Comparing eDNA metabarcoding primers for assessing fish communities in a biodiverse estuary.

Metabarcoding of environmental DNA is increasingly used for biodiversity assessments in aquatic communities. The efficiency and outcome of these efforts are dependent upon either de novo primer design or selecting an appropriate primer set from the dozens that have already been published. Unfortunately, there is a lack of studies that have directly compared the efficacy of different metabarcoding primers in marine and estuarine systems. Here we evaluate five commonly used primer sets designed to amplify rRNA barcoding genes in fishes and compare their performance using water samples collected from estuarine sites in the highly biodiverse Indian River Lagoon in Florida. Three of the five primer sets amplify a portion of the mitochondrial 12S gene (MiFish_12S, 171bp; Riaz_12S, 106 bp; Valentini_12S, 63 bp), one amplifies 219 bp of the mitochondrial 16S gene (Berry_16S), and the other amplifies 271 bp of the nuclear 18S gene (MacDonald_18S). The vast majority of the metabarcoding reads (> 99%) generated using the 18S primer set assigned to non-target (non-fish) taxa and therefore this primer set was omitted from most analyses. Using a conservative 99% similarity threshold for species level assignments, we detected a comparable number of species (55 and 49, respectively) and similarly high Shannon's diversity values for the Riaz_12S and Berry_16S primer sets. Meanwhile, just 34 and 32 species were detected using the MiFish_12S and Valentini_12S primer sets, respectively. We were able to amplify both bony and cartilaginous fishes using the four primer sets with the vast majority of reads (>99%) assigned to the former. We detected the greatest number of elasmobranchs (six species) with the Riaz_12S primer set suggesting that it may be a suitable candidate set for the detection of sharks and rays. Of the total 76 fish species that were identified across all datasets, the combined three 12S primer sets detected 85.5% (65 species) while the combination of the Riaz_12S and Berry_16S primers detected 93.4% (71 species). These results highlight the importance of employing multiple primer sets as well as using primers that target different genomic regions. Moreover, our results suggest that the widely adopted MiFish_12S primers may not be the best choice, rather we found that the Riaz_12S primer set was the most effective for eDNA-based fish surveys in our system.

1 Introduction

The monitoring of marine and aquatic communities requires accurate biodiversity assessments which are typically based on surveys conducted using nets, traps, cameras, or direct observation. Environmental DNA (eDNA) approaches are emerging as a tool for the characterization of marine biodiversity that can complement traditional surveys [1-4]. The initial eDNA studies in aquatic systems were published a decade ago and focused on species-specific primers and qPCR to detect invasive or endangered freshwater species [5-Conserv Lett. 2011 ">7]. Not long thereafter, researchers began exploring the use of eDNA in marine environments [8] including the use of metabarcoding approaches to identify and characterize whole communities [9-11]. This latter approach involves PCR amplification and library preparation using primers that are designed to amplify a barcoding gene (i.e., COI, 16S, 18S) across a specific taxonomic group; the breadth of which can vary from metazoans to a single genus. Subsequent sequencing of the metabarcoding libraries on high-throughput sequencing platforms such as an Illumina MiSeq generates millions of sequence reads that must then be sorted bioinformatically and assigned a specific taxonomy.

Prior to beginning a metabarcoding initiative, primers that target the group of interest must be selected from the literature or designed de novo. While there are a number of published primer sets already available [12], each suffers from unique biases making primer testing essential. For instance, primer-template mismatch or unequal amplification efficiency (PCR bias) can result in uneven amplification and false negatives [13-15]. This is particularly true for primers designed to work across broad taxonomic groups where primer mismatches are likely to occur. PCR bias can also result from the competition of unbound primers for DNA templates, in which case the most abundant templates are more likely to be amplified, resulting in false negatives for low copy number templates [16,17]. Furthermore, if PCR conditions allow for non-specific binding (low annealing temperature) or if poorly designed primers permit for ubiquitous template binding, preferential amplification of the most common templates can lead to sequencing runs dominated by non-target taxa such as bacteria, fungi, or algae at the expense of less common target DNA.

Cytochrome c oxidase subunit I (COI) has historically been the most common gene for DNA barcoding in animals [18-21]. As a result, there is extensive reference data available for COI in the public databases (National Center for Biotechnology Information, NCBI; the Barcode of Life Data System, BOLD). However, the conserved nature of COI, which makes it useful as a barcoding gene, also complicates the design of taxon-specific primers for metabarcoding [22-24]. Consequently, COI metabarcoding primers are designed with a high degree of base degeneracy [25,26] which results in the amplification of a wide breadth of taxa [27], but also often results in a large percentage of non-target sequence reads [28,29]. As a result, COI is not commonly used for eDNA metabarcoding [but see 18,30]. Instead, the mitochondrial 12S and 16S rRNA genes are generally favored [16,31-36]. Both genes include conserved regions for primer design as well as variable regions that allow for genus or species-level resolution. Moreover, reference sequences for these two genes, particularly for fishes, are well represented in the public databases [23,37].

Before beginning an experiment, the performance of newly designed primers can be evaluated in silico [16], however, there is no guarantee that they will perform as predicted when applied to field-based samples. The efficiency of metabarcoding primers varies with community composition and complexity [16,22,38], and as a result, it is recommended that candidate primer sets be tested on field samples [29]. Over the past several years, a number of studies have been published that evaluate and compare different primer sets for use in eDNA metabarcoding of fishes [16,29,39-43]. In a recent study, Zhang et al. [42] evaluated the efficacy of 22 published metabarcoding primer sets covering several gene regions and using both in silico and in vitro analysis on freshwater fishes in China. They found that, in general, 12S rRNA primers detected a greater number of species than either 16S rRNA or COI primers. Moreover, they found inconsistent results when comparing in silico and in vitro experiments [42]. Here we focus on five primer sets (12S and 16S), three of which performed well in the freshwater study of Zhang et al. [42], and include an additional 18S primer set not previously evaluated (but designed for fishes). We assess their suitability for detecting fishes in a species rich estuarine system in Florida (the Indian River Lagoon) and compare them in terms of 1) specificity (amplification of fishes at the exclusion of other taxa), 2) universality (amplification of a diversity of fish taxa), and 3) taxonomic resolution (ability to resolve taxa to the species level).

2 Materials and methods

2.1 Primer selection and sample collection

Here we selected five primer sets to evaluate. These included three 12S rRNA and one 16S rRNA primer set (Table 1) that had been employed in a number of aquatic eDNA studies targeting fishes. In addition, we selected one nuclear 18S rRNA primer set [44] that was designed for freshwater fishes but had not been evaluated for eDNA metabarcoding.

Table 1

Primer sets used in this study including primer sequence, annealing temperature used for PCR1, and expected average amplicon length.

Primer set	Locus	Original primer name	Primer sequence (5’-3’)	Annealing temperature (°C)	Amplicon length (bp)	Reference
MiFish_12S	12S	MiFish-U-FMiFish-U-R	GTCGGTAAAACTCGTGCCAGC CATAGTGGGGTATCTAATCCCAGTTTG	61.5	171	Miya et al. 2015
Riaz_12S	12S	12S-V5f12S-V5r	ACTGGGATTAGATACCCC TAGAACAGGCTCCTCTAG	55	106	Riaz et al. 2011
Valentini_12S	12S	L1848H1913	ACACCGCCCGTCACTCT CTTCCGGTACACTTACCATG	55	63	Valentini et al. 2016
Berry_16S	16S	Fish16sF/D16s2R	GACCCTATGGAGCTTTAGAC CGCTGTTATCCCTADRGTAACT	54	219	Berry et al. 2017
MacDonald_18S	18S	Fish_18S_1FFish_18S_3R	GAATCAGGGTTCGATTCC CAACTACGAGCTTTTTAACTGC	62	271	MacDonald et al. 2014

Two replicate 500 ml water samples were collected in June 2018 from each of six sites within the northern Indian River Lagoon (IRL) in central Florida (S1 Table). Sites were in close proximity and over a similar shallow water habitat known to support oyster reefs. Water samples were taken at the surface. Prior to water sampling, all collection bottles, forceps, scissors, and filter holders were sterilized with 20% sodium hypochlorite solution for at least 20 min, rinsed with reverse osmosis (RO) water and air dried. Water samples were collected in sterilized Nalgene bottles including two negative field controls consisting of sterile Nalgene bottles filled in the field with store-bought bottled water. As is true for much of the IRL, our water samples were highly turbid with high concentrations of suspended organic and/or inorganic material [45]. Following the recommendations of Kumar et al. [46], samples were stored on ice and transported back to the lab and filtered using 0.45 μm mixed cellulose ester (MCE) filters. Negative laboratory controls consisted of 500 ml of RO water filtered using the same protocol as our field samples. After filtration, filter membranes were stored in 3 ml tubes in Longmire’s buffer at -20°C until DNA extraction.

