DNA Barcoding of Stingless Bees (Hymenoptera: Meliponini) in Northern Peruvian Forests: A Plea for Integrative Taxonomy

Round 1

Reviewer 1 Report

It was a pleasure to evaluate this work. The presented manuscript deals with a biodiversity survey of stingless bees in northern Peruvian forest using an integrative taxonomic approach. The study is well conceived and properly executed. Biodiversity inventories are especially important now, when we live in the era of six mass extinction caused by human activity. Therefore, the authors make an important contribution to the field with publication of these results. I think the paper requires only a minor revision, and here are several comments which should be addressed by the authors.

-        >  The authors put considerable attention to the monophyly of the clades/species having the only NJ trees. This is a distance-based method for tree calculation and it results always with fully dichotomous trees. It has an advantage by being super-fast but in the other way is not very detailed. As the trees are not so big (the number of terminals is not large) I would suggest to use the prepared alignments and calculate the trees with the ML method that is now also much upgraded and time effective (see IQtree). Also, I couldn’t see the outgroups that should be used for the tree calculation and then rooting to give the tree a direction.

-       > The authors conducted molecular species delimitation in their work, but in my opinion the results are not sufficiently available for the readers. I would recommend to provide an additional section in the MS with a description and comparison of all the results from delimitation analyses.

-       >   One of the delimitation methods is bPTP which uses a non-ultrametric phylogenetic tree as the input data. Although trees of different methods can be used, bPTP was originally used for ML trees. The authors did not specify which trees were used for the analyzes. I recommend calculating ML trees with the IQ tree and using them for the bPTP analyses.

-        >  Figures comprising phylogenetic trees are inserted into the file as pictures and thus are of low quality. I would recommend inserting the tree figures into vector format to increase their quality and readability.

Author Response

Rev: The authors put considerable attention to the monophyly of the clades/species having the only NJ trees. This is a distance-based method for tree calculation and it results always with fully dichotomous trees. It has an advantage by being super-fast, but in the other way is not very detailed. As the trees are not so big (the number of terminals is not large) I would suggest to use the prepared alignments and calculate the trees with the ML method that is now also much upgraded and time effective (see IQtree). Also, I couldn’t see the outgroups that should be used for the tree calculation and then rooting to give the tree a direction.

Reply: As our paper was focused on the usefulness of DNA barcoding in identifying Peruvian Meliponini, we generated the Neighbour-Joining trees using the Kimura 2-parameter (K2P), following the standardized procedure routinely used by BOLD to classify the query sequences. These distance-based trees were generated solely to easily depict possible unreliability of “barcoding gaps”, evidenced by polyphyly. Thus, for the aim of our work, we did not generate phylogenetic trees sensu stricto, but “operative” barcode-phylogenetic trees to evaluate the usefulness of the barcoding procedure. In fact, barcode-phylogenetic trees are “not really” phylogenetic trees: in barcode-phylogenetic trees the differences between nucleotide characters are more important than the inferred evolutionary history. The main aim is to evaluate if sequences from the same species are clustered or not in a monophyletic branch and if the different ones are distributed in separated clades, beyond their evolutionary history. This also explain why we did not include outgroup taxa (the tree is, in fact, midpoint rooted). We clarified this point in M&M [lines 156-162 “To assess the relationships among our query sequences and their neighbouring reference sequences, we built consensus (distance-based) Neighbour Joining (NJ) phylogenetic trees (500 bootstrap replicates, K2P model, midpoint rooted) using MEGA 11 [64]. Barcode-phylogenetic trees generated for the various stingless bee genera were only built to evaluate the clustering of sequences from the same (predicted) species in monophyletic clades (not to trace evolutionary histories)”]. In any case, ML topologies (not shown, but used for PTP and bPTP analyses, see below) overlapped with those of NJ trees. We are currently working on two forthcoming papers aimed at reconstructing the molecular phylogeny of Trigona and Plebeia through ML and Bayesian approaches, including a larger number of markers, specimens and five outgroup taxa.

Rev: The authors conducted molecular species delimitation in their work, but in my opinion the results are not sufficiently available for the readers. I would recommend providing an additional section in the MS with a description and comparison of all the results from delimitation analyses.

Reply: Thanks for addressing this point. Species delimitation methods (SDMs) - routinely used in barcoding approaches to identify species/OTUs - were also applied to our datasets. However, due to the pervasiveness of polyphyly and the a priori taxonomic uncertainties on Peruvian Meliponini (as extensively discussed in the paper), the (conflicting) results achieved by distance- vs. tree-based SDMs were mostly uninterpretable and inconclusive. That is the reason why we did not emphasize SDM results. These, however, are available in Suppl. Materials (Table S1) for any comparison. We now specified the availability of SDMs results at line 190 of “Results” (…“see also results of SDMs in Table S1”). Some of SDM results were already included in the “Results” section when supporting the potential existence of some taxonomic “entities” which appeared well-distinguished from all other species/OTUs (see e.g. line 265-266; 369-370).

Rev: One of the delimitation methods is bPTP which uses a non-ultrametric phylogenetic tree as the input data. Although trees of different methods can be used, bPTP was originally used for ML trees. The authors did not specify which trees were used for the analyzes. I recommend calculating ML trees with the IQ tree and using them for the bPTP analyses.

Reply: Right point. We added this information in M&M [line 171-172: “Maximum Likelihood (ML) input trees for PTP and bPTP analyses were generated with MEGA 11 [64] (NNI heuristic method, K2P model)]

Rev: Figures comprising phylogenetic trees are inserted into the file as pictures and thus are of low quality. I would recommend inserting the tree figures into vector format to increase their quality and readability.

Reply: Thanks for this suggestion. We generated a vector format of both NJ trees to improve the quality of figures.

Reviewer 2 Report

Dear Author and Editor

This is a comprehensive study. The results of this study have certain implications for the application of DNA barcoding to identify species in the tribe Meliponini (Hymenoptera) in Peru.

However, there are still many unidentified samples. As well as a COX-based phylogenetic tree yielding many polyphyletic clades. This shows that more powerful markers need to be tested to provide a thorough answer to integrative taxonomy. 

The authors used mitochondrial Cytochrome Oxidase I (COI) as a marker, so what about nuclear markers such as 28S Ribosomal DNA?  Please discuss further the power of mitochondrial and nuclear makers in stingless bees barcoding.

Best wished

Author Response

Rev: […] There are still many unidentified samples. As well as a COX-based phylogenetic tree yielding many polyphyletic clades. This shows that more powerful markers need to be tested to provide a thorough answer to integrative taxonomy. The authors used mitochondrial Cytochrome Oxidase I (COI) as a marker, so what about nuclear markers such as 28S Ribosomal DNA?  Please discuss further the power of mitochondrial and nuclear makers in stingless bees barcoding.

Reply: Thanks for raising this point. Following this suggestion, we briefly point out on the usefulness of mitochondrial and nuclear markers in integrative taxonomy of stingless bee (see Discussion, line 553-555: “A combined use of the already available and tested mitochondrial and nuclear markers [75-76] would greatly improve taxonomy and identification of stingless bees”).