Response to these comments:

Reviewer #4 (Remarks for the Author):

1. The first sentence of the "main conclusions" part of the abstract reads: "Our findings suggest that Stygiocaris and Typhlatya may descend from a common ancestor that lived in the coastal marine habitat of the ancient Tethys Sea, and were subsequently separated by plate tectonic movements." This statement was not really backed up in the paper, apart from the sister relationship between these genera. For example, based upon the other information presented (geological events in Australia and molecular evolutionary rate estimates), are the divergence values compatible with this scenario? One potential way to address this issue would be to use a program such as BEAST to estimate divergence times and confidence intervals for specific deeper nodes. Also, does the branching order among species from all continents support this hypothesis?

RESPONSE:
We have now done this in Beast (see reviewer 2 point 2 above). We do not contend that Tethyan vicariance is responsible for the presence of all atyid species on all continents, but merely that the Stygiocaris/Typhlatya relationship is congruent with a Tethyan vicariant explanation. Many atyid distributions are more likely the result of dispersal (as we say in the text; lines 166-169, 443-446, 447-450).


2. Results. line 312. "... combined dataset unequivocally recovers S. lancifera and S. stylifera as sister taxa..." This statement is based upon Fig. 7, which includes no outgroup. An outgroup should be added to these analyses.

RESPONSE:
We did use an outgroup (Antecaridina sp. East Timor) for these analyses. We have now made this more clear in the text in Materials & Methods and Figure label.


3. Results. lines 334-335, talking about molecular clocks. "At the lower end, we can use the allozyme estimate of 3.8 MYA for this speciation." Estimating divergence times from a (typically) small number of allozyme loci is highly problematic. I suggest that this value is likely not to be a reasonable calibration point. However, if the authors wish to use it, it should be explained (both the methodology and sample size used in the source paper should be given) and justified.

RESPONSE:
We agree and have removed the allozyme based calibration date, and have retained the two geologically based dates (7 & 10 MYA).


4. Results. lines 336-345 and Table 4. Generally, discussions of "molecular evolutionary rates" are referring to a general evolutionary rate (at least within a given taxonomic group). The presented values instead refer to just one specific node. The standard errors are therefore very low, because they indicate the variability in divergences between pairs of sequences belonging to two sister clades. These values are fine for indicating error in the genetic divergence estimate at that node, but represent pseudo-replication in the context of estimating general evolutionary rates. This section should therefore be re-worded to more clearly reflect the fact that divergences are considered at just one node.

RESPONSE:
This proviso has been added. We have also now added the calculated mean rate for 16S derived from the new Beast analysis. This is not a single node calculation but is a general rate based on the whole tree and includes a relatively wide 95% confidence level.


5. This paper is generally well written and employs an enjoyable style, but there are a few awkward sentences that should be reconsidered. For example:

a) intro. sentence 2: "The various evolutionary and geographic pathways taken by subterranean biota can illuminate the complex processes and histories that have resulted in this biodiversity..." This sentence is sort of like saying "studying the evolutionary history of these species will reveal the evolutionary history of these species."

RESPONSE:
Not really. The sentence says that studying the evolutionary “pathways” (i.e. patterns) will reveal the evolutionary “processes”, which is a valid statement. We have changed the word “pathways” to “”patterns” to make this more clear.


b) intro. sentence 3: "The isolation and strong selective pressures inherent in the adoption of an underground life can lead in polar opposite directions..." (Can lead what in opposite directions? The meaning comes through later in the sentence, but this part should nevertheless be re-worded.)

RESPONSE:
Again we don’t really agree. The reviewer says that the meaning comes through later in the sentence. Yes is does; in the very next clause (“namely both genetic divergence and morphological convergence”). There is no stylistic reason to put that clause first, and then go on to say that these are opposites. The sentence works fine as is.


c) Intro. paragraph3. sentence 2. "The evolutionary relationships of this biota has not..." Change "has" to "have" because "relationships" is plural.

RESPONSE:
Changed.


d) Discussion. lines 430-431. awkwardly phrased sentence: "Putative 'molecular' species can set the boundaries and foster further, targeted morphological study, which can go on to discover new morphological characters."

RESPONSE:
Sentence reworded.


e) Intro. line 149. "...its hypothesized closest evolutionary relatives are all far distant..."