2.2 Library preparation and sequencing

All DNA extractions were carried out in a dedicated PCR free workspace at the University of Central Florida Marine Molecular Ecology and Evolution Laboratory. To prevent contamination, all equipment and bench spaces were cleaned before use with 10% sodium hypochlorite solution followed by 70% ethanol and irradiated with UV light for 20 min. All pipetting was conducted using sterile barrier filter tips. Prior to DNA extraction, filters were cut in half and one-half was archived in Longmire’s buffer and stored at -20°C. The other half was placed in a 1.5 ml Eppendorf tube for DNA extraction and cut into small pieces using sterilized scissors. DNA was extracted from each and eluted in 100 μL of buffer using the E.Z.N.A. Tissue DNA Kit (Omega Bio-tek, Inc., GA, USA), following the manufacturers’ protocol and which has been shown to perform well in previous experiments [45]. The resulting extraction was purified using a Zymo OneStep PCR Inhibitor Removal Kit (Zymo Research, CA, USA), and eluted in a final volume of 50 μl. DNA concentrations were determined using a Qubit 4.0 and the dsDNA High Sensitivity Assay Kit (Invitrogen, CA, USA).

Illumina libraries were constructed for each sample using a two-step PCR protocol following Kumar et al. [45]. PCR 1 (qPCR) was performed using custom primers that included both Illumina sequencing primers and locus specific primers (see Table 1 in [45]). Amplifications were carried out on a CFX96 Touch Real Time PCR System (Bio-Rad, CA, USA) in a total volume of 25 μl with each reaction containing 12.5 μl of 2× SsoAdvance Universal SYBR Green Supermix (Bio-Rad), 0.5 μl forward primer (10 μM), 0.5 μl reverse primer (10 μM), 2 μl of template DNA, and 9.5 μl of ultrapure water (ThermoFisher Scientific, MA, USA). The thermocycling profile included an initial denaturation step at 95°C for 3 min, followed by 30 cycles of denaturation at 95°C for 30 s, annealing at the primer annealing temperature (Table 1) for 30 s, and extension at 72°C for 30 s, followed by a final extension at 72°C for 5 min. Each qPCR run included a no-template control as well as an extraction control. To minimize false negatives (PCR dropouts), qPCRs were performed in duplicates. Following qPCR, duplicates were pooled before excess primers and dNTPs were removed using an E.Z.N.A. Cycle Pure Kit (Omega Bio-tek, Inc.) following manufacturer’s protocol. The purified PCR products were quantified using a Qubit 4.0 fluorometer and served as the template DNA for PCR 2.

PCR 2 was performed using primer pairs consisting of Illumina adaptors (P5 and P7), 8 bp Nextera index sequences, and an overhang sequence complementary to the Illumina sequencing primer (see S1 Fig in [45]). Amplifications were carried out using a Veriti Thermal Cycler (Applied Biosystems, CA, USA) and 25 μl reaction volumes containing 12.5 μl IBI Taq 2× Master Mix (IBI Scientific, IA, USA), 0.5 μl forward primer (10 μM), 0.5 μl reverse primer (10 μM), 2 μl DNA template, and 9.5 μl of ultrapure water. PCR cycling conditions were identical to PCR 1 except only 15 cycles were run and a universal annealing temperature of 55°C was employed. Final PCR products were cleaned using E.Z.N.A. Cycle Pure Kits, quantified on a Qubit 4.0, and pooled in equimolar concentrations (one pool for each primer set). Each pooled library was then size-selected based on expected fragment size using a PippenHT (Sage Science, MA, USA) and a 2% agarose gel cassette and quantified using a NEB Next Library Quantification Kit for Illumina (New England Biolabs, MA, USA). The library was adjusted to 4 nM and denatured following Illumina protocols. The denatured library was combined with 10% PhiX control and sequenced bidirectionally on an Illumina MiSeq at the University of Central Florida Genomics and Bioinformatics Cluster (GBC) Core Laboratory. Sequencing was conducted using a Nano 300 v2 (2 × 150) Reagent Kit for 2×111 cycles and a Nano 500 v2 (2 × 250) Reagent Kit for 2×251 cycles, depending on amplicon size.

2.3 Bioinformatic processing

The Illumina sequencing data was demultiplexed using the Illumina MiSeq software and downloaded onto an in-house server maintained by the Genomics and Bioinformatics Cluster (GBC) at the University of Central Florida. Individual FASTQ files were then filtered following a series of quality control steps using USEARCH v10 [47] and VSEARCH v2.14 [48]. First, the forward and reverse reads were merged using the fastq_mergpairs command in USEARCH with a minimum overlap of 100 bp for MiFish_12S, Berry_16S, MacDonald_18S; and 60 bp for Valentini_12S and Riaz_12S, and a maximum number of mismatches set at 3 bp. Sequences of unexpected length were discarded using the -fastq_minlen command of USEARCH to retain only those reads with maximum deviation of 10% from the minimum amplicon length. To locate and remove primers, the merged reads were sub-sampled to 5,000 sequences and then VSEARCH was used to remove primers. Next, we dereplicated sequences and discarded sequences with expected errors > 0.5 using VSEARCH. Finally, unique sequences were denoised using the UNOISE3 [49] option implemented in USEARCH. UNOISE3 generates zero radius operational taxonomic units (ZOTUs) by correcting point errors and filtering chimeric sequences [49]. To minimize the chance of spurious sequences being included in the final dataset, the minimum abundance of five reads were set to generate amplicon sequence variants (ASVs).

2.4 Taxonomic assignments

ASVs were compared against the NCBI GenBank nucleotide database using BLASTn with the default parameters. To retrieve the full taxonomic identity, we queried each of the “taxids” from the BLAST results against the NCBI database using the “taxonkit lineage” command in the program TaxonKit [50]. To reduce the uncertainty in taxonomic assignments, we discarded ASVs with a bitscore below 250 and/or query coverage below 100%. Each ASV was then assigned to the lowest taxonomic level based on the percent similarity to NCBI alignments. We recognize that the rate of evolution varies across genes and gene regions and so setting a single taxonomic threshold and applying it across all primer sets could impact interpretations of the results. Therefore, we examined results for three similarity thresholds for species level designations (99%, 98%, and 97%) and present these in supplemental materials (S2 Table). These results did not change our interpretation of the data and therefore in main manuscript we present the results for the following taxonomic thresholds for all primer sets: 99% for species; 97% for genus; 95% for family; 90% for order; 85% for class; and 80% for phylum following West et al. [35]. The resulting list of species was checked against a list of known species from the Indian River Lagoon. Species detected in our metabarcoding data but absent from that list were queried against FishBase (www.fishbase.org) to determine if they were present in the central-west Atlantic. If an ASV matched ≥ 99% with two closely related species but only one of them was known to occur in the Indian River Lagoon or eastern Atlantic, then that ASV was assigned to that known taxon. However, if both species were known to occur in the eastern Atlantic region, taxonomic assignments were collapsed to the genus level.

2.5 Statistical analyses

Unless otherwise specified, all statistical analyses were performed using R version 4.0.2 [51]. For data analyses, sequence reads from the two replicates taken at each of the six sampling sites were pooled. We computed diversity indices (species richness, evenness, and Shannon’s diversity) using the BiodiversityR package v. 2.12–3 [52] and standard deviations represent variation across the six sample sites. We ran an analysis of variance (ANOVA) using the program JMP Pro 12 (SAS Institute Inc., NC, USA) to determine if diversity indices differed among sample sites and across primer sets. When a significant interaction was detected, we performed a post-hoc Tukey-Kramer Honest Significant Difference (HSD) test to determine which group means were significantly different. Dissimilarity in species composition among the different primer sets were calculated by non-metric multidimensional scaling (NMDS) analysis using read abundance-based on Bray-Curtis coefficients. The NMDS analysis was conducted using metaMDS commands in the R package Vegan [53] and visualized in RStudio using ggplot2 [54]. An analysis of similarity (ANOSIM; [55]) was used to test for significance. When a significant difference was detected, we performed the similarity percentage analysis (SIMPER) in Vegan to determine which taxa were responsible for explaining most of the difference among groups. Finally, to visualize the number of common and unique species detected across primer sets, a Venn diagram was constructed using the R package VennDiagram v1.6.2 [56].

3 Results

3.1 Illumina sequencing

A total of 2.1 and 2.0 million paired-end sequence reads were generated from the 2 × 150 bp (Riaz_12S and Valentini_12S) and the 2 × 250 bp (MiFish_12S, Berry_16S, and MacDonald_18S) Illumina MiSeq runs, respectively. Across both runs ~ 92% of the paired-end reads had phred scores of ≥ 30 and in total 63.19% of the reads were retained after quality filtering. The percentage of reads retained after quality control was 71.02% for MiFish_12S, 63.74% for Riaz_12S, 60.10% for Valentini_12S, 78.74% for Berry_16S, and 72.25% for MacDonald_12S with the number of total reads retained for each primer set ranging from 295,277 to 711,351 (Table 2). The average number of reads per sample was the highest for Berry_16S (57,729 ± 14,331 reads) and lowest for MiFish_12S (33,711 ± 13,984) while the Riaz_12S and Valentini_12S primers resulted in 51,434 ± 13,944 and 48,522 ± 13,511 reads, respectively.

Table 2

Illumina sequencing results.

# Sequence reads	MiFish_12S	Riaz_12S	Valentini_12S	Berry_16S	MacDonald_18S
After quality filtering	497024	637771	676856	711351	295277
% assigned to fish
Class	94.12	99.99	94.49	99.95	0.094
Order	92.05	99.99	94.45	99.95	0.046
Family	84.88	98.31	88.51	98.11	0.00
Genus	84.88	98.31	88.51	98.11	0.00
Species	81.39	96.78	86.03	97.38	0.00
Non-target	5.88	0.005	4.75	0.055	99.99

Listed is the total number of reads retained after quality control and the percentage of those reads assigned to target (fishes) and non-target taxa (non-fish).

