Building a Robust, Densely-Sampled Spider Tree of Life for Ecosystem Research

Round 1

Reviewer 1 Report

The study by Macias-Hernández et al provides not only the most complete spider phylogeny to date, but also thoroughly evaluate a workflow to estimate and use phylogenetic diversity especially with regard to ecological studies. The manuscript is very well written and the figures are excellent. I enjoyed reading the paper, which makes certainly an important contribution to spider research and I have only few suggestions/comments.

- Since time-stamped trees play a major role in the suggested workflow, I was wondering whether the authors considered the recent paper of Magalhaes et al (2020, Biol Rev 95)? In this paper Magalhaes et al revised the fossil record especially with regard to calibrating trees. I suggest to explore this study as it present an updated (revised) dated spider tree of life showing different age estimations compared to Fernandez et al (2018). I have no idea whether it will change the results in the present study, but I think it should be tested in the present study! If the authors reached a different results it will be worth discussing the role of well justified calibrations points.

- I was missing a critical discussion on the barcode gene. Is it always "working" for spiders with regard on genus or family level? Furthermore, it is not suprising that the topologies of COI alone differ from phylogenies derived from more extensive data (see page 12, line 560). However, I missed a discussion about the reason for it and why the barcode gene is not sufficient for this kind of studies. It might be common knowledge, but given the goal of the paper it might be worth of a few explanatory sentences.

- Figure 1 shows the workflow used in this study (including the abbreviations of the matrices used, etc.). If possible I would suggest to provide another figure corresponding with the section "recommendations and suggested workflow" in the discussion. The strength of this paper is the critical evaluation of a variety of approaches and I think a clean scheme of the suggested workflow (maybe even with pro and cons) might be very useful for subsequent studies on PD in an ecological context.

- This might be a petty comment, but I would suggest to present the tree in Figure 2 in a clockwise direction. I would also suggest to provide the tree in a format that it can be explored by readers interested in the topology (e.g. as supplemental material).

Author Response

Dear reviewer1,

We appreciate the time and effort you dedicated to providing feedback on our manuscript. We agreed with most of your comments and suggestions (see details attached) and we have incorporated them in the manuscript. 

Thank you for all the useful comments, we think they really improved the quality of the manuscript. 

---------

The study by Macias-Hernández et al provides not only the most complete spider
phylogeny to date, but also thoroughly evaluate a workflow to estimate and use
phylogenetic diversity especially with regard to ecological studies. The manuscript is
very well written and the figures are excellent. I enjoyed reading the paper, which
makes certainly an important contribution to spider research and I have only few
suggestions/comments.
- Since time-stamped trees play a major role in the suggested workflow, I was
wondering whether the authors considered the recent paper of Magalhaes et al (2020,
Biol Rev 95)? In this paper Magalhaes et al revised the fossil record especially with
regard to calibrating trees. I suggest to explore this study as it present an updated
(revised) dated spider tree of life showing different age estimations compared to
Fernandez et al (2018). I have no idea whether it will change the results in the present
study, but I think it should be tested in the present study! If the authors reached a
different results it will be worth discussing the role of well justified calibrations points.
Magalhaes et al (2020) detailed revision of spider fossil data and subsequent reanalyses
of trancriptomic data is a remarkable and most welcome contribution to the field. At
certain point we did consider using Magalhaes et al inferred ages. However, as the scope
of our manuscript was not to infer absolute times but to transform the phylogenetic trees
into ultrametric trees, we decided to simplify the process using the same dated tree used
in Fernández et al 2018, which incorporates the same taxa as our backbone matrix and
topological constraint. To be honest, we doubt that the use of the calibration points from
Magalhaes et al (2020) will have an impact on the results of our study, and we think that
it hardly justifies the amount of work involved. Additionally, Magalhaes et al also indicates
that most of the 95% credible intervals of the mean age estimates broadly overlap with
other studies, and although some studies used inadequate calibrations the main
conclusions might not have been affected significantly. We have now included the
following sentence in the Material & Methods sections, Lines 276-280: “Recently
Magalhaes et al (2020) have published a time stamped tree of the life of spiders based
on a reexamination of the available fossil data. Although the location of some of the
calibration points differed from those used in Fernandez et al 2018, the inferred confident
intervals broadly overlapped”.
- I was missing a critical discussion on the barcode gene. Is it always "working" for
spiders with regard on genus or family level? Furthermore, it is not surprising that the
topologies of COI alone differ from phylogenies derived from more extensive data (see
page 12, line 560). However, I missed a discussion about the reason for it and why the
barcode gene is not sufficient for this kind of studies. It might be common knowledge,
but given the goal of the paper it might be worth of a few explanatory sentences.
We added references on the efficiency of COI at identifying spiders, and we added
comments on the different mutation rates of mitochondrial and nuclear genes and how
they affect shallow/deep nodes in line 604 in the “Tree topology” section of the
discussion.
- Figure 1 shows the workflow used in this study (including the abbreviations of the
matrices used, etc.). If possible I would suggest to provide another figure
corresponding with the section "recommendations and suggested workflow" in the
discussion. The strength of this paper is the critical evaluation of a variety of
approaches and I think a clean scheme of the suggested workflow (maybe even with
pro and cons) might be very useful for subsequent studies on PD in an ecological
context.
As suggested by the two reviewers, we made an extra Figure 3 explaining the
recommended workflow.
- This might be a petty comment, but I would suggest to present the tree in Figure 2 in
a clockwise direction. I would also suggest to provide the tree in a format that it can be
explored by readers interested in the topology (e.g. as supplemental material).
We agree and we modified the tree in Figure 2 so now it has a clockwise direction. We
also provide the final reference tree in .pdf format in supplementary material (Figure
S1) for an easier exploration, and it is also indicated in the main text. Additionally, all
trees are available in Zenodo (as it is indicated in the manusc

