration set 4 of 4.
The paper needs to be consistent. Either 4 calibration sets are used throughout, or 6 are
used throughout.
>>We agree and removed analyses [E]and [F] throughout.
L472-487. Repetition of earlier material.
>>As above. We think that these concluding remarks are quite helpful for the reader. We
would like to leave it to the editor’s discretion.
Figures, maps of the distributions of key taxa would be very helpful. It is difficult to
follow the geography in the Discussion.
>>Agree. We now include a map in Figure 2 showing the distribution of the key taxa.
Comments on the suitability of this paper for:
1. BMC Evolutionary Biology
Should be OK
2. Molecular Phylogenetics and Evolution
Should be OK
3. Journal of Evolutionary Biology
Perhaps the best choice, if the paper could be shortened considerably.
4. Journal of Biogeography
Could work, but there are no maps!