There was no indication of contamination in any of the extraction or PCR negative controls (samples did not amplify in PCR 1). However, two out of the six field negative controls did amplify and so these were included in library preparation and sequencing. Sequencing these negative field controls resulted in just 5,206 reads, 99.23% (5,166 reads) of which assigned to human DNA for Valentini_12S primers. Of the remaining 40 reads, 10 were assigned to Lutjanus spp. and 28 reads were assigned to Lutjanus griseus for the Riaz_12S primers, while the other two reads were assigned to Mugil curema for the Valentini_12S primers. No contaminating sequences were detected using the other three primers sets (MiFish_12S, Berry_16S, MacDonald_18S).

The MacDonald_18S primer set amplified taxa from 24 different phyla including Porifera, Cnidaria, Platyhelminthes, Nematoda, Mollusca, Annelida, Arthropoda, Echinodermata, and Chordata. Copepods (phylum Arthropoda) accounted for 28.57% of reads while green algae (phylum Chlorophyta) accounted for 24.87% of reads. Since only 278 reads or 0.09% of the total reads that resulted from the MacDonald_18S primer set were assigned to fishes (and none could be resolved to species level), we did not include data of this marker in subsequent analyses. Across the remaining four primer sets, the proportion of reads passing quality filters that were assigned to fish taxa were 94.12% for MiFish_12S, 99.99% for Riaz_12S, 94.49% for Valentini_12S, and 99.95% for Berry_16S (Table 2, Fig 1).

Fig 1

Proportion of reads assigned to each taxonomic level for the four metabarcoding primer sets tested here.

3.2 Taxonomic assignments and biodiversity estimates

While adjusting the sequence similarity threshold did change the number of species detected for each primer set, the interpretation of the overall pattern did not change. Regardless of which similarity threshold was applied the Riaz_12S primer set resulted in the most species detected with Berry-16S ranking second (S2 Table). There were no notable changes in taxa detected when comparing the 97% and 98% thresholds but there was predictably an increase in the number of species resolved if we applied the 98% threshold compared to the more conservative 99% threshold. Most notably, when we compared the best performing Riaz_12S using the 99% cutoff with the worst performing primers sets using a 97% cutoff the latter still resulted in a higher number of species detections (S2 Table). Similar patterns were found for the diversity indices (S3 Table). As a result, here we apply a single conservative similarity threshold of 99% for species-level designations and report these throughout.

Across the four primer sets, we detected 76 species of fish in two classes, 17 orders, 48 families, and 67 genera (Table 3). Although the vast majority of reads were assigned to class Actinopterygii (>99%; bony fish), we also detected sharks and rays with each primer set (class Chondrichthyes; Table 4), but with relatively low read counts that ranged from 7 to 2,581 reads. We detected the highest number of taxa (across all taxonomic levels) with the Riaz_12S primer set but the highest number of ASVs with the Valentini_12S primers (Table 3). We resolved a similar number of fish species with the Riaz_12S and Berry_16S primer sets (55 ± 5.59 and 49 ± 6.69, respectively). The MiFish_12S and Valentini_12S primers sets also resolved similar numbers of species (34 ± 3.13 and 32 ± 4.72, respectively) but significantly fewer than the Riaz_12S and Berry_16S primer sets (Table 5; Tukey-Kramer HSD P < 0.05). Estimates of evenness were low and were significantly different among primer sets (Table 5). Our estimates of Shannon’s diversity ranged from 1.01 ± 0.45 for MiFish_12S to 1.75 ± 0.47 for Riaz_12S. Similar to species richness, Shannon’s diversity values were roughly equal for the MiFish_12S (1.01 ± 0.45) and Valentini_12S (1.09 ± 0.35) primer sets and were not significantly different (Table 5). However, Shannon’s diversity values for the Riaz_12S primer set were significantly higher than the MiFish_12S primer set (Tukey-Kramer HSD, P = 0.049; Table 5).

Table 3

Number of fish taxa detected for each of the four metabarcoding primer sets using the following similarity thresholds: 99% for species; 97% for genus; 95% for family; 90% for order; 85% for class; and 80% for phylum.

Taxonomic rank	MiFish_12S	Riaz_12S	Valentini_12S	Berry_16S	Total
Class	2	2	2	2	2
Order	14	16	12	12	17
Family	28	37	24	29	48
Genus	38	53	34	43	67
Species	34	55	32	49	76
ASVs	161	159	247	140	-

Classes include Actinopterygii and Chondrichthyes.

Table 4

Species of cartilaginous fishes detected with each of the four metabarcoding primer sets tested in this study.

Class	Subclass	Primer	Species
Chondrichthyes	Elasmobranchii	MiFish_12S	Dasyatis say
		Riaz_12S	Aetobatus narinari Dasyatis sabina
			Dasyatis say
			Gymnura micruraHypanus sabinusRhinoptera spp.
		Valentini_12S	Aetobatus narinariDasyatis sayHypanus americanusRhinoptera spp.
		Berry_16S	Hypanus americanus

Table 5

Diversity indices including species richness, evenness, and Shannon’s diversity calculated using the R package BiodiversityR v. 2.12–3 for each of the four metabarcoding primer sets.

Primer set	Species richness	Evenness	Shannon’s diversity
MiFish_12S	34 ± 3.13 a,c	0.134 ± 0.054 a	1.01 ± 0.45 a
Riaz_12S	55 ± 5.59 b	0.163 ± 0.067 a	1.75 ± 0.47 b
Valentini_12S	32 ± 4.72 a	0.153 ± 0.054 a	1.09 ± 0.35 a,b
Berry_16S	49 ± 6.69 b,c	0.138 ± 0.064 a	1.30 ± 0.52 a,b

Standard deviations are based on the data from all six sample sites. Those comparisons that were significant using an ANOVA show superscripts with different letters indicating statistical difference at P < 0.05 using the post-hoc Tukey-Kramer Honest Significant Difference test.

Of the 76 species of fish detected in this study (S1 Fig), only 17 were common to all datasets with between 1–12 unique species detected by any single primer set (Fig 2). The numbers of reads assigned to the 17 common taxa were 514,041, 555,883, and 569,976 for the Riaz_12S, Valentini_12S, and Berry_16S primer sets, respectively (S4 Table). The MiFish_12S had the lowest number of reads across these 17 species (381,805) and accounted for just 16% of total read counts across all datasets (S5 Table) suggesting lower PCR efficiency. We resolved a high number of unique species using the Riaz_12S and Berry_16S primer sets (12 and 11 species, respectively), while only three unique species were detected using the MiFish_12S primers and only one was detected using the Valentini_12S primers. Based on the ANOSIM analyses, species assemblages differed significantly across primer sets (R = 0.4011, P < 0.001); a finding that was supported by the NMDS plots which showed clear separation among marker sets when either the read abundance data (Fig 3) or presence/absence data (S2 Fig) were analyzed. However, no significant differences were observed in the fish communities across the different sampling sites (ANOVA, F = 1.25, P = 0.33; Fig 3). The five most influential species contributing to these differences, based on SIMPER analyses, are given in Table 6 with the White mullet Mugil curema ranking first in all comparisons. This species was also the most dominant in terms of read count across primer sets.

Fig 2

Venn diagram representing the number of fish species detected across four metabarcoding primer sets.

The numbers shown in the areas of overlap reflect shared species.

Fig 3

Nonmetric multidimensional scaling plots (NMDS) based on read abundance for each of the species detected from six sampling locations in the Indian River Lagoon, Florida.

The numbers inside the circles represent sample sites. Metabarcoding data was generated for four primer sets designed to amplify fishes.

Table 6

Results of SIMPER analysis conducted using the R package Vegan.

Taxon	Average dissimilarity	% contribution	% cumulative	Overall average dissimilarity
Riaz_12S vs Berry_16S				43.26
Mugil curema	15.04	34.76	34.76
Leiostomus xanthurus	4.197	9.7	44.46
Bairdiella chrysoura	3.723	8.606	53.07
Pogonias cromis	3.026	6.994	60.06
Lagodon rhomboides	2.735	6.321	66.38
Riaz_12S vs MiFish_12S				49.51
Mugil curema	17.67	35.68	35.68
Bairdiella chrysoura	7.286	14.71	50.4
Pogonias cromis	4.15	8.382	58.78
Lagodon rhomboides	3.429	6.927	65.7
Ariopsis felis	2.824	5.704	71.41
Riaz_12S vs Valentini_12S				43.87
Mugil curema	15.77	35.95	35.95
Bairdiella chrysoura	6.508	14.84	50.78
Lagodon rhomboides	3.632	8.281	59.06
Pogonias cromis	3.394	7.738	66.8
Ariopsis felis	2.364	5.389	72.19
Berry_16S vs MiFish_12S				45.68
Mugil curema	20.79	45.51	45.51
Bairdiella chrysoura	5.267	11.53	57.04
Leiostomus xanthurus	4.125	9.031	66.07
Lagodon rhomboides	3.707	8.115	74.19
Mugio cephalus	1.392	3.048	77.23
Berry_16S vs Valentini_12S				34.12
Mugil curema	12.77	37.41	37.41
Bairdiella chrysoura	4.819	14.12	51.53
Leiostomus xanthurus	3.497	10.25	61.78
Lagodon rhomboides	3.185	9.335	71.12
Lutjanus griseus	1.098	3.218	74.33
MiFish_12S vs Valentini_12S				43.05
Mugil curema	21.11	49.03	49.03
Bairdiella chrysoura	8.486	19.71	68.74
Lagodon rhomboides	6.283	14.59	83.33
Mugil cephalus	1.528	3.549	86.88
Leiostomus xanthurus	1.296	3.009	89.89

The five species that contributed most to the dissimilarity among metabarcoding primer sets are listed.