--------

Best regards,

The authors

Reviewer 2 Report

General comment

This contribution constitutes a very fine piece of work that comes timely with a growing interest in computing metrics of phylogenetic diversity for arthropod assemblages. As mentioned in the ms, there are (very) few papers on the topic (especially dealing with spiders) and given the numeric and functional weight of this taxon in many ecosystems, this paper is expected to have an important impact on future research. The ms is well written (although I was a bit lost by places, see my comments below), and relies on a solid dataset and workflow. I have few things to do here, except congratulating the authors and providing a couple of minor suggestions.

Minor comments

L43 (and further in the ms): I missed how the choice of a backbone matrix was investigated.

L72-L89: Writing is super general (not to say elusive), and I was expecting more details on the cited papers (habitats, regions, taxa, etc.), at least in discussion but it didn’t come neither…(see below).

L83: is it a strong or weak influence?

L115: again, it is not clear (to my opinion) how the influence of backbone matrix was tested.

L120: please recall that this study is made for community ecology purposes.

L125: the Latin species name of the white oak is missing

L133: I missed how species were identified and where (voucher) specimens were stored.

Discussion: The authors do not mention other important related perspectives, such as beta diversity of phylogenetic relatedness (despite they studied 2 regions and several sites) and (abiotic) driving forces of PD (see e.g. Fichaux et al. 2019 // Oecologia).

L261-263: I suggest displacing this sentence just before the previous one.

L297: So you’re not expecting any interaction, or you’re not interested in testing it? That should be specified.

L314: please provide the software(s) used.

L319: I guess you mean 25% of species pool from each region, right?

L330-333: This is not discussed at all later.

L343-353: I was thinking, could a variance partition help in statistically quantifying differences between trees?

L356: PD was specified ‘PD alpha’, shouldn’t be the same for the first TD occurrence?

L388: Would this part fit earlier? I was somewhere lost in the ms structure, with a mismatch between sections and initial objectives.

L403: Maybe change these codes into reader-friendly names. Generally speaking, I guess an effort could be done for recoding trees, with a more meaningful combination of letters. As it stands, they look like workcodes, which is not very pedagogic.

L422: “were influenced by all factors”, but see “…had little, if any, effect…” on L352.

L531: a citation would be welcome.

L534-538: this is actually not the context of the study. I suggest either move this to the introduction (and then connect to e.g. the seminal paper by Thuiller et al. 2015 // Phil Trans R Soc B, but there are many others) or delete.

L539: I actually missed a proper workflow here. To be changed accordingly, e.g. by adapting and describing Figure 1.

L540-547: Is this necessary? I find these sentences quite redundant with L405-420.

L560-567: Idem.

L568-579: Could be placed before.

L598: Please support by a citation.

L596-600: these sentences could move to the introduction, but I’m pretty sure that’s something also considered, at some point!

L600: There is actually another study dealing with TD, FD and PD in spiders: Ridel et al. 2020 // BioRxiv (https://biorxiv.org/cgi/content/short/2020.06.19.161588v1).

L641-642: Please refer to Table 1 for codes (but see my comment above).

L645-646: Please add citation numbers in the caption.

L648: Need a more complete title.

L662: using which statistics? Good to recall at this stage.

There some typos through the ms (e.g. L8: Arnedo and M.A.->and Arnedo M.A., L38: affects->affect, L44: had only->only had, L51: phyllogenetic->phylogenetic), easy to fix with a careful reading / editing.

Author Response

Dear reviewer2,

We appreciate the time and effort you dedicated to providing feedback on our manuscript. We agreed with most of your comments and suggestions (see details attached) and we have incorporated them in the manuscript. 