RESPONSE:
Changed to “its hypothesized closest evolutionary relatives are all geographically remote…”


6. Methods. lines 240-241. Why was a model selected for all genes combined? These genes evolve at very different evolutionary rates and are even likely to have differences in nucleotide composition. Running "unlinked" models seems most appropriate, and a combined model should be justified.

RESPONSE:
We selected models both for each gene and for all combined. Each gene’s individual model was used for the single gene analyses. For the combined Bayesian analyses, “unlinked” models were indeed used as the reviewer suggests above, but for the ML analyses a single combined model was used because PHYML doesn’t currently allow different unlinked models.


7. Methods. line 244. A 50% burnin seems high when 2 million generations were run (how was this selected?), although the remaining 1 million generations is still a reasonable sample size.

RESPONSE:
This was just to make sure that only optimal trees were sampled, and, as the reviewer says, the 1 million generations is still a reasonable sample size.


8. The molecular clock analysis and calibration points are introduced in the Results, but should be set up in the Methods section.

RESPONSE:
Done (lines 271-291).


9. Discussion. paragraphs 3 and 4. As mentioned above, this part could include a more in-depth consideration of further lines of evidence for or against the Tethyan vicariance hypothesis.

RESPONSE:
We have done this (see reviewer 2 point 2 above).


10. Fig. 4 would be more informative if readily understandable locality codes (e.g. continent or country) were placed after each species name.

RESPONSE:
Ideally it would have been nice to do it this way, and we did originally make it like that. But, practically, it made the tree very complex and hard to read. This is why we opted to put all Aussie taxa in bold, and subterranean taxa with a symbol next to them, as these two categories are the most important for this paper. All of the localities can be found in Table 2.
Further, we sampled different taxa not to represent the entirety of the geographic ranges of their particular clades, but rather to sample the phylogenetic diversity. In other words, we used mostly Australian species to represent Caridina-like and Atya-like species (including all of the major Caridina clades), but this doesn’t mean that they are largely Australian, as they are in fact found all over the Indo-Pacific.


11. Fig. 6. The light grey box around Stygiocaris stylifera is currently placed around a region that also includes the node depicting the last common ancestor of all of the Stygiocaris species. The species boxes would be clearer just including each species-level clade.

RESPONSE:
Done for Figs 6 and 7.


12. Again, the trees shown in Fig. 7 do not include outgroups. Fig. 6 includes a larger sample size of 16S sequences than Fig. 7 and did not recover the S. lancifera/ S. stylifera sister relationship, conflicting with the "unequivocal" relationship claimed in the Results section.

RESPONSE:
Fig. 7 does have an outgroup (see point 2 above).


13. That is interesting that Halocaridina from Hawaii exhibits such a high rate of molecular evolution (an article in press was cited). Is this group part of a freshwater clade that secondarily invaded saline waters? In a study involving branchiopod crustaceans, secondary invasion of inland saline habitats was associated with elevated rates of molecular evolution. This could be an interesting case for comparison with atyid patterns.

RESPONSE:
We don’t want to get into this is this paper as it too would be an entire paper on its own. We have discussed this to some extent in our Journal of Biogeography paper and our in press paper in the Journal of the North American Benthological Society, and it is discussed a bit in papers by Scott Santos and others. One of the problems is with the idea of a “freshwater” taxon, as although atyids have no known close of distant marine relatives, many taxa, including Halocaridina have some salinity tolerance, often as larvae, which probably explains their presence in distant places like Hawaii.

So, I don’t think that Halocaridina’s elevated rates of molecular evolution is associated with secondary invasion of saline habitats, but is more likely linked to either 1) the geological dates used to relate to the genetic divergences in Halocaridina could inappropriate (I doubt this, and suspect that the geological dates do indeed relate to the geographically structured Halocaridina clades), 2) the difference between recent intraspecific rates and more ancient interspecific rates (this is more likely). This has received a great deal of interest in the various recent papers by Ho and those by Emerson, and would also require a whole other paper to discuss fully. That is why we made it clear in the text that our rates were “interspecific” and Halocaridina’s were “intraspecific” (lines 395-399). We have added a sentence about this and references to one each of Ho’s & Emerson’s papers to provide background to those interested in this topic.