4 Discussion

When designing eDNA studies, choosing metabarcoding primers is of critical importance as the initial monetary investment can be high and the decision will have a significant impact on project results [29,42]. Primers with insufficient taxonomic specificity can result in the loss of sequencing effort to non-target taxa as well as false negatives. This is particularly true in highly diverse study systems where non-target DNA (i.e., microbial and plankton communities) is abundant [29]. Despite the growing interest in the use of eDNA to assess fish communities, there has been a surprising lack of studies that have directly compared the efficacy of metabarcoding primer sets in marine and estuarine systems. Because the performance of eDNA metabarcoding primers will vary depending on the study system (freshwater, marine, or estuarine) and taxonomic composition, there is no guarantee that a primer set that performs well in a freshwater system will do so in marine or estuarine systems. For this reason, we included three primer sets that performed well in the freshwater study of Zhang et al. [42], to determine their effectiveness in a biodiverse estuarine system. Our study system, the Indian River Lagoon, which runs along Florida’s east coast is regarded as the most species rich estuary in the U.S. [57]. Our results show that the 12S rRNA primers of Riaz et al. [58] were the most taxon-specific with 99.99% of the resulting reads assigned to fish and 96.78% of reads assigned to the species-level. The 16S rRNA primers of Berry et al. [31] also performed well with 99.95% of reads assigned to fish and 97.38% assigned to species. Furthermore, a similar number of species were identified using these primer sets (Table 3) and both resulted in comparatively high Shannon’s diversity values (Table 5). The popular 12S primers of Miya et al. [59] did well in terms of the percent reads assigned to fish taxa (94.12%) and species (81.29%), but only 34 species were detected using this primer set, performing similarly to the Valentini_12S primer set. Taken together, these results indicate that the Riaz_12S and Berry_16S primer sets performed best in our biodiverse estuarine system and resolved the greatest number of target species. Moreover, of the 76 fish species identified in this study, 85.5% (65 of 76) were detected when we combined the three 12S primer sets while combining the Riaz_12S and Berry_16S primers detected 93.4% (71 of 76). This finding supports the importance of not only employing multiple primer sets but also using primers that target different genes.

The rate of evolution varies across genes and gene regions and therefore applying a single set of taxonomic thresholds (i.e., 99% sequence similarity for species-level designations) can be misleading. However, our analyses of the sequencing results from the four primer sets at three different species-level similarity thresholds (99%, 98%, and 97%) did not change our interpretation of the data. The Riaz_12S primer set either did as well or outperformed the others in terms of specificity (99.99% of reads assigned to fishes; Table 2), universality (greatest number of taxa amplified; S2 Table), and taxonomic resolution (96.78% of reads assigned to the species-level at the 99% threshold) regardless of which cutoff value was applied. Most notably, this was true even when comparing Riaz_12S using the conservative 99% cutoff against all other primers sets using a 97% cut off (S2 Table). Similar patterns were found for diversity indices (S3 Table).

Despite the fact that the 18S primer set of MacDonald et al. [44] was designed to specifically target fish and a high annealing temperature (62°C) was employed, less than 1% of the reads generated using these primers assigned to fishes and none could be assigned to the species level. Instead, >99% of reads were assigned to non-target taxa across a diversity of groups particularly copepods and green algae. This can likely be explained by the fact that these primers were designed based on the alignment of 18S gene sequences from only ten species of freshwater fish, and while they performed well in the original experiment where primer testing was performed on tissues from known fish species, they are not suitable for metabarcoding.

4.1 Detection of sharks and rays

All four primer sets tested here amplified DNA from both bony (class Actinopterygii) and cartilaginous fishes (class Chondrichthyes) with the vast majority of the reads (> 99%) assigned to the former (Table 4). Of the seven cartilaginous fish species detected, six were identified by Riaz_12S and four were detected by Valentini_12S. The MiFish_12S and Berry_16S detected only one species each. These results suggest that Riaz_12S primer set may also be suitable for the detection of sharks and rays. However, if elasmobranchs are the primary target group, it would be prudent to test primers using DNA extractions from species expected in the study area. Furthermore, there are a number of published primers that have been designed specifically for elasmobranchs that are worth exploring [12,60].

4.2 Reference databases and taxonomic assignments

Another factor that must be considered when choosing candidate primer sets for metabarcoding is the completeness of the reference databases [61,62]. For instance, while the COI reference databases for animals are robust, designing COI primers that are taxon specific is problematic, so primers are designed with high levels of degeneracy. This leads to the amplification of non-target taxa that can account for a large proportion of reads [28,29]. Furthermore, specific loci are often favored for some taxonomic groups. For example, 16S rRNA is most often employed to characterize bacterial communities, ITS is often used for fungi [63], and 18S rRNA is commonly employed for zooplankton [64]. For fishes, primers that amplify a portion of the 12S and 16S rRNA genes seem to provide a compromise between universality and specificity and are commonly used [65-67]. Incomplete reference databases also pose obstacles to taxonomic assignments [62,68]. Missing sequences can lead to misidentifications or the collapsing of assignments to genus or higher levels of classification. Over 20 years ago, when very few 16S rRNA bacterial sequences were publicly available, a 97% sequence similarity threshold was proposed for species-level assignments [69]. However, as publicly available sequence data increased exponentially, sequence similarity thresholds for species-level assignments have also increased and now typically range from 97–100% depending upon the target taxa and locus employed [27,35,70-74]. While there is no clear consensus on threshold criteria, recent studies have suggested that a 99–100% similarity threshold may be most appropriate for species-level assignments using 12S [73] and 16S rRNA markers [75].

4.3 Annealing temperature, human contamination, and other cautionary notes

Primer annealing temperature is an important factor in determining PCR success and specificity. At lower temperatures, just a partial match between primer and template can be sufficient to permit amplification. On the other hand, higher annealing temperatures require exact or nearly exact primer-template match and usually results in high specificity. The MiFish_12S primer set has been employed in several published studies with annealing temperatures ranging from 50°C to 65°C [16,36,42,59,76-78]. Initially, we tested the MiFish_12S primer set using an annealing temperature of 55°C based on Andruszkiewicz et al. [76]. Surprisingly, none of the resulting reads assigned to fish but instead were assigned to bacteria. Following the optimized protocol of Bylemans et al. [16], we raised the annealing temperature to 61.5°C for subsequent experiments which resulted in 94.12% of reads assigning to fish. Our initial results could have been exacerbated by the high bacterial loads in our estuarine samples as DNA primers will preferentially amplify abundant templates and sometimes fail to amplify low abundant target DNA [17]. However, a similarly high percentage of non-target reads was observed by Miya et al. [79] for the MiFish_12S primers employed in marine waters where target DNA was also scarce. This example highlights the utility of testing primers on a small number of samples prior to purchasing large primer sets and/or the bulk processing of samples.

Human DNA was detected in all of our sequencing libraries except those produced using the Berry_16S primer set. Because human DNA was amplified in our field negative controls but not our lab controls (extraction and PCR negative controls), contamination was most likely introduced during sample collection or filtering. The presence of human DNA in eDNA metabarcoding studies is common [59,80,81]. To alleviate this problem, human-specific blocking primers have been used [82,83], however, the use of blocking oligos has been shown to reduce the number of target species detected in metabarcoding studies [42]. Moreover, only a small percentage (1.48%) of our reads were assigned to humans. Therefore, the use of blocking primers is not advised if contamination levels are likely to be low.

Many studies show a positive correlation between animal abundance and/or biomass estimates and the number of reads obtained from eDNA metabarcoding studies (reviewed by [84]). However, the inconsistency across studies, our insufficient understanding of DNA shedding rates, and the current paucity of information on how biotic and abiotic factors influence eDNA detection rates, limits the utility of eDNA for estimating species biomass and abundance [85,86]. These issues need to be more fully addressed before the relationship between eDNA copy number and abundance can be accurately modelled. As a result, we use read abundance as a proxy for species abundance for our Shannon’s diversity value calculations but do so with great caution and refrain from over interpreting the results herein.

5 Conclusions and recommendations

The published data, including this study, demonstrate that for most aquatic systems no single primer set can capture all the diversity of any given community [10,87]. However, employing multiple primer sets may be cost prohibitive for some laboratories. In those cases where fish are the target the Riaz_12S primer set is a good option to consider. In our dataset from the biodiverse Indian River Lagoon in Florida, 99.9% of reads generated using this primer set assigned to fish and resulted in the greatest number of species detected. This is contrary to what seems to be a settling of some segments of the eDNA community on the MiFish_12S primers as the standard [88], which in our study did not perform as well. It should be noted that our results are based on limited sampling in a single, albeit highly diverse, estuary. Because fish communities vary significantly across space and time our findings may not be directly transferrable across systems. However, our results do add to the growing literature concerning eDNA protocols and best practices and will aid in the narrowing of primer choices for other researchers. Furthermore, our study highlights the importance of targeting different barcoding genes when possible. When we combined the results of the three 12S makers 85.5% species in the dataset were accounted for, whereas 93.4% of the species were detected by combining the Riaz_12S and Berry_16S primer sets. The failings of the MacDonald_18S primer set to identify target taxa and the mis-priming of the MiFish_12S primer set due to low annealing temperature highlights the importance of methods testing and optimization before large investments are made in any particular protocol.