Thank you for all the useful comments, we think they really improved the quality of the manuscript. 

-----

Minor comments
L43 (and further in the ms): I missed how the choice of a backbone matrix was
investigated.
We investigated the effect of the use of a backbone matrix by including the taxa from
the Wheeler et al (2016) matrix that contained 964 terminals, and it was chosen
because it represents the most complete matrix in terms of taxon sampling to date. The
analyses were performed including the backbone matrix (all matrices but Lc1 and
Lc128), or without including the backbone matrix (just the species collected from our
local communities; matrices Lc1 and Lc128). See matrices details in Table 1. We have
clarified this in the main text (line 245-248).
L72-L89: Writing is super general (not to say elusive), and I was expecting more details
on the cited papers (habitats, regions, taxa, etc.), at least in discussion but it didn’t
come neither…(see below).
As the reviewer suggested, we have added more information about taxa and regions in
this paragraph (L83-L89).
L83: is it a strong or weak influence?
We have added “a strong influence” to the main text. Line 92.
L115: again, it is not clear (to my opinion) how the influence of backbone matrix was
tested.
See above comment L43.
L120: please recall that this study is made for community ecology purposes.
Done. We specified phylogenetic diversity metrics in community ecology studies as the
purpose of the phylogenetic reconstruction guidelines. Line 138.
L125: the Latin species name of the white oak is missing.
All the sampled communities in the Iberian Peninsula were oak forests, but from
different species (Quercus petraea, Quercus faginea, Quercus subpyrenaica and
Quercus pubescens). We have now included the specific information. Line 143.
L133: I missed how species were identified and where (voucher) specimens were
stored.
We have added this information in the main text in line 155.
Discussion: The authors do not mention other important related perspectives, such as
beta diversity of phylogenetic relatedness (despite they studied 2 regions and several
sites) and (abiotic) driving forces of PD (see e.g. Fichaux et al. 2019 // Oecologia).
Yes, we do agree with the reviewer that there are many other aspects and indexes that
could be affected by phylogenetic reconstructions (e.g. beta PD, phylogenetic
relatedness or abiotic factors among others). In order to simplify the manuscript we just
focused on deciphering how different factors related to phylogenetic reconstruction
could ultimately affect the estimation of one of the most basic yet general indexes, this
is the alpha PD. We do acknowledge that all related indices should be probably
affected as well, but this was beyond the scope of the present paper. In a different
manuscript we will investigate the effect of abiotic factors and beta PD between the two
studied regions.
L261-263: I suggest displacing this sentence just before the previous one.
Done. We agree on this change of order, as the referred sentence is still introductory
and the other one explains the two possible cases.
L297: So you’re not expecting any interaction, or you’re not interested in testing it?
That should be specified.
We are not interested in testing it. As we are just including one predictor variable and a
random effect on each model, we should add at least two predictor variables for testing
their interaction. In this case, we were interested in testing how correlated each
community rank obtained with the different trees were regarding the different conditions
tested.
L314: please provide the software(s) used.
Done. We added that all statistical analyses were made with the software R and the
TSS test conducted by applying the formula from Allouche et al, 2006 in Excel. Line
421.
L319: I guess you mean 25% of species pool from each region, right?
We meant 25% of all species present in the two regions. We have clarified it in the
main text. Line 456.
L330-333: This is not discussed at all later.
The phylogenetic trees B_tc and B were inferred from the backbone matrix (Wheeler et
al, 2016) only to test the effect of using a different topological constraint than the one
used in Fernández et al, 2018. In the results section we discussed the main
differences, but we did not get into more details in the discussion section because it
was not the scope of the manuscript.
L343-353: I was thinking, could a variance partition help in statistically quantifying
differences between trees?
The reviewer refers to the section in which we compared the monophyly index between
the different trees. Maybe a variance partition could have been a more elegant way to
conduct comparisons, statistically speaking. However, we did decide to simplify tree
comparison by pruning taxa not present in the two trees being compared (pairwise
deletion).
L356: PD was specified ‘PD alpha’, shouldn’t be the same for the first TD occurrence?
Yes, we have modified the sentence accordingly. Line 509.
L388: Would this part fit earlier? I was somewhere lost in the ms structure, with a
mismatch between sections and initial objectives.
As the reviewer suggested we have reorganized the paragraphs to simplify the ms
structure. We have now changed the order of sections 2.5 Phylogenetic community