Number of reads (log transformed) for each of the 17 fish species detected by all four primer sets (MiFish_12S, Riaz_12S, Valentini_12S, and Berry_16S).

Read numbers are totals across the six sample sites.

(TIF)

Click here for additional data file.

Nonmetric multidimensional scaling plots (NMDS) using presence/absence data for the fish species detected at six sampling locations in the Indian River Lagoon, Florida.

The numbers inside the circles represent sample sites. Metabarcoding data was generated using four primer sets designed to amplify fishes.

(TIF)

Click here for additional data file.

Sampling sites used in this study. Two replicates of 500 ml water samples were collected in June 2018 from six locations within the Indian River Lagoon in central Florida.

Water samples were taken at the surface over shallow water habitat known to have oyster reefs.

(DOCX)

Click here for additional data file.

Total number of fish species detected for each primer set at three different species-level sequence similarity thresholds (99%, 98%, and 97%).

Also shown is the percentage of total reads assigned at the species-level (fishes) after quality control.

(DOCX)

Click here for additional data file.

Diversity indices including species richness, evenness, and Shannon’s diversity calculated using the R package BiodiversityR v. 2.12–3 for four metabarcoding primer sets designed to amply fishes.

Standard deviations were calculated based on the data across all six sample sites. Data for three different species-level sequence similarity thresholds (99%, 98%, and 97%) are shown.

(DOCX)

Click here for additional data file.

The number of reads obtained for each of the 17 fish taxa detected by all four primer sets used in this study.

Read numbers are totals across all the six sample sites.

(DOCX)

Click here for additional data file.

Number of reads that were assigned to fish species at the 99% sequence similarity threshold for each primer set.

Total number of reads across all markers and the % of those total reads attributed to each marker is shown. Also listed is the average number of reads per species with the range in parenthesis.

(DOCX)

Click here for additional data file.

4 Aug 2021

PONE-D-21-15188

Choosing the best eDNA metabarcoding primer set for assessing fish communities in a biodiverse estuarine system

PLOS ONE

Dear Dr. Kumar,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

The manuscript represents an exploration of Fish diversity in an interesting ecosystem. The reviewers highlight several major issues that need to be addressed.

The first one may be with the angle of the writing, should this focus on exploring diversity or compare methods to explore the diversity. As the reviewers highlight, then this (in current form) falls short of doing a thorough comparison of the utility of the methods. E.g. rev.1 stresses that comparison of those methods should contrast – 1. specificity, 2. universality, and 3. resolution, but that the ms in current form only tackles 1. You may choose to add in silioco work and tackle 2 and 3, or reorient the paper towards using the different methods to tackle you study system(s) (and tone down conclusions of general lessons on methods).

The reviewers also call for beefing up of the analyses, most importantly dropping the assumption that read-covereage correlates with abundance (see rev. 3)

Also add information on sampling and background biology in paper and abstract (rev 2).

Finally, add intermediate datatables as supplemental data or cite open repositories where they can be accessed (data.dryad.org, etc).

Please submit your revised manuscript by Sep 18 2021 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see:  http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols . Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at  https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols .

We look forward to receiving your revised manuscript.

Kind regards,

Arnar Palsson, Ph.D.

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1.Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

Additional Editor Comments (if provided):

The manuscript represents an exploration of Fish diversity in an interesting ecosystem. The reviewers highlight several major issues that need to be addressed.

The first one may be with the angle of the writing, should this focus on exploring diversity or compare methods to explore the diversity. As the reviewers highlight, then this (in current form) falls short of doing a thorough comparison of the utility of the methods. E.g. rev.1 stresses that comparison of those methods should contrast – 1. specificity, 2. universality, and 3. resolution, but that the ms in current form only tackles 1. You may choose to add in silioco work and tackle 2 and 3, or reorient the paper towards using the different methods to tackle you study system(s) (and tone down conclusions of general lessons on methods).

The reviewers also call for beefing up of the analyses, most importantly dropping the assumption that read-covereage correlates with abundance (see rev. 3)

Also add information on sampling and background biology in paper and abstract (rev 2).

Finally, add intermediate datatables as supplemental data or cite open repositories where they can be accessed (data.dryad.org, etc).

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: No

Reviewer #2: Yes

Reviewer #3: Partly

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: No

Reviewer #2: Yes

Reviewer #3: No

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: No

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: I have reviewed the manuscript by Kumar et al. entitled "Choosing the best eDNA metabarcoding primer set for assessing fish communities in a biodiverse estuarine system". The authors compared the usefulness of four primer sets for eDNA metabarcoding for fish at a model field of a lagoon in Florida. The four primer sites included three 12S (MiFish_12S, Valentini_12S, and Riaz_12S) and one 16S (Berry_16S) primers. The authors claimed that Riaz_12S and Berry_16S primer sets "detected" more numbers of species than MiFish_12S and Valentini_12S. Such comparison among metabarcoding primers is important for selection of assay systems, and in this sense, the manuscript has a merit to be published.

However, I have serious concerns on the methods for comparing primers. The usefulness of primers for eDNA metabarcoding should be evaluated the following three aspects, specificity (specific amplification of target taxa), universality (comprehensive amplification of target taxa without bias), and resolution (taxonomic resolution in the amplified region). The manuscript only evaluated primers based on the number of "detected" species. The criteria for species identification in this manuscript is ≧99% identity with sequences on the database, but this criteria is not fair. For example, the length of amplified region for Valentini_12S is short, and one base substitution make it genus-level identification according to the authors' criteria. Also, for MiFish_12S, many previous papers using this primer set adopt the identification threshold at ≧98.5% or lower, and from my inspection of these papers, I believe that the criteria of ≧99% identity is inappropriate. I recommend authors to set an appropriate threshold for each primer set, considering the difference of evolutionary speed among amplified regions. In addition, the accuracy of species determination may vary depending on the richness of the data in the database, so the results may vary greatly depending on the criteria set by the authors.

The above mentioned specificity, universality, and resolution can be tested via in silico test as well as using the data for field samples. Without such tests, the comparison among primer sets cannot be justified. From above reasons, I think the manuscript should be largely revised before its publication.

Reviewer #2: The manuscript submitted by Kumar et al. is a simple but well written piece focused on comparing different primer sets for fish eDNA metabarcoding. I would recommend the authors to improve the abstract and methods sections by including more necessary information, especially, regarding the sampling and studied area. For example, it is not very clear were exactly the samples were taken and if brackish or freshwater species are also expected to be recovered from those samples.

The abstract still needs to be modified by clarifying some important information (e.g., target taxonomic group, samples analysed, studied system, etc)

L27: Please include the referred ‘efforts’

L29: Include the target taxonomic group

L57: remove “a” in “a marine environments”

L87-89: The authors should better explore this information. The public databases might be more complete for some very few specific taxonomic groups. Therefore, when providing this information the target group might be included.

L114: Please include more information for the sampling sites

L166: How many runs?

L192-193: It is great to see the authors have adopted a conservative approach removing ASVs with query coverage below 100%.

L288: Insert a space between ‘and’’16S’

It is also important to note and discuss the lack of appropriate reference sequences. Despite being known that the combination of multiple primers targeting distinct gene is expected to increase taxonomic resolution and consequently, the number of species detected, the lack of references remains as a hindrance.

Reviewer #3: The authors performed a comparative study, aimed at evaluating the performances of different primer sets specifically designed to target fish species, in a highly biodiverse system: the Indian River Lagoon in Florida. They tested three 12S and one 16S rRNA primer sets, and a 18S primer set designed for freshwater fish that did not produce useful results. I found the manuscript clear, and I think its findings could be helpful for choosing the best primer sets for studying fish diversity using metabarcoding. However, I think that authors could make a bigger effort to present the differences among the tested primer sets in a more appropriate way. The figures also need to be supported by more precise legends.

Please, see below my comments:

Line 53-54: Here you cite some works aimed at describing environmental DNA metabarcoding in general. That is fine, except maybe for the paper of Kumar, Eble and Gaither (2020), that I find a little out of topic. I think that you should rather cite some empirical studies supporting your sentence. For example, the recent work of Aglieri et al. (2020) compared simultaneously three traditional sampling methods with environmental DNA metabarcoding, finding complementarity among the methods.

Line 98: duplicated “in”

Line 209: Except for species richness, the diversity indexes you used need abundance data. You used reads numbers as a proxy for abundance, but I am not so much a supporter of this approach. PCR biases can alter the proportion between the amount of initial DNA template and the final sequences yield. This is even more worthy of consideration when two PCRs are used to build the sequencing libraries, as you did. Moreover, the amount of genetic material in the water is not necessarily related to the abundance of individuals: different species have different dimensions, different shedding, etc. Several authors have made the attempt of relating sequence abundance to individual abundance, but the relation is very often poor. That is a current limit of metabarcoding, and at least you should explain that. Please, add a few lines to address this point.

Line 215: As in the previous comment. Bray-Curtis uses abundances. I would like to see also something made using presence/absence data, at least to verify the differences with the abundance related indexes you used. For example, you could show the same nMDS, but made with the Sørensen index. Please, show something more.