metrics and statistical analyses, and 2.6 Null models and ecological patterns in the
results section following the same order and headings as the discussion.
L403: Maybe change these codes into reader-friendly names. Generally speaking, I
guess an effort could be done for recoding trees, with a more meaningful combination
of letters. As it stands, they look like workcodes, which is not very pedagogic.
As the reviewer suggested we have recoded the matrices/trees with more readerfriendly
names. We further explain the meaning of each name in the legend of Table 1.
As: Table 1. Attributes of the matrices assembled in this study with information on their
taxa, number of genes and missing data. Matrices names are given according to the
data included, as follows: (B) taxa from backbone matrix (Wheeler et al. 2016,
Fernández et al. 2018) and (L) taxa from local communities (Iberian Peninsula and
Macaronesia) (=Taxa); genes added for local species, (c1): mitochondrial COI barcode
gene, (28): nuclear 28S rRNA gene (=Genes); (tc): topological constraint (=Topology);
(cal): time-calibrated trees (=Time). Terminals: number of terminals; % missing:
percentage of missing cells in the matrix; COI, 28S, 12S, 16S, 18S, H3: numbers of
positions (aligned) for the respective genes.
Tree names Alternative name
W B
W_b B_tc
WMc1_b BLc1_tc
WMc1_b_c BLc1_tc_cal
WMc1 BLc1
WMc1_c BLc1_cal
WMc128_b BLc128_tc
WMc128_b_c BLc128_tc_cal
WMc128 BLc128
WMc128_c BLc128_cal
Mc1 Lc1
Mc128 Lc128
L422: “were influenced by all factors”, but see “…had little, if any, effect…” on L352.
All factors tested had some effect, but some had a major effect than others. In the case
of the inclusion of missing data (as referred by the reviewer in L422 and L 352) it is
also discussed in detail in the discussion L430 (current version L584). We have now
explained it better.
L531: a citation would be welcome.
We are afraid there may not be a proper citation, since we made this statement out of
common sense. We consider that the use of the actual information of the species
present in the community is more reliable than using a surrogate or approximation by
matching the local species with something loosely resembling it in a databases or,
alternatively, using higher taxonomic ranks.
L534-538: this is actually not the context of the study. I suggest either move this to the
introduction (and then connect to e.g. the seminal paper by Thuiller et al. 2015 // Phil
Trans R Soc B, but there are many others) or delete.
As the reviewer suggested we have deleted this paragraph from the manuscript.
L539: I actually missed a proper workflow here. To be changed accordingly, e.g. by
adapting and describing Figure 1.
Following the suggestions made in the “Recommendations and suggested workflow”
section, we have now included a new figure (Fig. 3) based on the workflow in Figure 1,
but highlighting the factors that are more relevant to generate a phylogeny to estimate
PD (named Figure 3).
L540-547: Is this necessary? I find these sentences quite redundant with L405-420.
We agreed with the reviewer and we have partially deleted the redundant sentences.
L560-567: Idem.
As this paragraph refers to the final recommendations, we have compacted the writing,
removing some extra sentences, but keeping the main message.
L568-579: Could be placed before.
We are sorry, but in this case we do not agree with the reviewer. We have kept the
sentence in their original place.
L598: Please support by a citation.
We added the reference of the following study by Schmitz and collaborators (2000),
where they show the important role of spiders in regulating food webs through trophic
cascades. Line 115.
Schmitz OJ, Hambäck PA & Beckerman AP (2000) Trophic cascades in terrestrial
systems: A review of the effects of carnivore removals on plants. American Naturalist
155:141–153. https://doi.org/10.1086/303311.
L596-600: these sentences could move to the introduction, but I’m pretty sure that’s
something also considered, at some point!
We agree. We moved those sentences introducing the organism of study to the
introduction (L 113-117).
L600: There is actually another study dealing with TD, FD and PD in spiders: Ridel et
al. 2020 // BioRxiv (https://biorxiv.org/cgi/content/short/2020.06.19.161588v1).
Thanks for the suggestion, this manuscript was made available in BioRxiv after our
submission to Diversity. Now it has been included.
L641-642: Please refer to Table 1 for codes (but see my comment above).
Done.
L645-646: Please add citation numbers in the caption.
Done.
L648: Need a more complete title.
We added the code and information of the specific tree the figure depicts, as well as
information on what the different colours represent.
L662: using which statistics? Good to recall at this stage.
Done.
There some typos through the ms (e.g. L8: Arnedo and M.A.->and Arnedo M.A., L38:
affects->affect, L44: had only->only had, L51: phyllogenetic->phylogenetic), easy to fix
with a careful reading / editing.
We replaced “Arnedo and M.A.” by “and Arnedo, M.A.” in line 8.
We replaced “affects” by “affect” in line 38.
We replaced “had only” by “only had” in line 44.
We replaced “phyllogenetic” by “phylogenetic” in line 51.
We detected and corrected some more typos, which are indicated in the revised
version of the manuscript.

-------

Best regards,

The authors