Line 216-217 and Figure 3: The nMDS plot shows six points for each primer set. I don’t get what they represent, since you sampled three sites taking four replicates for each one. The plot should show three points, in the case you pooled the replicates for each site, or 12 points if you were showing the diversity for each replicate. Please explain better what the points represent.

Line 222: Since you decided to consider the abundance of reads to calculate diversity indexes, why don’t you also show the number of reads of the taxa in common among the different primer sets? This could be indicative of the amplification efficiency of each primer set for those taxa. This is just a suggestion, but I would like to understand more.

Line 232: Please, could you provide also the average number of reads for sample?

Line 261-262: Low evenness means big differences in reads number among taxa. This is quite common in metabarcoding studies, since several variables can influence the yield, resulting in high variability. Anyway, how different these yields are for each taxon? Please, could you provide something more? For instance, a plot with all the reads of each taxon ordered from the smaller to the higher. Alternatives to my suggestion are also welcome.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

Reviewer #3: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

6 Oct 2021

Response to reviewer’s comments

Reviewer #1

Comment: I have serious concerns on the methods for comparing primers. The usefulness of primers for eDNA metabarcoding should be evaluated the following three aspects, specificity (specific amplification of target taxa), universality (comprehensive amplification of target taxa without bias), and resolution (taxonomic resolution in the amplified region). The manuscript only evaluated primers based on the number of "detected" species.

Response: We agree that all three criteria are important and have clarified our goals in the introduction.

“Here we focus on the four best performing primer sets from the freshwater study of Zhang, Zhao (42) and include an additional 18S primer set not previously evaluated (but designed for fishes), and test these in a species rich estuarine system in Florida (the Indian River Lagoon). We compare the efficiency of these primers in terms of 1) specificity (amplification of fishes at the exclusion of other taxa), 2) universality (amplification of a diversity of fish taxa), and 3) taxonomic resolution (ability to resolve to the species level)”.

By this definition we do address specificity. Our target group is fishes. In Table 2 we show the percent of reads that were off target (non-fish). For all primer sets at least 94% of reads assigned to fishes with the exception of the MacDonald primers. We also address universality in terms of the taxonomic levels that were amplified. We can’t directly address the question of sequencing bias with this dataset and therefore we do not attempt to translate read number into relative abundances, but we do include ASV files with read counts and their corresponding sequences. These files have been uploaded to Data Dryad (DOI: https://doi.org/10.5061/dryad.70rxwdbzc). In Table 3 we address the universality issue by highlighting the number of fish taxa (from class to species level) that were amplified, and Fig. 2 shows a Venn diagram of overlapping detections.

Comment: The criteria for species identification in this manuscript is ≧99% identity with sequences on the database, but this criteria is not fair. For example, the length of amplified region for Valentini_12S is short, and one base substitution make it genus-level identification according to the authors' criteria.

Response: We agree that some primer sets are better candidates for species delimitation. Our goal with this experiment was to determine which primer set allows us to identify the greatest number of fish species). Some gene segments will be better as they are more variable and able to delineate taxa more clearly. This is a function of length but also how the gene evolves and probably of more importance how complete the public databases are. Our goal was not to compare gene regions or consider fragment length, but instead we considered each of these primer sets to be a tool with its assigned task being to identify species in a water sample. Therefore, we evaluated each primer set as a published and established tool (including the public databases) and our goal was to determine which is the best for characterizing fish diversity in our estuarine system.

Comment: Also, for MiFish_12S, many previous papers using this primer set adopt the identification threshold at ≧98.5% or lower, and from my inspection of these papers, I believe that the criteria of ≧99% identity is inappropriate. I recommend authors to set an appropriate threshold for each primer set, considering the difference of evolutionary speed among amplified regions. In addition, the accuracy of species determination may vary depending on the richness of the data in the database, so the results may vary greatly depending on the criteria set by the authors.

Response (L336-344): Regarding the threshold for the species identification, we used the following criteria for the taxonomic assignment: ≥ 99% similarity for species; 97% for genus; 95% for family; 90% for order; 85% for class; and 80% for phylum following West et al. 2020. The commonly used 97% sequence similarity threshold was proposed for species assignment in 1994 when only few 16S rRNA sequences were available in databases. With increase in the richness of sequences in databases over the years, a variety of sequence similarity threshold (97–100%) have been used in metabarcoding studies using different gene markers. Recent studies have suggested that a 99–100% similarity threshold may be more appropriate for species assignment using 12S (72) and 16S rRNA markers (74). Following these guidelines, we used ≥ 99% similarity cutoff for species assignment.

Comment: The above-mentioned specificity, universality, and resolution can be tested via in silico test as well as using the data for field samples. Without such tests, the comparison among primer sets cannot be justified. From above reasons, I think the manuscript should be largely revised before its publication.

Response (L98-104): In-silico testing of the primer sets used in our study (except McDonald_18S) was performed in Zhang et al. (2020). They found considerable discrepancies between the in-silico and in-vitro results including in the range and diversity of taxa amplified, fish community composition, and the power of each to discriminate species. Since the in-silico experiments had already been conducted (and the results were not always in agreement with in-vitro results), we went directly to in-vitro experiments testing the primers on field samples to evaluate their efficacy for fish biodiversity assessment. The explanation for this comment has been given in lines 98-104.

Reviewer #2

Comment: The manuscript submitted by Kumar et al. is a simple but well written piece focused on comparing different primer sets for fish eDNA metabarcoding. I would recommend the authors to improve the abstract and methods sections by including more necessary information, especially, regarding the sampling and studied area. For example, it is not very clear were exactly the samples were taken and if brackish or freshwater species are also expected to be recovered from those samples.

The abstract still needs to be modified by clarifying some important information (e.g., target taxonomic group, samples analysed, studied system, etc)

Response: We thank the reviewer for pointing out these omissions. We have now included more details about our sample sites, target taxa, etc. in both the abstract and methods sections. Also, a table with GPS coordinates are now including in supplemental materials (Table S1).

Comment: L27: Please include the referred ‘efforts’

Response (L26): Suggestion has been incorporated. Text now reads “The efficiency and outcome of these metabarcoding efforts are dependent upon…”

Comment: L29: Include the target taxonomic group

Response (L29): The target taxonomic group was fishes. The same has been added in line 29.

Comment: L57: remove “a” in “a marine environments”

Response (L60): Corrected

Comment: L87-89: The authors should better explore this information. The public databases might be more complete for some very few specific taxonomic groups. Therefore, when providing this information, the target group might be included.

Response (L88-89): We completely agree with the reviewer’s point of view. The markers used in this study are well represented for fishes in the public databases. The clarification has been given in line 89.

Comment: L114: Please include more information for the sampling sites.

Response: The details of sample information are now provided in the methods section and Table S1.

Comment: L166: How many runs?

Response (L176-178): The methods section now reads “Sequencing was conducted using a Nano 300 v2 (2 × 150) Reagent Kit for 2×111 cycles and a Nano 500 v2 (2 × 250) Reagent Kit for 2×251 cycles, based on amplicon size.”

Comment: L288: Insert a space between ‘and’’16S’

Response: Corrected

Comment: It is also important to note and discuss the lack of appropriate reference sequences. Despite being known that the combination of multiple primers targeting distinct gene is expected to increase taxonomic resolution and consequently, the number of species detected, the lack of references remains as a hindrance.

Response (L325-344): We completely agree that the effectiveness of eDNA metabarcoding studies are highly dependent on the completeness and accuracy of the relevant reference databases. We have discussed this issue in lines 325-344.

Reviewer #3:

Comment: The authors performed a comparative study, aimed at evaluating the performances of different primer sets specifically designed to target fish species, in a highly biodiverse system: the Indian River Lagoon in Florida. They tested three 12S and one 16S rRNA primer sets, and a 18S primer set designed for freshwater fish that did not produce useful results. I found the manuscript clear, and I think its findings could be helpful for choosing the best primer sets for studying fish diversity using metabarcoding. However, I think that authors could make a bigger effort to present the differences among the tested primer sets in a more appropriate way. The figures also need to be supported by more precise legends.

Response: We thanks the reviewer for their thoughtful comments. We have improved the manuscript by adding more detailed information on sampling and plotting nMDS using both read abundance data as well as presence/absence data. The figure legends have been elaborated to include all the necessary information.

Comment: Line 53-54: Here you cite some works aimed at describing environmental DNA metabarcoding in general. That is fine, except maybe for the paper of Kumar, Eble and Gaither (2020), that I find a little out of topic. I think that you should rather cite some empirical studies supporting your sentence. For example, the recent work of Aglieri et al. (2020) compared simultaneously three traditional sampling methods with environmental DNA metabarcoding, finding complementarity among the methods.

Response: We have removed the review papers here and added a few empirical studies that better support this statement including Aglieri et al. (2020).

Comment: Line 98: duplicated “in”

Response: Corrected

Comment: Line 209: Except for species richness, the diversity indexes you used need abundance data. You used to read numbers as a proxy for abundance, but I am not so much a supporter of this approach. PCR biases can alter the proportion between the amount of initial DNA template and the final sequences yield. This is even more worthy of consideration when two PCRs are used to build the sequencing libraries, as you did. Moreover, the amount of genetic material in the water is not necessarily related to the abundance of individuals: different species have different dimensions, different shedding, etc. Several authors have made the attempt of relating sequence abundance to individual abundance, but the relation is very often poor. That is a current limit of metabarcoding, and at least you should explain that. Please, add a few lines to address this point.

Response (L345-353): We largely agree with the reviewer here. Even though several studies have shown positive correlations between abundance estimates from eDNA metabarcoding data (reviewed by Rourke et al. 2020), the correlation is often weak. We have address this comment in lines 345-353.

Comment: Line 215: As in the previous comment. Bray-Curtis uses abundances. I would like to see also something made using presence/absence data, at least to verify the differences with the abundance related indexes you used. For example, you could show the same nMDS, but made with the Sørensen index. Please, show something more.

Response: As per the suggestion, we plotted the nMDS using abundance data as well as presence/absence data. Both the data set showed clear separation among primer sets. The nMDS plot using presence/absence data has been presented as a supplementary figure S2.

Comment: Line 216-217 and Figure 3: The nMDS plot shows six points for each primer set. I don’t get what they represent, since you sampled three sites taking four replicates for each one. The plot should show three points, in the case you pooled the replicates for each site, or 12 points if you were showing the diversity for each replicate. Please explain better what the points represent.

Response (L120-123): The six points in the nMDS plot represent the six sampling sites used in this study. From each site two field replicate water samples (500 ml per sample) were collected. The details of the sampling information has been provided in Materials and Methods section in lines 120-123 and Table S1.

Comment: Line 222: Since you decided to consider the abundance of reads to calculate diversity indexes, why don’t you also show the number of reads of the taxa in common among the different primer sets? This could be indicative of the amplification efficiency of each primer set for those taxa. This is just a suggestion, but I would like to understand more.

Response (L285-289): The number of reads of the taxa in common among the different primer sets have been provided in lines 285-289. The number of reads for the taxa in common among the Riaz_12S, Valentini_12S, and Berry_16S primer sets were comparable and 514,041, 555,883, and 569,976 reads were assigned to each primer set, respectively. The MiFish_12S has the lowest number of reads (381,805) assigned to species in common.

Comment: Line 232: Please, could you provide also the average number of reads for sample?

Response (L245-248): The average number of reads per sample for each marker set has been included in lines 245-248.

Comment: Line 261-262: Low evenness means big differences in reads number among taxa. This is quite common in metabarcoding studies, since several variables can influence the yield, resulting in high variability. Anyway, how different these yields are for each taxon? Please, could you provide something more? For instance, a plot with all the reads of each taxon ordered from the smaller to the higher. Alternatives to my suggestion are also welcome.

Response: As per the suggestion we plotted a graph of 17 common species detected across all four primer sets, with reads ordered from the smaller to the higher. The figure has been presented as a supplementary figure S1.

Submitted filename: Response to reviewers comment.docx

Click here for additional data file.

4 Nov 2021

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Arnar Palsson, Ph.D.

Academic Editor

PLOS ONE

Additional Editor Comments:

The manuscript has improved a great deal, but we opt for Major revision again because a major concern of reviewer 1 was not addressed (point 1). Also, because 2 reviewers backed out, I had to enlist two new ones, and one of them provided good recommendations that should improve the manuscript.

1. Change the threshold for species detection (>99%) is way to stringent. See rev 1. this AND previous comment! “I recommend authors to set an appropriate threshold for each primer set, considering the difference of evolutionary speed among amplified regions.”

2. Because of the differences in amplicon length, then the comparison of the primer sets is challenging.

3. Reviwer 5 offers suggestions on how to tackle this “I would suggest to use the proportions instead of the raw number of reads or at least transform the number of reads (e.g. forth square or logarithm) to make results more comparable. Comment: for a better comparison of the efficiency of the four different primer sets I would suggest the author to take into account not only the number of generated reads per marker but also their relative sequencing depth, especially for those sequences that were assigned to fishes”

4. Provide more details on the sampling sites.

5. Rev 5. “The authors state that they performed two replicates in each of the six sampling sites but then in the nMDS plot I can see only 6 points. Were the replicates pooled? How?... “

6. Tone down the title (rev. “As a methodology study, this study is not rigorous enough to reflect the value of the title.”

7. Also, the abstract ends with conflicting messages. You say that its best to use 2 or more primer sets, but then you conclude one set is better than another for your purposes?

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: (No Response)

Reviewer #4: (No Response)

Reviewer #5: (No Response)

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: No

Reviewer #4: Partly

Reviewer #5: Partly

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: N/A

Reviewer #4: Yes

Reviewer #5: N/A

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #4: Yes

Reviewer #5: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #4: Yes

Reviewer #5: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: The authors kept ≥ 99% similarity cutoff for species assignment regardless of the previous review comments. The threshold should be set for each primer set because the evolutionary rate varies even within the same gene. From my experience for example, The MiFish region clearly has a fast evolutionary rate, and one underestimates the species detection when a 99% threshold is adopted. The threshold value should be set carefully because it has a significant impact on the validity of primer selection.

Reviewer #4: In my opinion, the paper is mainly sound. However, the authors clarified the candidate primers were chosen according to Zhang (2021). Zhang (2021) is based on freshwater fish and suggested that the primer choice cannot be solely based on in silico evaluation. As the author said by themselves, primer selection experiments should do both in silico and vitro. So I think that's not enough to justify the choice of these candidate primers. The author should clarify this point in the discussion because marine samples are very different from freshwater samples. Otherwise, this study doesn't have a customer database to compare the outcome species, but it's fine as it's fair to all primers. They only use the inventory list alone is not enough to say that they did an excellent species resolution survey. In short, The authors show how they selected primers for their study in an estuarine environment. It's a good complement to the eDNA study. As a methodology study, this study is not rigorous enough to reflect the value of the title. The authors already account for the shortcomings of their experimental design in the revised edition. I suggest authors should also clarify the limitations of candidate primer selection.

Reviewer #5: Dear Editor,

I find myself in agreement with reviewer #1 and I think the manuscript still needs substantial revision before its publication, and also feel that reviewer 1 should have the opportunity to assess the authors’ responses.

I add some more personal comments to the previous reviewer’s ones.

Comment: I generally have serious concerns regarding the idea of comparing primers that have such a different amplicon length and especially I agree with reviewer #1 in the criticism about the criteria for species identification, using the same threshold for example for Valentini 12S that is 63 bp long and for MiFish 12S, which is nearly three times that length.

Comment: the authors used reads numbers as a proxy for abundance data and compared the results for the four primer sets. A lot of factors can affect the proportion between the initial amount of template DNA and the final numbers of sequence reads. This is even more an issue considering the different lengths of the fragment amplified and the fact that you used two PCRs to build sequencing libraries. I would suggest to use the proportions instead of the raw number of reads or at least transform the number of reads (e.g. forth square or logarithm) to make results more comparable.

Comment: for a better comparison of the efficiency of the four different primer sets I would suggest the author to take into account not only the number of generated reads per marker but also their relative sequencing depth, especially for those sequences that were assigned to fishes.

Comment: The authors state that they performed two replicates in each of the six sampling sites but then in the nMDS plot I can see only 6 points. Were the replicates pooled? How? I think this should be specified, as it will also affect species richness values within samples.

Comment: I agree with reviewer #2 and I think you should add more details on the sampling sites. I would suggest to add a map showing the geographic coordinates of each sampling site or at least add more details in Table S1. Latitude and longitude records should obviously include N/S and W/E annotations respectively.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #4: No

Reviewer #5: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

3 Feb 2022

Response to reviewer’s comments

Reviewer #1:

Comment: The authors kept ≥ 99% similarity cutoff for species assignment regardless of the previous review comments. The threshold should be set for each primer set because the evolutionary rate varies even within the same gene. From my experience for example, The MiFish region clearly has a fast evolutionary rate, and one underestimates the species detection when a 99% threshold is adopted. The threshold value should be set carefully because it has a significant impact on the validity of primer selection.

Response: We agree. As per the suggestion of the reviewer, we used three different similarity cutoffs (97%, 98%, and 99%) to see if the change in threshold criteria influenced species detection. An increase in the total number of detected species was observed when we used 98% cutoff as compared to 99% similarity threshold (Table S2; below). However, a comparison of 98% to 97% similarity threshold did not lead to a change (except one species for MiFish). Similar to taxonomic assignment, different cutoffs for the primers did not change the overall diversity indices (Table S3). Since the overall result did not change, we decided to use 99% similarity threshold to avoid any misidentification of species.

Table S2 Fish species detected for each primer set at each of three different similarity thresholds (99%, 98%, and 97%). Also shown is the percentage of total reads assigned at the species-level (fishes) after quality control.

Primer set 99% Similarity threshold 98% Similarity threshold 97% Similarity threshold

# Species % Reads assigned to species # Species % Reads assigned to species # Species % Reads assigned to species

MiFish_12S 34 81.39 40 84.71 41 84.91

Riaz_12S 54 96.78 61 97.65 61 97.65

Valentini_12S 32 86.03 37 87.43 37 87.43

Berry_16S 48 97.38 51 97.86 51 97.88

We’ve added this text to section 3.2

“While adjusting the sequence similarity threshold did change the number of species detected for each primer set the overall pattern did not change. Regardless of which similarity threshold was applied the Riaz_12S primer set resulted in the most species detected with Berry-16S ranking second (Table S2). There were no notable changes in taxa detected when comparing the 97% and 98% thresholds but there was an increase in the number of species resolved if we applied the 98% threshold compared to the more conservative 99% threshold. Most notably, when we compared the best performing Riaz_12S using the 99% cutoff with the worst performing primers sets using a 97% cut off the latter still resulted in high number of specie detections (Table S2). Similar patterns were found for diversity indices (Table S3). As a result, we have opted to apply a single conservative similarity threshold of 99% for species-level designations and report these throughout.”

Reviewer #4: In my opinion, the paper is mainly sound. However, the authors clarified the candidate primers were chosen according to Zhang (2021). Zhang (2021) is based on freshwater fish and suggested that the primer choice cannot be solely based on in silico evaluation. As the author said by themselves, primer selection experiments should do both in silico and vitro. So I think that's not enough to justify the choice of these candidate primers. The author should clarify this point in the discussion because marine samples are very different from freshwater samples. Otherwise, this study doesn't have a customer database to compare the outcome species, but it's fine as it's fair to all primers. They only use the inventory list alone is not enough to say that they did an excellent species resolution survey. In short, the authors show how they selected primers for their study in an estuarine environment. It's a good complement to the eDNA study. As a methodology study, this study is not rigorous enough to reflect the value of the title. The authors already account for the shortcomings of their experimental design in the revised edition. I suggest authors should also clarify the limitations of candidate primer selection.

Response: The choice of primer candidates has been further explained in the discussion section. Zhang et al. (2020) performed in silico experiments on a number of primer sets including 4 of the 5 selected here. For in vitro experiments they tested these same primer sets in a freshwater system. The title of the manuscript has been changed to reflect our study.

Our Discussion now reads

“Despite the growing interest in the use of eDNA to assess fish communities, there has been a surprising lack of studies that have directly compared the efficacy of metabarcoding primer sets in marine and estuarine systems. Because the performance of eDNA metabarcoding primers will vary depending on the study system (freshwater, marine, or estuarine) and taxonomic composition, there is no guarantee that a primer set that performs well in a freshwater system will do so in marine or estuarine systems. For this reason, we tested the four best performing metabarcoding primers from the freshwater study of Zhang, Zaho (42), to determine their effectiveness in a biodiverse estuarine system.”

Reviewer #5:

Comment: I generally have serious concerns regarding the idea of comparing primers that have such a different amplicon length and especially I agree with reviewer #1 in the criticism about the criteria for species identification, using the same threshold for example for Valentini 12S that is 63 bp long and for MiFish 12S, which is nearly three times that length.

Response: The reviewers point out the important point that the primers used in this study produce different amplicon lengths which may influence species detection. It is also important to take into consideration that when a lab picks up a new protocol, they are looking for clear guidance concerning thresholds and cuts-off but in general this is lacking in the literature. As a result, general rules of thumb are typically applied such as 99% for species level designations. With this in mind we now present some analyses using three different similarity cutoffs (97%, 98%, and 99%) to determine if changes in threshold criteria influence the total number of species detected. Overall interpretation of the results did not change, and our results section (Section 3.2) now reads:

“While adjusting the sequence similarity threshold did change the number of species detected for each primer set the overall pattern did not change. Regardless of which similarity threshold was applied the Riaz_12S primer set resulted in the most species detected with Berry-16S ranking second (Table S2). There were no notable changes in taxa detected when comparing the 97% and 98% thresholds but there was an increase in the number of species resolved if we applied the 98% threshold compared to the more conservative 99% threshold. Most notably, when we compared the best performing Riaz_12S using the 99% cutoff with the worst performing primers sets using a 97% cut off the latter still resulted in high number of specie detections (Table S2). Similar patterns were found for diversity indices (Table S3). As a result, we have opted to apply a single conservative similarity threshold of 99% for species-level designations and report these throughout.”

Furthermore, our discussion now reads

“Because the rate of evolution varies across genes and gene regions applying a single set of taxonomic thresholds (i.e., 99% for species; 97% for genus; 95% for family; 90% for order; 85% for class; and 80% for phylum; as applied here) can be misleading. However, our analyses of the sequencing results from the four primer sets at three different species-level similarity thresholds (99%, 98%, and 97%) did not change our interpretation of the data. The Riaz_12S primer set either did as well or outperformed the others in terms of specificity (99.99% of reads assigned to fishes; Table 2), universality (greatest number of taxa amplified; Table S2), and taxonomic resolution (96.78% of reads assigned to the species-level at the 99% threshold) regardless of which cutoff value was applied. Most notably, this was true even when comparing Riaz_12S using the conservative 99% cutoff against all other primers sets using a 97% cut off (Table S2). Similar patterns were found for diversity indices (Table S3). It’s important to note that the Berry_16S primer set also performed well in our experiments.”

Comment: the authors used reads numbers as a proxy for abundance data and compared the results for the four primer sets. A lot of factors can affect the proportion between the initial amount of template DNA and the final numbers of sequence reads. This is even more an issue considering the different lengths of the fragment amplified and the fact that you used two PCRs to build sequencing libraries. I would suggest to use the proportions instead of the raw number of reads or at least transform the number of reads (e.g. forth square or logarithm) to make results more comparable.

Response: We thank the reviewer for this suggestion. We have log transformed the sequence reads of the 17 common species among all the four primer sets used in this study and a figure presenting these data can now be found as Figure S2.

Our text now reads:

“Based on the ANOSIM analyses, species assemblages differed significantly across primer sets (R = 0.4011, P < 0.001); a finding that was supported by the NMDS plots which showed clear separation among marker sets when either the read abundance data (Figure 3) or presence/absence data (Figure S2) were analyzed.”

Comment: for a better comparison of the efficiency of the four different primer sets I would suggest the author to take into account not only the number of generated reads per marker but also their relative sequencing depth, especially for those sequences that were assigned to fishes.

Response: Percent of total reads per marker and average sequencing depth per species were calculated and data is now presented in Table S5. Our results were further corroborated as MiFish_12S resulted in the number of reads suggesting lower PCR efficiency for this primer set in our experiments.

Comment: The authors state that they performed two replicates in each of the six sampling sites but then in the nMDS plot I can see only 6 points. Were the replicates pooled? How? I think this should be specified, as it will also affect species richness values within samples.

Response: The sequence reads of replicates from the same sites were pooled to plot NMDS. This has been clarified in the text in line 230..

Comment: I agree with reviewer #2 and I think you should add more details on the sampling sites. I would suggest to add a map showing the geographic coordinates of each sampling site or at least add more details in Table S1. Latitude and longitude records should obviously include N/S and W/E annotations respectively.

Response: The details of our study sites has been added to Table S1 with GPS coordinates in decimal degrees (DD) which is the preferred format for GEOME where we archive our metadata. DD are not typically reported with N/S or W/E annotations.

Submitted filename: Response to reviewers comments_F.docx

Click here for additional data file.

8 Mar 2022

24 Mar 2022

Comment: The study compared the primer pairs but did not make an attempt to describe the differences in biological resolution. Do all primer pairs have the same power to capture differences between locations (or any other biological variable of interest, habitats, seasons, depth etc.)? It is not described in the methods (or I missed it) but do the 6 sites differ biologically?

This can be tackled easily; can you add location numbers to NMDS graph. Also, can you test for differences between sites?

Please bring this point also up in the discussion.

Response: Thank you for the positive feedback on our manuscript. Our sample sites don’t represent different habitats but instead are over a similar bottom type and are in close proximity. Therefore, we consider them to be replicates. However, that still begs the question of whether they are different. So, as you suggested, we performed the ANOVA to determine if the sites differ. Results showed no significant differences among the sites (p = 0.33), suggesting that the sampling sites are biologically similar. In the NMDS plot (figure 3), we have added the location numbers and clarified in the legend of figure. All these points have been clarified in the methods/results and mentioned in the discussion section.

Minor points.

Line 64

“breadth of which can vary from”

Response: Corrected

Line 98.

References to the Zhang paper, no need to include the first name, “Zhang et al” is better

“In a recent study, Zhang et al. (42)”

Response: Done

Line 209. Extra word in citation?

“following West, Stat (35).”

Response: Citation has been corrected.

Line 230-31. Another reference issue”

“An analysis of similarity (ANOSIM; ANOSIM; 55)”

Response: Duplicate ANOSIM has been deleted.

Line 272. Add “predictably”

“but there was predictably an increase in the number”

Response: “Predictably” has been added in the sentence.

Line 321.

Is there any guarantee that the primers found here will be best in other marine or estuarine systems?

Response: There is no guarantee that performance of primers in this study will be best in other marine or estuarine system as effectiveness of primers differ depending on the geographic regions and different ecosystems. The same has been discussed in the Conclusions and Recommendations section.

Line 350. Ref issues again, sort through entire manuscript and fix.

“MacDonald, Young (44)”

Response: Corrected

Line 365 Please rephrase.

“However, careful testing is still needed.”

Response: The section now reads “However, if elasmobranchs are the primary target group, it would be prudent to test primers using DNA extractions from species expected in the study area. Furthermore, there are a number of published primers that have been designed specifically for elasmobranchs that are worth exploring”

Submitted filename: Response to reviewer_F.docx

Click here for additional data file.

28 Mar 2022

Comparing eDNA metabarcoding primers for assessing fish communities in a biodiverse estuary

PONE-D-21-15188R3

Dear Dr. Kumar,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Arnar Palsson, Ph.D.

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

1 Apr 2022

PONE-D-21-15188R3

Comparing eDNA metabarcoding primers for assessing fish communities in a biodiverse estuary

Dear Dr. Kumar:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Arnar Palsson

Academic Editor

PLOS ONE